High resolution structure of an oligomeric eye lens β-crystallin

J. Mol. Riol. (1991) 222, 1067-1083

High Resolution Structure of an Oligomeric Eye Lens fI-Crystallin Loops, Arches, Linkers and Interfaces in pB2 Dimer Compared to a Monomeric y-Crystallin R. Lapattot,

V. Nalini, B. Bax, H. Driessen, P. F. Lindley T. L. Blundell and C. Slingsby

Laboratory of Molecular Biology and Imperial Cancer Research Fund Unit Department of Crystallography Birkbeck College, London University Malet Street, London WClE 7HX, U.K. (Received 10 May 1991; accepted 14 August 1991) J-Crystallins are polydisperse, oligomeric structural proteins that have a major role in forming the high refractive index of the eye lens. Using single crystal X-ray crystallography with molecular replacement, the structure of /IS2 dimer has been solved at 2.1 A resolution. Each subunit comprises an N and C-terminal domain that are very similar and each domain is formed from two similar “Greek key” motifs relat,ed by a local dyad. Sequence differences in the internally quadruplicated molecules, analysed in terms of their /?-sheets, hairpins and arches, give rise to structural differences in the motifs. Whereas the related family of y-crystallins are monomers, fi-crystallins are always oligomers. In the fiB2 subunit, the domains, each comprising two motifs, are separated by an extended linking peptide. A crystallographic 2-fold axis relates the two subunits of the dimer so that the N-terminal domain of one subunit of /?B2 and the C-terminal domain of the symmetry-related subunit are topologically equivalent to the two covalently connected domains of yB-crystallin. The intersubunit domain interface is very similar to the intradomain interface of yB, although many sequence differences have resulted in an increase in polar interactions between domains in PB2. Comparison of the structures of fiB2 and yB-crystallins shows that the two families differ largely in the conformation of their connecting peptides. A further extensive lattice contact indicates a tetramer with 222 symmetry. The ways in which insertions and extensions in the /?-crystallin effect oligomer interactions are described. The two kinds of crystallin are analysed for structural features that account for their different st’abilities. These studies are a basis for understanding formation of higher aggregates in the lens. Keywords: B,y-crystallins;

/3B2; eye lens; domain; molecular replacement

1. Introduction

found in the lens occurring as both homodimers and heterodimers, as well as in higher molecular weight aggregates (Slingsby & Bateman, 1990). /3-Crystallins belong to the same superfamily as monomeric y-crystallins (Driessen et al., 1981) and it was predicted that they would have a similar twodomain structure (Wistow et al., 1981; Inana et al., 1983; Slingsby et aE., 1988). X-ray structure analysis of /IS2 homodimer shows that this fi-crystallin is, indeed, composed of two similar domains, each domain being formed from two “Greek key” motifs (Bax et aZ., 1990). However, the connecting peptide was extended and the two domains were separated 7 Current address: Department of Medical Chemistry, in a way that was quite unlike the y-crystallins and University of Helsinki, Siltavuorenpenger 10, SF-00170 had not been predicted. The dimer interface is Helsinki, Finland. 1067

/?-Crystallins are a polydisperse group of oligomerit proteins found in eye lenses where, together with polymeric a-crystallins and monomeric y-crystallins, they contribute to the tightly packed protein matrix within fibre-like cells (Harding & Crabbe, 1984; Slingsby, 1985; Wistow & Piatigorsky, 1988). They are a multigene family comprising basic (/?Bl, /lB2, fiB3) and acidic @Al, fiA2, /?A3, PA4) subunits (Berbers et al., 1984). pB2 is a component of many /?-crystallin oligomers

0022-28:36/91/241067-17

$03.00/O

0 1991 Academic Press Limited

1068

R. Lapatto et al.

analogous to the intramolecular interdomain interface in y-crystallin. In the y-crystallins a single motif is used to form symmetrical domains (Blundell et al., 1981; Wistow et al., 1983; Chirgadze et al., 1986; White et al., 1989), the stability of which may have contributed to their selection as long-lived proteins essential for maintenance of lens transparency. In the oligomer, similar domains are connected in different ways as a result of different linker conformations; such knowledge is useful for the design of symmetrical oligomers. The high refractive index of eye lenses requires concentrated protein solutions; any tendency for sporadic aggregation, however, needs to be suppressed as fluctuations in the index of refraction will reduce transparency (Delaye & Tardieu, 1983; Benedek, 1971). The evolution of monomeric and various oligomeric structures of differing size from a simple repeating motif is probably selectively advantageous in providing an even packing of protein components within the lens of the appropriate refractive index. Here, we describe in detail the solution of the crystal structure of the symmetrical dimer of /?B2crystallin by X-ray analysis using yB-crystallin as a search molecule. The symmetry of the molecule and the differences in the connections between domains in the search molecule and flB2 posed very special problems, necessitating a careful validation of the model. Here we report a comparison of the fiB2 subunit with yB, which have about 30% identical residues. We consider first the conserved folded hairpin. We also compare topologically equivalent arches, in order to observe how insertions and deletions are incorporated into the domain structure. Members of the p-crystallin family are between 45 y. and 60% identical with bB2 (Berbers et al., 1984; Gorin & Horwitz, 1984; Quax-Jeuken et al., 1984; van Rens et aE., 1991) and, compared with y-crystallins, have several insertions in the loops that connect certain /?-strands. We assess the role that these sequence differences might play in domain interactions in the P-crystallin family. Previously, it had been assumed that the extensions at the N and C termini of /?-crystallins would hold the subunits together (Wistow et al., 1981; Slingsby et al., 1988) and some experimental data supported such a role for the N-terminal arm in flB2 (Berbers et al., 1983). However, the structure of the j3B2 homodimer indicates that only one extra residue in the C-terminal extension of j?B2 contributes towards stabilizing the dimer. Dimer stability must therefore be a consequence of two intersubunit P-sheet interfaces in BE!2 which we compare and contrast with the interface in y-crystallins. The main function of crystallins in eye lenses is to associate closely so that a high refractive index is achieved. The flB2 subunit not only self-associates to form dimers but also interacts with many other fi-crystallin subunits, forming heterodimers and higher complexes (Slingsby et aE., 1982; Slingsby & Bateman, 1990). In the crystal lattice, the fiB2 dimer associates with a second equivalent dimer to

give a tetramer with 222 symmetry (Bax et al., 1990). We describe the nature of this dimer-dimer interface, which provides a good model for building higher levels of interactions in the larger /?-crystallin aggregates. Knowledge of the modes of protein aggregation in the normal lens is a prerequisite for understanding the changes that occur in aging and cataract.

2. Experimental Methods and Results (a) Crystallization

and data collection

/?B2 was purified and crystallized as previously described (Bax & Slingsby, 1989). The crystals are orthowith cell dimensions rhombic, I222 or z212,2,, a = 77-812) A, b = B-6(2) .k, c = 109.2(3) A (1 A =Ol nm). with 1 subunit in the asymmetric unit. A total of 2 independent X-ray data sets were measured, the 1st to 265 A with I = 087 A and the other to 2.1 A with 1=089 A, each from 1 crystal, at the Synchrotron Radiation Source, SERC Daresbury Laboratory. The data were scaled and merged to give 10,510 unique reflections (37,200 measurements) at 2.65 A (98.2% of the unique data) with Rsymof 3.9% and 18,583 unique reflections (62,600 measurements) at 2.1 A (87.6% complete) with R,,, of 63%. (b) Molecular replacement The structure was solved (Bax et al.. 1990) by the method of molecular replacement using the yB molecular structure as a search molecule (Summers et al.. 1984). The sequence identity between jB2 and yB-crystallin is about 30%, similar to that internally between the N and C domains of /?B2 (or yB). The identity of the N-terminal domain of /lB2 with the C-terminal domain of yB and the C-terminal domain of j?B2 with the N-terminal domain of yB is also about 30%. These observations suggest that, the b and y-crystallins diverged at about the same time as the internal gene duplication which led to the 2 domains in each family. Therefore, special care had to be taken to distinguish the true solution from the pseudo solution in the rotation function. The yB search molecule was placed in a Pl cell with the pseudo-2-fold axis of the molecule aligned with the z axis and the long axis of the molecule roughly parallel bo 2. In all. 2 peaks related by approximately 180” in eulerian y angle were found on the /? = 30” section, indicating that the orientation of the 2 domains in the jlB2 crystal was similar to that in the yB search molecule. Translation functions were calculated in both possible space groups, 1222 and 12,2,2,. for both rotation function peaks using the T, function (Crowther & Blow, 1967) as implemented in the program TFSGEN (Tickle. 1985). The highest peaks in the resulting maps were analysed for packing constraints using computer graphics. In space group 12,2i2,, no acceptable solutions were found. Only the 2 solutions calculated in space group 1222 packed in an acceptable manner: essentially the only difference between them was a rotation of the yB molecule by 180” around its internal 2-fold axis, as expected. (c) Rejhement Following rigid body refinement using RESTRAIK (Driessen et al., 1989), an electron density map was calculated using all data between 10.0 and 2.65 A. The map. displayed on an Evans and Sutherland PS 390 using FRODO (Jones 1978). showed good connectivity for

Structure of an Eye Lens Oligomeric B-Cry&a&n

1069

and hence the pseudo rotation function solution had in reality been adopted, although it was slightly more prominent. This was particularly noticeable for residues Pro1 and Leu36 (Fig. l(a)) which both had extra density, indicating that they were really the topologically equivalent residues His88 and Arg125 from a C-terminal domain. There was insufficient density for Argl25 in the original model and it had adopted a very distorted conformation in the refinement trying to fit the density for Leu36. In all, 55 of the side-chains from residues 1 to 82 were replaced with those of residues 88 to 172, while the corresponding residues from 88 to 172 were replaced with the appropriate residues from 1 to 82. With the model this way round the insertions at 28A and 106A were better fitted into the density. This new model of /IB2, with the N and C domains swapped around, was subjected to further least-squares refinement with RESTRAIN. Maps calculated from the new model showed the expected electron density for side-chains (Fig. l(b)). However, the density of the interdomain connecting peptide and of the N and C-terminal extensions were unclear, as was the density of the main-chain in the region of the inserted residues 70A and 70B. However, the map did show a novel connection between the N and the C domains as shown in Fig. 2(a) and 2(b) (Bax et al., 1990). The model with the new connectivity refined well and had very good density. Extensive manual model building together with both conventional restrained least-squares and simulated annealing refinement (Briinger, 1988) were used to locate the double insertion region between residues 70 and 71, 2 of the residues ( - 1 and -2) from the N-terminal extension and residues 173 to 175 from the C-terminal arm. Following the placement of 106 water molecules, the final R-factor was 186% for all data between 2.1 A and 160 A.

3. Molecular

(b) Figure 1. The correct assignment of the N-terminal domain (residues 1 to 82) and C-terminal domain (residues 88 to 17 1) was made by examining the electron density of topologically equivalent residues from the 2 domains. (a) The electron density for Leu36 from the refined pseudo-solution, 21F01-lF,l, clearly shows extra density, whereas there was too little density for the topologically equivalent Arg125. (b) Following rotation of the model by 180” about the pseudo dyad between domains and reinterpretation of the connecting peptide conformation, the final refined structure showed good fit of model to density for Arg125.

much of the main-chain and for many side-chains. The side-chains of yB were replaced with those of pB2 and extra residues were inserted at 28A, 70A, 70B and 106A. The /?B2 model was subject to further refinement using RESTRAIN, including refinement of isotropic temperature factors, which reduced the R-factor to less than 30%. At this stage the electron density of side-chains in welldefined regions of the map was examined and the density clearly showed that the model was the wrong way round

Overview

The refined structure shows how each subunit of /IS2 folds into two domains that are very similar to the domains of y-crystallins (Fig. 2(a)). However, the two domains are not in contact due to a difference in conformation of the connecting peptide (Fig. 2(b)). The two subunits are related by a crystallographic 2-fold called P. This results in the N-terminal domain of subunit 1 interacting with the C-terminal domain of subunit 2 and the C-terminal domain of subunit 1 interacting with the N-terminal domain of subunit 2 (Fig. 2(c)). The dimer is related by two orthogonal dyads Q and R to a further dimer to give a tetramer with crystallographic 222 symmetry (Figs 2(d) and (e)). Each domain is formed from two Greek key motifs related by a pseudodyad. Each motif is formed from four antiparallel P-strands, a, b, c and d, with two motifs intercalating to form two b-sheets such that the a, b and d strands from one motif hydrogen bond with the c strand of the other motif (Figs 3 and 5(b)).

4. Structure (a) The alignment

of the Protomer and secondary structure

The structure of the j3B2 dimer shows that /I-crystallins, like y-crystallins, are constructed from four Greek key motifs each of approximately 40

1070

R. Lapatto

--

,. .I

R

(d

1

Fig. 2.

et al.

Structure of an Eye Lens Oligomeric #K’rystallin

1071

(e I

Figure 2. Stereo views of ribbon diagrams showing the path of the a-carbon atoms of the polypeptide chain with definition of axes and interfaces. (a) The yB molecule viewed approximately perpendicular to the pseudo 2-fold axis relating N and C-terminal domains. (b) Subunit 1 of /?B2 showing how the 2 domains are separated due to the extended conformation of the connecting peptide. (c) flB2 dimer viewed down the 2-fold P axis. The pseudo dyad (1) is shown relating the N-terminal domain of subunit 1 (continuous ribbon) to the C-terminal domain of subunit 2 (broken ribbon). (d) A tetramer of bB2 with subunits 1 and 3 in continuous ribbon and subunits 2 and 4 in broken ribbon. The dimer comprising subunits 1 and 2 is in the same orientation as in (c). N and C-terminal domains for all 4 subunits are indicated. (e) The tetramer viewed down the Q axis. amino key

acid

motifs

residues. The corresponding of bB2 and yB were compared,

main-chain

torsion angles at 2.1 A was calculated (Fig. 4(a)) and residues were assigned a conformation (Fig. 4(b)). The single letter codes for /3 confor-

Greek parti-

cularly in terms of their main-chain torsion angles and hydrogen bonding patterns, from which the alignment in Figure 3 was derived. There are 181 out of 204 residues of PBS visible in the electron density map; 13 residues from the NH,-terminal and ten residues from the COOH-terminal extensions are undefined. A Ramachandran plot of the -2

LO

mation (B or P), right-handed (A or a) and lefthanded (L or G) a helical, and glycine (E) conformation were also plotted onto the fiB2,yB motif alignment (Fig. 3). The plot shows that of the 65% of residues that are in fi conformation, 27% are in rather than extended, coiled polyproline-like,

20

LYA

B 7B

082

B

BPBBBBBABPLLBEPBBPBa

a.

PBaBaAa

b.

90

100

aLP

---

LOhA

aBPaBBBBaB

--

d, 11”

118h

AAaL

FRwRIYERDDPRGQwSEI~-D-DCPSLQDRFHL----TEVHSLNVtE HKITLYENPNFTGKKWEVIDD-DVPSFHAH-GY---QEKVSSVRVQS PBBBBBABPLLBBFBBBBBOP

PBaBAAaa

IL0

aB GB -

----

aPPBaBBBBAB

b, 40

7B 682

50

L30

7B 882

d, 60

7OA'"B

aaa

EB GC~~LYERPNYQGHQYFLR-RGDYPDYQQ-~~G--FNDSIRSCRLIPQHTGT GPWVGYEQANCKGEQPVFE-KG~YPRWDS-WTSSRRTDSLSSLRPIKVDSQE PEBBaBAAa BPBBBBABPLLEEPBBPBB

4 140

A

B

c,

B 7B 682

30 E

GKITPYEDRGPQGHCYECS-S-DCPNLQP-Y-F----SRCNSIRVDS LNPKIIIPEQENPQGHSHELN-G-PCPNLKE-TGV----EKAGSVLVQA

-

150 P

B

PP

PEP

~~~ALAB~PB~BBBPBBPBBPB

----

c,

GSWVLY~wPSYRGRQYLLR-PGEYRRYLD-wGA--wNAKVGSLRRVwDFY GTWVGYQYPGYRGLQYLLE-KGDYKDSGD-~~~--PQPQVQSVRRIRDWQW PEBBABAAA BBBBPBABPLLBEPBBPBB c. b. a.

80

dLP

-

d, 160

L

a

a

110

aP

BBPa

ABPBB~BBBPP~BB~ ----

-d.

Figure 3. Sequence alignment of the 4 motifs of bovine /IS2 and bovine yB-crystallin arranged so that topologically equivalent residues are aligned in vertical rows. Motifs 1 and 3 are grouped together as are motifs 2 and 4 to indicate their closer similarity. The numbering is based on yB. /?B2 has an additional 13 and 10 residues (not shown) at the NH, and COOH-terminal extensions, respectively. The letters in italics below the BB2 sequence refer to the conformation type as defined in Fig. 4(b). If the conformation in yB is different from flB2. it is indicated above the yB sequence in italics. 0 refers to a residue outside the boundaries of the Ramachandran plot of Fig. 4. The solid lines underneath the /IS2 sequence indicate the b-sheets of bB2 as derived from the hydrogen bonding pattern and the broken wavy line under the /?B2 sequence refers to the 3,, helical region of flB2.

R. Lapatto et al.

1072

Thr69 and Pro162 have conformations in between those of A, Lz and P conformations typical of turn regions. (b) /&Sheets and connecting peptides

(a)

6

a

I

Ll 1

4 (b)

Figure 4. (a) A Ramachandran plot of the backbone torsion angles of the refined co-ordinates of flB2. @, glycine; *, proline; x , all other amino acid residues. (b) The conformation of residues, defined in Wilmot $ Thornton (1990), is denoted by the 1 letter code (printed in the text and on Fig. 3 in italics) and is derived from its position in the Ramachandran plot. The boundaries are defined as by A. L. Morris. M. W. MacArthur. E. G. Hutchinson & J. M. Thornton. unpublished results.

conformation. Although 22 o/o of residues are in right-handed helical conformation, very few are consecutive with the exception of the region between c and d b-strands in each motif where a short stretch of CI 3,, helix occurs (Fig. 3). Any contribution made by the right-handed helical residues to the helical signal measured by circular dichroism in solution (Horwitz et al., 1986; Maiti et al., 1988) would be partially compensated by residues with left-handed helical conformations.

The four P-sheets in both /?B2 and ?B are very similar (Fig. 5). A residue is considered to be in a p-strand if either its amido or carbonyl group participates in a sheet hydrogen bond. This definition allows the conserved glutamate residues (positions 7. 94. 46 and 135) in the motifs to be considered as part of a /?-sheet through their peptide nitrogen, even though the residues are in righthanded helical conformation. The interdomain interface sheets are composed of fi-strands (cl, d2, a2 and b2) and (c,, d,, a4 and b4), whereas the solvent facing sheets are formed from (c,, d,, a, and b, ) and (~4, 4, a3 and b3) strands (Fig. 5). The c strands in the interface sheets contribute only one residue. whereas the outer sheets have a longer, highly distorted c strand due to the inserted glycine residues (positions 60 and 149) having E conformat’ion. The insert’ion of AsplOfiA, which has a high energy conformat,ion, in motif 3 of PB2 extends the b, strand and consequently the a3 strand in that sheet. In both proteins the d strands from solvent, facing sheets terminate with p-bulges that accentuate the local twist of the sheets (Richardson et al.. 1978), enabling the strands at t’he ends of these sheets to turn abrupt,ly> thus closing the barrel by joining up with the start of the interface sheet (d, aZ and d,a,). The twists are stabilized by the residues preceding the bulges (positions 37 and 126) hydrogen bonding with the additional glycinr residues of the solvent-facing c strands (Fig. 5(b)). By contrast. the ends of the d strands of the interface sheets lead into either the connecting peptide or C-terminal extension for which there is no need for such a twist and the residues that) are topologically equivalent to residues 38 and 127 are hydrophobic (positions XI and 170), contributing to the interdomain interfact,. Pro80 in /?B2, LeuHO in yB and Argl69 are in proline-like rather than the extended /3 conformkttion of the solvent facing strands (Fig. 3). Whereas all fl-crystallins have a proline residue at position 80, none of the y-crystallins does. and whereas all y-crystallins have a proline residue at position 82. this is not the case for any of the is whether these B,jj fi-crystallins. Th e question sequence specificities influence either the conformations of their connecting peptides or t,he disposit’ion of hydrophobic interface residues. Tn ;IB, most of the residues between positions 80 and 87 are in polyproline conformation with Gly86, allowing a sharp turn due to its E conformat,ion, whereas in BB2. many of t~he equivalent residues are in b extended conformation with Gln86 excluded from the E’ region (Fig. 3). Ramakrishnan & Brinivasan (1990) have recently reported that glycyl residues are poor formers of extended structures m globular proteins. Superposition of the N-terminal domains of fiB2 and yB shows that the presence of Leu80 in

Structure of an Eye Lens Oligomeric fi-Crystallin

1073

(a 1

(b) Figure 5. (a) A stereo view of the o!carbon atoms from residues - 2 to 87 from the N-terminal domain of subunit 1 and residues 88 to 175 from the C-terminal domain of symmetry-related subunit 2 of flB2 (-) superpo_sed on the complete molecule of yB (- - -). The superposition was performed with the program XS6 written by A. Sali using a method originally described by McLachlan (1982). Every 10th residue of bB2 is labelled. The view is approximately perpendicular to the pseudo dyad relating N and C-terminal domains. (b) A stereo view of an a carbon trace of residues comprising the N-terminal domain of jB2. The p strands of motif 1 are labelled (a1 b,c,d,) as well as the c2 strand that completes the solvent-facing /?-sheet. The C” positions of the topologically equivalent Glu7 and Glu46, and Gln38 and Ile81 are indicated. Residues equivalent to every 10th position of yB are also labelled.

yB compensates for the absence of side-chain at position 1, that the interface residues Ile81 in both proteins are closely equivalent and that the conformations begin to diverge only at the ti torsion angle of the residue at position 82 (Fig. 6). Presumably

the presence of Pro82 in y-crystallins ensures the polyproline conformation is maintained, whereas the extended conformation of Lys82 in fiB2 helps to place Va183 into the domain interface region compared with Gln83 in yB, which projects into

Figure 6. A stereo view of residues 80 to 87 and residue 1 from gB2 ( ) superposed on the equivalent residues from yB (- - - ) taken from a MNYFIT superposition (Sutcliffe et al., 1987) of residues - 2 to 87 of flB2 with residues 1 to 87 from yB.

1074

R. Lapatto

et al.

Figure 7. A stereo view of a section of the refined electron density map (contoured at lo) and superposed atomic model of the backbone region of residues 80 to 82 of /3B2 showing a water molecule (HOH176) interacting with N82 and the charged groups of the ion pair Arg79-Glu147. -. subunit 1: -~---, subunit 2: - - -, hydrogen bond. solvent (Figs 6 and 13(b)). Bax et al. (1990) have shown that the extended connecting peptide in pB2 deviates most markedly from the position of equivalent residues of yB after position 86 (Fig. 6). It was argued that the different sequences of the interdomain linkers led to either local or more distal interactions. In pB2, N82 hydrogen bonds with a water molecule that in turn bonds with the carboxylate group of Glu147 (Fig. 7); this position is always a glutamate residue in /?-crystallins, whereas in y-crystallins it is always an arginine residue. When PBS and yB are compared, it can be seen that remarkable sequence changes are allowed while the secondary structure is maintained. For example, all /3-crystallins have a small side-chain at position 44 on the a2 P-strand; in PBS this residue is a glycine residue, whereas in y-crystallins it is conserved as a leucine residue. This does not result’ in a hole in bB2, as sequence changes at positions 5, 57 and 78 compensate such that the core is closepacked without altering the main-chain conformaalthough on a tion (Fig. 8(a)). A similar observation, smaller scale, was observed when a core-filling hydrophobic residue was mutated to a smaller residue in the Greek key P-barrel protein superoxide dismutase (McRee et al., 1990). Leu80 in y-crystallins contributes to some extent towards filling space occupied by Pro1 in gB2 (Fig. 6). Tt is noteworthy that there are significant sequence differences in p-sheet strands between D and y-crystallins at core-packing residues (44, 57 and 80) that are adjacent in sequence to those hydrophobic positions (43, 56 and 81) that form the interdomain interface. Comparison of the C-terminal domain cores of BB2 and yB shows that very small rearrangements of side-chains compensate for glycine 133 (which is topologically equivalent to Gly44). However, due to major differences in the

conformation of t’he cd arch in motif 3, there is no hydrophobic position in fiB2 equivalent to Phell6 of yB (Fig. 3). In this instance. an edge j-sheet strand b, is allowed to move in order to place the side-chain of Met103 into the space occupied by Phell6 in yB (Fig. 8(b)). (c) Loops and arches In fl,y-crystallins t’here is one hairpin turn between a and b strands in each sheet. a short arch connecting b and c strands and a long arch connecting c and d strands between each sheet (Fig. 5(b)). Each domain has a short arch joining the two motifs (d, to a2 and d, to aJ. The hairpin and the motif-joining region are strictly conserved in length, whereas most of the insertions and deletions in the p,y-crystallin family occur in the arches within motifs (Fig. 3). A careful comparison of these structures was made in order to assess their contribution to structural divergence. (i) The folded hairpin The folded ah hairpin is a conserved feature of the Greek key motif. Analysis of eight loops from the structures of /?B2 and yB show that key conserved residues direct the folding of this stretch of polvpeptide chain. The hydrophobic stabilization ‘is provided by an aromatic residue in the hairpin interacting with an aromatic side-chain on the a strand. The hairpin consists of eight residues and using the numeration of motif 1 has the conformat’ion 7A, 8B, 9Y. lOL, 11 L: 12B, 13E, 14P. The end residues also contribute towards the B-sheets and residues 7; 13 and 14 form the hinge of the fold. The ,/&sheet between a and b strands halts at residue 7. The A conformation allows its carbonyl group to make a tertiary interaction by hydrogen bonding

Structure of an Eye Lens Oligomeric

1075

fl-Crystallin

Figure 8. (a) A stereo view comparing the packing of a region of the N-terminal domain core of /IS2 ( ) by superposing the N-terminal domain of yB (- - -) using the program MNYFIT. This region of the core is filled with hydrophobic residues that are part of /?-sheets; each protein has a different selection of side-chains that results in close packed cores without necessitating a change in 0:carbon positions. (b) A stereo view of a region of the C-terminal domain core of bB2 ( ) superposed on the C-terminal domain of yB ( - - - ). This shows how hydrophobic sequence differences in arch regions result in changes to the a carbon atoms (118) including movement of a sheet residue (103) to compensate for deletion of Phell6 in bB2.

with the arch immediately prior to the d strand. Although residues 9 and 14 in the hairpin have P conformation, position 9 is frequently populated by a proline residue, whereas proline occurs only once at the position equivalent to 14 in all j?,y-crystallin motifs, emphasizing the need for an NH hydrogen bond to the side-chain of the conserved glutamate residue at position 7. Residues 9, 10 and 11 form a distorted overlapping p-turn. Although there is a hydrogen bond across the hairpin, between N8 and 012,

there

are

no

hydrogen

bonds

within

p-turns. Instead, the turn region is stabilized hydrogen bonds to the c and d b-strands within sheet (involving the conserved serine residue position 34), showing how tertiary interactions supersede more local stabilization.

the

by the at can

(ii) The short bc arch In motifs 2 and 4 of /?,y-crystallins, this arch comprises two residues in BP conformation. The carbonyl groups of residues 58 and 147 hydrogen bond with the hydroxyl groups of Tyr62 and 151, effectively keeping the polypeptide in extended conformation. This interaction appears to prevent the strand from turning back into the same sheet (Fig. 9(b)). The charged side-chains at position 58 (147) make salt bridges between subunits in /?B2 (Table 1, Fig. 13(a)). In motifs 1 and 3, the positions topologically equivalent to glycine residues 60 and 149 are deleted and

the

tyrosine

hydrophobic

residues

are replaced

by

small

side-chains (22 and 109). Loss of the

1076

R. Lapatto

et al.

Table 1 Subunit

B&unit 1 (X>Y> 2) Pro 1 Pro41 Ala39 Va156 Ser70A Glu53 Ile81 ValB3 Gly40 N 40 0 40 GlIl54 OE154 NE2 54 Va156 N 56 Glu58 OEl 58 OE2 58 Ser70 OG 70 Arg79 NH2 79 Asp84 N 84 ODl 84 SerS.5 OG 85 Gln86 N 86

(b)

Figure 9. Comparison of short, arches in /3B2 ( ) and yB ( - - - ) by superposition of (a) N-terminal domains where the bc arches in motif 1 show structural variation as a result of hydrophobic sequence differences elsewhere in the domain; (b) C-terminal domains where the equivalent. arches in motif 4 are very conserved. Position 18 and Cys22 in motif 1 are topologically equivalent to Leu146 and Tyrl51 in motif 4, respectively. The hydroxyl group of Tyrl51 hydrogen bonds with a local backbone atom, thus conserving this arch conformation.

glycine residues results in the arch extending into the region that is topologically equivalent to the c strand of motifs 2 and 4. This places three residues in ABP conformation in motif 1 in fiB2 and yB and motif 3 in yB, whereas in bB2 motif 3 this region has quite a different structure due to the insertion at 106A. This residue extends the b b-strand that, in turn, helps to fix the side-chain of His88 in the connecting peptide. Even though the secondary structures of this arch in motif 1 of /?B2 and yB are similar (Fig. 3), the C atoms of the conserved Cys22 are almost 3 A apart when PB2 and yB are superposed (see also Fig. 5). This appears to be a consequence of the smaller side-chain of Thr28 compared with Tyr28 in yB; cysteine 22 in /?B2 moves to compensate for this (Fig. 9(a)).

041 581 0 Xl OEl 53 0 54 SE2 54 0 52 N 54 N 59 NH1 79 N 82 0 82 0 84

(iii)

interactions within the BB2 dimer: the PQ interface Subunit 2 (x. 1 -y. -z)

Interaction

Trpl75 He170

Hydrophobic Hydrophobic Hydrophobic Hydrophobic Hydrophobic Hydrophobic Hydrophobic1

ml174 Va1132 Leu142 Thrl30 Leu145 Ile170 Met173 Gln174 Trp175 OEl 174 N 174 Leu145 Phe157 N 145 0 157 Gln143 OEl 143 Arg168 NH1 168 NH2 168 Glv141 3; 141 Glu147 OEl 147 Gly129 Thr130 OGl 130 9 129 Argl71 0 171 His88 NE2 88

H-bond H-bond H-bond H-hond H-bond H-bond H-bond H-bond H-bond H-bond

H-bontl H-bond H-bond

NE1 175

HOH 327

Wat,er bridge

x 143

HOH 222

Water bridge

0 158

HOH 221

Water bridge

0 143 OEl 174 OEl 147

HOH 22’3 HOH 209 HOH 176

Water bridge Water hridge Water bridge

OGl 130 OE2 147 NH1 169

HOH 231

Water bridge

HOH 177 0 170 x 130

Water bridge

The long cd arch

This stretch of polypeptide crosses over the wide angle of the wedge-shaped domain. lMost of the insertions and deletions in j?,y motifs occur in this region. The first residue is commonly proline with a carbonyl group that hydrogen bonds to the overlapping /I-turn region of a folded ab hairpin from the other motif of the domain. In yB motif 1, the arch takes the shortest route, which consists of 11 residues in irregular conformation including a single turn of 3,, helix supported by a single hydrogen bond. The arch in motif 1 of j?B2 accommodates one insertion (Gly28A) by assuming L conformation (Fig. 3). The arch in motif 3 of yB accommodates a further residue in the helical region, which becomes whereas in j?~B2 the insertion is further along. a314,

Structure of an Eye Lens Oligomeric j34rystallin

1077

Figure 10. A stereo view of residues 65 to 73 of /?B2 ( ) superposed on the equivalent region of yB ( - - - ) showing how insertion of 70A and 70B in pB2 is accommodated by the conformational change at position 69 (methionine in yB and threonine in /?B2) allowing a type 1 /?-turn conformation between residues 69 and 70B in /IBZ.

The arches in motif

residues. The insertion poorly

defined

2 and motif

in the initial

maps and was not improved and remodelling,

4 have

15 or 17

region 70 to 71 of PBS was 2.1 !I electron

density

either by refinement

or by simulated

annealing.

A type

I p-turn was indicated by the sequence propensity tables of Wilmot & Thornton (1988, 1990). Simulated annealing of the modelled turn region gave continuous electron density for the main-chain atoms of the turn and the side-chains of Ser70, Ser7OA and Arg71, although the density for the side-chain of Thr69 and Arg70B remained poorly defined, indicating some disorder in this region. The turn has a hydrogen bond between the carbonyl group at the i position and i+ 3 amide group. Unlike many type I p-turns, where a serine residue can mimic a proline residue at the i+ 1 position (Wilmot & Thornton, 1988, 1990), Ser70 in the /YB2 structure forms a hydrogen bond at an interface of the oligomer (Table 1). The /?-turn is also stabilized by the OG of i +2 forming a hydrogen bond with the main-chain amide group of i + 4 of Arg71. A threonine residue at position 69 in PB2 probably prevents the adoption of left-handed helical conformation that occurs in this position in the other motifs of /?B2 and yB (Fig. 10). The side-chains from the long cd arch, which are inserted into the hydrophobic domain core, are remarkably conserved between /?B2 and yB. For example, the side-chains of Tyrl18 of fiB2 and Leul18 from yB are superimposed, even though the C” atoms are 3 A apart (Fig. 8(b)). Clearly, the arch can vary enormously in terms of length and conformation while still maintaining similar core-filling positions (Tramontano et al., 1989). By contrast, in b-sheets the side-chain positions of hydrophobic residues, particularly when the N-terminal domains can be remarkably variable are compared, (Fig. 8(a)), although the backbone is much more conserved. This region of the B,y-crystallin structure exemplifies how small additions to a basic domain can be accommodated whilst maintaining a well-packed core. Analysis of the polypeptide backbone of these

long arches shows how less common conformations such as an L and G allow insertions and deletions into

various

main-chain

stabilized by tertiary interactions.

structures

that

are then

and sometimes quaternary

(iv) The intradomain da arch A single glycine residue joins two motifs to form a domain. Whereas Gly129 in bB2 and yB and Gly40 in yB are all in E conformation, Gly40 in jB2 is in B conformation. This is due to a cisgroline residue at position 41, a proline that is conserved in basic but not acidic B-crystallins (Fig. 11). In the database 9% of cis prolines are preceded by a glycine residue (MacArthur & Thornton, 1991). Presence of cis proline in such a position in PBS may affect the kinetics of assembly of motifs into domains. (d) Structure and stability /?B2 readily unfolds in denaturing media (Maiti et al., 1988), whereas y-crystallins are remarkably resistant to unfolding (Mandal et al., 1987) and recently it has been shown that the N-terminal domain of yB is more stable than the C-terminal domain (Rudolph et al., 1990). A contribution towards the greater stability of yB over BB2 may

Figure 11. The conserved glycine residues at position 40 at the d,a, arch connecting motifs are in different conformations in order to compensate for cis proline 41 of /?B2 ( ~ ) replacing cysteine in yB ( - - - ).

R. Lapatto

1078

et al.

Figure 12. A stereo view of residues 84 to 88 from the connecting peptides of’/?SZ dimer close to the P axis showing the hydrogen bond (- - -) interactions.

come from the cd arches from motif 4. Although they are the same length and of similar conformation, this region in yB has three additional interacting hydrophobic residues at positions 154, 155 and 160 that extend the core. The replacement of Leu133 and Trp157 in yB by glycine and phenylalanine residues, respectively in pB2 might also contribute towards a decreased stability of fiB2 compared with y domains. The high stability of the N-terminal domain of yB may be due to its large number of aromatic-sulphur interactions (Summers et al., 1984).

5. Structure of the Dimer and Tetramer (a) The P& interface Although the domains of a /?B2 subunit are not in close contact, N and C-terminal domains from the two symmetry-related subunits within the dime1 are. The two subunits related by a crystallographic dyad called P (Fig. 2) interact such that a surface area, named PQ, of 2200 A2 per subunit is removed from solvent. Interactions between the antiparallel connecting peptides and between the connecting peptides and C-terminal extensions are related by the P axis which passes close to Gln86 (Fig. 12). The approximate axis of symmetry, relating the N-terminal domain of one subunit to the adjacent C-terminal domain of another subunit, does not lie parallel to any of the principal crystallographic axes that define the tetramer (Fig. 2(c)). Table 1 indicates that the majority of the subunit contacts that maintain the PQ interface in PB2 are hydrophilic: there are 26 hydrogen bonds and 16 water molecules between subunits. The extended conformations of the flB2 connecting peptides are stabilized by numerous main-chain-side-chain hydrogen bonds between the antiparallel chains (Fig. 12) and it is these interactions, unique to the dimer; that provide additional stabilization relative to a monomer. Of particular note is that the side-chain of His88

donates a proton to the carboxylate group of aspartate 106A whilst, accepting a proton from the backbone amide moiety of N86 close to the P 2-fold axis. All basic b-crystallins have the combination of insertion of Aspl06A and His88, but this is not the case for acidic p or y-crystallins. The connecting peptide, AsplO6A and the C-terminal extensions, which are all additional to the compact domains, are involved in interactions close to the dimer twofold axis. The stereo views in Figures 5(a) and 13 show the subunit interface together with side-chains formed from p-sheet 2 from subunit 1 associated with a-sheet 4 from subunit 2 of /?B2 compared with the intramolecular, interdomain association of b-sheets 2 and 4 from y-crystallin. Although the sheet orientations are very similar, the subst,itution in fl-crystallins of two glutamate residues for arginine residues in y-crystallins in positions such that, they are close to two arginine residues from the other sheet at the interface, results in two partially buried ion pairs in the dimer (Fig. 13(a)). The hydrophobic contribution to interface stabilization is redistributed slightly in that, B-crystallins have smaller side-chains at positions 43 and 56 compared with y-crystallin (Fig. 13(b)) and that Va183 and Met173 contribute in pB2, whereas Gln83 and Phe173 in y-crystallin are directed away from the interface. The hydrophobic contribution also involves the start of the C-terminal extension (Trp175) and the beginning of the first motif (Prol) and is influenced by the side-chain position at the end of motif 2, Pro80 (Figs 6 and 13(b)). Although the extensions of PB2 sequence are not primarily involved in holding the subunits together as was originally predicted (Wistow et al., 1981; Slingsby et al., 1988), they do make a limited contribution and in a region that might influence whether or not t,he OR2 domains make inter- or intradomain interactions. Superposition of equivalent N and C-terminal domain c” co-ordinates of PB2 and yR reveals the overall similarity of the P-sheets at the interface (Figs 5(a) and 13). The two polypeptides have very

1079

Structure of an Eye Lens Oligomeric /3-Crystal&

(a)

Figure 13. Stereo views of the domain interface drawn approximately perpendicular to the pseudo 2-fold relating the N and C-terminal domains. The N-terminal domain of subunit 1 and the C-terminal domain of subunit 2 of /?B2 (-) were superposed onto the complete molecule of yB (- - -) using the program XS6. (a) Replacement of Arg58 and Arg147 in yB with glutamate residues in pB2 results in 2 ion pairs stabilizing the intersubunit interface in flB2. (b) Differences in the interface hydrophobic residues: replacement of Met43 and Phe56 of yB with smaller valine residues in bB2 is compensated by the addition of Va183 and Trp175 in /IB2 to the interface.

up to residue 82 (Fig. 6). similar structures Although positions 83, 84, 85 and 86 differ locally in torsion angles, the equivalent C” positions are within 2.5 A. When the N-terminal domain of /IB2 is superposed on the complete molecule of yB, glutamate 87 from j3B2 is a great distance from threonine 87 of yB (Fig. 6). In the structure of j?B2, Glu87 is closely positioned next to Ile106. However, if the bB2 connecting peptide were to take the same course as it does in yB, then Glu87, superimposed on Thr87, would be in close contact with AsplOBA (Fig. 14). The allowance of 2 to 3 A displacement between equivalent positions in the two connecting peptides thus places the conserved glutamate residue of fl-crystallins next to He106 rather than another acidic side-chain. This repulsive interaction would doubtless destabilize monomer formation in PBS-crystallin.

(b) The QR interface The PBS dimer is related to a further dimer by two orthogonal dyads Q and R to give a tetramer with crystallographic 222 symmetry (Figs 2(d) and (e)). This is achieved by burying a contact area named QR of 1350 A2 per subunit. Table 2 indicates the main intermolecular contacts that stabilize this interface. There are interactions between the N-terminal domain of subunit 1 and the C-terminal domain of subunit 4, whereas the contacts between subunit 1 and subunit 3 are between corresponding domains.

The

heart

of

the

2-fold

contact

is

Tyr139, which has its side-chain running approximately parallel to the Q axis. The flat surfaces of Gly60, Pro63, Pro110 and Pro137 allow the close approach of subunits. There are less than half the number of hydrogen bonds in the QR compared

1080

R. Lapatto

Figure 14. A stereo view of the region around glutamate superposed superposed

using MNYFIT. on Thr87 would

If the conformation clash with

its inserted

residue

Table 2

Subunit 1 (x. y> 2)

interactions

within the /lB2 the QR interface

Subunit 4 (I-x,y, -z)

tetramer.

Interaction

A. Interactions between subunit I and subunit 4 Glu58 Lys59 Gly60 Glu61 Tyr62 Pro63 Ser67 Lys59 0 59 Mu61 061 S 61 Tyr62 OH 62 Arg64

Pro110 Gly138 His113 Ala114 His113 Pro137 Tyr136 Serlll OG 111 Gln165 NE2 165 OEl 165 Arg140 NH1 140 Tyr136 OH 136

0 37 0 59

ND1 113 N 113 OEl 165

Subunit 1 (x3 y> 2)

Subunit 3 (1 --x, 1 -y,

Hydrophobic Hydrophobic Hydrophobic Hydrophobic Hydrophobic H-bond H-bond H-bond H-bond H-bond HOH 183 HOH 198

U. Interactions between subunit I and subunit 3 Arg168 Tyr139 Arg70B NH1 7OB

Pro110 Gly138 Tyrl39 Ser7OA 0 70A OH 139 0 138 OH 143 NH1 168

OD2 108 0 139 NH2 140

Water bridge Water bridge

Interaction

-z)

HOH 259 HOH 191 HOH 190

87 from

,9B2 when residues

86 to 174 from

PI32 and yB are

of /?B2 followed t,he same course as yB. then 011187 from /?B2 when

with the PQ interface and ten water bridges. Tetramer formation results in the deeper burial of one of the PQ interdomain ion pairs (Glu58Arg168). Although the extensive area of the QR contact implies that PBS could form tetramers at very high protein concentration, many side-chains that are partially buried make no favourable interactions. This is consistent with the observation that

Dimer

et al.

Hydrophobic Hydrophobic H-bond Water bridge Water bridge Water bridge

of AsplOGA.

-

fiB2;

- - -

the molecules are dimers 70A, 70B is involved interaction. (c) Aolvent

yK.

in solution. this in

The insertion dimer-dimer

structure

There are 16 water molecules buried between the PQ interface (Table 1) and ten molecules in the QR interface (Table 2). Apart from three water molecules that are associated with lattice tetramertetramer contact,s, there are 45 water molecules located that participate in the first hvdration layer of the tetramer surface and three in the second layer. ln the first, hydration layer t)here are 17 contacts to main-chain oxygen atoms compared with 13 t,o main-chain nitrogen atoms and I6 contacts to side-chain oxygen atoms compared with 14 to side-chain nitrogen atoms. (d) The NH2 and CJOOH extensions Although there are 15 residues in the NH,terminal extension of BB2 and a further 11 residues from the COOH terminus compared with yB, only residues - I and -2 and Trp175 were visible in the electron densit,y map. The actual crystal from which the high resolution data were collected was subsequently subjected t)o sodium dodecyl sulphat,epolyacrylamide gel electrophoresis and compared with a sample of /lB2 that had received limited tryptic digestion as control (Kerbers et al.. 1983). The protein from the crystal clearly migrated as a complete molecule (0. A. Bateman, unpublished results) showing no evidence of cleavage after Arg’lOR, Arg71 or Arg178. Tt is concluded that, the extensions in most of the molecules in the crystal do not occupy fixed positions in the lattice. Additional regions of positive electron density were seen at the surface of the subunits in the difference maps but were generally disconnected; several attempts were made to model pieces of the arms into them but no convincing positions were found. A total of 29 water molecules were placed in positions too distal from the protein surface for hydrogen bonding inter-

Structure of an Eye Lens Oligomeric /?-Crystallin actions and these water positions may represent the apparently mobile or disordered extensions.

6. Discussion /?,y-crystallins demonstrate how evolution can result in selection of sequences that alter the size of oligomers while retaining not only the basic framework of the domains but also their interacting surfaces. This evolved system resembles the immunoglobulin domains with close interactions between domains in different chains and flexible linking connecting peptides, whereas the structure of y-crystallin has similarities to the engineered “single chain Fv fragments” (Winter & Milstein, 1991). The ability to vary aggregate size by modifying only a small region of sequence is probably a molecular adaptation to fit their packing role in the lens where refractive index gradients are required for optical properties (Fernald & Wright, 1983). Bearing in mind the very similar way in which N and C-terminal domains interact in p and y-crystallins, this raises the critical question of what stabilizes an entropically unfavourable dimer compared with two monomers. There are many sequence differences between monomers and dimers that result in more contacts between N and C-terminal domains in the fiB2 structure contributing to dimer stability. However, many of these interactions would also stabilize a monomeric form of PB2 and so cannot be responsible for favouring a dimer rather than two monomers. The interactions unique to an oligomer are from the extended connecting peptide and from the C-terminal extension. We have argued that sequence differences between /I and y-crystallins in the connecting peptides are primarily responsible for the different arrangements (Bax et al., 1990). In this analysis we have suggested a further structural feature, the insertion of an aspartate residue in the C-terminal domain of basic p-crystallins, that would destabilize a monomeric arrangement of domains. All residues in the connecting peptide of yB occupy similar positions to one or other of the two fiB2 subunits in the dimer, with the major structural difference stemming from one torsion angle difference in residue 86. This represents a most remarkable piece of natural protein engineering whereby maximum consequences in terms of exchange of domain partners is achieved with minimum change to the rest of the protein. This mechanism is absolutely dependent on exploitation of the high internal symmetry of these crystallins. Whereas it is obvious from the observed folded structures that sequence differences in the connecting peptides can have radical effects on domain organization, it is less easy to assess the structural consequences of sequence differences that affect the folding pathway. The hydrophobic side-chains that stabilize the /IS2 domain cores are widely dispersed through the sequence; in fact, major interaction sites are from the start and end regions of the motifs and domains. For example, hydrophobic residues at the start of

1081

the a strands make lateral interactions with hydrophobic residues from the end of d strands from the same sheet as well as packing opposite d and a strand residues from the pseudo symmetry-related sheet within a domain; proline 1 at the start of motif 1 interacts with Trp175 from the C-terminal extension of a symmetry-related subunit. Dill (1990) has reviewed the evidence that the dominant force in protein folding is hydrophobic in origin and argues that approximate tertiary interactions are essentially formed before the regular secondary interactions make their contribution to stability. If these a and d strands were the first-formed tertiary inter actions there would be some analogy with recently characterized folding intermediates. The helical ends of apomyoglobin have been proposed to interact in a stable intermediate form while the intervening region is unfolded (Hughson et al., 1990), as have the terminal helices in an early kinetic intermediate of cytochrome c (Roder et al., 1988) and also the terminal helices of lysozyme (Miranker et al., 1991). For a highly symmetrical /?-sheet protein such as flB2, if such a folding pathway occurred, it could be envisaged as being so highly co-operative that there would be little chance of trapping intermediate states. On the other hand, if one of the regular regions of secondary structure were first formed, then it might stand a chance of detection. Such a region would be the anti-parallel ab P-strands and enclosed p-hairpin turn. Because of the symmetrical conformat,ions of /IS2 and yB there are multiple copies of similar motifs that can be compared even though most amino acid side-chains can vary, In comparing the structures of the various motifs and domains, what is observed is the variety of ways in which sequence changes in one place are compensated by a small reorganization elsewhere. The analysis of these two proteins has shown that the /3-sheet secondary structure is remarkably resilient, particularly in the N-terminal domains, towards changes in the core packing hydrophobic side-chains. Presumably, this reflects the many contributing forces which stabilize P-sheet sandwiches. Where there are differences between the two proteins in the P-sheet structure (residues 22 and 103) they can be ascribed to hydrophobic changes in a long arch. In the case where b-strands are lengthened in a sheet in pB2, this is due to an adjacent insertion that appears to contribute to the dramatic conformational change of the connecting peptide. The four ab hairpins and two da arches from both /?B2 and yB do not allow any insertions or deletions and each has an absolutely conserved glycine residue. By contrast, the region where major structural variation can be tolerated is the long arch connecting c and d strands. Most of the insertions in the fl-crystallin family occur in this region (van Rens et al., 1991), at a site close to both the PQ and QR interfaces, suggesting a mechanism for the evolution of more complicated structures. Long loops in between p-strands are involved in variable domain dimerization in the immunoglobulin family,

R. Lapatto et al.

1082

whereas they are shortened in the monomeric CD4 structure (Wang et al.: 1990; Ryu et al., 1990). Where insertions are involved in contacts stabilizing the fiB2 dimer and tetramer, they are likely to contribute towards the specificity of heterooligomer formation in the same way that the insertion of aspartate 106A modifies domain interactions. The acidic subunit PA3, which is truncated at position 172 and has a longer N-terminal extension of quite different sequence, readily forms higher oligomers with fiB2 at low protein concentration (Slingsby & Bateman, 1990). It may be that par& cular combinations of insertions and extensions favour increased protein interactions that stabilize further aggregation. The financial support of the Medical Research Council is gratefully acknowledged. We thank Louise Morris for producing Fig. 4. We also acknowledge many useful discussions with our colleagues Orval Bateman, Linda Miller, David Moss, Shabir Najmudin and Ian Tickle.

References Bax.

B. & Slingsby, C. (1989). Crystallization of a new form of the eye lens protein pB2-crystallin. J. Mol. Biol. 208, 715-717. Bax. B., Lapatto, R., Nalini, N., Driessen. H.. Lindley. P. F., Mahedevan. I)., Blundell, T. L. & Slingsby, C. (1990). X-ray analysis of BBB-crystallin and evolution of oligomeric lens proteins. Nature (London), 347, 776-780. Benedek, C. B. (1971). Theory of transparency of the ryr. Appl. Optics, 10. 459-478. Berbers, G.. Brans. A.. Hoekman. W.. Slingsby, (1.. Bloemendal, H. & de Jong, W. (1983). Aggregation behaviour of t,he bovine /Gcrystallin Bp chain studied by limited proteolpsis. Biochim. Biophys. Actcr. 748, 213-219. Berbers. U. A. M.. Hoekman. W. A.. Bloemendal. H.. de *Tong, W. W., Kleinschmidt. T. & Braunitzer. (:. (1984). Homology between the primary structures of the major B-crystallin chains. Eur. J. Biochem. 139.

467-479. Blundell, T., Lindley. l’., Miller, I,., Moss. I)., Slingsby. C., Tickle, I.. Turnell, B. & Wistow. G. (1981). The molecular structure and stability of the eye lens: X-ray analysis of y-crystallin 11. BakurP (London), 289. 771-777. Briinger. A. T. (1988). Crystallographic refinement by simulated annealing: application to a 2% A resolution structure of aspartate aminotransferase. J. Mol. Riol. 203. 803-816. Chirgadze, Y. N., Nevskaya. N. A.. Fomenkova. N. I’.. Nikonov. S. V., Sergeev. Y. V.. Brazhnikov. E. V.. Garber. M. B., Lunin. V. Y.. Urzumtsrv, A. P. & Vernoslova. E. A. (1986). Spatial structure of gamma-crystallin IIIb from calf eye lens at 2.5 A resolution. Dokl. Akad. Sauk, (‘.f?.rS’.R. 290. 492-495. Crowther. R. A. & Blow. I). M. (1967). A method of positioning a known molecule in an unknown cry&al structure. Acta Crystallogr. 23, 544-548. Delaye. M. & Tardieu. A. (1983). Short-range order of crystallin proteins accounts for eye lens transparency. Nature (London), 302, 415-417.

Dill,

K. A. (1990). Dominant, forces in protein folding. Biochemistry, 29, 7133-7155. Driessen, H. P. C”.. Herbrink. P.. Bloemendal. H. & de .Jong, W’. W. (1981). Primary structure of the bovine P-crystallin Bp chain. Internal duplicabion and homology with ?;-crystallin. Eur. ,I. Hiochum. 121,

K-91. Driessen. H. P. C’.. Haneef. I.. Harris, (:. W.. Howlin. B.. Khan, (:. &, Moss, D. 8. (1989). R,ESTRAIN: restrained structure-factor least-squares refinement program for macromolecular structures. d. .4 ppl. Crystallogr. 22, 510-516. of Fernald, R. 1). &, Wright, S. E. (1983). Maintenance optical quality during cryst,alline lens growth. ,Vaturr (London), 301. 618-620. Gorin, M. B. & Horwitz, J. (1984). Cloning and characterization of a
Structure of an Eye Lens Oligomeric jh3ystallin repetitive protein structure. Proc. Nat. Acad. Sci., C.S.A. 75: 2574-2578. Roder, H., Eliive, G. A. & Englander, S. W. (1988). Structural characterization of folding intermediates in cytochrome c by H-exchange labelling and proton NMR. Nature (London), 335, 700-704. Rudolph, R., Siebendritt, R., Nesslaiier, G., Sharma, A. K. & Jaenicke, R. (1990). Folding of an all-p protein: independent domain folding in yII-crystallin from calf eye lens. Proc. Nat. Acad. Sci., U.S.A. 87. 4625-4629. Ryu, S.-E., Kwong, P. D., Truneh, A., Porter, T. G., Arthos, J., Rosenberg, M., Dai, X., Xuong, N.-h., Axel, R., Sweet, R. W. & Hendrickson, W. A. (1990). Crystal structure of an HIV-binding recombinant fragment of human CD4. Nature (London), 348, 419-426. Slingsby, C. (1985). Structural variation in lens crystallins. Trends Biochem. Sci. 7, 281-284. Slingsby. C. & Bateman, 0. A. (1990). Quaternary interactions in eye lens b-crystallins: basic and acidic subunits of j?-crystallins favor heterologous association. Biochemistry, 29, 6592-6599. Slingsby, C., Miller, L. R. & Berbers, G. A. M. (1982). Preliminary X-ray crystallographic study of the principle subunit of the lens structural protein bovine j-cr,ystallin. J. 1Mol. Biol. 157, 191-194. Slingsby. C.. Driessen, H. P. C., Mahadevan, D., Bax, B. & Blundell. T. I,. (1988). Evolutionary and functional relationships between the basic and acidic p-crystallins. Exp. Eye Res. 46, 375-403. Summers. L., Wistow. G., Narebor, M., Moss, I)., Lindley, P., Slingsby, C., Blundell, T., Bartunik, H. & Bartels, K. (1984). X-ray studies of the lens specific proteins: the crystallins. In Peptide and Protein Reviews (Hearn, M. T. W.. ed.), vol. 3, pp. 147-168, Marcel Dekker. New York. Sutcliffe. M. J., Haneef, I.. Carney, D. & Blundell, T. I,. (1987). Knowledge based modelling of homologous proteins, part T: three dimensional frameworks derived from simultaneous superposition of multiple structures. Protein Eng. 1, 377-384. Tickle, I. J. (1985). In 1MolecuZar Replacement (Machin,

1083

P. A., ed.), pp. 22-26, Science and Engineering Research Council, Daresbury, U.K. Tramontano, A., Chothia, C. & Lesk, A. M. (1989). Structural determinants of the conformation of medium-size loops in proteins. Proteins: Struct. Func. Genet. 6, 382-394. van Rens, G. L. M., Driessen, H. P. C., Nalini, V., Slingsby, C., de Jong, W. W. & Bloemendal, H. (1991). Isolation and characterization of the cDNAs of the last two acidic /?-crystallins, /IA2 and /?A4: heterologous interactions in the predicted fiA4-/?B2 heterodimer. Gene, 102, 179-188. Wang, J., Yan, Y.. Garrett, T. P. J., Liu, J., Rodgers, D. W.. Garlick, R. L., Tarr, G. E.. Husain, Y., Reinherz, E. L. & Harrison, S. C. (1990). Atomic structure of a fragment of human CD4 containing two immunoglobulin-like domains Nature (London), 348, 41-418. White, H. E., Driessen, H. P. C., Slingsby, C., Moss, D. S. & Lindley, P. F. (1989). Packing interactions in the eye-lens: structural analysis, internal symmetry and lattice interactions of bovine yIVa-crystallin. J. Mol. Biol. 207, 217-235. Wilmot. C. M. & Thornton, J. M. (1988). Analysis and prediction of the different types of j-turn in proteins. J. Mol. Biol. 203, 221-232. Wilmot, C. M. & Thornton, J. M. (1990). /?-Turns and their distortions: a proposed new nomenclature. Protein Eng. 3. 479-493. Winter, G. & Milstein, C. (1991). Man-made antibodies. Nature (London), 349, 293-299. Wistow, G. ,J. & Piatigorsky, J. (1988). Lens rrystallins: the evolution and expression of proteins for a highly specialized tissue. Annu. Rev. Biochem. 57. 479-504. Wistow, G., Slingsby, C., Blundell, T., Driessen, H., de Jong, W. 8t Bloemendal, H. (1981). Eye lens proteins: the three dimensional structure of fi-crystallins predicted from monomeric y-crystallins. FEBS Letters, 133, 9-16. Wistow. G.. Turnell, B., Summers, I,., Slingsby, C., Moss, D.. Miller. L.. Lindley, P. & Blundell, T. (1983). X-ray analysis of the eye lens protein y-11 crystallin at. 1.9 A resolution. J. Mol. Biol. 170. 175-202.

Edited by W Hendrickson