It has been known for some time that small peptides corresponding to certain sequences in proteins are capable of adopting conformations in aqueous solution that are similar to the conformation in the native protein (eg. Goodman & Kim, 1989, Biochemistry, 28, 4343-4347). Small peptides corresponding to sequences in actin and myosin, which are thought to be interaction sites between actin and myosin, are known to interact with the reciprocal protein partner in such a way that they interfere with the biological activity of the actomyosin system in assays both using the isolated muscle proteins (eg. inhibition of the actin activated Mg2+ATPase activity) and using intact myofibrils (eg. inhibition of force generation; eg. Keane et al., 1990, Nature 344, 265-268). The relationship between the structure adopted by these peptides and the biological activity has not previously been investigated.

In this thesis peptides from both actin and myosin have been identified which specifically bind to the protein partner (myosin S1 peptides corresponding to sequences from the 20 kDa domain 704-710 and 718-727, and from 50 kDa domain 609-618, each of which bind to F-actin; actin peptides: 16-41, 77-94, 29-58 and 96-117, the first two of which bind to the myosin S1 head). The binding residues within these peptides have been identified by the observation of broadening of resonances in NMR and, for some of the peptides, the site of interaction in the protein template has been determined by chemically crosslinking the radioactively labelled peptide to the protein. In addition for many of the peptides identified as binding sites, the solution structure of the isolated peptide was determined by NMR spectroscopy (and in one case by CD spectroscopy). In particular a helical peptide (actin 77-94) and extended peptide (myosin 718-727) were investigated. In each case an attempt was made to relate the solution structure to the observed binding properties. In the two peptides corresponding to actin 77-95 and myosin 718-727, transferred NOESY spectra were recorded in the presence of the protein partner, and some observations of the conformation adopted by the peptide when bound to the protein template were made. In addition the spatial relationship between the SH1 (707Cys) and SH2 (697Cys) thiols and the ATPase site in the S1 head was investigated by 19F labelling the thiol and using a spin labelled analogue of ATP. A region of the myosin S1 head corresponding to residues 601-635 was sequenced.

This thesis attempts to probe the location and structure of the protein interfaces between actin and myosin.

Peptides corresponding to sequences within actin or myosin, in the presence of the protein partner to which they bind, have been used in muscle research for many years to help in defining the molecular interfaces between actin and myosin. Biochemical techniques, such as the characterisation of the inhibition of the actin-activated myosin S1 ATPase in the presence of added peptide, chemical crosslinking of peptide to the protein partner, and in vivo inhibition of muscle contraction are well established methods of investigation, and have previously been applied to many of the peptides described in this study, eg. peptides Y941, Y933 (see at the end of this section for a summary of the synthetic peptides used in this study).

A further technique has been used to study such peptide-protein interactions, involving the observation of specific broadening of NMR resonances of peptide in fast exchange between the bound and free states. The work presented in this thesis is an attempt to extend the previous NMR measurements in order to further delineate various established acto-myosin contact sites (eg. peptides Y933, Y941), to identify new binding sites (eg. peptides Y935, Y847), and to probe the structure of the protein-protein interface (eg peptides Y847, Y933).

The principle strategies were as follows:

1. Several putative interface sites have been identified by biochemical experiments (often chemical crosslinking; see sections 1.10 & 1.11). Peptides corresponding to the regions indicated by these experiments were synthesized and attempts made to identify the precise points of interaction by observation of specific broadening in 1-dimensional NMR titration experiments (eg. peptides Y847 and Y935).

2. An attempt was made to 'hone down' the points of interaction of peptides with their protein partners in order to identify the 'crucial' few binding residues. For example, a relatively large (28 residues) S1 peptide, Y630, had previously been identified as containing a binding site for F-actin (see section 4.0). One of the sites of interaction identified in this peptide was at the C-terminus (as indicated by the broadening of resonances). A 10 residue peptide, and then successively smaller varieties of this peptide, were synthesized corresponding to this binding site. It was demonstrated that the characteristics of broadening by added F-actin were maintained in even the shortest of these peptide varieties (see chapter 5).

3. The recent work of Wright and others (eg. Dyson et al., 1988a & 1988b) has indicated that even very small peptides can adopt non-random/native conformations in aqueous solution. The free solution structures of several of the peptides implicated in binding to actin and S1 were determined. It is possible that these structures reflect either (or both) the structure of the corresponding region in the free protein, or the structure adopted at the protein-protein interface in the acto-myosin complex.

4. Attempts were made to determine the structures of two peptides when bound to the protein partner (peptide Y847 bound to S1A2 and peptide Y933 bound to F-actin, chapters 3 & 4 respectively) using the relatively new technique of transferred NOESY. Some information was provided about the structure of peptide Y847 when bound to S1 and a model was produced of the possible local changes in structure of residues approx. 77-95 of actin during the transition from free actin to the acto-myosin complex. For peptide Y933, the TrNOESY experiment is particularly difficult, and a strategy of careful examination of the resonances broadened in 2D-TOCSY experiments, and subsequent combination of this binding data with the free solution structure of the peptide was used to propose a model for the bound structure.

Although structural details of the free and bound interface sites was the ultimate goal of this work, other supplementary work was of course necessary. For example, the primary amino acid sequence of a region of the myosin S1 head, suspected of interacting with F-actin, was disputed among several authors, so it was necessary to first develop a preparation of a suitable digestion fragment of S1 and to sequence it, before peptides could be made that might be expected to reliably model the acto-S1 interface (peptide Y935, section 5.1).

Throughout the work attempts were made to cross-check the peptide NMR binding data by identifying the canonical binding site on the protein partner, and then attempting to demonstrate the reciprocal binding of a peptide from that site. The principle technique applied was chemical crosslinking of labelled peptide to protein, followed by fragmentation of the protein and sequencing, to identify the interaction site.

Peptides used in this study include:

Y941 (myosin S1 20 kDa 704-710) section 5.8

Y933 (myosin S1 20 kDa 718-727) chapter 4

Y935 (myosin S1 50 kDa 609-618) section 5.3

Y675 (myosin S1 50 kDa 592-616) section 5.2

Y630 (myosin S1 20 kDa 702-730) section 5.7 & 4.0

Y774 (actin: 96-117) section 5.4

Y655 (actin: 16-41) section 5.5

Y772 (actin: 29-58) section 5.6 & 5.7

Y847 (actin: 77-94) chapter 3.

Abbreviations:

1D NMR One-dimensional nuclear magnetic resonance

2D NMR Two-dimensional nuclear magnetic resonance

A1 / A2 Alkali light chain 1 or 2

ADP Adenosine 5'-diphosphate

AMP Adenosine 5'-monophosphate

ATP Adenosine 5'-triphosphate

ß and ßCH2 An amino acid sidechain ß-protons.

Bicine N,N'-bis(2-hydroxyethyl)glycine

CD Circular dichroism spectroscopy.

CDCl3 Deuterated chloroform

CNBr Cyanogen bromide

COSY 2D correlated spectroscopy

D and DCH2, etc. An amino acid sidechain delta proton.

D2O Deuterium oxide

DMF N,N-dimethylformamide

DNase-I Deoxyribonuclease-I (bovine pancreatic)

DQF Double quantum filtered spectroscopy

DTNB 5-5'-dithiobis(2-nitrobenzoate)

DTT Dithiothreitol

EDC 1-ethyl-3-[3-(dimethylamino)propyl] carbodiimide

EDTA Ethylenediaminetetraacetic acid.

EPR Electron paramagnetic resonance

EtOH Ethanol

Fmoc 9-Fluorenylmethoxycarbonyl

FRET Fluorescence energy transfer

FTIR Fourier Transform Infra-red spectroscopy.

G and GCH2, etc. An amino acid sidechain gamma proton.

GTP Guanosine 5'-triphosphate

HOHAHA Homonuclear Hartmann-Hahn 2-D spectroscopy

HPLC High pressure liquid chromatography

Hz Hertz

kDa Kilodalton

kHz kilohertz

MeOH Methanol

MOPS 3-[N-morpholino]propanesulphonic acid

nm Nanometres

NMR Nuclear magnetic resonance

NOE Nuclear Overhauser enhancement or effect

NOESY 2D nuclear Overhauser and exchange spectroscopy

Pi Inorganic phosphate

PMSF Phenylmethylsulphonyl fluoride

ppb parts per billion (part per thousand million).

pPDM N,N'-1,4-phenylenedimaleimide

ppm Parts per million

ROESY 2D Rotating Frame nuclear Overhauser and exchange spectroscopy

rpm Revolutions per minute

S1 Subfragment 1

S1A1 Subfragment 1 carrying the A1 light chain

S1A2 Subfragment 1 carrying the A2 light chain

sc the sidechain protons of an amino acid, beyond the ß-protons.

SDS Sodium dodecyl sulphate

SDS-PAGE SDS-Polyacrylamide gel electrophoresis

SL- Spin-labelled

SL-ADP Spin-labelled ADP

SL-ATP Spin-labelled ATP

TCA Trichloroacetic acid

TEA Triethanolamine

Tm Tropomyosin

TnC Troponin C

TnI Troponin I

TnT Troponin T

TOCSY 2D Total Correlated spectroscopy, see HOHAHA

TPCK Tosyl-L-phenylalanenylchloromethyketone

Tris 2-amino-2-(hydroxylmethyl)-1,3-propanediol

TrNOESY Transferred NOESY spectroscopy

TSS Sodium 3-trimethylsilylpropane sulphonic acid

uv Ultra-violet

For the amino acids the IUPAC-IUB one-letter or three-letter symbols are used.

Individual atoms are identified using the atomic symbols of related atoms where possible (NH is the backbone amide proton), together with a Greek prefix where necessary (αH, the backbone alpha proton; ίCH2, the sidechain ß-protons, etc). Atoms are named in amino acids according to the XPLOR format.

Proton distances and descriptions are described in the following fashion, with certain variation where it is appropriate.

dAB(i,j) is the distance from proton A in the amino acid residue i, to proton B in residue j, for example, dαί(i,i+3) or 9TyrßCH2-12ProαH.

All the other abbreviations are explained in the text.

Muscle tissue is the specialised organ in animals which performs the task of movement: The transduction of chemical free energy into motion. The behaviour of muscle can be characterised at a gross level by the mechanical work done. A physiologist would study muscle contraction in externally stimulated muscle fibres in terms of the generation of tension at a fixed length (isometric contraction) or a shortening of the fibres pulling against a fixed load (isotonic contraction). In vivo the load on muscle fibres changes continuously and the stimulus for contraction is driven by the nervous system. So these are artificial divisions of behaviour introduced in order to dissect the action of muscle. In recent years the dissection of muscle has proceeded to a molecular level. For example, it is now widely held that the mechanism of contraction involves the triggering of the hydrolysis of ATP, and that the means of transduction of the free energy so released is to drive changes in the interactions between the two principal muscle proteins. These proteins are actin and myosin. This study seeks to probe further the details of this crucial interaction that takes place at the interface between actin and myosin.

Vertebrate muscles are classified into two types, smooth and striated, with the latter classification being subdivided into cardiac and skeletal muscle. All three classes of muscle perform different functions. Skeletal muscle is used for short bursts of activity or prolonged contraction, and is under voluntary control. By acting through tendinous attachments to the bony skeleton the skeletal muscle achieves movements like running, swimming, and flying. Animals have different skeletal muscle fibre types which are used for different tasks, in order to optimise efficiency and power production. Slow fibres have a low maximum velocity of shortening and fast fibres have a large maximum velocity. Hence, slow fibres tend to be involved in the maintenance of posture and in slow motions, whilst as the speed of movement required increases the faster fibres are recruited. Indeed the fast fibres themselves are subdivided, so that as speed of movement increases then there is sequential recruitment of fast oxidative and fast glycolytic fibres (Rome et al., 1988). Cardiac muscle is found only in the heart, and is under involuntary control. It is required to function continuously throughout the life of the animal. Recently it has been suggested that the life-span of an animal is related to the heart rate, so that for example a mouse with a fast heart rate and an elephant with a slow heart rate live for a similar total number of heart beats. Smooth muscle is under involuntary control, and is used for a variety of bio-mechanical processes such as gut peristalsis and contraction of arteries. While muscle types differ in their function, structure, protein arrangement, the protein isoforms expressed, the metabolic pathways used to generate ATP, and the mechanism of regulation, nevertheless, the fundamental mechanism by which all muscles transduce chemical free energy into mechanical work is the hydrolysis of ATP to drive a mechanical interaction between actin and myosin protein filaments.

The muscles fibres, from which skeletal muscle is formed, consist of many separate cells fused into a single large polynucleate cell, which can be up to 50 cm long and 100 um in diameter. Most of its bulk is occupied by regular cylindrical arrays of protein which form the organelle of contraction. The actin and myosin in this organelle amount to 80% of total protein, thus providing a convenient source of pure proteins for biochemical investigations. Each muscle fibre is made up of a parallel bundle of up to a thousand myofibrils, which are long cylindrical elements 1 um to 2 um in diameter that extend the entire length of the cell. Fibres are linked together by collagenous connective tissue that comes together at the end of the muscle to form the tendonous attachment to the skeleton. Light microscopy reveals that myofibrils are composed of light and dark bands along the muscle axis, giving rise to repeating units along each myofibril about 2.5 um in length (hence striated muscle; figure 1.2.1). Each of these units is termed the sarcomere. The half-sarcomere is the fundamental repeating unit in striated muscle. The light bands are known as the I bands (isotropic), whilst the dark bands are called the A bands (anisotropic). At the centre of the I band is a thin dark line known as the Z disk. The region between two adjacent Z lines of a myofibril defines the sarcomere. Z lines tend to be aligned in axial register in adjacent myofibrils. Two types of highly ordered interdigitating myofilaments are seen in the electron microscope: the thin filament protein (actin, which forms most of the I-band) and the thick filament protein (myosin, which forms most of the A band). The thin filaments extend from the Z line and the actin composing these filaments is arranged in an antipolar fashion in successive sarcomers, with the polarity of the filament changing at the Z-band. The thick and thin filaments interdigitate. The region in which the thick and thin filaments overlap gives rise to the darker appearance of the extremities of the A band. A lighter region is observed at the central region of the A band (the H zone) that consists only of thick filaments and in which there is hence no filament overlap. A cross-section through the A-band shows that the thick filaments are arranged in a hexagonal lattice, and each thick filament is surrounded by six thin filaments. Each thin filament is surrounded by only three thick filaments. The M-line at the centre of the of the H-zone, holds the thick filaments in register. Passing throughout the sarcomere and extending into the extra-myofibillar space is the muscle cytoskeletal intermediate filaments. An elastic-like protein called connectin or titin (Wang et al., 1979) is found within the myofibrils. The role of titin is unclear, but it probably is involved in super-ordering within the myofibril because of its very high molecular weight (Trinick et al., 1984). It is the highly ordered arrangement of the actin and myosin filaments in skeletal muscle (not found in other muscle types) that makes skeletal muscle very amenable to a variety of spectroscopic (microscopic and X-ray diffraction studies) and mechanical studies. The first clues as to the mechanism by which the two fundamental protein interact with each other arose from simple light microscope studies.

The cell surface is specialised to receive electrical stimulation from the nervous system, which triggers action potentials at the membrane, that result in a sudden increase in intracellular calcium ion concentration. The influx of calcium in turn triggers the molecular events at the acto-myosin complex that result in contraction.

During contraction, the length of the of the I band and the H zone shortens, whilst the length of the A band remains constant. It was this observation which lead A.F. Huxley and A.E. Huxley separately (Huxley & Niedergerke 1954; Huxley & Hanson, 1954; Huxley, 1969) to formalise a sliding filament theory for muscle contraction in which it was suggested that stimulation of muscle causes the extent of overlap between the thin and thick filaments to increase. As a result of this the Z lines of each sarcomere are drawn closer together and concerted shortening of multiple sarcomeres, oriented along the fibre axis, leads to the contraction of the muscle as a whole. Importantly, in this model the length of the thick and thin filaments remains unchanged, and tension is generated by the sliding interactions between the thick and thin filament.

Electron microscopy of myofibrils, revealed that in the region of the sarcomere in which the thick and thin filaments overlap, the gap between the two filaments is bridged by projections at very regular intervals, called crossbridges. These crossbridges are an intrinsic part of the thick (myosin), rather than thin filament (actin). Myosin has the ability to hydrolyse Mg2+ATP (Szent-Gyorgyi, 1953) and interact with actin. The Mg2+ATPase activity of myosin is increased some 200 fold upon actin interaction (Eisenberg & Moos, 1968). It is believed that force generation in muscle contraction arises from the cyclical attachment to the crossbridges to the thin filament in the region where the thick and filaments overlap. A power stroke then occurs, which is accompanied by a change in configuration of the crossbridge attached to the thin filament and hydrolysis of ATP. This is then followed by detachment from the crossbridge from actin which are then ready to start a new cycle of muscle contraction. Hence the filaments travel laterally with respect to each other in a ratchet-like mechanism. Recently myosin head movements have been shown to be synchronous with the elementary force-generating processes in muscle (Irving et al., 1992).

Myosin is one of the two major proteins found in skeletal muscle (see review: Harrington et al. 1984). It is the principal component of the thick filament. In vertebrate skeletal muscle these filaments are approximately 1.6 um long and 14 nm in diameter (Page & Huxley, 1963). Approximately 1012 myosin molecules are found in each half-sarcomere unit. Myosin is a multisubunit protein consisting of two identical heavy chains each of molecular weight approx. 200 kDa, each with two associated light chains of molecular weight of approx. 20 kDa. Like other large filamentous proteins, myosin is a highly asymmetric molecule, which in the electron microscope can be seen to consist of two protruding pear-shaped heads (190nm long and 65nm maximum width) attached to a 140nm long tail (Elliot & Offer, 1978). Well separated myosin heads have recently been visualised protruding from the filament in the electron microscope in frozen sections of scallop myosin (Vibert, 1992). The N-terminal portion of the molecule forms the globular head, whilst the C-terminal half of the heavy chains associate to form the filament. Each globular head is associated with one of each of the two types of light chains, alkali and phosphorylatable (or regulatory).

As well as in muscle, myosin is also linked to motility in the cytosol. The most studied examples are in amoeba, where myosin II (analogous to skeletal muscle myosin) is involved in cytokinesis, and myosin I (single headed) is involved in chemotaxis and cytoplasmic organelle movement (Way & Weeds, 1990). Myosin heavy chains are expressed as different isoforms among different tissues at various stages of development. The genetics of expression is complex, involving multiple gene copies and differential gene splicing, somewhat reminiscent of immunoglobulin expression. The complete sequence of the rabbit skeletal myosin S1 heavy chain has now been published (Tong & Elzinga, 1990).

The study of myosin has only been made possible by examining the functions of its proteolytic fragments. Two sites in myosin are susceptible to proteolysis under mild conditions, situated in the tail region and at the junction between the head and tail (Weeds & Pope, 1977; figure 1.4.1). Selective cleavage at the site within the tail region can be achieved by digestion of monomeric myosin in solution at high ionic strength with α-chymotrypsin in the presence of divalent cations such as Mg2+, giving rise to light meromyosin (LMM, molecular weight 140 kDa, the C-terminal tail region), and heavy meromyosin (HMM, molecular weight 340 kDa, with a portion of the tail and the globular heads). Protection at the head-rod junction is afforded by the Mg2+ ion which binds to the phosphorylatable light chains on neck region of the myosin head. There are indications that there is a collapse of the S1 head on to the rod which affords a degree of protection to the S2/LMM hinge in the presence of Mg2+ (Persechini & Rowe, 1984). In the absence of Mg2+ (presence of EDTA) the phosphorylatable light chain is rapidly digested and so cleavage occurs at the head-rod junction. This gives rise to an isolated rod and a pair of separated globular heads. These head regions are commonly referred to as the S1 heads (subfragment 1) and contain the N-terminal 810 residues of the heavy chain associated with an intact alkali light chain. A different preparation of the S1 heads (papain digestion in the presence of Mg2+) has both the alkali and phosphorylatable light chains present, although the latter is slightly degraded at its N-terminus (Lowey et al., 1969). Digestion at the site within the rod region is also observed in this case and the C-terminus of the S1 head is somewhat heterogeneous.

Isolated myosin molecules at physiological ionic strength and pH self associate to form thick filaments (Huxley, 1963). The LMM domain forms the association site between the myosin molecules, which pack together in a regular staggered array from which three pairs of myosin heads (at 120° to each other) are projected from the surface of the myosin filament every 14.3 nm. Each set of three pairs of heads is staggered by 120° to the previous set, resulting in a helical repeating pattern of heads protruding from the thick filament, in which every third set of heads is aligned giving a 43 nm repeat. The myosin heads appear to lie at an angle of 45° to the filament axis. These globular S1 heads interact with actin, and are the crossbridges seen in the electron microscope, between the thick and thin filaments of muscle. The thick filament is bipolar which results in a bare central region where there are no crossbridges (150-200 nm long, Huxley, 1963) and the polarity of the filament is reversed. This bare zone is the site of initiation of filament formation where oppositely oriented myosin molecules come together.

The structure of the rod region has been shown to be an alpha helical coiled coil in which the two helical polypeptide chains are coiled around one another to form a rigid rod shaped molecule. The isolated LMM is a convenient fragment to study in solution. The amino acid sequence of the LMM domain shows a repeating pattern common to all coiled coils in which every alternate third and fourth residue is hydrophobic and forms the contact between the helices. Clusters of charged residues are found between these hydrophobic residues and are hence exposed on the outside of the coil. These charged residues are involved in the self-association of LMM at physiological ionic strength, probably into dimers (Harrington et al., 1972) which then interact electrostatically to build up the filament (McLachlan & Karn, 1982). These charged interactions are broken at high ionic strengths, and so myosin does not associate to form filaments and is hence soluble. Sharp lines are seen in the NMR spectrum of myosin (4-5Hz) at low concentration (25uM) which arise from the C-terminal residues and correspond to an unfolded area in the coiled-coil (Kalbitzer et al., 1991). These areas are implicated in association of myosin rod dimers into the thick filaments (Maeda et al., 1990). The subfragment 2 region of the rod, S2 has a similar coiled-coil seven residue sequence repeat pattern to LMM, but fails to form filaments at physiological ionic strength. The S2 region acts as an arm (55nm long) upon which the globular head can be held away from the thick filament in order to form tension generating crossbridges with the thin filament.

Some disruption of the coiled coil structure at the junction between LMM and S2 and between S2 and S1 is postulated, which would render these regions sensitive to proteolysis and allow them to act as flexible structural hinges. Bends at various angles confirming this assumption have been seen in electron micrographs of isolated myosin molecules (Elliot & Offer, 1978). In addition, monoclonal antibodies specific for a single site on the myosin head, are seen in electron micrographs to bind to both sides of the myosin head, indicating that rotation about the long axis of the heads (Winklemann & Lowey, 1986). These hinges provide myosin with segmental flexibility, which is thought to play a critical role in muscle contraction. The hinge between S1 and S2 provides the possibility that the S1 head interacts with actin differently before and after the power stroke, and the hinge between LMM and S2 allows considerable variation in the position of S1 relative to the thick filament, which may permit S1 to interact precisely with actin.

The S1 head is a component of myosin which appears to possess all of the properties of the myosin molecule required for the generation of motion in muscle contraction. Since it is soluble at low ionic strengths and fails to form filamentous aggregates, the isolated S1 head has been used in the majority of biochemical work on muscle contraction. There are other good reasons for seeking a small soluble species to study. For example, myofibrils which retain the filament structure of muscle, are difficult to work with since they are not anchored and thus experiments with them suffer from the difficulty that, at excess concentrations of ATP they overcontract to shorter than sarcomere lengths, and are hence denatured. In these situations even the ATPase activity of myofibrils has proven difficult to determine, although the ATPase burst during the initial portion of contraction has recently been measured (Houadjeto et al., 1991). A justification for the validity of using the soluble S1 head follows.

The S1 head is known to hydrolyse Mg2+ATP at equivalent rates to myosin, and to bind to actin with the same affinity as myosin, whilst the fragments LMM and S2 neither hydrolyse ATP nor bind to actin. This immediately indicates that the enzymatic function of myosin resides in the S1 head, and relegates the coiled-coil regions to the role of structural components of the filament. Since, in addition the actin affinity of S1 is strongly dependent on whether ADP or ATP is bound at the active site, and the ATPase activity of S1 is activated 200 fold by F-actin (Eisenberg & Moos, 1968), it is apparent that the rudiments of a power consuming crossbridge cycle are present entirely within the S1 head. In addition, direct evidence has recently been produced that the isolated head is sufficient to support contraction. Using in vitro motility assays, in which whole myosin molecules, HMM or S1 were attached to a fixed surface (glass or a nitrocellulose film), it has been demonstrated that each class of molecule can support the movement of fluorescently labelled actin filaments in the presence of Mg2+ATP at nearly the same velocity as intact myosin filaments (Toyoshima et al., 1987). The ATP-dependent force generated by the S1 head, in similar assays with active populations of 10-100 molecules of myosin or S1 bound to a silicon coated glass surface and a single actin filament attached to a flexible glass micro-needle (calibrated for the force required to bend it), has been measured and found to the same as that produced by a myosin head in a fibre during isometric contraction. Hence the S1 head contains the entire motor machinery necessary to generate both force and motion during contraction (Kishino & Yanagida, 1988; force generated is about 1pN per myosin molecule).

Although it would seem clear that the ability to reproduce all the characteristics of contraction in the isolated S1 head means that the remainder of the molecule is only playing a structural role, there are some indications that the tail is involved in smooth muscle myosin molecules. This is discussed latter in the context of the regulation of smooth muscle contraction. However, the hypothesis of Harrington (1979) that the S2 region has elastic properties that allow the melting of part of the rod (S2) during the production of motion by the crossbridge instead of the classic swinging crossbridge model are now fairly readily discarded.

S1 (chymotryptic) consists of the N-terminal 95 kDa of fragment of the myosin heavy chain (residues 1-810), associated with a single alkali light chain to give a total molecular weight of 115 kDa. The are two skeletal muscle isoforms of the alkali light chain called A1 (molecular weight 20.7 kDa, but migrates as 25 kDa on SDS-PAGE) and A2 (16.5 kDa). Rabbit fast skeletal muscle contains twice as much A1 light chain as compared to A2 light chain (Sarkar, 1972). Both hetero- and homo-dimers of single myosin molecules (as regards the complement of alkali light chains) are formed (Lowey et al., 1979) and hence the two S1 isoenzymes S1A1 and S1A2, will be present in any preparation. These two isoenzymes are readily separated by anion exchange chromatography (Trayer & Trayer, 1988).

In the absence of a crystal structure, which is proving somewhat difficult to produce (the first and only report of S1 crystals appears in Winkelmann et al., 1985), the overall topology of the S1 head has been inferred from a wide variety of evidence, although no consensus as to the structure of S1 has yet been formed (the probable structure of the S1 head is seen in figure 1.5.1). Several reconstructions of the S1 head have been made from electron micrographs. Tokunaga et al. (1987) showed the S1 head to have a bent shape, being divided into a main distal region containing the ATPase and actin binding sites (tip), and a region around the neck which forms the proximal two-thirds of the head. This comma-shaped molecule, was 7 nm wide at the tip, decreasing to 3.5 nm at the S1/S2 junction, with the head 17-20 nm in length from the tip to the neck region (Sutoh et al., 1986). In the complexes with actin, the area from the tip to a bent region (7nm from the tip) has an axis lying almost perpendicular to the helix axis, with the axis of the neck region almost parallel to the helix axis (Toyoshima et al., 1979; Wakabayashi et al., 1981). Similar structures for the S1 head have also been observed in avian S1 crystals (Winklemann et al., 1985) and in the electron microscope (Walker and Trinick, 1988).

Mild treatment of S1 with trypsin, cuts the heavy chain at two points giving rise to 3 heavy chain fragments of SDS-PAGE sizes of 27 kDa, 50 kDa and 20 kDa (Balint et al., 1978; Mornet et al. 1979). The N-terminal '27 kDa' domain contains residues 1-204 in rabbit skeletal myosin S1 (Tong & Elzinga, 1983; calculated molecular weight 23 kDa), the central 50 kDa domain residues 214-636 and the C-terminal 20 kDa domain residues 643-810. A variety of other proteases including Clostripain (Mornet et al., 1984), plasmin, elastase (Applegate & Reisler, 1983), and staph. aureus V8 protease (Chaussepied et al., 1983) also give rise to a similar pattern of digestion. These observations have lead to the suggestion that the myosin head is composed of 3 domains with independent structural integrity, joined together by surface exposed, proteolytically sensitive connector regions. The three fragments of the S1 heavy chain, remain tightly associated when S1 is digested in these connector regions unless denaturing conditions are used to separate the domains. Trypsin digested S1 is still able to hydrolyse ATP and bind to actin which suggests that the domains retain a native conformation and orientation even when not covalently bound (Mornet, 1981a). In the absence of a crystal structure little direct evidence can be available to support the idea of independent domain structures within the S1 head, but some other evidence has accumulated to support this structure. Electron microscopy studies of negatively stained S1, and its proteolytic fragments has indicated the presence of three main regions in the myosin head (Walker et al., 1985), three independently melting cooperative regions, presumed to correspond to the domains, have been resolved during thermal denaturation by differential scanning microcalorimetry (Levitsky et al., 1991), and all three fragments have been isolated and renatured (20 kDa, Muhlrad & Morales, 1984; 50 kDa, Muhlrad et al., 1986; 27 kDa, Muhlrad, 1989). In addition, analysis of the primary structure of many muscle and non-muscle heavy chains indicates that this proposed domain structure is a common feature for all myosin head sequences, and is supported by proteolytic analysis of a wide variety of muscle myosins (eg.. cardiac myosin, Applegate et al., 1984; smooth chicken gizzard myosin, Bonet et al., 1987).

The positions of the domains within the S1 head have been localised by the use of antibodies directed against specific segments of the heavy chain. The tip of the S1 head binds antibodies specific to the 50 kDa segment (Miyanishi et al., 1988), and the 27 kDa domain has been visualised in the electron microscope in the central part of the myosin head (Sutoh et al., 1987). The 20 kDa domain is represented schematically in the figure 1.5.1 as an elongated structure spanning nearly the entire length of the head. The two most reactive cysteine residues of S1, are present in this 20 kDa fragment (697Cys is SH2 and 707Cys is SH1) allowing labels specifically to be introduced into this region. Distance measurements have shown that the SH1 group is 12-13 nm from the head-rod junction, which is formed by the extreme C-terminal portion of the 20 kDa domain, and that the fast reacting thiol is on the same side of the myosin head as the actin binding site (Sutoh, et al., 1984). Reports tend to suggest that the 50 kDa region is the least stable of the domains of the S1 head, for example, in thermal denaturation by differential scanning microcalorimetry (Levitsky et al., 1991), and in the high suceptibility of this region to chemical cleavage (Audermard et al., 1988, review). NMR studies of myosin (Highsmith et al. 1979) and S1 (Prince et al., 1981; Highsmith & Jardetsky, 1980), have shown that a substantial proportion of mobile nuclei are located within the S1 region of myosin, and that perhaps 20% of the S1 structure has some internal motion. The overall picture of S1 is that it is not a rigid structure in solution, but one that possesses both backbone mobility (most of the amide protons are readily exchanged with D2O) and sidechains of differing mobility (differentially broadened NMR resonances). This detected mobility of the myosin head may provide S1 with the capacity to adopt a variety of conformations.

The A2 light chain has a sequence almost identical to the 141 amino acid C-terminal sequence of A1, differing only in the region corresponding to the first 8 residues of A2. However, the A1 light chain possess an N-terminal extension of 42 amino acids, not found on A2, consisting mainly of alanine, proline, and lysine. The N-terminal alanine residue of A1 is post-translationally trimethylated (Henry et al., 1982; Henry et al., 1985). The A2 light chain is N-acetylated. Both the A1 and A2 light chains are coded for by the same gene and differentially spliced at the level of pre-mRNA to give rise to either A1 or A2 light chain mRNA (Nabashina et al., 1984). The N-terminal segment of the A1 light chain adopts a well-defined rod-like configuration as a consequence of the Ala/Pro rich composition, with a segmental motion somewhat independent of the remainder of the S1 head (Bhandari et al., 1986). In addition to a single alkali chain each S1 head has one phosphorylatable light chain which is maintained in papain S1 (although partially degraded from 19 kDa to 17 kDa).

The alkali light chains of myosin are bound to the 20 kDa domain (Mitchell et al., 1986). The S1 20 kDa fragment associated with an alkali light chain has been isolated (Chaussepied et al., 1986c). The N-terminal tri-methylalanine of the A1 light chain was shown to be less than 15Å from the SH1 thiol (Trayer et al., 1987). Crosslinking studies have have shown that the alkali light chains are also in close proximity (11Å crosslinker) to the 27 kDa domain probably near 23Arg (Labbe et al., 1986). Thus the alkali light chains appear to span two domains of the S1 head. The 177Cys of the alkali light chain has been located about 50-100Å from the head-rod junction using an avidin-biotin system in the electron microscope (Yamamoto et al., 1985). The loss of the LC2 in chymotryptic S1 places a key interaction site for these light chains at the C-terminus of the myosin head in the head/neck region. When myosin is partially cleaved by thrombin (in the presence of EDTA) a 97 kDa fragment corresponding to chymotryptic S1 with a small 2-3 kDa C-terminal extension is formed. This binds actin a little tighter than chymotryptic S1, still in an ATP dependent manner. Despite the extension of the C-terminus, LC2 are lost and hence the essential binding sites for the LC2 light chains are located further into the S2 neck region, as found in papain S1 (Kerrache et al., 1990). The location of the LC2 light chains is further confirmed by immunoelectron microscopy with antibodies specific for the N-terminal region of A1 which locate this segment toward the base of the myosin head (Winkelmann & Lowey, 1984; Waller & Lowey, 1985). The location of the LC2 light chains suggests that they may play some role in controlling rearrangements of the heavy chain domains, but they have also been implicated in the assembly of skeletal myosin filaments and in determination of filament length (Margossian et al., 1987). The light chains probably have a dumbbell shape, by homology to the troponin C and calmodulin (see section 1.12).

Over several years it has become apparent that there is a functional difference between S1A1 and S1A2 in their interactions with actin. The difference between the isoenzymes must arise as a result of the structural difference between the alkali light chains. The N-terminal extension of the A1 light chain has been shown to form a direct contact with the C-terminal region of actin (Trayer et al., 1987). The earliest differences between isoenzymes were found in kinetic assays, where the measured Km of S1A1 for actin was lower than that of S1A2, and in affinity chromatography where S1A1 was found to exhibit the greater affinity for monomeric and polymeric actin (Winstanley et al., 1977). At sufficiently low ionic strengths, S1A1 binds to F-actin more tightly and binding induces a stronger activation of its Mg2+ATPase, in comparison with S1A2 [at low ionic strength K+/EDTA- and Ca2+-ATPase activities do not differ between the isoenzymes, but the actin-activated Mg2+ATPase of S1A1 is 2-2.5 times as high as that of S1A2; Chalovich et al., 1984]. However, under physiological conditions the difference in actin activated Mg2+ATPase is lost, while S1A1 still binds more tightly to F-actin (Trayer & Trayer, 1985). It is likely that an additional contact point with actin in S1A1 is responsible for the tighter binding of this isoenzyme (see section 1.14).

The contact between the A1 light chain and actin has been implicated in the ability of S1A1 to promote actin polymerisation much more readily than S1A2. Both isoenzymes appear to form a tight 1:1 complex with G-actin, but polymerisation with S1A2 only takes place over the period of many hours, as opposed to a few seconds with S1A1 (Chaussepied & Kasprzak, 1989; Rane et al., 1991). A ternary complex of S1 with two actin monomers has been detected, in which the S1A1 formed a tighter complex than S1A2 (Rane et al., 1991). The interaction of the S1 head with two actin monomers is probably the first stage in the polymerisation reaction. It seems possible that if a major contact on one such monomer was the A1 light chain, then the promotion of polymerisation would be retarded in the absence of this contact by the destabilisation of the ternary complex.

An interesting series of experiments examining the thermal disruption of the S1 structure has identified the three domains of S1 as separate entities, the most labile being the 50 kDa domain. This labile domain is disrupted at a lower temperature in S1A1, indicating that there is a structural difference within the S1 head depending upon the light chain carried. Properties such as thermally induced aggregation were different between the isoenzymes (significant at around 30°C for S1A1 and 40° for S1A2), while temperature dependent inactivation of the ATPase was independent of the light chain (K+/EDTA-ATPase inactivation begins at around 34°C). The structural changes were however linked to the differences in ATPase, since differences in the thermal denaturation of S1 are only seen in fresh preparations, disappearing after several days storage, with a parallel loss in the difference in actin activated ATPase at low ionic strength (Levitsky et al., 1991).

The A1 light chain is found only in skeletal and cardiac muscle, with smooth and non-muscle myosins having only the A2 light chain. This may have some significance for the function of the acto-myosin system in various tissues.

Actin is the second major protein component of muscle and is an important part of the cytoskeleton. G-actin consists of a single polypeptide chain (375 residues, 42.3 kDa), with a nucleotide (ATP or ADP) and a divalent cation (Ca2+ of Mg2+) bound to a specific site. The sequence of actin from various sources have been determined (eg. rabbit skeletal muscle actin; Elizinga, 1973; figure 1.7.1). Interesting features among the 375 residues of the rabbit sequence are the acetylation of the N-terminal Asp, methylation of 73histidine, and the presence of 5 free sulphydryl residues (10, 217, 257, 285, 374) of which 374Cys is the most exposed and hence easily labelled residue. Actin is conserved throughout nature and the largest difference found is 25% between rabbit skeletal actin and that from Tetrahymena (ciliate; Hirono et al., 1987). Among the higher animals variation is significantly less. The small number of differences between skeletal, smooth and cardiac actin should be noted. No functional differences have been found between skeletal and cardiac actin, but some difference in ATPase activation by smooth and skeletal actins has been correlated to amino acid substitutions at residues 17 and 19 (Mossakowska & Strzelecka-Golaszewska, 1985). The G-Actin monomer polymerises in the presence of salt to form F-actin (Lazarides & Lindberg, 1974). Actin is involved in the microfilament and cytoskeletal structure of the cytoplasm. The regulation of the state of polymerisation of actin in the cytosol has been shown to involve a host of protein components. Hence, a pool of monomeric actin is maintained by actin binding proteins like profilin, gelsolin is involved in the severing, capping and nucleation of actin filaments in response to the requirements of the cell and networks of actin filaments are arranged into superstructures or bundles by filamin or villin (Way & Weeds, 1990).

F-actin is a filamentous complex formed from G-actin upon the addition of traces of salt. Polymerisation occurs at both ends of the filament (at different rates) and is accompanied by the hydrolysis of bound ATP. One of the triggers for polymerisation is thought to be the binding of metal ions (eg. K+) at a series of low affinity binding sites (Zimmerle et al. 1987). F-actin is depolymerized in the presence of bovine pancreatic DNase I by the formation of a tight 1:1 complex trapping the nucleotide and cation in the actin molecule, the biological significance of which is in doubt. This complex has been the basis of the most advanced crystallographic study of actin,. Another complex with profilin is also under study (Schutt et al., 1989). The structure of the actin:DNase I complex was determined by Kabsch et al. (1990) at 2.8Å. This structure is described briefly below (illustrated in figure 1.7.2).

Actin is a flat molecule (width 35Å) with a roughly square profile (side 55Å). There are two nearly equally sized domains called the 'small' and 'large' domains between which the nucleotide and divalent cation bind, thus contributing to the stability of the structure. The small domain contains both the N and C-termini and is split into subdomains, 1 and 2 (residues: 1-32, 70-144, 338-372 subdomain 1 and 33-69 subdomain 2) while the large domain is split in subdomains 3 and 4 (residues: 145-180, 270-337 subdomain 3, and 181-269 subdomain 4). A degree of similarity in the structure of the domains suggests that they may have arisen from gene duplication. Both subdomains 1 and 3 consist of a 5 stranded ß-sheet formed from a ß-meander and a right handed ßαß-unit similar to that seen in hexokinase. The ß-sheet is surrounded by five solvent exposed helices in subdomain 1 and three helices in subdomain 3. One of these helices (338Ser-348Ser) in subdomain 1 has solvent accessible hydrophobic residues which may be involved in protein-protein contacts either with myosin or tropomyosin. Subdomain 2 consists of a three stranded anitparallel ß-pleated sheet with a helix contacting the two edge strands, while subdomain 4 contains a two stranded anti-parallel ß-pleated sheet and four helices. The four subdomains are held together mainly by the nucleotide and by salt bridges.

The bound nucleotide fits into a pocket between subdomains 1 and 3. Specific contacts are found for the ribose ring and the phosphate groups. The bound Ca2+ ion (probably Mg2+ in the cell) is associated with the ß-phosphate. Occupancy of this site by a divalent cation may promote polymerisation by reducing the net negative charge of the actin monomer. The interaction between actin and DNase I in the complex involves residues 39-45, 61 & 63 (subdomain 2), with minor contact points at residues 207Glu & 203Thr (subdomain 3). Residues 42Gly, 43Val and 44Met are incorporated into a parallel sheet type contact with DNase I. The structure of DNase I is distorted at the contact points by as much as 1.8Å in the Cα positions and its seems likely that the structure of the binding loop of actin is similarly distorted by the contact. In the profilin complex the contact point is in subdomain 1 near 364Glu (Vanderkerckhove et al., 1989) and so an undistorted view of subdomain 2 could soon be available. The close identity of the structures of the ADP- and ATP-actin DNase complexes may also be an artifact, with DNase binding across the top cleft locking the G-actin structure in to one conformation. Recent evidence (Janmey et al. 1990) suggests that ADP-Actin can be polymerised to form a loose filament, and hence it is likely that the hydrolysis of ATP during polymerisation will change the conformation to provide a more rigid polymer. However, the large number of interactions between actin monomers in the filament (see below) would suggest that cooperative allosteric interactions between actin subunits might be a means of change between the two filament states. However any changes in structure within the actin monomer, and particularly within the small domain, are expected to be minor. For example, Miki (1991) has recently shown that the binding of S1 (or DNase I or the Tn/Tm complex or polymerisation) to actin do not appear to cause any great internal rearrangement of the actin domains, with fluorescent labels within the subdomains 1 and 2 not changing their separation during these interactions.

The regions in the actin molecule with large temperature factors are 1-7, 95-100, 183-276, 306-329 and 348-373. Uncertain regions are 1-6, 232-235 and 366-373. The C-terminal residues 373-375 are not present in the crystalline preparation.

F-actin cannot be crystalised, but can be oriented as a gel in capillary tubes to yield X-ray fibre diffraction patterns of up to 8Å resolution. Holmes et al. (1990) have produced a model of the actin filament based on this data, by fitting it to all the possible orientations of the G-actin monomer. Refinement of the model involved allowing the subdomains to move relative to each other. A maximum movement of 6° was modelled for subdomain 2 with the other domains only moving by 2-3°. The close similarity of the F-actin model and the structure of the actin:DNAse complex could mean that G-actin has a very similar structure to F-actin, or that the contacts (some of which mimic the subunit contacts in F-actin) with DNAse hold the G-actin in an F-actin-like conformation.

The model of the filament is illustrated in figure 1.7.3. The filament is 90-95Å in diameter, with the large domain in the centre of the filament and the small domain on the surface and readily available for binding to other proteins. The filament forms a two stranded left-handed helix. Each of the 4 subdomains of the actin subunit, are involved in interactions with adjacent actin monomers. Each actin subunit in the filament interacts with 4 other actin monomers. There are a large number of subunit-subunit interactions involved in stabilizing the actin filament. Interactions along the helix axis are 322-325 with 243-245; 286-289 with 202-204; 166-169 & 375 with 41-54. Contacts along the genetic [genetic = diagonal whereas longitudinal is long pitch of the helix] helix axis are between residues 110-112 and 195-197. Strong contacts may also exist between loop 262-272, which was rebuilt to form a ß-sheet finger which pokes into a hydrophobic pocket formed by 166, 196, 171, 173, 285, 63, 64, 40-45. This latter interaction is believed to be the heart of a hydrophobic core holding together the filament, with other electrostatic interactions and hydrogen bonds contributing to stability.

Much evidence of the structure of F-actin has been accumulated by biochemical techniques, and agrees largely with the model. Holmes et al. (1990) summarise the evidence and state that it would have been sufficient to determine the orientation of the actin monomer in the filament without recourse to the X-ray diffraction pattern of oriented actin. Many types of solution data had implicated the amino acid sequences found in subdomains 1 and 2 as being those involved in the interaction with several actin-binding proteins (see later).

The groups of Mannherz and Weeds (Cambridge; Abstract Alpbach muscle meeting 1992) have recently crystalised G-actin in a complex with gelsolin segment 1 (residues 1-130). A comparison of the actin structures in the two complexes did not show any gross change in the relative positions of the four subdomains. Interestingly there was little evidence for electron density in the DNase I binding loop (residues 40-50), indicating that this region is poorly structured in the absence of inter-molecular contacts (eg. those formed in F-actin or with DNAse I). In addition the C-terminal residues were not cleaved from the actin in this complex and hence the structure of this region has been identified.

The S1 head possesses a single nucleotide binding site composed entirely of the heavy chain. The alkali light chains are not required for ATP binding or hydrolysis (Sivaramakrishran & Burke 1982; Audemard et al., 1988). The myosin head has three distinct triphosphatase activities, a monovalent cation activity (K+ or NH4+/EDTA), a divalent cation activity (Ca2+ or Mg2+), and an actin activatable Mg2+ATPase activity. All three domains of the S1 head are thought to be involved in forming the ATP binding site. Much of the evidence has comes from the crosslinking of reactive analogues of ATP (affinity and photoaffinity labels) to the S1 heavy chain. Different regions of the ATP molecule have been placed close to selected regions of the S1 head, such that the adenine ring contacts at least the 27 & 20 kDa domains, and the ribose group contacts the 50 & 27 kDa domains. [A photosensitive group on the adenine moiety was shown to bind to the 27 kDa fragment (130Trp; Szilagyi et al., 1979; Okamoto & Yount, 1985). Photosensitive groups on the ribose moiety were shown to bind to the 50 kDa (324Ser; Mahmood & Yount, 1984), or to the 27 kDa fragment (within 10 residues of 130Trp, Sutoh, 1987). A bifunctional crosslinking analogue of ATP, modified on both the adenine and ribose positions, was able to crosslink the 27 and 20 kDa domains (Martuka et al., 1990). An analogue of ATP with the adenine ring in predominantly the syn conformations (mimicking ATP) was crosslinked to S1 in the 20 kDa domain (within 50 residues C-terminal of 660Leu), with an analogue forming anti-conformation crosslinked to the 27 kDa domain (Martuka et al. 1989)]. This appears to place the ATP binding site at a confluence between the domains, such that ATP binding might affect the relative orientations of the domains. Indeed some data has been obtained which suggests that the S1 head changes shape in the presence of Mg2+ATP, presumably influencing the orientation of the domains. For example, electron microscopic studies (Tokunaga, et al., 1991) and low angle X-ray scattering studies (Wakabayashi & Tokunaga, Alphbach 1992) indicate a contraction of the S1 head in the complex S1.Mg2+ATP and a re-arrangement in the neck region of the S1 structure. In passing it is worth noting that many changes in the susceptibility of proteolytic sites in skeletal MHC have been found in the presence and absence of ATP (Yamamoto, 1989). Protease sensitive sites are revealed in the region 28-32 residues from the N-terminus of skeletal and smooth muscle from various sources, upon addition of ATP, and the tryptic susceptibility of the 27 and 50 kDa fragments is affected by the binding of nucleotide, promoting their degradation to a 22 and 45 kDa fragments respectively (Muhlrad & Hozumi, 1982; Mornet et al., 1985). The significance if these changes is still not clear.

The ATP binding site on the myosin head has been located in relation to other landmarks within the structure. In one of the best experiments, Tokunaga et al. (1987) photoaffinity labelled S1 near 130Trp with an ATP analogue similar to Sutoh (1987, see above), but containing a biotin group attached by a six membered methylene spacer chain to the adenine ring. The biotinylated ATP could be visualised in the electron microscope, when bound into the myosin active site, using avidin, and the S1 was still able to interact with F-actin. Three dimensional image reconstruction from electron micrographs revealed that the ATPase site of the myosin head was about 5 nm from the tip of the myosin head, and 4 nm away from the actin-binding site of S1, 14nm from the head-rod junction (Sutoh et al, 1986), and placed the actin binding site and the ATPase site on opposite sides of the myosin head. Fluorescence resonance energy transfer (FRET) has been used to measure the distance between fluorescent ADP derivatives bound to the S1 active site and the SH1 thiol labelled with fluorescent reagents. Despite inconsistencies in the data from various sources it is clear that the nucleotide binding site and the reactive thiols are quite close in space. [Tao & Lamkin (1981) and Cheung et al. (1985) estimate the distance between an analogue labelled on the ribose moiety of ADP and a fluorophore on the SH1 group to be 4.0-1.5 nm, Aguirre et al., (1988) found that the distance separating etheno-ADP and SH1 to be between 2.0-3.8 nm, and Perkin et al. (1984) entrapped etheno-ADP in the active site by crosslinking the SH1 and SH2 thiols with a fluorescent crosslinker, and estimated the distance separating the two fluorophores as 2.3-2.6 nm.] This data is at odds with that of Sutoh et al. (1986) which places the actin site and nucleotide binding sites on the opposite side of the S1 head, which is interpreted (see figure 1.5.1) as indicating that the ATP binding cleft probably traverses the whole of the S1 molecule.

The most important nucleotide binding site on S1 is the phosphoryl site. The sensitivity of myosin to ATP hydrolysis must be linked to the structural contacts of the phosphate moieties of ATP. Some localisation of this phosphoryl subsite has been made. It is possible to trap Mg2+ADP into the active site of S1 with a vanadate ion (which acts as a gamma phosphate analogue, Muhlrad et al. 1991) and then to cleave the S1 heavy chain in close proximity to the vanadate by irradiation with uv light. Cleavage occurs at 23 kDa (180Ser), 31 kDa, and 74 kDa away from the N-terminus of the S1 heavy chain (Grammer et al., 1988; Cremo et al., 1988). The 23 kDa site occurs in the middle of the glycine rich conserved ATP binding consensus sequence on the 27 kDa domain (see below). Specific photocleavage at this site inhibits all the of the S1 ATPase activities (Muhlrad et al. 1991). In contrast, cleavage at the 74 kDa site (corresponding roughly to the C-terminal 6 kDa of the 50 kDa domain of the S1 heavy chain) only results in a decrease in the actin-activated Mg2+ATPase activity. The 23 kDa site may just be involved in binding the terminal phosphate of ATP, whilst the 74 kDa site which forms a different part of the same phosphate binding site, must also play a role regulating the myosin-actin interaction. Other evidence for the involvement of region around the 50-20 kDa junction comes from the inhibition of the actin-activated Mg2+ATPase when the site is specifically cleaved by trypsin, resulting from a decrease in actin affinity (the NH4+/EDTA-ATPase activity remains intact; Mornet et al. 1981a; Botts et al, 1982). Covalently crosslinking a peptide to this connector region between the 20 & 50 kDa domains, had no effect on the intrinsic NH4+/EDTA- and Ca2+-ATPase activities of S1, but considerably inhibited the actin-activated Mg2+ATPase activity of S1 (Chaussepied & Morales 1988). The tryptic digestion of the 50 kDa fragment into a 45 kDa derivative only in the presence of ATP may be related. It thus appears that the ATPase site may be in close communication with the 20-50 kDa junction and thence with actin. Another site in the 50 kDa domain has been implicated in the phosphoryl subsite. Thrombin cleavage of the S1 heavy chain (560Lys-561Ser, Chaussepied et al., 1986a) inactivated all the S1-ATPase activities, without abolishing the binding of Mg2+ATP (decreased affinity). The data suggests that this region of the 50 kDa domain may be important in the tight binding and hydrolysis of nucleotide.

Sequence analysis in recent years has revealed several sequence homologies between nucleotide binding proteins, suggesting that certain regions of these proteins have a common function in binding and hydrolysis of nucleotide. The region of S1 which bears the best homology to the "glycine-rich" consensus sequence lies between residues 178-185 in the 27 kDa domain, although a similar site is found at the 50-20 kDa junction between residues 632-638 (GXXXXGK; Fry et al., 1986; , Walker et al., 1982). Another homology has also been identified between the sequence from 702Glu-719Asp in the 20 kDa domain containing the SH1 thiol, and the triphosphate binding site consensus sequence seen in F1-ATP-synthase enzymes and adenylate kinase (Burke et al., 1990; Eto et al., 1990). These predictions serve to focus attention on the possible ATP binding regions of S1, and to provide clues as to the conformations of these regions by comparison to solved crystal structures (eg. phosphoglycerate kinase, adenylate kinase). The common motif among such nucleotide binding proteins is the nucleotide fold, which consists of a conserved rigid glycine-rich loop, held between α helix and ί sheet (Walker et al., 1982), with the α and ί phosphates of the nucleotide fitting into an anion hole which is formed by the backbone of the residues of the glycine rich consensus sequence (Saraste et al., 1990). In the absence of a crystal structure of S1, such comparisons will in the future be the starting points for structural studies of the nucleotide binding site in the S1 head. Recent expression systems for S1 should allow the conserved lysine residue to be identified by mutation studies (mutation of the conserved lysine lowers the catalytic activity of adenylate kinase without abolishing nucleotide binding; Saraste et al., 1990), and perhaps the identification of residues in this region in NMR studies of 15N labelled S1 (the backbone amide of the conserved lysine is shifted downfield to 10.5-11 ppm by hydrogen-bonding to a phosphate oxygen in N-ras p21 and elongation factor; Lowery et al., 1990).

There are indications that the hydrolysis of ATP leads to changes in the S1 structure. The electron microscopic (Tokunaga et al., 1991) and low angle X-ray scattering studies (Wakabayashi & Tokunaga, 1992) mentioned above indicate a contraction of the S1 head in the order S1.Mg2+ATP < S1.ADP.Pi < S1.ADP+Pi or S1.ADP < S1 or S1-pPDM. It should be noted that many other groups have reported that there are no large changes in the head shape in the presence or absence of Mg2+ATP or Mg2+ADP (eg. Vibert, 1988), and there are no gross changes in the NMR spectra of either myosin or S1 when Mg2+ATP is added (Highsmith et al., 1979; Prince et al., 1981). However, as described in the next section, subtle changes can be detected upon nucleotide binding, by reagents which crosslink different areas of the myosin molecule.

In summary, it appears that the nucleotide binding site on S1 resides in a region where all three domains are closely apposed. It seems likely that this would require the nucleotide to bind in a cleft between the domains in a similar manner to other ATP binding proteins (phospho-D-glycerate kinase, Bryant et al., 1974; actin, Kabsch et al., 1990; adenylate kinase, Schulz et al., 1974), although the gross domain movements seen in phosphoglycerate kinase are probably not present in S1. Such a structure may explain how the loss of a single phosphate group from bound nucleotide can have a profound effect upon the actin binding affinity of a large protein like S1. Bound nucleotide could readily regulate the contacts between the domains and so enable them to form a binding or non-binding conformation, especially since the phosphoryl subsite may be located at the junction of several domains. It is significant that all three domains of S1 also are involved in binding to actin.

The two principal functions of the S1 head are the hydrolysis of ATP and the consequent change in the actin binding affinity of S1. A dissective approach has been used to identify the relationship between the elements of these processes, in which the S1 head is subfragmented and the functional abilities of these fragments assessed. The object was the definition of the smallest region of the heavy chain capable of acting as an intercommunicating system between actin and the polyphosphate chain of ATP.

A C-terminal 30 kDa fragment has been isolated with associated alkali light chain (Chaussepied et al. 1986a & 1986b; thrombin cleavage at 560Lys-561Ser to yield 68 & 30 kDa fragments). The fragment was found to bind to actin with the affinity of the interaction regulated by nucleotide (affinity was decreased 10 fold in the presence of ATP and PPi, but only to a much lesser extent by ADP; Chaussepied et al., 1986b). A similar result was seen for the renatured 26 kDa C-terminal/light chain fragment reported by Griffiths & Trayer (1989; cleaved between 600Asp-601Pro with formic acid). In contrast the isolated and renatured C-terminal 20 kDa domain fragment of S1 with associated alkali light chain was found to bind to actin in an ATP and PPi independent manner (Chaussepied et al., 1986b). These data suggest that the 26 kDa fragment is the minimal fragment containing the crucial functions of S1, that is an actin binding site capable of regulation by nucleotide, and a nucleotide binding site capable of distinguishing ATP from ADP. Since all three fragments contain the entire 20 kDa region, a crucial site for S1 function must to be located between the 20 kDa and 26 kDa C-terminal digestion sites (between residues 601-643). It is interesting to note that ATP interacts with the 30kDa-LC1 and 20kDa-LC1 complexes, and is displaced by the addition triphosphate (Bonet, et al., 1990). The report of Okamoto & Sekine (1987) of an N-terminal 75 kDa fragment of S1, which lacks the 20 kDa domain, but also binds to actin and is able to hydrolyse ATP in an actin activated manner is somewhat dubious since recent reports (and my own experience in trying to reproduce their data) suggests that the preparation was contaminated with sufficient intact S1 head to account for the observed activity. The isolated and renatured N-terminal 27 kDa domain, and a smaller N-terminal 21 kDa fragment have also been found to bind to actin in an ATP dependent manner (Muhlrad, 1989; Muhlrad, 1990).

The region of the 20 kDa domain around the two reactive thiols, SH1 and SH2, has been shown repeatedly to be a central point in the structural changes occurring when nucleotide binds to S1. The SH1 is readily and specifically labelled with a large number of relatively hydrophobic thiol directed reagents. Labelling of this SH1 results in a complete inactivation of the NH4+/EDTA-ATPase activity and a fourfold increase in the Ca2+-ATPase. The ATPase activities of S1 are altered in the same way if the SH2 thiol alone is specifically modified. However, if both the SH1 and SH2 thiols are labelled, then all the S1 ATPase activities are inactivated (Reisler et al., 1974). Modification of the SH1 thiol leads to a reduction in rigor binding of S1 to actin and an enhancement of the binding to actin in the presence of ATP (Greene & Eisenberg, 1980; NEM-modified S1 has a 2.6 fold higher affinity for actin in the presence of ATP than unmodified S1, Root et al. 1991). The ability of ATP to dissociate actomyosin is greatly reduced when the region around the SH1 is selectively cleaved (Chaussepied et al., 1986). The addition of Mg2+ADP to S1 is observed to partially protect the SH1 thiol from labelling, whilst greatly increasing the susceptibility of the SH2 thiol to various reagents (Reisler et al., 1974; Sekine & Kielley, 1969). The same characteristics are maintained in muscles fibres, where the SH2 group can be specifically labelled taking advantage of the increased reactivity of the SH2 in the presence of Mg2+ADP and the decreased reactivity of the SH1 in the presence of actin (Ajtai & Burghardt, 1989; Duke et al., 1976).

There is much evidence that there are changes among all of the domains of S1 upon binding of nucleotide and that many of these structural changes are clustered within close proximity to the reactive thiols. The idea that the ATPase site is close to the reactive thiols thus presents itself. In the presence of Mg2+ATP, a conformational change on the myosin head occurs, resulting in the SH1 and SH2 thiols being brought together. The thiols can be crosslinked with a variety of thiol crosslinking reagents such as pPDM. In the absence of nucleotide no crosslinking is seen. The distance between the SH1 and SH2 thiols can be as much as 14Å, depending in the length of the crosslinking reagent used (Wells & Yount, 1982), and under certain conditions disulphide bridge can be formed between directly between these thiols. It must also be significant that when the thiols are crosslinked, the nucleotide remains trapped in the S1 nucleotide binding site, and the myosin head has a much reduced affinity for actin. A pincer-like movement of domains around the nucleotide with the region of the reactive thiols as a hinge is a tempting picture. Since pPDM.S1-nucleotide binds to actin with a similar affinity to S1.ATP and has essentially the same binding characteristics, it is considered to be a long lived trapped analogue of S1.ATP (pre-power stroke conformation, ATP or ADP,Pi bound, 90° crossbridge; Trayer & Trayer, 1988). A similar trapped weak actin-binding pre-power stroke conformation can be formed when Mg2+ADP is trapped into the active site of S1 with vanadate (Goodno, 1979; Goodno & Taylor, 1982). Vanadate appears to be a phosphate analogue that adopts a stable pentavalent-bipyrimidal geometry allowing it to mimic the geometry of the gamma phosphate of Mg2+ATP during the hydrolysis of Mg2+ATP to Mg2+ADP,Pi (transition state; a further stable complex with tetrahedral BF3- mimicking c-phosphate in ATP prior to hydrolysis or the bound Pi following hydrolysis can also be formed).

The reactive thiols can be crosslinked to various areas of the S1 head and the crosslinking is often nucleotide dependent and inhibits ATPase activity and traps nucleotide. Figure 1.9.1. summarises the data, which implies that both the SH1 and SH2 move with respect to sites in the 27 & 50 kDa domains of S1 in response to nucleotide binding. The crosslinking of the SH2 thiol to either 184Lys or 189Lys of trypsin digested S1 (3-4.5Å span; Sutoh & Hiratsuka, 1988; independent of the presence of nucleotide) confirms the proximity of the glycine rich loop (residues 178-184) in the 27 kDa domain which may be involved in ATP. In interpreting crosslinking experiments it must be remembered that a protein may have several different structures in solution. A particular crosslinking agent spanning a given distance may only react with one particular protein structure in which the reactive groups are correctly apposed. On the other hand, physical methods such as FRET probably only measure weighted-average distances. Several such distances measured by FRET are included in the figure 1.9.1.

The changes in the thrombic nicked S1 head are of interest. When the SH1 and SH2 thiols are crosslinked by pPDM, a local conformational change is induced in the 50 kDa region allowing thrombin to cleave within this region (Chaussepied et al., 1986a & 1986b). The nicked species is unable to bind Mg2+ATP irreversibly, unable to form a stable ADP-vanadate complex, has virtually no ATPase activity, rigor binding of actin is much weaker, and EDC will crosslink to the 20 kDa region but not to the 50 kDa region. Mg2+ATP does still dissociate the acto-nicked S1 complex. The ATPase cycle is disrupted and appears to be unable to undergo the transition to the stable M.ADP.Pi state, and the tight binding state is hence apparently disrupted.

Finally, it has been found that in the presence of nucleotide the crosslinking of alkali light chains to the 27 kDa domain was markedly reduced (Pliszka, 1990), indicating a movement of the light chains and possibly the associated region of the 20 kDa fragment away from the 20 kDa domain.

In the above description, the reciprocal effects of SH1 modification, actin binding, nucleotide binding have been demonstrated. The data suggest that the SH1 region is important in the coupling of nucleotide and actin binding. The 20 kDa thiol modification experiments suggest that there is a conformational change in the 20 kDa region upon ATP binding. There must be considerable flexibility in the SH1-SH2 region. These residues are only separated by ten peptide bonds, with the maximum extended length is close to the 29Å measured by Botts et al. (1984; figure 1.9.1), and the separating can shrink to the minimum value where a disulphide bond is formed. The components of this process, nucleotide sensitive actin binding, may reside in the 26 kDa C-terminal fragment, and may coalesce to a single site in space. This fragment is unable to perform a competent ATPase cycle and hence as expected, the remainder of the molecule is involved in the process.

In order to understand muscle contraction at the molecular level, it is essential to localise the interfaces in the acto-S1 complex, both before and after the power stroke. Three methods have been used repeatedly to localise these sequences, EDC crosslinking, use of antibodies directed against sequences in the actin molecule, and the use of peptides corresponding to the actin sequence.

The water soluble carbodiimide reagent EDC acts as a zero-length crosslinking reagent, but the method relies on the availability of a convenient pair of lysine and carboxyl groups placed close together in the two proteins. This introduces the possibility of trapping infrequent contacts which have the required reactive groups available, and excluding the principle binding sites where such contacts are not available. A recent example of the problems associated with crosslinking reagents is the crosslinking of caldesmon to both the N- and C-termini of actin, which are close in the actin structure. It appears that the site at the C-terminus of actin is spurious (Crosbie et al., 1991; Graceffa et al., 1991). Even so this method remains an important first step in localisation of reciprocal sites in any interface.

Antibodies have been used to block putative interaction sites on actin, but some possible problems arise because of the relatively large size of the molecule (Fab about 50 kDa), which results in the possibility of indirect steric hindrance of binding.

Peptides from one protein have been used to map the interaction on a second protein. Although most peptides do not assume their native protein-like structure in solution, it likely that the sequence contains sufficient information for the peptide to adopt this structure in the presence of a suitable template, that is when bound to the protein partner. In this state the peptide should be capable of very specific blocking of interactions, and may trigger conformational changes in the protein template. However, the inherent problem with peptide binding is the possibility of a non-specific interaction of the peptide with the protein template, which could have the same characteristics as those anticipated for a native-like interaction. Where the binding site between proteins is split between several non-contiguous regions no single peptide will mimic the all of the properties of the intact protein.

The principal region that has been implicated in the binding of S1 to actin is that containing the N-terminal acidic residues of actin. Sutoh (1983) found that these residues of actin (1-11) were crosslinked by EDC to sites in the 50 & 20 kDa domains of S1. These interactions have also been picked out by other means, for example using other crosslinking reagents (Bertrand et al., 1988), mutation of specific residues in expressed actin to eliminate the binding site (replacement of 3&4Asp with Lys; Aspenstrom & Karlsson, 1991), observation of interactions at labelled sites in NMR (residue 10Cys labelled with 19F is broadened by interaction with S1; Barden et al. 1989), and by the observation of broadening in residues in the amino terminal region by 1H-NMR (Moir & Levine, 1986; Moir, et al., 1987).

There have been indications that this N-terminal segment of actin is important in the interaction with S1 only in the presence of nucleotide, but that it contributes little to the rigor (strong) binding state between actin and myosin (Chaussepied & Morales, 1988; Bertand et al., 1989). The region would hence be a contact site in the weakly bound acto-S1 complex. For example, Miller et al. (1987) used an antipeptide antibody to residues 1-7, which in the rigor complex (absence of nucleotide), did not interfere with the binding of S1 to actin, and indeed the antibody and S1 appeared to be able to bind to the same actin monomer (DasGupta & Reisler, 1992). In the presence of ATP (and ADP or PPi), these antibodies to the first seven N-terminal residues of actin were able to inhibit the binding of S1 to actin (DasGupta & Reisler, 1989, 1991, 1992). Other evidence also seemed to confirm that the region was important in the interaction of S1 and actin in the presence of nucleotide. For example, the introduction of ATP induces a shift among the crosslinking lysine residues of S1 (sequence 632-642) in their interaction with the N-terminus actin (Yamamoto, 1990). However, doubts were cast on the importance of this region as a direct contact site, when it was found that the lost binding of S1 to actin was recovered at high concentrations of added S1, without displacement of the Fab fragment (DasGupta & Reisler, 1992). It appears that the antibody may be inducing a conformational change in the actin filament which regulates binding at remote sites, and that recovery of binding is due to interaction with actin monomers not occupied by Fab. Hence, it was not surprising when it was found that the inhibition of actin-activated ATPase by this Fab fragment does not recover in parallel to the binding of S1 (DasGupta & Reisler, 1992). In this case, the inhibition of acto-S1 ATPase does not require the displacement of S1 from actin, and so the bound S1 can be inhibited by the binding of Fab to the N-terminus of actin at remote sites. While the data at present does not indicate that the Fab fragment and S1.nucleotide complex can occupy the same actin monomer, and hence does not exclude an interaction between S1 and actin in the N-terminal region of actin, the binding here may be only of minor importance, and the site may instead be important for regulating conformational changes in the actin filament which are transmitted indirectly to S1. It is noteworthy that the regulatory proteins TnI (Levine et al., 1988) and caldesmon (Levine et al., 1990a; Adams et al., 1990) interact with the N-terminal segment of actin. The antibody may be reproducing a comformational change in the actin filament which is important during regulation. It is worth noting that this region has been crosslinked to many other proteins, such as gelsolin (Rouayrenc et al., 1986), fragmin (Sutoh & Hatano, 1986) and others, and the N-terminal portion is particularly flexible and could readily move into the vicinity of any protein bound in general area. The N-terminus is also highly negatively charged, and hence taking in to account all these factors, it is likely that interactions detected in this region could be relatively non-specific.

The implication of the above data is that at least in the rigor state, and perhaps throughout the ATPase cycle, the acto-S1 interaction must be dominated by contact sites other than residues 1-11 of actin. Some indications of the other contact sites on actin have been seen. Mejean et al. (1986, 1987) identified a second antiprotein antibody recognising an epitope in the region 18-28, which completely blocked the interaction of actin and S1. The mutation of residues 24&25Asp to His in Dicotstelium actin has implicated this region in binding since the maximum turnover rates (Vmax) for actin activated ATPase are diminished, but as above the effect is divorced from actin binding which is not significantly reduced (Sutoh et al., 1991). The peptide 1-28 of actin was shown to form a 1:1 complex with S1, to be displaced from the complex by actin, and to activate the ATPase activity of S1 in a manner similar to F-actin, up to a maximal activity at a peptide:S1 ratio of 0.5:1 (van Eyk & Hodges, 1991). The ability to mimic a portion of the actin function in the absence of the filament structure does indicate direct binding in this region and so there may be a site at the C-terminus of the region 1-28.

A further site was indicated by the crosslinking of the rigor complex with EEDQ (carboxyl directed zero-lenght crosslinking reagent), and with glutaraldehyde to both the regions 1-28 and the region 40-113 of actin (Bertrand et al., 1988). This site is consistent with the pattern of protection of G-actin from limited proteolysis by the binding of S1A2, which identified the segment 61-69 as protected from tryptic cleavage (Chen et al., 1992) and with the mutation of the residues 99&100Glu to histidine, which results in a partial loss of actin-activated ATPase activity.

One site which can be excluded from the interaction appears to be the loop 39-52, which in involved in DNase I binding. This loop is not protected from enzymatic digestion in the presence of S1, and a ternary DNAse I-G.actin-S1A2 complex can be formed with only a slightly diminished affinity of S1A2 for G-actin (H.Trayer, this lab).

Finally, the C-terminal residues of actin (360, 362 & 363) were shown to crosslink to the A1 light chain (Sutoh, 1982).

The earliest indication of the location of the actin-binding sites on S1 was the observation that in the presence of actin, that the connector region between the 20 & 50 kDa domains was protected against trypsin proteolysis, in absence of protection at the junction of the 27 & 50 kDa domains (Mornet et al., 1979). Many subsequent investigations relied upon the use of EDC to crosslink S1 and actin. Mornet et al. (1981), found that in the rigor complex both the 20 kDa and 50 kDa domains of S1 crosslink to actin, and subsequently Sutoh (1983) localised these sites to the N-terminal region of the 20 kDa domain (between residues 633-663 of the S1 heavy chain) and the C-terminal region of the 50 kDa domain (between residues 550-587). Under optimal conditions in the absence of nucleotide, the rate of EDC crosslinking of actin to the 20 kDa fragment was three fold greater than at the 50 kDa. Under all conditions the 20 kDa site was heavily favoured and thus appears to be the high affinity site (Chen et al., 1985). Similar 50 and 20 kDa sites also appear to be present in smooth, molluscan, and non-muscle myosins (Marianne-Pepin et al., 1985; Labbe et al., 1986; Atkinson & Korn, 1986).

Myosin peptides from around the reactive thiols have implicated this region of the S1 head in binding to actin. The myosin peptide 702-708 binds to F-actin and completely inhibits the acto-S1 ATPase (Suzuki, 1987; Eto, 1990; Suzuki, 1990). Keane et al., (1990) have located a sequence on the 20 kDa domain of S1, involving residues 695-725, which binds strongly to actin. A variety of peptides in this region (but not in the flanking regions), were found to bind to actin, in both NMR experiments and by direct binding studies. Furthermore, peptides from this region inhibited the S1 actin activated Mg2+ATPase. Studies carried out using muscle fibres showed that these peptides would also compete with myosin crossbridges for actin, when the myosin head was in a strong actin-binding conformation (ie. post-power stroke intermediate), but not when the myosin head was in a pre-power stroke conformation (Keane et al. 1990). These studies suggest that the region around the SH1 group on the 20 kDa domain the myosin head is involved in the strong rigor interaction between S1 and actin.

The region around the SH1 thiol has been implicated in actin binding by other data. A 19F probe on the SH1 thiol is broadened greatly by the addition of actin, whereas the same label on the SH2 thiol is much less broadened (Barden et al., 1989), and a nitroxyl label on the SH1 group places this site within 15Å of F-actin (Keane et al., 1990). Recently the SH1 has also been crosslinked to actin using the crosslinking reagent MBS (m-maleimidobenzoyl-N-hydroxysuccinimide ester; forms crosslinks Lys-Cys and Lys-Lys). MBS was used to treat G-actin, forming intra-molecular crosslinks, which appears to lock the actin in a G-actin conformation which is somewhat resistant to polymerisation (Bettache et al., 1989), although it does form stable polymers in the presence of phalloidin. MBS-actin has a free reactive group available for reaction with either a cysteine or lysine, and it is this which has been used to crosslink S1 stociometrically to either the G- or F-actin forms. Under various conditions the crosslinking can be exclusively to the 50 kDa (Bettache et al., 1989), or to both the 50 kDa and the SH1 in the 20 kDa domain (Bettache et al., 1992). This places the SH1 thiol about 10Å from the crosslinking site on actin (so far unidentified). The SH2 thiol is not crosslinked, so this places it further away in the MBS-acto-S1 complex. In contrast to the above data, the attachment of a bulky 68 kDa avidin moiety to SH1, did not inhibit the cosedimentation of labelled S1 with F-actin (Sutoh et al., 1984).

Peptide mimetic studies have shown a peptide from the C-terminal region of the 50 kDa domain of S1, corresponding to 610-616 (and possibly C-terminal to this sequence as well) binds strongly to actin (Trayer et al., 1991; this thesis).

All 3 tryptic domains of S1 have been isolated and renatured, and all of them have been found to bind to actin [20 kDa (Muhlrad & Morales, 1984); 50 kDa (Muhlrad et al., 1986); 27 kDa (Muhlrad, 1989)]. A smaller N-terminal 21kDa fragment (1-180) also binds to actin (Mulhrad, 1990). The 20 and 27 kDa domains bind strongly to actin (Kd=1 uM & 0.1 uM respectively), whereas the 50 kDa has only a very weak affinity (Muhlrad & Morales, 1984).

The renatured 27 kDa fragment cosediments with F-actin, forming a tight complex which is competed out by S1, which is disrupted at high ionic strength, and is dissociated by ATP, ADP & PPi binding. The interaction of neither of the other tryptic fragments (20 & 50 kDa) were weakened by the addition of Mg2+ATP. The fragment binds to F-actin when immobilised on a nitrocellulose membrane, as does the subfragment 130-240 (Mulhrad, 1989). The finding that the N-terminal 27 kDa domain, interacted with actin was unexpected, since no crosslinking studies have indicated the participation of this domain with actin, but this may be as a result of the lack of a suitable crosslinking pair of residues. However, F-actin does inhibit thermolysin and papain cleavage at the 27-50 kDa junction (Applegate & Reisler, 1983). From the high sensitivity of the antipeptide antibody binding to ionic strength an electrostatic interaction was postulated and has been suggested to arise from the interaction of a conserved highly positively charged region (143RGKKR147Q) with the negative N-terminus of actin (Dan-Goor & Muhlrad, 1991). This interaction has been investigated with a peptide corresponding to this region and with an antibody raised against this peptide and reacting with the intact 23 kDa region. The interaction of the antibody with the surface exposed region corresponding to the peptide in S1 inhibits the interaction with actin, and the antibody reduces the actin-activated ATPase activity of S1 without any effect on the intrinsic ATPase activities of the S1 (Mg2+ or K+(EDTA)-ATPase activities). The peptide cosediments with actin (Kb=105 M-1), and is crosslinked to F-actin by EDC. Hence this may form one of the binding sites for actin on S1.

An "antipeptide", has been designed to specifically crosslink to the lysine rich 50-20 kDa junction region of S1, between residues 633 and 642 (Chaussepied & Morales, 1988). This connector region has been mentioned earlier in this section.

Myosin 633 642

G G K K G G K K K G

Peptide Cys G G D D G G D D D G

This peptide was covalently attached to the target, with no effect in the intrinsic ATPase activities or ATP binding (Chaussepied & Morales, 1988) and likewise binding of nucleotide did not change the rate of the crosslinking reaction. The crosslinking did not modify thiol reactivity, so the site is independent of that around the reactive thiols. The affinity of the crosslinked S1-antipeptide product, for actin in the presence of Mg2+ATP, was markedly reduced, but not completely inhibited. On the other hand, it was found that the interaction of the S1-antipeptide product for actin in the absence of ATP was largely unaffected. In the presence of ATP, S1 has a low affinity for actin (Kd=10-4 M) and is considered to be in a pre-power stroke conformation of muscle contraction (in the absence of nucleotide, high affinity with Kd=10-8 M, in the post-power stroke conformation). Therefore, Chaussepied & Morales (1988) suggest residues 633-642 of the S1 heavy chain, represent part of the site of S1 which interacts with actin when the myosin head is in a weak actin-binding pre-power stroke conformation. DasGupta, et al. (1990) used some similarities between the behaviour of this site and that in the N-terminus of actin to imply that the negatively charged N-terminus of actin interacts with positively charged myosin sequence 633-642. As described in the previous section, the validity of any electrostatic interaction with the N-terminus of actin is suspect.

The EDC crosslinking of the region 635KKGGKK643K to F-actin was investigated in the presence and absence of ATP, by sequencing from a convenient upstream elastase digestion site (Yamamoto, 1989). A difference was detected, such that in the weakly binding state (with ATP) the crosslinking is much reduced and only a single Lys residue is involved in the crosslinking reaction, whereas in the rigor complex the entire region is involved in a considerably stronger interaction.

Finally, Sutoh (1982) found that the A1, but not the A2 light chain crosslinked to the C-terminal residues 360, 362 & 363 of actin. NMR studies of peptides derived from the N-terminal region of the A1 light chain, have located the actin-binding site in A1 light chain as residing in the N-terminal 1-37 peptide (Trayer et al., 1987), probably between residues 3-8 ( KKDVKK; Z. Murad, MSc thesis, Birmingham, 1990). The interaction is electrostatic in nature. The spin labelling residue 374Cys of actin was observed to enhance the relaxation of residues in the N-terminus of the A1 light chain, in particular the N-terminal tri-methylalanine, placing this group within 15Å of the 374Cys.

In summary, there appear to be multiple sites of interaction on S1 for actin. This multiple-docking system would be precisely the type of system in which the interaction of nucleotide between the domains of the S1 head, could be visualised to regulate the affinity the of the S1 head for actin, by regulating the affinity of each of the of these different sites. In the strong binding rigor conformation of S1, all of the actin-binding sites on S1 might be complexed to actin, whilst in the pre-power stroke conformation only a few these sites on S1 (eg. 50-20 kDa junction and perhaps surrounding region) would be involved in the binding interaction. In such a model, the power stroke of muscle contraction would involve a rearrangement of the myosin head, which would be accompanied by an increase in the number of sites on S1 for actin.

Of the other components the contractile system the structure of troponin C (Herzberg & James, 1985) and tropomyosin (Phillips et al., 1986) have been determined. Tropomyosin (68 kDa) is a dimer in which two helical peptides possessing a regular 7 amino acid repeat, associate with one another to form a coiled coil structure, similar to that found in the myosin tails. The troponin complex consists of a calcium binding subunit (TnC, 17.8 kDa), a subunit binding tightly to the 400Å long coiled coil tropomyosin molecule (TnT, 30.5 kDa), and a subunit inhibiting the Mg2+-ATPase of actomyosin (TnI, 20.9 kDa). TnC has four Ca2+ binding sites, two with high affinity sites (Ka about 107 M-1) that bind Mg2+, and two with lower affinity sites (Ka 105 M-1) specific for calcium and lying in the N-terminal domain of the molecule. These lower affinity Ca2+ specific sites are responsible for the regulatory function of the protein. They are only fully occupied when the free calcium concentration rises above 10 uM, a situation in vivo that only arises upon activation of the muscle cell by the nervous system. Each calcium binding site adopts the EF hand helix-loop-helix structure which is common to Ca2+ binding proteins. TnC consists of two domains joined by a nine-turn helix with no direct intramolecular interactions between the domains (N-terminal domain residues 1-85; C-terminal domain residues 97-162). The central helix has a region in which eleven amino acids are completely solvent exposed (K87EDAKGKSEEE), and appears to be highly flexible. A similar dumbbell structure connected by a linker helix has been observed in calmodulin (Kretsinger et al., 1986). Some resonances of both cardiac and skeletal TnC have been assigned together with calcium-induced chemical shift changes (MacLachlan et al., 1990; Hincke et al., 1981). TnT is highly asymmetric with a length of 18.5 nm (Flicker et al., 1982). TnI is a small globular protein.

The interaction of metal ions with peptides analogues of the Ca2+ binding site of TnC have been extensively investigated (loop and helix-loop-helix). Gariepy et al. (1985) used a thirteen residue peptide corresponding to the contiguous sequence forming the whole of the EF calcium binding loop, together with gadolinium, to examine the ion binding properties of the region (the ion Gd3+ has the effect of broadening resonances which lie within about 5-10Å of the ion, but has a different coordination as compared to Ca2+). Careful examination of the differential changes in relaxation times of individual groups of the peptide, as Gd3+ was gradually titrated into the solution, especially after correction for chemical exchange (Marsden et al. 1989) allowed the precise mapping of the binding site. Later experiments by Shaw et al. (1991) showed that if longer peptides were used, including regions not directly involved in the ion-protein interaction, then the structure adopted by the peptide was increasingly like the intact protein and the binding constants increased sufficiently to allow the direct binding of the peptides to calcium to be investigated. The experiments were aided by exchange which was slow on the chemical shift timescale in these tighter binding peptides, and hence separate resonances and NOESY patterns were detected for free and bound peptides. These experiments are among the few where specific mutations in the peptides have been shown to affect binding and can be shown to be comparable to mutations in other intact expressed calcium binding proteins. The series of experiments has recently been further extended by Shaw et al. (1990) who used a 34 residue peptide corresponding to half of the C-terminal domain of TnC (the whole helix-loop-helix EF hand). This was shown to adopt the structure seen in the intact protein in the presence of saturating calcium, and in addition dimerised via the formation of a ß-sheet between the two peptide units to form a structure very similar to the entire calcium binding domain in the intact protein. It should be noted that the effects of gadolinium induced 1H relaxation are much longer range than NOE measurement (5Å) and hence the methods are complimentary.

The thin filament has two fundamental components. The F-actin polymer forms a rigid filament against which myosin heads can pull to generate the force which results in the thin filaments moving towards the centre of the sarcomere. In additional the F-actin provides a sensitive cooperative system with which the regulatory components tropomyosin and the troponin complex can interact to control contraction in response to the calcium signal.

The thin filament, or 'regulated actin', is composed of F-actin together with tropomyosin and the troponin complex. The elongated tropomyosin molecule (see section 1.12) binds around the two grooves of F-actin, with one tropomyosin dimer associated with 7 actin monomers on the thin filament. The troponin complex interacts with the thin filament with a stoichiometry of one troponin complex for every seven actin subunits, and thus to one tropomyosin dimer. Electron micrographs of whole troponin complex, show it to be composed of a globular domain and a tail (Flicker et al., 1982). The regular disposition of the troponin complex along the thin filament, results from the specificity of the N-terminus of TnT for the end-to-end junction of tropomyosin molecules. TnT lies along the surface of tropomyosin bound to the actin filament. TnI acts to inhibit the actin-activated Mg2+ATPase of myosin by direct contact with the actin. The inhibition by TnI is regulated through direct contacts to TnC, the calcium binding component. TnC interacts with TnI and TnT (White et al., 1987).

The interactions of TnI with actin and TnC have been extensively investigated by using peptides corresponding to the TnI sequence. The C-terminal region of peptide corresponding to the sequence 96-116 of TnI was shown to interact with actin, while the N-terminal region of this peptide 96-116 together with the peptide 1-21 were identified as interaction sites for TnC (Grand et al. 1982; Van Eyk and Hodges, 1988). The inhibitory consequences of interaction of TnI with actin can apparently be mimicked by the peptide 96-116 (Talbot & Hodges, 1981). The effect is enhanced by binding of the peptide to tropomyosin, in a manner reminiscent of the intact protein, and the peptide binds to TnC in a calcium dependent manner resulting in the release of the inhibition (Cachia, 1983, 1986; van Eyk & Hodges, 1987). The binding site on actin was shown to consist of the regions 1-7 and 19-44 of actin (probably 24D and 25D; Levine et al., 1988). The region 104-115 was identified as the minimum sequence from TnI capable of inhibiting the ATPase activity (Talbot & Hodges, 1981). The peptide 104-115 competes with TnI and inhibits the force development in skinned muscle fibres (Ruegg, 1989). The 3D-structure of TnI 104-115 bound to TnC in the presence of Ca2+ has also been determined (Campbell and Sykes, 1990).

Cryoelectron microscopy, image averaging, and difference calculations were used to analyse F-actin and combinations of actin decorated with S1(A1 or A2) heads, tropomyosin and in some cases having labelled the actin 374Cys with undecagold providing images of 25-30Å resolution (Milligan et al., 1990). The structure of F-actin agrees with the Holmes et al. (1990) model and confirms that the front back, top and bottom surfaces of the outer domain of actin are accessible, whist only the front face of the inner domain is accessible. The maximum filament diameter is 95-100Å, with the outer and inner domains centred at 26 and 17 Å respectively from the filament axis. This cryoelectron microscopy data has been used to fit together the higher resolution structures of the individual components of the contractile system. The positions of various components of the acto-myosin-tropomyosin complex are illustrated in figure 1.14.1.

The myosin head interacts with the actin filament in a tangential manner (resulting in the characteristic arrow head appearance which has been used for definition of filament polarity) obscuring the front concave face of the actin monomer extending towards the point of attachment of tropomyosin (see below), so that S1 may also bind to the inner domain of actin. The major contact is with the small domain of actin (in agreement with the biochemical data, see section 1.10), and the line of contact would involve the residues 1-4, 24-28, 93, 95 in the actin sequence. In addition to this main site of interaction there is a thin bridge of density extending between the main body of the S1 and the top of the outer domain of the adjacent long pitch monomer. A single S1 head hence appears to make contact with two actin monomers. This bridge lies at a radius of 40Å. The site of interaction agrees with data from previous three dimensional image reconstructions of electron microscopic images of the S1-actin rigor complex (Amos et al. 1982; Moore et al. 1970).

The N-terminal extension of the A1 light chain is known to interact with C-terminal portion of actin (residues 360-363; Sutoh, 1982; Trayer et al., 1978). This region was visualised in the actin structure by labelling 374Cys with undecagold, which allowed this residue to be located at a distance of 27Å from the axis filament. The N-terminal extension was located by calculating the difference between images of S1A1 and S1A2 decorated actin. The binding site was located under the bottom of the outer domain of the actin monomer close to both the genetic helix and long pitch helix contacts between the monomers. In this position it is easy to imagine how the binding of A1 might stabilise the monomer contacts of actin and hence enable S1A1 (but not S1A2) to promote actin polymerisation (Chaussepied & Kasprzak, 1989). This may also explain why S1A1 has a higher affinity for actin than S1A2 at low ionic strength (Chalovich et al., 1984).

When both Ca2+ and S1 are present the tropomyosin molecule follows the long pitch of the helix of the F-actin, contacting the inner domain of the monomer near the genetic helix contact, lying at a radius of 38Å. This position would correspond to a position in the Holmes et al. (1990) model to tropomyosin running in the groove between subdomains 3 and 4 with contacts around 215K and 307P and the helix 222D-233S. Tetrahymena actin differs from the rabbit sequence in the region 222-233 and does not bind tropomyosin (Hirono et al., 1990)

The image reconstruction data of Milligan et al., (1990) implied that a single S1 head might bind to two actin monomers. There has been previous biochemical evidence that this was the case. There is no doubt that the overall stochiometry of the F-acto-S1 complex is 1:1 (Chen et al., 1985), however this leaves the possibility of molecular stochiometry of two actin monomers to a single S1 head, if the S1 lies in a staggered fashion along the actin filament. Fluorescence polarisation experiments support the 1:1 stochiometry (Chaussepied & Kasprzak, 1989). Two actin binding sites on the S1 head were identified by Chen et al. (1985), by EDC crosslinking to the 20 & 50 kDa domains, although the isolated reaction products only contained actin crosslinked to a single site, so that binding at the two sites appeared mutually exclusive. Workers have failed to convincingly isolate the trimeric species crosslinked to both sites on a single S1 head, although claims have been made (Mornet et al., 1981; Sutoh, 1982). However, recently several reports of such trimeric complexes have appeared. Actin polymerisation and the interaction with myosin can be detected by changes in fluorescent labels attached to actin at the C-terminus 374Cys or 373Lys (eg. F-acto-S1 exhibits a six-fold enhanced pyrenyl fluorescence over G-actin). Light scattering is also used to detect acto-S1 interactions. These techniques have been used to detect two complexes of G-actin with S1, where the actin:S1 ratios are 1:1 and 2:1, formed at low ionic strength and in the absence of ATP. These complexes appears to be intermediates in the S1-mediated promotion of the polymerisation of actin, since in neither complex is the actin-bound ATP hydrolysed (Rane et al., 1991). It appears that the binary and ternary complexes described above represent a nucleation state for actin polymerisation, after which actin forms a filament and during which process the bound ATP is hydrolysed. The rigor state of F-acto-S1 may bear some structural relationship to this ternary complex, since both are formed in the absence of added ATP. A crosslinked trimeric complex has also been isolated (Arata, 1991) formed between the S1 head and actin which had been pretreated with the crosslinking reagent pPDM. In the 1:1 actin monomer:S1 head complex, the 27-50 kDa junction was protected against proteolytic digestion, and in the 2:1 complex an additional second protected site was seen at the 50-20 kDa junction.

In the model of the F-actin structure, there appears to be a very limited amount of space on the exposed outer domain of actin for all the proteins known to bind to the actin filament to possess individual binding sites. Therefore, this gives credence to the ideas of Moir & Levine (1986) and Keane et al. (1990) that the exposed domain of actin acts as a docking area for various proteins which must compete with each other for the available surface binding sites. Actin is highly conserved, so the variability in the interaction of proteins with actin must therefore reside in the protein partners.

Section 1.15: The kinetics and thermodynamics of the actin activated Mg2+ATPase cycle.

Analysis of the transient kinetics of the hydrolysis of Mg2+ATP by S1 and acto-S1 has allowed intermediates during the crossbridge cycle to be studied. In these studies S1 and Mg2+ATP are mixed rapidly and the resulting reaction mixture analysed on the millisecond time scale. The development of caged compounds has allowed similar millisecond processes to be observed in muscle fibres. The slow diffusion through a muscle fibre is overcome by rapidly photo-releasing molecules such as ATP, Ca2+ or Pi from inert and light sensitive precursors (Homsher & Millar, 1990). Both steady state and transient kinetic studies of acto-S1 ATPase by Lymn & Taylor, 1971; Stein et al., 1979; Geeves et al., 1984, and others have lead to the formulation of a several kinetic models, based around a central theme, which is described below and illustrated in figure 1.15.1.

In solution, the binding of actin to myosin can be measured at different stages of the ATP hydrolysis cycle. Such measurements in the absence of ATP (the rigor state) or in the presence of ADP, show that S1 (myosin) and actin bind very tightly (Kd=0.01 uM). These are the strong states indicated in the figure 1.15.1. In the presence of ATP or the complex ADP.Pi, S1 and actin bind only weakly (Kd= approx. 100 uM; weak states). Thus as acto-S1 progresses through the ATP hydrolysis cycle it switches between weakly and strongly bound conformational states, with a change in binding affinity of 3-4 orders of magnitude. The hydrolysis of bound Mg2+ATP is both very rapid and readily reversible. The products of the reaction are Mg2+ADP and Pi, both bound to the active site. The subsequent step, the release of Mg2+ADP and Pi from the active site is the rate limiting step in S1 Mg2+ATPase activity (Lymn & Taylor, 1970). F-Actin activates the steady state S1 Mg2+ATPase activity by 200 fold, to 20 s-1 (Eisenberg & Moos, 1968). The activation of the S1 Mg2+ATPase activity by actin is of course achieved by increasing the rate at which the slowest step in the process proceeds (the release of Mg2+ADP and Pi products from the S1 active site). Using viscosity and turbidity measurements, it was shown that most of the S1 remains dissociated from actin when the Mg2+ATPase activity is close to Vmax (Stein et al., 1979). Hence, in solution actin appears to activate the S1 Mg2+ATPase by only a transient interaction with S1, which is provides a route to by-pass the slow product release step. In the weakly bound states, there is rapid equilibrium with the corresponding detached states, and so the acto-myosin system cannot bear force. In these weakly bound and detached states, hydrolysis of Mg2+ATP bound to the S1 active site occurs rapidly (k=100-144 s-1), and reversibly (Keq=10). Both S1.Mg2+ATP and S1.Mg2+ADP.Pi are in rapid equilibrium with their actin-bound counterparts, and may attach and detach many times in the timescale of a single Mg2+ATP turnover (Stein et al., 1979). In the strongly bound states, the equilibrium is such that detachment is unfavouable. Since the detached states during release of the hydrolysis products are the rate limiting step, then the attached state forms the route to overcome the inhibition of S1-ATPase activity. Accordingly, at very high actin concentrations, when S1 remains essentially always complexed to actin, no inhibition of the Mg2+ATPase activity is observed (Stein et al., 1979). It will be noted that in vitro, the Mg2+ATPase cycle has no point at which it is essential for the myosin head to dissociate from actin (unlike the Lymn & Taylor, 1971 model where dissociation of the myosin head from actin upon binding of Mg2+ATP was a compulsory event in the Mg2+ATPase cycle). Obviously, crossbridge detachment is a prerequisite for muscle contraction in order to allow the thin and thick filaments to slide past each other.

The power stroke in the cycle is the point at which chemical energy can be converted to mechanical work, and is thought to correspond to the phosphate release step. In vitro, this step is accompanied by a change in the conformation of the myosin head attached to actin (Trayer & Trayer , 1988), and an increase in affinity of S1 for F-actin corresponding to a transition from weak to strong binding states. This is the step which is assumed to result in a change in angle of the myosin head attached to actin, resulting in the thin filament moving with respect to the thick filament, although no such reorientation has ever been observed to take place. The rate limiting step for the acto∙S1 Mg2+ATPase activity does not correspond to the phosphate release step at the power stroke, since the rate of Pi release occurs at a five-fold greater rate (k=100 s-1) than the maximum steady state ATP turnover (20 s-1; Stein et al., 1979). The rate-limiting step is represented by the transition from the 'refractory state' to the 'non-refractory state'. This is assumed to consist of a structural isomerisation of the protein while in the state A.M.ADP.Pi or M.ADP.P. This isomerisation precedes the power stroke.

Many other kinetic schemes exist in which the mechanism is broken down into more intermediates. Of these the most interesting modification is that proposed by (Geeves et al., 1984; Geeves & Jeffries, 1988; Geeves, 1991) and Woodward et al. (1991). Light scattering and fluorescence evidence suggests that S1 binds to actin in two steps, and hence two actin-bound states may be introduced into the kinetic scheme for each intermediate of the crossbridge cycle. It is postulated that in one form (state I) the equilibrium constant for attachment of S1 to actin is insensitive to the form of nucleotide bound to the active site of S1. The alternate attached acto-S1 complex (state II) arises as a result of isomerization of the first attached state. This isomerisation was proposed to be very sensitive to the type of nucleotide bound. The scheme calls for weak-binding conformations of acto-S1 to be largely in state I, and the strong-binding conformations to populate mainly the state II. There is kinetic difference AM.ADP in state I & II, such that the former releases ADP 103-fold more slowly (AM.ADP in state I is equivalent to M.ADP). Thus, actin effectively accelerates ADP release from the attached crossbridge, getting it ready for the next cycle of muscle contraction more rapidly.

It is instructive to look at the energetics of the crossbridge cycle, in order to see how closely the intermediates described above in a kinetic scheme are applicable to an energetic scheme. A large input of free energy is required to cause a conformational change in the myosin head, sufficient enough to reduce the affinity of the protein for actin by 3 to 4 orders of magnitude. This free energy is supplied by the formation of strong bonds between the Mg2+ATP and the myosin head (Myosin binds to Mg2+ATP with a Kd=10-11 M; Geeves, 1991). In this state, the release of bound Mg2+ATP from the active site of the myosin head would be hence be energetically unfavourable (positive free energy change). However, hydrolysis of Mg2+ATP to Mg2+ADP & Pi is accompanied by very little change in free energy (Eisenberg & Hill, 1985) and can therefore proceed. The products of this reaction have a much lower affinity for S1 (Mg2+ADP Kd = 10-6 M; Geeves, 1991) and they can hence be released from the active site. This is probably because the free energy of ADP and Pi in solution is much lower than that of Mg2+ATP. Therefore, the free energy of binding Mg2+ATP to the myosin head is effectively used for the transition of the crossbridge from a weak to strong actin-binding conformation. The role of Mg2+ATP hydrolysis in muscle contraction is to change the chemical form of Mg2+ATP so it can be released from the enzyme at no extra cost in free energy.

It is apparent that there are three functionally important stages in the crossbridge cycle, pre- and post- power stroke and the active S1 head. The crossbridges in these states must have different conformations and mobility. The states have been studied in both intact fibres and in solution. In rigor, the myosin heads are in a strong actin-binding conformation (end of the power stroke), which can be reproduced in a stable form for study, in the absence of Mg2+ATP or in the presence of Mg2+ADP (post-power stroke). In the presence of AMPPNP, muscle fibres exhibit structural and mechanical properties intermediate between those of rigor and relaxation, and myosin is believed to be trapped in a strongly bound intermediate state (Berger & Thomas, 1991). In solution the weak binding state may be formed for example by the intramolecular crosslinking of the SH1 & SH2 reactive thiols of S1 with pPDM in the presence of ADP (Audermard et al., 1988, review). In muscle fibres the weak binding state (pre-power stroke conformation) exists in the presence of Mg2+ATP (Mg2+ADP & Pi) but in the absence of Ca2+ (or low Ca2+). Under these conditions the fibre is 'relaxed' and contraction is inhibited. ATPcS is an analogue of ATP that is hydrolysed 500 times more slowly than ATP. In the presence of ATPcS myosin hence accumulates in a weakly bound pre-hydrolysis intermediate state of the ATPase cycle. The active conformation may be maintained in a muscle fibre by fixing it at either end, followed by activation with Mg2+ATP and Ca2+ to produce isometric tension.

Stiffness measurements on muscle fibres provide evidence for the weak binding (Schoenberg, 1988) and the strong binding (Brenner et al., 1986) crossbridge states. Muscle fibres in rigor are very stiff, presumably reflecting the number of crossbridges attached to the thin filament, although stiffness is not necessarily a linear function of actin-attached myosin heads (crossbridges). In an isometrically contracting muscle fibre stiffness is high (70-80% of the rigor value). Under conditions in which the muscle fibres are trapped in the weak binding state (eg. in the presence of ATPcS) muscle fibres develop stiffness without developing tension, indicating that these myosin heads early in the ATPase cycle can bind to actin, but cannot generate force (Danzig et al., 1988). Rapid stiffness measurement have indicated that the number of attached crossbridges in relaxed fibres at low ionic strength could be > 50% of the number bound in rigor (Brenner et al., 1982).

X-ray diffraction studies of fibres activated from relaxation show that there is a considerable movement of mass from the vicinity of the thick filament, towards the thin filament (Hazelgrove & Huxley, 1973), attributable to as much as 40% of the mass of the myosin heads. Such studies have concluded that an isometrically contracting fibre consists of weakly attached crossbridges, and strongly attached crossbridges. Time resolution X-ray studies indicate that the myosin heads bind to the thin filament before tension development occurs (Huxley & Kress, 1985). This has been interpreted as non-force generating weakly attached crossbridges binding to the thin filament before the power stroke of muscle contraction occurs.

Orientational EPR experiments of spin-labelled crossbridges (for example labelled on the SH1 thiol with an ESR probes) show that the rigor crossbridges bound to the thin filament are highly oriented. When the fibres are stretched till the thick and thin filaments no longer overlap, the uniform arrangement of crossbridges is seen to collapse, confirming that the regular arrangement of crossbridges originates from the myosin crossbridges bound to the thin filaments (Thomas & Cooke, 1980; Barnett et al., 1986). Similarly, X-ray diffraction studies show that in rigor muscle the layer lines corresponding to the thick filament are absent, while those corresponding to the thin filament are strongly enhanced, suggesting that myosin heads are fully attached to the thin filament (Holmes et al., 1980). In similar X-ray studies, in relaxed fibres the myosin crossbridges are regularly arranged around the thick filament, in a similar fashion to that observed for isolated thick filaments (Squire, 1972). Using EPR orientation measurement in isometrically contracting muscle fibres it has demonstrated that most of the myosin heads (>80%) are in a highly disordered state, comparable to that observed for myosin heads in relaxation. The other 20% of the myosin heads have an orientation similar to that observed for crossbridges in rigor (Cooke et al., 1982; Fajer et al., 1990b).

Time-resolved phosphorescence of labelled S1 has shown that isometrically contracting muscle fibres undergo rotational motions that are distinct from those in rigor (no nucleotide) or relaxation (Stein et al., 1990). When spin-labelled S1 was trapped in weakly and strongly attached states, then saturation transfer-EPR experiments were used to show that in the weakly bound states, S1 undergoes microsecond rotational motions, while in the strongly bound states no such motions were seen (Berger & Thomas, 1991). Similar microsecond rotational crossbridge motions have been identified in isometrically contacting muscle fibres (Fajer et al., 1990b), in the attached population of crossbridges. The pre-power stroke weakly attached state is currently considered to be a loosely attached state, perhaps involving a flexible and internally mobile form of the myosin head, where the head can adopt several angles of attachment to actin. This may be true for contracting muscle where the geometric and mechanical constraints of filament lattice will operate. It is less likely to be true in solutions, as the weakly bound S1.ATP might be expected to bind in an orientation representing the lowest free energy state. The fast microsecond motions of the weakly attached state are likely to be involved in the molecular mechanism of muscle contraction, but the precise role is unclear.

A tilting mechanism for the whole myosin head (with some exceptions) is the generally accepted hypothesis for the molecular mechanism of muscle contraction. Hence, cross-bridge orientations are fundamental to models of muscle contraction. The rotating crossbridge model suggests that the crossbridges rotate while attached to actin to produce a shortening against load (Huxley, 1969; Huxley & Kress, 1985). Many experiments have however failed to demonstrate tilting. A change in the structure of the S1 head has also been proposed to accompany the transition between pre- and post-power stroke states.

Some evidence has accumulated for conformational differences between S1 heads when strongly and weakly bound to actin. For example, NMR studies of S1 binding to actin, show that binding results in a restriction in the motion of the mobile portion the S1 heavy chain (Prince et al., 1981). FRET measurements show that the distance between 374Cys on actin and 177Cys on the A1 light chain, increases from 3 nm in the pre-power stroke conformation (weak binding state modelled by pPDM crosslinked S1 with trapped ADP or by S1.ATP in the presence of regulated actin/Tm/Tn and the absence of Ca2+) to 6 nm apart in rigor (modelled in the absence of nucleotide; Trayer & Trayer, 1988). However, these must be subtle differences since, when Curmi et al. (1988) used neutron scattering to compare the shape of S1 in free solution (in the presence of MgPPi at low ionic strength, which is thought to mimic the initial state of attachment to actin), with S1 bound in the rigor state, to deuterated actin (which can be made 'invisible' by arranging for it to scatter at the same level as the solvent), then no structural differences were found between the two states (resolution 25Å).

Early electron microscopy studies of muscle fibres in rigor showed that the crossbridges were bound to the thin filament in a highly oriented fashion at about 45° to the filament axis, whereas in relaxed muscle that crossbridges are attached at 90° (Reedy et al., 1965). It was these images that gave rise to the swinging crossbridge model of muscle contraction. Later experiments have not been so clear in the conclusions drawn as to the orientations of the crossbridges, and the point remains controversial. Myosin crossbridges in the muscle fibre labelled at the SH1 with both fluorescent and ESR probes, show that when myosin is bound to actin the SH1 thiol maintains more than one orientation relative to the actin filament. A fluorescent probe on the SH2 of myosin in glycerinated fibres distinguishes three different ordered states of the fibre for crossbridges in rigor and in the presence of Mg2+ADP, for relaxed crossbridges and for crossbridges during contraction (Ajtai & Burghardt, 1989). However, not all probes have detected a significant displacement of the crossbridge. This may mean that some probes are sensing only a local change in mobility rather than a global orientation change in the crossbridge. If the orientation of a probe is such that the transition dipole is aligned with the axis of rotation of the protein then no angular displacement could be measured, which may account for the failure to observed movement with some probes. Ajtai & Burghardt (1987) have developed a method using fluorescence polarisation spectroscopy in which the absorption dipole of the probe can be rotated within the molecular frame and so the probe is no longer limited by its physical orientation. This method should provide more information on the movement of crossbridges in the future. An additional problem arises in the interpretation of spectroscopic crossbridge studies based upon the assumption that SH1 labelling does not alter the normal crossbridge cycle. Crowder and Cooke (1984) claimed that SH1 in myofibrils could be labelled with impunity without changing the crossbridge cycle. However, functional differences have been found between SH1 labelled S1 in solution and labelled myosin in the fibres. In particular, changes in the shortening velocity and ATPase activity of modified fibres and myosin extracted from these is non-linear with increasing extent of labelling, whilst the effects on isolated S1 heads from these preparations are linear. This holds in the fibres even when the degree of labelling is small (Harrington et al., 1975; Root et al., 1991). It appears that the labelled molecules may fail to undergo a proper crossbridge cycle, and that cooperative effects exist with other myosin molecules in the filament whereby neighbouring unmodified molecules compensate for the loss of a normal cycle in the labelled molecules. Hence the conclusion of studies on the orientation and motion of molecules which are measured in this way are called into question.

When S1 is crosslinked to actin by EDC, via both the 50 and 20 kDa domains, it has an enormously activated Mg2+ATPase activity (Mornet et al., 1981), of similar magnitude to that seen during the actin-activated ATPase cycle (Biosca et al., 1985). The superactivated state does not appear to be associated with the intramolecular crosslinks trapping an active S1 conformation, but with the attachment of actin to S1 (Mornet et al., 1981) since intra-molecular crosslinking appears reduced in the presence of actin. Muscle contraction appears to be the result of a cyclic interaction of parts of the myosin molecule with sites on the actin filament which involves a significant reorientation of the myosin head bound to actin (corresponding to the force generating step). These motions are probably restricted in the crosslinked species by the immobilisation of the interface surfaces. Hence the ability of the crosslinked species to undergo the same orientation changes as the native acto-S1 system, during the ATPase cycle, has probably been lost. The crosslinked system must have important implications on the coupling of ATPase activity and orientational changes of S1 attached to actin, but the significance is not clear.

The Eisenberg & Hill (1985) model of the crossbridge cycle is illustrated in figure 1.17.1 and discussed below.

Before the power stroke in muscle contraction, the myosin heads are weakly attached to actin and are in rapid equilibrium with the unattached state. The myosin heads in this pre-power stroke state bind to the actin filament with a different orientation to the crossbridges in the post-power stroke (rigor) complex. The orientations are usually assumed to be 90° (weakly attached) and 45° (strongly attached) relative to the thin filament axis. The power stroke of muscle contraction thence consists of a change in the orientation of the crossbridge from the attached 90° conformation to the attached 45° conformation, in an "anti-rowing" motion.

In solution, the next steps in the process can occur without the myosin head dissociating from the actin filament. In the muscle fibre there must be a mechanism to promote dissociation of the myosin crossbridge from the actin filament at the end of the power stroke. In the absence of such a mechanism, the subsequent binding of Mg2+ATP to the rigor crossbridge at the 45° orientation would result in the crossbridge changing into a weak actin-binding conformation at 90°, thus reversing the work carried out during the power stroke. This mechanism is probably simply a consequence of the arrangement of the contractile apparatus in the fibre, where the myosin crossbridges is in a fixed lattice of actin and myosin filaments. A single myosin head is not isolated from the behaviour of the neighbouring crossbridges. The sliding of filament as a whole will pull apart the defunct weak binding conformation, which can then reorientate and attach to the next suitable actin monomer which is presented by sliding of the filament, and another cycle of muscle contraction can then take place. In solution studies there are no such strains on S1. It is interesting to note that the rate constants between the intermediates in the crossbridge cycle in the intact fibre will be sensitive to the overall strain on the crossbridge and therefore are not necessarily equivalent to those measured in solution (Stein, et al., 1979).

Direct evidence of the crossbridge cycle is not available. Some of the supporting evidence has been described in the previous section. One of the best pieces of evidence is that addition of the non-hydrolysable ATP analogue, AMPPNP, to a muscle fibre in rigor caused the fibre to lengthen slightly, presumably because AMPPNP causes the crossbridges to rotate from 45° to 90°, thereby increasing the length of the fibre (Marston et al., 1976). Recent mechanical measurements on single myofibrils (population of about 106 myosin molecules) have provided evidence of force fluctuations smaller than those predicted by the model of cyclic reattachments and rowing motions (Iwazumi, 1987). The type of in vitro motility assays of Kishino & Yanagida (1988) should enable mechanical measurements to soon provide evidence at the level of individual molecules and hence confirm or extend the model described above.

The observation of Milligan et al. (1990) and others that there may be two actin monomers attached to a single myosin head, enables the model to be placed into more precise molecular terms. It is possible that in the weak actin-binding conformation the myosin head is only attached to a single actin monomer, whilst in the strong post-power stroke conformation the myosin head is attached to two adjacent actin subunits. If this were the case, the force generating step in muscle contraction would thus be the transition of the myosin head from an interaction with a single actin monomer to an interaction with a pair of adjacent actin monomers.

Schutt & Lindberg (1992) have proposed a completely new model of muscle contraction. There is no particular evidence to prove the model, but it is an indication of the lack of informative experimental data on the true nature of the crossbridge cycle that there is no evidence to disprove the model. Briefly, based on the tendency for the profilin-actin crystal to pack in an untwisted fashion (not compatible with helicity), the authors assert that there are two states of the actin filament, the helix which is familiar to all workers and a ribbon-like conformation which lacks the twist. They propose that changes in the actin structure driven by the interaction with the S1 head are responsible for the generation of lateral displacement and tension. So, myosin binding in the ADP,Pi state would induce the ribbon conformation in the local actin unit. This is accompanied by a change in length of the local actin unit (helix to ribbon unwinding), and the tension develops as the units contract back to the helical state. The attachments of tropomyosin to the filament would be used to summate the displacements from along the filament, since the actin would no longer be a rigid unit against which the system could pull.

The Schutt & Lindberg model is somewhat extreme, but the crossbridge cycle model of Eisenberg & Greene may be too simple in the way in which it treats actin as simply a rigid rod against which the system can pull. There have been numerous studies suggesting that the binding of S1 to the F-actin filament induces a long-range conformational change along the filament (eg. Yanagida et al., 1984; Rouayrenc et al., 1985). A torsional motion of actin subunits in the filament has been reported (Thomas et al. 1979; Yoshimura et al. 1984) which is thought to correspond to the disorder seen sometimes in images of actin filaments in the electron microscope. There is some evidence that the disorder in the actin structure, which leads a variable twist can be modulated by actin binding proteins (Stokes & DeRosier, 1987). This may be related to the mode of regulation of muscle contraction, cf. the effects of antibody Fab directed against residues 1-7 of actin (section 1.10). Recently, polymeric crosslinked MBS-actin has been shown to be highly potentiated, so that it activates the Mg2+ATPase of S1 more strongly than native F-actin, with a partial loss of Ca2+ regulation in the presence of the Tm/Tn complex (Miki & Hozumi, 1991). The residue 336Lys at the cleft region between the two domains of actin, near the ATP site has been implicated in this potentiation. Hence, cooperative allosteric interactions between actin subunits might be a means of communication between proteins binding in one of the subunit in the filament to adjacent subunits.

The contraction event in muscle is initiated by a nerve impulse which causes a rapid increase in the intracellular Ca2+ ion concentration from around 10-7M (resting muscle) to 10-5M. In vertebrate skeletal and cardiac muscle the regulation of contraction and relaxation is at the level of the thin filament, via the protein complex troponin/tropomyosin. Following the influx of Ca2+ ions the low affinity calcium binding sites of TnC become occupied, which leads to a poorly understood allosteric structural change in the interactions of TnC with both TnI and TnT. The result is that TnI comes to interact more strongly with TnC and less strongly with actin (and hence the inhibition by TnI is overcome) and the altered TnC-TnT interaction results in the tropomyosin filament moving towards the centre of the thin filament. The effect of this calcium induced rearrangement of the thin filament proteins is to allows the crossbridge cycle to proceed (Leavis & Gergely, 1984). The role of the TnI, the inhibitory component of the troponin complex which binds to the N-terminal region of actin, may be similar to the allosteric effect described for an antibody against the N-terminal residues 1-7 of actin described by DasGupta & Reisler (1992, section 1.10).

The earliest proposed mechanism for regulation of muscle contraction was the steric blocking model. In this model, in the absence of Ca2+, the tropomyosin molecule blocks the site on actin where the crossbridges bind, so that the thin filament is switched off and the muscle is relaxed. When Ca2+is released it binds to troponin, which causes tropomyosin to move, thus exposing the binding site on actin and switching on the filament. Hence in the presence of Ca2+ the crossbridges can attach and contraction proceeds. This hypothesis is simple, but has been questioned in the light of considerable evidence which contradicts the proposed mechanism.

X-ray studies have confirmed that tropomyosin moves deeper into the groove on the actin filament in response to Ca2+ binding at TnC, and time resolved X-ray studies have shown that tropomyosin movement into the groove of F-actin helix precedes by several milliseconds an increase in the number of myosin heads attached to actin (Kress et al., 1986; in addition note that while there is no troponin present in this molluscan muscle the tropomyosin filament still moves during the contraction cycle, Vibert et al., 1972).

Evidence against the steric blocking model came from the calcium sensitivity of the binding of the S1 head to regulated actin. The binding of the strong actin attached states was blocked in the absence of Ca2+ as expected (Greene & Eisenberg, 1980), but surprisingly the binding of the weak states were not affected by changes in the Ca2+ concentration, under conditions at which ATPase was inhibited by upto 95%, as compared to the rate in the presence of calcium (Chalovich et al., 1981). In the absence of calcium, S1.ATP forms a relatively long lived weak-binding acto-S1 complex, with no force generation (see section 1.17, this state is detected by a small stiffness in the muscle under these conditions) and so the myosin crossbridge can bind to actin (albeit weakly) even with the thin filament switched off. Hence tropomyosin was not simply blocking attachment of S1 to actin. In addition there were several other pieces of evidence which did not fit the model. For example, the binding of myosin to F-actin significantly increases the affinity of tropomyosin for F-actin (Sobieszek & Small, 1981), and Tm/Tn binding to F-actin results in a threefold strengthening of the myosin interaction (Williams & Greene, 1983). Also, the binding of myosin heads to regulated actin in the presence of Ca2+ contributes cooperatively to the activation of the thin filament. That is, the binding of S1 to regulated F-actin in the presence of Ca2+ is transmitted to several neighbouring actin subunits via allosteric conformational changes. This alters the conformation of the adjacent actin subunits in a way which enhances actin-activation of the Mg2+ATPase activity of myosin crossbridges which bind to neighbouring subunits (Greene, 1982).

In the light of this data, the next possibility investigated was that the means of regulation might be kinetic, so that in the presence of Ca2+, the release of Pi from the acto-myosin complex M.ADP.Pi might be stimulated (Chalovich et al., 1981). Hence the rate limiting weak-to-strong (or force generating) step, which is coupled to Pi release, would fall under the control of Ca2+, via some allosteric mechanism transmitted through the actin filament. This idea was supported by mechanical experiments in skinned muscle fibres, where the force generating step can be observed by measuring the rate of tension redevelopment following a quick release. This rate decreased gradually as the Ca2+ concentration was reduced, supporting the idea that the Ca2+-regulatory system was controlling the kinetics of the force generating step (Brenner, 1988).

However, data to contradict this hypothesis appeared when the Pi release step was probed more directly, by changing the Pi concentration, using caged species to increase the time resolution of the experiment. It was shown by Millar & Homsher (1990) that the rate of the Pi release step in active muscle was not changed by reducing the Ca2+ concentration, so that the kinetics of the force generating step was not affected by Ca2+ (the active crossbridge is not Ca2+-sensitive).

At this point, the data of Milligan et al. (1990) should be considered. The myosin crossbridge appears to be able to interact with two adjacent actin monomers, and it is possible that the weak binding site resides on one monomer and the strong binding site on the other. This spatial displacement of the S1 binding sites means that tropomyosin movement may block the strong binding site, whilst the weak binding site remains unaffected. This mechanism amounts to a modified steric blocking model, in which the tropomyosin only blocks the strong binding state.

Regulation of contractile activity in smooth muscle involves phosphorylation and de-phosphorylation of myosin light chains (Harshorne, 1987) by myosin light chain kinase at 15Ser (Perrie et al., 1973). Phosphorylation of the RLC is a prerequisite for the triggering of contraction in vertebrate smooth muscles (not in skeletal or cardiac muscle; Kendrick-Jones & Scholey, 1981). MLC kinase is regulated by Ca2+ levels, with calmodulin as an intermediary. An increase in phosphorylation is associated with an increased actin activated ATPase activity. De-phosphorylation is believed not to be regulated. At low ionic strengths and in the presence of ATP, smooth muscle myosin exists in a folded state, where the myosin tail loops back upon itself to interact near the head-neck junction and the heads are directed back towards the tail (Sommerville et al. 1990). At increased ionic strengths the extended conformation is restored. The equilibrium between these states is shifted towards the extended conformation by phosphorylation. It is not clear whether this change is physiologically important in regulation. The extended-folded transition results in changes in segmental mobility within the S1 head that may be involved in the activation of actin binding. These changes are seen in myosin and HMM, but not in the isolated S1 head (Sommerville et al. 1990). Hence they must arise from high order structures, either in the S2 section, or involving the interaction of two heads.

Whilst smooth muscle contraction is primarily regulated by calmodulin stimulation of myosin light chain phosphorylation as described above, there is evidence that caldesmon is also involved in calcium regulation. Caldesmon is a long thin flexible monomeric molecule (470Å, 756 residues) which binds at its C-terminal, along the long-axis of actin in smooth and non-muscle cells. Binding inhibits actin activated ATP hydrolysis and acto-S1 binding. The inhibition is reversed by the binding of Ca2+-calmodulin to caldesmon. Whilst the precise role of caldesmon is unclear, it does appear to behave somewhat like a smooth muscle analogue of TnI (stochiometry of 1 TnI molecule per 7 actin monomers, or 1 caldesmon molecule per 28 actin monomers; binds to the N-terminus of actin; Crosbie et al. 1991; Graceffa & Jancso, 1991). The N-terminal portion of caldesmon is also able to bind to myosin, which might play a role in the development of the latch state of smooth muscle. In the latch state tension is maintained while Ca2+, myosin phosphorylation, and ATP hydrolysis are low.

The regulatory light chain is the main regulatory component in myosin-linked regulation in molluscs, during which the Ca2+ ion trigger binds directly to the RLC (Vibert et al., 1985). This mechanism is not seen in the regulation of vertebrate muscle and despite the homology of the sequence of the light chains to other calcium binding proteins, they do not bind calcium.

4-Aminobenztrifluoride,

15N-leucine, 98 atom % 15N,

tris(hydroxy-d-methyl) amino-d2-methane, 99%,

anhydrous magnesium sulphate, guanidine hydrochloride.

Sodium 125Iodide.

ANSA (1-amino-2-napthol-4-sulphonic acid), 2-mercaptoethanol, enzyme grade ammonium sulphate, HPLC acetonitrile, spectrosol trifluoroacetic acid, Aristar formic acid, ammonium persulphate (electran), sodium dodecyl sulphate.

Urea.

Aquacide (grade IIA), thrombin.

Sp-TrisAcryl.

Fluorenylmethyl succinimidyl carbonate, Fmoc-valine, Fmoc-leucine.

Iodo-gen.

Freeze-dried α-chymotrypsin, Dowex-50, molecular weight standards for SDS-PAGE, TLCK-treated trypsin, 15N-Valine, bicine, trizma base, tricine.

Deuterium oxide, methanol-d4, trifluoroethanol-d3.

NMR Tubes (507-PP-7), Coaxial tubes (516-CC-5).

All other reagents were of the highest available purity and used without further treatment.

Mass spectra were obtained with the aid of A.J.Pemberton on a Kratos MS80RF instrument operated in fast atom (Krypton 84) bombardment mode. CsI calibration was used (positive ion mode; m/z where z=+1). Samples of 0.5umole in a volume of 20-40ul were used, with the carrier matrix 50% aqueous glycerol (mass 92). The signals were averaged over the 10-20 scans with best signal to noise ratio.

Photographs of actin were generated from the protein databank entry for actin-DNase I ATP, kindly provided by W.Kabsch, in Quanta 3.0 and 3.2 (Polygen Inc) on an Iris 4D-120GTX workstation operating under Unix. Molecular dynamics simulations were performed using CHARMm on the Iris workstation via the Quanta interface. For structural determinations the simulated annealing protocol of Nilges (1990) was followed using the NMR modules of XPLOR. Crystal coordinates for other proteins were taken from the Brookhaven Protein Data Bank (Blevins, & Tulinsky, 1985).

Protein sequencing was carried out by Dr. J.Fox of Alta Bioscience on an AIB 473A sequencer. Peptides were transferred to a polybrene support for sequencing and fragments blotted from gels were sequenced directly on PVDF membrane.

Unless otherwise stated all protein manipulation were carried at at 4°C.

Fast twitch white muscle myosin was obtained from freshly excised and minced longissimus dorsi and the white muscle of the thigh of the New Zealand White rabbit. The method used was based upon that of Trayer & Perry (1966).

Minced muscle (500 ml approx. volume of solid) was extracted on ice for 20 min with 0.6 M NaCl, 50 mM 2-mercaptoethanol, pH 6.5 (1.5 L). This solution was then centrifuged (8 K rpm for 10 min), and the supernatant filtered, under low suction, through a glass wool pad previously washed with 0.6 M KCl. The filtrate was then slowly poured into cold distilled water (20 L) containing 25 mM 2-mercaptoethanol, and the resulting mixture stirred for 2 min. The myosin precipitate was centrifuged (8 K rpm for 15 min), and solid KCl (67 g) was added to the precipitate (1.5 L) to bring the salt concentration up to 0.6 M, and 2-mercaptoethanol to 25 mM. The resulting mixture was then stirred on ice until all the myosin had dissolved, and was then poured into 25 mM 2-mercaptoethanol (20 L). The precipitated myosin was collected again by centrifugation (8 K rpm), and redissolved by adding solid KCl as above. The myosin solution was centrifuged (80 Kg, 40 K rpm for 60 min), and the supernatant filtered under low suction through a glass wool pad pre-washed with 0.6 M KCl. The concentration of myosin was calculated by subtracting A320nm from A280nm to correct for light scattering, and ε280-320nm(myosin) = 0.57 (mg.ml-1)-1.cm-1. A total of 15.8 g of myosin was obtained at a concentration of 20.2 mg.ml-1. Myosin was stored in a sealed bottle for up to five days at 20 mg.ml-1, 4°C in 0.6 M KCl, 0.5 mM DTT, pH 6.8.

The myosin solution (20 mg.ml-1) was divided into 3 batches of 200 ml (some left over) that were each dialysed overnight against filament buffer [20 mM sodium phosphate, 0.12 M NaCl, 1 mM EDTA, 0.25 mM DTT, 0.03% NaN3, pH 7.2] (5 L). The three batches of myosin filaments were digested separately in a conical flask equilibrated 25°C on a water bath. The myosin filaments (200 ml, 20 mg.ml-1) were then stirred at room temperature on a magnetic stirrer and freshly prepared α-chymotrypsin solution (ε280nm{α-chymotrypsin} = 2.04 [mg.ml-1]-1.cm-1;9.4 mg.ml-1) was added to this myosin filament solution at 50 ug per ml of myosin solution. The digestion reaction mixture was stirred for 8 min, and then phenyl methyl sulphonyl fluoride (PMSF, 70 mg, 2 mM final concentration) dissolved in dry ethanol (1 ml) was added to inactivate α-chymotrypsin, and the solution stirred on ice for a further 2 min. The resulting mixture was centrifuged (40 K rpm for 1 hour, 40 K rpm 6x100) and the supernatant dialysed on a a rocking dialyser for two hours against 5 mM MOPS, 0.5 mM PMSF, 0.5 mM DTT, 0.03% NaN3 pH 7.0 (1 L), and for a further two hours against 5 mM MOPS, 0.2 mM PMSF, 0.5 mM DTT, 0.03% NaN3, pH 7.0 (1 L), and finally against 5 mM MOPS, 0.1 mM PMSF, 0.5 mM DTT, 0.03% NaN3, pH 7.0 (1 L) for 2 hours. Finally the preparation was centrifuged at 80 Kg for 2 hours. The crude S1 concentration was determined by the A280nm (ε280nm = 0.8 [mg.ml-1]-1cm-1; 5.7 mg.ml-1; total vol. 123 ml; 702 mg mixed crude isoenzymes). It is possible to shell freeze the preparation at this stage for storage.

This mixture was then applied to a cation exchange sulphoxypropyl (SP Tris-Acryl) column (4.5 cm x 50 cm, pre-equilibrated with 10 mM MOPS, 0.5 mM DTT, 0.03% NaN3, pH 7.0, flow rate 70 ml.hr-1). The column was then washed with 10 mM MOPS, 0.5 mM DTT, 0.1 mM PMSF, 0.03% NaN3, pH 7.0 till A280nm was zero, and the bound protein is eluted with a gradient of 0 to 0.2 M NaCl (each 750 ml, in 10 mM MOPS, 0.5 mM DTT, pH 7.0). Fractions were collected and monitored by A280nm and SDS Gel Electrophoresis (figure 2.2.5). The central portions of peaks containing pure S1(A1) and S1(A2) were collected separately, and concentrated in aquacide IIA to 5-10 mg.ml-1, and then dialysed overnight in 25 mM TEA-HCl, 0.25 mM DTT, pH 7.0, and finally centrifuged (40 K rpm for 1 hour). After determination of concentration (yields from one column: 23.5 ml of 11 mg.ml-1, 260 mg S1(A1) and 18.5 ml 11.6 mg.ml-1, 215 mg S1(A2) ), solid DTT was added to a final concentration of 2 mM, and 4 mg of sucrose is added per mg protein. The samples were divided into aliquots containing 20 mg of protein, and then shell frozen for storage at -20°C.

Further purification of the S1(A1) or S1(A2) isoenzymes was sometimes performed using an affinity column procedure, that selects only the protein in the preparation able to bind actively to an ATP analogue fixed to the column matrix. The preparation was applied to a HAPP column prepared by H.Trayer (4 x 20 cm; pump at 35 ml.hr-1; 5 mM TEA / 5 mM DTT / pH 7.5; elute with 0-0.2 M NaCl gradient) and a yield of approx. 60% was recovered. The basic preparation has some minor bands, that do not correspond to the expected protein contents, that can be seen in a Coomassie stained SDS-PAGE gel. These were largely removed in affinity purified preparations, but in all experiments tried no difference was observed between these two classes of preparation.

Rabbit skeletal muscle F-actin was prepared from muscle acetone powder (Barany et al., 1957) by the method of Spudich & Watt (1971).

350 g of skeletal muscle residue from which the myosin had been extracted (usually stored frozen at -20°C until required) was extracted for 15 minutes on ice in 2 L (3 vols. excess) 50 mM sodium bicarbonate. The residue was collected after centrifugation. (6 x 500, 8 K rpm 10-20 minutes.) and resuspended (not re-extracted) in 10 volumes excess (approx. 5 L) of 1 mM EDTA. The residue was again collected by centrifugation. This procedure was repeated twice using distiled water as the respuspension medium. The resuspension and centrifugation process was then repeated with cold acetone was added , a total of three times, with 2 volumes (approx. 1 L), 2 vols. and finally 1 volume excess of acetone. The resulting fibrous mass was laid out to dry in the air overnight between filter paper pads. The yield of muscle acetone powder is approx 35 g.

To prepare actin, 5 g muscle acetone powder was extracted for 30 minutes, with 100 ml of buffer A (5 mM TEA-HCl, 0.2 mM CaCl2, 0.2 mM ATP, 0.25 mM DTT pH 8.0), on ice using an overhead stirrer. The supernatant was collected after centrifugation (30 minutes, 100 Kg, 30 K rpm, 8x50). Solid KCl was added to the supernatant to 50 mM, and 1 M MgCl2 to 2 mM, final concentrations. The actin polymerised within 2 hours, after which the concentration of KCl was increased to 0.6 M. The dissociation of tropomyosin was complete within 2 hours. The solution was centrifuged (3 hours, 100 Kg, 40 K rpm, 8x50) and the polymerised actin collected as a clear gel-like pellet. The pellet was removed, broken up, and resuspended in at least 20 ml of buffer A. (It is important to dilute the preparation sufficiently here so that all the actin depolymerises, since only G-actin will be harvested in the next step.) The actin was depolymerised by dialysis on a rocking dialyser for approx. 36-48 hrs, and the resulting G-actin was clarified by centrifugation (100 Kg, 2 hours, 50 K rpm, 10x10). Typically 70-80 mg of freeze-dried powder was collected which was approx. 45% protein.

G-actin dialysed in buffer A (5 mM TEA-HCl, 0.2 mM CaCl2, 0.2 mM ATP, 0.25 mM DTT pH 8.0) was freeze-dried and stored at -20°C. F-actin was reconstituted by dissolving 10-20 mg of the freeze-dried powder (≈50% protein) in buffer A (1-2 ml) on ice (stir gently for 30 mins) followed by dialysis against buffer A (2 L, overnight). The G-actin solution was clarified by centrifugation (35 K rpm for 30 minutes), and its concentration determined by uv absorption (ε290nm{G-actin} = 0.63 [mg.ml-1]-1cm-1). Polymerisation to F-actin was achieved by dialysing the G-actin against buffer A containing 50 mM KCl, 2 mM MgCl2 (1 L) (ε290nm{F-actin} = 0.66 [mg.ml-1]-1cm-1).

In most cases proteins were prepared in the standard fashion and then dialysed into three changes of 1 Litre of 5 or 10 mM phosphate pH 7.5 (Na2HPO4 0.596 g.L-1; KH2PO4 0.109 g.L-1), 0.25 mM DTT, 2 mM MgCl2, (MgCl2 was usually omitted for S1 preparations), 0.02% NaN3, followed by dialysis into the equivalent solution in either 90%H2O / 10%D2O (1 change only required), or 99.9% D2O (3 changes of 25 ml, the final of which was a fresh solution and the others rotated from the previous preparation).

The pH of protein preparations was never adjusted directly, and instead the pH of a dialysate was manipulated. 0.1 mM DSS was usually added as a standard. When phosphate buffers were not appropriate, then borate (around pH 8) or bicarbonate (around pH 7) were used.

Separated myosin alkali light chains 1 and 2 (A1 and A2) and phosphorylatable light chain (PLC) were prepared by the method of Henry et al., (1985) from skeletal muscle myosin, using a procedure involving urea-mediated dissociation of the light chains and purification by ethanol fractionation and on DEAE-Sephadex A25 ion exchange chromatography.

This protein fragment was prepared from S1A1 or S1A2 by a method loosely developed from the preparation of a similar fragment, the 20 kDal fragment of S1 (Chaussepied et al., 1989). The formic acid digestion of S1A2 and S1A1 was described by Griffiths and Trayer (1989) and Sutoh (1983).

S1A1 or S1A2 prepared from rabbit myosin was freeze-dried from 10 mM sodium phosphate buffer, pH 7.2 [Na2HPO4=5.11 g NaH2PO4.2H2O=2.18 g for 5 L] with solid DTT added to 2 mM and sucrose to 4 mg per mg of protein. The freeze-dried powder was dissolved in 98% Aristar formic acid and then diluted to 70% to give a final concentration of protein of 5 mg.ml-1. The digestion was allowed to proceed for 18-40 hrs at 37°C, at which time the reaction was quenched by dilution 1:1 with water and then dialysis against 5 mM HCl (5 L) for at least 5 hrs. It was important to keep the solution cool (≈15-20°C) at this stage to avoid precipitation. The solution was freeze-dried at this stage and could then be stored at -20°C.

The preparation was soluble in 70% formic acid (5 ml per 60 mg) and could then be diluted out with water (total volume 15 ml per 60 mg) and dialysed overnight at 15-20°C in about 30 times volume excess 8 M urea A.R., 50 mM sodium phosphate, pH 7 (adjusted with 1 M phosphoric acid). The separation of the digestion fragments then began with the ion exchange chromatography on a Mono-S FPLC column. The pH of sample was adjusted prior to use from about 4-5 to 7, with 3 M NaOH. A typical loading was 2 ml. The column buffers were A: 50 mM sodium phosphate, 8 M urea, pH 7 (low uv absorbance urea was used, Fluka) B: 50 mM sodium phosphate, 8 M urea, 1 M NaCl, pH 7. The column was pre-equilibrated in buffer A (flow rate 1 ml.min-1) and the loaded sample was washed for 5 minutes in buffer A before application of the gradient. The gradient used was 0-50%B in 24 mins. The output was monitored at 220nm (the fragment of interest does not have any tryptophan residues). It was important to filter the sample prior to loading through a loosely packed glass wool pad and to wash the column well after each run in buffer B followed by 3 x 1 ml loadings of 70% formic acid, all to prevent column blockage. Two peaks were collected, one running through the column during the wash-on phase which consisted of formic acid, light chain and 26 kDal fragment. The peak eluting at 12-13 mins contained 26 kDal fragment (by SDS-PAGE). This peak containing the fragment was dialysed against 5-10 L 5 mM HCl at 15-20°C for at least 5 hrs and then freeze-dried for storage or used directly (redissolve in 70% formic acid and dialyse against 5 L of 5 mM HCl).

The solution was diluted to A280nm ≈ 2 against dialysis buffer and excess DTT and guanidine (1 g.ml-1) added. The solids were dissolved by gentle swirling and the sample the incubated for 30 mins at 37°C. 3 volumes of 95% ethanol (pre-incubated at -20°C) were then added and the mixture incubated at -20°C overnight. The mixture was spun in ethanol resistant plastic tubes at 20 K rpm at 4°C for 4 hrs to pellet the unwanted contaminant digest product. The ethanol was removed from the decanted supernatant by rotary-evaporation at room temperature, using a hair-drier to warm the flask as necessary. All the ethanol was gone when some precipitation of the guanidine due to water loss was seen. A little water was added to redissolve the precipitated guanidine and the solution dialysed against 5-10 litres 5 mM HCl (2 changes). The purity of the final product was checked by SDS-PAGE. The freeze died sample at this stage was soluble in either 70% formic acid or 100 mM Tris, pH 7.0. The final yield from 100 mg of freeze-dried protein was 5 mg of 26 kDal fragment which was approximately 90% pure by SDS-PAGE.

The fragment was then further purified by preparative SDS gel electrophoresis before sequencing.

The solubility of this fragment in most buffers was very low. In general the fragment could be dissolved in formic acid and then dialysed into other buffers like 0.1% SDS, 5mM Tris, 5mM HCl, 6M Guanidine HCl.

The protein (Myosin S1 26 kDal fragment) was used as a freeze-dried powder. The protein fragment was purified to better than 99% homogeneity by electroelution from a polyacrylamide gel. I am indebted to Mark Sckhiel, MRC Cambridge for help with this procedure.

A known amount of protein powder (approx. 0.5 mg) was dissolved in 50 ul of loading buffer and allowed to equilibrate at room temperature for approx. 15 minutes. The sample was then electrophoresed on Tris/12-22% gradient gel/3% stacking polyacrylamide mini-gel. The gel was stained and destained as briefly as possible and the band of interest excised using a razor blade on a clean glass sheet. The gels were always overloaded, but if there had been contaminant protein bands close to the band of interest a smaller loading of protein would have been used to increase resolution. The gel pieces were cut up into small cubes and the protein eluted from the polyacrylamide in an Isco Little Blue Tank apparatus using freshly prepared 0.1 M ammonium bicarbonate/0.1% SDS as the elution buffer. The final volume of approx. 1 ml solution (contents of microcup and washings) was dried in a Speedivac apparatus to 50 ul and the protein then ethanol-precipitated by addition of 19 volumes of ice-cold 95% ethanol, followed by incubation at -20°C for 20 hours. The precipitate was spun down in a bench centrifuge for 15 minutes/13 K rpm and the ethanol / water mixture carefully removed with a pipette to 50 ul. The residual fluid was removed on a Speedivac apparatus. If the protein fragment had not previously been carboxymethylated at cysteine residues, then complete reduction was ensured by dissolving the residue in 50 ul of 70% formic acid with 2 ul of 2-mercaptoethanol added, incubating the mixture under nitrogen for 30 minutes at room temperature and then diluting 1:10 with water before drying in a Speedivac apparatus.

The fragment was then sub-digestion with cyanogen bromide.

The protein fragment was dissolved in 70 ul of 70% Aristar formic acid and three small crystals of fresh cyanogen bromide added. The reaction was allowed to proceed at 37°C over a water-bath, in the dark and under nitrogen. During the reaction time (24-48 hours) further cyanogen bromide was added at about 4 hours and 24 hours. The reaction was terminated by diluting 1:10 with water and drying the mixture in Speedivac apparatus. The sample was redissolved in water and dried again to ensure complete removal of cyanogen bromide and formic acid. The subfragments so produced were analysed or purified by running on a 16.5%T/6%C/6 M urea/3% stacking polyacrylamide gel using the method of Schägger et al. (1987) followed by elecrtoblotting onto PVDF membrane as required.

The separation of small peptide fragments in the molecular weight range 100kDa to 1.5kDa requires a special SDS-PAGE system. The method is summarised in table 2.2.10.b and consists of a 4%T,3%C stacking/10%T,3%C spacer /16.5%T,6%C,6 M urea separating gel. Horse heart myoglobin cyanogen bromide digest fragments were used as markers (16950, 14400, 8160, 6210, 2510 kDa). The best separation was seen over the molecular weight range 10-30 kDa, but some degree of resolution was seen down to 1.5 kDa, as expected. The gel system is suitable for electroblotting of peptide fragments onto PVDF membrane for subsequent sequence analysis. The electroblotting method is summarised in table 2.2.11.

The sequencing of protein fragments has been found to often be made easier by pre-reduction and alkylation of cysteine residues in the protein fragment. The method described below was used in the preparation of the 26kDa fragment for sequencing.

80 mg of S1A2 previously digested in 70% formic acid for 20 hours, 37°C and then dialysed against 5 mM HCl before freeze-drying was dissolved in 20 ml 6 M guanidine-HCl, 500 mM Tris, 10 mM EDTA, pH 8.6 (adjusted with HCl) by stirring at room temperature for 30 minutes. 55 mg solid DTT (50 fold molar excess over cysteine residues) was added and the mixture stirred to dissolve for 2 hours. 160 mg of iodoacetic acid (1.1 fold molar excess over total thiols present) in 1 ml of 0.5 M NaOH was added and the reaction mixture stirred in the dark at room temperature for 1.5 hours. The reaction was quenched by addition of 2-mercaptoethanol to 1% and stirring. The mixture was dialysed to 10 L of 5 mM HCl overnight and then freeze-dried before use. The sample was first ethanol fractionated before electroelution from a polyacrylamide gel and subdigestion with cyanogen bromide.

Peptides were synthesized by the Merrifield solid phase method by Dr. J.Fox at Birmingham Univerisity Alta Bioscience on a Biotech Instruments BT7600 automated peptide synthesizer using Fmoc-amine (and occasionally t-Boc) protected amino acids. The coupling efficiency was monitored at each step. Typically 100-200 umoles were synthesized. For smaller quantities a multi-well system developed in-house was used. In some cases additional deblocking procedures were used to remove protective groups from the reactive peptide functions, for example Y933, the Y933 series, Y936 were additionally deblocked using 1:1 v/v dichloromethane:TFA mixture, in which the peptide was dissovled and left at room temperature for 12 hours.

Crude peptides were purified principally by HPLC in ≈1-5 umole batches (Vydac C-18, 5 um 300Å, 25 cm x 1 cm, flow rate 4.5-7 ml.min-1, with the detector at 210-230 nm) using a shallow gradient ≈0.5% min-1 increase in solvent B (solvent A: 0.1% trifluoroacetic acid in water; solvent B: 0.1% trifluoroacetic acid in acetonitrile), the precise gradient being tuned to the particular peptide. Other purification techniques were applied as required. It was often necessary to re-purify peptides to ensure better than 95% purity. This was either achieved by re-applying the partially purified peptide to the reverse phase C18 column, using the same gradient as before, or by pre-application to a ion exchange column (Fracrogel carboxyl column, in 10-50 mM ammonium acetate, eluted with upto a 1 M NaCl gradient). Either method appeared to work equally well. It was noticed that different preparations of peptides behaved identically in the various purification steps, but the quantity of contaminants could vary considerably between preparations. All peptides were freeze-dried and stored in 1-5 umole batches at -20°C. Peptide concentration was estimated from the absorbance at 210mn in pure samples where interfering compounds were not present (ε210=20.5 mg.ml-1cm-1; the peptide bond n_π* absorption at 210-220 nm swamps that of the side chains which absorb in this region) or by amino acid analysis (hydrolysed in 6 M HCl for 24 hours at 115°C) using a LKB 4400 analyser. Absorption at 210nm showed excellent correlation to quantative amino acid analysis.

In general during the preparation of these peptides it was noticed that the quality of the final product depended critically upon the quality of the synthetic product. The purification methods were poor at separating minor peptide based components from the principal product. Separation on HPLC was based mainly on the size of the peptide rather than its composition or sequence, except where an abnormally large number of either hydrophobic or charged residues were present. Hence peptides of an altered sequence (due to truncation during synthesis) were not efficiently separated except by cutting unresolved peaks (usually obtained from ion exchange chromatography) into subfractions and then analysing these to identify the fraction with the smallest quantity of contaminant.

For the purpose of illustration a typical example of peptide purification is given for peptide Y933.

Peptide sequence:

1Ala-2Asp-3Phe-4Lys-5Gln-6Arg-7Tyr-8Lys-9Val-10Leu

the peptide has a free amino terminus and an amidated carboxyl terminus

Peptide Y933 was synthesized by the standard F-moc chemistry using standard blocking groups to protect the functional groups of the reactive amino-acid residues (Asp, o-butyl; Lys, t-Boc; Arg, PMC free acid; Tyr, o-butyl). Deprotection was performed in 90% TFA/5% Phenol/5% water (1 hour) followed by an overnight treatment in 1:1 v/v dichloromethane:TFA mixture.

The first step was gel filtration of the crude synthetic product.

Approximately 100 umole of peptide was loaded in 5 ml of 0.1% TFA onto a G-10 gel filtration (2.5 x 160 cm = 785 ml) column at room temperature. Fractions of 5 ml volume were collected (flow rate 25 ml.hr-1). The peptide was detected by reading absorbance at 230nm and salt detected by measurement of conductivity. Following assessment of the suitability of the method by analytical HPLC (Vydac C18, 4.6 mm x 25 cm), Y933 was further purified by semi-preparative reverse phase HPLC (Vydac C18, 1 cm x 25 cm) using a 0.5% min-1 gradient with solvent A=0.1%TFA /water & solvent B=0.1%TFA /acetonitrile flow rate 4.5 ml.min-1. 5 umoles peptide dissolved in 1 ml of solvent A was loaded and the outflow monitored continuously at 220nm. The middle portion of the largest peak on the profile was collected, the initial and final portions of this peak being discarded because of contaminant peaks contained under the main peak. The collected fractions were freeze-dried directly. The purity of this fraction was assessed by running part of the sample on an analytical column under equivalent conditions (1.5 ml.min-1 0.5%.min-1 gradient solvents and solvent profile as above). No significant minor contaminants were identified. A final gel filtration of the sample (G-10, 0.1%TFA, 1.8cm x 52cm, 25ml.hr-1) was followed by freeze-drying and storage. Analysis of purified product by additional means was necessary in this case. The positively charged peptide was run on a Mono-S ion exchange FPLC column [strong cationic exchanger -CH2SO3- Pharmacia] (50 mM sodium phosphate pH 7.0) and fractions taken across the main peak to be analysed by 1-dimensional 1H NMR. As described in chapter 4 the expected impurities in this peptide were easily resolved from the peptide resonances. Fast atom bombardment mass spectroscopy was used to assess the identity of the species and the level of impurity. 2-dimensional 500 MHz proton NMR HOHAHA spectra in 90%1H2O / 10%D2O (60 msec MLEV-17 mixing time) were found to be very sensitive to impurities consisting of minor (5-20%) species of peptide. Specifically the backbone amide protons in minor species were usually shifted slightly from the resonances of the intact peptide. The absence of additional crosspeaks in the amide to sidechain region of this spectrum are a good indication of acceptable purity. Peptides were also routinely sequenced by N-terminal Edman degradation methods.

The purification procedures for other peptides are given below.

Y847: actin 77-94, NH2-TNWDDMEKIWHHTFYNEL-CONH2.

After application to a G10 gel filtration column (2.5 x 160 cm, room temperature, 25 ml.hr-1) the peptide was purified by HPLC (conditions: C18 Vydac 5 um 1 cm x 25 cm; Flow=4.5 ml.min-1; A=0.1% TFA in water; B=0.1% TFA in acetonitrile; Elution gradient: 24-40%B over 32 mins. The major peak eluted at 20 mins). Re-purification of the peak collected by re-application of the same gradient was necessary to ensure better than 95% purity from peptide contaminants. The peptide was freeze-dried from water three times to remove the residual acetonitrile. No final gel filtration step was necessary.

Y933/10 (Myosin 718-727, NH2-A-D-F-K-Q-R-Y-K-V-L-COOH )

Y933/8 (Myosin 720-727, NH2-F-K-Q-R-Y-K-V-L-COOH )

Y933/6 (Myosin 722-727, NH2-Q-R-Y-K-V-L-COOH )

Y933/4 (Myosin 724-727, NH2-Y-K-V-L-COOH )

Y975A (Myosin 724-727, NH2-Y-K-V-L-CONH2 )

Y975B (Myosin 724-727, CH3-CO-NH2-Y-K-V-L-CONH2 )

Y933AsnGlu D6 (Myosin 718-727, NH2-A-N-F-K-E-R-Y-K-V-L-CONH2 )

Y933AspGlu D5 (Myosin 718-727, NH2-A-D-F-K-E-R-Y-K-V-L-CONH2 )

Y1023A (Myosin 724-727, NH2-Y-K-V-L-COOH );15N-Val/15N-Leu

Y1023B (Myosin 724-727, NH2-A-D-F-K-Q-R-Y-K-V-L-COOH );15N-Val/Leu

The Y933 series peptides were prepared from a single peptide synthesis by removing a portion of the resin after addition of 4, 6, 8, and 10 residues. The Y975 series from were prepared from a single preparation, split before acetylation of one peptide. The Y670 series were synthesized in the 5 umol 'block' method. They all were purified by HPLC using the following conditions: C18 Vydac 5um 1 cm x 25 cm; Flow=4.5 ml.min-1; A=0.1% TFA in water; B=0.1% TFA in acetonitrile.

The individual details are:

Y933/10; Gradient: 10-30%B in 40 mins; major peak at 21 mins

Y933/8 ; Gradient: 10-30%B in 40 mins; major peak at 19 mins

Y933/6; Gradient: 5-20%B in 40 mins; major peak at 20 mins

Y933/4; Gradient: 5-25%B in 40 mins; major peak at 20 mins

Y975A; Gradient: 6-22%B in 32 mins; major peak at 19 mins

Y975B; Gradient: 9-29%B in 40 mins; major peak at 22 mins

Y670AsnGlu; Gradient: 10-30%B in 40 mins; major peak at 20 mins

Y670AspGlu; Gradient: 10-30%B in 41 mins; major peak at 20 mins

After application to a G10 gel filtration column (1.5 cnm x 35 cm), room temperature, 20 ml.hr-1, 12 ml fractions collected, peptides generally eluting in fractions 25-35. These peptide were freeze-dried.

Y983: Actin 338-348, NH3+-SVWIGGSILAS-CONH2.

The solubility of this peptide was very low in aqueous solution, but it was soluble in a 1:1:1 mixture of water/ dimethylformamide/ methanol. Interestingly upon standing for a minute the solution became jelly-like, but the gel could be redissolved by vortexing. The peptide was purified by HPLC using the following conditions: C18 Vydac 5 um 1 cm x 25 cm ; Flow=4.5 ml.min-1 ; A=0.1% TFA in water ; B=0.1% TFA in acetonitrile; Elution gradient: 15-55% B over 42 mins. The major peak eluted at 26 mins. The aqueous solubility of this peptide was reduced after purification and so it was not used further. The peptide was resynthesised, with the addition of several hydrophilic residues from the actin sequence, included to aid solubility. Together with A. Eastwood, this peptide has subsequently been shown to be alpha helical.

Y935: Myosin 50K 608-617, NH3+-GLYQKSAMKT-CONH2.

The peptide was purified by HPLC using the following conditions: C18 Vydac 5 um 1 cm x 25 cm ; Flow=4.5 ml.min-1 ; A=0.1% TFA in water ; B=0.1% TFA in acetonitrile, Elution gradient: 4-21% B over 34 mins. The major peak eluted at 18 mins.

Y941-M: Myosin 20 kDa 704-710, NH3+-IRITRKG-CONH2.

This peptide has been examined by Morita et al. (see section 5.8). The peptide was exclusively present as the monomer in the original preparation (acidic after deblocking). Upon dissolving part of the preparation in 0.1% TFA, the major species after 24 hours was found to be the dimer. The monomer was restored by addition of a large excess of 2-mercaptoethanol, which was then removed from the peptide solution within 30 mins. by application to a HPLC column. This peptide was purified by HPLC using the following conditions: C18 Vydac 5 um 1 cm x 25 cm; Flow=4.5 ml.min-1; A= 0.1% TFA in water; B= 0.1% TFA in acetonitrile, Elution gradient: 0-40% B over 40 mins. Two major peaks eluted at 17 and 22 mins, the latter peak being identified as the dimer since it was removed, with a concomitant increase in the first peak, after treatment with 2-mercaptoethanol (peptide at 1 mg.ml-1 mixed with an equal volume of neat 2-mercaptoethanol, the solution buffered by 100 mM Sodium Phosphate, pH 6.8, and left to stand for 30 mins. at room temperature, diluted 1:5 with water immediately before application to the column). For preparative loadings the major peak eluting at 27 mins. was collected (0-20% over 40 mins) which corresponds to the monomer in the original. An alternative purification was developed using ion exchange chromatography under the following conditions: Fractogel carboxyl column; A=10 mM ammonium acetate, pH 4, B=10 mM ammonium acetate, 1 M KCl, pH 4. 5.25 mg in 300 ul peptide in buffer was loaded. The gradient was 0-99% over 100 mins. The monomer and dimer eluted at 15 and 21 mins respectively.

A spin-labelled form of this peptide was produced.

2 umoles of peptide (by A210nm) were dissolved in 2 ml 100 mM sodium phosphates pH 7.9 and a 4 molar excess of 4-maleimido-tempo added in two equal aliquots from a 70 mM stock in dry-ethanol, 4 hours apart. The reaction allowed to continue for 12 hrs. The reaction mixture was fractionated by gel-filtration [conditions: G10 1.8 x 52 cm ; flow rate 25 ml/hr ; buffer 100 mM sodium phosphates pH 7.9 ; 5 min fractions collected (monitored over range 210-300 nm); fractions 28-42 were pooled. Another larger peak eluted in fractions 44-60. The product was purified by HPLC [ conditions: C18 Vydac 5um 1cm x25 cm; Flow=4.5 ml/min; A=0.1% TFA in water; B=0.1% TFA in CH3CN; Elution gradient: 0-23% over 46 mins. The peak containing the spin-labelled peptide was identified as eluting at 31 mins. Many other peaks were seen over the elution profile arising from the spin-label and impurities and decomposition products of the label (principal peaks at 22 and 41 mins).

Y941: Myosin 20 kDa 704-710, NH3+-IRITRKG-CONH2.

A variety of Y941-M was prepared with the cysteine replaced by a threonine residue. The peptide was purified by HPLC as above with an elution gradient 2-20% over 36 mins. with the major peak eluting at 18 mins.

The synthesis follows the method provided by Milligen/Biosearch with minor modifications and the solvent system for TLC is from Calbiochem (1990 Catalogue pp.392-405).

Valine-15N (100mg, 1.31 mmoles) was dissolved in saturated sodium carbonate (5.3 ml), and added to acetone (3.9ml), and the resulting solution was stirred in ice. Fluorenylmethyl succinimidyl carbonate (FMOCsu) was gradually added as a solid to the cooled stirred amino acid solution, over 10 min [normalised ratios: 100 mmol of free acid; 200 ml acetone; 267 ml sodium carbonate; 105 mmol FMOCOSu].

The reaction mixture was left stirring on ice for two hours, and then left to stir at room temperature for several more hours or overnight. The progress of the reaction was monitored by TLC (85% chloroform:10% methanol:5% acetic acid; K6F Whatman silica gel plates with a fluorescent indicator; commercial Fmoc-Val with Rf 0.6 as TLC standard). Ethyl acetate (20ml) and water (20ml) was added to the reaction mixture and 6M HCl was carefully added until the pH of the aqueous phase was 2.0. The ethyl acetate phase was separated from the aqueous phase in a separating funnel and the aqueous phase was extracted with another 20ml of ethyl acetate. The combined ethyl acetate phases containing the Fmoc 15N-valine were washed with acidified water of pH 2-3 (3 x 20ml), dried with magnesium sulphate and rotary evaporated to dryness. The crude product was recrystallised from ethyl acetate and petroleum ether 40-60. This was achieved by dissolving the crude Fmoc-valine product in the minimum amount of hot ethyl acetate over a 70°C water bath and adding warm petroleum ether to the solution until it becomes cloudy. The resulting mixture was then left at room temperature overnight. The crystals obtained were filtered and washed with petroleum ether and dried in a vacuum desiccator. Multiple further recrystallisations of the product was necessary to achieve a purity of better than 95% as assessed by TLC.

Fmoc-15N valine (48% yield) was obtained, and stored desiccated in the freezer.

Fmoc-leucine-15N was produced in a similar manner with a yield of 52%. These products were then used directly for peptide synthesis without activation of the carbonyl function as a pentafluorophenyl ester.

N-[4-(Trifluoromethyl)phenyl]iodoacetamide [p-CF3-Ar-NH-CO-CH2I] was prepared by the method of Shriver and Sykes (1982). Iodoacetic acid (2.7g, 14.5mol) was dissolved in dry ethyl acetate (15ml), and dicyclohexylcarbodiimide (1.53g, 7.4mmol, dissolved in 10ml of ethyl acetate) was gradually added with a funnel over a period of 2 minutes. The resulting solution was stirred for one hour at room temperature, (CO2 produced released via a drying tube) and the dicyclohexylurea was removed by filtration. The iodoacetic anhydride-ethyl acetate solution was immediately added to 4-aminobenzotrifluoride (920ul, 7.3 mmol) and left at room temperature for one hour. TLC analysis (K6F Whatman silica, with fluorescent indicator, 50% chloroform:50% petroleum ether) showed that the starting material (Rf=0.5) had been quantitatively converted to a new product (Rf=0.1) by this time. The solvent was then removed by rotary evaporation, and the residue dissolved in benzene (50ml), and washed 4 times with 0.1M sodium hydroxide solution. The benzene was dried with anhydrous magnesium sulphate, and removed by rotary evaporation. The residue was twice recrystallised from chloroform, twice from a petroleum ether / chloroform mixture, and finally from toluene. White crystals were obtained which were dried under vacuum to yield 300mg (14% yield) of product which was homogeneous by TLC and NMR. The 1H-NMR, 13C-NMR and 19F-NMR spectra were consistent with the structure of the product.

o-Iodosobenzoic acid [formula Ph(COOH)(IO)] cleaves proteins on the carboxyl side of tryptophanyl residues, leaving a free N-terminus for the succeeding peptide (Mahoney et al., 1981). Commercial o-Iodosobenzoate contains o-iodoxybenzoic acid (Mahoney et al., 1981) as an impurity which is known to promote cleavage at some tyrosyl residues. This side reaction can be decreased by preincubation with p-cresol, which because of its chemical similarity undergoes a similar reaction to tyrosine. o-Iodosobenzoic acid is virtually insoluble in 80% acetic acid and so the reaction vessel must be constantly agitated. This is conveniently done in an Eppindorff tube clipped to a rotating vertical wheel. This of course provides a means of removing reagent at the end of the reaction, since it can be simply spun out. Alternately guanidine hydrochloride can be added to the reaction mixture until o-iodosobenzoate easily dissolves (eg. 4M). Urea should not be used (Mahoney et al., 1981).

o-Iodosobenzoate was preincubated with p-cresol (1:2 p-cresol:o-iodosobenzoate, molar ratios, 1mg o-iodosobenzoate= 0.5ul p-cresol) in 80% acetic acid (10 mg iodosobenzoate / 100ul 80% acetic acid) for 4-6 hours. Actin was dissolved in 80% acetic acid (5mg.ml-1) and added to the appropriate amount of preincubated o-Iodosobenzoate (actin:o-iodosobenzoate 1:2, by weight) and the digestion allowed to proceed at room temperature in the dark for 16 hours. The digestion was terminated by spinning out the reagent in a bench centrifuge. The pellet was resuspended and washed in 80% acetic acid and the pooled supernatants diluted 1:10 with water and freeze-dried. 2-mercaptoethanol or dithiothreitol will reduce unreacted reagent and terminate the reaction if guanidine hydrochloride is added. The dry residue was then analysed by SDS-Page.

The steady state ATP hydrolysis rates of S1A2 and S1A1 were measured over a period of 10 minutes by the rate of Pi release (Fiske & Subbarrow, 1925). The total assay volume was 1.0ml. The solutions used were as follows:

ammonium/EDTA ATPase: 25mM TEA-HCl pH 7.5, 0.5M NH4Cl pH 7.5, 10mM EDTA pH 7.5, 0.005mg S1A1, 4mM ATP pH 7.5.

Calcium (high salt) ATPase: 50mM TEA-HCl pH 7.5, 0.6M KCl, 10mM CaCl2, 0.0165mg S1A1, 4mM ATP.

Actin-activated MgATPase: 25mM TEA-HCl pH 7.5, 2.5mM MgCl2, 0.006mg S1A1, 2mM ATP, F-actin varied over the range 0 to 50uM.

The assays always included appropriate controls and were made up stored on ice with the nucleotide added as the final component to start assay. The assay was terminated after up to 10 minutes incubation at 25°C by addition of 20% trichloroacetic acid (0.2ml), and vortexed and then spun in a microfuge for 5 minutes to remove the precipitated protein. The assay mixtures (1.2ml) were then added to a solution of dilute acid molybdate (3.8ml), followed by reducer (0.2ml), and the solution mixed and left to stand at room temperature for 10 minutes. The absorbance at 730 nm was measured for each tube, and the amount of phosphate liberated in each assay calculated from a phosphate calibration curve constructed with a phosphate standard of known concentration (eg. 0-1umol.ml-1). The calibration curve tended to vary with the age of the reagents used and so was repeated daily.

The dilute acid molybdate (0.36 M H2SO4, 0.329% ammonium molybdate), was made up by dissolving 1.65g ammonium molybdate in 490 ml water with 10ml conc. H2SO4. The reducer solution was prepared by dissolving 15g sodium metabisulphite (Na2S2O5) and 0.5g sodium sulphite (Na2SO3) in 100ml water. 0.25g ANSA (1-amino-2-napthol-4-sulphonic acid) was then added, and the suspension stirred in the dark for 15 minutes before filtration. The solution was stored in a dark bottle to protect it from light.

The method used was based on that of Shriver & Sykes (1982). Shell frozen S1A2 (3ml, 6mg.ml-1) was thawed and dialysed against two changes of 50mM TEA-HCl pH 8.0 (5L) for 5 hours. The S1A2 solution was then centrifuged for 30 minutes at 50 Krpm, and the concentration of the supernatant determined (5.5mg.ml-1;ε280nm = 0.8 [mg.ml-1]-1.cm-1). The S1A2 was then dialysed into 50mM TEA-HCl, 2mM tetrasodium pyrophosphate, 4mM MgCl2, pH 7.9, and was labelled by the addition of 1.5 fold molar excess of a 50mM solution of N-[4-(Trifluoromethyl)phenyl]iodoacetamide dissolved in dry ethanol. The reaction was allowed to proceed for 10 minutes at 0°C and then terminated by the addition of 200 fold molar excess of DTT. Ammonium sulphate (0.74g to give a 60% saturated solution) was added to precipitate the S1A2, and the resulting suspension was centrifuged at 30 Krpm for 5 mins and the pellet resuspended in 50mM TEA pH 8.0 (10ml). The S1A2 was precipitated again by adding ammonium sulphate (0.37g), and collected by centrifugation. The pellet was redissolved in 50mM TEA-HCl pH 8.0 (2ml) and applied to a G-25 Sephadex gel filtration column (1cm x 60cm, pre-equilibrated and run in 50mM TEA-HCl pH 8.0, flow rate 25ml.hr-1). Fractions were collected and the protein detected by absorbance at 280nm. A single sharp protein peak was observed and the peaks pooled and concentrated to around 25mg.ml-1 in dialysis tubing with acquacide, followed by dialysis against 10mM sodium phosphate pH 7.5 (2 x 5L) for 5 hours. The labelled S1A2 showed a 2.7 fold elevation of calcium ATPase activity relative to unlabelled S1A2, whilst the ammonium / EDTA ATPase was 80% lower than that of the unlabelled protein. This indicated that ≈80% of the protein had been labelled on the SH1 reactive thiol.

The same procedure was used to label the S1A2 on the SH2 reactive thiol except that the protein in the labelling buffer was pre-treated with 1.5 fold molar excess N-ethyl-maleimide. This reaction was allowed to proceed for 40 minutes on ice, and then terminated by addition of a 200 fold molar excess of DTT. The N-ethyl maleimide SH1 labelled S1A2 was purified by ammonium sulphate precipitation and gel filtration, and then treated with 1.5 fold molar excess of N-[4-(Trifluoromethyl)phenyl]iodoacetamide, in 50mM TEA-HCl, 2mM MgADP on ice for 45 minutes, before being quenched by DTT (Reisler et al, 1974). The preparation was then purified as above.

Peptides Y847 and Y935 each contain a single tyrosine residue (91Y) which was readily iodinated with 125I sodium iodide. Iodo-gen (2mg) was dissolved in dry chloroform (0.5 ml) in an Eppindorf tube and then dried by blowing nitrogen in to the tube. This left the iodo-gen evenly coated around the sides of the tube. Typically a 5mM solution of peptide in 50mM sodium phosphate pH 7.4 (50ul, 0.25umol) was added to the dried iodo-gen followed by 300ul of water and 2ul of source 125I (original activity 100 mCi.ml-1, 3 months old). The tube was sealed and left at room temperature for 15 minutes, with frequent shaking. 50mM potassium iodide (25 ul) was added to ensure complete iodination of the peptide and the reaction allowed to proceed for a further 10 minutes. The iodination mixture was filtered through a syringe microfilter to remove free floating iodo-gen and then applied to a Waters SEP-PAK C18 cartridge (1ml bed volume, pre-wetted with water). The cartridge was washed with water until the gamma count of the eluate was much reduced. At this point it was assumed that most of the free iodide which had not been incorporated into the peptide had been removed. Generally 10-20ml of water were required and about 50% of the original counts (as estimated with the gamma counter) remained bound to the column. The peptide was eluted with 50% acetonitrile / 50% water (≈10 ml), the eluate being collected until no further counts were lost from the column. The column was still substantially radioactive, but the vast majority of the counts had been removed. The peptide was freeze-dried. Iodine (I3-) acts as an electrophilic reagent and mono- or di- substitutes at the 3,5 positions of tyrosine. (Means & Feeny, 1971) The extent of iodination was not determined.

13C and 125I Autoradiography was carried out according to the procedures described in the Amersham Review 23: Radioisotope detection by fluorography and intensifying screens.

A one-step crosslinking procedure using the water-soluble carbodiimide crosslinking reagent EDC was used to crosslink peptides to actin and S1.

Shell frozen S1A2 was thawed and dialysed against 10mM HEPES-NaOH pH 7.0 (2 x 5L, No DTT added), and centrifuged at 30 K rpm for 30 minutes. The S1A2 was diluted to 2 mg.ml-1 and a portion of this solution (2ml, 17.5uM, 35nmol S1A2) was added to approximately 100nmol of iodinated Y847 peptide and preincubated on ice for 5 mins. Then a freshly prepared solution of EDC (final concentration 4.7mM) in 0.5ml of 100mM HEPES (pH with NaOH), pH 7.0 was added. The crosslinking reaction was allowed to proceed at room temperature for 1 hour, and then terminated by addition of 2-mercaptoethanol (10ul, 143umol) to quench the unreacted EDC. The solution could then be stored at 4°C.

In some early experiment imidazole buffer (10mM) was used, but the yields were reduced as compared to those using HEPES, presumably because of some reaction of the EDC with the buffer.

A similar procedure was used to crosslink peptide Y935 to F-actin. The final concentration of F-actin used in the reaction mixture was 2mg.ml-1. When a final concentration of 5mg.ml-1 F-actin was used crosslinking of actin momomer in to dimers was observed. The molar ratio of peptide to actin used in the reaction mixture was always 1:1

Sulpho-NHS was routinely added to the reaction mixtures (N-hydroxysulphosuccinimide freshly made up, at a ratio of 1:1 EDC:Sulpho-NHS). Sulpho-NHS has been reported to enhance the yeilds obtained in water-soluble carbodiimide-mediated coupling reactions (Staros et al., 1986) by prolonging the lifetime of the activated carboxyl group in aqueous solution.

Crosslinking reaction mixtures were analysed by SDS-PAGE and autoradiography.

TLCK (L-1-(tosylamido)-2-phenyl chloromethyl ketone) treated trypsin as a freeze-dried powder (1-2 mg) was dissolved in 1 mM HCl (1 ml), and the concentration determined (ε280(trypsin) = 1.43 (mg.ml-1)-1 cm-1). Trypsin was added at the ratio 1:200 to S1A2 (2 mg.ml-1, 0.2-0.5 ml). The digestion was allowed to proceed at room temperature and at various time intervals aliquots (20 ul) were removed and quenched by addition to a mixture containing 10% SDS (5 ul), 2-mercaptoethanol (2 ul), and 1:1 bromophenol blue:glycerol (3 ul). The solution was boiled for 3 minutes and analysed by SDS-PAGE and autoradiography.

Following the procedure described above, S1A2 (100 ul, 2 mg.ml-1) was freeze-dried, and the residue dissolved in 70% formic acid (40 ul) and the digestion allowed to proceed for 24 hours at 37°C. The mixture was diluted to 1 ml with water and freeze-dried, and the residue redissolved in 0.1 M Tris, 0.1 M Bicine, pH 8.3 (40 ul) and 10% SDS (20 ul), 2-mercaptoethanol (8 ul), and 1:1 bromophenol blue:glycerol (12 ul). The solution was boiled for 3 minutes and analysed by SDS-PAGE and autoradiography.

Following the method of Chaussepied et al. (1986a & 1986b) S1A2 was also digested with thrombin. S1A2 (1.5 ml, 2mg.ml-1, 27 nmol) was dialysed overnight against 40 mM HEPES-NaOH pH 8.0 (5 L). A fresh 10 mM solution DTNB (5,5'-dithiobis(2-nitrobenzoic acid) pH 8.0 in 40 mM HEPES-NaOH was prepared and 29 ul (29 nmol) added to the S1A2 solution, with 0.1 M MgCl2 (10 ul), 43 mM ADP (23 ul), and the reaction allowed to proceed for 16 hours at 4°C. The DTNB labelled S1A2 was purified by two rounds of ammonium sulphate precipitation (as described above) and then dialysed at 3 mg.ml-1 for 2 hours against 50 mM HEPES-NaOH pH 8.0 (3 x 1 L). The protein was the equilibrated on a 25°C water bath and 20 ul of freshly prepared 1 mg.ml-1 thrombin added. A time course of the digestion was again performed.

Peptides Y847 and Y935 were covalently crosslinked with EDC to S1A2 and actin respectively following iodination with 125I. The interaction site was then investigated by digestion of the cross-linked products, followed by separation on SDS-PAGE, electroblotting, and sequencing. All the procedures are described above.

NMR is simply a form of spectroscopy. Certain nuclei possess the property of spin, providing these nuclei with a magnetic moment that can interact with applied magnetic fields. For a nucleus of spin I, there are 2I+1 different energy levels (nuclear spin states) corresponding to a different interaction of the magnetic moment with an applied magnetic field. NMR is usually confined to nuclei with spin numbers of 1/2, for simplicity of interpretation of the spectra, since only two spin states exist for each nucleus (1H, 13C, 15N, 31P). In one of the two energy states, the magnetic moment is aligned with the applied magnetic field. This is the low energy state. In the high energy state the magnetic moment is aligned against the applied magnetic field. Transitions occur between the two energy states and can be excited by an applied oscillating electromagnetic field. The energy of transition between these two energy levels (resonance condition) is given by the Plank equation (DE=hv becomes DE= chB0/2P, where c is the magnetogyric ratio characteristic of each nucleus, and the resonance frequency is given by cB0/2P). In the NMR experiment, each nucleus, at a particular field strength (B0), absorbs a characteristic frequency during the transition process, this frequency lying in the radio region of the electromagnetic spectrum. The population difference between the two energy states at equilibrium is very small, so that the net absorption of energy in the NMR experiment is small, and the technique is relatively insensitive. The population difference is greater at higher field strengths.

The NMR experiment is best described in terms of a classical physical process. The interaction of the nuclear magnetic moment with the applied magnetic field gives rise to a torque (angular force), which results in the precession of the magnetic moment around B0. The nuclear magnetic moment precesses at the Lamor frequency, w0 (w0=cB0). In any population of nuclei, many individual spins are precessing, providing a net resultant magnetisation (vector) parallel to the applied magnetic field. At equilibrium the population difference causes this resultant to be aligned parallel to B0. In an NMR experiment the signal perpendicular to B0, in the xy plane is observed. At equilibrium, there is hence no NMR signal. The signal is induced by applying a second magnetic field, B1 in the xy plane. The net magnetisation interacts with B1, a torque is induced, and the net magnetisation hence precesses about B1. If the B1 field is applied for a suitable length of time, the net magnetisation is flipped into the xy plane (a 90° pulse). When the field B1 is switched off, the net magnetisation precesses about B0 at the Lamour frequency. If this NMR signal is detected from a single direction, the signal hence oscillates between a positive maximum and a negative minimum. The time domain of this signal may then be Fourier transformed to provide a frequency domain NMR spectrum. In time the signal decays as the system relaxes back to equilibrium.

The precise position of a given nucleus in the spectrum is dominated by the applied B0, but local fields induced in the electron clouds of adjacent nuclei by the applied B0 (opposing B0, hence a shielding effect), change the effective field experienced by any particular nucleus, and hence the precise absorption frequency. Hence the NMR spectrum is generated, by sensitivity of nuclei to their local magnetic environments. The spectrum is described by a chemical shift scale, d = VREF-VOBS/VREF x 106, in ppm, which is independent of field strength. The chemical shift of a particular nucleus (absorption frequency) is contributed by the primary shift arising from the covalent character of its environment (eg. methyl group) and by a secondary shift arising as a result of neighbouring groups, both covalently bonded and through space (eg. aromatic rings). Non-isolated non-equivalent nuclei are coupled together by weak interactions of their magnetic fields, communicated by electrons across a chemical bond. This leads to peak multiplicity. A nucleus coupled to N spins is split in to 2N+1 lines, with the intensity of the individual lines described by binomial coefficients. The coupling constant, the chemical shift value between split resonances is related to factors such as bond angle and the number of bonds over which the coupling is communicated. Peak intensity, the area under a given resonance is proportional to the number of nuclei contributing to the peak. Peak shape is related to the relaxation rate (width at half-height Hz= 1/ PT2) superimposed on a Lorenzian lineshape.

The NMR spectrum is observed because there is a net absorption of energy by the system, corresponding to a change in the population distribution between the high and low energy states. If the population in each state becomes equal then the system is saturated, and no NMR signal can be observed. However, relaxation processes provide a means of energy leakage from an individual nucleus, so that the system tends to return to equilibrium, in which there is a population difference, and hence saturation is avoided. T1, the longitudinal relaxation time, is the time constant that describes the recovery of the Mz component of net magnetisation (aligned with B0) to the equilibrium value of M0 after a perturbation. T2, the transverse relaxation time is the time constant that describes the decay of the Mxy components (in the xy plane) to the equilibrium value of zero. Relaxation occurs via fluctuations in magnetic fields caused by random Brownian motion. These fields arise from dipole-dipole interactions between pairs of nuclear magnetic moments. As magnetisation recovers in Mz, the vector in xy must decay, and hence T1 would be expected to be the same as T2. However, the Mxy components may decay by other routes, principally because of local inhomogenities in the magnetic environment, so that in general T2 processes are more efficient (T2<T1).

The standard 1-dimensional 1H sequence is seen in figure 2.3.1. An approximately 50° pulse was routinely used with a presaturation pulse of 1.5 seconds. The 90° pulse was calibrated on the water solvent signal by division of a 180° pulse. Spectra were usually collected with the water solvent peak at the carrier (quadrature detection). In aqueous solutions the spectrometer was locked to added 5-10% 2H2O, and in d4methanol or d3TFE mixtures the signal of the organic solvent was used for the lock.

The cpa sequence was also routinely used (figure 2.3.2). In this sequence less mobile and hence broad components are filtered out of spectra, for example added protein in peptide mixtures. This filtering arises because during the refocusing period the broad signals (short T2) have largely relaxed. To modulate the spectrum by the proton-proton coupling constants, a refocusing delay of 60 msec was used (1/2J, J≈7.5 Hz) and hence doublets were of opposite sign to singlets and triplets. This simple pulse was very useful in initial assignments of peaks, for example the only methyl triplet is from isoleucine, among the other methyl doublets.

Spectra were generally collected on a Bruker AMX500 instrument using a dedicated 5mm 1H probehead or broadband inverse 5mm probehead (equipped with 1H coil and X (15N to 31P) coil tunable and matchable in both frequency ranges) with the temperature controlled by a Haake cooling bath. Some spectra were collected on a GX270-JEOL instrument equipped with matrix-shim and a fluorine transmitter, using either a 1H/13C 5mm (tunable only) probehead or a 10mm multinuclear probehead (proton coil tuned to 1H or 19F, and broadband coil tuneable and matchable from 15N to 31P). Temperature was reduced below ambient if required using a home-made (KJS) cooling system based on a chromium-nickle coil in a dry ice-acetone bath.

A Lorenzian line-broading of between 0.3-1 Hz was routinely applied to spectra. Sample volumes were typically 2.5ml (10mm probe), 0.6ml (5mm Jeol probe), 0.51 ml (Bruker 5 mm probes). Generally sodium 3-(trimethlysilyl) propionate was used as the internal standard for proton spectra in water (δ=0 ppm), or occasionally dioxane (3.71 ppm). 13C spectra were referenced against glycine Cα (42 ppm) or TSS (0 ppm). 19F spectra were referenced against an external sample of trifluoroacetic acid, the deuterium lock being used as an intermediate reference point (Shriver & Sykes, 1982). The chemical shift CD3OH is 3.30 ppm with respect to TMS.

2 dimensional spectroscopy is a technique to aid the resolution of peaks which overlap in the 1 dimensional spectrum. The 1 dimensional spectrum lies along the diagonal of the 2 dimensional spectrum, and off-diagonal crosspeaks are seen for peaks on the diagonal which are related either by spin-spin coupling or close proximity, depending upon the experiment. The collection of the FID (free induction decay) provides one dimension (F2), with F1 being provided by a delay period which is regularly incremented from row to row. Information concerning the behaviour of spins during this delay period is present in the time domain (FID) only as phase or amplitude modulation of the normally observed spectral frequencies. After transformation of the rows collected, the columns may be seen to correspond to a second series of FIDs, which are themselves transformed to provide the 2-dimensional spectrum.

All spectra were collected on a Bruker AMX500 instrument. The individual pulse schemes are discussed below, with particular emphasis placed on the relatively new experiments TOCSY and ROESY, and upon NOESY which is the basis of most of the structural determinations. All spectra were collected in the phase sensitive mode with quadrature detection in F1 using TPPI (Dronby et al., 1979; Bodenhausen et al., 1980). Most spectra were baseline corrected in both dimensions using a third-order polynomial baseline fitting routine after tranformation and phasing. All spectra were processed using UXNMR on a Bruker X32 computer.

To supplement the assignment of peptides a COSY spectrum was generally collected. The familiar pulse scheme is seen in figure 2.3.3. If NH-CHα coupling constants were to be determined the resolution in the F2 dimension was as high as possible (8 K points) and 512 increments collected. In cases where signal to noise would be sufficiently high a double quantum filtered COSY (DFQ-COSY; Piantini et al., 1982) experiment was used because of the in-phase nature of the diagonal peaks. Briefly, the first 90° pulse creates transverse magnetisation after which, the systematically varied delay allows the magnetisation to evolve. The second 90° pulse ends the evolution period by returning the transverse magnetisation back to the z-axis. Those magnetisation components (vector) that are parallel to the second pulse are not affected by the second 90° pulse, and are left in the x-y plane for observation. Thus the second 90° pulse causes a redistribution, or mixing, of the population among the energy levels of the spin system by returning only some of the magnetisation back to the z-axis. The systematic variation of the time delay provides the second frequency domain between rows, and the interaction between coupled spins is detected since the evolution is modulated by coupling to neighbours. Hence the coupling partner of any peak is detected as a crosspeak with its own frequency in the F2 dimension and the modulating frequency of the partners in the F1 dimension. Coupling is felt reciprocally by both partners and so the spectrum is symmetrical about the diagonal.

This 2-dimensional experiment is known as HOHAHA or TOCSY (total correlation spectroscopy, Bax & Davis, 1985) The experiment involves the correlation of all protons within a scalar coupling network and is used for resonance assignment as an alternative to COSY and Relayed-COSY experiments. Coherence is efficiently transferred through multiple bonds and crosspeak intensity is maximized by the in-phase character of the multiplets, but the latter may reduce resolution of closely overlapping peaks as opposed to a phase sensitive experiment, like DQF-COSY. Some of the important aspects of the experiment are discussed below.

The pulse scheme used during this work is illustrated in figure 2.3.4. Magnetisation transfer, between spins that are scalar coupled, during the mixing period occurs via the Hartmann-Hann effect. During the spin-lock pulse magnetisation is exchanged between coupled spins (the transfer is coherent and could be described as spin flipping), but the Hartmann-Hahn transfer requires that these spins have little effective difference in chemical shift, such that difference in the field experienced by them is much less than the effective coupling. This condition is met by either the application of a very strong rf field (impractical since too high a field would need to be generated) or a relatively strong composite pulse similar to those used in heteronuclear decoupling. It is pertinent to the later discussion of the ROESY experiment to note that, for the special condition where the carrier is midway between the two resonances being considered, that a relatively weak rf will also achieve magnetisation transfer (also note that the magnitude of transfer must therefore be strongly dependent on the resonance offset). An MLEV-17 scheme was used to achieve mixing (pulses must be coherent with the receiver and so the transmitter is used to deliver the pulses, with the ecoupler rapid power switch being used to reduce the transmitter power). It is often recommended that a trim pulse be added before and after the mixing period (approximately 2 msec) to eliminate unwanted magnetisation which is preserved by the MLEV-17 sequence. However the antidiagonal artifact which is expected in the absence of trim pulses was never observed in practise during this study, and the excitation of the water solvent by the trim pulse was considered to create too great a response in the final spectrum to be of benefit. The MLEV-17 sequence is relatively forgiving of inexact calibration of the low power 90° pulse. For 10 ppm at 500 MHz a rf field 10-12.5 kHz is required. Typically a power level (about 10-13 dB) was chosen. The pulse was cycled to give a mixing time of 60 msec in most instances which provided a virtually complete connectivity of amide to sidechain protons. Heating of the sample was not considered a problem, but approx. 5 minutes of dummy scans was run to allow equilibration. No evidence was ever seen for transverse NOE effects generated during the spin lock (these would be of the opposite sign to the Hartmann-Hann peaks).

The pulse was implemented with a low power pulse for presaturation of the solvent in most studies and later, with a 1-1 jump return pulse for non-excitation of the water peak. Saturation of peaks directly below the solvent peak caused bleaching of these α-proton resonances, but a little way from the solvent, although peaks were substantially reduced in size by saturation, the magnetisation transfer was found efficient enough that peaks were strong in the 2-dimensional spectrum. Some bleaching of amide resonances in the rapidly exchanging amide proton region may have occurred. At mixing times reduced below 60msec the transfer was reduced to one bond (15msec) but the sensitivity was much reduced and so a COSY-type experiment was preferred. At 60msec the efficient transfer of magnetisation, even for the small NH-CHα couplings in an alpha-helical peptide or between the ring protons and ί-protons in histidine sidechains, gave excellent sensitivity The recycle time used for experiments was usually about 1.7 sec (1.5 sec presaturation and 0.2 sec acquisition).

Cavanagh et al. (1990) have given expressions for isotropic mixing behaviour and have produced a catalogue of intensity of crosspeak vs. mixing time for various amino acids. The upshot is that, the behaviour oscillates irregularly over mixing times up to 250 msec according to residue, topology and coupling constant, so the lack of a peak in any individual case may reflect a minimum in a curve. This can account for the loss of some peaks from TOCSY spectra. Hence, the use of increasing mixing times to identify subsequent spins in a particular spin system may be dangerous since the intensities of some crosspeaks at certain mixing times may fall to zero while others are unusually large. Hence this technique was never used, and COSY or Relayed COSY were used as required.

The most important property which is measured in NMR is interatomic distance, since this allows structural and conformational information to be obtained about the molecule under investigation. Distance is measured by the nuclear Overhauser effect, with a maximum range of about 5Å. The NOE is measured using two classes of 1-dimensional experiment. In the truncated driven or steady state NOE (figure 2.3.5) a resonance line is saturated by a weak, selective irradiation during a preirradiation time, followed by a non-selective observation pulse and acquisition of an FID. The NOE is manifest as a small change in the intensity of resonances of nuclei which are close in space to the preirradiated nucleus. The small change is detected by subtraction of a reference spectrum in which the preirradation is offset into an empty region of the spectrum. In the second experiment, the transient NOE, the long preirradiation period is replaced by a selective 180° pulse and a delay interval.

The dominant relaxation mechanism for spin 1/2 nuclei arises from dipolar interactions with other spins. Hence the fluctuating magnetic fields required for relaxation (in the frequency range corresponding to the energy of the transition involved in the relaxation process) are provided by the tumbling of protons. The NOE is a change in the intensity of an NMR resonance because the population distribution between the two energy states has been perturbed by the transfer from a nearby nucleus (cross-relaxation), the transfer arising as a consequence of the modulation of dipole-dipole coupling between the different nuclear spins by the Brownian motion of molecules in solution. It is clear then that the motion of molecules in solution will have important consequences for the way in which the NOE develops, and that the NOE will be distance dependent (the NOE is related to the distance (r) between the nuclei involved as a function of r-6, so that doubling the distance results in a decrease in cross relaxation rate (σn) by a factor of 64).

The efficency of the two relaxation processes T1 and T2 varies with molecular rotation (Tc) as shown in the figure 2.3.6. T2 processes (spin-spin relaxation) are more efficient at relatively low tumbling frequencies (energy conserving exchange of magnetisation between spins close in energy). These conditions are found in large molecules where Tc is long compared to w0-1 (rotational motion slow). T1 processes (spin-latice) are most efficient (minimum T1) when the frequency of molecular tumbling which promotes relaxation is close to the Lamour frequency w0 (w0-1≈Tc; loss of energy to the solvent). In the case of extreme motional narrowing in which the Tc is very short (eg. methanol≈10-12sec, w0≈10-9 for 500 MHz 1H) the proportion of the molecules rotating in frequencies around w0-1 is low and hence while T1 relaxation processes dominate they are inefficient and T1 is long (T1=T2; both increase in proportion to Tc-1).

During the steady state NOE experiment, one spin is saturated (there are an equal number of nuclei in high energy and low energy states), and following a pulse to excite neighbouring spins, the system relaxes. The saturation of the first spin changes the equilibrium magnetisation state of a second nucleus which is close in space, and hence changes the intensity of the signal observed for the second spin (the NOE). According to the dominant relaxation process, the NOE may be positive (an increase in the intensity of the non-irriadiated peak, for a rapidly rotating molecule) or negative for a slowly tumbling molecule.

The precise dependence of the cross-relaxation rate σn on molecular motion is shown below, hence where,

for w0Tc = √5/2 then σn = zero

for w0Tc > √5/2 then σn = negative

for w0Tc < √5/2 then σn = positive

Hence for molecules of intermediate size, the NOE response falls to near zero. Peptides of around ten residues are in this class of molecule.

In the steady state experiment, if the saturating irradiation is continued for a long time, then eventually a steady state will be reached in which the relaxation matches NOE buildup. In the homonuclear case where dipole-dipole relaxation is the only relaxation process, the value of this steady state enhancement will be +0.5 for rapid motion and -1 for very slow motion, irrespective of the distance between the nuclei. However the rate at which the enhancement develops is proportional to the cross relaxation rate and hence distance related. Incoherent exchange of magnetisation by cross relaxation (phase is not preserved) is a slow process, often occurring over a second or more, so that the buildup of the NOE can be followed by using different mixing periods, up to a maximum of around 1 second.

The principal problem encountered in NOE spectra is spin-diffusion, especially for slow, isotropic motion of molecules, for which cross-relaxation is an energy-conserving process, so that the magnetisation can be propagated via intermediate spins. The observed intensity of an NOE can thus be entirely due to, or altered by, a serial transfer of energy via an intermediate nucleus, rather than to a genuine short internuclear distance. In the steady state experiment, spin diffusion can lead to the loss of signal for the entire molecule (NOE= -1). NOEs arising from spin diffusion can be readily identified from buildup curves since they show a lag-period of up to several hundred milliseconds, corresponding to the time needed to build up the previous NOE transfer in the series. If the initial buildup rate is determined then, since this is directly related to the distance between closely related spins, the effects of spin diffusion can be eliminated (Noggle & Schirmer, 1971). Spin diffusion limits the mixing time that may be used for NOE type experiments since many of the NOEs observed at long mixing times will be due to spin diffusion, rather than to direct NOEs with slow buildup rates (long distance).

Where more than two spins are involved in the relaxation process the same principles apply, but since relaxation is a competitive process, the observed NOEs measure the relative contribution of cross-relaxation from each particular spin (the nearer nucleus will be the dominant relaxation pathway). This means that when distances, are measured only relative distances are available. Strictly speaking for two pairs of protons at distances ra and rb in a rigid structure then,

σa/σb = r6b/r6a.

This expression is only valid when the relative motions of the nuclei are identical, which is unlikely to be the case in small peptides. The function describing the cross relaxation rate actually depends on the motion of the vector connecting the nuclei in space, and hence will be affected by overall motion of the molecule and the independent segmental motions within the molecule. Slow internal motions within the molecule actually extend the upper limit of the range over which NOEs can be observed to at least 5 Å by changing the characteristics of the function describing cross relaxation rates (Braun et al., 1981), whereas high frequency intramolecular motions contribute to the effective correlation time of the nuclei involved, so that for a molecule rotating in negative-NOE range, the effects of these motions are to reduce the intensity of the NOE (Braun et al., 1981). Hence it is possible that an NOE may be quenched by molecular motions. It should be noted that dipole-dipole coupling is not the only contribution to the T1 relaxation of the spins involved in an NOE, and so the NOE can be quenched by the presence of other relaxation pathways such as a close paramagnetic centre. It is alway dangerous to interpret the absence of an NOE in terms of a minimum separation between nuclei.

In this study the transient NOE was never used because the steady state experiment is inherently more sensitive. In addition the time after inversion that gives the maximum response in the transient experiment (governed by the cross relaxation rate and the T1) is difficult to predict, so that the experiment is also more difficult to use. The data of 1-dimensional experiments may be falsified by the limited selectivity of the initial pulses in crowded regions of the spectrum. This is not a problem in the NOESY experiment.

In the NOESY (Macura & Ernst, 1980) experiment all pairs of protons which are close in space are detected simultaneously. The familiar pulse scheme is seen in figure 2.3.7. It is equivalent to the selective transient NOE applied simultaneously at all frequencies. The first 90° pulse puts the magnetisation in the xy-plane. During the subsequent evolution time (which is varied, F1) the spins precess in the rotating frame according to their chemical shift. The second 90° pulse causes the vector component of the magnetisation which has not precessed to align with the -z axis, equivalent to an overall 180° pulse. Hence spin magnetisation at the beginning of the mixing time is aligned parallel to the external static magnetic field. The NOEs are built up during the mixing time which separates the second and third 90° pulses. The relevant component is detected by the final 90° pulse.

In the NOESY experiment, the system is not continually being pumped up by irriadation, as in the steady-state NOE, and hence the buildup curve seen for NOESY shows a decay at long mixing times. Cross-relaxation (spatial proximity) is indicated in the NOESY spectrum by a crosspeak. Crosspeaks are negative in phase sensitive spectra when they are the same sign as the diagonal. As in the 1-dimensional experiment, the rate of buildup of cross-relaxation contains the distance information.

Unknown interatomic distances were determined relative to the size of a crosspeak from a fixed distance transfer. To account for crosspeak intensity arising from spin diffusion, analysis of NOESY buildup curves in the time range 50-500 msec was used in this study, the aim being to determine the initial rate of buildup of the NOE (50 msec was generally the shortest mixing time to give a decent signal to noise level). In practice much of the analysis was somewhat subjective and based on interpreted observations rather than detailed buildup curves for each crosspeak in the spectrum. Hence the new crosspeaks arising as a result of obvious spin diffusion pathways at longer mixing times were eliminated from the analysis. The phase sensitive NOESY provides a good method to demonstrate the potential for problems arising from internal segmental motions within small peptides. In general the peptide is assumed to be a rigid body undergoing isotropic motion. In practice this was observed not to be the case, since it was often observed that the sidechains of the peptides had positive crosspeaks, whereas the backbone had negative crosspeaks. The NOESY was hence sensitive to the independent motion of the sidechains, over that of the backbone to which they are attached.

The addition of small random variations to the mixing time (±10%; Macura et al., 1981) in order to eliminate zero quantum coherence transfer was not implemented on our AMX500, although of course double quantum coherence was eliminated by phase cycling. Towards the end of the study, water suppression was improved by using a modified NOESY pulse sequence where the final 90° pulse was replaced by a jump-return sequence (Guillet et al., 1984) consisting of two 90° pulses, with the phase of the second pulse 180° from that of the first, the two pulses seperated by a delay period. The first pulse brings all the spins along x, during the waiting period they fan out in the xy plane, but at O1 the solvent resonances remain along x. The second pulse returns spins from xy to zy, and the solvent to equilibrium. A fifth order baseline correction was applied to the FID before transformation which also had the effect of reducing the apparent water signal.

The NOESY experiment described above can be extended to observe the structure of small ligands bound to proteins. NOEs which develop between protons in the bound species are transferred by chemical exchange to the free species for easy observation. The bound structures of many small molecules have been determined by this technique, including cofactors bound to enzymes (eg. 1d: NADPH bound to fatty acid synthetase, Leanz et al., 1986; nucleotide bound to aspartate transcarbamylase, Bannerjee et al., 1985) and a hexapeptide substrate bound to elastase (2d: Meyer et al., 1988).

The theory of time dependent transferred NOEs (TrNOEs) has been described by Clore and Gronenborn (1982,1983) and by Campbell and Sykes (1991). The method allows the distances (cross relaxation rates) between protons in a bound ligand, or a ligand and the protein to which it is bound to be determined. The linewidth of the bound ligand will be very large, because of the restriction of its rotation by association with the protein, making it too broad for direct observation. However advantage is taken of chemical exchange between the bound and the free ligand, to transfer the information into the easily observed free ligand (usually the averaged resonance, see section 2.3.8), in which large negative NOEs will be observed (positive, zero or very small negative NOEs would otherwise be expected). Because the cross-relaxation rates in the ligand-protein complex are directly proportional to the correlation time of the complex, then the sensitivity of detection increases with the molecular weight of the protein, and so very small molar ratios of protein compared to ligand may be used. Just as in a regular NOE experiment, the initial rate of buildup is a good measure of direct cross relaxation, and the ratios of cross relaxation rates provide a measure of relative distances within the complex (the relationship r6 still applies when it is calibrated using a known distance within the exchanging system). The cross relaxation rate observed is a weighted average (see section 2.3.8) of the cross relaxation rates in the free and bound ligand, but the falsification of the measured distances by the contribution of the free ligand is usually small because of the reduced mixing time used (and is reduced as the size of the protein template is increased). Spin diffusion can obviously be a great problem, so the mixing time used is reduced significantly as compared to that appropriate to the free peptide. The effects of mixing time and the fraction of ligand bound are additive: a larger fraction ligand bound means that the peptide spends a larger proportion of the time in the bound state where cross relaxation is more effective, and a longer mixing time increases the opportunity for cross relaxation to take place. Hence, increases in mole fraction bound, in size of the protein and mixing time, increase the effects of spin diffusion.

In designing a TrNOE experiment the inter-related parameters which may be varied are, mixing time, mole fraction of protein added, mole fraction of ligand bound, and the binding constant and the exchange rate for the ligand which are accessible via manipulations of temperature and ionic strength. Campbell and Sykes (1991) have investigated TrNOESY and have shown that the following practical considerations apply. The optimum mole fraction of bound ligand is below 5%, and falls to less than 1% if the ligand is small and the protein is large. The mixing time used should be less than around 100msec in most cases, since for large proteins, the intensity of the TrNOE falls off rapidly above 50-100 msec, and the TrNOE buildup rate is increasingly non-linear with cross relaxation rate above 20 msec, especially for short interproton distances. The contribution to the NOE from the free peptide conformation was shown to be negligible even when very small mole fractions (2%) of quite a large peptide ligand (25 residues) was bound. An experiment should be designed to trade off positive and negative aspects of the variables, so that a mixing time of reasonable length may be used to give an acceptable signal to noise.

Most cases where TrNOEs have been applied (see above) use a 1-dimensional procedure involving a presteady state NOE experiments (presaturation time about 0.5 sec) in which a series of spectra are recorded with the presaturation frequency moved at 10-20 Hz intervals through the region of interest. The NOE is analysed by plotting the intensity of a particular peak as a function of the irradiation frequency to yield an 'action spectrum'. This special method is required to detect where overlap of peaks and spin diffusion is causing a decrease in selectivity of the experiment. With the advent of the NOESY-type experiment no such special technical precautions are required.

The observation of NOE effects using the conventional NOESY often fails for molecules of intermediate size (small peptides) because the tumbling rate is close to that at which the maximum possible NOE passes through zero. Using the rotating frame ROESY experiment this problem is overcome.

The ROESY experiment, like NOESY, measures cross-relaxation rates. The pulse scheme is seen in figure 2.3.8. and the similarity to the spin-locked TOCSY experiment should be noted. A strong on-resonance rf (B1) is applied perpendicular to the external static field (B0) and the spin magnetisation becomes spin-locked parallel to this rf. Cross-relaxation occurs in this rotating frame (as opposed to the laboratory frame for NOESY), that is, in the xy-plane of the standard pulse representation. Since cross-relaxation occurs in the rotating frame, the effective field in which the cross-relaxation process occurs is B1 (this field is much smaller than B0). The relationship between Tc and the magnitude of the NOE (figure 2.3.6.c) is shifted leftward as magnetic field strength is reduced, so that at the very small effective field B1, it is readily appreciated that the cross-relaxation rate in the rotating frame is always positive (opposite sign to the diagonal) independent of the molecular rotation, and for medium size molecules where σn≈ 0 the sensitivity of the ROESY experiment is much greater than NOESY.

In principal interatomic distances can be measured precisely by the ROESY experiment, since they are related by the familiar r6 relationship. As with the NOESY experiment, potential problems arise since the ROESY cross relaxation rate (σr) is affected by the non-isotropic motions of a flexible peptide. Unfortunately, the σr is sensitive to internal and anisotropic motions, for which overall tumbling times are close to the Lamour frequency w0 (ie. for medium-sized molecules). This means that for molecules in the molecular weight range where the ROESY experiment is most useful, internal motions can change the size of σr severalfold (while still maintaining the dependence on r6) and the interatomic distances measured may be very different from those expected. Hence, quantitative use of interatomic distances based on these experiments in medium-sized molecules may be unwise. [It is possible to make allowances for these motions in calculations of interatomic distances, for example by measuring the σr/σn ratio, taking advantage of the non-equivalence of dependence of these rates on the motions, and by measurement of σr at several magnetic field strengths, but methods are complicated and were not used in this study. The interpretation of crosspeak intensity is further complicated by the offset of the resonance from the carrier frequency, since the amount of magnetisation spin-locked is reduced further away from the carrier.]

One of the principal problems of the NOESY experiment is spin diffusion. In the ROESY experiment spin diffusion is attenuatated, because the route of cross-relaxation is not an energy conserving process. If spin diffusion does occur in ROESY it can usually be easily identified, since crosspeaks arising by spin diffusion are negative [strictly the sign of the crosspeak is modulated according to (-1)n+1 where n is the number of steps involved in the spin diffusion pathway, but since such transfer is attenuated it is likely that only a two step process will occur, so in the pathway A_B_C, the AB crosspeak will be positive and the AC crosspeak negative]. Hence all positive cross-peaks in a ROESY experiment are likely to arise from a one step magnetisation transfer. As in NOESY, spin diffusion can be eliminated if short mixing times are used (eg. 50 msec), although the signal to noise will also be reduced. In this study no problems with spin diffusion in ROESY were observed.

The only other common artifact in the ROESY experiment arises from coherent Hartmann-Hann type transfer during the spin locking period, and can also be readily identified since it gives rise to negative crosspeaks. However, multistep processes which may contain contributions from more than one transfer mechanism and which may be common during the long mixing times used for ROESY experiments can cause problems. As mentioned above, the sign of the crosspeak is a product of the individual steps, and hence a crosspeak arising from a Hartmann-Hahn transfer followed by cross-relaxation would be positive and so indistinguishable from a one step cross-relaxation process. In this study HOHAHA-type peaks were seen in early spectra, especially where the carrier was midway between the two resonances contributing to the crosspeak, but after after some investigation it was found that this was not a problem if the spin-locking power was substantially reduced below that used for the TOCSY experiment. In practise the high power output of the ecoupler on the AMX500 can only be attenuated by the software by 26 dB (usually corresponding to a field strength of approximately 3.2 kHz), at which spurious peaks were never observed.

As with NOESY, incoherent processes other than cross-relaxation can give rise to magnetisation transfer. Chemical exchange gives rise to negative crosspeaks (with the diagonal negative). It was noted that where NOESY crosspeaks could be obtained with mixing times of intermediate length (200-400msec), that the sensitivity of the ROESY experiment was lower than that of NOESY. Hence NOESY remains the method of choice in all but the most unfavorable cases.

Chemical shifts, coupling constants and relaxation rates are all subject to averaging due to time dependent processes. Consider for example, a proton which has two resolvable chemical shifts, each corresponding to a different environment. Then exchange between the environments can be classified into three time dependent groups. For slow exchange, the exchange rate between the two environments is much less than the difference between the shifts of the resonances in hertz, for fast exchange the rate is much greater than the difference, and for intermediate exchange they are of similar magnitude. The exchange regime between the two environments can be judged by the appearance of the spectrum. For fast exchange a single resonance line is observed, with a chemical shift that is the weighted average of those for the individual species.

Shiftobs = (p) shiftA + (1-p) shiftB.

where p is the fraction of the population A.

For slow exchange, separate resonances are seen at the chemical shift of each individual species. The situation for intermediate exchange is more complicated. As the exchange rate increases from slow exchange, the resonances move together and broaden, and then come to overlap at the position of the weighted average, at which point they are very broad (intermediate exchange). A further increase in the exchange rate causes the single peak to sharpen.

In intermediate exchange, the rate of the process contributes to the relaxation of the nucleus, since as a nucleus replaces another in a specific environment, it senses the magnetic environment of its predecessor, and is relaxed by it. In slow exchange, the relative populations and hence an equilibrium constant may be determined, by integration of the two resonances, and the upper limit of the exchange rate is given by the frequency difference. In fast exchange, no such parameters can be determined. Similar concepts apply to coupling constants and relaxation rates, and so these parameters are sensitive to different time scales. Each process is also sensitive to the spectrometer frequency.

Fast exchange offers an ability to observe binding between macromolecules and ligands. NMR lines of the ligand are normally narrow and those of the bound ligand are likely to be broader by several orders of magnitude, because of the decrease in segmental mobility of the peptide in the bound state. Hence, the broadening of the average resonance observed for a proton between the bound and free states will be appreciable, even if only a few percent of the ligand exists in the bound state. The linewidth is essentially reporting the T2 relaxation time of the ligand in bound and free states. It is this method which is used to assess the binding of peptides to protein templates throughout this thesis (width at half height is proportional to 1/T2, where T2 is the half life for spin-spin relaxation of the excited nucleus). In general, if a peptide has a Kd of 1 mM or less for a protein, resonances of the peptide will be broadened by the interaction of the peptide with its protein template (Levine et al., 1990).

The NOE technique for determination of spatial relationships has been described above, but similar information may also be obtained by measurement of the proximity of a given proton, for example in a peptide, to a nitroxyl group attached to a specific site on a protein. The magnetic dipole moment of the unpaired electron of a nitroxyl group is about 1800 times larger than that associated with a proton. Paramagnetic perturbations of nearby protons result from strong through space dipole-dipole interactions with this unpaired electron. In the same way as the NOE, the strength of the interaction is distance dependent and falls off at a rate which is proportional to 1/r6 (r is the distance between the nitroxyl group and the nucleus of interest). In practise, any nucleus within 20Å from the unpaired electron will be relaxed by the dipole interaction, and so the corresponding proton signal will be broadened in a distance dependant manner. If the correlation times and the association constants of all the components under study are known, the exact distance separating a nitroxyl group and a broadened nucleus can be calculated (Wien et al., 1972). In the absence of such parameters, and with the likelyhood of overlapping poorly resolved spectra in 1-dimensional experiments, precise distance measurements are unusual, but qualitative measurements of the relative distances of protons from a nitroxyl spin-label are quite feasible.

Assignments were based on recognition of spin coupling patterns in TOCSY experiments and the sequential assignment strategy of Wuthrich (Wuthrich et al., 1984) using NOESY experiments in H2O where short sequential distances like dαN(i,i+1), dßN(i,i+1), dNN(i,i+1) are identified and used to assign adjacent regions in the primary sequence. The D-methylene protons of Pro were used in place of the amide proton.

Chemical shifts were measured from 1-dimensional, HOHAHA, COSY and DFQ-COSY spectra. When measuring shifts in 2-D spectra, where possible the F2-dimension was used because of the higher digital resolution in this dimension. No correction was made for the temperature dependence of the DSS standard (0 ppm).

Much work has been published on the correlation between proton chemical shifts and secondary structure, and some attention was paid to this in analysis of peptide structures (eg. Pastore and Saudek 1990). Because of the dependence of amide chemical shifts on small changes in pH of the solution, analysis was generally restricted to alpha-proton chemical shifts. For regular α-helix, alpha-protons tend to shift upfield (typically 0.4 ppm), while for ß-sheet alpha-protons move downfield (+0.4 ppm), as compared to random coil values. Ring current shifts were also analysed, to extract structural information. An analysis program was supplied by colleagues at Mill Hill, MRC.

The peptide backbone and sidechain has several possible rotations about the bonds which may be characterised to aid in the structural determination. The latter allows determination of rotamer populations of sidechains, and thus constraints may be applied to stereospecifically assigned ß-protons, to increase the precision of calculated structures. The backbone rotation about the peptide bond is restricted to 180° (trans) and 0° (cis) by its double bond character. The backbone angle phi (φ; rotation about N-Cα) is related to the coupling constant 3JHNα. The angle psi (rotation about Cα-CO bond) cannot be derived from proton coupling constants, but can be estimated from the intensity of the Hαi-HNi+1 NOEs [Otter et al., 1988; for a trans peptide bond, the Hαi-HNi+1 NOE is maximal if the Cα-Hαi bond is approximately trans to the carbonyl bond(i+1), so that the allowed range is centered at +120°, while weak NOEs suggest a cis relationship between these bonds centered at -60°]. The use of spin-spin coupling constants for studies of molecular conformation relies upon application of the Karplus equation (Karplus, 1959; figure 2.4.1) which describes the relationship between the size of the three bond spin-spin coupling constant and the intervening torsion angle. The solution of this equation produces at least two torsion angles, sometimes four. However the dihedral angles phi for all amino acid residues except glycine are concentrated in the range -30° to -180° (Richardson, 1981) which allows some solutions to be discarded.

The 3JHNα coupling constants were measured as the doublet splittings of amide peaks in 1-dimensional spectra in cases where where the amide peaks were resolved (generally 64K points over 11 ppm). These values were converted to phi dihedral angles and used as input for the simulated annealing protocol in XPLOR (usual applied precision ± 60°). This procedure was only meaningful for identification of secondary structure if value of 3JHNα was greater than 8 Hz or less than 5 Hz. The value of the coupling constant indicates the average conformation about the backbone torsion and hence is dominated by the series random conformations adopted by most peptides in solution. Hence a 3JHNα value only a few Hz below the random coil value (7 Hz) could well indicate a small proportion of helix among the population of peptide conformations (eg. Motta et al., 1991; the 3JHNα value may fall to 3 Hz in an alpha helical polypeptide segment which populates this conformation almost exclusively). In general a series of 3JHNα coupling constants of less than 5-6 Hz is seen only in regions of α-helix, but isolated individual, or pairs, of such coupling constants are found in other regions outside α-helices. It should be noted that in cases where the termination or initiation of a helix is uncertain then the magnitude of the 3JHNα values may allow these positions to be tied down.

Coupling constants were measured from the splitting between antiphase doublet components of crosspeaks of phase sensitive COSY or DQF-COSY spectra acquired with a high digital resolution in F2 (typically 8 K x 0.5 K points). The amide to alpha proton coupling was measured from the fingerprint connectivities between these protons by examining rows from the 2-D spectrum (the amide frequency manifested in F2). When measuring the coupling constant in this manner, the apparent coupling constant is likely to be only an estimate of the true value (upper limit). When antiphase peaks come close together their mutual cancellation causes the apparent separation to be increased relative to the true separation, placing a lower limit on the separation which can be observed to 0.576 times the line-width (Neuhaus et.al., 1985). Below this value no apparent reduction in separation will be seen, but there will be a reduction in peak height (possibly to zero for very small coupling constants in alpha helix) as the antiphase components come closer together. [Although not used here, the problem may be avoided by methods such as PE-COSY, in which only one of the anitphase partners is present. No allowances were made for the effects of the window function (eg. the simulation technique of Redfield & Dobson, 1990).]

Sidechain rotations around the Cα-Cί bond are related by a Karplus type equation to the coupling constant 3Jαί. Only approximate values of the 3Jαί coupling constants were determined to aid with the assignment of beta-protons. Splitting patterns of the α-ß crosspeaks from phase-sensitive COSY were analysed according to principles set out by Neuhaus et al. (1985; usually crosspeaks from above the diagonal were used, where the small Hα multiplet shows less internal cancellation). These sterospecific assignments were used to provide dihedral angle constraints as input for the simulated annealing protocol of XPLOR, and to provide explicit NOE constraints for each ß-methylene proton.

The analysis of chi1 is only meaningful if the sidechain is locked in a unique conformation. The energetically most favoured conformations are those with the atoms around this bond in a staggered conformation. The ß-protons are termed gauche or trans with respect to the α-proton. There are three rotamer positions g2g3(chi1=60°; Hß2=g-, Hß3=g+), g2t3(chi1=180°; Hß2=g+, Hß3=t), t2g3(chi1=-60°; Hß2=t, Hß3=g-), with the prochiral proton positions described as Hß2 and Hß3 as defined according to the IUB-IUPAC conventions (1970). The ideal values for the coupling constants are 3Jt=12.9Hz and 3Jg=3.4Hz (Bundi & Wuthrich, 1979).

The rotomer populations were identified by a combination of coupling constant and NOE data. If both Jαί coupling constants are smaller than 5 Hz then the predominant sidechain conformer has chi1 near 60°. One strong/medium and one weak intraresidue NH-CHß NOE identifies CHß3 and CHß2 respectively (this rotamer state is the least populated for all residues except serine, probably because it is the only one in which the sidechain is directed away from the backbone and into the solvent; Bundi & Wuthrich, 1979). The other two conformers with chi1 near 180° or -60°, both have one of the two coupling constants larger than 10 Hz and the second smaller than 5 Hz. These cannot be distinguished prior to stereospecific assignment. In cases where stereospecific assignment cannot be achieved, it is often possible to confirm or exclude chi1=60° on the basis of these coupling constants alone. Strong intraresidue NH-CHß and NH-CHß' NOEs indicate chi1 near 180°(g2t3), so that CHß=CHß2 and CHß'=CHß3. Otherwise, a strong NH-CHß' NOE and a weak NH-CHß NOE identifies CHß2 and CHß3 respectively, with chi1 near -60°(t2g3). For residues with only a single ß-proton (Val, Thr, Ile) the coupling constants alone allow the the trans position and two gauche positions of this ß proton to be distinguished (eg. if 3Jαί > 10Hz for valine then the predominant rotamer about the Cα-Cß bond is g+). Thr and Ile are equivalent to ß2 and Val to ß3 (Kraulis et.al., 1989). Where non-degenerate ß-methylene protons have both 3Jαί coupling constants near 6-7Hz, this is indicative of motional averaging resulting from conformational heterogeneity about the Cα-Cß bond (a mixture of rotamer populations) and no stereospecific assignment were made.

Interproton distances were be measured from NOESY experiments. The size of an NOE was estimated by counting contour levels (exponentially spaced) for each plotted peak. The upper allowed ranges of various geometrically related protons were used to calibrate in a semi-quantative manner the distance range 'observed' by the experiment (eg. distances between protons on aromatic rings). The tabulated NOEs were then classed according to size (number of contours counted) and hence related to an upper distance limit for the interaction. Four classifications of NOE were made for structural determinations:

very weak 6.0Å weak 5.0Å

medium 3.3Å strong 2.7Å

(no lower limit for the NOE was specified).

Pseudo-atom parameters were used for corrections of inter-proton distances where stereospecific assignments could not be made as follows: methylene (-CH2-, 1.0Å), methyl (-CH3, 1.0Å), methyls as in Val & Leu (-(CH3)2, 2.4Å), aromatic ring (D* and E*, 2.0Å)

(Wagner et al., 1987). Two restraints were used to define a backbone hydrogen (NHi-COj; rN-O restricted to 2.8-3.3Å and rNH-O to 1.8-2.3Å).

Backbone and sidechain amide protons in peptides are labile. The exchange of labile protons is either acid (H3O+, and H2O) or base (OH- and H2O) catalysed in aqueous solution (other routes are also available if buffers are present). The process involves the formation of a hydrogen bonded complex with the catalyst and then a release of the amide proton associated with the catalyst. The amide exchange rate is a minimum at about pH 4 and increases rapidly at higher pH's (the rate is proportional to [OH-] and so exhibits a 10 fold increase for each pH unit rise). For this reason most NMR on peptides in protonated aqueous solutions is done at around pH 4, so that the signals from the amide protons will be as strong as possible. Under other conditions the resonance intensity is reduced and the linewidth increased by exchange processes. All spectra in this study were collected at around neutral pH, because of the limitations imposed by addition of proteins to the peptide solutions. Since most of the structural information about peptides involves the amide resonances, it was important to try to maximise intensity and sharpness of the peaks. In some peptides the amide protons were afforded protection from the solvent, so that even at high pH's (pH>7) and high temperatures (>35°) they gave very strong amide resonances. This was intrepreted in terms of the structures adopted by the peptides, since hydrogen bonds and steric exclusion will both provide solvent protection (see section 3.2). However, in most cases a suitable precaution was reduction of the temperature of the solution (eg. 4-8°C) in order to retard the exchange of amide protons. As an added bonus, this aided the observation of transient structures adopted by the peptide, because of the reduced mobility of the peptide and the increased viscosity of the water at reduced temperatures. Both these factors served to change the correlation times of groups within the peptide into ranges in which interactions could be observed (over the temperature range 25°-4°C the Tc is expected to increase by a factor of 0.53).

The NOEs for a flexible peptide are a population-weighted average of the conformational states available to the peptide. Depending on the backbone conformation adopted by the peptide, certain distances are short enough to give observable NOEs between residues. For example, dNN(i,i+1) are only seen when the backbone dihedral angles lie in the helical region of (phi,psi) space. Other NOEs such as sequential dαN(i,i+1) connectivities are used only for assignment purposes in peptides, since they arise primarily from the substantial population extended chain conformations of the peptide in the unfolded state. The type of NOEs which indicate structure in peptides and the typical inter-proton distance ranges associated with these conformations are seen in figure 2.4.2.

Both NOE volumes in NOESY spectra, and proton coupling constants are readily interpreted in terms of the structures of peptides in solution. Interproton distances and dihedral angles were estimated from these parameters and used as input for a modified multistage simulated annealing protocol in the XPLOR program (Nilges et al., 1988; BrÜnger, 1988). Distance and dihedral constraints were applied as forces tending to oppose the violation of the constraints during dynamics calculations. Each constraint was allocated a target value, and when satisfied no further force was applied.

Briefly, the sequence of the peptide was used to generate a topology file defining the atoms and their covalent connectivities, with the backbone dihedral angles (phi & psi) set to random values. The calculation was begun by a brief period of energy minimisation to ensure correct residue geometry, followed by 15psec of dynamics at 1000 K. The properties of the peptide were modified in the calculation so that a large time step could be used (the planarity and chirality force constants were reduced allowing a 2psec timestep under the SHAKE algorithm), and the trajectories of atoms were allowed to pass through each other during evolution under the applied constraints (repulsive van der Waals energy term was very low). The NMR constraints were introduced as a gentle force acting on the peptide, so that when approximately satisfied no further force was applied. Even quite large deviations from the target only introduced a relatively small energy penalty. Evolution under the applied constraints was slow, with the period of dynamics sufficiently long to massage the structure into a position in which the constraints were loosely satisfied. The high temperature ensured that the conformational space was adequately sampled.

During the next period of dynamics the force generated by violations of the constraints was increased, in a looped fashion every 1psec of dynamics, and the van der Waals force also incremented to give atoms of a more realistic size. By this process the peptide structure was coerced into fulfilling the applied constraints more closely.

In the final period of dynamics, the parameters for atoms and interactions were reset close to those which mimic real situations and the temperature of the dynamics calculation was reduced gradually to 300 K. The cooling process was slow enough to prevent the trapping of local minima. At the end of the calculation, a brief period of energy minimisation, with constraints turned off, was again used to ensure correct residue geometry.

The whole procedure was repeated about fifty times from fifty different random backbone starting conformations, so that all possible structures which might satisfy the applied constraints were generated. Some structures were discarded since they had a high total energy (after minimisation of structure in CHARMm under QUANTA) or a large residual violation of the NOE constraints. These presumably corresponded to false minima. The remaining structures were aligned by a least squares deviation fitting procedure, which served to highlight differences and similarities between the calculated structures. The quality of the derived structures was assessed on two grounds, the degree to which the structure fitted the experimental data (small violation of constraints) and the variablity of the calculated structures (average RMSD for the atomic coordinates of the overlaid structures). If all the structures were overlain satisfactorily, then sufficient parameters were present to determine a solution structure, but any regions which failed to align were assumed to be relatively mobile (inadequate structural constraints for precise definition of a conformation).

Peptides are mobile molecules, and NMR only detects a weighted average of the structures adopted. The above protocol assumes that all the NOE constraints are satisfied simultaneously, which may not necessarily be the case. Some difficulty was found in certain peptides in simulating this inherent mobility in the peptide conformation in the structures generated by XPLOR, with certain NOE constraints being mutually exclusive, and so violated alternately in different structures. In general, difficulties were avoided by seldom assigning any constraint to a short interproton distance (medium and weak only). Torda et al. (1990) have recently reported a time-averaged NOE distance constraint which is applied to the simulation in a way allows the constraint to be violated briefly by small amounts without any drastic energy penalty, as long as the constraint is satisfied on average over the time course of the dynamics simulation trajectory. This allows the mobility of the molecule to be modelled more faithfully, but was not available in the routines used in this study. Likewise, methods of accounting for falsification of NOE derived distances by spin diffusion (Boelens et al., 1989 [IRMA], Summers et al., 1990; the full relaxation matrix is calculated from a starting structure and structure then refined by an iterative comparison of experimental and calculated data), for changes in cross relaxation rates as a consequence of internal or local motion of residues (Koning et al., 1990; dynamics parameters are incorporated following long simulations water as predictors of local correlation times), were not available.

In globular proteins approximately one-quarter of all residues are found in an alpha helical conformation. The average protein helix is twelve residues in length and contains eight intrahelical amide to carboxyl backbone hydrogen bonds, lacking intrahelical partners for the first and last four groups. Average dihedral angles in the protein database are phi=-63.8°±6.6° psi=-41.0°±7.2° (Presta and Rose, 1988.) Unlike a ß-stand where the stabilisation of secondary structure is extrasegmental, the alpha helix forms stabilizing hydrogen bonds within the body of the structure and hence it not surprising that populations of helical structures can be detected in water for protein fragments spanning helical regions of native protein chains. Helices might also be expected to act as nucleation centres during the folding process of a protein. The importance of helices in the structure of proteins and in the mechanism of their action is well known.

Several peptides which exist in transiently formed alpha helical secondary structure occurring in aqueous and/or organic solution have been studied in recent years (eg. residues 47-58 in the C-terminus of BPTI which is 8-23% helical at 0°C, Goodman & Kim, 1989), although many studies of natural and synthetic peptides have shown that alpha helical regions, when excised from proteins, fail to form a stable helix (eg. Panijpan & Gratzer 1974, Wagman et al. 1980, Hammes & Schullery 1968). The theory of alpha-helix formation is relatively advanced and has been supplemented and tested by many experiments, although most theoretical studies are limited to a two state helix-coil transition, which is usually a poor approximation (Schellman, 1958). It is widely held that the empirical factors derived by Chou and Fasman (Pα, 1978), by Zimm and Bragg (1959) and by later workers (complex statistical surveys sensitive to position, eg. Presta and Rose, 1988) give a good indication of the relative helix forming capacity of various residues, but that they may be dominated in individual systems by other interactions (eg. salt-bridges, steric factors, helix-helix contacts, dipole moment). Despite much effort, no-one knows how to predict the native secondary or tertiary structure of a peptide or protein from its amino acid sequence, unless that sequence has significant homology with another protein of known structure, or unless a large number of protein sequences are known for single family of proteins.

Zimm and Bragg proposed a theoretical model to describe the helix to random coil transition in oligopeptides (Zimm & Bragg, 1959). Their two thermodynamic parameters, σ and s, describe the nucleation (difficult) and subsequent propagation (easier since it is cooperative) of the helix in terms of helix length (more helix formation in longer peptides ) and residue nature. Unfortunately, the measured factors (s≈1, except proline and possibly glycine, with all measured values <1.3, and σ≈10-4) indicate that all short peptides (less than 20 residues) should show little alpha-helix formation in water. This is patently incorrect. One of the problems is that most theoretical studies have been restricted to long chain homo-polymers, as opposed to short hetero-polymers, the latter representing most natural examples. For example, the values of the Zimm & Bragg constants have been estimated by the host-guest method in which the test amino-acid is placed in a random copolymer of hydroxybutyl- or hydroxypropyl-glutamine or lysine and glutamate, and the tendency to stabilise/destabilise the helix present quantitated by circular dichroism (Seuki et al., 1984). Bierzynski et al. (1982) gave a useful formalisation of the Zimm-Bragg relationship:

σsn-1

y=--------

(s-1)2

where n is the number of residues (<20), σ is the helix nucleation constant (≈10-4), s the average helix stability factor (<1.3) and y is the fraction of molecule in a helical, as opposed to a random coil, state. Using these values for a thirteen residue peptide, y <0.02 (often a thousand fold underestimate) and substitution shows that the s value would have to exceed 1.7 to obtain 30% helix (as seen in ribonuclease C peptide 1-13 at 273K, pH 5.0, 0.1M NaCl, Shoemaker, 1987). Short peptides containing helix-forming amino acids are hence predicted to fall just below the threshold of exhibiting measurable helical structure. Other methods of estimating the propagation factors have recently been used (the effect of the nature of the host peptide in host-guest studies has been questioned by Chakrabartty et al., 1991 and a template which is constrained to nucleate helix has been used by Kemp et al., 1991), and these contest the magnitude of s, but none shows the ability to adequately predict the extent of helix which is observed in peptides. Problems in the Zimm & Bragg formalisation arising from the simplification that the propagation factor for a residue is independent of its position have been addressed by Vasquez et al. (1987), who have produced an alternate set of position dependent propagation factors. They argue that previous measured values only take account of the intrinsic properties of each residue, whereas the long range electrostatic interactions that may arise from the type of residues which are near neighbours in the sequence have been averaged out in the host-guest copolymers. Further improvements may be made using alternative formalisations of the helix coil transition, eg. the Lifson-Roig theory (Lifson & Roig, 1963) together with such position dependent factors, thus providing a better representation of cooperativity and position dependent interactions in the helix (Chakrabaty et al., 1991; Strehlow et al., 1991).

The Chou and Fasman factor Pα (the relative frequency with which an amino acids occurs in a protein helix) is the most widely used indicator of the relative tendency of a sequence of amino acids to form a helical secondary structure (in general, if the mean Pα factor for a region of amino acids is greater than 1.05 then helix is indicated, but the Pß factor should also be taken into account). These workers correlated various residues to their presence in secondary structures as identified in X-ray crystal coordinates of proteins to identify nucleation regions for secondary structure and provided simple rules to predict how far the secondary structure would be propagated in either direction. Other similar methods have become available (eg. Robson-Garnier algorithm, Garnier et al., 1978, which applies information theory to extract the secondary structure information in a primary sequence), but in each case the secondary structure predictions are only weakly correlated with actual structures because a smoothing window over 3-20 residues was applied in the analyses, giving rise to a statistically solid, but relatively weak preference (the best accuracy of prediction from information contained in the local sequence is about 65%, Gibrat et al., 1991). There is a correlation between Pα and the helix propagation factor s, but this is not perfect (Creighton, 1984). Further extensions to examine the predilection of residues to occur outside the termini, inside the termini and within the body of a helix, broken down by individual positions or based on detailed searches of secondary structure pattern (Presta and Rose, 1988) have given better insights into the nature of helix initiation or termination. Most preferences, like the N-terminal capping by asparagine (see below), can be explained in structural terms, but some like the clustering of the hydrophobic residues leucine, phenylalanine and methionine towards the interior of a helix or the presence of tryptophan near the N-terminus (eg. Mertuka et al., 1991) are largely anecdotal. New procedures are under development in which correlations to local structural elements (helix, turn etc) are used in structural prediction routines, neural networks lending themselves particularly readily to this type of analysis (Cohen, 1991). It is interesting to note that the limit of accuracy of prediction does not rise significantly as the sophistication of the analysis increases.

In summary, the best use of any predictive method might be in the prediction of regions of relative instability in potentially alpha-helical peptides, because specific stabilising interactions are expected to be involved in the formation of most short helices in water.

The most studied structural factor which will affect the helicity of a peptide is the electrostatic interaction between charged peptide sidechains. The best known case is the "charged group effect" for the 2Glu- to 10Arg+ salt bridge in the ribonuclease C peptide (aka. the C helix). This has been identified by the pH dependence of the stability of the helix, which reflects the pKa's of both residues, and also by the systematic mutation of the charged residues within the peptide (Shoemaker et al., 1985, 1987). It is important to recognise that the interaction seen is also present in the native protein crystal structure of bovine pancreatic ribonuclease (Nelson & Halperin, 1985). Another possible interaction, that between 2Glu- to 12His+, showed similar characteristics of helix stabilisation, but upon further examination the effect at each residue was shown to be independent and additive, and hence arises from the dipole moment effect described later (Shoemaker, 1987). Zimm and Bragg parameters predict that the charged states for the residues involved in this interaction will destabilise the helix (eg. His+ and Glu- are helix breakers in Zimm-Bragg), whilst the opposite effects were actually observed in the peptide. This type of long range interaction is unlikely to be a general feature among other peptides and an (i,i+4) sidechain contact is more often significant (see statistical survey of Maxfield & Sheraga 1975). Examples are seen in the interaction between the rings of 12His+ and 8Tyr in the C helix (Shoemaker et al., 1990) or the Glu-i-Lys+i+4 [ i+4(E,K) ] salt bridges which stabilise the helical propensity of alanine based peptides (Marqusee & Baldwin 1987a). Although less effective, (i,i+3) salt bridges are also seen to stabilise helix in peptides. While all these interactions are usually salt bridges, they may also be present as singly charged hydrogen bonds at extremes of pH, that can be almost as strong as a salt bridge (eg. Glu-...Lys0; Marqusee and Baldwin, 1987). Naturally the effectiveness of these charged-based interactions will be reduced by the solvent exposure in short peptides.

A fundamental property of the helix is its dipole moment. A positive/negative charge of 0.48x│e│, in a low dielectric constant environment, is located at the centre of the Cα-N bond of the amino acid at the amino/carboxyl terminus of the helix (Wada & Nakamura 1981; Wada, 1976). In peptides which are exposed to an aqueous solvent the dipole moment effect is likely to be reduced. The origin of the dipole moment is the alignment of the peptide bond dipoles nearly parallel to the helix axis, together with a contribution by the hydrogen bonds in the alpha helix. (Hol, 1978; the peptide bond itself is polarized by the trans arrangement of the partially negatively charged carbonyl group and the partially positively charged amide group). The strength of the field associated with the helix only increases significantly up to a length of about 10Å (seven residues or two turns). The dipole moment of a helix has often been implicated in its stability. In proteins, stabilisation by the dipole moment is manifest in the propensity for negatively charged sidechains to cluster at the amino terminus, and positively charged sidechains at the carboxyl terminus, thus tending to neutralise the dipole (Hol, 1978). The presence of unblocked and hence charged termini in peptides will conversely have a destabilising effect by strengthening the dipole (Marqusee & Baldwin, 1987). The simple addition of an acetyl group to the N-terminus will stabilise a helix (Ho and Degrado, 1987; Shoemaker et al., 1987). It is destabilising to place negatively charged residues near the N-terminus or positively charged residues near the C-terminus of the structure (Shoemaker et al., 1985; Kim et al. 1982). The effect is not limited to the ends of the peptide, and so the helix stabilisation by Glu--Lys+ ion pairs separated by (i,i+4) is greater than Lys+-Glu- ion pairs (Marqusee & Baldwin, 1987) and the block copolymer Ala20Glu20Phe shows a greater helical tendency in the polyalanine region than Glu20Ala20Phe, tendency being reversed if glutamate is replaced by lysine (Ihara et al., 1982).

Steric factors have been implicated in the nature of helices. These have usually arisen as an explanation of the observation that a particular residue is found more often in a certain position. Presta and Rose (1988) put forward the hypothesis that one of the dominant factors in the formation of alpha helices was the presence of residues at either terminus, which possess sidechain groups suitable to hydrogen bond to backbone amide and carboxyl groups within the helix. Since it was the sidechains of these polar flanking residues which were involved in the hydrogen bonding, the regular helical hydrogen bonding pattern was terminated (or initiated). The most common residues at the N-cap position are Ser, Asn, Gly, Asp and Thr, in order of decreasing frequency (Richardson & Richardson, 1988). For example, for asparagine it is the sidechain carbonyl which can adopt the correct position with respect to a backbone amide, and in a similar fashion serine, aspartate and threonine can also take part in helix capping or initiation (glutamine, glutamate and tyrosine do not have the correct geometry). The ability of residues to act as N-caps, with their sidechains becoming involved in H-bonds to the backbone, is determined by the surrounding residues, which may sterically disallow the formation of the capping interaction. In cases where N-caps are seen the backbone angle psi falls into small ranges (Ser & Thr: 150-180°; Asn or Asp: 60-150°) and the ability of Asn and Asp to be involved is improved by the greater chain length and the additional freedom to find an interaction route that this affords (Bell et al., 1992). N-terminal capping of helix by glycine has been attributed to its ability to form a left-handed helical twist which can satisfy two successive backbone carbonyl groups, while turning the chain in a direction that prevents continuation of the helix. A similar twisted backbone conformation accounts for the frequent occurrence of glycine at the C-cap position. Within the body of the helix steric factors also play a role. Proline is very rare within a helix because of difficulty of accommodation of the ring and the disruption of at least two hydrogen bonds in the regular pattern by the lack of an amide proton. Barlow & Thornton (1988) have shown that while proline can be accommodated in helices, it generally causes a bend. The backbone conformation of proline is suited to the first turn of a helix where the backbone hydrogen bonds the amide of this residue are not required to be formed. The ring of proline fixes the backbone dihedral phi angle at the preferred alpha helical value of -60°, and leaves the psi angle flexible, but hindered. In initiation of a helix the entropy loss for proline by enforced ordering of the backbone is reduced by the presence of only one rotatable backbone angle (Strehlow et al., 1991; Mertuka et al., 1990). The use of unnatural amino acids in a recent study has indicated that helix stability is markedly reduced if branching of alkyl sidechains occurs at the ß-carbon, since this reduces the conformational space available to the sidechains. Those sidechains which have less restriction on sidechain torsion angles (especially at the ß-carbon) are better able to fit into restrictions of the compact alpha-helix. (eg. torsion angles found in the g- (+60°) Cα-Cß rotamer cause a clash between CGi and the carbonyli-3 groups in an alpha helical conformation.) For natural amino acids alanine and leucine are found to be helix stabilising and isoleucine and valine are weakly destabilising (Lyu, 1991; Padmanabhan et al., 1990). Small sidechains along the hydrophobic face have been shown to facilitate the compression of ampipathic alpha helices into a slightly curved structure. The driving force arises from bifurcated hydrogen bonding of carbonyl groups on the hydrophilic face, which form an additional interaction to either a water molecule or sidechain (Blundell et al., 1983; Baker & Hubbard, 1984).

A steric factor largely restricted to the behaviour of peptides is fraying of the helix towards the ends of peptides. This can be accounted for by the loss of restraints provided by the remainder of the intact polypeptide chain present in a protein. A residue at the end of a helical segment must also be more easily converted to the random coil state, since it lacks the interlocking hydrogen bonds which are found in the interior of the helix (Strehlow et al., 1991).

An important factor to be considered is the presence of tertiary and quaternary structure. Long range interactions are known to stabilise helix in both proteins and peptides (Lau et al., 1984), but most such interactions that are present in proteins cannot be modelled in peptides (often not sufficiently large for the structures to be stable) or accounted for in statistical searches. Some factors are however of interest. Workers have always attempted to show that peptides remained monomeric in the concentration range of interest and hence that the extent of helix formation was independent of concentration. Conditions are however common where this does not apply. The C helix peptide begins to aggregate at concentrations exceeding 4mM (Brown & Klee, 1971; 0.1M salt at room temperature where sedimentation equilibrium experiments indicated aggregates of 10-30 peptides) while at lower concentrations in deionised water (Brown & Klee 1971) helicity is promoted by low temperature aggregation (absent at room temperature). The C helix is studied at 0.1M ionic strength (reduced or room temperature), where helicity is promoted, but not by aggregation. Hence the conditions of the experiment may be crucial in determining the factors involved in stabilisation. Similar helix promotion by aggregation has been noted elsewhere, for example in peptides of the bomolitin series (Bairaktari et al., 1990b). Interestingly, as in this case, non-specific aggregation often fails to give rise to intermolecular cross-peaks in NOESY experiments (even when very long mixing times are used). Marqusee and Baldwin have designed alanine based peptides with glutamate and lysine residues spiralling around the surface of the helix in order to prevent amphipathic aggregation. It should be noted that although a knowledge of the aggregation state of peptides is important to any investigation, that aggregation could actually serve as a crude model of the state of the peptide in its native environment. In globular proteins the sidechains of residues in helices often alternate from hydrophobic to hydrophilic with a periodicity of three to four, giving the helices an amphipathic character (Richardson, 1981). Since helices are commonly located along the outside of proteins, they are stabilised by interactions of the polar groups on their outer face with the solvent, and by the packing of apolar groups on their inner face in the protein interior, often to a beta sheet underlying the helix.

In some circumstances specific interaction between peptides is an interesting property. Short polyheptapeptides (20-30 residues, Lau et al., 1984; monitored by molecular weight determinations and C.D. spectroscopy) have been shown to form very stable alpha helical coiled coils at low concentration in aqueous solution. These have repeating units, with a sequence similar to the parallel in-register coiled-coil structures seen in tropomyosin, consisting of a 3-7 hydrophobic repeat. The repeating sequence puts interacting residues on the same face of the helix so they are able to cluster together in the dimer (a periodicity is seen in the rates of hydrogen-deuterium exchange for amide protons, with the protons at the hydrophobic interface showing the most protection from exchange). The pair of alpha helices are wrapped around each other with a slight left handed twist, which is caused by a short helical H-bond followed by a long H-bond, every seven residues (Goodman & Kim, 1991; a corresponding periodicity of the chemical shifts of amide protons is seen). Smaller peptides in the series have some alpha helical character, but the longer peptides appear to be entirely helical and even fraying at the peptide ends is absent, stabilised by the strand to strand interaction. The principal stabilizing factor is the transfer of the alpha helix surface from an aqueous to a hydrophobic environment (the dimer interface) upon association. The geometrical requirements for intimate sidechain contacts have been investigated by Richmond and Richards (1978). Leucine is often the residue involved in hydrophobic contacts as seen in the leucine 'zipper' (leucine zipper sequences are known to be important for mediating dimerisation of many DNA binding proteins; Landschultz et al., 1988). In proteins, other interhelical contacts are also seen, for example the 4-alpha helical bundle (Sheridan et al. 1982). This motif usually shows an antiparallel arrangement which is stabilised by the helix dipole. The helices are of equal length and packed in an array of square crossection, with adjacent helix axes inclined at an angle and most closely approaching each other midway along their axes. Minimal peptide models of such structures have been constructed (Ho & Degrado, 1987) and used to form primitive frameworks for catalytic groups (Hahn et al., 1990).

The structures of alpha helical peptides is dependent on the temperature, ionic strength and dielectric constant of the solvent. All the common effects were observed in early studies on the C helix peptide (Brown and Klee, 1971). The helical state is favoured by low temperature, increasing ionic strength and mixtures of water with organic cosolvents such as trifluoroethanol, methanol, and ethylene glycol (the latter pair allowed temperatures to be reduced as low as -13°C). Maximal helix formation is seen at low temperature for most peptides and unfolding observed at higher temperatures. This indicates an enthalpy-driven process. Negative enthalpy changes upon hydrogen-bond formation and hydrophobic interactions (entropy driven at room temperature, but increasingly enthalpy driven at lower temperatures; Baldwin 1986) are probably common to all alpha helical peptides. The thermal transition temperature is a parameter often used to quantify helix formation (eg. Ihara et al., 1982). The screening effects of counterions at a high ionic strength on charged interactions has been shown (eg. Marqusee and Baldwin, 1987), but is often complicated by the driving of hydrophobic interactions under such conditions. Many peptides, for example the C helix, require 0.1M salt to optimize helix formation.

It is very common to investigate peptides in aqueous mixtures with trifluoroethanol. This is considered to be a non-interacting (inert) solvent (Lotan et al., 1972) which is known to stabilise helix and destabilise higher order structures such as coiled-coil (Lau et al., 1984 where it promotes monomeric single-stranded alpha helix). The mechanism of stabilisation of helix by TFE is unknown. Some peptides show a linear gain in helicity with increasing concentration of added TFE , whilst others show a cooperative rise in helicity (Nelson & Kallenbach, 1989) that may be very abrupt with the transition between non-helical and helical states occurring within a few percent of added TFE. Reed & Kinzel (1991) describe a peptide which undergoes a very abrupt transition to α-helix at 60%TFE, and a variety in which a Trp residue is replaced by Val that shows a very gradual increase in helicity. Cooperativity is usually reduced at higher temperatures. For each peptide a threshold value is reached at which further additions have no further effect, which may be as low as 10% or as high as 70% added TFE. It is known that the magnitude of the charged-group effect of the ribonuclease C-helix does not increase in TFE solutions (Nelson & Kallenbach, 1989) and so despite a dielectric constant of about one third that of water, enhanced electrostatic interactions are not the mechanism. In addition the helix 'stop signals' in the S helix peptide (at residue 12His ) do persist in TFE (Kim et al., 1984) and so helix is not promoted where there is not a prior tendency. Solvent may have a very dramatic effect on the structure adopted by a peptide. Certain of the peptides of Reed & Kinzel (1991), referred to above, exhibit a bistable structure forming a ß-sheet in dilute buffer, and α-helix in apolar solvent, the transition being sharp and reversible. Other cosolutes are used, for example SDS is used as a mimetic for biological membranes (eg. Bairaktari et al., 1990a).

The principle methods for investigation of alpha helical peptides have been circular dichroism, NMR, and recently molecular dynamics simulations. CD offers two advantages over NMR in that the relative and absolute amount of helix in the population of peptide can be quantified and the concentrations of peptide required are very low (typically 5-100uM). The disadvantage of CD is that the average content of helix over the whole peptide is monitored, so that localisation of the residues involved in the helix for a particular peptide is only available by NMR NOE-type experiments. Regular secondary structure, including α-helix, in peptides has been shown to exhibit characteristic amide absorption bands in FTIR (Kennedy et al., 1991).

The peptide Y847 is an FMoc-based synthetic peptide corresponding to residues 77-94 in the skeletal muscle actin sequence . The sequence of the peptide is :

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

NH2--T--N--W--D--D--M--E--K--I--W--H--H--T--F--Y--N--E--L--CONH2

77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94

The peptide was purified by HPLC and three rounds of purification of the peak collected was necessary to ensure better than 95% purity from peptide contaminants. It has low solubility (<250µM) in the range pH 3 - pH 7.5, but is soluble to more than 15mM at pH 7.8. The calculated isoelectric point is 4.6 and molecular mass 2377.

The calculated secondary structures according to several algorithms are seen in figure 3.1.1. The Robson-Garnier prediction, as shall be shown latter, is a very accurate representation of the structure of the peptide in water, but both prediction routines indicate that the majority of the residues would be expected to be in a helical conformation in a globular protein. The helix-coil initiation factors indicate that helix should begin at 3W, and a tendency towards helix termination is indicated by both initiation and propagation factors at 11H and 12H, for uncharged histidine, but this tendency is reversed for charged histidine+. Hence, helix stability might be dependent upon the ionisation state of these histidine residues. If helix is present in the C-terminal portion of the peptide then it might be expected to extend to residue 13T, 16N, or 18L, depending upon the predictor used. As will be seen later, the helix formation in the peptide is very stable, so algorithms based upon statistical predictors might be expected to be fairly accurate. It should be noted that since the crystal structure of actin was only recently published, actin was not included in the databases from which these various parameters were determined.

Some of the preferences of residues for certain positions in alpha helices, as investigated by Richardson and Richardson (1983) are relevant to peptide Y847. The residue 2N is expected to act as an N-terminal capping residue, particularly since 3W has a preference for the first position within the helix, and 4D the second. Residues such as 6M, 7E, 8K, etc. are common within the centre of the helix. Based on these rules, however, there is no obvious C-terminal capping residue in this peptide. The clustering of positively charged residues (84K and possibly 87H/88H) towards the middle of the helix and the negatively charged 93E at the C-terminus might be expected to destablise the helix somewhat over the later half, as will the presence of multiple large aromatic residues in the C-terminal half.

In the published X-ray crystal structure of actin (Kabsch et al., 1991) the region 79W-91Y is a surface exposed helix, and has several features common to right-handed alpha-helix: the helix is terminated by underwinding to give a short region of π-helix, the characteristic alpha helical hydrogen bonding pattern ( CO(i), NH(i+n) with n=+4 ) is seen, but non-regular bifurcated hydrogen bonds of the type (i,i+4) and (i,i+3) are also present for residues 85I-90F and a potential i,i+4 electrostatic interaction between 80D and 84K and may stabilize the alpha helix by formation of a salt bridge (figure 3.1.2). When the region extracted from the crystal coordinates is compared to the helical wheel projection (figure 3.1.2) it can be seen that the residues which are displaced from regular helical positions are 77T, 91Y, 92N, 93E and 94L. This has the effect of placing 94L and to some extent 91Y within the body of the hydrophobic face of the helix and removing 93E into the hydrophilic area. The fast motions, which are sampled later in the dynamics simulations, leave their trace in X-ray diffraction data as they cause atoms to occupy more space on average than in a rigid molecule, smearing out the electron density as determined by these studies. The isotropic temperature factor B, which is related to the square of the mean fluctuation of the atom, (Post et al., 1989) is indicated in figure 3.1.2a. No obvious flexible region is seen within the body of the helix, although these temperature factors rise for residues 92-94, which are in a 5-turn conformation.

The sequence of the helix 77-94 is somewhat unusual in that the are many very bulky residues accommodated within the body of the helix ( particularly 10W, 11H, 12H, 14F, 15Y). Examining the solved structures in the Brookhaven PDB database this situation is unique although there are several examples of helices with fewer bulky residues (eg. G helix 460-478 in acetlylcholine esterase, 1ACE). The geometry of the helix tends to impose constraints on the packing of sidechains so that bulky residues are expected to point largely to one end of the helix or to point out perpendicular from the helix axis. In the helix 77-94 the histidine residues (87 & 88) point towards the C-terminus while the remainder of the residues point towards the N-terminus. This is not surprising since it allows alignment of the sidechain dipole with the helix dipole, but does create a 'hole' in the sidechain packing with the sidechains changing direction between 86W and 87H.

In order to determine the structure of the excised peptide several 1-D and 2-D experiments were run under various conditions.

The spectrum of peptide Y847 is seen in fig.3.2.1. Peaks are annotated according to the assignments in 3.2.11.

In figure 3.2.2. the difference between the chemical shifts for the alpha proton resonances for random coil conformation and those recorded for peptide Y847 are shown. The chemical shift of HA resonances is sensitive to the dihedral angle between HA and the carbonyl (Pardi et al., 1983) such that, in the absence of other effects, a helical conformation leads to an upfield shift of HA resonances as compared to random coil (and a ß-sheet to a downfield shift). The pattern observed is consistent with the assertion that this peptide adopts an alpha helical secondary structure in solution. The mean changes in chemical shift are slightly larger in the N-terminal half of the peptide than the C-terminal half, indicating a larger tendency to helix in the N-terminal portion of the peptide. The differences are small and changes between individual residues are difficult to interpret. The large shift for residue 1T may reflect the sensitivity of this N-terminal residue to the small pH changes due to the heat of ionisation of the peptide over the temperature range used.

Spectra (as described in the figure 3.2.3 legend) were collected at various temperatures between 280K-308K in order to determine the temperature dependence pattern for the shifts of the backbone amide protons. The normal temperature dependent shift of an amide in a random coil model flexible peptide in water lies in the range -6 to -10 ppb/K (Jimenez et al., 1986). Any reduction in this coefficient is taken as indicating reduced solvent accessibility, usually reflecting the formation of a hydrogen bond, although a reduced temperature coefficient can be due to shielding from the solvent by a local structure (solvent exclusion). The amide chemical shifts of the peptide decrease linearly with temperature in the range 280K-308K. The amide temperature coefficients were determined by regression analysis. Several amide resonances for peptide Y847 have a reduced temperature dependence coefficient which can be interpreted in as indicating strong hydrogen bonding. These are seen in the N-terminal half of the peptide (4D-9I), disappearing at 10W and being partially and irregularly regained for residues 11T-18L. These indications of hydrogen bonding were used in the input for structural determination in XPLOR, defining likely hydrogen-bonds in the alpha-helical staggered fashion. The smallest coefficients can be taken as corresponding to a more populated alpha helix (Jimenez, 1988). Protection of the amide at an individual residue reflects the formation of a helical structure over a wide range of neighbouring residues. Hydrogen bonding is staggered over four residues (NHi-COi-4), and so the stability of the helix in this region, displaced from the residue in question, is important. In a helical conformation the sidechains may be quite tightly packed, and this provides some degree of solvent exclusion for the backbone amide protons. Hence the amide protection at a particular residue also be affected by local factors. The amide protons of the first four residues of a helix do not have intra-peptide hydrogen bonds, and hence should show little protection. Later NOE data suggests that the helix is propagated from residue 3W. If the data is interpreted in terms of the staggered hydrogen bonding, it suggests that the N-terminal 1T residue is possibly in helix. On balance, the lack of protection at 3W probably arises from the absence of local solvent exclusion, as a result of a largely extended random conformation in residues 1T & 2N, with helix extending from residue 3W. The very low amide shift coefficient for residue 4D remains unexplained. Since inter-molecular association will also result in solvent exclusion, the possibility that any aggregation in this peptide may begin from the N-terminus cannot be excluded. Based on other evidence (see later), the lack of protection at the amide of residue 10W probably reflects the breakdown of helical structure at residue 10W, rather than the extent of helix at residue 6M (to which it would be hydrogen bonded). It should be noted that even though the helical hydrogen bonding inferred from the amide protection at an individual residue is staggered, the best indication of helical content at the peptide applies locally to the particular residue involved.

Temperature dependence of sidechain chemical shifts arises from the temperature dependence of the folding of the peptide into an alpha-helical secondary structure, and so ring current shifts which would be expected to arise from the adoption of this structure should be reduced by increasing the temperature of the solution provided that this disrupts the alpha helix (Jimenez, 1988). Hence the aliphatic proton chemical shifts were also monitored (figure 3.2.4). The calculated ring current shifts of groups in a peptide of structure corresponding to that of residues 77-94 from the Actin-Dnase I crystal indicates that a particularly large ring current shift is expected as a result of the interaction of 6Met with 10Trp (up to +1.4ppm for the methyl of 6Met). A small downfield shift for the 6Met resonance with increasing temperature is observed, which is no different in magnitude to other methyl resonances in the peptide. This indicates that either this region of the peptide fails to adopt an ordered alpha helical structure or that no change in the population of helix occurs over the temperature range 280K-308K, the latter of which is unlikely.

In order to determine whether the hydrogen bonds in the backbone were very stable an attempt was made to observe slow exchange rates by a deuterium exchange experiment (Spios et al., 1991). The peptide Y847 was freeze-dried from 1H2O at pH 7.8 and then dissolved in ice cold D2O, and a spectrum taken within 4 minutes (data not shown). No protons were observed for the exchangeable amides, and hence no protection by hydrogen bonding was observed. Hence the exchangeable protons must be exposed to the solvent and are lost rapidly from the spectrum by replacement with 2H. However, at pH > 7.5 most peptides show no amide resonances because the exchange with the solvent is too rapid under these conditions. As the rate of exchange increases the amide proton signal disappears to zero intensity. As described previously depression of pH (eg. pH 4-6) and temperature reduces the exchange rates so that these amides can be observed, although usually at less than unit intensity. The presence of all (except the N-terminal charged amide, that of 2N, and 15Y) at high pH (7.8-8.1) is in itself an indicator of amide protection in some degree. It may be significant that the amide proton of residue 15Y is absent from the all spectra. The formation of helix appears to be only very weak in the C-terminal half, but the total absence of protection may arise from the breakup of the helical structure at residue 11H, with the loss of the helical hydrogen bond to the amide of 15Y. In general, it was noted that the amide resonances which have the largest integrated area also show the lowest temperature coefficient. However Marsden et al. (1988) have shown that the correlation between these data are not strong and care should be taken in interpreting exchange or temperature coefficients in terms of a single cause such as hydrogen bonding.

Any increase in stability of hydrogen bonds will affect the resonant frequency of the amide protons (Spios et al. 1991; note the effect of alpha helical conformation on amide shifts is opposite to that of hydrogen bonding, the former causing an upfield shift and the latter causing a downfield shift). In this case several residues are duplicated within the peptide providing an internal control, as illustrated below:

280K 4D 8.20 7E 8.32 3W 8.39 11H 8.12

280K 5D 8.05 17E 8.49 10W 8.22 12H 8.05

299K 4D 8.17 7E 8.25 3W 8.24 11H 7.99

299K 5D 8.00 17E 8.37 10W 8.06 12H 7.95

In each case an upfield shift as compared with the duplicated residue is expected to correspond to a residue which is in a more alpha helical environment. The helical character is thus expected to follow the pattern 5D > 4D, 7E > 17E, 10W > 3W, 12H > 11H.

In addition to the 1-d spectrum and CPA spectrum seen in figure 3.2.1, a TOCSY (mixing time 60msec, pH 7.8, 280 K), DFQ-COSY (D2O, pH 7.8, 280 K), and NOESY spectra with mixing times 50, 200, 400msec, (H2O, pH 7.8, 280 K) were collected (figure 3.2.7-10). The 400msec NOESY was used for assignment purposes only.

The assignment process was relatively straightforward and followed the pattern described latter in section 3.3. It was possible to identify 15 out of 18 possible observable amide proton resonances in the NH-CHα fingerprint region of the HOHAHA or COSY spectra in water. (1T, 2N, and 15Y amides were not seen). Only one set of resonances was seen consistent with a single population of structures among the peptide. When the peptide was dissolved to concentration of 15mM, pH 7.8 the peaks seen were very broad (figure 3.2.6). Upon dilution the spectrum sharpened considerably to that shown in figure 3.2.1. No further change in the spectrum was shown as the peptide was diluted. The chemical shifts of resonances (especially amide protons since these are the most sensitive to structural changes) were monitored in the range 6mM to 200uM, and no significant systematic changes were seen, although a slight variation of pH during the dilution (samples were made up from a stock solution rather than by serial dilution) did have some minor effects (less than 0.05 ppm changes; data not shown). The line widths and alpha proton shifts were followed in D2O solution to 50uM with no significant changes. This indicates that no change in the aggregation state of the peptide was observed over this concentration range. The assignment is shown in great detail in the NOESY (mixing time 200msec) in figure 3.2.8 and recorded in table 3.2.11. Stereospecific assignments of the beta protons of 18Leu, 8Lys, 6Met and 3Trp were made, based on the COSY and NOESY spectra, in a similar fashion to that described in section 3.3.

As discussed above, aggregation of the peptide was seen at high concentrations. In order to investigate the possible nature of this aggregation two spectra were collected (HOHAHA/60msec and NOSEY/500msec mixing time 280K/pH7.8/90%H2O-10%D2O/15mM). Some assignments are seen in figure 3.2.5. By comparison of the changes in chemical shift between the spectra collected at 280K at 5mM and 15mM it is seen that apart from the fact that in general the reduced concentration leads to a small downfield resonance shift (increased helical content is thus indicated) no pattern emerges for the nature of the shifts. This implies that at higher concentrations the peptide self-associates with some rearrangement of structure, otherwise a simple pattern should be seen for example the involvement of hydrophobic residues. Unfortunately no further clues are available about the structural changes since the NOESY spectrum shows a great deal of spin diffusion indicative of the formation of a high molecular weight aggregate. No inter-molecular NOEs were distinguished. It should be noted however that there are only a single set of resonances, indicative of a single conformation of the peptide (or of averaging of multiple conformers, fast on the NMR timescale). It is not surprising that an amphipathic alpha helix self-associates in aqueous solution by hydrophobic attraction. As described latter, aggregation is promoted by increased ionic strength (section 3.7) and reduced pH (section 3.6).

The region corresponding to the peptide in the crystal structure of actin shows definite hydrophobic and hydrophilic faces, corresponding to the amino acids which are hidden within the body of the protein and exposed to the solvent respectively. The structure determined for the peptide (see latter) however shows a spreading of the hydrophobic residues out over the surface of the peptide. The change in winding of the alpha helix in the free peptide thus destroys the potential for an amphipathic helical conformation in which aggregation would be more likely to occur. In the case of aggregation the rationale would be to 'hide' the hydrophobic residues from the aqueous solvent. The spreading out of these hydrophobic residues would be expected to have the same effect, by placing them under the umbrella of charged residues. The likely structural changes upon aggregation would be to reverse these changes back towards the amphipathic helix.

From the NOESY experiments (280K/pH7.8/90%H2O-10%D2O/5mM 50msec/200msec) distance constraints were determined by the presence of NOE peaks. The NOEs identified are listed in figure 3.2.12. In general intra-residue constraints are taken from the 50msec NOESY and inter-residue constraints from the 200msec NOESY. Spin diffusion was not considered to be a problem except in the 400msec experiment. These connectivities were used as input for the XPLOR simulated annealing routine. The ROESY experiment (250msec mixing time) was of poor quality, with only the largest crosspeaks being seen. The much better quality of the NOESY experiment can be understood from the alpha helical nature of the peptide. Y847 is a peptide of 18 residues, towards the upper end of the range where ROESY is expected to have as good signal to noise ratio as NOESY, and in addition the peptide was moving as a partially rigid helical structure that will have taken the effective correlation time far outside the favourable range for ROESY. Hence no constraints were determined from the ROESY experiment.

Several dNN connectivities are quite strong (6M-10W) indicating a relatively high population of helix around these residues. Some others are present, but are much weaker (4D-6M) and so a lower helical population can be inferred around these residue. None of the long range alpha helical NOEs are seen [dαN(i,i+2), dαN(i,i+3), dαί(i,i+3)], indicating that the helix formed here is of the nascent type described by Dyson et al. (1988). The intensity of the CαH-NH cross peaks is 3 to 5 fold more intense than NH-NH, indicating a considerable population of extended conformations in the peptide solution.

No coupling was evident in the backbone amide region of the 1-dimensional spectrum (the apparent coupling resolved was due to partial overlap of amide peaks) because the amide protons were exchanging rapidly and thus the resonances were broadened by the chemical exchange (or by motional vibrations). It was not possible to reduce the exchange rate by reducing the pH of the solution because of the reduced solubility of the peptide. No coupling pattern was seen in the fingerprint region of a phase-sensitive COSY in 90%H2O/10%D2O. Hence no 3JHNαH coupling constants were determined.

The distance and hydrogen-bonding constraints were used as input for the simulated annealing protocol of XPLOR as described previously. Fifty-six structures were generated from peptides with initially randomised phi and psi backbone angles. From these, the 39 structures with the lowest residual NOE violations were selected.

In the calculation of this peptide structure, as in other cases, no short range NOEs were entered into the XPLOR routine (instead they were entered as medium range NOEs). This was found necessary in order to model the mobile structure of a peptide in a program which uses rigid distance target constraints. In all the peptides in this study, if short range NOEs were introduced, then a number of small to medium sized violations were seen for each calculated structure. The violated constraints usually varied significantly between the calculated peptide structures, so that they did not represent mistaken assignments. The structures calculated in the presence of short range distance constraints were found to be 'rigid' and to align fairly well with each other, but this was not a result of additional information introduced by the short range NOEs, and instead arose because many of the NOEs were conflicting and could not be satisfied easily by a single conformation. In this situation the calculated structure was a mean structure, stretched rigidly between the multiple conflicting constraints. This is not too surprising in a peptide which must, because of its mobile nature, adopt many conformations each of which will be characterised by an individual subset of short 1H-1H distances. In the strategy adopted, conflicting NOEs do not violate each other. Instead, depending upon the random starting point, one NOE will dominate in one subset of calculated structures, and the second NOE will dominate in other structures. Hence the final conformation generated will have sampled the conformational space available to the peptide under its constraints and produced an ensemble of structures which may be aligned to form an envelope into which all the calculated conformations fall. It was observed that the peptide structure calculated in this way was a better representation of the data, but a less 'precise' structure.

In the absence of the hydrogen-bond distance constraints the peptide could not be constrained to any form of alpha helical structure over the latter half (residues 11H - 18L). Under these conditions the calculated peptide showed a degree of helical character in the N-terminal portion, but the Kabsch and Sander (1983) algorithm failed to assign the conformation of many residues. Closer inspection indicated that the best description of the conformations adopted was somewhere between an extended and a 4-turn structure. These structures aligned poorly at the backbone level and not at all when the sidechains were considered. Hence the calculated structures are not illustrated here. This was expected since the only constraints applied which were exclusively helical were from the NHi-NHi+1 connectivities, which are insufficient to define helix.

With the inclusion of the hydrogen bonding data, the peptide was helical over the corresponding region (N-terminal residues 3W - 9I). However, the structures calculated in this way showed other short range 1H-1H distances that were not observed in the experimental data, and hence the calculation was considered unreliable. Ordering of the sidechains in these calculated structures was still poor, reflecting a lack of distance constraints from the NOESY experiment. It is apparent that the hydrogen bonding data, obtained from the protection of the corresponding amide protons from the solvent, as indicated by the reduced temperature shift coefficients, is much more sensitive to the presence of helical structure than the observation of helical (principally i,i+3) NOEs. The observation of the NHi-NHi+1 NOEs is likewise relatively sensitive to the presence of helix, but these alone these provide constraints which are insufficient to define helical structure.

Some interesting intra-residues NOEs were observed for residues 9W and 10H. For 11H the observed NOE, 11HCHD2-11HNH, defined the orientation of the histidine ring and tended to suggest that the ring was orientated forward, towards the N-terminus of the peptide. A similar NOE was absent for residue 12H excluding this preferred orientation for the sidechain ring of this residue. For 10W the observed NOEs, 10WCHD1-10WNH 10WCHE3-10WNH, again defined an orientation for the sidechain ring in which the ring was pointing towards the N-terminus of the peptide (the preferred rotamer in Trp-X peptides was also that with the indole near to the N-terminus, Chen et al., 1991). In this case other very weak inter-residue NOEs confirmed the orientation of the sidechain (eg. 10WCHE3-7ECH2B, 10WCHD1-9ICH3G2). For both 11H and 10W, these intra-residue NOEs do not appear to arise as the result of spin diffusion, since they are selective and are consistent with a single conformation of the residue ring. The observation of these NOEs in the absence of similar intra-residue NOEs for residues 3W and 12H suggests that residues 10W and 11H are conformationally restricted, and that the structure adopted by these sidechains may define the end of the helical conformation in the N-terminal half of the peptide. It may be significant that this is the region where the backbone NHi-NHi+1 NOEs are lost.

Some vague indications are seen of helical ordering of the sidechains as indicated by very weak medium-range sidechain to sidechain connectivities (eg. 3WCHD1-6MCH2B, 10WCHE3-7ECH2B), but these are not too reliable as indicators of secondary structure. Two unusual NOEs were observed (6MNH-7EαH, 5DNH-6ECH2ß) that probably arose as a result of spin-diffusion. They occur in the region which shows helical structure and so may represent increased rigidity of the backbone in this area. No evidence for the maintenance of the π-helix observed in the crystal coordinates, at the C-terminus of the peptide was found.

In conclusion it is clear that the peptide Y847 shows considerable helical character over the N-terminal portion of the peptide, as indicated by NHi-NHi+1 NOEs (4D-10W) and reduced amide temperature shift coefficients (4D-9I). The sidechains remain disordered and so the peptide cannot be said to have a significant helical structure. The data is similar to that obtained by Dyson et al. (1988), and such a structure was dubbed as 'nascent helix'. This means that individual pairs of residues show a tendency to adopt a structure in which the backbone forms a turn (populates the α-helical region of phi,psi space). If these pairs of turns occur simultaneously in adjacent residues, then the structure becomes helical, and longer range NOE connectivities are observed (eg. i,i+3). However, in nascent helix, there is no ordering of the structure along the peptide chain, and so only local indications of the turns are observed. In the C-terminal portion of the peptide (10W-18L), the indication of helix is even more tentative, with no observed NHi-NHi+1 NOEs, but slightly reduced temperature shift coefficients. Helix is indicated, but the population of conformations will contain a much lower proportion of the local turns, and will hence be principally extended random conformations.

A further series of experiments were performed at an elevated temperature (HOHAHA 60msec mixing time H2O pH7.8 299K; NOESY 200msec mixing time H2O pH7.8 299K) to determine any changes that had taken place in the structure of the peptide. The assignments made and selected NOESY connectivities found are listed and the amide-amide and fingerprint regions of the NOESY are seen in figure 3.2.13 (assignments were simplified because no significant changes, except upfield shifts of the amides, had taken place with the increased temperature from 280 K). It was noted that the 200msec NOESY showed many fewer crosspeaks than that at 280K. This applied not only to connectivities between residues (eg. Hαi-NHi+1), but also to those within the residues. Those NOEs which were preserved were in the N-terminal half of the peptide between residues 4 and 11 (data not shown). All the NOEs in NOESY spectra at both 280K and 299K were negative. At the higher temperature the loss of NOE crosspeaks can be accounted for by assuming an increased mobility of the peptide and thence a transfer from a structured to an unstructured conformation, the molecule moved from the 'negative-NOE' towards the 'zero-NOE' rotational regime. The most dramatic changes were in the C-terminal half of the peptide, which was the least structured region in water at a lower temperature. The alpha helix, over the region 4D-10W, appears to be preserved, but the intensity of the peaks (which is proportional to the % of alpha helix in the population, since the protection of the amide protons from exchange arises from helical backbone hydrogen bonding) is much reduced. These characteristics are consistent with a larger proportion of the peptide population adopting an unstructured conformation. Insufficient time was available to run a ROESY spectrum, but it is expected that this would have shown the short range sequential connectivities that are lacking in the NOESY spectrum. (eg. Hαi-NHi+1), so that the peptide would be moving in the 'zero NOE' correlation time frame.

Chemical shifts of the resonances were also determined by HOHAHA at 299K for comparison with values at 280K. The trend of downfield shifts of the HA resonances, away from an alpha helical conformation, towards the C-terminus of the peptide as the temperature is increased, is illustrated in table 3.2.14, but the shifts observed are very small. An opposite trend of similar magnitude is shown for alpha protons in residues 2N-8K, but this is unlikely to indicate an increase in helicity at the higher temperature, but instead a structure independent temperature dependent shift. A similar comparison of the amide chemical shifts illustrates a loss of helical character (increased rate of downfield shift as the temperature increased ) for residues 10W to 18L (figure 3.2.3).

In conclusion, the helical character in the N-terminal portion of the peptide appears to be significantly reduced and any small indications of helix in the C-terminal portion of the peptide are lost.

The promotion of alpha helical peptide structure by addition of TFE has been referred to in section 3.1. With this in mind the procedure described in section 3.2. was repeated for peptide Y847 in 50%TFE/50%H2O at 4mM at 299K.

The annotated 1-dimensional spectrum of Y847 in the TFE mixture is seen in figure 3.3.1. In a similar way to the analysis of chemical shifts in aqueous solution, it is informative to examine the difference in chemical shifts between water and the TFE mixture. There are no large solvent dependent shifts associated with the change in solvent (typically a shift of +0.04 to +0.07 ppm may be expected; Nelson & Kallenbach, 1989), and so any shifts which do occur (figure 3.3.2) can be correlated to changes in the secondary structure. Nelson & Kallenbach (1989) reported that regular changes in chemical shift arising from increasing helicity in TFE are demonstrated by alpha protons [-0.06 to -0.25 ppm], while ß-protons may show a downfield shift, and other sidechain protons are much less sensitive to structure and show less regular patterns of change.

Increasing helicity is indicated by the changes in the alpha proton chemical shifts from water to water/TFE at 299K for all residues, except 1T-5D and 17E, which show only solvent dependent shifts. The largest shifts are seen towards the C-terminus of the peptide for residues 9I-11H, 13T-14F. There is an obvious break in the trend of the α-proton chemical shifts at residue 11H. It is interesting to note, that as in water, there is no observed amide proton for residue 15Y, and it is the carbonyl of residue 11H which would be hydrogen bonded to the amide of 15Y in a regular pattern of helical hydrogen bonding. It is probable that these two pieces of data reflect a poor degree of helical secondary structure at around residue 11H. Based upon the ß-proton chemical shifts there is a substantial increase in helical character detected through the sidechain protons of residues 7E-12H. The ß-proton pattern breaks down at residue 13T suggesting that either the disruption of helix is between residues 11H-13T, or that the structural characteristics detected by ß-protons are offset by one or two residues from the source. The pattern in amide chemical shifts is similar to that already described, but because of the sensitivity of the shifts to other factors such as pH, these are viewed as less reliable. Despite the helical ordering of the sidechains which is demonstrated below, the 6M fails to show the shift anticipated in 3.2.4b, so it is unlikely that the 6M is oriented pointing towards 10W, as seen for the corresponding residues in the actin coordinates. Residue 82M is buried in G-actin, but must necessarily be solvent exposed in the peptide.

In summary, it appears that the helical character of the peptide is overall increased in the TFE/H2O mixture, as compared to water, but that the increase in helicity of residues 3W-8K is small. This could suggest that these residues 3W-8K already populate helix to a great, perhaps maximal, extent for the peptide in water. In section 3.2, these residues were seen to be those which showed the largest degree of helical character. It may be that the nascent helix 'turns' are populated in a majority of the conformations adopted by the peptide in water, and that the significant population of such structures does not require ordering of the sidechains. Of course it is possible that this crude method is insensitive to changes beyond a certain threshold. In contrast, the residues in the C-terminal portion of the peptide, possibly 9I-10W & 12H-14Y, have considerable scope to increase their vague helical character indicated in water, and so large changes in the chemical shifts are seen in these areas.

Based upon several 2-dimensional experiments the assignments seen in table 3.3.6 were made. The spectra collected were COSY (collected in 8K x 0.5K points), TOCSY (mixing time 60msec), NOESY spectra with mixing times 50, 200, 400msec (the 400msec NOESY was used for assignment purposes only), all in 50%d3-TFE/50%H2O, 4mM, pH 7.8, 299K, 500MHz).

Virtually complete assignment of resonances in the peptide was achieved by the method of Wuthrich et al. (1982). Briefly, the technique involves establishing the identity of the amino acid spin systems from COSY (which manifests connectivities between protons separated by up to three bonds), TOCSY (remote connectivities) and NOESY (through space connectivities). Identification of these spin systems is begun by locating those with unique coupling of chemical shift properties (eg. Ala, Thr). The search is then widened to include AMX systems (such as Asn, His, Tyr) and finally the long sidechains of Leu, Lys, Glu and Gln. Subsequently individual amino acid spin systems are placed in the amino acid sequence by identification of sequential interresidue NOE connectivities (typically NHi-NHi+1, CαHi-NHi+1 and Cßi-NHi+1). Longer range NOE connectivities arising from secondary structure elements such as alpha helices also aid in confirmation of the assignments.

In the present case it was possible to identify 17 out of 18 possible observable amide proton resonances in the NH-CHα fingerprint region of the TOCSY or COSY spectra. Only one set of resonances was seen consistent with a single population of structures among the peptide. No change in the spectrum was seen as the peptide was diluted, so that the chemical shifts of resonances (especially amide protons since these are the most sensitive to structural changes) showed no significant systematic changes in the concentration range 4mM to 100 uM (a slight variation of pH during the dilution did have some minor effects, data not shown). This indicates that no change in the aggregation state of the peptide was observed over this concentration range.

A brief description of the method of assignment follows. The relevant spectra are illustrated in figures 3.3.3-3.3.5. The assignment is shown in great detail in the NOESY (mixing time 200msec) in figure 3.3.5 and recorded in table 3.3.6.

The spin systems of several amino acids are easily identified. The two threonine methyl resonances are readily resolved (between 1.0-1.2ppm) in the 1-dimensional spectrum. 1threonine is adjacent to the free N-terminus and consequently the αH resonance is shifted upfield of the ί-protons. The αH connectivities to the sidechain protons for this residue are readily identified in the TOCSY spectrum. The identity of the residue corresponding to these crosspeaks is confirmed by the lack of an αH-NH connectivity in the fingerprint region of either the COSY or TOCSY spectra (the N-terminal amide of 1threonine was again lost to rapid exchange with the solvent). In contrast the residue 13threonine shows α, ί & c connectivities to its amide and so is readily identified as compared to 1threonine. The identity of this residue is confirmed by sequential connectivities to neighbouring residues (eg. 12HαH-13TNH). The familiar connectivities of 8lysine E to -α,-c,-d are readily seen in the TOCSY spectrum, where indeed the whole spin system found to be connected except α-e. In COSY it was clear that both the ί-protons were well resolved, but that two resonances for both the d & c protons were degenerate. The d1-triplet methyl of 9isoleucine is easily identified in the 1-dimensional (especially in the CPA spectrum where is pointing in the opposite direction sign to remainder of the methyl protons) and from this start the entire spin system can be identified in both COSY and TOCSY. The c1 protons are resolved but could not be stereospecifically assigned. Once 9isoleucine was identified then 18leucine was readily assigned. d1 and d2 methyl protons were resolved, but not assigned. The aromatic regions of tryptophan are readily identified in the COSY and TOCSY spectra. The 10tryptophan amide was indentified in the fingerprint region by connectivities to the adjacent residues in NOESY (9IαH-10WNH, 10WαH-11HNH). This allowed identification of the 10tryptophan NH-ßCH2 connectivity in TOCSY. The overlapping ßCH2 resonances of 10tryptophan shows a crosspeak to the d1 and e3 proton of the ring, which are well resolved from the corresponding resonances 3tryptophan. From this basis the entire spin-system can be easily followed. 3Tryptophan is assigned in a similar fashion, with this time the sequential connectivity being 3WßCH2-4DNH, and both ß-protons being resolved. The aromatic rings of 14phenyalanine and 15tyrosine are easily resolved in the TOCSY and COSY spectra and connectivities the ß-protons in NOESY to the rings allow the entire spin system to be identified. In TFE/water, unlike in water alone, the backbone amide proton of 15tyrosine was seen and hence sequential connectivities were observed fro this residue in the fingerprint region of the NOESY spectrum. 11/12histidine d2 and e1 crosspeaks were identified as compared to the amide crosspeaks of 2/16asparagine since the latter are absent in D2O. The 11&12Histidine d2&e1 crosspeaks are resolved in the aromatic region of the TOCSY and COSY spectra. As with the other aromatic residues the connectivities of the ring to ß-protons and sequential backbone connectivities in NOESY allowed identification in the whole spin system. 6Methionone was readily recognised in TOCSY and the sequential connectivities in NOESY confirmed it identity. The last six residues to be considered are from three pairs of amino acids. In water the backbone amide resonance of 2asparagine was not seen due to exchange with the solvent, but in TFE/water the backbone amide of 2asparagine was present in the COSY spectrum where the amide shift overlapped with the amide proton 11histidine (and an impurity commonly seen in all spectra). In NOESY and TOCSY spectra resonance was absent. The alpha proton of 2asparagine lies directly under the water resonance. Assignment of the alpha and amide backbone resonances of 16asparagine according to sequential connectivities was easy, and hence by default these resonances for 2asparagine were assigned. The sidechain amide resonances of these residues proved a little more difficult. In each case the peaks were identified in the 1-dimensional CPA spectra, and then assigned according to connectivities to ß-protons in NOESY. The downfield sidechain amide resonances overlapped. The amides of 7glutamate and 17glutamate were very well resolved and so sequential connectivities allowed easy identification of these spin systems.(eg. 17EαH-18LNH, 7EαH-8KNH). In a similar fashion, the amide resonances of 4aspartate and 5aspartate were very well resolved and sequential connectivities allowed easy unambiguous identification. Stereospecific assignments of the beta protons of 18leucine, 8lysine, 6methionine and 3tryptophan were made according to COSY coupling patterns and NOESY connectivities.

The NOESY spectra yielded 149 connectivities within residues and 85 inter-residue connectivities, of which 41 were for interactions greater than 1 residue apart. No long range connectivities were seen greater than four residues apart. These constraints were used as input into the simulated annealing protocol of XPLOR. They are summarized in figure 3.3.7. Most of the crosspeaks listed were present at 50msec mixing time in NOESY, but most of these crosspeaks were very weak. Considerable spin diffusion was identified when the mixing time was increased to 400msec. The inter-residue crosspeaks listed in the figure are largely taken directly from those observed at 200msec mixing time with only a few exclusions made due to spin diffusion.

The typical alpha helical character of the peptide is illustrated well in figure 3.3.5 (200msec mixing time NOESY) in which strong amide to amide connectivities are seen for the peptide extending along the whole chain, (very small for 10W-11H, absent for 11H-12H and 12H-13T and obscured for 13T-14F by close proximity to the diagonal). The helical content of a peptide may be estimated from the ratio of the intensities of sequential dNN NOEs for each residue (Bradley et al., 1990). On this basis helix in the peptide is strongest for residues 6M-10W, 14F-15Y, 17E-18L, 4D-6M and 15Y-16N in approximately descending order, very weak for the region 10H-11H and probably absent in the region 11H-14F. In the fingerprint region, alpha to amide connectivities show precisely the pattern expected for an alpha-helical backbone conformation. For example, the alpha resonance of 9I shows a large (i,i+1) connectivity to the amide of 10W (arising from the presence of a large amount of random coil extended backbone conformation in the peptide), a large (i,i+3) crosspeak to the amide of 12H (arising from groups in close proximity in an alpha-helical conformation), and progressively smaller intensity crosspeaks for (i,i+4) and (i,i+2) connections (characteristic of a minor population of non-regular alpha helical conformations or spin diffusion in regular helix). The αNH(i,i+2) distance in alpha-helix is relatively long, but is significantly shorter in 310 helix. If the connectivities seen for residues 7E-9I and 9I-11H do not arise from spin diffusion, they may represent a small fluctuation between α- and 310 helix in this region of the peptide, with the hydrogen bonding of the backbone carbonyls alternating between i, i+3 & i+4 amides (Ni et al., 1992). Sidechain to sidechain connectivities seen of the type (i,i+3) and (i,i+4) which were observed indicate the ordering of the sidechains into an alpha-helical conformation. The temperature dependence of the backbone amide protons was not measured in the TFE mixture, but it is probable that hydrogen bonding of the backbone, in an alpha-helical fashion, was present where large amide to amide connectivities between adjacent residues were present. However it was found that it was not necessary to include hydrogen bonding constraints in the XPLOR dataset in order to generate a good alpha helical structure. Unlike the dataset for nascent helix in water, the NOE constraints included long range sidechain constraints (eg. dαN(i,i+3), dαί(i,i+3)),) which are sufficient to define an alpha helical structure. In the TFE/water mixture augmentation of the NHi-NHi+1 NOEs and a decrease in CαHi-NHi+1 NOE cross peak intensities, relative to the spectrum collected in water, indicate an increased proportion of the helical conformations for each residue, with a concomitant decrease in the unstructured (random coil) component of the peptide. Several new dNN NOE connectivities were seen in TFE/water in the C-terminal half of the peptide (14FNH-15YNH, 15YNH-16NNH, 17ENH-18LNH). However, dNN connectivities were not promoted in the region 10W-13T. The specificity of these changes suggests that TFE merely stabilises an incipient structural preference observed in water and does not induce helix in regions of the peptide which lack such a preference. Helix is hence favoured in residues 4D-10W and 14F-18L with a disruption at 10W-13T and possibly 16N-17E.

It is interesting to note that the while the histidine rings of residues 11His & 12His show some restricted mobility, (CHD2-CH2B NOEs observed, Billeter et al., 1982), the intra-residue NOEs described for residues 10W and 11H in section 3.2.6. are not found. This tends to indicate that, as compared to the structure in water, the sidechain rings of these residues no longer have a significant preference for an orientation in which they point towards the N-terminus of the peptide (see figure 3.3.8).

3JNHαH coupling constants below 6Hz and above 8Hz fall into a range that is structurally informative for short peptides. The classical alpha helical value is 3.9Hz. All the apparent coupling constants measured from the COSY spectrum fall between 5.8 and 7.7 Hz (figure 3.3.6.) and so are consistent with a random coil structure for the peptide. The amide proton peaks showed no fine structure in the 1-d spectrum due to exchange broadening with the solvent. There are two possible explanations for the lack of the small coupling constants characteristic of alpha helix in a peptide which otherwise shows an alpha-helical disposition. The peptide is likely to be sufficiently flexible to be constantly forming alpha helical structure and then re-entering a random coil state, this being the case for individual residues, to a certain extent independent of their neighbours, rather than an all or non process for the peptide as a whole. The coupling constant is averaged throughout all the structures adopted by the peptide, and so a coupling constant lying within the random coil range is simply a reflection of the large amount of time the peptide spends in this random-coil state relative to the helical state. In addition, the method used to measure the coupling was simply the examination of the apparent splitting of the amide to alpha crosspeaks in cross-sections of the fingerprint region of the COSY spectrum. This is known to give a value which represents the upper limit of the coupling (Neuhaus et al. 1983, cancelling of peaks in phase-sensitive COSY by their partners tends to increase the apparent splitting). The relative magnitudes of the coupling constants are not very informative, but it was noted that the 3JNHαH values are smallest in the region 4D-7E, where the helix is relatively strong according to other parameters described elsewhere. The magnitudes of the coupling constants may suggest that the termini of good helix lie at residues 4D and 10W. The apparent 3JNHαH coupling for 2Asn was 2.5Hz. This is probably artificially low, arising from the partial saturation of the alpha proton resonance which lies directly under the solvent peak.

The structure of the peptide Y847 was calculated using the simulated annealing protocol in XPLOR with the constraints shown in figure 3.3.7 as input. Hydrogen bonding for the backbone was not introduced into the calculation. The coupling constants measured were used to provide upper limits to the backbone torsion angle phi. Stereospecific assignments were confirmed by examination of the structures. Those NOE connectivities which had been present in the NOESY spectra, but remained unassigned due to spectral overlap were assigned by model building, although these were not used in the calculation. A total of fifty-six structures were calculated and the thirty-nine structures with the lowest energy were selected for further study. The structures are illustrated in figure 3.3.8 and the statistics for the dataset seen in figure 3.3.9. The total set of selected structures showed no violations greater than 0.4Å. and the maximal sum of all violations for a single structure was 5.9Å. Six systematically violated restraints were relaxed early in the calculation by 0.4Å. The RMS deviation between the main chain atoms is illustrated in figure 3.3.9. The general trend is to higher deviations at the termini and lower deviations towards the centre of the peptide, as would be expected. The small deviation at residue 3W (figure 3.3.9.b) may define the N-terminus of the helical conformation. The high deviation for residue 4D is puzzling, but is not seen in the position of the peptide-bond nitrogen atom. A plot of the statistical standard deviation of the RMSD from the mean is included since this illustrates a trend for less well defined structure at the termini and a steadily increasing flexible structure from the N- to C-termini. The content of alpha helix among the thirty-nine structures is readily accessible in the Ramachandran plots for each individual residue (figure 3.3.9a.) The backbone appears poorly defined in residues 2N and 3W, and thereafter the backbone conformation lies increasingly in the alpha helical region of the plot up to residue 10W. The helix, although present in residues 11H-12H and 15Y-17E, is less tightly defined. In residue 13T the conformation is equally distributed between the helical and extended regions of the plot. Hence this is structurally a breakpoint in the helix. The pictures in figure 3.3.8. show a tendency to helix along the whole peptide. However good alignment of the structures could only be achieved if groups of residues towards the N-termini, C-termini or centre of the peptide are aligned (details in the figure). This leads to a 'bunch of flowers' view of the peptides when aligned from either terminus, which can be accounted for by a partially curved helix (commonly seen, eg. Barlow & Thornton, 1988) and by the variability of the conformation around 13Thr.

Macromolecular motions cover a large range of both distance (up to tens of angstroms) and time (sub-picosecond to second). In intact proteins these motions can be involved in the delivery of ligand to the binding site (oxygen binding to haemoglobin, where the individual motions of numerous groups form an overall breathing motion which opens a transient pathway to the binding site) or in the creation of a functional enzymatic site (hexokinase where the binding of glucose causes the formation of a protected catalytic site by the closure of two domains of the protein around the ligand). These motions are difficult to study directly, but mathematical simulations allow the details to be studied and predictions to be made. These predictions can often be readily confirmed experimentally.

Molecular dynamics simulations involve a description of a molecular system in terms of the energies associated with atoms at particular coordinates in space. A molecular dynamics simulation will assess the full range of conformational space available to a molecular system and record the potential energies associated with each molecular conformation so that global energy minima can be determined. The dynamic behaviour of the system is calculated by assigning velocities to these coordinates (using a Maxwellian distribution of temperatures or kinetic energies, the maximum energy of the system is defined by the temperature) and solving Newtons laws of motion for each atom as a function of time. Covalent bonding terms (correct residue geometry) and non-bonding terms (van der Waals interactions and electrostatic contributions; these make up the most important contribution to the description of the system) are considered for contributions to the calculated forces. For example, bonds are simply treated as springs (Hookes law is used, a good approximation to real behaviour at small displacements of bonded atoms) which can undergo stretching and bending harmonic motions. The force on each atom as a result of interaction with its neighbours is calculated, and this is used to alter the trajectory of the atom in the next time period of simulation. The timestep of the calculation is very small, typically 1 femtosecond, so that the change in potential energy of the system is small in each time period (30x10-15sec is equivalent to one carbon-carbon bond vibration). This is very much a 'brute force' method with the individual calculations being relatively simple, but repeated for the equivalent of several hundreds of picoseconds of dynamics simulation. Usually the dynamics calculation is broken up into three stages: the heating where the "temperature" of the system is increased gradually to that required, an equilibration period during which the velocities of the atoms are periodically rescaled to enable the temperature to stabilise (usually carried on until the temperature throughout the system is homogeneous and stable for longer than 10 psec), and the simulation period in which the experiment actually takes place and the trajectory of the atoms is recorded.

The non-bonded interaction consists of terms for the Lenard-Jones 6-12 potential (repulsive as atoms interpenetrate and attractive at longer distances) and electrostatics in which atomic charges are treated as coulombic interactions (either repulsive of attractive depending on the sign of the two charges) with an "effective dielectric constant" for the medium between the two charges. Hydrogen bonding is described by a combination of these two classes of interaction. In some recent studies a full atomic representation of the solvent has been included (a box of water molecules with periodic boundary conditions imposed is added to the system). However since the number of non-bonded interactions in such a system increases with the square of the number of atoms, the calculation becomes overwhelming. A shorthand representation of the solvent environment was used in this study with 'in vacuo' calculations using a distant dependent dielectric constant. Interactions are more significant if atoms come into close approach, mimicking the expected reduced significance of solvent in such situations, while well separated atoms experience solvent shielding. In addition ionised atoms are assigned scaled partial atomic charges in an attempt to mimic the solvent screening. The absence of specific solvent molecules creates errors due to 'physical effects' like motions into the voids left by the missing solvent molecules, a lack of viscous damping, and the impossibility of forming water mediated hydrogen bonds and water separated ion pairs, but more serious are the errors in non-bonded interactions which arise from the failure to adequately treat the dielectric discontinuity between the polypeptide and the solvent. The solvent screening in peptide systems (fully exposed to the solvent) is often under-estimated (Gilson & Honig, 1988) particularly because induced polarization, long range non-bonded interactions (subject to a cutoff distance) and ionic strength (non-zero ionic strengths require complex terms; Harvey, 1989) are neglected in the calculations. Hence any effects observed in molecular dynamics calculations that can be attributed to electrostatic interactions must be interpreted with great caution. The choice of the value for the dielectric constant is dependent upon this individual system. It is relevant to consider the magnitude of the dielectric constant of the bulk solvent (water 80) and the mean dielectric constant within a protein (estimated in the range 2-4, but subject to large local variations when charged groups are "solvated" deep within a protein by other protein moieties ; Harvey, 1989) when choosing a value. The dielectric constant at the polypeptide interface is likely to be less than in bulk water because the closest water molecules are considered to form a loosely associated hydration sphere. In this study several values for the dielectric constant were used for each simulation condition, so that the effects of electrostatic interactions were highlighted.

In order to predict the stability of the alpha-helical secondary structure seen for the peptide Y847 a series of unrestrained molecular dynamics simulations were run in the program CHARMm. The direction of unfolding was chosen because of the practical impossibilities in exploring the multiple minima in the folding pathway. The protocol is described in the legend to figure 3.4.1. The figure shows the data for a series of simulations in which dielectric constants of 1, 8, 20, 30, and 80 were used at 283K. The table shows a coarse-grained history for the main chain hydrogen bonds during the simulations, but these 'snapshots' actually give quite a good representation of the full dynamics simulations.

At low dielectric constants the whole helix remains remarkably stable and in particular the C-terminus shows extensive helical character, although the 5-turns of which it was originally composed were not entirely lost. At higher dielectric constants the N-terminal to central portion of the helix (extending between residues 3-9) was maintained, although less rigourously, throughout the simulation. However there was extensive variability of the structure in the C-terminal region. For example at a dielectric constant of 80 the 5-turn is extensively populated. Throughout the period of the simulation under any of the applied conditions the structures observed were quite stable with time. As expected there was an unwinding of the helix at both extremes, representing fraying of the helix. Finally the change from helix to unstable turn structure, going from the N- to C-termini, was located around 10W and was seen as a unassigned or irregular backbone structure.

It is tempting to over interpret simulation data, but the following points do emerge. At lower dielectric constants the model is more representative of the peptide structure in TFE/water, while at higher values the structure is a somewhat better representation of the situation in water. It is plain however that the extensive degree of mobility of the peptide backbone in water is not adequately modelled. The structure calculated is too stable. The persistence of the structures over 500psec is indicative of the stability of the real peptide structure. Experience has shown that peptides folded into low energy conformations in simulations do not persist if there is no inherent physical tendency in the sequence to adopt that structure (Dyson et al., 1988). There may also be an indication that good alpha helix can form in the C-terminus from a 5-turn structure and that since the formation is sensitive to dielectric constant that it is an ionic interaction (enhanced at low dielectric constant).

Given the tendency of helices to behave as rigid bodies that are disrupted by sidechain packing reorientations, and the possibility that a charge interaction might be involved in maintaining helix stability, together with the observed discontinuity in regular structure seen in the region of 10W-12H in the above simulations, it was decided to vary the ionisation state of the histidine sidechains and observe the effects on the calculated structure. In the above simulations the HIS residue topology (uncharged and protonated on the ND1) was used. A preliminary investigation was started into the effect of changing these histidine residues to the HSC topology entry (fully protonated and charged). A dielectric constant of 80 and temperature 283K were chosen as suitable conditions for the simulations. No data is presented, but it was clear that the stability of the helix was greater in the trajectory of the fully charged species. This short series of simulations was superseded by real experiments which they prompted. These involved testing the pH dependence of the CD spectra of the peptide. The fully charged histidine pair corresponds to the situation seen at low pH.

Further work on this peptide will require simulations using full atomic representations for both the peptide and the solvent thus creating a more realistic environment for the model of the peptide and more accurate structural calculations. The orientation of the histidine sidechains and for example the interaction of the charged ring with the dipole moment of the helix will be of particular interest in such calculations.

Plane polarised light consists of sinusoidally varying electric and magnetic fields confined to mutually perpendicular planes which are both perpendicular to the direction of propagation. Plane polarised light can excite regular absorption (promotion of an electronic transition in a molecule when the wavelength/energy of the light is the same as that of the transition), but in addition if the chromaphore is optically active it has a asymmetric magnetic field associated, then the excitation of the electron in this field will have a directional component arising from the interaction with the oscillating magnetic field, and will be more efficient when the exciting light field is of a given orientation. Since plane polarised light may be resolved into two circularly polarised components, either mathematically or by optical equipment, which are moving clockwise and anti-clockwise, then it can be considered as the sum of two helical paths of circularly polarised components. On interaction with an appropriate optically active chromaphore, one component will be preferentially absorbed over the other, and the resultant sum is now elliptically polarised (an elliptical screw; both components are still circularly polarised but the amplitude or radius of one is less than the other). A measure of this is given by the molar ellipicity [Θ] which is related to the difference between the extinction coefficients for the two components (ε=εL-εR; [Θ]=3300ε). If left circularly polarised light is absorbed more efficiently than right then the value of [Θ] is positive. Since [Θ] is a difference its value is often small and usual units are 10-5 for optical densities of less than 1.

Chromphores such as the peptide bond are optically active, with the absorption being dependent upon the secondary structure adopted by the residue. A strong minimum is associated with the alpha helical CD spectrum of proteins and peptides at 222nm. This is only seen as a very weak absorption band and corresponds to the parrallel-polarized component of the (n_π*) transition with a maximum at 222nm and range 210-230nm (promotion of an electron from an oxygen non-bonding orbital to an antibonding molecular orbital involving the oxygen, carbon, and nitrogen atoms; Holzwarth & Doty 1965). The transition is sensitive to helical conformations with the magnitude of this CD band proportional to the time averaged proportion of helix at each residue in the system. Other alpha helical transitions are seen close to 208nm (minimum) and at 193nm (maximum). At 222nm ß-sheet has a small minimum (but a maximum at 200nm) and random coil has a small maximum (with a minimum at 200nm; Reed & Kinzel, 1991). Specific sidechain chromaphores may also be optically active depending upon conformation. If the rotation of the chromphoric group is unrestricted then optical activity will cancel because the chromaphore is free to adopt many orientations relative to its neighbouring charges and dipoles. However, a degree of asymmetry and restriction of movement of the chromophore can result in varying amounts of manifested optical activity.

The helical content of a peptide is most readily measured at 222nm, where helical content is taken as directly proportional to the mean residue ellipicity. The precision of the estimate may be improved by considering the other alpha helical bands, but at lower wavelengths the absorbance is larger and hence the measured ellipicity is subject to greater error. In order to calculate the helix content, it is necessary to account for other possible sources of absorption at 222nm arising from other backbone conformations in the peptide and contributions from sidechain chromophores. Of particular interest in peptide Y847 are the far uv absorptions of tyrosine (approx. molar ellipticity: 200-210nm +7200) and especially tryptophan (195nm -30400; 223nm maximum with ellipicity +21300±500deg. cm2.dmol-1 at 222nm; Auer, 1973).

The fraction of helix is most commonly calculated according to the formulae in 3.5.3. Allowance for the optical activity of tryptophan is more difficult. It should include an analysis of the orientation of the chromophore. In this case the close proximity of 11H and 12H residues to 10W means that care has to be taken in assessing the effects of pH changes since these groups will ionise at low pH, although this can probably be neglected as the rings of tryptophan and the histidine point towards opposite termini of the peptide (Auer, 1973). The method of Mertuka et al. (1991) is followed in which a simple subtraction will provide adequate allowance, since a single indole chromophore should contribute no more than +1200 deg.cm2.dmol-1 to the mean residue ellipicity of an 18-residue peptide at 222nm (Auer,1973; Brahms & Brahms, 1980). It is convenient for the calculation that allowance for the coil absorbance is roughly cancelled by the contributions of two tryptophan residues.

C.D. spectra were collected for peptide Y847 in water and in water/TFE mixtures. TFE (50%, 25% or 14% v/v) was added to aqueous solutions of peptide in buffer (10mM sodium phosphate, 0.3mM DTT). The pH of the mixtures was adjusted before addition of TFE. DTT added to the buffer did not interfere with the C.D. spectrum (as found by Mertuka et al., 1991). Spectra were recorded from 260 to 180 nm on a Jasco J600 spectropolarimeter in cells with 1mm pathlengths (200 µl volume) with pH 3.0, 5.0, 7.6 or 9.2., with final peptide concentrations of 70, 80 or 120uM and normalised to a single concentration. Spectra were collected at 20°C or 4°C. Three spectra were collected and averaged for each sample. Data are reported as molar C.D. absorption coefficients (ε), calculated using a mean residue weight of 110. Spectra of the buffer background were collected and subtracted. In addition several near uv. spectra were collected as above using a 2mm pathlength cell, from 340-250nm for peptide Y847 in water, for S1A2 and for S1A2 with added Y847.

In figure 3.5.1a. spectra are seen for peptide Y847 in water at pH 7.6, 5.0 and 3.0 at both 20°C and 4°C. There is no apparent difference in the spectra collected at different temperatures, but the negative ellipticity at 222nm increases from pH 7.6 to pH 5.0. (An anomalous result is seen for pH 3 with an ellipticity approaching that for pH 7.6). Spectra in figure 3.5.1.b show the equivalent experiments to those above (4°C) in a 14% TFE/water mixture. The negative ellipicities at 222nm fall in two classes, smaller at pH's 9.1 and 7.6 and larger at pH's 3.1 and 5.1. The negative ellipicity at 222nm is considerably greater in the 14.3% TFE/water mixture than in water alone. When the proportion of TFE was increased (figure 3.5.1.c, 4°C, 25 & 50% TFE ) the negative ellipicity was again increased. At 4°C in a 50% TFE/water mixture, negative ellipicities at pH 7.6 and 5.2 were similar, with the magnitude at pH 3.1 being larger. An isodichroic point (indicating a two state transition for each residue between helix and coil; Shoemaker et al., 1987) is seen at around 200-202nm in for the 25% & 50%TFE/water mixtures. The isodichroic point occurs at increasingly lower wavelengths in 14.3% TFE/water and then water alone.

The CD spectra for the peptide in water have a broad negative shoulder at 222nm and a low intensity band below 200nm indicating that the peptide that lacks well defined structure, but is not completely random. Correction for the positive ellipicity of tryptophan at around 220nm could possibly amount to a 5-12% additional helical content over that indicated in figure 3.5.1.d. There is a good reason to assign an increased helical content to Y847 as an allowance for tryptophan optical activity. The negative absorption of tryptophan at 190-195 nm, occurs at a somewhat lower wavelength than the large absorption of the alpha helix at about 205 nm, and so as the proportion of helix falls (in water) the isodiochroic point will be 'pulled' to a lower wavelength and disrupted by the increasing relative contribution of tryptophan. The effect will be much less as the proportion of helix increases (with increasing added TFE). This type of effect is seen for peptide Y847 and so it is assumed that the tryptophan makes a similar contribution to the spectra around 222nm. No attempt has been made in calculations to allow for end effects (eg. The lack of helical H-bond partners for the first four NH and last four CO groups).

It is clear that the peptide has considerable helical content in water (around 10%) and that the helix is greatly stabilised by increasing concentrations of TFE (up to nearly 80%). This confirms the data from NMR. In addition the helical content is increased by reducing the pH over a range in which the histidine residues will ionise (7.6-5.1, the random coil pKa for histidine is 6.1, but in the present case two adjacent histidine residues would be expected to reduce each others pKa). It is not clear whether this amounts to additional residues (1-2) becoming helical or to an increase in the proportion of helical conformers adopted by the peptide as a whole (this latter explanation is more likely). It is clear however that ionisation of the histidine residues can trigger a structural change in the peptide. The increased alpha helicity may be linked to the reduced solubility of the peptide at low pH, since increased helicity would be expected to lead to increased inter-peptide aggregation.

Figure 3.5.2 shows the near uv CD spectra for peptide Y847 alone and for the mixture of Y847 and S1A2. The spectrum of S1A2 in this region was negligible on this scale (not shown, S1A2 contains only 5 Trp residues, but multiple Tyr and Phe. Since the net absorptions are approximately zero, the conformations must be such that circular dichroism is cancelled.) The absorptions of tryptophan, phenylalanine, and tyrosine are indicted. In this region, the magnitude and sign of the absorptions of aromatic residues are very dependent on conformation and solvent exposure. In is expected that any changes that occur in the spectrum over this region arise from the peptide (although this is not proven). No changes are seen for tryptophan, but there are changes in the spectra of both phenylalanine and tyrosine, and so these must be involved in some unspecified conformational change.

The addition of S1A2 to Y847 can be seen to result in the broadening of several resonances of Y847. A summary of the affected resonance as seen in figures 3.6.1-3.6.3 is given below.

In the aromatic region 11HDCH, 12HDCH, 3WHZ2, 10WHZ2 14FDCH 15YECH, were all seen to broaden to differing extents. As the concentration of added S1A2 is increased 400 µM solution of Y847 (figure 3.6.1, in the absence and presence of 20 and 40 µM S1A2) these resonances were further broadened. The apparent order of the extent of broadening is as follows: 10WHZ2 > 3WHZ2 > 11H and 12H > 15YDCH > 14FDCH.

A greater degree of broadening may indicate a greater affinity of binding of a particular residue, but the difference is small. In a peptide adopting a rigid rod-like structure, interaction at any one point is likely to cause some degree of non-specific broadening in neighbouring residues.

In the aliphatic region several peaks are broadened to differing extents by the addition of S1A2 to the peptide, including resonances corresponding to 9IGCH3 & 9IDCH3, and to a lesser extent the resonances corresponding to 6MECH3, 1TGCH2 and 13TGCH2. At a higher pH where the threonine GCH2 peaks are resolved, the doublet of 13T is broadened more rapidly than that of 1T (data not shown). Peaks attributable to 18LDCH3 are not broadened at all. Certain peaks which are not well resolved are also significantly broadened by addition of S1A2 to Y847. The 8KBCH2 component of the peak at 1.7 ppm is probably that which is broadened, since the residue corresponding to the other resonances at this position (18LBCH2) is not affected by S1A2. Similarly the resonances at 2.8 ppm and 3.1 ppm which are made up of the groups of 14FBCH2, 15YBCH2 and 8KECH2 and 11,12HBCH2 respectively are broadened, but in this case it is not clear which components are involved in the broadening. Clear broadening is also seen at the resonances of 4DBCH2 and 1TACH. The resonances of 7,17EGCH2 appear to be largely unaffected by the addition of S1A2. These observations confirm the broadening seen in the aromatic region and may also indicate that the C-terminal residue of the peptide is not binding to S1A2.

Because of overlap in the backbone αCH region, and decrease in resolution brought about by the addition of S1A2 to Y847 only two alpha proton resonances provide any information. The 6MACH resonance appears to be broadened to a greater extent than the 18LACH resonances.

The data described above, for the binding of peptide Y847 to S1A2, establishes a specificity of binding in which only certain resonance are greatly broadened. All the residues between 1T and 15Y which give rise to resolvable NMR signals namely 1T, 3W, 4D, 6M, 8K, 9I, 10W, 11H, 12H, 13T, 14F, 15Y broaden to varying extents when S1A2 is added to Y847. The implication is that some or all of the resonances in this region are involved directly in the binding of Y847 to S1A2. The NMR resonances of Y847 arising from 3W and 10W are most broadened by the addition of S1A2 to the peptide, and are therefore good candidates for the site of the interaction.

In figure 3.6.1 increased binding of the peptide Y847 as pH is decreased from 7.8 to 7.35 is shown. Examining the spectra of the peptide alone, it is clear that there are second resonance positions for many of the peaks at the lower pH, so that the spectrum looks apparently somewhat less clean and sharp with a decreased signal to noise ratio. The decreased solubility of the peptide at pH 7.35 is probably the cause of these changes with aggregation of the peptide leading to the formation of multimeric species. It is clear however that the binding of the peptide to S1A2 is considerably enhanced at lower pH. The CD data implied a greater population of helix at lower pH, and so it may be that the peptide binds to S1A2 in a helical conformation and that binding is hence easier when the peptide alone adopts this conformation more readily.

It is important to examine the binding of peptide Y847 to S1A2 in the presence of actin (figure 3.6.3). The loss of intensity of the signal as increasing amounts of actin are added to the Y847:S1A2 system indicate that the interaction of S1A2 with actin has not been blocked by the presence of peptide Y847. This intensity loss is probably caused by the decreased inter-segmental mobility of S1A2 when bound to F-actin. In addition, the binding of peptide Y847 to S1A2 is tighter in the presence of increasing concentrations of actin (eg. see the 9IGCH3 & DCH3 resonances). This is difficult to reconcile with a specific interaction of Y847 with S1A2 that models the acto-S1 interaction. It would have been much more likely that actin would have displaced Y847 from S1A2 in a competitive fashion. However, it is possible that a tight binding site occurs on actin elsewhere and that this is sufficient to provide the acto-S1 interaction which then traps the peptide Y847 bound to a second binding site on the S1 head.

The NOE is a method of both demonstrating the proximity of protons in space and determining their separation. The TrNOE involves an extension to an exchanging system, making use of chemical exchange to transfer magnetic information concerning cross-relaxation between the protons in the bound ligand, from the bound to the free state of the ligand. In this manner negative TrNOEs arising from cross-relaxation between protons in the bound state are visible in the free ligand. For chemical exchange fast on the chemical shift scale, a single set of exchange broardened averaged ligand resonances is observed. The 1-dimensional data suggests that the binding of peptide Y847 to S1A2 does occur within a fast exchange time frame (section 3.6). The NOESY data for the free peptide (section 3.2) was collected with a mixing time of around 200msec. In the TrNOESY experiments the mixing time was much reduced to below 50msec. In the free peptide, crosspeaks were still observed at 50msec mixing time (at 280K), but they were much reduced in intensity. In addition the TrNOESY experiment was run at 290K, at which temperature crosspeaks in the free peptide were very weak even using a 200msec mixing time. Hence in the TrNOESY experiment the crosspeaks arising from free peptide should be weak relative to those arising as a result of transfer from the bound state. The concentration of added S1 was too low for any crosspeaks detected to arise from the protein template. An example has recently appeared in the literature in which alpha helical structure was promoted in an unstructured peptide (peptide from the N-terminus of rhodanase binding to Chaperonin GroEL, Landry & Gierasch, 1991). It was anticipated that a similar effect would be observed upon the binding of Y847 to S1A2.

In principal the transferred NOESY experiment is simple (see chapter 2), but there is a requirement to achieve suitable exchange conditions for observation of NOEs indicative of structure of the bound peptide. The solution properties of the peptide Y847 cause some difficulties in varying experimental conditions in order to see these 'structural' transferred NOEs for the binding to S1A2. The parameters which may be changed in order to manipulate the proportion of bound peptide and the exchange rate between free and bound states are: absolute concentrations of peptide and S1, ratio of concentrations of S1 to peptide, temperature, ionic strength (by addition of potassium chloride), and pH of the solution. However the following restrictions apply in the present case. The peptide concentration must be above 3mM for ease of observation in 1H2O and in the presence of S1, somewhat below 8mM to prevent broadening of the signals due to inter-peptide association. The ratio of S1 to peptide may be varied, in the peptide concentration range 3-5mM, between 1:100 (very slight broadening of peptide resonances) to 1:20 (still possible to recognise specific broadening of the 9I10W resonances). The ionic strength must be kept low in order to avoid aggregation of the peptide. At 5mM the peptide alone forms a high molecular weight aggregate in 10mM sodium phosphates/ 100mM KCl/ pH 7.8, which has virtually no NMR signal and within hours forms a 'jelly-like' solution. The same aggregation is seen with added 25mM KCl at 287K, but is not seen at temperatures of 297K and above. In the absence of added KCl (10mM sodium phosphates/pH 7.8) no such problems are seen. The S1 head is a somewhat labile protein, which is perfectly stable for 36 hours at 287K or below, but which denatures gradually at elevated temperatures, with an accompanying precipitation of the protein and loss of ATPase activity. A spectrum can be collected at 297K in 10 hours, but at temperatures above this the protein denatures too rapidly to be of use. The pH of the solution is also restricted since the solubility of the peptide is dramatically reduced below pH 7.6-7.8, and above pH 8.3 the rate of exchange of the backbone amides becomes very rapid, hence making these signals difficult to observe. Suitable conditions are pH 7.8-8.1. The signal to noise ratio in 1-D or NOESY spectra is improved if a jump-return pulse is used instead of the observation pulse so that the level of presaturation of the solvent may be greatly reduced.

The best NOESY spectra (figure 3.7.1) were collected at 5.1mM peptide, with S1A2 added at a ratio of 1:43 at 290K in 5mM phosphate at pH 8.1 (no added KCl). Under these conditions selective broadening of resonances as described in section 3.6 was maintained. NOESY spectra at ratios of 1:60 and 1:30 showed similar effects to that illustrated in figure 3.7.1. except they were more prominent in the latter case and less prominent in the former, as would be expected for a exchanging system (data not shown). The NOEs observed for peptide Y847 in the presence of S1A2 are summarised in figure 3.7.2. The assignments are made according to spectra described in section 3.2. The relatively broad linewidths meant that it was not possible to verify the assignments by TOCSY. The pH used for the TrNOESY was a little higher than for the experiments with free peptide, because this appreared to reduce the strength of interaction of the peptide with S1A2. The most significant pH dependent movement was for the resonance 3WNH (from 8.39 ppm at pH 7.8 to 8.30 ppm at pH 8.1). This probably reflects close proximity of the free N-terminus of the peptide with the backbone at residue 3W, rather than a binding dependent shift. The NOE crosspeaks observed in the TrNOESY spectra can be assigned to intramolecular NOEs within the bound peptide, as opposed to peptide to protein crosspeaks, because the shifts of both components of the crosspeaks correspond to shifts in the free peptide and as the ratio of peptide to protein is decreased, the observed NOE is decreased (peptide to protein NOEs show an opposite dependence on ratio).

The 1-dimensional spectrum of peptide in the presence of protein is broadened as compared to the spectrum of the free peptide, as seen in section 3.6., arising as a result of exchange broadening from the bound state of the peptide. Resonances from the region of the peptide which interact with the protein are observed to be specifically broadened to a much greater extent than the remainder of the peptide. This broadening indicates a graduated change in T2 for various residues in the peptide. Cross-relaxation rates are increased for a reduced T2 (Noggle and Schirmmer, 1971). Hence a representation of the binding information seen in the 1-dimensional spectrum translated into the NOESY spectrum might be that the intensity of crosspeaks corresponding to residues in the peptide which are binding would be increased, whilst those residues which are not directly involved in the interaction will be effected to a much lesser extent (such highly specific effects have been observed for by Ni et al., 1992). In addition, the relaxation rates in binding residues are expected to be so large that they might show some local spin diffusion (seen only in the free peptide at very long mixing times). Hence, even in the absence of NOEs indicating anything about the bound structure of the peptide, it would be expected to see transferred NOEs locating the binding residues in the peptide.

Comparison of the spectra in figure 3.7.1 with figures 3.2.8. & 3.2.13 will show that these TrNOE experiments were successful within the limitations on the experimental system. No improvement of the observed spectra could be made despite much effort. In the TrNOE experiment a suitable control is usually assumed to be a spectrum of peptide alone, collected under the same experimental conditions (with a longer mixing time) and since few significant NOEs are seen for most residues under these conditions (figure 3.2.13d) any crosspeaks which appear in NOESY are assumed to arise from transferred NOEs. In the present case many, but not all, of the NOEs observed are characteristic of the free peptide, albeit at a much longer mixing time (200msec , not 35msec), lower temperature (280K, not 290K) and lower pH (7.8, not 8.1), so that there is a great similarity between the TrNOESY spectra and the NOESY spectra at 280K in water. However, the NOESY spectra on the peptide were collected with 3-4 times the number of scans as compared to the TrNOESY spectra. It was necessary to reduce the collection time for each experiment with S1A2 added to around 8 hours in order to be sure that a high proportion of active protein would still present at the end of the experiments (two separate experiment with different mixing times were run on the each sample). Although spectra of peptide alone were never collected under these conditions, it would be expected that the signal to noise level would have been reduced significantly, leading to the loss of many crosspeaks. Taking the additional factor of collection time into consideration, the assignment of the observed crosspeaks to TrNOEs, even though they appear in spectra of the peptide alone, is valid.

Several of the alpha helical backbone to sidechain NOEs observed in the TrNOESY spectrum which were never seen in any spectra in water, even at lower temperatures (αißi+3). It is not difficult to assign crosspeaks not observed in any spectra in water as transferred NOEs indicating the bound structure of the peptide. These NOEs are restricted to the N-terminal half of the peptide (residues 5D-12H) and indicate that the bound structure of the peptide is good helix in this N-terminal region, and that this region of the peptide is involved in the interaction with S1A2. It is worth noting that the additional αi-ßi+3 crosspeaks are seen in the lower portion of the spectrum only, so that the spectrum is no longer symmetrical in this region. A similar observation was made by Zilber et al., (1990), and probably reflects the different rates of relaxation for protons in the bound and free states of the peptide.

For other information on the binding of the peptide, slightly more tenuous indications had to be taken into account. Little data was seen to indicate the structure of the C-terminal half of the peptide when bound. This may indicate that binding in this portion is weak, despite the contacts which were observed in the 1-dimensional spectrum (eg. 15Y).

Additional crosspeaks, not observed in the free peptide, are seen in the aromatic-sidechain to alpha-proton region of the TrNOESY spectrum (see table 3.7.2). These are all intra-residue crosspeaks or crosspeaks between adjacent residues, and may partly arise from spin diffusion. In the 1-dimensional experiments, the sidechains of 10WHZ2, 3WHZ2, 11H, 12H, 15YDCH, 14FDCH were all broadened to some extent (shown in descending order). The additional crosspeaks in this area tend to confirm a restricted mobility for these residues in the bound complex, and in a similar descending order. More than half of these additional crosspeaks are seen for residues 10W or 11H, so that restricted mobility of these residues in the the bound complex in particular is confirmed. It was noted in section 3.2.6 that, in the solution of peptide alone in water at 280K a restricted conformation exists for residues 10W and 11H, so that crosspeaks 10WN-10WD1/E3 and 11HN-11HD2 were observed. These same crosspeaks appear in the TrNOESY spectrum, but not in the spectra in water at 294K nor in TFE/water mixture, except at very long mixing times. Since, from the data above it seems likely that residues 10W and 11H are immobile in the bound complex, the structure of the free peptide in water at low temperature may be very much like the structure in the bound state. Other similarities are seen, such as the low population of helix towards the N-terminus of the peptide. It is noteworthy that other residues such as 9I, which are dramatically broadened in the 1-dimensional spectrum, do not show especially large NOE crosspeaks nor spin diffusion in the TrNOESY spectrum and so 9I may not be directly immobilised in the bound peptide, but instead may be broadened as a result of the interaction of neighbouring residues with S1A2.

The key question to be answered by this experiment is not whether the peptide is helical when bound to S1A2, since this would almost certainly be the case, but which residues are those principally involved in the binding. It is clear that the region 3W-11H is involved in the interaction, and that helical ordering of the sidechains is promoted in this region of the peptide by the interaction with S1A2. The failure of the TrNOESY spectrum to detect any binding in the C-terminal portion of the peptide, beyond residue 12H, would tend to exclude this as a major interaction site. The 1-dimensional spectra have indicated several residues which are specifically broadened by the interaction of S1A2 with Y847 (section 3.6), but it is not conceivable that so many consecutive residues are directly involved in the interaction since in a helix the residues spiral around the surface, but little definition of the specfic residues involved in the interaction is apparent in the spectrum. Instead it is likely that certain residues are broadened as a result of restricted mobility arising from the interaction of their neighbours. The residues which are principal contact sites might be expected to show spin diffusion even at very short mixing times (the NOESY spectrum is sensing a large change in the correlation time for these residues) and to have additional connectivities that are not present in the free peptide. In particular at the C-terminus of the binding region the data suggests that residues 11H and 10W, but not 9I are directly involved in the interaction with S1A2. The extensive broadening of residue 9I may arise as a result of restriction of the sidechain motions due to helical packing in the bound state.

The helical character of the bound peptide in residues 3W-9I is increased over that observed in the free peptide. It is possible that these are also key residues in the binding interface between peptide Y847 and S1A2. However, a helix is somewhat rod-like and hence behaves as a rigid body throughout. Again , all the residues in this region of the peptide cannot be involved in the interaction with S1A2. In the absence of any well defined data it is only possible to conclude that the region 3W-9I is helical in the bound state, and that some residues within this region must be contacting the surface of S1A2. It is interesting to note that the resdiues 10W-11H lie at the C-terminus of the binding region. It seems likely that since interaction between proteins often involves not only the sidechains, but also hydrogen bonding to the backbone, that where a helix is involved in the interaction between proteins, the only 'spare' backbone hydrogen bonding capacity is at the end of the helix. Hence the tendency for termination of the helix at resiudes 10W-11H in the peptide alone may directly reflect the function of the region in binding to S1A2.

The key points to emerge from these experiment are that the bound structure of the peptide will contain good alpha helix in the residues 3W to 9I, and that residues 10W and 11H appear to be principal contact sites.

Recently Rao and co-workers (Landy et al., 1992 and personal communication) have raised the question of crosspeaks arising in TrNOE experiments as a result of non-specific interactions. It was observed that upon binding of nucleotide to various enzymes, crosspeaks from specific binding at the active site were observed together with crosspeaks indicating the structure of the nucleotide in a non-specific complex. These non-specific crosspeaks were only eliminated at substrate to enzyme ratios approaching 1:1. Such conditions are not available in the present case, since the linewidths of the peptide at such ratios are very large, so that the observation of mixed specific and non-specific crosspeaks cannot be ruled out.

The result of the transferred NOE experiments provided information on the bound structure of the peptide Y847. It was decided to use molecular dynamics to examine the consequences of binding in this region within the actin crystal coodinates. The information provided by the TrNOESY was that the binding region, from approximately 3W to 11H, adopted a well ordered alpha helical structure, and that perhaps residues 10W and 11H were important contact sites. In the X-ray coordinates of G-actin the residue 87H is exposed for binding, but 86W is hidden behind the helix within the body of the actin structure. Based on this observation, data about the binding of residues 10W and 11H was not introduced into the calculation, although as will be seen latter, the calculation reveals much about the likelihood of such an interaction. It was found difficult to introduce any information about the promotion of helix in the N-terminal portion of the bound peptide, since this region is alpha helical in the crystal coordinates. Hence this area was simply left alone. The only remaining data was that the C-terminus of the peptide fails to interact with S1A2, with little structural information about this region being present. Hence it was decided to model possible changes in the actin structure which might be responsible for exposing residue 86W in order that it become available for binding by altering the structure of this C-terminal portion. In the crystal structure this region exists as a π-helix. In the absence of any data about how to modify the structure, this region was wound up into good α-helix. There is a precedence for this in the structure observed for the peptide alone in the TFE/water mixture. In any case, it was reasoned that the structural consequences of any small changes in the C-terminal area for the remainder of the region would indicate the general possibilities for structural motions, even if the details were not valid.

The actin crystal coordinates were imported into QUANTA 3.2 and hydrogens added to all residues using CHARMm. The residues 150 to 321 were fixed (subdomains 3 & 4) and phi/psi dihedral constraints applied to the residues 91-94 of actin causing the original 5-turn structure to be wound rigidly into good alpha helix. After 50 rounds of steepest descents minimisation a dynamics simulation was run in the same fashion as that described in section 3.4 (dielectric constant of 4, heating and equilibration periods of 5psec, simulation of 25psec). Coordinates were stored every 10 steps (10 fsec). The total calculation time on the IRIS 120GTX was 24 days. The histidine residues during the simulation described were in the uncharged state.

The results are shown in figures 3.8.1, there are several points to be noticed. The exposure (calculated as the Conolloy exposed surface area with a 1Å probe) of the residues 86W in the starting actin structure was negligible and this residue can be considered as hidden and not available for binding. The exposure of the 86W residue during the simulation period was much increased. The exposed area was generally around 20Å2 and never fell significantly below this value. Fluctuations in the precise exposed area and the relative proportions of hydrophilic and hydrophobic groups exposed indicate some movement of the residue within a small cleft of roughly stable size. Most interesting are the two periods during which there was a significant increase in exposure of the residue 86W. The periods amount to around 10% of the simulation time. The structure illustrated in the figure 3.8.1. is that at 15.2psec with residue exposure over twice the average. It is apparent that given the driving force to wind up the residues 91-94 into good alpha helix that it is quite possible that the residue 86W will be available for binding and that other changes in the structure of the C-terminal region may also expose this residue.

In order to assess the other structural changes which occured in the actin molecule during the dynamics simulation, several views of the actin molecule are included in figure 3.8.1. It is essential that the structural integrity of the actin molecule is not compromised by the changes observed. Examining the figure it is clear that the C-terminal residues of the helix (91-94), which are subjected to the helical dihedral constraints, have poor attachment to the underlying structure of actin and are hence easily moved. The residues 77 to 90 are however more intimately involved in the structure and hence any movements in these must be small. In the starting structure, all along the helix length, the contacts to the underlying body of the actin structure are composed of a layer of hydrophobic residues. The aromatic rings of Trp and Phe residues lie parallel to the helix axis and thus present a 'flat' face to the underlying surface. The arrangement of these hydrophobic residues during the dynamics simulation is slightly different. A line of hydrophobic residues available to make contact with the same underlying residues from the body of actin is maintained, but they turn to present the edge of the rings to the underlying residues. Other similar accomodating rearrangements are seen in the remaining hydrophobic resdiues. These movements do not disrupt any contacts and hence would not be expected to result in the denaturation of actin.

The residue 91Y is solvent exposed in both the starting structure and throughout the dynamics simulation. It changes orientation, originally pointing to the N-terminus of the helix, and during the simulation rapidly coming to point towards the C-terminus, with a change in the torsion angle around CB-CG. In these positions residue 91Y remains available for interaction, but the ring moves away from the resdiue 86W. Residue 90F (which shows little tendency to bind to S1A2) is locked into a hydrophobic cage formed by residues 127F and 98P, and hence is able to act as a anchor point for reorientation of the helix during the structural changes. The role of 89T is unclear, but in the original actin structure the 5-turn at the C-terminus causes a crowding of the residues, such as 92N, so that contacts are formed with the methyl group of 89T. This crowding is relieved when the helix is elongated by application of dihedral constraints during the simulation. The residue 89T remains buried.

Examination of the changes in the actin structure as a whole, well away from the region corresponding to peptide Y847, reveal vast rearrangements in the subdomain 2. This region contains a solvent exposed loop (38P-52S) which in the actin-DNase I complex is a major point of inter-protein contact. The model building of the F-actin structure (Holmes et al., 1990) requires rebuilding of this region in order to fit into the filament. Hence rearrangements in this region during the dynamics simulation are expected and probably are more representative of the structure in F-actin (coordinates not yet available) than in the actin-DNase complex.

It is apparent that there must be a driving force and trigger to cause the rearrangements described above within the region of actin corresponding to peptide Y847, upon binding of the S1 head to actin. A possible mechanism presents itself upon careful examination of the structural changes during the simulation. In the starting actin structure there is a tripartite interaction between 88H and both of 93E (i,i+5) and 92N (i,i+4). Throughout the dynamics simulation the distances for both interactions increases, with a greater increase for the interaction 88H-93E. The maximal solvent exposure of the residue 86W corresponds to the maximal disruption of these interactions. It is easy to isolate the cause of these changes. The underwinding of the residues 92-94 in the starting actin structure leads to a smaller distance between these residues (92-94) and those immediately preceding them (87-91). The helix is stretched by winding under the application of dihedral constraints during the simulation, and so residues are further apart along the helix axis. It is interesting that residue 95R which is pointed out into the solvent in the original actin coordinates swings into the body of actin and forms an ionic interaction with 93E. This could be a means of 'pulling away' the residue 93E from its interaction with 88H. The significance of this change may be doubtful. The dielectric constant during the simulation was 4 and this would be expected to cause collapse of solvent exposed residues into the surface of the protein. However this may be a fortuitous indicator of the situation when the two proteins, actin and the myosin head, come into close contact. It is worth noting that the integrity of 95R appears to be essential for the binding of tropomyosin to actin.

Putting together these changes, it is possible to construct a trigger pathway. A change in the ionisation state of the residue 88H would be expected to disrupt its interactions with 93E and 92N sidechains. Examination of the crystal coordinates of actin, indicates that the interactions 88H - 93E/92N are likely to be those which maintain the 5-turn structure in the C-terminus of the region corresponding to peptide Y847. When these interactions are weakened it would be quite feasible for the C-terminal residues (92-94) to rewind into good alpha helix (or indeed into some other structure). The final result of such changes is to expose of the residue 86W, so that it would be available for binding.

The observed binding (tranferred NOE, section 3.7) involves 86W and 87H (and to a lesser extent 91Y, section 3.6), but 88H is less readily effected by the binding. The rearrangement in the actin structure (figure 3.8.1) during the dynamics simulation, exposes the residue 86W, maintains the exposure of 87H and hides residue 88H. Residue 91Y is turned away from the centre of the helix to point towards the C-terminus, but remains solvent exposed. The interaction of S1 with adjacent resdiues 86W and 87H is possible, there would be a steric hinderence to the interaction with 88H, and 91Y is available but turned away from the interaction site. The N-terminal residues (79-85) are able to remain in regular helix, in accordance with the transferred NOE data. Several charged residues are available in this region for possible interaction with S1 (8K, 7E and 5D). It is tempting to conclude that this conformation will be that adopted by the region in actin corresponding to peptide Y847.

EDC is water soluble bifunctional crosslinking reagent that generates a 'zero-length' crosslink (amide bond) between amine and carboxyl groups, and hence is used to form covalent linkages between basic (Lys) and acidic (Asp/Glu) residues of polypeptide chains that come into close proximity. The reagent EDC will only result in the formation of a crosslink if these residues come close to each other in space, either forming a specific ionic interaction with one another, or held close by other interactions elsewhere in the polypeptide chains. A specific noncovalent interactions between two regions of the polypeptide chains is hence trapped covalently. It was anticipated that peptide Y847 would be covalently linked to the region on the myosin head with which it interacts using this reagent, enabling the binding site on the S1 head to be identified.

In order to aid visualisation of the crosslinked peptide, Y847 was iodinated with 125I at the single tyrosine residue (91Y). (All methods for this section are given in chapter 2; all crosslinked and digestion products were analysed by SDS-PAGE using a 10% Bis-Acrylamide gel and autoradiography.) 125I-Y847 was crosslinked to S1A2 using EDC in 10mM-HEPES pH 7.0. The contents of the crosslinking reaction mixture after 1 hour at room temperature are illustrated in figure 3.9.2. The S1A2 heavy chain band (100K) becomes radioactively labelled, indicating that some 125I-Y847 had become crosslinked to the heavy chain. The A2 light chain was not radioactively labelled. Two radioactive bands were observed on the gel which could not be attributed to visible protein bands on the Comassie blue stained SDS-PAGE gel, with apparent molecular weights equivalent to around 20 kDa and 10 kDa. The lower molecular weight radioactive band (10 kDa) was also seen in a control experiment in which the 125I-Y847 peptide was run alone on an SDS-PAGE gel (data not shown), and therefore appears to correspond to free radioactively labelled peptide which is migrating more slowly on the gel than otherwise might be expected for an 18 residue peptide (probably because it maintains an alpha helical structure which hampers progress through the gel matrix). The other higher molecular weight band (20 kDa) is presumably an EDC crosslinked oligomer of 125I-Y847 (not proven). In a control experiment carried out under identical conditions, but in the absence of EDC, no radioactivity was found to have been incorporated into the heavy chain of S1, although a small amount of radioactivity was seen in the band corresponding to free peptide. No effect on the level of radioactivity incorporation into the S1 heavy chain was seen when the crosslinking experiment with 125I-Y847 and S1A2 was repeated in the presence of 200mM KCl. This observation is in agreement with NMR experiments described by D.R.Alessi in which the addition of KCl to a concentration of 0.5 M, did not decrease the observed interaction between Y847 and S1A2 (figure 3.10.1).

In order to localise the region of the S1 heavy chain to which the peptide Y847 had been crosslinked, limited enzymic and chemical digestions of S1A2 which had been crosslinked to 125I-Y847 were performed. These digests are illustrated in figure 3.9.2 and summarised below.

A limited tryptic digestion cleaved the S1 heavy chain at the two expected sites, resulting in 3 proteolytic fragments of molecular weights of 27 kDa, 50 kDa and 20 kDa corresponding to the domains of the S1 head. A 15 min time point from a time course of the digestion is shown in figure 3.9.2. The 20 kDa, 27 kDa 50 kDa fragments, and a partially digested 70 kDa band corresponding to the undigested 50 kDa-20 kDa fragment can be seen. The autoradiograph of the gel clearly shows that the only the 70 kDa and the 50 kDa bands are radioactive. Importantly the 20 kDa and 27 kDa fragments are not radioactive, and hence the peptide Y847 appears to have crosslinked exclusively to the 50 kDa domain of the myosin head. A control trypsin digest using S1A2 not treated with EDC, under identical conditions, gave an equivalent time course trypsin digestion pattern. The chemical treatment with EDC and the crosslinking of peptide Y847 hence did not appear to have disrupted the domain structure of the S1 head as revealed by trypsin digestion. Chaussepied & Morales (1988), have reported that treatment of S1A2 with double the concentration of EDC which we used above, under otherwise identical conditions, apparently does not alter the monovalent, divalent, and actin activated ATPase activities of S1A2.

Formic acid cleaves the S1 heavy chain at a single site, between the acid labile 600Asp-601Pro peptide bond in the 50 kDa domain (Griffiths & Trayer, 1989; see chapter 5), to yield an N-terminal 75 kDa fragment and a C-terminal 26 kDa fragment. Upon formic acid digestion of S1A2 crosslinked with 125I-Y847, it is the 75 kDa formic acid fragment which is radioactively labelled and not the 26 kDa fragment. Hence the crosslinking site of Y847 is localised to the N-terminal 44 kDa region of the 50 kDa domain of S1 (figure 3.9.2)

Chaussepied et al. (1986a) have found that a single site in the 50 kDa domain of S1, located between residues 560-561, becomes sensitive to thrombin digestion after disulphide crosslinking of the SH1 and SH2 thiols. The products of thrombin digestion are hence an N-terminal 68 kDa fragment and a C-terminal 30 kDa fragment. The 1 hour time point from a time course of a thrombic digestion of S1A2 crosslinked to 125I-Y847 is shown in figure 3.9.2. Cleavage products are a pair of ≈30 kDa fragments, and a single 68 kDa fragment. Only the band corresponding to the N-terminal 68 kDa thrombin fragment, as well as the undigested heavy chain were radioactively labelled. The pair of ≈30 kDa fragments were not radioactively labelled. Hence the crosslinking site for peptide Y847 is further localised to the 40 kDa N-terminal portion of the 50 kDa domain of S1, between residues 214-560 of the S1 heavy chain. In a control experiment in which non-crosslinked S1A2 was digested with thrombin only a single 68 kDa and 30 kDa fragment was obtained (data not shown). The reason why two ≈30 kDa fragments, neither of which were radioactive were obtained by thrombic digest of S1A2 crosslinked with 125I-Y847 was not investigated further, but it may be related to the additional cleavage point for thrombin detected by Chaussepied et al. (1986a). In addition to the cleavage point at 560Lys-561Ser, these authors also detected a lower yield digestion site at 553Lys-554Tyr. Chaussepied et al. (1986a) observed a 68 kDa doublet as a result of this second digestion site, but under slightly different digestion conditions this might give rise to the ≈30kda doublet observed in the preparations described here.

In order to confirm the localisation of the 50 kDa domain of S1A2 as the crosslinking site for peptide Y847, and to aid in further more precise localisation of the site, the 50 kDa domain of S1A2 crosslinked to 125I-Y847 was isolated from a trypsin digest, by electro-elution of the fragment corresponding to this domain from a 10-20% bis-Acrylamide gradient gel. This fragment was prepared in better than 95% purity as judged by SDS-PAGE. This fragment was subjected to iodosobenzoic acid (Mahoney & Hermodson, 1979) and CNBr digestion, and these digestion mixtures were analysed on an SDS-PAGE Schagger peptide gel (Schagger & Von Jarrow, 1987). Several of the bands obtained from these gels were found to be radioactive, and were electro-blotted onto PVDF membrane for amino acid sequencing. Unfortunately, these bands were either found to contain more than one peptide, or there was not enough peptide present in order to obtain a sequence. However the preliminary data (not shown) suggests that a CNBr fragment of the S1 50kDa domain beginning with the residues FPKA?DT.. was labelled (in the myosin sequence of Tong & Elzinga (1990), the sequence from residue 541 onwards is FPKATDT... ). This places the binding site of the peptide Y847 between residues 541 to 615 in the 50kDa domain of the myosin head. Since residues 560 onwards are excluded by the thrombin digest fragments, confirmation of this data would localise the binding site of peptide Y847 to residues 541-560, in which there are four residues which could be involved in the EDC crosslinking reaction. These preliminary experiments were intended to provide subdigestion fragments of the S1 heavy chain 50 kDa domain, and by characterisation of the radioactively labelled fragments to enable the precise residues of the S1 heavy chain involved in the crosslinking reaction to be identified. These experiments remain incomplete, but the relatively high yields of the crosslinked 50 kDa preparation and its high purity point the way to future work. Separation of the subdigestion fragments of the crosslinked 50K by HPLC should provide a cleaner preparation of the subdigestion fragments. Alternative methods of purifying the 50K fragment have been described (Muhlrad et al., 1986). The classical approach of Chaussepied & Morales (1988; see section 1.11) in which extensive digestion of S1A2 to which a radioactively labelled peptide had been crosslinked within the heavy chain, with trypsin and pepsin, followed by the isolation of a pure radioactively labelled peptide dimer, using a combination of gel filtration, ion exchange, and HPLC chromatography is an alternative and has the advantage of processing large amounts of the S1A2 crosslinked with peptide, so that in principal large yields of the digestion peptides may be prepared for amino acid sequencing. However since the interaction site has been localised within the 50K fragment, it appears to be sensible to use preparation of this fragment as a first step in any purification process. The ratio at which Y847 was crosslinked to S1A2 and the ATPase properties of S1A2 which had been crosslinked with Y847 also remain to be studied.

Crosslinking studies (eg. Sutoh, 1983) have implicated regions between residues 550-587 in the C-terminal of the 50 kDa domain of S1 as binding to actin. NMR studies have found that a peptides comprising a sequence between 610-616 (and C-terminal to this sequence) of the 50 kDa S1 heavy chain bind to F-actin (Trayer et al., 1991 & reported here in chapter 5). Formic acid cleaves the S1 heavy chain at a single site, between the acid labile 600Asp-601Pro peptide bond in the 50 kDa domain (Griffiths & Trayer, 1989; chapter 5), which is conveniently situated between these regions of S1. Therefore a formic acid digest of S1 is able to differentiate between potential binding at these sites. The data above excludes the possibility that peptide Y847 crosslinks at the site C-terminal to the 600Asp-601Pro digestion site (residues 610-616). The crosslinking site must be in the the N-terminal 44 kDa region of the 50 kDa domain of S1. The data from the thrombic digestion (site between residues 560-561 in the 50 kDa domain; Chaussepied et al., 1986a), further localises the binding site to the 40 kDa N-terminal portion of the 50 kDa domain, between residues 541-560 of the S1 heavy chain. These results do not exclude the possibility that Y847 is crosslinked to the N-terminal region of the actin binding site on S1 identified by Sutoh (1983), between residues 550-560 of the S1 heavy chain.

The experiments described in this section were carried out with D.R.Alessi. Additional experiments performed by D.R. Alessi with peptide Y847 and reported in his PhD thesis (Birmingham) are described in figure 3.10.1

The structure of the isolated peptide Y847 has been examined based upon three NMR datasets under different conditions of temperature and solvent. In water at low temperature the peptide is helical at the N-terminus (residues 3W to 11H), with a breakdown of the helix at around residue 12H-13T, and a weak reformation of the helix towards the C-terminus. The sidechains are disordered in water and show little helical character. The peptide becomes less helical as the temperature is raised in water, but more helical in TFE/water mixtures. The TFE/water mixture tends to stabilise the helical character of the peptide by increasing the ordering of the sidechains, so that well ordered helix is seen in both sidechains and backbone for the regions 3W-11H and 13T-18L. Some indications of helix are seen in the middle portion of the peptide (11H-13T) but the stability is significantly reduced as compared to the remainder of the helix, so that a 'breakpoint' in the helix is still maintained even in TFE. Since the formation of helix is cooperative, and once begun tends to carry on fairly readily, there must be a powerful sequence code for termination of helix in this area.

Evidence for alpha helix came firstly from the presence of dNN(i,i+1) connectivities in NOESY spectra. These are only seen when the backbone dihedral angles lie in the helical region of (phi, psi) space. In the peptide in water, in the absence of medium range NOEs (i,i+3), these dNN(i,i+1) connectivities indicated only a tendency to alpha helix in the backbone, this tendency being exhibited somewhat independently of the structure of neighbouring residues (nascent helix, Dyson et al., 1988), whereas in TFE/water mixtures the presence of medium range NOEs meant that neighbouring residues were helical at the same time and that ordering of the sidechains also took place. As evidenced by the different relative intensities of the NN(i,i+1) NOEs, the population of helical conformers varied along the peptide, rising to a maximum in the region 6M-10W (the distance dNN does not vary greatly over for all possible alpha helical conformations so intensity differences reflect differences in the population of helical conformers).

In regions where helical NN(i,i+1) NOEs were not observed, presumably because the population of helix was too small, some helical tendencies were still noted. Helical hydrogen bonding was measured according to the temperature shift coefficient of amide and alpha protons in the peptide, the probability that residue i was in a helical conformation being related to the hydrogen bonding between the backbone amide proton of residue i+2 and the carbonyl oxygen of residue i-2. Helix formation thus measured was much more sensitive to small populations of helix than the detection of NOEs. At the most sensitive level, helix was indicated in most of the peptide simply by the presence of almost all the backbone amide protons at an appreciable intensity even at pH 7.8-8.2. Direct amide exchange from a helix can be neglected and so the intensity of the amide resonance is related (although rather loosely) to the population of helix, so that where exchange was seen (in residue 15Y) this was related to the lack of a hydrogen bond (in this case related to the loss of helix at residue 11H). The presence of helix in the peptide was confirmed by circular dichroism measurements and the estimated mean population of helix in the peptide found to be around 10% in water and around 70% in a 50%TFE/50%water mixture. Both CD and NMR detect equilibrium populations of folded and unfolded structures, so that helices are fully formed on the timescales of the experiment. The CD gives no information about which residues are in helix, whether the 10% is full helix in 10% of the residues, or a 10% helical population throughout the peptide. Neither technique gives any information about the timescale of the process the folding and unfolding processes. The difficulty of addressing these processes experimentally has lead some workers to use molecular dynamics simulations, and the timescale of unfolding has been shown to take place in the time range of hundreds of picoseconds to several nanoseconds (Tobias et al., 1991). An interesting extension to the experiments described in this chapter would be to follow the methods of Miick et al. (1991) in which the peptide is labelled (spin labelled peptide is used and the reorienataion of the peptide examined by ESR) so that the global reorientations of the peptide on a nanosecond timescale could be studied. It is difficult to justify describing a peptide as helical when the only information available is about the mean conformation. The existence of a helical conformation for a substantial portion of the time is implicit in such a statement, but not proven by simple NMR and CD experiments.

For a significant proportion of the time, peptide Y847 must adopt extended structures, as evidenced by the near random coil 3JαHNH values and the presence of large αN(i,i+1) crosspeaks, so that the population of helix present, especially in water, is fairly low. The transition between helix and extended structure has been examined by several workers using solvated molecular dynamics simulations (Rives & Jorgensen, 1991; DiCapua et al., 1990). The consensus opinion is that helix destabilisation is accompanied the insertion of a water molecule to take the place of a backbone hydrogen bond. The pattern of unfolding of main-chain hydrogen bonds is α-helix, through 310-helix to the final extended structure. The hydrogen bonding of an amide progresses from the carbonyl of the i-4 residue, to that of the i-3 residue and finally to the solvent. Molecular dynamics on peptide Y847 in a solvated environment would allow investigation of similar processes for this peptide, with special attention being paid to the region around 10W-13T.

In the crystal coordinates (see figure 3.1.2.c), there is a tendency for the sidechains of residues in the N-terminal half of the peptide (upto residue 11H) to point towards the N-terminus of the peptide, and for the sidechains of residues C-terminal to point towards the C-terminus of the peptide (note 15Y, which dominates in the figure, is an exception). The effect of this reorientation of the polarity of the sidechains at around 12H is to leave a small 'gap' in the space occupied by the sidechains of the helix (see figure 3.8.1). This gap appears to be important in the binding of the peptide to myosin S1. The manifestation of the gap in the isolated peptide appears to be the breakpoint in the helical conformation in this area. The orientation of the sidechains of residues 10W and 11H was maintained in water at low temperature and in the TrNOESY experiment. It is interesting to note that the structure of the peptide in water at low temperature is very much like the structure obtained for the peptide bound to S1A2.

The structures of several alpha helical peptides have been published recently with a common theme running through these structures, in that they have two alpha helical regions with a flexible hinge region between them. Examples include cecropin A (Holak et al., 1988), melittin (DeGrado, 1981), pardaxin P-2 (Zagorski et al., 1991) and a signal peptide involved in the initiation of protein translocation across membranes (Karslake et al., 1990). In each case there is a definite break in the helical conformation that is correlated with the appearance of proline at, or near the breakpoint. A similar structure has been reported for the gut hormone motilin (Edmonson et al., 1991) which adopts a helical structure in hexafluoro-2-propanol, with an ill-defined hinge region towards the middle of the helix, but in this case the residues at the hinge point are FTYGE. Although in Y847 there is no proline in the helix, a region of instability in the helix is seen around residues 10W and 11H. The helix is stable in water for residues extending from the N-terminus to around 11H, and certainly continues in the TFE/water mixture for residues 13T-18L. In water the evidence for helix in this latter region is slight, but temperature dependence factors of the backbone amides, etc suggest a slight tendency towards helix. The alpha helix is a good structural element and can act as a rigid rod in transmission of binding information around a protein (eg. haemoglobin), but the common pattern arising from these structures is that where a helix alone is involved in biochemical function, there is often a breakpoint in the helix which is intimately involved in this function. The role of the breakpoint is not immediately apparent, but the increased flexibility or the availability of backbone atoms for hydrogen bonding outside the helix may be important.

The interaction of peptide Y847 with S1A2 was investigated. The peptide was shown to crosslink to a single site on the S1 head. In 1-dimensional titration experiments several residues (particularly 3W, 6M, 8K, 9I, 10W, 11H, 12H, 13T, 15Y) were implicated in the interaction. The interaction was also investigated by the transferred NOESY technique. The data indicated a promotion of helical ordering of the sidechains of residues 5D-12H and hence the occurrence of binding residues in this region. Together with molecular dynamics calculations, the TrNOESY experiment indicated that residues 10W and 11H may also be important contact points in the interaction. In the crystal coordinates of actin the residue 86W is hidden within the body of the actin structure, whereas the residue 87H is exposed and available for interaction. The possibility thus arises that the interaction observed is not representative of that seen in the native acto-myosin system. The non-specific interaction of a large hydrophobic residue like tryptophan with a protein is not unlikely, but there is some independent evidence supporting specificity and validity of the observed interaction.

The most compelling evidence is that of Milligan & Holmes (see figure 1.14.1, this combined data is as yet unpublished). Data obtained from cryo-electron microscopy combined with the model for the filament structure of actin identified the region around 83E (which does not appear to interact with S1A2 in 1-dimensional titration experiments) as an interaction site on actin for the S1 head. In the figure 1.14.1.a there is an obvious 'arm' of atomic density emanating from the S1 head and contacting a second actin monomer unit, in addition to the expected interaction of the S1 head around the N-terminus of actin. The higher resolution of the actin filament data allows the site of interaction to be localised, and these authors identify the centerpoint of the interaction as around 83E. The 'arm' of the S1 head stretches down over the front face of the actin monomer (see figure 1.7.2), to contact the N-terminal portion of the helix 77-95 from above. The data would be consistent with any interaction in the region 79W-88H. The well established interaction of S1 with actin at the N-terminus of actin would also involve an interaction on the front face of the actin monomer. Other evidence from our own laboratory (preliminary data) suggests that helix 107-125 does not interact with the S1 head. It is significant that this lies on the rear surface of the actin monomer and hence would not be able to contact the S1 (incidentally this is a good control for the interaction of peptide Y847, since helix 107-125 is an amphipathtic helix of similar type to Y847, but it fails to interact despite the presence of several hydrophobic residues).

In deciding the precise point of interaction of S1 with the helix 77-95, Milligan & Holmes relied on the partial sensitivity of the weak interaction between the S1 head and actin to increases in ionic strength (the affinity of the acto-S1 complex increases 104 fold at physiological ionic strength, but only 102 fold at zero ionic strength during the transition from acto-S1-MgADP,Pi to acto-S1-MgADP, Highsmith & Murphy, 1992). The residue 86E was hence chosen as a convenient exposed charged residue in the area of contact. The interaction of peptide Y847 with S1 does not show any such susceptibility to increasing ionic strength. If the interaction of peptide Y847 were composed only of electrostatic interactions then the interaction would be reduced at increased ionic strength, if the interaction were purely hydrophobic then the interaction might be expected to be driven by an increase in ionic strength, and hence a mixed interface involving both charged and hydrophobic residues may be indicated for peptide Y847. In a similar case, the interaction of a hirudin peptide with thrombin has been shown to involve a hydrophobic cluster of residues brought together at the binding interface by the folding of the peptide (TrNOE study, Ni et al., 1992) and at least one charged residue (by assessment of biological activity in mutagensis studies). In the case of the hirudin-thrombin interaction, the binding of the peptide appears to be dominated by the hydrophobic interactions, but the activity of the peptide is also dependent upon the interaction of a charged residues after binding. The involvement of exposed charged residues on both the peptide Y847 and S1A2 is indicated by the ability to specifically crosslink these two species with EDC.

In order to interpret the data for peptide Y847 interacting with S1A2 in terms of the binding of actin to myosin, it was necessary to incorporate the evidence available into a model building exercise. Such model building must involve some degree of intuition in order to place the single interaction site observed in the experiment into a model of a real system in which there are multiple contacts that are likely to induce many reciprocal conformational changes in the proteins. In the model described in section 3.8, residue 87H is a surface residue which is exposed for easy interaction with the S1 head. The same is true of residue 91Y which is a minor contact point that lies in close proximity to 87H. However residue 86W which is at first buried within the body of actin behind the helix (77 to 94), must be exposed by a structural change in the helix. It was shown in section 3.8 that changes in the backbone torsion angles of the C-terminal residues of the helix were easily accommodated in the actin structure and lead to the exposure of residue 86W for binding. The other contacts between the peptide and the S1A2 lie in the N-terminal portion of the peptide, which appears to remain as a well ordered helix in the bound structure.

To propose the exposure of a previously hidden residue upon binding of two proteins in order to make this residue available to take place in the interaction is quite bold. However, there is considerable circumstantial evidence to back up the claim. In the region of the residue 86W the helix is known to have a 'gap' in the space occupied by the sidechain residues. A rearrangement of structure is obviously more readily accommodated in a region where there is the physical space to allow it to occur. In this area there are a great many bulky residues (unusual in a helix), but buried at the rear of the helix in the actin structure is the relatively small residue 89T. This small residue may be involved as a pivot point in any structural change (in the free peptide in TFE/water the helical backbone is least constrained in this region). Hence it would seem that changes in the local structure in this region in order to expose residue 86W for binding might be quite plausible. It may be significant that one of the only residues varying between skeletal and cardiac actin sequences is 89Thr -> 89Ser. In cardiac muscle actin, the destabilisation of the helix at around residues 12H & 13S may be greater than in skeletal muscle actin. Serine has no hydrophobic character and so is less likely to be accomodated at the rear of the helix, and in addition is known to make hydrogen bonds with the carboxyl of residue n-3 (10W; Fasman, 1989), thus destabilising the regular helical pattern of hydrogen bonding.

The model of the structural changes predicted for the region 77-95 of actin upon binding to S1A2 has been described in section 3.8. The structural changes to expose residue 86W were promoted in the dynamics calculation by constraining the C-terminal residues of the region away from π-helix towards the dihedral angle expected for good α-helix. The principal evidence in favour of this change was that the C-terminal region adopts an alpha helical structure in the TFE/water mixture. In addition, the peptide was shown to bind more tightly to S1A2 at lower pH in 1-dimensional titration experiments and this reduction in pH was correlated to an increase in the helical content of the peptide by CD. However, there are indications that the structure adopted in this area in the bound complex might not be helical. Since a minor contact point is seen at 15Y, if helix were present then it might have been expected to be somewhat immobilised by the contact with S1A2 and to have been observed in the TrNOESY experiment (helix is a rigid rod). Hence the increase in helicity of the C-terminal region of the peptide upon binding to S1A2 is by no means certain, and it is possible that the structural change in actin upon binding S1 could be that the C-terminal area corresponding to peptide Y847 'melts' to adopt an extended structure. However, doubts about the precise structure adopted by this C-terminal area do not change the general conclusions about the possibility of the structural changes around residue 86W occurring in the acto-S1 complex. As was described in section 3.8, the key action in the triggering of the proposed change is the breaking of the interactions between 88H and 93E/92N, which are accompanied by an increase in length of this region upon transition from π- to α-helix. If the C-terminal region of corresponding to peptide Y847 were in fact to 'melt' into an extended structure, then these contacts would be even more readily broken and the trigger mechanism would still be valid. It is perhaps instructive to recall the manner in which changes in winding of super-coiled DNA are achieved. An increase in the winding of one area leads to a change in the winding of a remote area. In the actin structure, the region corresponding to peptide Y847 is 'trapped' within the rigid framework of the remainder of the protein, so that an increased winding from π- to α-helix might be expected to unwind another portion of the local sequence (the occurrence of 310 or overwound helix has been noted by Kabsch and Sanders, 1983, at the termini of many helices in proteins, so such a function might be widespread). Note that a small increase in length of the region corresponding to peptide Y847 is readily accommodated in the loops at either end of the helix.

At this point in the discussion it should be noted that the modelling of the change in backbone dihedral angles at the C-terminus of this region as a cause of the exposure of residue 86W for binding, does not necessarily imply that in the interaction of S1A2 with actin that this acts as the driving force of any structural change. This is simply a readily accessible parameter to be altered during the modelling process. In reality, this is much more likely to be a change which occurs to accommodate movement induced by contacts further towards the N-terminus of the helical region as well as elsewhere within the acto-S1 interface.

The next question to be addressed is whether there is any precedent for the conformational changes which have been proposed. The data in the protein structure databases on protein-protein interactions is very sparse, and in general limited to very long-lived complexes such as multimeric proteins and antibody-antigen complexes. However, Thornton and coworkers have made interesting observations about the likely nature of more transient protein-protein interactions based upon the solved structures of the serine-protese/inhibitor complexes (Hubbard et al., 1991). The inhibitors are idealised substrates for the serine-proteases, and when bound to the protease have an open loop region close to the active site which adopts the same conformation in every case. The best inhibitors maintain the structure of this region even in the free inhibitor. However, an examination of the protease cleavage sites in solved crystal structures of proteins which are digested by serine proteases shows that there is little correspondence between the binding loop expected and the nicksite conformation. Indeed some nicksites are helical and thus totally unsuitable for cleavage by these enzymes. The polypeptide backbone will however have to be presented in a similar manner to the open loop structure observed in the inhibitor complexes in order to be cleaved, so that there must be large structural changes in these proteolytic sites upon interaction with the protease. These will consist of both changes in the sidechain orientations and changes in the backbone structure of the substrate. This description of the recognition of nicksite in proteins by the serine proteases has a lot in common with the proposed mechanism of recognition of S1A2 by actin.

The key effect in the trigger mechanism described in section 3.8 is the change in protonation of the residue 88H. Considerable changes in local structure have been shown by other workers to result from changes in the protonation state of single residues (eg. deprotonation of the N-terminal residue and of a histidine residue in mouse epidermal growth factor, Kohda et al., 1991). The means by which a change in the ionisation state of histidine could occur are also well known. The pKa of histidine (and thus the ionisation state of a histidine residue, since this will fall at around physiological pH) can vary according to direct interactions at the histidine residue, for example an interaction of a carbonyl moiety with the protonated form of histidine will stabilise the positively-charged form and hence increase the pKa of the residue. Likewise an interaction with the uncharged form, for example a hydrogen bond involving the uncharged nitrogen lone pair will result in a decrease of the pKa (Sancho et al., 1992). An interaction with a cationic species (such as another histidine) may also reduce the pKa (eg. a pKa of below 6, Pertuz et al., 1985). Even in the absence of a direct interaction, the pKa of histidine can be influenced. Where the imidazole ring is protected from the solvent, this poor solvation will destabilise the protonated form of histidine and so will lower the pKa. Hence, a change in solvation upon the interaction of two proteins in the region of the histidine residue could be responsible for a change in ionisation state. The protein-protein interaction might simply exclude water or trap a water molecule at the interface.

It is difficult to obtain evidence to support this type of mechanism. In the peptide Y847 alone, it is difficult to investigate the protonation state of the histidine residues by the usual means of pH titration because of the low solubility of the peptide over most of the pH range, and this is doubly true in the case of the bound peptide where the pH resistance of the protein has to be taken into account, as well as the requirement for high concentrations of peptide. However Banchovchin (1986) has collated the chemical shifts of ND1 and NE2 resonances in 15N-labelled histidine as a function of the protonation state and involvement in hydrogen bonding, so in an 15N labelled peptide the protonation states of the peptide in free and bound forms would be readily available (eg. Lodi & Knowles, 1991). The situation would of course be somewhat complicated in the presence of protein, since the exchange rate is such that only averaged resonances between free and bound forms would be observed. Measurement of the protonation state of histidine residues in the unlabelled acto-S1 complex is impossible, so that molecular dynamics of the system in a solvated environment is the only means of investigation.

This peptide corresponds to the region 718-727 in the 20 kDa domain of the S1 head of myosin. The sequence is:

718 7 8 9 0 1 2 3 4 727

N A D F K Q R Y K V L amide C

1 2 3 4 5 6 7 8 9 10

Considerable biological work with this peptide has been previously undertaken by A.M.Keane and is reported in her PhD thesis (Birmingham, 1990) and in Keane et al. (1989, 1990).

The previous work with this peptide is summarized below:

1. In assays of the steady-state Mg2+ATPase activity of acto-S1, the peptide Y933 was shown to act as an inhibitor of the actin-activated Mg2+ATPase activity. The inhibition observed showed all the characteristics of mixed inhibition, with Ki=55uM (measured by liberation of inorganic phosphate at 25°C, 25mM TEA-HCl, 5mM MgCl2, 5mM ATP, 0.6uM S1A1, pH 7.5, with an actin concentration ranging between 1.5-6uM, and a peptide concentration range of 0-120uM).

A similar mixed inhibition was observed for several peptides in the region 702-730.

Peptide Y669 myosin 707-717 (RKGFPSRILY) Ki=44uM

Peptide Y668 myosin 699-709 (GVLEGIRITR) Ki=244uM

Peptide Y630 myosin 702-730 (EGIRICRKGFPSRILYADFKQRYKVLNAS) Ki=5uM

2. In assays of the interaction of the peptide Y933 with F-actin by 1H-NMR, resonances of the peptide were shown to be gradually broadened by additions of F-actin, the resonances broadened most rapidly being those towards the C-terminal end of the peptide (conditions, pH 7, in 5mM phosphate buffer, 25°C, 270MHz, with a concentration range of peptide between 300uM and 1mM and an actin:peptide ratio of between 1:10 and 1:1).

3. In assays of the interaction of the peptide Y933 with F-actin, monitored by direct binding in the ultracentrifuge, the binding was shown to be hyperbolic and to occur in ratio 1 mol of peptide bound per mol of actin. The dissociation constant Kd (32uM) agreed closely with the Ki value (the peptide was labelled by 3H-N-acetylation, and the quantity of peptide that was cosedimented with added F-actin was measured). In the same way, the other peptides mentioned above in note 1, were shown to form a 1:1 complex with F-actin, and the measured binding constants (Kd) were Y669 = 40uM, Y668 = 65uM, Y630 = 9.6uM (only the Kd value for peptide Y668 was significantly different from the Ki value).

4. In assays of ability of Ca2+ to stimulate the Mg2+ATPase activity of S1 in the presence of regulated F-actin (actin/ tropomyosin/ troponin), the peptide was virtually unable to influence the calcium activation, whereas other peptides from the N-terminus of the region 702-730 (peptides Y668 & Y669) were able to shift the pCa-activity relationship significantly to the left, thus acting as calcium sensitisers.

5. In assays of force generation after activation by Ca2+, in chemically skinned rabbit psoas muscle fibres, the peptide Y933 was shown to reduce isometric tension in a dose dependent manner, at both a maximal activating concentration of Ca2+ (pCa=4.5) and at a submaximal Ca2+ concentration (pCa=5.5). In a similar manner peptides from the N-terminus of the region 680-727 (Y668, Y669) were shown to inhibit force development at maximal activating Ca2+, but at submaximal Ca2+ to cause an increase in force development, and thus a leftward shift in the pCa-force relationship. The region of S1 corresponding to peptide Y933 was thus implicated in force development in vivo, while peptides from the C-terminus of the region 680-727 appear to be involved additionally in the Ca2+ sensitivity of the muscle (conditions; psoas fibres chemically skinned with 1% v/v Triton X-100 with peptide added in the concentration range 20-200 uM added to the bath containing the fibres). The data implies that adjacent sequences in the region around the reactive thiols in the 20kDa domain of the S1 head, have different specific functions in the process of interaction of S1 with actin.

All of the above work was unfortunately performed of a somewhat impure preparation of the peptide Y933, as described in section 4.1. Further work was done on a purified preparation of the peptide by A.M. Keane, who found that the conclusions drawn in her work were essentially valid (no significant changes in the binding constants and inhibition constants were found). The work presented here describes the characterisation of a method of preparation of pure peptide, the determination of the solution structure of the peptide and includes a further investigation of the characteristics of the interaction of this peptide with actin.

It will be useful bear in mind some of the properties of ß-strand structures in proteins during this chapter. The extended ß-sheet is a common type of repeating secondary structure, in which the backbone dihedral angles lie in the upper left quadrant of the Rhamachandran plot (near -120°, 140°). As in the α-helix, the hydrogen bonding capacity of the backbone groups is satisfied entirely within the body of a ί-sheet, but since the hydrogen bonds are from one strand to another, the ß-structure is expected to be only stable within the body of a ß-sheet, as opposed to in an isolated ß-strand. That is, the ß-strand pair is the basic structural unit (hence in linear peptides where a ß-strand structure has been detected, this is usually associated with a antiparallel hairpin loop). The strands may be hydrogen bonded in either an antiparallel or a parallel arrangement, and in both of these arrangements the sidechains point alternately up and down from the plane of the sheet, with the inter-strand hydrogen bonds lying in the plane of the sheet. The sidechains on adjacent strands are in register. Among the solved crystal structures, parallel ß-sheets are usually buried within the interior of a protein and so tend to be formed from hydrophobic residues except at the edges, where the sheet is solvent exposed (Fasman, 1989). Antiparallel ß-sheets are typically buried only on one side, while the other side is exposed to the solvent, so that amino acid types tend to alternate from hydrophobic to hydrophilic down the strand (Fasman, 1989). The regular ß-strand is fully extended, but several distortions of the structure are common in proteins. In an antiparallel pair of ß-strands the effects of these distortions would usually be to cause the strands to wrap around each other in a very long pitch, open, right-handed spiral. The driving forces for such distortions can be, for example, to accommodate large residues or cluster hydrophobic residues on the inside of the spiral. The antiparallel ß-sheet is often curved with the solvent exposed residues on the convex face.

The peptide Y933 was originally synthesized by standard t-BOC chemistry and purified essentially as described in the materials and methods (section 2.2.13), except that the initial and final gel filtration steps (G10 column) were performed in 5mM hydrochloric acid and the purified peptide was freeze-dried from this solution prior to storage. As is described below, the peptide prepared in this manner was partially degraded by cleavage of a peptide bond, and also contained a small proportion of unidentified contaminants. Later, the peptide was re-synthesied using Fmoc chemistry, which significantly reduced the contaminant content of the preparation, and degradation of the peptide was prevented by performing all gel filtration steps in a 0.5% trifluoroacetic acid solution.

The NMR spectra for the first preparation of peptide Y933 are shown in figures 4.1.1, 2 & 3 (details described in the figures). Despite the contamination and degradation of the sample, it was found that these spectra provided important information on the structure of the peptide. This information was later confirmed by preparation of the individual components in the mixture. To make identification of the peptide sample under discussion easy, the first preparation containing degradation products is referred to as the 'Y933 peptide mixture'.

By the examination of the TOCSY spectrum for the peptide Y933 mixture (figure 4.1.1; amide to sidechain connectivities) it is plain that, for several amino acid residues, which appear only once in the sequence of the peptide, there are two principal chemical shifts for the backbone amide proton. These are most readily seen for the NH-CH3D connectivities of 10Leu and the NH-CH3G connectivities of 9Val. Based on this initial indication, several peaks were grouped into pairs, with different amide proton chemical shifts and similar or identical sidechain chemical shifts. In most cases where a pair of peaks was identified, one appeared more intense than the other in the TOCSY spectrum. The sequential assignment for the peptide then proceeded as usual, but looking for two species of the peptide, one present in a larger amount in the solution than the other. The residue type for each peak was determined tentatively based upon the chemical shifts and, where pairs of peaks had been identified, then one was assigned as the major (more intense crosspeak in the TOCSY, species I) and the other the minor species of peptide (species II). The next step was to confirm these assignments in the ROESY spectrum (figure 4.1.2). In the fingerprint region, a full sequential connectivity was obtained for the major species (I) from 3PheNα to 10LeuNα, and for the minor species (II) a partial connectivity was seen for residues 3PheNα to 6ArgNα and for residues 8LysNα to 10LeuNα. The assignments made are recorded in figure 4.1.3. The only difficulty in the assignment arose for the residues 4&8Lys and 6Arg, which have similar sidechain chemical shifts. There might have been expected to be six amide chemical shifts associated with these residues, but only five amide protons were identified. Careful examination of the fingerprint region of the ROESY spectrum failed to find a connectivity for 8LysNH of the minor species, but the remainder were identified. The alpha proton chemical shift of residue 8Lys of the minor species (II) was assigned by default, all the other five spin systems having been identified from the fingerprint region connectivities. The identification of the spin systems for residues 1Ala & 2Asp of both species, and of residues 7Tyr & 8Lys of the minor species (II), for which amide connectivities were not found, was confirmed by repeating the spectra on this sample at pH 4.6, with the reduced exchange rate at this low pH allowing the full sequential assignment to be made in the fingerprint region, except for residue 1Ala, for which the backbone amide resonance was still absent (data not shown). The identities of certain protons were confirmed in a COSY spectrum, in particular the sidechain protons in residues 10Leu, 4&8Lys and 6Arg.

In the 1-dimensional spectrum, the major and minor species are readily resolved at several positions, for example the 5GlnCH2G resonances (minor II 2.39 ppm; major I 2.31 ppm) and 7TyrCHE (minor II 6.84 ppm; major I 6.81 ppm). By comparison of the integrated areas under these peaks, the minor (II) species appears to account for approximately 30% of the peptide in the sample.

Based upon these observations the following interpretation was made. In all the other peptides in this study (at around pH 7) the amide protons for the N-terminal one, or usually two residues, were not seen, presumably because of the accelerated exchange rate of these protons at neutral pH, leading to broadening of the resonance, reduction in intensity and increased loss of the proton signal due to saturation transfer from presaturation of the solvent. The same pattern was observed for the residues 1AlaNH & 2AspNH of both major and minor species of the present sample of peptide. The loss the amide resonances for residues 7TyrNH & 8LysNH of the minor species suggests that these may also be N-terminal residues. That is, a partial degradation of the peptide Y933 at the peptide bond 6Arg-7Tyr was indicated. In this case then the minor (II) species identified would consist to the two peptides YKVL and ADFKQR, with major species being the intact peptide ADFKQRYKVL (note that the two degradation products will be referred to throughout the text as though they were a single minor species, although of course they are actually two independent peptides).

To account for the possibility that the second species identified in this sample was simply a different conformer of the intact peptide, with the two conformers in slow exchange, the behaviour of the peptide at several temperatures was observed. There was no difference in the ratios of the integrals in 1-dimensional spectra of the resonances of 5GlnCH2G for the two species between 8°-65°C (data not shown). The equilibrium between two conformers would have been expected to be perturbed by this type of temperature change. Likewise, in a long mixing time NOESY (500 msec mixing time, data not shown) no crosspeaks arising from chemical exchange were seen (if these two species were simply different conformers of the same peptide, then there would have been exchange crosspeaks between equivilant groups, such as 5GlnGCH2, in the two species), indicating that if two conformers were present then they must be very stable, and had not exchanged within the period of the mixing time. While this is possible, the only case where it would be expected in a short peptide would be in trans-cis isomerisation around proline. Hence, the possibility of two conformers giving rise to the pairs of different amide chemical shifts along most of the peptide was excluded.

The NMR spectra in water (90%H2O/10%D2O) and 10 mM phosphate buffer, and phosphate buffer with added 100mM KCl were essentially identical (data not shown). In the TOCSY and ROESY spectra in figure 4.1.1.& 4.1.2. several other crosspeaks were seen which were not assigned. These crosspeaks probably arose from other minor short peptide species in the preparation. These crosspeaks were not seen when the two principal species described above were once again identified in the peptide synthesized by FMoc chemistry (section 4.5.1) and afterwards purposely degraded by treatment with 5mM HCl. Therefore these crosspeaks were taken to be unidentified impurities.

The peptide mixture Y933 as shown in figure 4.1.1. & 4.1.2. (purified in 5mM HCl) was sequenced as described in the section 2.2.3 and the results are shown in figure 4.1.4. From the data it is apparent that an additional N-terminus was present in this preparation of peptide Y933, corresponding to a degradation product of Y933.

Intact peptide ADFKQRYKVL

Degraded peptide YKVL

The hypothesis that the peptide Y933 had partially degraded, by cleavage of the peptide bond 6Arg-7Tyr, was hence confirmed. The ratio of approximately 30% minor degraded species in the sample was also confirmed (figure 4.1.4).

To prove the assertion that the peptide Y933 had broken down, it was necessary to separate the fragments, which had been characterised by sequence analysis and NMR spectroscopy, from the mixture. Unfortunately, our standard peptide separation technique of HPLC chromatography on C18 vydac reverse phase columns, in a water/ acetonitrile/ 0.1%TFA solvent system (section 2.2.13), was not successful in resolving the components of the mixture (see figure 4.1.5; this is why the components of the peptide mixture were not detected in the preparation by A.M. Keane until after the 2-dimensional spectroscopic investigations reported in this chapter). In general this method of peptide purification, used in many laboratories, relies heavily on the quality of peptide synthesized (eg. the efficiency of the coupling reaction), and is poor in separating truncated peptide products (it is however a good method in the separation of the intact peptide from other products of the synthesis reactions, and from partially deblocked species).

Based on the predicted differences in charge in the fragments it was obvious that the preferred means of separation was ion exchange chromatography.

ADFKQR Net charge +1.

YKVL Net charge +2.

ADFKQRYKVL Net charge +3.

The figure 4.1.5 describes the separation of the fragments on Mono-S FPLC ion exchange chromatography. The spectra illustrate the contents of the starting products and fractions taken from the separation on Mono-S. The most convenient resonances to monitor the separation process were the well resolved 5GlnCH2G peaks from the two species. It is apparent that the fragment corresponding to the N-terminal six residues of peptide Y933 was separated from the mixture and hence the degradation of the peptide was established.

The method of preparation of the peptide Y933 as described in figure 4.1.5. was successful in isolating the pure non-degraded product. However, after a final gel-filtration purification step in a solution of 5mM HCl, the second set of resonances re-appeared in the spectrum of the product. It was clear that this final purification step was responsible for cleavage at the 6Arg-7Try peptide bond. A brief investigation of the conditions under which the peptide was cleaved was undertaken.

After standing for several weeks in solution at 1mM, pH 7.1, 5mM phosphate buffer, 4°C, the purified peptide product was found to be stable, so that very little breakdown was detected by 1-dimensional NMR (using the 5GlnCH2G resonances to monitor the breakdown). Upon standing at 1mM peptide, 5mM HCl, 4°C, a small amount of degradation (approx. 10%) was seen over 2-3 weeks. When a sample under the same conditions was frozen, a much larger amount of degradation was seen (approx. 40%). The level of degradation upon freeze-drying a sample, within hours of dissolving it in 5mM HCl, was about 10-15%. The degradation was eliminated by freeze-drying from 0.5% trifluoroacetic acid (v/v).

The data tend to suggest that there is a hydrolysis of the peptide bond 6Arg-7Try in acid. The behaviour of the peptide upon freeze-drying may be understood if it is assumed that, during the freeze-drying process the water-ice is more volatile than the added acid, and so that, as the volume of the sample was reduced, then the concentration of acid in the sample was increased. It appears that in the case of HCl the increase in concentration of acid, as a result of freeze-drying is sufficient to cause hydrolysis of the peptide backbone, while the more volatile TFA does not become concentrated in the sample and hence does not build up to a sufficient concentration to cause hydrolysis of the peptide bond. A similar report exists in the literature, where freeze-drying of peptide samples from various acidic solutions was observed to result in the hydrolysis of amide moieties from the sidechains of Gln and Asn residues and from blocked amidated C-termini (Kortenaar, et al., 1990). Again, the rate of hydrolysis was linked to the volatility of the acid used in the particular sample. The increased rate of hydrolysis when the sample was frozen must arise as a result of the mechanism of hydrolysis. On most occasions the rate of a hydrolysis reaction is decreased as the thermal energy available to the reaction is reduced (lower temperature). In the present case, the increased reaction rate in ice can most readily be explained if a structure adopted by the peptide is responsible for the hydrolysis mechanism, and that this structure is fixed in the frozen solution. The likely structures adopted by the peptide are discussed in sections 4.2-4.4.

Certain other peptide bonds are unusually labile to acid cleavage, for example the Asp-Pro bond, which may be cleaved by treatment by formic acid (see the preparation of the S1 26 kDa fragment, section 5.1), but no such instability has been reported for the Arg-Tyr bond. Hence, it is likely that cleavage in this case arises as a result of a specific catalytic mechanism arising from the structure adopted by the peptide. There are precedents in the literature for this type of specific cleavage of peptide bonds, for example the cleavage of human epidermal growth factor (EGF) upon storage in solution. This naturally occurring peptide is spontaneously cleaved at the 11Asp-12Gly bond (Nascimento et al., 1990; 3Asp-4Ser is also labile). The common factor among these reported labile peptide bond cleavage reactions is the residue aspartate. Both the amino acids asparagine and aspartate have the capability of folding back to the mainchain and interacting via the sidechain carbonyl (C=O) with the amide of the adjacent residue. Various mechanisms for this reaction have been proposed, including the attack of the aspartyl sidechain carbonyl carbon on the peptide bond nitrogen of the adjacent residue, and the attack of the ionised carboxyl oxygen on its own backbone carbonyl carbon to form a non-symmetrical anhydride from which the protonated amine is a leaving group (Piszliewicz et al., 1970). Schultz (1967) described preferential cleavage at all Asp or Asn residues in dilute acid (high temperatures are required), resulting in the release of aspartate, by cleavage at the neighbouring peptide bonds on either side of the residue. In the present case there are no adjacent Asp or Asn residues, but 5Gln is close to the cleavage site. The amino acid glutamine has an additional methylene group which gives it an additional degree of freedom and extra length as compared to asparagine. Like asparagine the sidechain is able to to hydrogen bond to the backbone at neighbouring residues. Despite the additional length of the sidechain, the 5Gln residue would however be of insufficient length to enable the sidechain carbonyl to reach the i+2 backbone amide of 7Tyr in an extended structure (this appears to be the conformation adopted by the peptide Y933 in water, see section 4.2). However, if a helical-type conformation is adopted by the backbone, then the 5Gln sidechain carbonyl may contact the 7TyrNH, providing a possible explanation of the autohydrolysis of the peptide backbone. As is seen later, the peptide Y933 does adopt a partially helical conformation in a water/ methanol mixture. It may be the case that the conformation in which the peptide becomes degraded has a backbone which adopts a series of 4-turns, that this structure is weakly populated in water and that the population is too rapidly changing for hydrolysis to take place. In ice however, this structure could be favoured to some extent, and might be much more long lived. This would account for the increase in reaction rate in ice. However if the situation were this simple, then the rate of breakdown would have been expected to be enhanced in the methanol/ water mixture, whilst in fact no breakdown was observed. There is no significant structural evidence to favour the involvement of the 5Gln sidechain (see next section), so the mechanism of cleavage proposed above may be incorrect and some other means of hydrolysis of the peptide bond, perhaps involving the three positively charged sidechains in the peptide may instead be correct.

Note, that although not seen in figure 4.1.4., a minor sequence was found for the degraded peptide beginning at residue 3Phe. This was seen at a lower level than the major background peaks. I am informed that while lysine and arginine are common background contaminants in the sequencing reaction, that phenylalanine is not often seen (J. Fox, Birmingham, personal communication). In addition the full sequence of the peptide, in step with a possible 3Phe N-terminus, could be picked out from the HPLC traces. The existence of 3Phe as an N-terminal residue would suggest that cleavage has occurred as a result of the action of residue 2Asp.

The ROESY spectrum for the mixture of the peptide Y933 and its degradation products is seen in figure 4.1.2. Additional ROESY spectra were collected for the purified intact peptide (figure 4.2.1). The inter-residue connectivities defining secondary structure in peptide Y933, as observed in the ROESY spectrum, are summarised schematically in figure 4.2.2. Only certain columns from the ROESY spectra are illustrated in figure 4.2.1. The following points are worth noting:

- The existence of very strong Hαi-NHi+1 NOEs suggests an extended structure is adopted by the peptide

- Many of the connectivities detected are of the type (i,i+2), which are consistent with an rigid extended structure. For example in the peptide mixture (figure 4.1.2) the particular long range connectivity 7TyrCHE-9ValGCH2, was present in both the intact peptide and the degradation product YKVL.

- The amide to sidechain region in the Y933 peptide mixture (figure 4.1.2) was somewhat overcrowded. However, several poorly resolved inter-residue crosspeaks were seen in this spectrum, and again, they were seen for both species (in particular 9ValCHB-10LeuNH). The same crosspeaks were seen in the ROESY spectrum of the purified peptide. Although this NOE is seen for all regular secondary structures, ßN (i,i+1), the absence of the NOE for most of the remainder of the residue pairs in the peptide Y933, does imply a restricted rotation of the peptide in the same region 9Val-10Leu for both the intact peptide and the degradation product.

- No inter-residue crosspeaks were seen aliphatic region the ROESY spectrum.

The figure 4.2.2 shows a schematic summary of the NOEs observed for peptide Y933, superimposed upon a sketch of the likely structure adopted by the peptide. The most interesting NOEs are of the type i,i+2 (see the columns illustrated in figure 4.2.1; note in particular 7TyrE-5GlnG & 7TyrE-6Argα, 7TyrE-9ValG & 7TyrE-9Valα and also note 3Fφ-1Alaß).

These NOEs tend to define a structure in which the sidechains of alternating residues are close to each other, in the absence of other contacts between sidechains. Hence the sketch in figure 4.2.2. is of a peptide in an extended ß-strand conformation. It should be noted that the majority of the NOE connectivities observed are between sharp methyl resonances and aromatic rings. NOEs between such groups build up more rapidly than for other groups, and hence the association of these residues may take place over a longer distance or may occur for a shorter time than would otherwise be expected (some of the connectivities do not involve a methyl group, eg., 7TyrE-6Argα & 7TyrE-9Valα, so this is not a feature of all the observed NOEs). Note also that the size of the NOE from 7TyrE to 9ValGCH3 is greater than that for 7TyrE to 5GlnGCH2, and the NOE from 7TyrE to 9Valα is larger than the NOE 7TyrE-6Argα (see columns taken from the ROESY spectrum in figure 4.2.1), so that the aromatic ring of residue 7Tyr may spend a greater amount of time close to residue 9Val, or be closer to residue 9Val on average, than to residue 5Gln. It should be emphasised that although in the sketch in figure 4.2.2, all the NOEs to the residue 7Tyr appear to be satisfied simultaneously, that this is probably not the case, and the side of 7Tyr, as well as the other sidechains, are quite at liberty to swing into two or more positions in which only a portion of the NOEs are satisfied at any one time.

The sidechain to sidechain i,i+2 NOEs observed for the peptide Y933 are somewhat unusual, and the possibility that they arose as a result of an artifact in the experiment was examined. In the ROESY experiment crosspeaks can arise as a result of a a combination of coherent (Hartmann-Hahn) and incoherent (NOE) magnetisation transfers. However, such crosspeaks can be distinguished from real NOE crosspeaks by variation of the carrier offset and the spin-lock power (Neuhaus & Keeler, 1986). The problem of Hartmann-Hahn type transfers is greatest when resonances are approximately symmetrical about the carrier frequency. While the resonances 9ValGCH3 and 7Tyrφ, and 1AlaßCH3 and 3Pheφ are not precisely symmetrical about the carrier, it was considered worth the effort of checking for these artifacts. Hence, in order to confirm the validity of these observed NOE crosspeaks, several experiments were run under different conditions. In one experiment, the transmitter offset frequency was changed by 2 kHz, placing the carrier downfield of the amide region, which should strongly attenuate any Hartmann-Hahn contribution in the original experiment. Since the same connectivities were still observed in this experiment, the NOEs recorded in figure 4.2.2 appear to arise as a result of genuine short inter-proton connectivities (data not shown). The validity of the results was emphasised by the observations in ROESY spectra with a shorter mixing time (150msec) and with a reduced radio frequency power applied as the spin lock. In each case, although the signal intensity was reduced, due to inefficient spin-locking, the same crosspeaks were seen (data not shown).

A NOESY experiment (mixing time 500msec) was run under identical conditions to the ROESY experiments described above. This NOESY yielded a very limited number of reliable crosspeaks (that is, above the noise level) and was therefore not used in the structural determination. All the intra-residue crosspeaks were present, but much reduced in intensity, and many of the inter-residue crosspeaks were almost absent. However, upon careful examination of the NOESY spectrum, the most intense inter-residue NOEs as seen in the ROESY experiment (from the aromatic rings of 3Phe and 7Tyr to the methyls of residues 1Ala and 9Val, respectively) were identified as present in this NOESY experiment. The differences between the NOESY and ROESY experiments can simply be accounted for by the differences in intensities of the crosspeaks, arising as a result of the different dependence of the rate of buildup of the NOEs in these experiments upon molecular rotation.

The signs of the NOESY crosspeaks for the peptide mixture are recorded in figure 4.1.3. For a peptide moving close to the 'zero-NOE' condition the NOESY spectrum is often able to detect a change in the correlation time of the sidechains for certain residues, so that a freely rotating sidechain gives intra-residue crosspeaks which are positive, whilst a sidechain with a restricted motion (i.e. moving in about the same time-frame as the peptide backbone) tends to give negative intra-residue crosspeaks. From the table in 4.1.2 it is clear that where the sign of a NOESY crosspeaks varies between the intact and the degraded species, the intact species is always negative and the degraded species always positive. This can be taken as simply confirming the assignments made in section 4.1, since the degraded peptide species will be smaller, and hence more rapidly moving than the intact species. However, more interesting are the cases in which the sidechains showed negative crosspeaks in both intact and degraded species (in particular for residues 7Tyr and 9Val). For these residues, this indicates a restriction of the motion of the sidechains to rotation regime close to that of the backbone. These residues of restricted sidechain motion are clustered at the C-terminus of the peptide, indicating that this region of the peptide is structured.

The stability of the structure observed in the peptide was examined in a ROESY experiment at 25°C (rather than 10°C). All NOEs indicating secondary structure were lost at the higher temperature, whereas the intra-residue crosspeaks were still seen. The loss of crosspeaks was presumably due to a reduced population of structured forms, or to an increased interconversion rate between folded and unfolded states. The reduction in the intensities of the exchangeable amide resonances at high temperature cannot account for the loss of NOEs between non-exchangeable protons.

Other parameters from the spectra of peptide Y933 showed indications of structure in the peptide. The 3JαHNH coupling constants for both the intact peptide and the degradation products were measured in a COSY spectrum. All peptides are subject to motional averaging, which tends to force the coupling constant values close to the random coil values (about 7-7.5Hz). The coupling constant values for the intact peptide were appreciably larger than the random coil values (up to 9Hz), but in the degradation products (YKVL & ADFKQR) the coupling constants were close to (actually a little smaller than) these random coil values. This tends to suggest a more rigid structure in the intact peptide, as compared to the less structured and more mobile degradation products. The large magnitude of the coupling constants for the intact peptide Y933 tends to indicate that the peptide adopts a rigid extended structure (for example, in parallel ß-sheets (phi=-119°), 3JHNα approx. 9 Hz; Pardi et al., 1984). The relative magnitudes of the coupling constants also contain information about the population of an extended structure at each individual residue in the peptide. A particular residue might be expected to adopt an extended conformation more frequently if its associated 3JαHNH coupling constant is larger. The values recorded in table 4.1.3 would tend to suggest a greater population of extended structures towards the C-terminal end of the peptide. As described in section 2.4.4, it is somewhat difficult to measure absolute values of coupling constants without taking account of corrections of the apparent values for window functions, etc. However, the measurement of larger coupling constants for the intact peptide in relation to the random coil values of the degradation products in a single COSY spectrum of the peptide Y933 mixture, provides a high degree of confidence in relative sizes of the measured coupling constants. Examination of the α-ί connectivities of the COSY spectrum showed no marked tendency to adopt a preferred rotamer conformation in any of the sidechains (data not shown).

An analysis of the chemical shift values of peptide Y933 is shown in figure 4.2.3. Examination of the alpha proton chemical shifts shows in every case, upfield shifts as compared to random coil values. This is the opposite shift to that expected for an extended or ß-strand structure in the peptide, and can be accounted for only by assuming a systematic error in the chemical shifts caused by the difference in the conditions under which the two set of shifts were measured. Some useful data can however be gleaned if the variation of the difference in chemical shift from the median value (around -0.15ppm) is examined (see the sketch graph in figure 4.2.3). It is seen that the difference for residues 3Phe, 7Tyr, 8Lys and 10Leu is significantly smaller than the median value, indicating an increased tendency to ß-strand conformation in these areas (especially towards the C-terminus of the peptide) as compared to the remainder of the peptide. A more useful comparison was that made between the shifts of the intact and the degraded species of peptide Y933. Only small differences in shift are seen in the N-terminal half of the peptide, but those differences which do occur tend to indicate a slightly higher population of ß-strand structures in the degraded product (ADFKQR) than in the intact peptide (ADFKQRYKVL). The shift for 6Arg is unusual since in the degraded peptide this is a free carboxyl terminus. In the C-terminal portion of the peptide, 7Tyr is a free N-terminus in the degraded peptide (YKVL) and hence the since the alpha protons in the intact and degraded species overlap precisely, then this can be taken as indicating an underlying downfield shift in the intact species. A similar and large (0.13ppm) downfield shift is seen for residue 8Lysα. Hence, for residues 7Tyr and 8Lys this data must be interpreted as an increased tendency for the conformation adopted to be ß-strand like in the intact species of peptide (comparison of peptide to random coil values also gave this result above). There is no shift in the alpha protons of residues 9Val and 10Leu and hence no structural change is indicated here between the intact and degraded species. This is discussed in section 4.3.2.

The temperature shift coefficients for peptide Y933 are seen in figure 4.2.3. The temperature dependence of the amide protons was high over the whole molecule (between 7.0-10.8 ppb/°C) indicating that there were no hydrogen bonds present. The lowest measured coefficient was for residue 8LysNH (7.0ppb/°C). Despite, the apparent lack of protection of the backbone amide protons from exchange with the solvent by hydrogen bonding, eight of the possible ten backbone amide protons were present at an appreciable intensity in the peptide Y933. The amide intensities for residues 9Val and 10Leu were especially large, particularly taking account of the high pH of the solution (pH 7.1), and high temperature shift coefficients of the amide protons (9Val = 9.2ppb/°C; 10Leu = 10.6ppb/°C). The protection of amide protons in the absence of hydrogen bonding has been discussed in chapter 3. Briefly, the slow exchange of the amide protons (indicated by the large intensity of the amide resonances) must correspond to protection from the solvent by steric hindrance, preventing the approach of the solvent to the exchangeable amide. This most probably takes the form of protection of the backbone amides by the sidechains of adjacent residues, particularly hydrophobic sidechains (see section 4.3.2). No protection of the amide protons by the formation of hydrogen bonds is expected in an extended conformation.

The chemical shift change of 5GlnCH2G associated with cleavage of the peptide backbone (species I 2.31ppm; species II 2.39ppm; table 4.1.3) may be accounted for in terms of the contact observed between 5GlnGCH2 and 7TyrE. After cleavage of the peptide backbone, then this contact must be lost since the respective residues are then found in separate peptides. Similar large downfield shifts (towards the random coil values) were seen for other resonances in the 5Gln residue between the intact and degraded species (eg. one of the 5GlnDNH2 and one of the 5GlnßCH2 resonances are shifted downfield, table 4.1.3). The likely orientation of the residues 5Gln and 7Tyr in the peptide Y933 is illustrated in figure 4.2.2. A shift in the alpha proton of 6Arg is also seen between the intact and cleaved peptides consistent with the observed NOE 7TyrE-6Argα. In the other connectivities for 7Tyr (7TyrE-9ValG & 7TyrE-9Valα), no change in the chemical shifts of the respective 9Val groups were seen. These residues remain on the same peptide in the degraded fragment and must maintain the same structure in the intact and degraded peptides.

The schematic summary of the NOE connectivities in the peptide seen in figure 4.2.2 has been described previously in this section. It is clear that the connectivities seen are consistent with an approximately rigid extended structure in the peptide, and that there are many more connectivities clustered at the C-terminal end of the peptide. The structure of the peptide illustrated accounts for NOE, chemical shift, coupling constant, and exchange data. The peptide is shown as an extended chain, with peptide sidechains alternating in direction down the length of the peptide. This arrangement is indicated by i,i+2 connectivities (see above), which occur over the entire length of the peptide. The larger than random coil values of the coupling constants, 3JαHHN (about 9Hz), suggest a degree of rigidity of the extended structure. A much more mobile extended backbone conformation would tend to give 3JHNα coupling constants closer to the random coil value (7-7.5Hz).

In the figure, the extended backbone was drawn with a slight curvature, creating a concave and convex face for the peptide. The hydrophobic residues (and hence the i,i+2 connectivities) lie on the concave surface of the structure. The spacing of the sidechains of the residues on the convex surface (4Lys, 6Arg, 8Lys) is greater, in keeping with the lack of i,i+2 NOEs seen for these residues. It should be noted that the illustrated conformation puts the three positively charged residues 4Lys, 6Arg, 8Lys on the same face of the structure. These positive charges are expected to repel each other, which might well be considered to be one of the driving forces to produce a curvature in the extended backbone (arrows on the diagram; another driving force might be the clustering of hydrophobic residues on the concave face). A similar simplistic interpretation can explain why no i,i+2 connectivities are seen between residues 3Phe and 5Gln. The electrostatic attraction of the 2Asp- sidechain for the 4Lys+ sidechain will tend to increase the distance between the residues 3Phe and 5Gln.

At the C-terminus of the peptide, the same crosspeaks are seen in the isolated peptide YKVL (section 4.3.) as for the C-terminal region of the peptide Y933, indicating that the structure of this region was maintained even when the peptide was restricted to only four residues in length. The sequence of this region must contain sufficient information to define an approximately extended structure as described in section 4.3. Since the extended structure of the region YKVL in the peptide would appear to be the same as that calculated for the peptide Y975b (acetyl-YKVL-amide) in section 4.3.2 the structure is not discussed further here.

The conformation consistent with the above data can best be described as extended throughout the peptide. This does not imply the existence of one fixed structure, but rather a strong preference for extended over folded conformations (a random coil conformation contains folded as well as extended structures). Coupling constant and chemical shift data suggest that there was a significant population of extended structures in the ensemble of conformations adopted by the peptide, and that the population of extended structures was probably highest towards the C-terminus of the peptide. The diagrammatic representation of the structure of peptide Y933 in figure 4.2.2. is somewhat simplistic, as was the reasoning behind the structure presented. For example, the curvature of the backbone, is a piece of artistic license, since sidechain motions, of residue 7Tyr, alternatively causing the aromatic ring to contact 9Val and 5Gln would probably best account for the observed NOEs. However, the diagram is a fairly good summary of the data collected to define the solution structure of the peptide. Of course, with the rather limited amount of NOE and other information, several other possible structures could also be proposed. As a result of this lack of conclusive structural data, no XPLOR calculations are presented in this section. In the simulated annealing protocol, the 'one-sidedness' of the NOE constraints (restricted to the concave face in figure 4.2.2) tended to pull the peptide into tight curled structures, which failed to converge, and in which multiple other NOEs should have been observed. The XPLOR routine is not capable of simulating a peptide in which the NOEs are satisfied on a time average basis, but in which at any one time only certain of the observed NOEs are satisfied. The peptide Y933 appears to be a mobile peptide which would benefit from this type of treatment in calculation of the solution conformation.

The mechanism of cleavage of the 6Arg-7Tyr peptide bond discussed in section 4.1.4. relies on the formation of an interaction between the 5Gln sidechain carboxyl group and the 7TyrNH. There is no evidence of protection of the 7TyrNH by hydrogen bonding as detected by the temperature shift coefficient, and no evidence of a significant non-systematic shift of the 7Tyr amide resonance away from the random coil value. Hence, the interaction which is a pre-requisite for the proposed mechanism was not observed at around neutral pH. At pH 4.6, there were no significant change in shift of the amide resonance of 7Tyr (nor of any other alpha or amide resonance of a non-terminal residue) and hence there is no indication that the structure of the peptide changes at low pH. However, the contact of sidechains of residues 7Tyr and 5Gln has been demonstrated above so the mechanism cannot be totally ruled out.

The data in section 4.5 (later) indicated that the binding site for F-actin in the peptide Y933 lies at the C-terminal end of the peptide. Since the C-terminal peptide YKVL was readily cleaved from the intact peptide (section 4.3), and much of the structure observed in peptide Y933 was clustered at around the C-terminus (section 4.2), it was decided to investigate the F-actin binding properties and structure of the isolated peptide 7Y-8K-9V-10L (N.B. throughout this chapter the numbering of residues in varieties of peptide Y933 has been kept consistent with the original peptide Y933 for easy comparison). Some data on this peptide had already been collected in the spectra of mixture of intact and cleaved product (the C-terminal cleaved product is NH3+-YKVL-CONH2, section 4.1). The data was described in the previous section (4.2), and briefly, from chemical shift, NOE, and temperature coefficient data it was concluded that the structure seen for the peptide Y933 was maintained in the four residue peptide YKVL. Since the region of interest lies at the C-terminus of the peptide, a cheap and effective method of preparation of varieties of the peptide Y933 was to synthesize a series of peptides (synthesized from the C-terminus) corresponding to 4, 6, 8 and 10 residues from a single preparation, taking out a measured proportion of the resin from the reaction cup after the addition of every second residue for deblocking and purification. The peptides prepared were:

Y933/4 N Y K V L C

7 8 9 10

Y933/6 N Q R Y K V L C

5 6 7 8 9 10

Y933/8 N F K Q R Y K V L C

3 4 5 6 7 8 9 10

Y933/10 N A D F K Q R Y K V L C

1 2 3 4 5 6 7 8 9 10

The structures adopted by the latter three are discussed in section 4.3.3. In addition, a variety of the four residue peptide with both N- and C-termini blocked was prepared:

Y875b N acetyl Y K V L amide C

7 8 9 10

The spectrum of peptide Y933/4 is shown in figure 4.3.1. annotated with the assignments according to the 2-dimensional spectra described in figure 4.3.2. The assignment process was trivial in such a short peptide. It is significant that the amide protons of 9Val and 10Leu, the latter at almost unit intensity, are clearly seen in the spectrum, despite the shortness of the peptide and the high pH of the solution. The ROESY data for the peptide Y933/4 is illustrated in figure 4.3.2. Several crosspeaks indicating inter-residue contacts were detected (these are tabulated in the figure 4.3.2). These NOEs are consistent with an approximately extended structure. These observed NOEs were used, together with the additional NOEs observed for peptide Y975b, to calculate a structure for the sequence YKVL (section 4.3.3). A series of 1-dimensional experiments was performed at temperatures between 4°-32°C to examine the temperature dependent shifts of various protons in the peptide (figure 4.2.3). Both amide protons of 9Val & 10Leu were present over the whole temperature range, indicating some degree of protection, although the temperature shift coefficients are consistent with full exposure to the solvent (around 10 ppb/°C).

In order to investigate the structure of the sequence YKVL further, the peptide Y975b was synthesized with N- and C-termini blocked by acetyl and amide groups respectively. As has been seen elsewhere, any structure in a peptide is detected most readily by connectivities to backbone amide protons. It was anticipated that blocking the N-terminus of the peptide would result in a decrease in rate of amide exchange for residues 7Tyr and 8Lys, allowing the amide protons to be observed. In order to aid with this objective the pH was reduced to pH 6.6. A portion of the ROESY spectrum collected for peptide Y975b is illustrated in figure 4.3.5. together with a 1-dimensional spectrum and the assignments made for the peptide. It is clear that all the backbone amides were present in this peptide, and there are several more crosspeaks detected as compared to peptide Y933/4. However, little additional structural information was detected. The NOEs observed were as tabulated in figure 4.3.2 (peptide Y933/4) with the addition of:

7YD*-Ac, 7YN-Ac, 7YA-8KN, 8KA-9VG*.

The backbone 3JαHNH coupling constants are consistent with a random extended structure. The structure of the peptide was calculated in the simulated annealing protocol of XPLOR, and the calculated structure is shown in figure 4.3.6. The structure is discussed in section 4.3.4.

As described above, the peptides Y933/6 (NH3+-QRYKVL-COO-) and Y933/8 (NH3+-FKQRYKVL-COO-) were prepared in order to investigate the relationship between the structure adopted by peptide Y933 (section 4.2) and the length of the peptide. TOCSY and ROESY spectra were run on concentrated samples of these peptide (9 and 10 mM respectively) at pH 7.1 in order to investigate the structures.

The inter-residue NOEs observed in peptide Y933/6 were the same as those seen in peptide Y933/4, and since they were limited to the region YKVL they provided no additional information. The 3JαHNH coupling constants measured from the 1-dimensional spectra were consistent with a mostly random coil population of structures (9Val = 7.55Hz, 10Leu = 7.60Hz). For peptide Y933/8, likewise the 3JαHNH coupling constants measured were close to the random coil values (9Val = 7.75Hz, 10Leu = 7.95Hz). These coupling constants are slightly larger for peptide Y933/6 as compared to peptide Y933/4, suggesting a slightly higher population of extended structures in the former. As for peptide Y933, several key structural NOEs were seen in the ROESY spectra,

9ValαH-7TyrE, 9ValG-7TyrE, 9ValG-7TyrD.

9ValßH-10LeuNH, 10LeuNH-9ValG.

The quality of the ROESY spectrum was rather poor despite the high concentration of peptide in the sample, with for example with amide to sidechain NOEs almost completely absent for residues 6Arg and 5Gln, even though intra-residue connectivities for these residues were very strong in TOCSY. Probably the peptide was aggregating at the high concentration conditions used for these experiments (the experiments for peptide Y933 described in section 4.1 & 4.2 were performed at 2.6 & 1.3 mM). It may be significant that the most structured portion of the peptide is at the C-terminus (7YKV10L), but that connectivities to these residues were quite clearly visible in the ROESY spectra, indicating a rapid overall rotation in this region, and so not consistent with aggregation in this area.

The NOEs observed for the sequence YKVL in peptides Y933, Y933/10, Y933/8, Y933/6, Y933/4 & Y975b have common features which suggest that the sequence defines the same structure independent of both the length of the peptide and the presence or absence of terminal blocking groups.

The calculated structure of the region YKVL is illustrated in figure 4.3.6. The alignment of the peptide backbone and sidechains in the calculated structure is quite good. The alignment of the backbone is worst at the C-terminal end, reflecting the reduced number of NOE constraints in this area. The principle differences between the aligned structures are the rotations of the sidechain of residues 7Tyr and 10Leu. That is, the sidechains of these residues are able to point in different directions which are approximately symmetrical about an axis along the peptide backbone (to the left or right of the molecule, illustrated best for 7Tyr in photos 1 & 2, and for 10Leu in photo 4, figure 4.3.6). The principle feature of the structure is contact between the sidechains of 9Val and 7Tyr (photo 2, figure 4.3.6). This contact places the residues 9Val and 7Tyr in a hydrophobic cluster to the bottom of the peptide in the photos, together with residue 10Leu, while residue 8Lys lies at the top of the molecule (connectivities like 8LysαH-9ValGCH3 and 8Lysß-9ValNH define the orientation of the 8Lys sidechain). The backbone dihedral angles for the structures (illustrated in figure 4.3.6) fall in the ß-region of the Rhamachandran plot for residues 7Tyr and 8Lys, but in each peptide one or both of the residues 9Val and 10Leu falls in the α-helical region of the plot. This 'twist' in the backbone in the calculated structure arises as a result of satisfying all the applied NOE constraints simultaneously. Since no violations are recorded for the calculated structure, the twisted backbone cannot be ruled out as a valid conformation, but in reality the structure shown in figure 4.4.6. is probably only poor indication of the average structure adopted by the peptide in this regard. It is more likely that swinging of the sidechains between the observed contacts (as mentioned elsewhere in this chapter) would account for the observed NOEs in a more plausible fashion. The failure of the structure to converge to one or other twisted backbone arrangement tends to agree with this suggestion. It should be remembered that the expected 3JαHNH coupling constants for residues 9Val and 10Leu would be reduced considerably below the values observed in the peptide Y933 if such a mobile and twisted backbone structure were adopted (figure 4.1.3). There is scant evidence to support the arrangement of the sidechain of 10Leu on the same face of the peptide Y933 as 9Val as illustrated in the sketch in figure 4.2.2 (such an arrangement would require a twist in the backbone in this region), since no NOEs are observed between the sidechain of residue 10Leu and the sidechain of 9Val (some of the possible crosspeaks are between methyl resonances which would be too close to the diagonal for easy observation). For a conformation which places Leu in this position, a twist in the backbone away from the extended structure would be required.

The signs of the NOE crosspeaks in phase sensitive NOESY experiment are recorded in figures 4.1.3 & 4.3.2. For small peptides moving close to the 'zero-NOE' condition, it will be recalled that the sign of the intra-residue crosspeaks is sensitive to the independent motion of the sidechains as compared to the backbone. All the NOEs associated with residues 7Tyr and 8Lys are positive in the shorter peptide varieties, and for residue 8Lys this holds in even the longer species, peptide Y933. The high mobility and indeed the spatial disposition of these residues probably reflects their role in maintaining the solubility of the short YKVL peptides, particularly for lysine. Among protein structures lysine is by a very wide margin the most mobile of the sidechains, even when the backbone is well ordered, the sidechain is able to adopt the most possible conformations of any amino acid, and the charged group is involved in multiple, rapidly changing interactions with water (Fasman, 1989). It is likely that the peptide Y975b, with no other charged groups present, owes it very high solubility (>25mM) principally to residue 8Lys. Some negative crosspeaks are seen for residues 9Val and 10Leu even in the shorter peptides, suggesting a partially restricted mobility of these residues.

The 3JαHNH coupling constant data for residues 9Val and 10Leu (eg. figure 4.1.3) in various of the peptides indicates that largest values are seen in the peptide Y933. This may correlate with a reduced population of the rigid extended structures in any peptides shorter than original Y933 (that is, all the residues are required to define the rigid extended structure). The ability to observe of many of the same NOEs in the shorter peptides as in the peptide Y933, despite the reduced population of folded structures, must be correlated to the increased concentration of peptide used in experiments with shorter peptides (maximum concentration used for peptide Y933 is 2.6mM, all other peptides studied at around 10mM).

In the structure shown in figure 4.3.6 (photo 5) the amide protons of residues 10Leu and 9Val are somewhat protected. The 10LeuNH is protected by the sidechain of residue 9Val, and the 9ValNH protected by the sidechain of resdiue 8Lys, which would account for the resistance of these amide protons to exchange with the solvent. Conversely, the backbone amide protons of 8Lys and 7Tyr are relatively exposed and hence are not seen in the peptide Y933/4, and disappear fairly readily from the other peptides in the series as the temperature is raised.

One interesting feature of the temperature dependence studies are the shift coefficients of the non-degenerate pairs of methyl resonances for residues 9Val and 10Leu. One of the 10LeuDCH3 resonances is stationary in Y993/4, but moves downfield in the larger peptides (see figure 4.2.3, max shift is 0.11ppb/°C for peptide Y933/6), while the other 10LeuDCH3 resonance shows a stable upfield shift of 0.48 ppb/°C in all peptides. This indicates that these two methyl resonances are in different environments, with the former falling under the influence of a nearby group which is able to affect its chemical shift. The likely identity of this group is the carbonyl group (C=O) at the free C-terminus (the negative charge at the terminus is excluded since it would be expected to have a shielding influence; the deshielding influence of a carbonyl group lies along the plane of the C=O bond). Examination of the structure in the figure 4.3.6. (photo 5) shows that one of the methyl groups of 10Leu would be more readily influenced by the C-terminus than the other, since it lies closer to this group. The difference between the methyl groups may suggest a restricted motion of this residue sidechain. In peptide Y933/4, these methyl groups are degenerate upto 32°C, suggesting an unrestricted motion of the sidechain in this peptide. The same kind of effect is seen for residue 9Val, where the two methyl resonances m