Báo cáo Y học: The evolution of monomeric and oligomeric bc-type crystallins potx

REVIEW ARTICLE The evolution of monomeric and oligomeric bc-type crystallins Facts and hypotheses Giuseppe D’Alessio Dipartimento di Chimica Biologica, Universita ` di Napoli Federico II, Naples, Italy The case of homologous monomeric c-type and oligomeric b-type crystallins has been described and analyzed in evo- lutionary terms. Data and hypotheses from molecular gen- etics and structural investigations converge and suggest a novel three-phase model for the evolutionary history of crystallin-type proteins. In the divergent cascades of mono- meric and oligomeric crystallins, a pivotal role was played by alterations in the gene segments encoding the C-terminal extensions and the intermotif or interdomain linker peptides. These were genomic hot spots where evolution experimented to produce the modern variety of bc-crystallin-type quater- nary structures. Keywords: crystallins; evolution; quaternary structure; introns late; introns early. The question of how oligomeric proteins evolved has gained renewed interest in the last few years [1–9]. Although the possibility cannot be excluded that some proteins emerged first as functional aggregates and later dissociated into functional monomers, the available evidence suggests that divergent evolution more often used the association of protein protomers into oligomers to vary and enrich the cell repertoire of structures and functions. Evidence for this evolutionary path can be seen in the Ôhydrophilic effectÕ recorded at intersubunit interfaces [3], i.e. a surprising, significant presence of polar and charged residues at oligomeric interfaces. This can be readily interpreted as the result of the association of previously exposed, hydro- philic surfaces (from a monomer) into solvent-excluded interfaces (in an oligomer). It has been argued that the alteration of a protein surface to render it adhesive for the generation of oligomers, would be too long an evolutionary process, as it would require multiple mutations in the gene encoding the ancestral monomer [9]. It was therefore proposed that evolution used pre-existing interdomain interfaces that after a ÔswapÕ of domains between monomers would be readily reconstituted as intersubunit interfaces. This would induce the association of monomers into oligomers without the need for a lengthy process of substituting one residue after another to build an adhesive interface. However, it has been noted that a swap of domains between monomeric ancestors is not an evolutionary event per se, but rather the outcome of one or more mutational events in the monomeric ancestor: these events could then prime a swap of domains [3]. Monomeric proteins have been transformed artificially into dimers by inducing the displacement of terminal helices, which deter- mined the helix segments between two monomers [10–12]. However, to make the swap permanent and the dimers stable, mutations had to be engineered into the cDNAs encoding the proteins [11,12]. These were, naturally, experiments of in vitro evolution, in which a single genetic alteration was sufficient to induce oligomerization. When we compare a present-day set of homologous proteins, one monomeric the other oligomeric, what we see when we compare the amino-acid sequences of the two proteins are merely amino-acid substitutions. Some of these may not related at all to the monomer to oligomer transition, and it is difficult and risky to discern the changes presumed to be significant for the transition. However, if we could have observed the entire process of evolution of a monomeric protein into a dimer, we would have assigned to each gene alteration responsible for the evolutionary transition a different status in the evolutionary mechanism. A ÔprimaryÕ mutation would top the hierarchy, as the single event responsible for the step of no return towards the new, oligomeric structural organization. Although such a primary event would have been essential, it may not have been sufficient to engender oligomerization. On the other hand, it may not be easy, or even possible, to decipher in the structure of a present-day oligomer what was the primary mutation originally responsible for oligomerization. Besides investigations of mutational events as revealed by amino-acid substitutions in homologous proteins, another tool might be useful to shed light on putative ancestors of present-day protein oligomers. It has been surmised [3,8] that the analysis of the refolding mechanism by which denatured, unfolded polypeptide chains fold back into oligomers may shed light on the evolutionary history of the oligomers, as this might be recapitulated in the pathway of oligomer refolding. The monomeric c-crystallins and the evolutionarily related dimeric b-crystallins provide an interesting case study in the discussion of the evolutionary transition from monomeric to oligomeric proteins. They are one of the Correspondence to G. D’Alessio, Dipartimento di Chimica Biologica, Via Mezzocannone, 16, 80134 Napoli, Italy. Fax: + 39 081 5521217, Tel.: + 39 081 2534731, E-mail: dalessio@unina.it Abbreviations: EDSP, epidermis differentiation-specific protein; TKR, tyrosine kinase receptor. (Received 17 December 2001, revised 8 April 2002, accepted 17 May 2002) Eur. J. Biochem. 269, 3122–3130 (2002) Ó FEBS 2002 doi:10.1046/j.1432-1033.2002.03004.x present-day sets of monomeric and dimeric homologous proteins on which the Ô3D domain swapping modelÕ has been based [9], and special attention has been given to their evolutionary history [13–15]. Crystallins were so named when they were recognized as the proteins that provide the crystalline lens of the vertebrate eye with its indispensable transparency and unique refract- ive properties (reviewed in [16]). They are long-lived proteins, as lens cells live as long as their host organisms. They also have lived a very long evolutionary history, as their primitive ancestors can be traced back to the diver- gence of protozoa. Crystallins are not confined to the lens; in various taxa, crystallin genetic material was ÔrecruitedÕ in different tissues to encode proteins serving functions, as diverse as those of enzymes and antistress proteins [17–19]. Three major classes of crystallins are common to the eye lens throughout the vertebrates: the a-, b-andc-crystallins. The latter two classes are made up of homologous proteins, and constitute the superfamily of bc-crystallins. For c-crystallins, six genes (cAtocF) have been identified, encoding 21-kDa monomeric proteins, and a gene for a cS crystallin, previously classified as bS-crystallin. b-crystallins, encoded by five to seven genes, depending on species, may form aggregates of up to 200 kDa that consist of acidic type (bA1 to bA4) and basic type (bB1 to bB3) subunits, 23–33 kDa. The c-crystallins have short C-terminal peptide extensions, whereas b-crystallins possess long N-terminal extensions, and the basic b-type subunits also have C-terminal extensions. In this review, only genetic and structural aspects of monomeric or oligomeric bc-crystallins likely related to their evolutionary origin will be discussed. For other aspects, the reviews cited above should be consulted. STRUCTURAL FEATURES OF MONOMERIC AND DIMERIC CRYSTALLINS To date, the available 3D structures are those of bB2-crystallin [20–22], and of cB- [23–25], cE- [26], and cC-crystallin [27]. Formerly, the latter were called cII-, cIIIB- and cIV-crystallin, respectively. cB-crystallin is monomeric, as are all c-crystallins; bB2-crystallin is a dimer in solution, although its structural unit in the crystal lattice is a tetramer, made up of two dimers, and the likely assembly of this protein in the lens is that of higher heteroligomers. However, monomeric cB-crystallin and dimeric bB2-crys- tallin will be considered here as the monomeric and dimeric prototypes for the respective families of c-andb-crystallins, andsimplyreferredtoasc-type or b-type crystallins, respectively. Both monomeric c-type crystallin and the subunit of dimeric b-type crystallin are composed of two domains, an N- and a C-terminal domain (termed N- and C-domains). Each domain is made up of two homologous ÔGreek keyÕ b strand motifs; motifs M1 and M2 in the N-domain, and motifs M3 and M4 in the C-domain. In both the c-monomer and the b subunit, the four motifs and the two domains are organized symmetrically, with local intermotif and interdomain pseudo-dyads. However, in dimeric b-crystallin the topological equivalents of the two domains of monomeric c-crystallin are domains from Fig. 1. The structures of: (A) bB2-crystallin (PDB 1BLB), the dimeric b–type prototype, and (B) cB-crystallin (PDB 4GRC). The monomeric c-type prototype. Fragments of cB-crystallin are shown to illustrate schemat- ically the structures of (C) two-motif/one- domain, and (D) one-motif putative crystallin ancestors. The interdomain linker peptides are coloredinred. Ó FEBS 2002 On the evolution of crystallins (Eur. J. Biochem. 269) 3123 different subunits (see Fig. 1). Two types of dimers may thus be viewed in a b-type crystallin: the orthodox dimer, made up of the two subunits with the two N • CandN¢•C¢ domains covalently linked through linker peptides ( |)in antiparallel fashion: and a pseudo-dimer made up of noncovalently associated N • C¢ and C • N¢ domains from the two subunits, which reproduce the topological association of the N- and C-domains of c-crystallin (see Fig. 1). The linker peptide segments that connect the N- and C-domains have very different conformation in the c-monomer and in the b subunits. In the monomeric c-type crystallins, the linker peptide bends to reach from the N-domain through the C-domain as in N • C. In dimeric b-type crystallin instead, the two linker peptides have an extended conformation; as in the scheme above, they run antiparallel on either side of the pseudo twofold axis relating the two-domain pseudo- dimeric structure made up of N • C¢ and C • N¢ (see Fig. 1). When present, N- and C-terminal extensions are not entirely defined in the structure of crystallin proteins, as they are flexible, without unique conformations, with the excep- tion of the proximal segments of the C-terminal extensions. MOLECULAR GENETICS STUDIES: FACTS AND HYPOTHESES Owing to the stringent necessity to conserve the critical function of providing the lens, by an appropriate arrange- ment of protein aggregates, with the precision of an optical measuring instrument, lens crystallins have been subjected to severe selective pressure in the course of their evolution. This is indicated by the very low substitution rates registered in the vertebrate crystallin genes, especially in those coding for b-crystallins, and by the unusually very similar substi- tution rates recorded for internal and surface regions of these proteins [14]. The latter finding can be interpreted as indicative of the importance of surface, intermolecular interactions among the lens proteins. A striking exception to this general sequence conservation rule are the high substitution rates that have been recorded only for the sequences encoding the interdomain linker peptides and the N- and C-terminal extensions. These findings certainly have an evolutionary significance. In both b-andc-type crystallin genes, the sequence coding for the interdomain linker peptide is interrupted by an ÔinterdomainÕ intron. In the b-type crystallin genes, ÔintermotifÕ introns are also present. Thus in b-type genes, each motif (M1, M2, M3, M4) is encoded by a separate exon, whereas in the c-type genes the pairs of adjacent motifs (M1/M2 and M3/M4) are each encoded by a single exon (Fig. 2). Sequence similarities are higher when motif M1 is compared with M3 (termed A type motifs), or M2 with M4 (B type motifs), which results in a ABAB pattern [35]. The structural similarity and topological equivalence between motifs and between domains, and the significant degree of sequence identity between domains, even higher than between motifs, led to the proposal [23] that the evolutionary path of c-crystallin started with a one-motif ancestor. Then, upon gene duplication followed by fusion, a two-motif/one-domain protein evolved, to be followed, after a second duplication-fusion step, by the two-domain c-type proteins. Subsequent findings from protein sequence, gene sequence, and structural studies [7,13,14,20] have strengthened and expanded this view. This evolutionary path based on primary and tertiary structure homologies, and consisting of two main events of gene duplication, each followed by fusion, is supported by the identification in distant phyla of homologous genes encoding proteins that can be related to putative crystallin ancestors. A one-domain crystallin-like fold has been found in a protein (spherulin 3a) from a slime mould, with a significant sequence identity and a high structural similarity with c-crystallin domains [61,65]. Interestingly, in the amino-acid sequence of spherulin 3a motif M1 is not N-terminal as in bc-crystallin sequences, but C-terminal to motif M2 (Fig. 2). Another case of a one-domain crystallin fold has been identified [30] in Streptomyces metallo- proteinase inhibitor (SMPI), with a clear relationship in three-dimensional structure to bc-crystallins. In this protein, a significant, albeit weak, sequence similarity has been detected between its N-terminal motif and M1 motif of bc-crystallins, but no similarities were found between its Fig. 2. A scheme of the arrangements of motif encoding gene sequences in crystalline-type genes. SPHE-, STRE-, S-, C-, b-andc-type nota- tions indicate motif arrangements in: spherulin 3a, Streptomyces protease inhibitor, S-protein-, G. Cydonium protein, b-andc-type crystallin, respectively. Motifs are shown as boxes and their numbers (M1 through M4) are those typical of both b-type and c-type crys- tallins, assigned to the other genes on the basis of homologies. Two- motif domains are formed by adjacent motifs. Thin and thick bars represent intermotif and interdomain introns, respectively. Dotted line segments between domains or motifs indicate that it is not known if an intermotif or an interdomain intron is present in that gene. 3124 G. D’Alessio (Eur. J. Biochem. 269) Ó FEBS 2002 C-terminal domain and any other known crystallin-type motif sequences (in Fig. 2, this motif is marked as MX). A crystallin-type one-domain fold has also been proposed for a yeast toxin [31], and for a Streptomyces toxin-like protein [32]. However, in these cases the possibility of convergent evolution may not be excluded [33]. Two-domain crystallin-like folds have also been found. One was identified in protein S from the spore coat of a bacterium [34], another long-lived protein (like spherulin 3a and the crystallins). Interestingly, in this two-domain protein, the four homologous motifs are not arranged as in bc-crystallins (M1-M2-M3-M4), but in a reversed pattern (M2-M1-M4-M3) (Fig. 2). This prompted the suggestion [13] that the two evolutionary lines of the bacterial crystallin-like protein and the vertebrate crystallin ancestor diverged at the one-motif stage. Another two-domain, evolutionarily related member of the bc-crystallin superfamily has been identified in the epidermis differentiation-specific protein (EDSP) from an amphibian, Cynops pyrrhogaster [35]. The N-terminal portion of this protein contains four crystallin-type motifs that appear to be arranged in the M1-M2-M3-M-4 pattern typical of the b-andc-type lens crystallins. More recently, a two-domain crystallin signature has been identified in a protein sequence from a sponge of the genus Geodia [36]. In this protein too the four Greek key motifs are arranged in the same order (M1-M2-M3-M4) as in the vertebrate bc-crystallin genes. Another impressive addition to the bc-crystallin super- family is that proposed for AIM1, a protein encoded in a human gene whose expression has been related to melan- oma suppression [37]. The 3¢ terminal region of this gene codes for a protein sequence comprising 12 crystallin-type motifs arranged in the M1-M2-M3-M4 order. Trimeric protein models have been constructed connecting the six two-motif domains with either the bent c-type or the extended b-type interdomain linkers. The gene, however, appears to code for protein domains more closely related to b-type than to c-type crystallins. This conclusion is based on the following elements: (a) the linker peptide sequences are closer to those typical of b-type crystallins; (b) the gene contains intermotif introns as the b-type genes; (c) the interdomain intron positions are homologous to those of the b-type crystallins introns. As for the evolution of dimeric b-type crystallins, the possibility that a c-type gene encoding a monomeric crystallin was the immediate ancestor to a b-type gene encoding a dimeric crystallin has been excluded [14], based on the absence of intermotif introns in c-crystallin genes and their presence in b-type genes (Fig. 2). The lack of these introns in c-crystallins has been attributed to an intron loss occurred in a two-motif/one-domain crystallin ancestor. The loss would have occurred in the c-type genes only after the divergence of the evolutionary paths leading to c-type and b-type genes, respectively. This because it was deemed unlikely that an identical mutational event, the intron loss, could have occurred twice in the evolution of two homol- ogous one-domain genes after their divergence and before their fusion into four-motif/two-domain encoding genes. In fact, the opposite argument may be valid. The probability that a certain type of gene alteration occurs (an insertion, a deletion) depends on extrinsic (e.g. nature of the mutagen, environmental conditions) and on intrinsic factors: the base sequence, the consequent secondary and supersecondary structures, as well as the topology of the DNA region in which the event takes place. For homolog- ous genes we may assume that they share most of the intrinsic and extrinsic elements. Thus to evaluate the probability that a certain gene alteration occurred in evolution in homologous genes, we should use very similar probability factors. In conclusion, the likelihood that a gene alteration, such as an intron loss or insertion, could occur twice in the evolution of homologous genes in a certain gene family is greater the closer these genes are in evolution, i.e. if their divergence was a recent occurrence. Furthermore, the proposal that an intermotif intron loss occurred at the two-motif/one-domain stage would have as a consequence that a single two-motif/one-domain-enco- ding gene, in which the loss would have occurred, was the common ancestor to all subsequently diverged one-domain genes. But this does not appear to be the case, as indicated by the different motif orders in different crystalline-type genes. As mentioned above, in cb-type crystallins homol- ogous M1 motifs are N-terminal to motif M2, whereas in spherulin 3a and in protein S they are C-terminal to M-2 (see Fig. 2). Hence, in the evolutionary path of crystallin- type proteins it would seem unlikely that a single two-motif/ one-domain ancestor duplicated and diverged while also undergoing a switch of motif-encoding sequences to gener- ate different motif arrangements in the various descendant genes. A more general argument in favour of a late insertion of introns in crystallin-type genes, as opposed to a late deletion of pre-existing introns, may be based on the intron-late theory, originally proposed to explain the presence of spliceosomal introns in eukaryotes, and their absence in archea and in eubacteria [38–40]. Over recent years, a vast amount of data has been interpreted as supporting this theory [41]. In particular, the results of a statistical analysis [42] of pairs of gene paralogs may only be interpreted to favour intron gains rather than intron losses in these genes. Recent data [43] in support of the theory is the finding that in the sponge Geodia the gene encoding the extracellular and transmembrane domains of the tyrosine kinase receptor (TKR) has no introns. In homologous vertebrate TKR genes instead several introns are present. As for the late insertion of introns in crystallin-type proteins, it has been recently found (A. Di Maro, M. V. Cubellis & G. D’Alessio, unpublished results) that there are no introns in the gene encoding the crystallin-type protein from Geodia (see above). It should be noted that Geodia sponges are very primitive organisms that diverged more than 500 million years ago (some 300 million years earlier than mammals), whose crystallin genes have a full comple- ment of introns. This finding is in support of late gains of introns, rather than introns loss in the evolution of crystallin-type genes. An alternative model may therefore be proposed for the evolution of crystallin-type genes, clearly evolved from the previous models reported above [13,14,23]. In this model an early one-motif crystallin ancestor gene duplicated and diverged into several one-motif genes, whose combinatorial fusion engendered several two-motif pairs. This would accommodate all the motif arrangements (M1-M2, M1-M3, etc.) identified in present time crystallin-type proteins (Fig. 3). In this scenario there would be no loss of intermotif Ó FEBS 2002 On the evolution of crystallins (Eur. J. Biochem. 269) 3125 introns in the evolutionary pathway to c-type genes, but rather the acquisition of an interdomain intron in the evolution of c-crystallins, and of both interdomain and intermotif introns in the evolution of b-type crystallin genes (Fig. 3). The likelihood of late-in-evolution intron insertion, or that identical mutational events could have occurred in evolutionarily proximal genes has been discussed above. It should be noted that the DNA sequences coding for C-terminal and N-terminal extensions in one-domain ancestral crystallin genes would be the most likely candi- dates for the formation of intron sequences between motifs or between domains. This can be based not only on their intermotif and interdomain topologies, but also on their very high substitution rates (see above). Thus the available evidence from molecular genetics studies on crystallin-type genes may be interpreted as illustrated in Fig. 3, in which three main phases are summarized. In phase 1, duplications and divergence of the putative earliest, one-motif M ancestor occurred. In phase 2, the diverged duplicates fused in different combi- nations and underwent further divergence. Upon fusion, an intron formed between motifs in the evolutionary path of the ancestors toward the b-type, but not in that toward the c-type genes. In phase 3, the two-motif/one-domain enco- ding genes duplicated and fused, with the formation of an interdomain intron, possibly after the divergence of verte- brates. It should be noted that the scheme illustrated in Fig. 3 provides parallel, independent evolutionary paths for c-type monomers and for b-type oligomers. Thus, as previously proposed [14], oligomeric b-type crystallins did not evolve from monomeric c-type crystallins, although here this conclusion is based on different considerations. Naturally, and in line with previous analyses [2,3], such conclusion excludes the possibility that a dimeric b-type crystallin evolved from a monomeric c-type crystallin through a 3D domainswap[9]. The molecular genetics studies described above also suggest an important evolutionary role of the DNA regions encoding the interdomain linker peptides and the terminal extensions, as they are regions: (a) with high substitution rates; (b) where intron insertions or deletions occurred. STRUCTURAL STUDIES: FACTS AND HYPOTHESES When the question of crystallin evolution is examined from a structural viewpoint, the most impressive data is the high conservation of hydrophobic patches at inter–domain interfaces [20,23]. In c-crystallin, the hydrophobic residues Met43, Phe56 and Ile81 from motif M2 interact with the homologous Val132, Leu145 and Val170 from motif M4. Identical or analogous interactions occur at the b-crystallin interface between the triad of Val55, Val68, and Ile92, and that of Val143, Leu156, and Ile181. Then the C-terminal extensions have also been suspected to have a role in the evolution of domain association, as suggested by the interdomain hydrophobic interactions observed between the C-terminal extensions of the b-C-domain and the surface of the N-domain from the partner subunit [20], and by the peculiar behaviour [44] of the isolated c-C-domain altered at its C-terminal extension (see below). Finally, the strikingly different conformations of the interdomain linkers, bent or extended in c-andb-crystallins, respectively, could certainly not escape attention. Thus, the key structural features to focus upon while analyzing the evolution of two-domain or two-subunit bc- crystallins (i.e. the determinants of interdomain association, intramolecular or intermolecular), are the hydrophobic interdomain patches, the interdomain linker peptides, and the terminal extensions. These have been the precise targets selected by the London and Regensburg research groups in their investigations on the structural determinants and the evolution of the b-type and c-type crystallins [7,25,44–51]. The burying of the hydrophobic patches at the interdo- main interfaces, intramolecular in the c-type structure, intermolecular in the b-type, has been early recognized as the apparent driving force for domain association [20,23]. Fig. 3. A schematic summary of the main events in the evolutionary paths leading to pre- sent-day crystallin-type genes. Mdenotesa monomeric putative ancestor encoding a one- motif (Greek key) protein, hyphenated to indicate duplication and divergence of genes, with the numerals 1–4 indicating motif typologies. Subscripts F, S, C, G, and B denote the respective evolutionarily com- mitted ancestors of: spherulin 3a, protein S, the G. cydonium protein, the c-type, and the b-type crystallin genes. The segments con- necting the boxed M motifs indicate the pres- ence of intermotif (thin bars) and interdomain (thick bars) introns; the lack of separation lines between motifs or domains indicate that inthosecasesthepresenceorabsenceof introns has not been determined. 3126 G. D’Alessio (Eur. J. Biochem. 269) Ó FEBS 2002 However, an impressive network of H-bonds and ion pairs between Glu and Arg residues is also evident in these structures at the interdomain interfaces [52]. It is therefore tempting to conclude that the polar or charged side-chains involved in these contacts are remnants of the ancestral, solvent exposed surfaces of single-domain crystallins, now buried at interdomain interfaces of present day crystallins. As they concur to the interface stabilization, we can suggest that a Ôhydrophilic effectÕ [3] apparently concurred in stabilizing the interfaces of crystallins that evolved into higher order structures. As for the hydrophobic patches, many experiments have been performed to investigate their importance in the determinism of domain association, some of them with contradicting results. It has been reported that isolated c-crystallin domains, perfectly equipped with their hydro- phobic triad, either obtained through proteolytic cleavage [53], or as recombinant proteins [54], do not associate spontaneously into c-like domain dimers, and behave as stable monomeric proteins. These results would lead to conclude that the hydrophobic effect is not the only determinant of domain association. Yet, they may simply suggest that covalent interdomain linkers are essential to raise the local concentration of interdomain surfaces and engender the hydrophobic effect [46]. On the other hand, the substitution of a single residue (Phe56, replaced by Ala, Asp or Trp) in the triad responsible for the hydrophobic patch proved sufficient to destabilize c-crystallin domains to the point of rendering them incapable of engaging into a stable association [48]. Different results have been obtained with the isolated N-domain of rat bB2-crystallin, found to associate in solution [51], and with the isolated N- and C-domains of c-S-crystallin, for which a tendency to associate into heterodimers has been reported [55]. It should be noted that c-S-crystallin is very similar to b-crystallin, and that for a long time it was labelled as a b-crystallin. Recently, the structure of dimeric N-domains from rat bB2-crystallin has been solved [52] and shown to be maintained essentially by the canonic hydrophobic contacts described above, and by the polar interactions mentioned above. The apparent discrepancy between the two sets of data may be reconciled by the conclusion that c-type domains, once dissociated cannot re-associate, whereas domains of b-type and b-like c-S-type crystallins do not need a high local concentration of structural elements to build up the interface. Hence, in c-type crystallins the interdomain hydrophobic patches may not be the only determinant for domain association, whereas they are determinant and sufficient in b-crystallins. This conclusion may not be surprising if we consider the radically different conforma- tion of the interdomain linker peptides, bent and extended, respectively, in c-type and b-type crystallins. In the former case, the bent linker seems to be essential to drive the association at the interface, whereas in the latter the extended, spatially distant linker is not involved in the association. The role of the linker peptides in the determination of monomeric vs. dimeric structures, has also been investigated by protein engineering, with apparently contradicting results. One early conclusion had been that the linker peptides have no role in determining domain association. This was based on the following findings: a c-type protein remains monomeric when its c-type linker is replaced by a b-type linker [46]. Likewise, a b-type protein remains a dimer when its original linker is replaced with a c-type linker [56], as described previously [49]. In these experiments, the exchanged sequences comprised residues 82–87, as under- lined in the alignment of c- and b-type crystallins (Table 1.) However, when the latter experiment was carried out [47] by replacing the linker of the b-type protein with a longer c- type peptide sequence that included two extra residues at the N-terminus (Pro80 and Ile81 in the alignment above), the engineered b-type protein did become monomeric. Thus, if the linker peptide connecting motifs M2 and M3 of the protein is defined as the sequence comprising residues 80–87 [20], the linker sequence does appear to have a role as a determinant of the dimeric structure. It must be noted that the Pro residue at position 80 is strictly conserved in b-type crystallins, whereas in c-type proteins a Leu is found at that position (with the single exception of a Ser in cA-crystallin). This suggests that the presence of a Pro at position 80 can force the linker into an extended conformation, that typical of b-type crystallins, which does not allow for a sufficiently high local concentration of interdomain interacting residues [23]. In the absence of Pro80, these residues can interact and the two domains associate into a c-type monomer. It is tempting to propose that a key amino-acid substitution (a Ôprimary mutationÕ) in the evolution of c-type and b-type crystallins from their common ancestor was the insertion of a Pro residues at that position in the b-type sequences, and of a hydrophobic residue in c-type crystallins. Contrasting results were obtained in another laboratory, showing that a recombinant b-crystallin variant is isolated as a dimer also when its linker is replaced with a c-type linker [57]. Although the b-crystallin used in the latter experiment was rat b-B3, instead of bovine b-B2, and the replaced fragment was two residues longer, the replacing linker was from the same c-B crystallin as in the experiment cited above [46]. The insertion of a C-terminal Tyr residue in the substituting fragment, and the presence of a Ser instead of a Thr, may explain the contrasting results. If these were both confirmed, we may only surmise that in these types of engineering experiments only limited areas of the protein structure under test are narrowly illuminated, while other effects of the engineering on other areas of the protein structure remain in the dark, and may affect the interpret- ation. However, the overall conclusion that the linker peptides did have a role in the evolution of monomeric vs. oligomeric crystallins is convincing. In this respect, it would be interesting to determine the structure of the crystallin-type protein from the sponge gene [36], in which a short (only three residues) interdomain linker peptide has been identi- fied, i.e. with a length typical of c-type crystallins linkers. As for the terminal extensions, they are mostly flexible and mobile [58] and do not seem to play any roles in folding and domain association [44,59,60]. The proximal stretch of Table 1. Alignment of c- and b-type crystallins. The exchanged sequences comprise residues 82–87 (underlined). 80 87 bB2 crystallin linker PIKVDSQE cB crystallin linker LIPQHTGT Ó FEBS 2002 On the evolution of crystallins (Eur. J. Biochem. 269) 3127 the C-terminal extension in the b-type structure instead is not flexible, and has been suggested to mimic a noncovalent interdomain linker because it introduces its Trp175 residue in a hydrophobic pocket on the surface of the N-domain from the partner subunit [20]. When the whole C-terminal extension, including Trp175, is removed, b-type crystallin can still associate into dimers and tetramers [47]. But the terminal extensions, although apparently not a determinant in the structural chemistry of present-day crystallins, may have instead had key roles in the evolu- tionary modular assembly of these proteins. It has been found that although isolated, recombinant c-type C-domains cannot associate into noncovalent structures to mimic a c-type crystallin [7], yet they will associate after the removal of the terminal Tyr residue from their C-terminal extensions. In the 3D structure of this des-Tyr-c-C-domain, the C-terminal extension hinders the association of the two domains by interacting with the hydrophobic interdomain interface. This destabilizing effect would not be exerted when the covalent interdomain linker is in position and displaces the peptide extension out into the solvent. These results suggest that the extended form of the linker peptide, characteristic of b-crystallins, could have evolved directly from the C-terminal extension of a two-domain ancestor [7]. An independent experimental approach has led to similar conclusions. The C-terminus of the C-domain extension of rat bB2 crystallin has been fused by protein engineering with the N-terminus of the N-domain from the partner subunit [50]. Because the engineering also discontinued the interdo- main linkers in both subunits, a circularly permuted structure was obtained. In this structure, the C-terminal extension was turned into an interdomain linker. The resulting expressed protein was still a dimer, but differed from the wild-type bB2 dimer, in that its domain pairing was that typical of c-crystallin. These experiments support the proposal that the exten- sions may have been the evolutionary precursors of interdomain linkers, but also confirm the crucial role played in evolution by the linkers themselves. They hint at the possibility that circular permutation may have been one of the mechanisms employed in the evolution of new crystallin structures [15,49]. In modular constructions, structural variation depends on the different ways modules are assembled, i.e. on the different types of structural elements connecting and pairing the modules. Domain extensions could well have been exploited by evolution to generate a variety of linkers in order to get the creative advantages inherent in modular assemblies. It has been proposed that another experimental approach to obtain insight to the evolutionary history of an oligomeric protein is to investigate its unfolding/refolding [3,8]. This is based on the idea that the folding pathway of an oligomer might reiterate its evolutionary pathway. Thus, it may be of interest to analyze the results of unfolding/ refolding experiments carried out on crystallin-type pro- teins. Spherulin 3a [61], the single-domain crystallin-type pro- tein, unfolds in a highly cooperative fashion with a two-state transition [2,62,63]. Two-domain proteins, such as protein S [64] and a c-type crystallin [45], unfold instead with three- state transitions, just as a b-type crystallin does [51]. It should be added that the isolated N- or C-domains, prepared by recombinant technology unfold cooperatively with two-state transitions [54]. The intermediates in the unfolding pathway of both protein S and c-type crystallin have been described as presenting a still folded N-domain and a fully unfolded C-domain. In contrast, in the unfolding pathway of b-type crystallin the N-domain unfolds first while the C-domain remains folded. Interestingly, the isolated b-type C-domains are monomeric, whereas isolated N-domains associate. Based on these results, and on the findings described above, we can envisage that single-domain crystallin-type proteins natural as Spherulin 3a, or artificially produced as the isolated domains from c-andb-type crystallins resemble the evolutionary ancestors of two-domain crystallin. Hence, we may regard these one-domain proteins as stable mono- mers. Once rendered unstable through mutations in their encoding genes, they could find a new stable conformation only upon gene fusion leading to domain association. This evidently happened along distinct, parallel evolutionary paths, for c-type and protein-S crystallin-type proteins, and b)type crystallins, respectively. Thus, the results of the unfolding/refolding experiments and their interpretation are in support of the evolutionary pathway illustrated in Fig. 3. CONCLUSIONS It appears that the findings described above, based on structural and protein engineering studies or on molecular genetics analyses, lead to the same conclusions. Both sets of data indicate that a series of gene alterations and fusions led from crystallin ancestors coding for proteins made up of a single Greek-key motif to two-motif/one-domain proteins, to two-domain c-type crystallin monomers, or two-domain/ two-monomer b-type dimers. A key role in the evolutionary cascade was apparently played by the gene sequences encoding the C-terminal extensions downstream to the motif encoding exons in one-motif and one-domain ances- tors. These are the sequences involved in the gene fusion molecular events and especially marked by high substitution rates. In the present day, postfusion two-domain crystallin genes, homologous sequences encode the interdomain linker peptides. These DNA sequences were the hot spots in the ancestral crystallin genes, where evolution intensely experi- mented to generate protein sequences that independently evolved into two distinct paths, leading to different linker conformations for c-andb-crystallins, hence to monomers and dimers, respectively. ACKNOWLEDGEMENTS I am grateful for comments and criticism on the manuscript to J. F. Riordan (Harvard Medical School), G. Wistow (NIH), M. Riley (MBL, Woods Hole), and my colleagues in Naples: M. V. Cubellis, A. Di Maro, T. Giancola, R. Piccoli, and A. Russo. The rendering of molecular graphics for Fig. 1. was provided by M. V. Cubellis. Figures 2 and 3 were drawn by A. Di Maro; I am very grateful to both. REFERENCES 1. Park, C. & Raines, R.T. (2000) Dimer formation by a ÔmonomericÕ protein. Protein Sci. 9, 2026–2033. 2. Jaenicke, R. & Lilie, H. (2000) Folding and association of oligo- meric and multimeric proteins. Adv. Prot. Chem. 53, 329–401. 3128 G. D’Alessio (Eur. J. Biochem. 269) Ó FEBS 2002 3. D’Alessio, G. (1999) The evolutionary transition from monomeric to oligomeric proteins: tools, the environment, hypotheses. Prog. Biophys. Mol. Biol. 72, 271–298. 4. Xu, D., Tsai, C J. & Nussinov, R. (1998) Mechanism and evo- lution of protein dimerization. Protein Sci. 7, 533–544. 5. Ciglic,M.I.,Jackson,P.J.,Raillard,S.A.,Haugg,M.,Jermann, T.M., Opitz, J.G., Trabesinger-Ruf, N. & Benner, S.A. (1998) Origin of dimeric structure in the ribonuclease superfamily. Bio- chemistry 37, 4008–4022. 6. Beintema, J.J., Breukelman, H.J., Carsana, A. & Furia, A. (1997) Evolution of vertebrate ribonucleases: ribonuclease A superfamily. In Ribonucleases: Structures and Functions (D’Alessio, G. & Riordan, J.F., eds), pp. 245–269. Academic Press, San Diego. 7. Norledge,B.V.,Mayr,E.M.,Glockshuber,R.,Bateman,O.A., Slingsby, C., Jaenicke, R. & Driessen, H.P. (1996) The X-ray structures of two mutant crystallin domains shed light on the evolution of multi-domain proteins. Nat. Struct. Biol. 3, 267–274. 8. D’Alessio, G. (1995) Oligomer evolution in action? Nat. Struct. Biol. 2, 11–13. 9. Bennett, M.J., Schlunegger, M.P. & Eisenberg, D. (1995) 3D domain swapping: a mechanism for oligomer assembly. Protein Sci. 4, 2455–2468. 10. Crestfield, A.M., Stein, W.H. & Moore, S. (1962) On the aggregation of bovine pancreatic ribonuclease. Arch. Biochem. Biophys. 1S, 217–222. 11. Green, S.M., Gittis, A.G., Meeker, A.K. & Lattman, E.E. (1995) One step evolution of a dimer from a monomeric protein. Nat. Struct. Biol. 2, 746–751. 12. Russo, A., Antignani, A. & D’Alessio, G. (2000) In vitro evolution of a dimeric variant of human pancreatic ribonuclease. Biochem- istry 39, 3585–3591. 13. Wistow, G.J. & Piatigorsky. J. (1988) Lens crystallins: the evolu- tion and expression of proteins for a highly specialized tissue. Annu.Rev.Biochem.57, 479–504. 14. Lubsen,N.H.,Aarts,H.J.M.&Schoenmakers,J.G.G.(1989)The evolution of lenticular proteins: the b-andc-crystallin super gene family. Prog. Biophys. Mol. Biol. 51, 47–76. 15. Slingsby, C. & Clout, N.J. (1999) Structure of the crystallins. Eye 13, 395–402. 16. Augusteyn, R.C. & Stevens, A. (1998) Macromolecular structure of the eye lens. Prog. Polym. Sci. 23, 375–413. 17. Wistow, G. (1993) Lens crystallins: gene recruitment and evolu- tionary dynamism. Trends Biochem. Sci. 18, 301–306. 18. Piatigorsky, J. & Wistow, G. (1991) The recruitment of crystallins: new functions precede gene duplication. Science 252, 1078–1079. 19. Piatigorsky, J. (2000) Review: a case for corneal crystallins. J. Ocul. Pharmacol. Therap. 16, 173–180. 20. Bax, B., Lapatto, R., Nalini, V., Driessen, H., Lindley, P.F., Mahadevan, D., Blundell, T.L. & Slingsby, C. (1990) X-Ray analysis of bB2-crystallin and evolution of oligomeric lens pro- teins. Nature 347, 776–780. 21. Lapatto, R., Nalini, V., Bax, B., Driessen, H., Lindley, P.F., Blundell, T.L. & C.S. (1991) High resolution structure of an oli- gomeric eye lens b-crystallin. J. Mol. Biol. 222, 1067–1083. 22. Nalini, V., Bax, B., Driessen, H., Moss, D.S., Lindley, P.F. & Slingsby, C. (1994) Close packing of an oligomeric eye lens b-crystallin induces loss of symmetry and ordering of sequence extensions. J. Mol. Biol. 236, 1250–1258. 23. Blundell, T., Lindley, P., Miller, L., Moss, D., Slingsby, C., Tickle, J., Turnell, B. & Wistow, G. (1981) The molecular structure and stability of the eye lens: X-ray analysis of c-crystallin II. Nature 289, 771–777. 24. Wistow, G., Turnell, B., Summers, L., Slingsby, C., Moss, D., Miller, L., Lindley, P. & Blundell, T. (1983) X-ray analysis of the eye lens protein c-II crystallin at 1.9 A ˚ resolution. J. Mol. Biol. 170, 175–202. 25. Najmudin, S.N.V., Driessen, H.P.C., Slingsby, C., Blundell, T.L., Moss, D.S. & Lindley, P.F. (1993) Structure of the bovine eye lens protein cB(cII) crystallin at 1.47 A ˚ . Acta Crystallogr. D49, 223–233. 26. White, H.E., Driessen, H.P.C., Slingsby, C., Moss, D.S. & Lind- ley, P.F. (1989) Packing interactions in the eye lens. structural analysis, internal symmetry and lattice interactions of bovine gIVa-crystallin. J. Mol. Biol. 207, 217–235. 27. Chirgadze, Y.N., Sergheev, Y.V., Fomenkova, N.P. & Oreshin, V.D. (1981) Polypeptide chain pathway in g-crystallin IIIb from calf lens at 3 A ˚ resolution. FEBS Lett. 131, 81–84. 28. Wistow, G. (1990) Evolution of a protein superfamily: relationship between vertebrate lens crystallins and microorganism dormancy proteins. J. Mol. Biol. 30, 140–145. 29. Rosinke,B.,Renner,C.,Mayr,E.,Jaenicke,R.&Holak,T.A. (1997) Ca 2+ -loaded spherulin 3a from Physarum polycephalum adopts the prototype c-crystallin fold in aqueous solution. J. Mol. Biol. 271, 645–655. 30. Ohno, A., Tate, S., Seeram, S.S., Hiraga, K., Swindells, M.B., Oda, K. & Kainosho, M. (1998) NMR structure of the Streptomyces metalloproteinase inhibitor, SMPI, isolated from Streptomyces nigrescens TK-23: another example of an ancestral bc-crystallin precursor structure. J. Mol. Biol. 282, 421–433. 31.Antuch,W.,Gu ¨ ntert, P. & Wu ¨ thrich, K. (1996) Ancestral bc-crystallin precursor in a yeast killer toxin. Nat. Struct. Biol. 3, 662–665. 32. Ohki, S., Kariya, E., Hiraga, K., Wakamiya, A., Isobe, T., Oda, K. & Kainosho, M. (2001) NMR structure of streptomyces killer toxin-like protein, SKLP: further evidence for the wide distribu- tion of single-domain betagamma-crystallin superfamily proteins. J. Mol. Biol. 305, 109–120. 33. Clout, N.J., Slingsby, C. & Wistow, G.J. (1997) An eye on crys- tallins. Nat. Struct. Biol. 4, 685. 34. Wistow, G., Summers, L. & Blundell, T. (1985) Myxococcus xanthusspore coat protein S may have a similar structure to vertebrate lens bc-crystallins. Nature 315, 771–773. 35. Wistow, G., Javorski, C. & Vasantha Rao, P. (1995) A non-lens member of the bc-crystallin superfamily in a vertebrate, the amphibian Cynops. Exp. Eye Res. 61, 637–639. 36. Krasko, A., Mu ¨ ller, I.M. & Mu ¨ ller, W.E.G. (1997) Evolutionary relationships of the metazoan bg-crystallins, including that from themarinespongeGeodia cydonium. Proc. R. Soc. Lond. B. 264, 1077–1084. 37. Ray,M.E.,Wistow,G.,Su,Y.A.,Meltzer,P.S.&Trent,J.M. (1997) AIM1, a novel non-lens member of the bc-crystallin superfamily, is associated with the control of tumorigenicity in human malignant melanoma. Proc. Natl Acad. Sci. USA 94, 3229–3234. 38. Rogers, J.H. (1989) How were introns inserted into nuclear genes? Trends Genet. 5, 213–216. 39. Palmer, J.D. & Logsdon Jr, J.M. (1991) The recent origin of introns. Genet. Dev. 1, 470–477. 40. Cavalier-Smith, T. (1991) Intron phylogeny: a new hypothesis. Trends Genet. 7, 145–148. 41. Logsdon Jr, J.M. (1998) The recent origins of spliceosomal introns revisited. Curr. Opin. Genet. Dev. 8, 637–648. 42. Cho, G. & Doolittle, R.F. (1997) Intron distribution in ancient paralogs supports random insertion and not random loss. J. Mol. Evol. 44, 573–584. 43. Gamulin, V., Skorokhod, A., Kavsan, V., Muller, I.M. & Muller, W.E.G. (1997) Experimental indication in favor of the introns-late theory: the receptor tyrosine kinase gene from the sponge Geodia cydonium. J. Mol. Evol. 44, 242–252. 44. Norledge, B.V., Trinkl, S., Jaenicke, R. & Slingsby, C. (1997) The X-ray structure of a mutant eye lens bB2-crystallin with truncated sequence extensions. Protein Sci. 6, 1612–1620. Ó FEBS 2002 On the evolution of crystallins (Eur. J. Biochem. 269) 3129 45. Rudolph, R., Siebendritt, R., Nesslauer, G., Sharma, A. & Jae- nicke, R. (1990) Folding of an all- b protein: independent domain folding in c-II-crystallin from calf eye lens. Proc. Natl Acad. Sci. USA 87, 4625–4629. 46. Mayr, E.M., Jaenicke, R. & Glockshuber, R. (1994) Domain interactions and connecting peptides in lens crystallins. J. Mol. Biol. 235, 84–88. 47. Trinkl, S., Glockshuber, R. & Jaenicke, R. (1994) Dimerization of bB2-crystallin: the role of the linker peptide and the N- and C-terminal extensions. Protein Sci. 3, 1392–1400. 48. Palme, S., Slingsby, C. & Jaenicke, R. (1997) Mutational analysis of hydrophobic domain interactions in c-B-crystallin from bovine eye lens. Protein Sci. 6, 1529–1536. 49. Wright, G., Basak, A.K., Wieligmann, K., Mayr, E.M. & Sling- sby, C. (1998) Circular permutation of betaB2-crystallin changes the hierarchy of domain assembly. Protein Sci. 7, 1280–1285. 50. Wieligmann, K., Norledge, B., Jaenicke, R. & Mayr, E.M. (1998) Eye lens betaB2-crystallin: circular permutation does not influence the oligomerization state but enhances the conformational stabi- lity. J. Mol. Biol. 280, 721–729. 51. Wieligmann, K., Mayr, E.M. & Jaenicke, R. (1999) Folding and self-assembly of the domains of betaB2-crystallin from rat eye lens. J. Mol. Biol. 286, 989–994. 52. Clout, N.J., Basak, A., Wieligmann, K., Bateman, O.A., Jaenicke, R. & Slingsby, C. (2000) The N-terminal domain of bb2-crystallin resembles the putative ancestral homodimer. J. Mol. Biol. 304, 253–257. 53. Sharma, A.K., Minke-Gogl, V., Gohl, P., Siebendritt, R., Jaenicke, R. & Rudolph, R. (1990) Limited proteolysis of gamma II-crystallin from calf eye lens. Physicochemical studies on the N-terminal domain and the intact two-domain protein. Eur. J. Biochem. 194, 603–609. 54. Mayr, E M., Jaenicke, R. & Glockshuber, R. (1997) The domains in cb-crystallin: identical fold-different stabilities. J. Mol. Biol. 269, 260–269. 55. Wenk, M., Herbst, R., Hoeger, D., Kretschmar, M., Lubsen, N.H. & Jaenicke, R. (2000) Gamma-S-crystallin of bocine and human eye lens: solution structure, stability and folding of the intact two-domain protein and its separate domains. Biophys. Chem. 86, 95–108. 56. Trinkl, S. (1995) Einfluss von Strukturelementen auf Assoziation und Stabilitaet von bb2-Kristallin. Ph D Thesis, University of Regensburg, Regensburg. 57. Hope, J.N., Chen, H C. & Hejtmancik, J.F. (1994) aggregation of bA3-crystallin is independent of the specific sequence of the domain connecting peptide. J.Biol. Chem. 269, 21141–21145. 58. Carver, J.A., Cooper, P.G. & Truscott, R.J.W. (1993) 1 H-NMR- spectroscopy of bB2-crystallin from bovine eye lens. Conforma- tion of the N- and C-terminal extensions. Eur. J. Biochem. 213, 313–320. 59. Coop, A., Goode, D., Sumner, I. & Crabbe, M.J.C. (1998) Effects of controlled mutations on the N- and C-terminal extensions of chick lens bB1 crystallin. Graefe’s Arch. Clin. Exp. Ophtalmol. 236, 146–150. 60. Kroone, R.C., Elliot, G.S., Ferszt, A., Slingsby, C., Lubsen, N.H. & Schoenmakers, J.G.G. (1994) The role of sequence extentions in b-crystallin assembly. Protein Eng. 7, 1395–1399. 61. Kretschmar, M., Mayr, E. & Jaenicke, R. (1999) Homo-dimeric spherulin 3a: a single domain member of the bc-crystallin super- family. Biol. Chem. 380, 89–94. 62. Jaenicke, R. (1987) Folding and association of proteins. Prog. Biophys. Mol Biol. 49, 117–237. 63. Jaenicke, R. (1999) Folding and stability of domain proteins. Progr. Biophys. Mol. Biol. 71, 155–241. 64. Wenk,M.,Baumgartmer,R.,Holak,T.A.,Huber,R.,Jeanicke, R. & Mayr, E. (1999) The domains of protein S from Mixococcus xanthus: structure, stability and interactions. J. Mol. Biol. 286, 1533–1545. 65. Clout, N.J., Kretschmar, M., Jaenicke, R. & Slingsby, C. (2001) Crystal structure of the calcium-loaded spherulin 3a dimer sheds light on the evolution of the eye lens bc-crystallin domain fold. Structure 9, 115–124. 3130 G. D’Alessio (Eur. J. Biochem. 269) Ó FEBS 2002 . c-andb -crystallins, andsimplyreferredtoasc-type or b-type crystallins, respectively. Both monomeric c-type crystallin and the subunit of dimeric b-type crystallin. Italy The case of homologous monomeric c-type and oligomeric b-type crystallins has been described and analyzed in evo- lutionary terms. Data and hypotheses