J. iL’o1.Riol. (1992) 227, 441-452

Structure of Tropomyosin

at 9 kqgstroms

Resolution

Frank G. Whitby’, Helen Kent*, Francesca Stewart’, Murray Stewart*, Xiaoling Xie3 Victoria Hatch’, Carolyn Cohen3 and George N. Phillips Jr’? ‘Department of Biochemistry and Cell Biology Rice University, Houston, TX 77251. U.S.A. 2MRC Laboratory of Molecular Biology Cambridge, CB2 2QH, England 3Rosenstiel Center for Basic Medical Sciences Brandeis University, Waltha,m, MA 02254, U.S.A. (Received

9 December

1991; accepted 24 March

1992)

We have used molecular replacement followed by a highly parameterized refinement, to determine the structure of t.ropomyosin crystals to a resolution to 9 A. The shape, coiled-coil structure and interactions of the molecules in the crystals have been determined. These crystals have C2 symmetry with a = 2597 A, b = 55.3 A, c = 135.6 A and /? = 97.2”. Because of the unusual distribution of intensity in X-ray diffraction patterns from these crystals, it) was possible to solve the rotation problem by inspection of qualitative aspects of the diffraction data and to define unequivocally the general alignment of the molecules along the (332) and (3-32) directions of the unit cell. The translation function was then solved by a direct, search procedure, while electron microscopy of a related crystal form indicated the probable location of molecular ends in the asymmetric unit, as well as the anti-parallel arrangement’. The structural model we have obtained is much clearer than that obtained previously with crystals of extraordinarily high solvent content and shows the two alphahelices of the coiled coil over most of the length of the molecules and establishes the coiledcoil pitch at 140( + 10) A. Moreover, the precise value of the coiled-coil pitch varies along the molecule, probably in response to local variations in the amino acid sequence, which we have determined by sequencing the appropriate cDNA. The crystals are constructed from layers of tropomyosin filaments. There are two molecules in t’he crystallographic asymmetric unit and the molecules within a layer are bent into an approximately sinusoidal profile. Molecules in consecutive layers in the crystal lie at an angle relative to one another as found in crystalline arrays of actin and myosin rod. There are three classes of interactions between tropomyosin molecules in the spermine-induced crystals and these give some insights into the molecular interactions between coiled-coil molecules that may have implications for assemblies such as muscle thick filaments and intermediate filaments. In interactions within a layer, the geometry of coiled-coil contacts is retained, whereas in contacts between molecules in adjacent layers the coiled-coil geometry varies and these interactions instead appear to be dominated by the repeating pattern of charged zones along the molecule. Keywords: tropomyosin;

coiled coil; muscle filament; fiber diffraction; X-ray crystallography

long rod-like tropomyosin molecules contain two chains arranged predominantly alpha-helical parallel and in register. In thin filaments these molecules lie in the grooves of the actin helix, spanning actin subunits, and are joined end to end. Together with troponin, tropomyosin forms a &sensitive switch that regulates contraction in many muscles and appears to involve a change in the azimuthal position of tropomyosin relative to

1. Introduction Tropomyosin is a key component of thin-filament based regulation in vertebrate skeletal muscle and is also found as a component of actin filaments in a range of muscle and non-muscle cells (for a review, see Phillips et al., 1986). The 420 A (1 A = 61 nm) t Author to whom all correspondence should be addressed. OO22~-2836/92/180441-12

$08.00/O

441

0

1992 Academic.

Press Limited

442

F. G. whitby

et al.

Table 1 Summary

of spermine-induced

crystal forms

Space Form x Alternate

cell I

a

b

c

B

259.7

553

135%

97.3

(‘2

791.2 7936

5k3 42.9

277.2

1708

7800

530

6’2 cmm cmm

7604

t

t

Method of crystal analysis

Pattern

FP”P

Staggered Single-doublet-singlet Complex

X-ray diffraction X-ray diffraction EM negative stain Cryo-EM

broad

EM negative

Staggered

II

t Because of the paracrystalline be determined with certainty.

nature

of the specimen,

the actin helix (Haselgrove, 1972; Huxley, 1972; Kress et al., 1986; Parry & Squire, 1973). Early fiber X-ray diffraction studies indicated that tropomyosin was one of the “kmef” class of alpha-helical proteins (Astbury, 1947; for a review, see E’raser & MacRae, 1973) and the presence of a strong meridional arc at 51 A (rather than 5.4 A) indicated that it was based on a coiled-coil conformation (Cohen & Holmes, 1963; Crick, 1953). Previous low resolution X-ray crystallographic studies confirmed the existence of the coiled coil and indicated that the average pitch, or rate of winding of the two alpha-helices around each other (Phillips et al., 1986), was of the order of 140 A as previously predict’ed (Parry, 1975; Stewart & McLachlan, 1975). The amino acid sequence of tropomyosin has a distinctive heptad repeat of hydrophobic residues consistent with a coiled-coil structure (Stone et al.. 1975). In each successive group of seven residues, a-b-c-d-e-f-g, residues u and d tend to be hydrophobic. Because there are roughly two turns of the alpha-helix every seven residues, this pattern produces a hydrophobic seam along one side of each helix and the two heliees in the coiled coil join along this seam in a “knobs-in-holes” arrangement (Crick. 1953). Salt bridges between charged residues in positions e and g appear to add further stability and specificity of chain interaction to the coiled-coil structure (McLachlan & Stewart, 1975). A wide variet,y of proteins have been shown t’o contain coiled coils based on sequence or structural information (for a review, see Cohen & Parry. 1990). Examples include myosin (McLachlan & Karn. 1983), laminin B2 (Beck et al., 199(I), intermediate filament proteins (for a review, see Stewart, 1990), “leucine zippers” and domains of a variety of traw scription factors (Lupas et al., 1991). Influenza virus hemagglutinin (Wilson et al., 1981) has been shown to contain a triple-stranded coiled coil. Of particular interest is the leucine zipper domain of the yeast transcription factor, GCN4, which has recently been solved to atomic resolution (O’Shea et al., 1991). Tropomyosin remains the archetype of this class of proteins because of its long, uninterrupted hept’ad repeat, the lack of other domains of significant size and the role it has played in our understanding of this protein motif.

doublet the b dimension

and the symmetry

stain cannot

The distribution of acidic residues in the tropomyosin sequence is strongly periodic so that they are clustered into 14 groups (McLachlan & Stewart, 1976; Parry, 1975). There is probably also a periodicity in hydrophobic residues located on the coiledcoil surface. These clusters of residues are thought to play an important role in the interaction of tropomyosin with actin and the coiled-coil (bonformation of tropomyosin results in the formation of four groups of seven binding sites on its surface, consistent with it interacting with seven actin subunits. There are slight differences between alternate binding zones along the tropomyosin sequence (McLachlan & Stewart, 1976) and molecular models suggest that one of these series of zones is more regular than the other and so more likely to represent the primary actin binding sites on tropomyosin (Phillips et al., 1986). Studies using a range of novel tropomyosin molecules with internal deletions of all or part of one of these zones produced by sitedirected mutagenesis (Hitchcock-DeGregori & Varnell, 1990) appear consistent with the presence of two sets of seven quasi-equivalent actin binding sites on each tropomyosin chain, but have not been able to establish the precise role of the different sets of sites in the regulation of contraction. Tropomyosin is remarkable in its ability to form a wide array of ordered aggregates. At least’ 15 different crystalline or paracrystalline forms have been identified, including a number of forms that are produced by precipitation with the polycation, spermine (Table 1). The form with the best order is produced by a stoichiometric titration with spermine (Phillips et al., 1987) and shows diffraction to at least 5 a in all directions. Weak reflections have been seen to 3 A in directions parallel to the molecular axes. In conjunction with observations on the location of the molecular ends by electron microscopy, X-ray diffraction data from the crystals have been phased to 9 A resolution allowing the molecular packing and coiled-coil structure to be investigated in greater detail and to establish the precise pitch and its regularity.

2. Methods and Materials Crude described

tropomyosin was prepared as previously (Phillips et al., 1979) from adult porcine ven-

Structure

of Tropomyosin

at 9 Angstroms

tricles and from mature rabbit hearts (Pel-Freez Biologicals). The tropomyosin was purified with the use of a hydroxyapatite column (Eisenberg & Kielly, 1974), with the alpha-alpha fraction chosen for crystallization. Purity was verified by SDS/PAGE and the ratio of abaorbances at 277 and 260 nm. Porcine cardiac tropomyosin crystals were grown in capillaries by liquid diffusion (Phillips, 1985). Briefly, a solution with a low spermine concentration (typically 6040 to 9065 M) was layered above 50 to 100 ~1 of protein solution at 0*070 m-spermine (pH 74). The open capillaries were then immersed in the same low spermine buffer inside a sealed test tube. The conditions found to be most suitable for crystal growth include an initial protein concentration of 20 to 30 mg/ml, and a temperature of 15 to 175°C. Crystals were typically plate-like and grew up to 1.5 mm x I mm x 65 mm in 3 to 10 weeks. Rabbit cardiac tropomyosin crystals were grown by’vapor diffusion with drops at concentrations of 16 to 20 mg/ml in 6070 M-spermine (pH 7.4) and 1 to 1.25% PEG 1OOOt equilibrated at 18°C against wells containing 6020 to 0945 M-spermine. Crystals grew in 3 weeks and were slightly smaller than the porcine cardiac tropomyosin

crystals. X-ray diffraction data from the porcine cardiac tropomyosin crystals were collected at room temperature using a Siemens area detector rotating anode system equipped with Franks double focusing mirrors to obtain a finely focused beam. Detector-to-crystal distance was 140 mm and a helium path was used to reduce air scattering of X-rays. The beam-stop size was minimized and it was suspended on a thin mylar strip within a few millimeters of the detector surface to allow very low resolution reflections to be recorded. Typically frames were collected as scans of 65 to 1.0” for 90 to 300 s. The data were processed using the XDS software (Kabsch, 1988). Data were collected from 4 crystals and scaled using the program XSCALE (W. Kabsch, personal communication). A total of 111,248 measurements were made of 11,296 unique reflections to 4.5 A resolution. The data were greater than 997; complete to 4.5 A. IZmergcfor data with intensities greater than 3 sigma to 4.5 A was 1@7%. Intensity data for the rabbit cardiac tropomyosin crystals were collected to 7.8 A at 17 “C using a Xentronics area detector on an Elliott GX-13 rotating anode generator using Franks double focusing mirrors. The frames were in the range of 0.2 to 625”. The data were processed using BUDDHA (Blum et nl., 1987) and scaled using the CCP4 package. The data were greater than 99% complet,e and the Rmarge to 9 A resolution was 12.6% for 11,650 total reflections comprising 1149 unique reflections. Rmerge for data with intensities greater than 3 sigma was 10% Samples for electron microscopy were prepared in conventional negative stain or in vitrified ice as previously described (Cabral-Lilly et of., 1991). Micrographs were recorded using a Philips EM420 electron microscope operated at 100 kV. The method of least-squares refinement of the model follows that described previously (Phillips et d., 1986). The coiled coil was represented as strings of point scatterers placed at small enough intervals (2.5 A) along each of the paths of the alpha-helices to approximate a uniform line of densit,y at low resolution. The positions and weights of these point, scatterers were determined by various refinable parameters. The path of the molecule t Abbreviations used: PEG 1000, polyethylene glycoi 1000; kb. 1 x IO3 base-pairs; EM, electron microscopy.

Resolution

443

was expressed as a series of sine and cosine terms that determined the displacement of the coiled coil of each of the molecules along orthogonal directions perpendicular to the overall direction of run of the molecular strands. Isotropic temperature factors and scattering strengths

were alsoplacedin serialform, sothat theseaspectsof the molecular structure could be refined. To obtain the predicted amino acid sequence of the

alpha subunit of porcine cardiac tropomyosin, a commercial porcine cardiac lambda gtl0 library (PL1004a obtained from Clontech, Palo Alto, CA) was screened using a 1.2 kb chicken alpha-tropomyosin clone (Gooding et al., 1987) j2P-labeled using the random oligonucleotide priming method. A 1 kb EcoRI positive transcript was subcloned into bacteriophage Ml3 mp18 and mp19 (Messing, 1983) and sequenced using the dideoxy method of Sanger et al. (1977) with 3 synthetic oligonucleotide primers (designed from partial sequence data and synthesized by Jan Fogg and Terry Smith of the MRC Laboratory of Molecular Biology) in addition to universal primer. Both the coding and non-coding strands were sequenced and full nucleotide and predicted amino acid sequences have been deposited with the EMBLI data base under accession number X66274. All microbiological manipulations were as described by Maniatis et nl. (1982).

3. Results (a) X-ray crystallography The overall distribution of intensity in the X-ray diffraction pattern of spermine-induced tropomyosin crystals is unusual in that it is concentrated in defined regions with only weak spots in between. The intensity distribution closely resembles that seenin fiber diffraction patterns, indicating that the molecules in the crystals are all aligned roughly parallel to one another. Thus, there is a strong concentration of density in the 10 A equatorial region and a strong meridional arc of density in the 5 b region. Between these two concentrations of density the pattern is remarkably weak. One view of the tropomyosin crystalline diffraction pattern perpendicular to the molecular axes (Fig. l(b)) is remarkably similar to the fiber diffraction pattern from hydrated molluscan catch muscle (Fig. 1(a)). The diffraction from these muscles is dominated by scattering from coiled-coil paramyosin molecules aligned parallel to the muscle fiber axis (seeCohen & Holmes, 1963; Fraser & MacRae, 1973). This correspondence confirms that both paramyosin in muscles and tropomyosin in crystals comprise strongly axially oriented, alpha-helical coiled coils. The concentrations of intensity correspond broadly to different levels of structure in the crystals: the data out to about 9 ip are dominated by the overall shape of the molecules: their packing in the crystals and the conformation of the coiled coil. The next concentration of intensity, in the 5 A region, is dominated by the internal structure of the helices (especially the track of their backbone) and the amino acid sequence of tropomyosin. This remarkably distinct demarcation between different structural levels indicated that our structural investigation of the crystals should proceed in analogous stages. Accordingly. we have (*omen-

P’. (2. Whitby

444

(a)

et al.

(b)

Figure 1. On the left (a) is a fiber diagram from hydrated molluscan “catch” muscle containing large amounts of thr cGlrd-coil protein paramyosin (Cohen & Holmes. 1963). Kotr the 5.1 A meridional arc and strong near-equatorial layer line related to the Pitch of the coiled coil. On the right (b) is a,n unscreened 6.5” precession photograph of porcine car&c tropomyosin near the (h01) zone. trated first on establishing the structure to a nominal resolution of 9 A (using molecular replacement and highly parameterized refinement together with electron microscopy) and determining the precise amino acid sequence of the molecule before proceeding to use isomorphous replacement, to analyze t,he higher resolution data. The general paths of the molecules were previously inferred from the positions of characteristic 5.1 a coiled-coil reflections seen in still and precession film photographs (Phillips et al., 1987). These and other features are clearly seen in higher resolution pseudo-precession plots of area-detector data (Fig. 2). Narrow meridional arcs are present in the (h01) plane of reciprocal space, and a pair of arcs can be seen in the (OkZ) plane. Tropomyosin polymerizes head-to-tail to produce continuous, head-totail filaments. This property fixes the molecular directions to be integer multiples of the unit cell and the positions of the arcs determines the directions to be the (332) and the symmetry-related (3-32) direct,ions. Thus, because of the unique properties of the diffraction pattern of the tropomyosin molecule and the fortunate arrangement of the molecules in t,hese crystals, we were able to solve the crystallographic rotation problem without the need for search techniques. The clear presence of a sinusoidal fringe function in the distribution of diffraction intensity in the (h01) plane of reciprocal space (Fig. 2, arrows) further supports the above solution for the structure by indicating that molecules are arranged in layers with an interlayer separation of about 22 A. The space group of the crystals is C2, with a = 259.7 WI

b = 55.3 is, c = 135% A and /? = 97.2”. However, for convenience in modeling, viewing and comparison with the sheet structures seen by electron microscopy (Phillips et al., 1987; Stewart, 1984), a nonstandard alternative C2 unit cell has been chosen with a = 791% A, b = 553 A, c = 277.2 A and /? = 17@8”. Tn this alternative system, the molecules run in the (130) and (l-30) directions, and the molecular layers are parallel to planes containing the (100) and (010) axes. This alternative cell is about 44 A (277.2 a x sine(17080)) thick in the direction perpendicular to the layers. Because space group C2 has a 2-fold axis along the (010) direction, producing two layers per unit cell, this value is in keeping with the minimum in the fringe function at a spacing corresponding to 22 A. Thus, the alt’ernative unit cell contains two layers of molecules, related by the crystallographic 2-fold axis, with one layer containing molecules running parallel to the (130) direction, and the other layer running parallel to the (l-30) direction. All further discussion in the paper will use the alternative, rather than the standard, crystallographic unit cell definition. The translation problem was solved using carbonalpha co-ordinates from the previous low resolution structure determination (Phillips et al., 1986; entry 2TMA in the Protein Data Bank at Brookhaven National Laboratory). These co-ordinates were placed along the crystallographic (130) direction, as previously inferred, and XPLOR (Brunger et al., 1989) was used to search for translation solutions using the correlation coefficient between observed and calculated intensities as the target function. A clear solution was found. At t,his point, the

Structure

of Tropomyosin

(a)

f-2)

at 9 Angstroms

Resolution (b)

(d)

F:igr Ire 2. Pseudo-precession plots of the measured X-ray diffract’ion intensities for the 3 principle crgstallo~ ;raphir plai nes of reciprocal space (a) (Ok& (b) (hOZ), (c) (MO) and (d) t.he calculated intensities for the f/&l) plane. The helical layc?r 1ine structure of the measured pattern (b) is well matched by those calculated from the model (d). The c:urrent mot zlel does not account, however, for the meridional reflections at about 10 -4 resolution or the strong mctridion Ial and equ ate trial features arising from alpha-helices in the 5 A regiou of the (hOZ) pat~trrn.

molecule was broken into four segments of equal length and this model was subjected to rigid-body minimization using XPLOR. The correlation coefficient increased dramatically from 0.4 to 67.5, although the standard crystallographic R-factor was no better than 0.60. On the basis of the high correlaThe tion coefficient, a 2F,- F, map was calculated. map showed not only the molecule included in the phasing model. but a second molecule in the asymmetric unit, running essentially parallel to the first one. Thus the asymmetric unit contains two coiledcoil molecules (4 polypeptide chains) with a total relative molecular mass of 132,000. At this point, alternative directions such as the (120) and (140) were examined as possible solutions of the structure, but rigid-body minimization of these starting struc-

tures badly fragmented the molecule. Thus, the solution seems to be unique. Once the general paths of the molecules in the cell were determined, t,he models were refined by a specially devised reciprocal space least-squares refinement. The method is essentially the one used previously to investigate the Railey “kite” crystals et al., 1986), modified for of tropomyosin (Phillips space group C2 and two molecules per asymmet,ric unit. Each of the long alpha-helices of the coiled coil is modeled as a uniformly dense, curved rod whose path through space is determined by two sets of orthogonal Fourier coefficients. Terms are also present for the radius and azimuthal origin of the coiled coil. temperature factors that vary smoothly This along the molecule and an overall scale factor.

F. G. Whitby et al.

440

Table 2 R-factor as a function of resolution (A)

All data Data > lOa

3OC20

20-15

15-12

12-10

10-9

Overall

0.29 @26

0.41 0.37

0.44 0.38

0.42 038

@46 0.45

0.40 035

Table 3 Re$nement results Molecule Average radius of coiled coil Amplitude of lateral displacement the layers Amplitude of displacement perpendicular to the layers Average B-factor (not based on atomic movements)

I

Molecule

51 A

54A

98A

87 A

2.7 A

@9A

990 A2

1240 a2

2

in

model is satisfactory for investigating the overall shape of the molecule and its coiled-coil pitch, but will obviously break down once the internal structure of the helices and the arrangement of sidechains begins to dominate the intensity distribution. Since the data to 9 A resolution contain only small contributions from the protein sequence and the internal structure of the alpha-helices, this model was considered adequate to this resolution, but would clearly not be so when investigating the higher resolution data. Refinement of the structure using a model with constant scattering power along the molecule yielded a crystallographic R-factor of 94 to 9 A resolution using all the data and ti35 using data with I > lOa (Table 2). A comparison of the refined values for some of the properties of the two molecules in the asymmetric unit is given in Table 3. The handedness of the structure was fixed by setting the coiled coil to be left-handed, as it must be if formed from right-handed alpha-helices (Crick, 1953). The curvature of the molecule is different from that in the previously studied Bailey crystal form, and is better described as sinusoidal than helical (Fig. 3). The refined radii of the coiled coils are slightly smaller than the values obtained from the solution of the Bailey form (5% A), but it is not clear what the error limits on these refined numbers are. The atomic structure of the GCN4 fragment (O’Shea et al., 1991) confirms that these values for the radii are reasonable. The R-factor for the strongest one-half of the data (those with I > 100) was 035. Considering the localization of intensities in the diffraction pattern, this number suggeststhat the fit is quite reasonable for regions of reciprocal space that contain the largest contribution from the coiled-coil transform. A good fit in other regions of reciprocal space will require the inclusion of amino acid side-chain and alpha-helix stereochemical information in the molecular model. The 2F,- F, electron density map at 9 A resolu-

Figure 3. A 2F,-F, map showing the molecular packing within 1 layer of the structure. The asymmetric unit contains 2 molecularfilaments.

tion clearly shows the paths of the molecules in the layers and the two molecules in the asymmetric unit (Figs 3 and 4). The map also shows clear regions of coiled-coil structure (Fig. 4), although there are other small regions that are not yet well defined. The separation between the molecular filaments

Figure 4. Diagram showing the packing of the molecular filaments in the lattice. A crystallographic 2-fold axis relates the top layer (white) to the bottom layer (violet) and runs through the middle of the structure in the horizontal direction. The 420 A-long molecules makenumerouscontacts betweenthe layers, but contacts within a layer appear to be restricted mainly to areas where the 2 coiled coils come close to one another, roughly every 70 A.

Structure

of Tropomyosin

at 9 ~rq.strom.s

Resolution

447

Figure 5. Stereo pair of the 2F,- Fe map of a part of 1 layer, showing clearly the coiled-coil structure of the filaments. The map suggests that the coiled coils may not be perfectly regular, but vary somewhatalong their length.

within one layer is 25 A, whereas the separation between layers is 22 A. The diameter of the coiled coil is about 22 A at the widest point, so that tight contacts are made between the layers, and these layers are criss-crossed by the crystallographic 2-fold operator parallel to the b axis. The electron density map shows that adjacent molecules within a layer are in close contact over part of their length but separated by a few angstroms over the remainder. Each molecule is associated more closely with one neighbor for about half of one molecular length, and more closely associated with the neighbor on the opposite side for the other half. Although the relative phases of the coiled coils of adjacent molecules within layers vary because of the sinusoidal paths of the molecules, the relative stagger is roughly a quarter pitch for the neighbor on one side and an eighth of a pitch for the neighbor on the other side. In directions perpendicular to the layers, the staggering is not simple to define because of the angles at which the molecules cross (Fig. 5). Even at a resolution of 9 A, the pitch of the coiled coil can be seen to vary along the molecule. Although a detailed analysis must await the higher resolution structure, clearly the distribution of amino acid side-chains has an influence on the parameters of the coiled coil. (b) Electron

microscopy

Electron microscopy has provided additional information to complement the results from X-ray

crystallography. Electron micrographs of spermineinduced tropomyosin crystals have been studied to reveal at least two closely related forms, neither of which is identical with the form examined by X-ray crystallography. Although we have examined thin crystals from the same vial as large, X-ray quality crystals, all the forms thin enough for transmission electron microscopy showed an 800 L! centered repeat (see Table 1). Attempts to crush large single crystals also resulted in sheets with an 800 A centered lattice. Thus far, it has not been possible to get thin crystals for electron microscopy that match the unit cell size and symmetry of the form examined by X-ray diffraction. There is, however, a basic packing relationship among all these forms that can be related to the crystals examined by X-ray diffraction. Stewart (1984) reported the formation of crystalline sheets of tropomyosin by titration with spermine as a counter ion, and produced a Fourier filtered image showing a centered lattice with unit cell dimensions of about 800 A x 50 A and a striking singlet-doublet-singlet band pattern when viewed in negative stain (Table 1, Form I). More recently, a number of the images from this form have been obtained that exhibit “fringes” at the ends of the crystalline sheets that clearly reveal the positions of the molecular ends in the structure. In contrast to the original model (Stewart, 1984), we have shown that the molecular ends are located at each bright stain-excluding band. This kind of pattern can be generated by packing four layers of

P’. G.

et al.

14Thitby

crystals. The molecular ends within layers of related crystal forms induced by spermine can be determined from electron micrographs to have a separation of 130 A. Since the molecular filaments in the asymmetric unit of the X-ray crystal form have features in 2&F, maps that appear to have discontinuities with a similar separation, we have assigned these features to be the locations of the head-to-tail joining regions. We have also begun an examination on Form I by cryo-electron microscopy. In these images the molecular ends are much less pronounced than in negative staining, but the curvature of the molecules in projection can be deduced from the images at about 18 a resolution. Using co-ordinates derived from the molecular filaments in the layers determined by X-ray crystallography, a satisfactory simulation of the images is obtained. (c) Amino acid sequenceof porcine cardiac tropomyosin

filaments with each layer similar to that in the alternative unit cell of the X-ray form. but dif?erently disposed. The ends of the two anti-parallel molecules in the asymmetric unit) in the layer are staggered 1)s about one third of the molecular length. Not.e t.hat this kind of molecular arrangement in a layer is similar to t,ha,t seen in t.he common divalent, cation paracrystal of tropomyosin (Cohen & Longley, 1966; Cohen et al.. 1971: Stewart.. 1981). Another frequent’ form is characterized by pairs of bright stain-excluding bands spaced at about 400 A (Table 1. Form TI). This form may also be generated from a packing of four layers as described above, but with a different, shift between successive layers. (Some variants of Forms 1 and II may also be accounted for by other shifts of the layers). The molecules of Form I probably run in an anti-parallel manner. in keeping with t,hrir arrangement in paracrystals induced by divalent cations. On this hasis, thta molecules in t,he as:mmet.ric unit in the Y-ray crystal would also run in an arrti-

parallel manner The locations of the molecular ends in the layers cannot yet be determined with certainty from the X-ray data alone, but have been tentatively assigned based on the images of negatively stained

To confirm the general similarity between rabbit and porcine tropomyosin (and so validate our assumption that their structures could be used interchangeably at least to 9 a resolution) and as a prerequisite for extending the resolution of our study to higher resolution. we determined that primary structure of porcine cardiac alpha tropmJ,osin. R,at her than sequence the prot.ein tlir~~c+Jy, we instead sequenceda cI)NA cnloneobtained from ;I porcine cardiac lambda gt 10 library probed with randomly primed chicken alpha-tropomyosin cal)?u‘A ((iooding ef al.. 1987). Thr? predict,ed arnino acid sequence differs from rabbit alpha tropomyusin at onl? four positions out of 284: and includes substitut.lon of aspartic acid for glutamic acid at position 23. glutamic. acid for glutamine at position 24. alanine for glycine at position 52 and phenylalanine for valine at position 95. Except’ for the inclusion of a slightly la,rger hydrophobic residue a,t a hydrophobic core position. these substitutions are on the surface. Purthcrmore, all of them are germall\: c%onservativeand one would not expect any signif;cant differences in the three-dimensional structures of rabbit and porcine alpha tropomyosins. The similarity of the rabbit and porcine molecules is horn out by the similarities of their diffraction patterns. besides the sequence similarities. The dat.a from crystals of porcine cardiac, alpha t~ropom~osin. rabbit1 skeletal alpha tropotnyosin a.nd rahhit csrdisc

tropomyosin

are

indist~inguishahle

at

!4 AA

resolution.

4. Discussion (a) Molecular struct we Although it is premature to interpret th(* fine details of this electron density map precisely, these cryst,als allow an accurate determination of the value and the regularity of the pitch of the coiled coil in the tropomyosin molecule. Much discussion

Structure

of Tropomyosin at 9 Angstroms Resolution

has been offered on the “regular” attachment of repeating motifs of tropomyosin to actin, requiring an average coiled-coil pitch of 137 a (Parry, 1975; Stewa.rt & McLachlan, 1975; Phillips et al., 1986). Our maps support this model strongly and continue to be consistent with an average pitch close to 140( + 10) .k for tropomyosin. The coiled coils of myosin (Quinlan & Stewart, 1987), paramyosin (Elliott et al., 1968) and mantis egg ootheca protein (Rullough & Tulloch, 1990) seen to have a very similar value. although initially reported to have a pitch of 180 i-r (O’Shea et al., 1991). the GCN4 leucine zipper protein also has a pit’ch of about 140 -4 (Phillips, 1992; J. Sew & C. Cohen. unpublished results). Resides establishing that the average value of the tropomyosin coiled-coil pitch was very close to 140 A, this study has also identified a variability in the coiled-coil pitch as one moves along the molecule. It. is quit.e clear from Figure 3 that the local value of the pitch varies and probably reflects the influence of the local amino acid sequence. In addition to changing the pitch of the coiled coil. it appears likely that the radius of the coiled coil varies along the molecule, probably to accommodate the different sizes of residues in positions a and n of the coiled coil that form the hydrophobic seam. This variability in both the pitch and radius of the coiled coil is probably a general property of coiled coils and may account for the rather broad and diffuse character of those layer line reflections (most especially the layer line at about 70 A) that are associated with the coiled coil. In fiber patterns. such as that shown in Figure 1, these reflections are broadened substantially in the 2 direction as would be expected if the coiled-coil helical tracks were perturbed. The close correspondence of both t)he amino acid sequence and the diffraction patt’erns of the porcine and rabbit alpha tropomyosins is in keeping with the observed homology amongst all vertebrate tropomyosins (Lees-Miller & Helfman: 1991). Knowing the primary structure of porcine tropomyosin will now allow model building to proceed wit’h the correct amino acid sequence and will allow correlation of the irregularities of t,he pitch with part.icular amino ac*ids. (i)) Crystal

structure

Although the unit cell contains about SSyA water by volume, the molecules are very tightly packed. Within each layer the filaments are separated by 25 .A, only slightly greater than the widt,h of a coiled coil (see Fig. 4). The sepa.ration between layers is even less, 22 A, and most of the contacts between molecules occur between t’he layers. There are three different types of interaction between tropomyosin coiled coils in spermine-induced crystals: those between the two molecules in the asymmetric unit; those between molecules in adjacent asymmetric units in t’he plane of the layers; and those between molecules in adjacent layers. These interactions

449

may give clues to general features of the interactions between coiled coils that could have implications for their assembly into macromolecular aggregates such as muscle thick filaments and intermediate filaments. Although the precise geometry of interaction between molecules in the plane of a layer will vary slightly because they follow a roughly sinusoidal path, there are some general features that are apparent, and that give some clues to the interactions involved. The geometry of the interaction between molecules in adjacent asymmetric units in the plane of a layer (Fig. 3) appears to be broadly similar to those seen in many other assemblies where the coiled coils are staggered by roughly an odd number of quarter pitches. resulting in “node to anti-node” packing. Model building studies (Rudall, 1956: Longley. 1975) have indicated that such a form of packing allows the closest contact between adjacent coiled coils if they are considered as being made of t’wo uniform cylinders twist,ed round one another. Direct visualization of such a geometry has been obtained, for example, with myosin subfragment-2 crystals (Quinlan & Stewart. 1987) and mantis egg shell protein (Bullough & Tulloeh. 1990). In the case of the tropomyosin crystals, however, the spacing between neighboring molecules within a layer is somewhat variable. Some contacts are very closr to a stagger of one quarter of a pitch, but others are noticeably less (Fig. 3). The curvature of the filaments in the layers precludes regular contacts. An explanation for the curvature of the tropomyosin molecules within layers must await t)he higher resolution structure. Since the curvature we have observed for the spermine crystals is very different. from t.hat observed in the Bailey crystals (Phillips et al., 1986), it seems likely that this curvature is not an intrinsic property of the molecule but instead result,s from its deformat,ion by the forces of its interaction with its neighbors. The flexural rigidity of coiled coils is relatively low (Hvidt rt CLI., 1982. 1983: Stewart. et al., 1987; G. X. Phillips and S. Chacko. unpublished results) so that only comparatively small forces would be required to achieve the degree of bending observed in Figure 3. The alphahehces from adjacent molecules would be in contact only for comparatively short’ stretches or zones once every half-pitch. If the coiled-coil pitch is not an int’egral multiple of alpha-helical pitches (137 a is, for example, roughly 253 alpha-helix pitches of 5.4 x), the phases of the interacting alpha-helices will gradually change from one zone to the next. One way to maintain the phase of the int)eraction would be to translate one molecule slightly relative to its neighbor. Such a relative translation caould be achieved if t)he molecules were to follow a curved path: the molecule at the higher radius would then be retarded slightly relative to the one at lower radius. To prevent the molecules following a circular path. all that would be required would hr for the direction of curvature to be reversed periodically. and so such a mechanism would a(.colmt for the

450

F. G. Whitby

roughly sinusoidal path followed by the molecules. Since the molecules in the myosin subfragmentcrystals are also curved, it appears likely that the generation of some molecular bending may be a general feature of interactions between coiled coils. The geometry of interaction between coiled coils in different layers varies considerably as a result of both the angle at which the individual molecules cross and their curved profile in the plane of the layers. Essentially all relative staggers of coiled coils are observed, which indicates that the geometry of coiled-coil interaction is of secondary importance in this molecular interaction. Because of the curvature of the filaments and the roughly 22” rotation of the molecular axis in one layer relative to its neighbor, the distance along an individual molecule of contact points between it and molecules in adjacent layers varies somewhat, but several are close to 58 A, the spacing of alternate zones of negatively charged residues (McLachlan & Stewart, 1976; Phillips et al., 1986). This observation suggests that the interactions between the layers may derive primarily from positively charged spermine molecules bridging between zones of negative charge on interacting molecules. Because of the size of spermine. it is likely that it can bridge a considerable distance, and so the precise azimuthal geometry of the two interacting coils is less important than when other types of contact (such as those between the molecules within a layer) are made. Tt is likely that spermine facilitates the formation of these sheets because of its size compared with that of, for example, magnesium ions. Clearly there will be a number of possible translations of one layer relative to its neighbor that will enable bridging of charged zones, and so this form of interaction would offer at least a partial explanation for the series of polymorphic forms observed by electron microscopy in which the layers take up different relative positions in forming sheets. (c) Summary Tropomyosin is well known for its polymorphic behavior upon aggregation. Taken together, the X-ray diffraction and electron microscopy results indicate that there is a close relationship between all the forms of tropomyosin produced by spermine. The study also emphasizesthe need for co-ordinated X-ray crystallography and electron microscopy for the analysis of large molecular complexes at this resolution. It seemsplausible that all of the densely packed forms observed thus far are variations on the packing of closely related layer structures with two anti-parallel molecules staggered by approximately 130 A. The well-known divalent cationinduced paracrystals appear to arise from this pairing of staggered filaments (Cohen & Longley, 1966; Stewart, 1981). A detailed analysis of the electron microscopic forms and their correlation with the form studied by X-ray crystallography will be described subsequently (C. Cohen et al., unpublished results).

et al.

Our 9 A resolution map of spermine-induced crystals of tropomyosin has enabled us to obtain information on the molecular shape, molecular interactions between coiled coils and the mechanism by which a range of polymorphic forms can be produced. The coiled coil has beeen seen clearly and has enabled not only its pitch to be established reliably, but has also revealed perturbations in its regularity. Moreover, the molecular interaction geometry in the paracrystals has given clues to features that may have implications for interactions between coiled-coil proteins in macromolecular assemblies formed in viwo, such as muscle thick filaments and intermediate filaments. The distortion of the molecules into a curved path by molecular interactions such as those we have observed in the t,ropomyosin crystals is probably a general feature of coiled-coil interactions and the prospect of having bent molecules needs to be considered in other models of coiled-coil filament structure. Such curvature may, for example, account for the irregularities seen in naturally occurring paramyosin crystalline arrays in molluscan thick filaments (Elliott & Bennett, 1982) or in vertebrate skeletal muscle thick filaments (Stewart & Kensler, 1986). Solving the crystal to this structural level has given us important new insights into the structural biology of tropomyosin and related molecules, but) we clearly need to advance to higher resolution t,o understand the disposition of side-chains and helix order more fully. To advance to this next level of tropomyosin structure it will be necessary to supplement our modelling work with a measure of empirical phase information, such as that provided by a heavy atom derivative and/or non-crystallographic averaging and solvent flattening. Given the constraints of the coiled-coil architecture and thr knowledge of the preferred geometry of side-chains in the core, as seen in the gcn4 leucine zipper structure, it then should be possible to obtain a refined, near-atomic model of tropomyosin. This work was supported by NIH grants AR32764 and AR17346, the Robert A. Welch Foundation, the W:. M. Keck Center for Computational Biology and the Muscular Dystrophy Assobiation. F.G.W. is supported by an NIH Houston Area Molecular Biophysics training grant. We thank Steve Harrison and Don Wiley at. Harvard University for use of their Xentronics area detector system, Florante Quiocho at Baylor College of Medicine for use of his area detector for initial data sets and Augustin Avila-Sakar, Mike Schmid and Wah Chiu at Baylor College of Medicine for additional electron micrographs.

References Astbury, W. T. (1947). On the structure of biological fibres and the problem of muscle. Proc. Roy. SOC.ser. B, 134, 303. Beck, K., Hunter, I. & Engel, J. (1990). Structure and function of laminin: anatomy of a multidomain glycoprotein. FASEB J. 4, 148-160. Blum, M., Metcalf, P., Harrison, S. C. & Wiley, D. C. (1987). A system for collection and on-line integra-

Structure

of Tropomyosin

tion of X-ray diffraction data from a multiwire area detector. J. Appl. Crystallogr. 20, 235-242. Brunger, A., Karplus, M. & Petsko, G. A. (1989). Crystallographic refinement by simulated annealing: application to crambin. Acta Crystallogr. sect. A, 45, 5661. Bullough, P. A. & Tulloch, P. A. (1990). High-resolution spot-scan electron microscopy of microcrystals of an alpha-helical coiled-coil protein. J. Mol. Biol. 215, 161-173. Cabral-Lilly, D.. Phillips, G. N,, Jr, Sosinsky, G. E., Melanson, L., Chacko, S. t Cohen, C. (1991). Structural studies of tropomyosin by cryoelectron microscopy and X-ray diffraction. Biophys. J. 59, 8055814. Cohen, C. & Holmes, K. C. (1963). X-ray diffraction evidence for alpha-helical coiled coils in native muscle. J. Mol. Biol. 6, 423-432. Cohen. C. & Longley, W. (1966). Tropomyosin paracrystals formed by divalent cations. Science, 152, 794. Cohen, (1. & Parry, D. A. D. (1990). Alpha-helical coiledcoils and bundles: how to design an alpha-helical protein. Proteins Struct. Funct. Genet. 7. l-15. Cohen, C., Caspar, D. L. I)., Parry, D. A. D. & Lucas, R. M. (1971). Tropomyosin crystal dynamics. Cold Spring Harbor Symp. Quant. Biol. 36, 205-216. Crick, F. H. C. (1953). The packing of alpha helices: simple coiled coils, Acta Crystullogr. 6, 689-697. Eisenberg, E. & Kielly, W. W. (1974). Troponintropomyosin complex. J. Biol. Chem. 249, 4742-4748. Elliott, A. &z Bennett, P. (1982). Structure of the thick filaments in mollusean adduetor muscle. In Basic Biology of Muscles: A Comparative Approach (Twarog, B. M., Levine, R. J. C. & Dewey, M. M., eds). pp. 11-28, Raven Press, New York. Elliott, A., Lowry. J., Parry, D. A. D. & Vibert, P. J. (1968). Puzzle of the coiled coils in the alpha-protein paramyosin. Nature (London), 218, 656-659. Fraser, R. D. B. & MacRae, T. P. (1973). Conformation oj Fibrous Proteins. Academic Press, New York. Gooding, C., Reinach, F. C. & Macleod, A. R. (1987). Complete nucleotide sequence of the fast-twitch isoform of chicken skeletal muscle alpha tropomyosin. Nucl. Acids Res. 15, 8105. Haselgrove, J. C. (1972). X-ray evidence for a conformational change in actin-containing filaments of vertebrate striated muscle. Cold Spring Harbor Symp. Quant. Biol. 37, 341-352. Hitchcock-Degregori, S. E. 6 Varnell, T. A. (1990). Tropomyosin has discrete a&in-binding sites with sevenfold and fourteenfold periodicities. J. Mol. Biol. 214, 885-896. Huxley, H. E. (1972). Structural changes in the actin- and myosin-containing filaments during contraction. Cold Spring Harbor Symp. Quant. Biol. 37, 361-376. Hvidt, S., Nestler, H. M., Greaser, M. L. & Ferry, J. D. (1982). Flexibility of myosin rod determined from dilute solution viscoelastic measurements. Biochemistry, 21, 40644073. Hvidt, S., Ferry, J. D., Roelke, D. L. & Greaser, M. L. (1983). Flexibility of light meromyosin and other coiled-coil alpha-helical proteins. Macromolecules, 16, 740-745. Kabsch, W. (1988). Evaluation of single-crystal X-ray diffraction data from a position-sensitive detector. J. Appl. Cystallogr. 21, 916924. Kress, M., Huxley, H. E., Faruqi, A. R. L Hendrix, J. (1986). Structural changes during activation of frog

at 9 Angstroms

Resolution

451

muscle studied by time-resolved X-rav diffraction. J. Mol. Biol. 188, 325. Lees-Miller, J. P. & Helfman, D. M. (1991). The molecular basis for tropomyosin isoform diversity. Bioessays, 13, 429-437. JAongley, W. (1975). The packing of double heliees. J. &Zol. Biol. 93, 11 l-l 15. Lupas, A.. Van Dyke, M. & Stock, J. (1991). Predicting coiled coils from protein sequences. Science, 252, 116221164. Maniatis, T., Fritsch, E. F. & Sambrook, J. (1982). Molecular Cloning: A Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. McLachlan, A. D. & Karn, J. (1983). Periodic features in the amino acid sequence of nematode myosin rod. J. Mol. Biol. 164, 605-626. McLachlan, A. D. & Stewart, M. (1975). Tropomyosin coiled-coil interactions: evidence for an unstaggered structure. J. Mol. Biol. 98, 293-304. McLachlan, A. D. & Stewart, M. (1976). The 14-fold periodicity in alpha tropomyosin and the interaction with actin. J. Mol. Biol. 103, 271-298. Messing, J. (1983). New Ml3 vectors for cloning. Methods Enzymol. 101, 20-78. O’Shea, E. K., Klemm, J. D., Kim, P. 6. 8r Alber, T. (1991). X-ray structure of the GCN4 leucine zipper, a two-stranded, parallel coiled coil. Science, 254, 539-544. Parry, D. A. D. (1975). Double helix of tropomyosin. Nature (London), 256, 346-347. Parry, D. A. D. & Squire, J. M. (1973). Structural role of tropomyosin in muscle regulation. 1. Mol. Biol. 75, 33-55. Phillips, G. N., Jr (1985). Crystallization in capillary tubes. Methods Enzymol. 114, 128-131. Phillips, G. N., *Jr (1992). What is the pitch of the alphahelical coiled coil? Proteins: Struct. Funct. Genet. In the press. Phillips, G. N., Jr, Lattman, E. E., Cummins, P., Lee, K. Y. & Cohen, C. (1979). Crystal structure and molecular interactions of tropomyosin. Nature (London), 278, 413-417. Phillips, G. N., Jr, Fillers, J. P. & Cohen, C. (1986). Tropomyosin crystal structure and muscle regulation. J. Mol. BioE. 192, 111-131. Phillips, G. N., Jr, Cohen, C. & Stewart, M. (1987). A new crystal form of tropomyosin. J. Mot Riot 195, 219-223. Quinlan, R. A. & Stewart, M. (1987). Crystalline tubes of myosin subfragmentshowing the coiled-coil and molecular interaction geometry. J. CPU Biol. 105, 403-415. Rudall, K. M. (1956). Proteins, ribbons and sheets. Lectures Sci. Basis Med. 5, 217-230. Sanger, F., Nicklen, S. & Coulson, A. R. (1977). DNA sequencing with chain-terminating inhibitors. Proc. Nat. Acad. Sci. U.S.A. 82, 5463-5467. Stewart, M. (1981). Structure of alpha tropomyosin magnesium paracrystals. J. Mot. BioZ. 148, 41 I425. Stewart, M. (1984). Crystalline sheets of tropomyosin. J. Mol. Biol. 174, 231-238. Stewart, M. (1990). Intermediate filaments: structure, assembly and molecular interactions. Curr. Opin. Cell Biol. 2, 91-100. Stewart, M. & Kensler, R. W. (1986). Arrangement of myosin heads in relaxed thick filaments from frog skeletal muscle. J. Mol. Biol. 192, 831-851.

F. G. Whitby Stewart, M. & McLachlan, A. I). (1975). Fourteen actinbinding sites on tropomyosin? Nature (London) I 257. 331-333. Stewart. M., McLachlan. A. II. & Calladinr. (:. R. (l!#S). :i model Taoaccount for the elastic element in must+ crossbridge in terms of a bending myosin rod. Proc. Rvy. Aw. ser. B, 229, 381-m413. St,onr, I).. Sodek. .J.. .Johnson, I’. & Smillir. L. H. (197.7).

Edited

et al.

Tropomyosin: correlation of amino acid sequence and structure. Proc. I?i FEBS Meeting (Budapvt). vol. 31. pp. 125-136. North Holland Publishing (lo.. Amsterdam. Wilson, T. A., Bkehel. ,J.