Association of Structural Repeats in the α-Actinin Rod Domain. Alignment of Inter-subunit Interactions

J. Mol. Biol. (1995) 252, 227–234

Association of Structural Repeats in the a-Actinin Rod Domain. Alignment of Inter-subunit Interactions G. Flood1, E. Kahana2, A. P. Gilmore1, A. J. Rowe1, W. B. Gratzer2 and D. R. Critchley1* 1

Department of Biochemistry University of Leicester University Road, Leicester LE1 7RH, UK

2

MRC Muscle and Cell Motility Unit, King’s College London, 26-29 Drury Lane London, WC2B 5RL, UK

Fragments of the rod domain of chicken a-actinin, which comprises four spectrin-like repeat sequences, have been prepared by expression in Escherichia coli. Electron microscopy reveals that all products containing three or four complete repeats are rod-like. Self-association of fragments was detected by chemical cross-linking and analytical equilibrium sedimentation. The intact rod domain forms a stable dimer, which does not dissociate measurably in the accessible concentration range. Elimination of either terminal repeat (repeat 1 or repeat 4) greatly diminishes the extent of dimerisation. The fragment comprising repeats 1-3 dimerises appreciably, with an association constant estimated from the sedimentation equilibrium distribution of approximately 5 x 105 M−1. The fragment made up of repeats 2-4 dimerises to a small extent, but also forms aggregates at high concentrations. The results are most easily reconciled with an aligned structure for the rod domain in solution, in which repeat 1 associates with repeat 4 of the partnering chain, and repeat 2 with repeat 3, rather than with a staggered structure, in which one of the terminal repeats does not participate in dimerisation. Possible explanations for the apparent difference observed between the a-actinin rod structure in solution and in two-dimensional crystalline arrays are examined. 7 1995 Academic Press Limited

*Corresponding author

Keywords: a-actinin; spectrin-like repeats; dimer formation

Introduction Structural repeats of the type first identified in erythroid spectrin (Speicher & Marchesi, 1984) are a feature of the group of cytoskeletal proteins that includes non-erythroid spectrins (such as neuronal spectrin, or fodrin), dystrophin (Davison & Critchley 1988), utrophin (Tinsley et al., 1992) and the muscle and cytoplasmic a-actinins (reviewed by Blanchard et al., 1989). In spectrin most of the repeats contain 106 residues, while those in other members of this family of proteins are, in general, somewhat longer. Structural predictions indicated that each repeat contains three segments of a-helix in the form of a three-stranded coiled-coil (Speicher & Marchesi, 1984), and more detailed analyses suggested a repeating pattern in which one long helix is divided Present address: A. P. Gilmore, Department of Cell Biology and Anatomy, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-7090 USA. Abbreviations used: GST, glutathione S-transferase; DMS, dimethylsuberimidate; PCR, polymerase chain reaction; SEM, standard error of estimate of mean value. 0022–2836/95/370227–08 $12.00/0

between two adjoining repeats (Cross et al., 1990; Koenig & Kunkel, 1990; Parry & Cohen, 1991). The crystal structure of a single repeat of Drosophila spectrin (Yan et al., 1993) has confirmed the accuracy of the prediction. The boundaries of the repeats, which constitute the minimum folding unit, have been established for spectrin (Winograd et al., 1991), dystrophin (Kahana et al., 1994) and chicken gizzard a-actinin (Gilmore et al., 1994). The quaternary structure of dystrophin and utrophin has still to be established, but the other proteins of the spectrin class all form rod-like, or at least highly elongated, antiparallel dimers. In the spectrins these are (ab) heterodimers and in the a-actinins homodimers. Since the isolated a-actinin rod domain exists in solution as a highly stable dimer (Imamura et al., 1988; Kahana & Gratzer, 1991), and defined complementary repeats of the erythroid spectrin a and b-chains also form a stable 1:1 complex (Speicher et al., 1992), it is implied that the structural repeating elements interact specifically by pairs along the rod. Thus, in the case of the a-actinin rod, which consists of four repeats, the antiparallel dimer should be stabilised by interactions between repeats 7 1995 Academic Press Limited

228

Association of a-Actinin Structural Repeats

Figure 2. Chicken gizzard a-actinin repeat domain constructs expressed in E. coli. The amino acid residues which define the N-terminal boundary of each repeat are indicated (Gilmore et al., 1994). Successive structural repeats begin and end near the centre of the long continuous a-helix that is partitioned between any two contiguous structural repeats. Thus, the N-terminal helical segment in the three-helix bundle is helix B in the terminology of Parry & Cohen (1991), or helix A in the terminology of Yan et al. (1993). Figure 1. Aligned and staggered models for the a-actinin homodimer. a-Actinin contains an N-terminal actin-binding domain (ABD), four spectrin-like repeats, and two C-terminal EF-hand calcium binding motifs (Baron et al., 1987). A, Aligned model in which there are four pairwise interactions between repeats. B, Staggered model which predicts only three pairwise interactions, namely between repeats 2 and 4, and between the aligned repeat 3 in each subunit. C, Alternative staggered model which predicts an interaction between repeats 1 and 3, and between the aligned repeat 2 in each subunit.

1 and 4 and between 2 and 3. However, on the basis of image reconstructions from two-dimensional crystals Taylor & Taylor (1993) concluded that the chicken gizzard a-actinin structure is staggered, so that either repeat 1 or repeat 4 is free (or may interact with a terminal globular domain of the other subunit). Dimerisation of the rod would then result from only three pairwise interactions, namely repeats 2, 3 and 4 of one chain with 4, 3 and 2 of the other, or alternatively repeats 1, 2 and 3 of the first chain with 3, 2 and 1 of its partner (Figure 1). We have used deletion mutants of the a-actinin rod domain in an endeavour to discriminate between the aligned and staggered models.

Results The a-actinin rod-domain constructs that we have used in this study are shown in Figure 2. The positions of the N and C termini were selected to ensure that the sequence encompassed an integral number of homologous repeats, based on the phasing established earlier, which corresponded within a few residues to that observed in spectrin and dystrophin (Gilmore et al., 1994). The repeats have been shown to constitute independently folding units (Gilmore et al., 1994). a-Actinin polypeptides spanning repeats 1–4, repeats 1–3 and repeats 2–4 were expressed in Escherichia coli as soluble glutathione S-transferase (GST)-fusion proteins, purified on glutathione-agarose beads, and liberated

from GST by proteolytic cleavage with either thrombin or factor Xa. The polypeptides isolated in this way were >90% pure as analysed by SDS-PAGE (Figure 3 and Figure 4, lanes a, e and i) and the CD spectrum of the polypeptides was characteristic of proteins with a high a-helix content (Table 1) and similar to that of the intact rod domain liberated from native a-actinin by proteolysis (Imamura et al., 1988). The bacterially expressed rod domain polypeptides were also relatively resistant to degradation by trypsin (Figure 3), a characteristic of the intact repeat domain liberated from native a-actinin (Imamura et al., 1988; Davison & Critchley 1988; Gilmore et al., 1994). These results suggest that the bulk of the expressed a-actinin polypeptides are correctly folded. On reaction with the bifunctional reagent dimethylsuberimidate, the complete rod domain (repeat 1–4) showed extensive cross-linking to a doublet, migrating in the region expected for dimers (Figure 4). The proportion of covalent dimer was essentially unchanged by a tenfold increase in protein concentration (Figure 4, lanes b to d, and Table 1), and was the same within error as that formed under identical conditions by the rod domain prepared by proteolysis of natural chicken gizzard a-actinin (Kahana & Gratzer, 1991). Both truncated rod fragments, containing repeats 1–3 and 2–4, respectively, gave rise to very much smaller proportions of covalent dimer in the cross-linking reaction, the extent of which rose with increasing protein concentration (Figure 4, lanes e to l, and Table 1). Whether the cross-linked material represents an equilibrium proportion of dimer or non-specific aggregation cannot be determined from these results. Figure 4 also shows conversion of the monomer zone in the SDS-gel into a doublet or even a triplet by the cross-linker. Such electrophoretic heterogeneity is often generated by bifunctional reagents and results presumably from intra-chain cross-linking, which can perturb both SDS binding and the dimensions of the hydrodynamic particle

Association of a-Actinin Structural Repeats

229 Figure 3. Trypsin-resistance of the expressed a-actinin repeat domain polypeptides. Purified a-actinin polypeptides were exposed to trypsin (enzyme/substrate ratio 1:1000) for the times shown (minutes), in 100 to 150 mM NaCl containing 1 to 2.5 mM CaCl2 , and the resulting digests analysed by SDS-PAGE. (A) Repeats 1–4; (B) Repeats 1–3; (C) Repeats 2–4. The positions of molecular mass standards (kDa) are indicated.

migrating in the gel (Griffith, 1972; Nakamura et al., 1991). That this is observed only in the truncated fragments and not with the intact rod domain we take to be a consequence of the high probability of inter-chain cross-linking in the stable dimer. These results show that the elimination of a repeat from either end of the rod greatly reduces dimerisation. If the repeats in the antiparallel dimer are aligned, then any association between the polypeptides containing repeat 1–3 or repeat 2–4 would, in both cases, result from two of the four pairwise associations prevailing in the complete rod domain (Figure 1). In a staggered model (Taylor & Taylor, 1993), in which repeat 1 does not participate in inter-chain interactions (Figure 1B), fragment 2–4 would be stabilised by the same three pairwise interactions as occur in the intact rod and should therefore have equal stability. In fragment 1–3 there could be only one such interaction, namely the isologous association between repeats 3 from either chain (Figure 1). The converse would apply for the

inverse stagger, with repeat 4 unpaired (Figure 1C). The results of the cross-linking experiments are thus compatible with an aligned structure for the a-actinin dimer but not, in the absence of other structural features, with a staggered structure. The cross-linking procedure may not give a quantitative reflection of the monomer-dimer equilibrium, since on the one hand the reaction is not fully efficient, in consequence probably of competing intra-chain reactions, and on the other it could displace the equilibrium. We have therefore undertaken an analysis of the association by sedimentation equilibrium. The results shown in Figure 5(A) and summarised in Table 2 indicate that the complete rod domain is predominantly or entirely dimeric. The weight-average mass (108 kDa) is a little lower than the calculated mass of the dimer (114 kDa). The small discrepancy could be due to (1) a difference between the true and calculated partial specific volumes, (2) the presence of a small proportion of residual, incorrectly folded monomeric material or (3) perceptible dissociation of the dimer in accordance with a reversible thermodynamic equilibrium. To discriminate between these possibilities a careful analysis was performed on the equilibrium distribution for the highest total polypeptide concentration to determine the apparent values of Mr at a series of radial positions down the solution column. Table 1. Formation of covalent dimers by chemical cross-linking of a-actinin polypeptides expressed in E. coli

Figure 4. Analysis of the association of a-actinin polypeptides expressed and purified from E. coli by chemical cross-linking. Varying concentrations of the a-actinin polypeptides were exposed to the cross-linking reagent dimethylsuberimidate (DMS), as described in Materials and Methods, and the products of the reaction were analysed by SDS-PAGE. The positions of molecular mass standards (kDa) are indicated. Lanes a to d, a-actinin polypeptide containing repeats 1–4, untreated (lane a) and treated with DMS at protein concentrations of 0.28 (b), 0.14 (c) and 0.028 (d) mg ml−1. Lanes e to h, a-actinin polypeptide containing repeats 2–4, untreated (lane e) and treated with DMS at protein concentrations of 0.4 (f), 0.2 (g) and 0.04 (h) mg ml−1. Lanes i to l, a-actinin polypeptide containing repeats 1–3, untreated (lane i) and treated with DMS at protein concentrations of 0.4 (j), 0.2 (k) and 0.04 mg ml−1.

Fragment

Helix − [u]222a content

Concentration (mg ml−1 )

Cross-linked dimer (%)

Repeats 1–4b

23,800

66c

0.28 0.14 0.03

67 69 63

Repeats 1–3

24,800

69

0.40 0.20 0.04

11 9 5

Repeats 2–4

24,900

69

0.40 0.20 0.04

16 16 7

Molar residue ellipticity (deg. cm2 dmol−1 ) at 222 nm. Cross-linking under identical conditions of the intact rod domain, formed by proteolysis of natural chicken gizzard a-actinin, gave 69% covalent dimers (Kahana & Gratzer, 1991). c Approximate helicity based on a value of −36,000° for 100% a-helix (Greenfield & Fasman, 1969). a

b

Association of a-Actinin Structural Repeats

230

Figure 5. Sedimentation equilibrium distributions of a-actinin polypeptides. Protein concentrations, measured by absorbance at 280 nm, as a function of distance from the centre of rotation at sedimentation equilibrium. (A) Repeats 1–4 fragment at 0.43 mg ml−1; (B) repeats 1–3 fragment at 0.28 mg ml−1; (C) repeats 2–4 fragment at 0.27 mg ml−1. Points are experimental and curves are calculated best fits for a monodisperse solute of Mr 108 kDa (A), monomer-dimer equilibrium with Mr 70 kDa for the monomer and association constant 5 × 105 M−1 (B), monomer-dimer equilibrium with Mr 60 kDa for the monomer and association constant 105 M−1 (C). In (C) the fit was computed for the upper 0.5 mm only of the solution column, so as to disregard the high molecular weight aggregates appearing near the cell bottom.

Within experimental error Mr was invariant throughout the column. This would seem to eliminate significant contributions from (2) and (3), but we do not exclude a small effect from either, the ensuing polydispersity masked possibly by concentration-dependence of the apparent Mr : this is necessarily small, however, and obviously so in view of the lack of any perceptible dependence of measured Mr values on initial protein concentration (Table 2), implying a negligibly small second virial coefficient. By contrast, the polypeptide spanning repeats 1–3 showed evidence of a concentration-dependent association; the increasing molecular mass towards the bottom of the solution column gives a satisfactory fit to a monomer-dimer equilibrium model with an Table 2. Association of a-actinin fragments determined by sedimentation equilibrium analysis Concentration (mg ml)−1

Dimer formation (%)

Rod (Repeats 1–4)

0.81 0.43 0.03

80 89 77

Repeats 1–3

0.86 0.37 0.28 0.19 0.04

71 64 61 57 53

Repeats 2–4

0.43 0.24 0.20 0.11 0.05

56 53 45 45 25

Fragment

The results shown for repeats 2–4 were obtained by analysis of the top half of the solution column, so excluding heavy aggregates from calculated weight-average molecular mass values.

association constant of approximately 5 × 105 M−1 (Figure 5(B)). The polypeptide spanning repeat 2–4 is largely monomeric, but the molecular mass again increases towards the bottom of the cell (Figure 5(C)). The distribution can be fitted by a monomer-dimer equilibrium, but the apparent association constant increases with initial protein concentrations, suggesting that higher aggregates are formed. The sedimentation data thus indicate again a qualitative difference in association state between the intact rod and its fragments, and are thus difficult to reconcile with a simple staggered model. Examination of the a-actinin polypeptides in the electron microscope after rotary shadowing (Figure 6) reveals that all three are predominantly rod-shaped. A limited extent of end-to-end association of the rods is seen in the truncated polypeptides. The images of the repeat 2–4 polypeptide appear to have appreciably smaller width than those of the other two polypeptides. In the complete rod domain and the repeat 1–3 polypeptides, many of the particles appear to contain two laterally aligned components. The average contour length of the intact rod domain was 22.5(22.1) nm, close to that (25 nm) of the equivalent polypeptide liberated from chicken gizzard a-actinin by proteolysis (Imamura et al., 1988). The length of the repeat 1–3 polypeptide was determined to be 20.6(21.7) nm and that of repeat 2–4 was 14.8(22.9) nm (2standard error of estimate of mean value (SEM); N = 20 in all cases).

Discussion The cross-linking and sedimentation equilibrium data show a large difference in the extent of self-association between the intact rod domain (repeat 1–4) and the deletion mutants comprising

Association of a-Actinin Structural Repeats

Figure 6. Electron micrographs of rotary shadowed a-actinin polypeptides. A, Repeats 1–4; B, repeats 1–3; C, repeats 2–4. The bar represents 20 nm. The arrows indicate examples of apparent linear association.

repeats 1–3 or 2–4. Even though the results for 2–4 must be regarded as only semi-quantitative, in consequence of a tendency of the molecule to aggregate (a process probably precluded in the intact rod domain by formation of the stable dimer), the

231 data argue against a staggered model for the rod dimer. The results of electron microscopy are less conclusive. The dimensions of the intact rod-domain dimer (Figure 6) correspond to a length of about 5.5 nm per repeat. This is close to the value inferred by Taylor & Taylor (1993) from electron micrograph projections of two-dimensional a-actinin crystals. For erythroid spectrin (Shotton et al., 1979) the dimer contour length is 100 nm, and that of the tetramer 200 nm, which is also the length of the neuronal spectrin tetramer (Glenney et al., 1983); thus with 22 complete repeats in the a-chain and 17 in the b-chain, the lengths per repeat are 4.5 and 5.9 nm, respectively, but this makes no allowance for the terminal and other non-homologous elements in the sequences (Sahr et al., 1990; Winkelmann et al., 1990; Wasenius et al., 1989). A further uncertainty is whether the proteins prepared for shadowing electron microscopy in solutions containing 50% glycerol, rising to 100% during drying, attain their maximally extended lengths in these circumstances. This condition is evidently fulfilled in the case of erythroid spectrin, as determined by its extension at low ionic strengths (McGough & Josephs, 1990). By hydrodynamic criteria skeletal muscle a-actinin behaves in the same manner (Kuroda et al., 1994). Particle lengths therefore do not give definitive information on the number of repeats between the ends of the a-actinin rod dimer. On the other hand, the thickness appears uniform from end to end. It is also possible that, if these proteins can be represented as two-stranded helices with variable pitch (McGough & Josephs, 1990), the maximum pitch could be different in spectrin and a-actinin. Meyer & Aebi (1990) have demonstrated extensibility in a-actinins from two sources (Dictyostelium and Acanthamoeba); both types change their lengths when they bind to F-actin. The similar lengths of the intact rod-domain polypeptide (22.5(22.1) nm) and fragment 1–3 (20.6(21.7) nm), taken together with the smaller length (by the equivalent of one repeat) of fragment 2–4, are of course compatible with the staggered model shown in Figure lB, but in this case a stable antiparallel dimer of repeats 1–3 would need to form through a single-paired interaction between repeat 3 of the two chains. This is intrinsically unlikely and in any event the cross-linking and sedimentation data rule out complete dimerisation in any accessible concentration range, as would be implied by this interpretation of the electron microscope observations. The relation of our data to the appearance of the a-actinin dimers in the two-dimensional crystals described by Taylor & Taylor (1993) thus remain to be clarified. The possibility of a more complex structure for the rod domain than a contiguous or nested succession of identical repeats cannot be disregarded. The exceptional stability of the isolated rod dimer (Imamura et al., 1988; Kahana & Gratzer, 1991) excludes a dominant contribution from interactions between the N and C-terminal domains of a-actinin to dimerisation. On the other hand, the remarkable

232 extensibility of the dimer, which reveals itself in the Acanthamoeba protein by a 26% increase in length when it binds at both ends to F-actin (Meyer & Aebi, 1990), could reflect slippage of the constituent subunits relative to each other or a change in pitch, if the rod conforms to the helical model put forward by McGough & Josephs (1990) for spectrin (compare also the evidence for ionic strength-dependence of the length of muscle a-actinin in solution (Kuroda et al., 1994)). At all events, a structural difference between the molecule in dilute aqueous solution and when constrained in a two-dimensional crystal lattice (or indeed when partially dehydrated in high glycerol concentrations before shadowing) is not unlikely. A switch in the pattern of inter-chain interactions between two conformational states of the dimer in equilibrium with one another could also in principle give rise to distinct cross-linked products on reaction with bifunctional reagents; such a phenomenon is exemplified in actin, which can generate two electrophoretically distinct dimers, representing interactions in the genetic-helix and long-pitch helix modes, respectively (Millonig et al., 1988). The work of Speicher et al. (1992) on erythroid spectrin has demonstrated a wide variation in the strengths of pairwise interactions among repeats in the antiparallel ab-dimer, with strong association between the first four repeats at the N-terminal end of the a-chain and the corresponding repeats at the C terminus of the b-chain. The complementarity must be assumed to be specific; thus, for example, no hybridisation occurs between the polypeptide chains of erythroid spectrin and the neuronal spectrin, fodrin (Glenney et al., 1983). In a-actinin, if the antiparallel chains are aligned, there can be only two types of interaction: those between repeats l and 4 and between repeats 2 and 3. In the aligned structure the absence of one repeat would thus eliminate two paired interactions. That self-association of the truncated rod domains is much weaker than that of the intact domain is entirely to be expected, since multiple interactions along the polypeptide chains must lead to high co-operativity of binding and a favourable contribution from the reduced entropy of mixing on dissociation. If the negative free energy of association of repeats 1 and 4 is DG14 , and that between repeats 2 and 3 DG23 , then that for formation of a dimer by the integral rod sequence (assuming negligible statistical weights for partially associated chains) should be 2DG14 + 2DG23 + 3TDSm , where T is the absolute temperature, and DSm , the entropy of ideal mixing, represents the entropic advantage of an association between a pair of reactants already apposed to one another by virtue of the association of an attached pair. This disregards other (especially rotational) entropic terms, which could also make a contribution. Then DSm = R ln xo , where xo is the mole fraction (1/55.6) of water in a protein solution sufficiently dilute that the mole fraction of protein can be disregarded. (The thermodynamic basis of the cratic entropy correc-

Association of a-Actinin Structural Repeats

tion has recently been questioned (Holtzer, 1995). In such a case as the present a favourable translational entropy contribution must nevertheless exist.) Thus, for values of the partial free energy contributions so small that no association could be detected in the accessible concentration range (say 4 kcal mol−1, corresponding to an association constant of 103 M−1 ), a notional free energy of dimerisation for the entire domain of some 24 kcal mol−1 (association constant of 1017 M) might be generated. On this basis elimination of one interacting pair of repeats would reduce the association constant by 4 to 5 orders of magnitude. The integral rod domain does indeed form a very stable dimer (Kahana & Gratzer, 1991), with an association constant greater than about 1012 M−1. A greatly reduced stability for dimerisation of a three-repeat chain is also then predicted. Quantitative determination of the contribution of each interaction to the free energy of association will require additional studies on further constructs.

Materials and Methods Expression and purification of a-actinin polypeptides cDNAs encoding a-actinin repeat 1–4 (amino acid residues 267 to 749), repeat 1–3 (residues 267 to 615) and repeats 2–4 (residues 385 to 749) were generated by amplifying selected regions of the full-length chick a-actinin cDNA C17 (Baron et al., 1987) by polymerase chain reaction (PCR). Because there is a BamHI site in repeat 1, constructs encoding this repeat were generated, using primers containing an EcoRI site, and the PCR products were subcloned into the EcoRI site of pGEX-3X. The repeat 2–4 construct was amplified using a 5' primer containing a BamHI site and a 3' primer containing an EcoRI site and the PCR product was subcloned into BamHI/EcoRI-cut pGEX-2T. Recombinant a-actinin fragments were expressed in E. coli and purified on glutathione-agarose beads (Sigma), essentially as described by Smith & Johnson (1988). The a-actinin polypeptides were liberated from the GST either by thrombin (pGEX-2T) or factor Xa cleavage (pGEX-3X) with the fusion proteins still bound to the glutathione-agarose, as described (Gilmore et al., 1994). The purity of the liberated polypeptides was assessed by SDS-PAGE (Laemmli, 1970) and protein folding was monitored by circular dichroism (Jobin Yvon Dicrogaphe CD6). Protein concentrations were determined spectrophotometrically with molar absorptivities calculated from the contents of the chromophoric amino acids (Perkins, 1986) or by the BCA colour reaction (Pierce) using the intact a-actinin rod as the calibration standard. Evaluation of a-actinin dimerisation by cross-linking Cross-linking was carried out with dimethylsuberimidate (Davis & Stark, 1970) at 24 mM in 0.2 M triethanolamine hydrochloride (pH 8.5) for one hour on ice. Protein concentrations were varied between 0.03 and 0.40 mg ml−1. Unreacted cross-linker was quenched with an excess of Tris (pH 7.6) and the samples were analysed by SDS-PAGE and densitometry of the stained gels.

Association of a-Actinin Structural Repeats

Sedimentation analysis Sedimentation equilibrium measurements were performed in the Beckman Optima XL-A analytical ultracentrifuge, using ultraviolet absorption optics between 220 and 280 nm. Protein concentrations were in the range 0.01 to 0.90 mg ml−1 in 100 to 150 mM NaCl containing 1 to 2.5 mM CaCl2 . A solution column of 1.2 mm was centrifuged to equilibrium at 10,000 revs/min for 11 hours at 5°C. Plots of protein concentration against distance from the centre of rotation were fitted to a single-term sedimentation equilibrium relation, assuming a negligible second virial coefficient term at these low protein concentrations. Where a deviation from the distribution for a monodisperse solute was observed, the curves were fitted to a monomer-dimer equilibrium. This yielded an association constant, corresponding to the non-linear least-squares fit. Electron microscopy For electron microscopy, protein samples at 0.1 mg ml−1 in 50% (v/v) glycerol were sprayed onto mica, shadowed with platinum at an angle of 5° and coated with carbon. The replicas were transferred to grids and examined in a Siemens 102 transmission electron microscope at 80 kV accelerating voltage. Particle dimensions were corrected for metal cap by using tobacco mosaic virus as an internal standard.

Acknowledgements This work was supported in part by a Cancer Research Campaign grant to D.R.C., a Muscular Dystrophy Group grant to W.B.G. and a BBSRC grant to A.J.R. in support of the National Centre for Macromolecular Hydrodynamics. G.F. was supported by a University of Leicester research studentship.

References Baron, M. D., Davison, M. D., Jones, P. & Critchley, D. R. (1987). The sequence of a chick a-actinin reveals homologies to spectrin and calmodulin. J. Biol. Chem. 262, 17623–17629. Blanchard A., Ohanian, V. & Critchley, D. R. (1989). The structure and function of a-actinin. J. Muscle Res. Cell Motil. 10, 280–289. Cross, R. A., Stewart, M. & Kendrick-Jones, J. (1990). Structural predictions for the central domain of dystrophin. FEBS Letters 262, 87–92. Davis, G. E. & Stark, G. R. (1970). Use of dimethyl suberimidate, a cross-linking reagent, in studying the subunit structure of oligomeric proteins. Proc. Natl Acad. Sci. USA, 66, 651–657. Davison, M. D. & Critchley, D. R. (1988). a-Actinins and DMD protein contain spectrin-like repeats. Cell, 52, 159–160. Gilmore, A. P., Parr, T., Patel, B., Gratzer, W. B. & Critchley, D. R. (1994). Analysis of the phasing of the four spectrin-like repeats in a-actinin. Eur. J. Biochem. 225, 235–242. Glenney, J. R., Glenney, P. & Weber, K. (1983). Mapping the fodrin molecule with monoclonal antibodies. A general approach for rod-like multidomain proteins. J. Mol. Biol. 167, 175–293.

233 Greenfield, N. & Fasman, G. D. (1969). Computed circular dichroism spectra for the evaluation of protein conformation. Biochemistry, 8, 4108–4116. Griffith, I. P. (1972). The effect of cross-links on the mobility of proteins in dodecyl sulphate-polyacrylamide gels. Biochem. J. 126, 553–560. Holtzer, A. (1995). The ‘cratic correction’ and related fallacies. Biopolymers 35, 595–602. Imamura, M., Endo, T., Kuroda, M., Tanaka, T. & Masaki, T. (1988). Substructure and higher structure of chicken smooth muscle a-actinin molecule. J. Biol. Chem. 263, 7800–7805. Kahana, E. & Gratzer, W. B. (1991). Properties of the spectrin-like element of smooth-muscle a-actinin. Cell Motil. Cytoskel. 20, 242–248. Kahana, E., Marsh, P. J., Henry, A. J., Way, M. & Gratzer, W. B. (1994). Conformation and phasing of dystrophin structural repeats. J. Mol. Biol. 235, 1271–1277. Koenig, M. & Kunkel, L. M. (1990). Detailed analysis of the repeat domain of dystrophin reveals four potential hinge segments that may confer flexibility. J. Biol. Chem. 265, 4560–4566. Kuroda, M., Kohira, Y. & Sasaki, M. (1994). Conformational change of skeletal muscle a-actinin induced by salt. Biochim. Biophys. Acta, 1205, 97–104. Laemmli, U. K. (1970). Cleavage of structural proteins during the assembly of bacteriophage T4. Nature, 227, 680–685. McGough, A. M. & Josephs, R. (1990). On the structure of erythrocyte spectrin in partially expanded membrane skeletons. Proc. Natl Acad. Sci. USA, 87, 5208–5212. Meyer, R. K. & Aebi, U. (1990). Bundling of actin filaments by a-actinin depends on its molecular length. J. Cell Biol. 110, 2013–2024. Millonig, R., Salvo, H. & Aebi, U. (1988). Probing actin polymerization by intermolecular cross-linking. J. Cell Biol. 106, 785–796. Nakamura, S., Sakurai, T. & Nonomura, T. (1991). Determination of the native subunit pattern of gizzard tropomyosin using antibodies specific to each subunit. J. Biochem. (Tokyo), 109, 758–762. Parry, D. A. N. & Cohen, C. (1991). Structure of the spectrin superfamily: a three-a-helix motif. In The Living Cell in Four Dimensions (Paillotin, G., ed.), pp. 367–377, vol. 226, AIP Conference Proceedings, American Institute of Physics, New York. Perkins, S. J. (1986). Protein volumes and hydration effects. The calculation of partial specific volumes, neutron scattering matchpoints and 280-nm absorption coefficients for proteins and glycoproteins from amino acid sequences. Eur. J. Biochem. 157, 169–180. Sahr, K. P., Laurila, P., Kotula, L., Scarpa, A. L., Coupal, E., Leto, T. L., Linnnebach, A. J., Winkelmann, J. C., Speicher, D. W., Marchesi, V. T., Curtis, P. J. & Forget, B. G. (1990). The complete cDNA and polypeptide sequences of erythroid a-spectrin. J. Biol. Chem. 245, 4434–4443. Shotton, D. M. Burke, B. & Branton, D. (1979). The molecular structure of human erythocyte spectrin. Biophysical and electron microscopic studies. J. Mol. Biol. 130, 303–329. Smith, D. B. & Johnson, K. S. (1988). Single step purification of proteins expressed in E. coli as fusion proteins with glutathione-S-transferase. Gene, 67, 31–40. Speicher, D. W. & Marchesi, V. T. (1984). Erythrocyte spectrin is comprised of many homologous triple helical segments. Nature, 311, 177–180. Speicher, D. W., Weglarz, L. & DeSilva, T. M. (1992). Properties of human red cell spectrin heterodimer

Association of a-Actinin Structural Repeats

234 (side-to-side) assembly and identification of an essential nucleation site. J. Biol. Chem. 267, 14775– 14782. Taylor, K. A. & Taylor, D. W. (1993). Projection image of smooth muscle a-actinin from two-dimensional crystals formed on positively charged lipid layers. J. Mol. Biol. 230, 196–205. Tinsley, J. M., Blake, D. J., Roche, A., Fairbrother, U., Riss, J., Byth, B. C., Knight, A. E., Kendrick-Jones, J., Suthers, G. K., Love, D. R., Edwards, Y. H. & Davies, K. E. (1992). Primary structure of dystrophin-related protein. Nature, 360, 591–593. Wasenius, V.-M., Saraste, M., Salve´n, P., Era¨maa, M., Holm,

L. & Lehto, V.-P. (1989). Primary structure of the brain a-spectrin. J. Cell Biol. 108, 79–93. Winkelmann, J. C., Chang, J.-G., Tse, W. T., Scarpa, A. L., Marchesi, V. T. & Forget, B. G. (1990). Full-length sequence of the cDNA for human erythroid b-spectrin. J. Biol. Chem. 265, 11827–11832. Winograd, E., Hume, D. & Branton, D. (1991). Phasing the conformational unit of spectrin. Proc. Natl Acad. Sci. USA, 88, 10788–10791. Yan, Y., Winograd, E., Viel, A., Cronin, T., Harrison, S. C. & Branton, D. (1993). Crystal structure of the repetitive segments of spectrin. Science, 262, 2027– 2030.

Edited by J. Karn (Received 2 February 1995; accepted in revised form 28 June 1995)