Introduction

The ability to explore one’s environment by vision is one of the major innovations during evolution and provides a clear selective advantage. In the vertebrate camera-type eye, it is the lens that determines visual acuity and the properties of the vertebrate lens thus correlate closely with lifestyle: aquatic and nocturnal animals have hard, round lenses, while diurnal animals have flatter and softer lenses (Land and Nilsson 2002). All vertebrate (gnathostome) eye lenses studied thus far contain the so-called ubiquitous crystallins, members of the α-, β- and γ-crystallin protein families (Bloemendal et al. 2004; Bloemendal and de Jong 1991; Wistow and Piatigorsky 1988). The properties of the lens are, amongst other factors, determined by the composition and properties of these water soluble proteins present at very high concentrations (Delaye and Tardieu 1983). At first glance, the crystallin composition of the lens appears to be rather variable. The lenses of many species contain in addition to the ubiquitous crystallins other proteins, the so-called taxon-specific crystallins, at high concentrations, while lenses of other, closely related species, do not (Bloemendal et al. 2004; Piatigorsky 1989; Wistow 1993). To what extent the taxon-specific crystallins are functional in the sense of determining the exact optical properties of a lens is not clear. For example, in the gecko ι-crystallin, a taxon-specific crystallin, is likely to have been recruited as a UV filter (Werten et al. 2000). The taxon-specific crystallins are, however, merely superimposed on a common and highly conserved theme, namely the three ubiquitous crystallin protein families. The α-crystallins belong to the small heat-shock protein family (Horwitz 2003). Members of this family are found in virtually all organisms and a scenario for the evolutionary origin of the lenticular α-crystallins can easily be envisaged (de Jong et al. 1993). In contrast, the precursors of the present day vertebrate β- and γ-crystallins have been difficult to trace. The (multimeric) β- and the (monomeric) γ-crystallins are structurally and evolutionarily related: they are both built up out of four Greek key motifs organized into two domains. The basic distinction between these two gene families in gnathostomes is in the location of the introns. In the β-crystallin genes, each motif is encoded by a separate exon, while in the γ-crystallin, each domain, i.e. two motifs, is encoded by an exon (for reviews, see Bloemendal et al. 2004; Bloemendal and de Jong 1991; de Jong et al. 1993; Lubsen et al. 1988). One member of the γ-crystallin family, γN, has an intermediate gene structure, with the N-terminal domain being encoded by a single exon (γ-crystallin gene like), while each of the two motifs of the C-terminal domain is encoded by a separate exon (β-crystallin gene like) (Weadick and Chang 2009; Wistow et al. 2005). For all β- and γ-crystallins (an) additional exon(s) encodes a N-terminal extension including the start methionine.

Combining information on βγ-crystallin gene organization with their encoded three-dimensional protein structures provides clues about their evolutionary origins. In the vertebrate βγ-crystallins, the sequences of motifs 1 and 3 are more similar to each other than they are to motifs 2 and 4. The same holds true for motifs 2 and 4: these are more closely related to each other than they are to motifs 1 and 3. This is considered to reflect the origin of the two-domain protein from an ancient gene duplication, with present day βγ-crystallin domains pairing about an approximate dyad recapitulating an ancient two-domain dimer (Blundell et al. 1981). Supporting evidence for this scenario comes from three-dimensional studies: engineered single domains can form symmetric homodimers (Norledge et al. 1996; Basak et al. 1998; Purkiss et al. 2002; Clout et al. 2000). An essential attribute of a lens protein is the ability to pack inside a lens fibre cell without forming discontinuities on the scale of half a wavelength of light. Domain pairing in β- and γ-crystallins is the first higher level of protein organization for this superfamily. The characteristic domain pairing interaction is mediated by three key residues, donated by motifs 2 and 4. The symmetry of βγ-crystallin domain pairing allows two versions of this domain assembly interaction: intramolecular pairing to form monomeric γ-crystallins, and intermolecular domain pairing by domain swapping to form certain β-crystallin dimers (Bax et al. 1990; Smith et al. 2007).

The double Greek key fold is used as the sole basic building block in the β- and γ-crystallins but also in a number of bacterial, archaeal and fungal proteins, such as in the four motif Protein S (Myxococcus xanthus; Wistow et al. 1985), the two motif protein from Methanosarcina acetivorans (Barnwal et al. 2009) or the two motif Spherulin 3a (Physarum polycephalum; Kretschmar et al. 1999; see also Jaenicke and Slingsby 2001). None of these proteins are potential orthologs of the vertebrate β- and γ-crystallins. More closely related invertebrate proteins are a four motif βγ-crystallin-like protein in the sponge Geodia cydonium (Di Maro et al. 2002) and the two motif βγ-crystallin of the urochordate Ciona intestinalis (Shimeld et al. 2005). The gene for the Ciona protein is likely on the evolutionary route to the vertebrate βγ-crystallin genes: it is β-crystallin-like in that the two motifs are encoded by separate exons (in contrast, the Geodia βγ-crystallin gene lacks introns) and its promoter region drives expression to the vertebrate lens (Shimeld et al. 2005). The single domain Ciona protein lacks the set of hydrophobic residues involved in domain pairing on its motif 2, and it does not dimerise in solution or in the crystal lattice (Shimeld et al. 2005). The Ciona protein further differs from the vertebrate protein in that it has characteristic calcium-binding residues in each motif: by exploiting the approximate dyad symmetry of the domain fold, a pair of half-sites combines to form two full sites on the double motif domain. Three-dimensional studies have shown that microbial βγ-crystallin-like proteins (PDB id 2k1w, Barnwal et al. 2009; PDB id 1hdf, Clout et al. 2001; PDB id 1nps, Wenk et al. 1999) have very similar calcium-binding sites as the Ciona protein (PDB id: 2bv2). The double Greek key fold can also be fused to other protein domains as in the epidermal differentiation-specific protein (EDSP) like proteins, thus far only detected in amphibians (Liu et al. 2008; Wistow et al. 1995), or in the Absent In Melanoma 1 (AIM1; a protein associated with suppression of malignancy of melanomas), a non-lenticular protein found in gnathostomes (jawed vertebrates). AIM1 has six βγ-crystallin like domains (Ray et al. 1997) flanked at the N-terminal side by an as yet poorly defined filament-like region and at the C-terminal side by a ricin-type β-trefoil domain. In the AIM1 gene, introns are found between motif coding regions, as in the β-crystallin genes.

In gnathostomes, expression of the β- and γ-crystallin genes is mostly restricted to the lens, and tracing the evolution of these genes would thus also shed light on the evolution of the lens. It is therefore of interest to close the gap between the two motif/one domain Ciona gene and the four motifs/two domain vertebrate genes. Here, we show that the gene duplications leading to the present day vertebrate β- and γ-crystallin genes preceded the divergence between the cyclostomes (jawless vertebrates) and the gnathostomes and must thus have occurred very early in vertebrate evolution. We further show that the genome of a member of the cephalochordates, Branchiostoma floridae (amphioxus), does encode βγ-crystallin-like protein domains, which at the protein level are closely related to the Ciona βγ-crystallin, but which have a rather different gene structure.

Materials and Methods

Searching Databases for β- and γ-Crystallin Related Sequences

The preliminary genome assembly of the Petromyzon marinus (http://pre.Ensembl.org/Petromyzon_marinus/index.html) and the second genome assembly of Branchiostoma floridae (http://genome.jgi-psf.org/Brafl1/Brafl1.download.ftp.html) as well as the EST databases were searched for sequences encoding β- and γ-crystallin related proteins using as query the vertebrate β- and γ-crystallin protein sequences, the Ciona βγ-crystallin sequence as well as the sequences of EDSP and AIM1. Platypus (Ornithorhynchus anatinus), opossum (Monodelphis domestica) and armadillo (Dasypus novemcinctus) γ-crystallin sequences were as annotated in Ensemble and manually curated to remove some errors (for example, a Q encoded by a splice acceptor site).

Cloning of P. marinus β- and γ-Crystallin cDNAs

Total P. marinus lens RNA was reverse transcribed using the first-strand cDNA synthesis kit for RT-PCR (Roche Applied Science) according to the manufacturer’s instructions with either oligo(dN)6 primers or an oligo(dT)15GC primer (5′-CCGCCGCCTTTTTTTTTTTTTTT-3′). Putative β- and γ-crystallin transcripts were amplified using the primers listed in Table 1 and the Expand High Fidelity polymerase kit (Roche Applied Science). Products were ligated into the pGEM-T Easy vector (Promega) and individual inserts were sequenced using BigDye terminators and a 3730 DNA analyzer (Applied Biosystems). Pm-βA2, -βB, -γA and -γB are available in the Genbank database under accession numbers GQ355899, GQ355900, GQ355901 and GQ355902, respectively.

Table 1 List of primers used in the amplification of β- and γ-crystallin cDNAs from Petromyzon marinus eye lenses
Full size table

Phylogenetic Analysis

Protein sequences were aligned using Muscle (Edgar, 2004) at default settings at the Wageningen Bioinformatics webportal (http://www.bioinformatics.nl). In the alignment, the AIM1 sequences were split into three four motif/two domain segments. A phylogenetic tree was inferred from the alignment using the maximum likelihood method as implemented in PhyML v3.0 (Guindon and Gascuel 2003) using the WAG substitution model, four substitution rate categories, an estimated proportion of invariable sites and gamma shape parameter. Nodal support was estimated by bootstrap analysis with 500 replicates. The sequences and accession numbers are given in the supplementary material.

Results and Discussion

β-Crystallin Related Sequences in the Lamprey

Searching the P. marinus (sea lamprey) genome assembly for β-crystallin related genes identified a number of contigs containing one or more β-crystallin-like exons. By sequence comparison, the predicted amino acid sequences encoded by these exons were classified as βA- or βB-crystallins (Table 2). Since none of the contigs encompassed a complete β-crystallin gene, we tried to amplify complete β-crystallin coding sequences from eye lens cDNA using primers based on the DNA sequences of the various exons encoding first and last motifs. A PCR using oligos derived from the exons on contigs 12940 and 23617 (Table 1) resulted in the amplification of a transcript representing a βA-crystallin gene (see below). In a similar fashion, the exons on contigs 59572, 50016 and 80908 could be linked by PCR resulting in an amplified transcript representing a βB-crystallin gene. The exons encoding the N-terminal arms of the proteins could not be located with any degree of confidence in the genomic sequence and are thus not included in the amplified sequences.

Table 2 List of Petromyzon marinus genome contigs showing βγ-crystallin related sequences
Full size table

γ-Crystallin Related Sequences in the Lamprey

γ-Crystallin related sequences could also be readily detected in the lamprey genome. The contigs harbouring such sequences are listed in Table 2. On most contigs two exons were found each encoding a γ-crystallin domain and separated by an intron of approximately 1.5 kb in size. The exception is contig 17429 on which the two exons are separated by an intron of only 217 bp. As for the β-crystallin genes, the (first) exon, which encodes a 3–7 amino acid N-terminal arm in the gnathostomes could not be detected. Expression of the putative γ-crystallin genes located on contigs 1382, 1488 and 17429 in the lens was tested by PCR on eye lens cDNA. For the genes located on contigs 1488 and 17429 the corresponding transcript was found; we failed to detect a transcript from contig 1382, which could be due to differentiation and/or developmental specificity of expression. We did not test for expression of the γ-crystallin genes located on the other contigs as these are virtually identical in sequence to the γ-crystallin gene on contig 1488 and represent either very recent duplications or assembly errors.

The predicted amino acid sequences encoded by the γ-crystallin-type motifs 1 and 2 (M1.M2) exon on contig 106517 and by the two β-crystallin-type motif 3 (M3) and 4 (M4) exons linked on contig 17603 were most similar to the corresponding motifs of γN-crystallin. We could not amplify transcripts of either the single exons or combinations thereof from eye lens cDNA, presumably because γN-crystallin is not expressed at a high level (Wistow et al. 2005). Alignment of the predicted γN-crystallin sequence (Pm_γN) with putative orthologs from zebrafish and mouse showed that the Pm_γN would have an insert of one amino acid in motif 3 (Fig. 1).

Fig. 1
figure 1

The P. marinus γN protein. The predicted sequences of Pm γN-crystallin and of the putative γN-crystallin-type single domain protein (Pm_sdG) are aligned with the D. rerio and M. musculus γN-crystallin sequence. Structurally important residues are in bold, the conserved tyrosine and tryptophan corners in blue, and the three hydrophobic residues in motifs 2 and 4 involved in domain pairing in green. The predicted sequence of the C-terminal extension of Pm_sdG is given below the alignment. The alignment is split in the odd and even motifs

Full size image

The P. marinus genome contains a second region encoding a γN-type N-terminal domain, on contig 23770 (denoted Pm_sdG in Fig. 1). We could not detect a splice donor site either at the 3′ end of the second exon region (the sequence here reads gtctg) or further downstream. The putative splice acceptor site of this potential γN exon is directly preceded by an ATG initiation codon, suggesting that this region of the P. marinus genome could encode a single domain γN-like protein lacking an N-terminal arm but with a long C-terminal tail. This putative single domain γN-crystallin protein would have the hydrophobic domain pairing residues that allow it to form a homodimer.

Other βγ-Crystallin Related Sequences in the Lamprey Genome—the AIM1 Gene

The search for β- and γ-crystallin sequences in the P. marinus genome yielded significant hits on contig 332. However, the predicted protein sequences aligned only poorly with fish or mammalian β- and γ-crystallins. We therefore repeated the search using AIM1 and EDSP sequences. The AIM1 sequences also matched sequences on contig 332; no potential EDSP ortholog was found. Closer inspection of contig 332 identified a total of 13 exons together encoding AIM1-like βγ-crystallin motifs 1-11 and two linker regions (Fig. 2). The exons encoding the twelfth βγ-crystallin motif, the C-terminal ricin domain and the N-terminal half of the AIM1 protein could not be found. The intron positions and phases of the P. marinus AIM1 exons are identical to the ones in the human AIM1 gene. An alignment of the 11 Pm AIM1 βγ-crystallin motifs with the corresponding motifs from the D. rerio and, for comparison, motifs 1 and 2 from human AIM1 clearly shows Pm AIM1 is the ortholog of the AIM1 gene in gnathostomes (Fig. 2, see also Fig. 3). The large insert in the second Pm βγ-crystallin motif between positions 40 and 70 is present in all second AIM1 βγ-crystallin motifs. Determination of the structure of the first domain of human AIM1 protein shows that the tertiary structure stays intact despite the extra bulge within the second motif (Aravind et al. 2008). The seventh Pm βγ-crystallin motif does deviate substantially from the others since it lacks most of the conserved residues that specify the folded hairpin of the Greek key motif-fold (shown in bold black typeface in Fig. 2). The presence of an AIM1 gene in P. marinus shows that the origin of this gene must predate the cyclostomes–gnathostomes split. We could not find an AIM1 gene in B. floridae or in C. intestinalis (Shimeld et al. 2005; unpubl. res. and see below). This suggests emergence of the AIM1 gene after the urochordate and vertebrate divergence.

Fig. 2
figure 2

Alignment of B. floridae βγ-crystallin related sequences. The predicted protein sequences of B. floridae Bf-bg1, -bg2, -bg3 and -bg4 (acc.nr. FE565747, BW704196 and BW723025) are aligned with the C. intestinalis βγ-crystallin (Ci_bg) and the D. rerio βB2- (Dr_bB2) and γSa-crystallin (Dr_gSa). The motifs are indicated as M followed by the number. Structurally important residues are indicated as in Fig. 1; calcium-binding residues are in grey. The sequence in italics in Bf-bg3 M1 was derived from the genomic sequence; the C-terminal extensions of Bf-bg1 and Bf-bg2 are shown below the alignment. The # indicates the position of the HVNPANT insert in Bf-bg1 M2 and & that of PPSNM in Bf-bg1 M4. The alignment is split in the odd and even motifs

Full size image
Fig. 3
figure 3

Evolution of the vertebrate βγ-crystallins. A maximum likelihood tree based on the alignment of both vertebrate and invertebrate βγ-crystallin sequences (see supplementary material) was constructed. Species are indicated as Mm M. musculus, Dr D. rerio and Pm P. marinus. Invertebrate βγ-crystallins were included in building the tree; for clarities sake the invertebrate branches have been collapsed (arrow). The Pm_gN is predicted from genomic sequence and marked with an asterisk, other Pm βγ-crystallin sequences are as validated by RT-PCR. The Dr and predicted Pm AIM1 sequences were divided into three four motif, two-domain segments as indicated. Numbers adjacent to nodes indicate percentage bootstrap support; only values larger than 70% (of 500) are shown

Full size image

Phylogenetic Analysis of the P. marinus βγ-Crystallins and Related Proteins

On the basis of an alignment of the deduced sequences of the P. marinus β- and γ-crystallins with representative gnathostome β- and γ-crystallins sequences as well as AIM1 motifs and invertebrate βγ-crystallins (see below and supplementary material) the phylogenetic tree shown in Fig. 3 was inferred. The P. marinus AIM1 βγ-crystallin domains and the vertebrate (lens) βγ-crystallins form separate clades. This strongly suggests that these proteins originated independently from an ancestral βγ-crystallin sequence. Within the βγ-crystallin branch, the two Pm β-crystallin sequences cluster with βA2-crystallin and with βB1-crystallin, respectively. The tree shows that the gene duplications leading first of all to the acidic and basic β-crystallins and subsequently to the subdivision in βA2- and βA3/βA4-crystallin within the acidic β-crystallins and βB1- and βB2/βB3-crystallins within the basic β-crystallins must have preceded the divergence between cyclostomes and the gnathostomes. One expects then to find genes for a βA3/βA4-crystallin and a βB2/βB3-crystallin in the P. marinus genome as well. The genome sequence is about 80% complete (pre.ensembl.org/Petromyzon_marinus/Info/StatsTable) and those genes could thus still be missing from the sequence. Furthermore, the lamprey genome is extensively rearranged in somatic cells during development and genes could thus be present in the germ line genome but lacking from the DNA of somatic cells (Smith et al. 2009) and from the present genome assembly. The orphan βA-crystallin-type motif 4 on contig 41226 (Table 2) does suggest that genomic information is still missing. The alternative is that the genes have been lost again during the evolution of the cyclostomes.

The phylogenetic tree also shows that the putative Pm γN-crystallin groups with the mouse and fish γN-crystallin, whereas the Pm γA- and γB-crystallin genes form a separate clade. The P. marinus genome appears to lack the γS-crystallin gene common to gnathostomes. Either this gene originated after divergence of the cyclostomes and the gnathostomes or it is located in that part of the P. marinus genome that still needs to be sequenced and assembled. It has been previously shown that the γS-crystallin genes in the gnathostomes are orthologs. Similarly, the γN-crystallin genes all descended from the same ancestral gene (see also Fig. 4). In contrast, the other γ-crystallin genes have repeatedly expanded and contracted (Lubsen et al. 1988; Weadick and Chang 2009; Wistow et al. 2005). To determine when the mammalian crygA-F genes originated, we added platypus (O. anatinus), opossum (M. domestica) and armadillo (D. novemcinctus), Xenopus leavis, rat (Rattus norvegicus) and human (Homo sapiens) γ-crystallin sequences to the phylogenetic tree. As shown in Fig. 4, the mammalian γA-F radiation from a single γ-crystallin gene preceded the earliest split in the mammalian lineage, that between Prototheria and Theria. The selective pressure on the variable subclass of the γ-crystallins is thought to result from the adaptation of the lens shape and properties to the lifestyle. A high γ-crystallin level correlates with low lens water content (for review, see Bloemendal et al. 2004) and thus round, hard lenses suitable for aquatic or nocturnal life. The soft chicken lens, for example, contains only γN- and γS-crystallin and lacks the variable subgroup of γ-crystallins altogether (but does have large amounts of a taxon-specific enzyme crystallin). Fish eyes have steep, symmetric gradients of refractive index, approximately parabolic in shape, to increase focusing power whilst correcting for spherical aberration (Land and Nilsson 2002). They are often multifocal, which corrects for chromatic aberration, allowing high-resolution colour vision (Kröger et al. 1999). The positive and negative corrections for spherical aberration, which facilitate spectral tuning to the retinal photopigments, are considered to stem from small perturbations to the shape of the lens refractive index gradient. Multifocality may be of ancient evolutionary origin in vertebrates (Karpestam et al. 2007), and recently it has been shown that in the adult stage, the P. marinus has a multifocal lens (Gustafsson et al. 2008). High levels of sulphur containing residues may contribute towards a high power lens by increasing the protein refractive index increment and/or the ability to close pack. The fish γM proteins contain high levels of sulphur containing residues, particularly methionine (Chang et al. 1988): for example for D. rerio, out of the (177) amino acid residues of γM1 (excluding the start methionine), 15.2% are cysteine or methionine, of γM2a 18.9%, of γM3 13.3%, of γM4 11.0%, of γM5 13.0%, of γM6 10.2% and of γM7 13.8%; however, in human lens the sulphur level for γC is 7.5% and for γD is 5.8%. By comparison, in the lamprey 172-residue γA sequence, the sulphur level is 9.9%, and for γB it is 8.8%, while of the two elasmobranch lipshark γM sequences, M1 has 21.2% sulphur and M2 has 9.7%. The lens focal length (normalized for lens radius) of the deep water P. marinus (2.31R) is in the lower range for teleost lenses, which is 2.2–2.8R (Gustafsson et al. 2008). It would be useful to collect clade specific γ-crystallin sequences, measure their refractive index increments, and compare these with the measured lens optical properties.

Fig. 4
figure 4

Evolution of the γ-crystallins. A maximum likelihood tree based on an alignment of γ-crystallins (supplementary material) was constructed. Species are indicated as in the legend to Fig. 3. In addition, sequences were used from platypus (Ornithorhynchus anatinus Oa), opossum (Monodelphis domestica Md), armadillo (Dasypus novemcinctus Dn), rat (Rattus norvegicus Rn), Xenopus laevis (Xl) and man (Homo sapiens Hs). Sequences and accession numbers can be found in the supplementary material. For clarities sake, some branches have been collapsed (arrows). gA, gB and gC denote the corresponding genes from Hs, Mm, Rn, and Dn, gS the genes from Hs, Mm, Rn, Md and Oa, gN the genes from Hs, Mm, Rn, Md, Dn and Oa. Numbers adjacent to nodes indicate percentage bootstrap support; only numbers larger than 70 are shown

Full size image

βγ-Crystallin Related Sequences in Invertebrates

We have previously identified a single domain βγ-crystallin in the urochordate C. intenstinalis (Shimeld et al. 2005). To determine whether cephalochordates also contain a similar βγ-crystallin gene, we searched the genome assembly of the cephalochordate Branchiostoma floridae (amphioxus) for sequences encoding βγ-crystallin related motifs. This search did not yield potential orthologs of AIM1 and EDSP but did yield four βγ-crystallin related genes. Three of these, Bf-bg1, Bf-bg2 and Bf-bg3 (Fig. 5), are each supported by a single EST. Bf-bg2 and Bf-bg3 are closely linked head to tail. Gene models predict an alternative transcript of this genome region, which would contain the first three exons of Bf-bg2 and the first exon present in the Bf-bg3 EST (see Fig. 6; note that the Bf-bg3 EST is likely to be incomplete at the 5′ end, see below). This fourth possible βγ-crystallin coding sequence is here denoted as Bf-bg4. The EST coverage of this region is sparse and insufficient to exclude the possibility of Bf-bg4. Furthermore, there may still be problems with the genome assembly as evidenced by the fact that the Bf-bg3 region is inverted in the second assembly of the genome relative to the first assembly. The Bf-bg1, Bf-bg2 and Bf-bg4 protein sequences have four βγ-crystallin motifs encoded by four exons. However, unlike in the Ciona, or vertebrate β- and γ-crystallin genes and the βγ-crystallin-like region in the AIM1 gene, introns 1 and 2 are located within and not between the motif coding regions (see below). If the AUG codon in exon one is indeed the initiation codon, then these proteins would also lack the short N-terminal arm encoded by a separate 5′ exon in all the vertebrate β- and γ-crystallins genes as well as in the Ciona βγ-crystallin gene. The Bf-bg3 sequence, as far as it can be deduced from the corresponding EST, is very similar to that of Bf-bg2, but would be missing the first motif and is probably incomplete.

Fig. 5
figure 5

Alignment of P. marinus AIM1 βγ-crystallin motifs. The AIM1 βγ-crystallin motifs found in the P. marinus (Pm) genome assembly are aligned with those of the D. rerio (Dr) AIM1 protein and with the first two βγ-crystallin motifs of human (Hs) AIM1. The alignment is split in the odd and even motifs. Structurally important residues are indicated as in Fig. 1

Full size image
Fig. 6
figure 6

Schematic representation of the genomic arrangement of the exons of Bf-bg2, Bf-bg3 and Bf-bg4. The ESTs mapped to nucleotide region 394006–397596 of contig 106 (version 2 of the genomic assembly) are shown with the exons indicated by the thick lines and the introns by the thin lines. The arrowheads show the direction of transcription, the numbers indicate the exon numbering as referred to in the text and the ? indicates that this exon is likely incomplete at the 5′ end

Full size image

An alignment of the Bf-bg protein sequences shows the same pattern of motif similarity as vertebrate βγ-crystallins: motifs 2 and 4 are more similar to each other than to motifs 1 and 3 (Fig. 5). However, the divergence between Bf-bg2 or Bf-bg3 motifs is less than in the vertebrate β- and γ-crystallins, suggesting a recent duplication from a one domain to a two domain encoding gene. The critical Greek key motif residues are also conserved in the Bf-bg proteins (Fig. 5, in bold) with few exceptions, such as the highly conserved G replaced by an N in motif 3 of Bf-bg3, and the highly conserved S replaced by C in Bf-bg2 motif 2 (Fig. 5). Lens βγ-crystallin domains have a conserved tyrosine and tryptophan corner, both residues being contributed by motifs 2 and 4 for each domain. These residues are conserved in equivalent motifs in Ciona protein and the Bf-bg proteins (Fig. 5, in blue). The calcium-binding residues are also conserved in both motifs 3 and 4 of Bf-bg2 and 3 (indicated in grey in Fig. 5), making it likely that the C-terminal domains in these proteins bind calcium, as does the single domain Ciona βγ-crystallin. The Bf-bg4 sequence lacks the motif 4 part of the calcium-binding site and would thus be predicted not to bind calcium. The second and fourth motifs of Bf-bg1 have an insert; that in the second is at the same position as the insert in the second motif of the first βγ-crystallin domain of AIM1 (Fig. 2).

The two domain Bf-bg proteins lack the domain pairing hydrophobics in motif 4. Motifs 2 of Bf-bg2 and Bf-bg3 each do have the full set, leading to the possibility that these N-terminal domains might pair with each other (to form homo- or heterodimers), leaving their C-terminal domains unpaired.

A protein with a sequence distantly related to βγ-crystallins has been characterized from the phylum Porifera, the sponge Geodia cydonium (Krasko et al. 1997; Giancola et al. 2005). The completed genome of the cnidarian sea anemone, Nematostella vectensis, encodes a distant βγ-crystallin relative that groups in the phylogenetic tree with the sponge sequence, as does another cnidarian hexacorrallia sequence, reported from the coral Montipora capitata (Fig. 7). The cnidarian sequences attributed to motifs 3 and 4 each have the characteristic amino acid residues for the Greek key hairpin fold, with N. vectensis having both the tyrosine and tryptophan corner residues conserved in motif 4, thus providing good evidence that the cnidarian C-terminal domains will have the double Greek key βγ-crystallin fold. The sponge motif 3 sequence is short and in the absence of three-dimensional information, it is unclear how this motif completes the double Greek key fold. The motif 2 sequences (from the three N-terminal domains) have most of the conserved residues, whereas the motif 1 sequences have hardly any, making an alignment unreliable. It is interesting that the camera eye of the cnidarian jelly fish recruited enzymes for a lens role (Piatigorsky and Kozmik 2004) even though βγ-crystallin-like proteins were present in the phylum.

Fig. 7
figure 7

Evolution of the βγ-crystallins. A maximum likelihood tree based on the alignment of both vertebrate and invertebrate βγ-crystallin sequences (see supplementary material) was constructed. Species are indicated as Ci C. intestinalis, Bf B. floridae, Gc Geodia cydonium, Mc Montipora capitata and Nv Nematostella vectensis. For clarities sake, the vertebrate branches of the tree have been collapsed; this part of the tree is shown in Fig. 3. Numbers adjacent to nodes indicate percentage bootstrap support; only values larger than 70% (of 500) are shown

Full size image

Origin of the Vertebrate βγ-Crystallin Gene Family

The phylogenetic trees presented in Figs. 3 and 7 clearly show that the present day vertebrate β- and γ-crystallin gene family must already have been present in the last common ancestor of the cyclostomes and the gnathostomes. If we superimpose the gene structure on the phylogenetic tree (Fig. 8) the obvious hypothesis is that the ancestral vertebrate βγ-crystallin gene was a single domain C. intestinalis like gene, with each motif being encoded by a single exon. Overall, the comparison of the βγ-crystallin coding sequence and gene structures is in line with the tree calculated from 146 nuclear genes showing that urochordates are closer to vertebrates than cephalochordates (Delsuc et al. 2006).

Fig. 8
figure 8

Changes in structure of the βγ-crystallin genes during evolution. Exonic regions encoding a motif are indicated as a boxed M, introns are shown by a line. Only the exonic regions encoding the domains are shown. The gene structure of the Mc βγ-crystallin gene is unknown; the sequence reported is an EST

Full size image

The common theme in βγ-crystallins is a four motif/two domain structure. Protein structure and sequence similarity between domains strongly suggest that the gene encoding the two domain protein originated from two successive duplications: the duplication of a motif coding segment to a single domain encoding gene and duplication of the single domain gene to a two domain gene (Fig. 8). The only known eukaryotic representatives of the presumptive ancestral single domain gene are the C. intestinalis βγ-crystallin and P. polycephalum Spherulin 3a genes. Our hypothesis is that gene expansion from a Ci-type βγ-crystallin single domain gene accompanied by fusion and shifts in intron position has occurred repeatedly in the various eukaryotic lineages. Even though intron positions tend to be conserved, lineage specific loss and gain is known to have happened during eukaryotic evolution (Carmel et al. 2007; Rogozin et al. 2003; Scott and Gilbert 2006).

A Ci βγ-crystallin-like gene is not only the likely ancestor of the vertebrate βγ-crystallin genes but also of the AIM1 gene: the intron positions correspond exactly. In the ancestral vertebrate, the ancestral βγ-crystallin gene underwent an explosive series of gene duplications and gene fusions to yield an AIM1 gene, an ancestral β-crystallin gene, an ancestral γ-crystrallin gene and a γN-crystallin gene. It has been suggested that the γ-crystallin gene is a descendant of a β-crystallin gene, which underwent successive loss of the between-motif introns. The γN-crystallin gene, with a γ-crystallin-type exon encoding the N-terminal domain and two β-crystallin-like exons encoding the C-terminal domain would be the retained intermediate (Wistow et al. 2005). Our results do neither support nor negate this hypothesis: both types of genes were already present in the ancestral vertebrate. Our results also do not support or negate the alternative hypothesis, namely that the intron between motifs was first lost from a Ciona type single domain encoding gene, which then duplicated and fused to form a γ-crystallin gene. The single domain intron-less γ-like gene in the P. marinus genome could represent a remnant of a putative single domain γ-crystallin precursor gene; it could equally well be the result of a mutation in the splice donor site of a γN-crystallin gene. The β-crystallin gene in the ancestral vertebrate must have duplicated and diversified further to the family of β-crystallin genes found in all vertebrate genomes. In contrast, the phylogenetic tree (Figs. 3, 4) suggests that the γ-crystallin gene remained single and duplicated only after divergence of the various vertebrate lineages. If so, the ancestral vertebrate lens is likely to have had a low level of γ-crystallin, which, extrapolating backwards from the properties of the present day vertebrate lenses, would indicate that the lens had a high water content, was soft and only a bit higher in refractive index than non-lens cells.

βγ-Crystallin-like genes encoding double Greek key folds are found in several microbial forms including bacteria, archaea and a eumycetozoa, but have not yet been found in several of the major animal phyla such as nematoda, arthropoda, mollusca, platyhelminthes, enteropneusta and echinodermata. It may be that despite the three-dimensional structural similarities between microbial and vertebrate βγ-crystallins, the genes did arise independently in the various lineages. Alternatively, loss of these genes may have been a not uncommon event in prevertebrate evolution. Although vision is common in animals, what is innovatory in vertebrates is a camera-type eye with the retina derived from ciliary photoreceptors, something we share with jellyfish, but they lack the brain (Lamb et al. 2007). The genome sequences show that there has been massive expansion of the βγ-crystallin family in vertebrates, with exon encoding of the motifs providing the necessary flexibility. The driving force behind this expansion would be the remarkable adaptations in the optical systems for different lifestyles, even amongst different teleosts that eat different kinds of food, at different depths, under different light levels (Karpestam et al. 2007). The clade-specific γ-crystallins contribute at some level to the required variations in refractive index gradients; the ability to vary the levels of specific proteins, something in which lenses appear to excel, is useful as well.