Reconstitution of Ancestral Green Visual Pigments of Zebrafish and Molecular Mechanism of Their Spectral Differentiation
http://www.100md.com
《分子生物学进展》
Department of Integrated Biosciences, Graduate School of Frontier Sciences, University of Tokyo, Chiba, Japan
Correspondence: E-mail: kawamura@k.u-tokyo.ac.jp.
Abstract
We previously reported that zebrafish have four tandemly duplicated green (RH2) opsin genes (RH2-1, RH2-2, RH2-3, and RH2-4). Absorption spectra vary widely among the four photopigments reconstituted with 11-cis retinal, with their peak absorption spectra (max) being 467, 476, 488, and 505 nm, respectively. In this study, we inferred the ancestral amino acid (aa) sequences of the zebrafish RH2 opsins by likelihood-based Bayesian statistics and reconstituted the ancestral opsins by site-directed mutagenesis. The ancestral pigment (A1) to the four zebrafish RH2 pigments and that (A3) to RH2-3 and RH2-4 showed max at 506 nm, while that (A2) to RH2-1 and RH2-2 showed a max at 474 nm, indicating that a spectral shift had occurred toward the shorter wavelength on the evolutionary lineages A1 to A2 by 32 nm, A2 to RH2-1 by 7 nm, and A3 to RH2-3 by 18 nm. Pigment chimeras and site-directed mutagenesis revealed a large contribution (15 nm) of glutamic acid to glutamine substitution at residue 122 (E122Q) to the A1 to A2 and A3 to RH2-3 spectral shifts. However, the remaining spectral differences appeared to result from complex interactive effects of a number of aa replacements, each of which has only a minor spectral contribution (1–3 nm). The four zebrafish RH2 pigments cover nearly an entire range of max distribution among vertebrate RH2 pigments and provide an excellent model to study spectral tuning mechanisms of RH2 in vertebrates.
Key Words: zebrafish ? RH2 opsins ? visual pigments ? gene duplication ? spectral differentiation ? ancestral sequence
Introduction
Vertebrate visual pigments consist of a protein moiety, an opsin, and a chromophore, either 11-cis retinal or 11-cis 3,4-dehydroretinal. The pigments reside in rod (for dim light vision) and cone (for daylight and color vision) photoreceptor cells in the retina. Opsin is a member of the G-protein–coupled receptor family with seven transmembrane (TM) domains. The retinal chromophore is bound to the opsin by a Schiff base linkage at lysine residue 296 located in the seventh TM domain (residue numbers hereafter follow those in bovine rod opsin [rhodopsin]). Free 11-cis retinal with protonated Schiff base linkage absorbs light maximally at 440 nm in organic solvents (Kito et al. 1968). Opsin perturbs the electric environment of the retinal chromophore and hence tunes its absorption spectra over a wide range of wavelengths, with the peak absorption spectra (max) ranging from 360 to 560 nm (Yokoyama 2000; Ebrey and Koutalos 2001). This spectral effect of opsins on the retinal chromophore is called "opsin shift."
Vertebrate visual opsins can be classified into five phylogenetic groups: RH1 (rod opsin or rhodopsin), RH2 (RH1-like, or green, cone opsin), SWS1 (short wavelength–sensitive type 1, or UV-blue, cone opsin), SWS2 (short wavelength–sensitive type 2, or blue, cone opsin), and M/LWS (middle to long wavelength–sensitive, or red-green, cone opsin) (Yokoyama 2000). RH1 has been a major subject of study for biophysical and biochemical properties of visual pigments. Approximately 20 amino acid (aa) residues responsible for RH1 spectral tuning have been identified (Takahashi and Ebrey 2003). M/LWS and SWS1 pigments have also been well investigated for spectral tuning mechanisms by site-directed mutagenesis.
The max values of the M/LWS pigments are mostly determined by 5 aa residues at positions 164, 181, 261, 269, and 292 (Yokoyama and Radlwimmer 1998, 1999, 2001). In typical "red" opsins, these sites are occupied by serine, histidine, tyrosine, threonine, and alanine, respectively. The aa replacements from serine to alanine at 164 (denoted S164A), H181Y, Y261F, T269A, and A292S shift the max values toward the shorter wavelength by 7, 28, 8, 15, and 27 nm, respectively (Yokoyama and Radlwimmer 2001). Effects of these aa replacements on the max shift can be detected separately and are nearly additive.
As for SWS1, spectral differentiation between the UV (max 360 nm) and violet (max 410 nm) pigments is achieved by combinations of 10 aa replacements, but individual spectral effects appear to be negligible and are largely dependent on background aa sequences, with combined effects being nonadditive and synergistic (Wilkie et al. 2000; Yokoyama, Radlwimmer, and Blow 2000; Shi, Radlwimmer, and Yokoyama 2001; Fasick, Applebury, and Oprian 2002; Shi and Yokoyama 2003; Parry et al. 2004). For example, when 7 of the 10 aa residues of the mouse UV pigment (max 358 nm) are individually replaced to the corresponding aa residues of the orthologous human blue pigment (max 414 nm) (F46T, F49L, T52F, F86L, T93P, A114G, and S118T), they do not cause any max shift, but when all mutations are introduced simultaneously, the max shifts by 50 nm (Shi, Radlwimmer, and Yokoyama 2001).
Compared to RH1, M/LWS, and SWS1 pigments, much less is known about the spectral tuning mechanisms of RH2 pigments. Two aa replacements, E122Q and M207L, result in 13- and 6-nm blueshift, respectively, to the coelacanth RH2 pigment, and the two effects are nearly additive (Yokoyama et al. 1999). This is the only report where the effect of aa replacements on the absorption spectra was experimentally tested in RH2 pigments. We previously reported that zebrafish (Danio rerio) have four tandemly duplicated RH2 opsin genes (RH2-1, RH2-2, RH2-3, and RH2-4) (Chinen et al. 2003). The reconstituted photopigments for the four genes with 11-cis retinal showed a wide variety of absorption spectra, with the max being 467, 476, 488, and 505 nm, respectively (fig. 1). Gene expression patterns differ among the four zebrafish RH2 opsin genes in quantity, temporal order, and locality in the retina. In adult retina, expression of RH2-2 is most abundant among the four RH2 opsin genes (Chinen et al. 2003). RH2-1 and RH2-2 are expressed in the central to dorsal area of the adult retina, whereas RH2-3 and RH2-4 are expressed in the peripheral area, especially in the ventral side, and in the larval retina RH2-1 is the first to be expressed (M. Takechi and S. Kawamura, unpublished data). In the present study, we aimed to clarify the evolutionary process of the max differentiation among the four zebrafish RH2 pigments and to identify the relevant aa substitutions to the spectral shifts by inferring ancestral aa sequences of the zebrafish RH2 opsins with likelihood-based Bayesian statistics and by applying site-directed mutagenesis to the reconstituted photopigments.
FIG. 1.— Normalized absorption spectra of the zebrafish RH2-1, RH2-2, RH2-3, and RH2-4 photopigments reconstituted in vitro with 11-cis retinal by Chinen et al. (2003).
Materials and Methods
Estimation of Ancestral Sequences
All fish RH2 gene sequences currently known were considered in this study: cichlid (Dimidiochromis compressiceps, GenBank accession number AF247121) (orthologous sequences from two other cichlid species, Metriaclima zebra and Labeotropheus fuelleborni, are known but were omitted in our analyses because of their close evolutionary relatedness to D. compressiceps [Carleton and Kocher 2001]), tilapia (Oreochromis niloticus, GenBank accession number AF247124), fugu (Takifugu rubripes, GenBank accession number AF226989), halibut (Hippoglossus hippoglossus, GenBank accession number AF156263), mullet (Mullus surmuletus, GenBank accession number Y18680), sand goby (Pomatoschistus minutus, GenBank accession number Y18679), medaka (Oryzias latipes, GenBank accession number AB001603), zebrafish RH2-1 (GenBank accession number AB087805), zebrafish RH2-2 (GenBank accession number AB087806), zebrafish RH2-3 (GenBank accession number AB087807), zebrafish RH2-4 (GenBank accession number AB087808), goldfish GFgr-1 (Carassius auratus, GenBank accession number L11865), goldfish GFgr-2 (GenBank accession number L11866), and Mexican cavefish (Astyanax fasciatus, GenBank accession number AH004622). Additionally, RH2 genes of coelacanth (Latimeria chalumnae, GenBank accession number AH007713), two species of lizards, American chameleon (Anolis carolinensis, GenBank accession number AH007735) and gecko (Gekko gekko, GenBank accession number M92035), and two species of birds, chicken (Gallus gallus, GenBank accession number M88178) and pigeon (Columba livia, GenBank accession number AH007731), were included in the analysis as representative out-groups. For zebrafish RH2-2 and RH2-3, cDNA sequences used for our previous spectral analysis differ from their corresponding genomic sequences registered in GenBank at 2 aa sites (phenylalanine in the genome and tyrosine in the cDNA at residue 198 [denoted F198Y] and E332D in RH2-2 and S166A and V173F in RH2-3, Chinen et al. 2003); therefore, the cDNA sequences were used in the study. The cDNA/genome difference did not affect the reconstructed phylogenetic tree topology and estimation of ancestral sequences.
Given the tree topology in figure 2 (determined in Chinen et al. 2003), the ancestral opsin sequences at each phylogenetic node were inferred by using PAML computer program with a likelihood-based Bayesian method (http://abacus.gene.ucl.ac.uk/software/paml.html) (Yang, Kumar, and Nei 1995; Yang 1997). In the computation, the empirical substitution matrix of Jones, Taylor, and Thornton (1992) (JTT model) and that of Dayhoff, Schwartz, and Orcutt (1978) (Dayhoff model) were used as mathematical models of a substitution. The certainty of each of the inferred aa was expressed as a posterior probability.
FIG. 2.— The phylogenetic tree topology of the vertebrate RH2 opsins used for the estimation of their ancestral sequences. The max values (in nm) are indicated in parentheses for the opsin photopigments of which absorption spectra have been measured in vitro. The max values of the ancestral opsins at nodes A1 (Ancestor 1), A2 (Ancestor 2), and A3 (Ancestor 3) based on the JTT model are also indicated. The branches A, B, and C denote the branches A1-A2, A2–zebrafish RH2-1, and A3–zebrafish RH2-3, respectively, and are emphasized with thick lines. The max shifts that occurred along the branches A, B, and C are indicated.
Reconstitution of Ancestral and Mutant Pigments
Point mutations were introduced into the cDNAs of the zebrafish four RH2 opsins (Chinen et al. 2003) cloned into the pMT5 expression vector (Khorana et al. 1988) by using QuikChange site-directed mutation kit (Stratagene, La Jolla, Calif.). Various chimeric pigments were constructed by ligating the restriction fragments from the cDNA clones using restriction sites for AseI, HindIII, and PstI (fig. 3). All mutagenized cDNAs were sequenced to confirm that no spurious mutation was incorporated. Sequencing was carried out by using Thermo Sequence cycle sequencing kit (Amersham, Piscataway, N.J.) with dye-labeled primers and the LI-COR 4200L-1 automated DNA sequencer.
FIG. 3.— Amino acid sequences of the ancestral RH2 opsins estimated with the JTT model of the aa substitution and the contemporary zebrafish RH2 opsins. For phylogenetic positions of the ancestral opsins (Ancestors 1–3) see figure 2. The seven TM regions and the second extracellular loop (E2) region (Palczewski et al. 2000) are indicated with solid and dotted lines, respectively. Amino acid pairs considered for site-directed mutagenesis are boxed. Amino acid sites surrounding the 11-cis retinal are depicted in red. Position 122 is highlighted with boldface letters. Positions of AseI, HindIII, and PstI restriction sites are indicated with arrows.
The pMT5 expression vector contains the last 15 aa of the bovine rod opsin necessary for immunoaffinity purification by 1D4 monoclonal antibody (Molday and MacKenzie 1983). Each pMT5-cDNA clone was transfected into cultured COS-1 cells (RIKEN Cell Bank, Tsukuba, Japan); the resuspended cells were incubated with 5 μM 11-cis retinal (Storm Eye Institute, Medical University of South Carolina, Charleston, S.C.) and solubilized with 1% dodecyl maltoside. The resulting pigments were purified using the immobilized 1D4 (The Cell Culture Center, Minneapolis, Minn.) as in Kawamura and Yokoyama (1998). The UV-visible absorption spectrum of each pigment was recorded from 250 to 650 nm by 1-nm intervals using the Hitachi U3010 dual beam spectrometer at 20°C for a minimum of five times in the dark and for a minimum of five more times after 3 min of light exposure (with a <440-nm cutoff filter) as in Kawamura and Yokoyama (1998). The Savitzky-Golay's least squares smoothing (Gorry 1990) was carried out for each absorbance curve with 100 repetitions to eliminate spurious spikes. All max values in this study were taken from the dark spectra.
Results
Ancestral Pigments of Zebrafish RH2
We previously analyzed phylogenetic positions of the zebrafish RH2-1, RH2-2, RH2-3, and RH2-4 opsin genes among teleost RH2 opsin genes (Chinen et al. 2003), and figure 2 summarizes the tree topology among them. The ancestral opsin of the four zebrafish RH2 opsins was designated Ancestor 1 (A1), that of RH2-1 and RH2-2 opsins was designated Ancestor 2 (A2), and that of RH2-3 and RH2-4 opsins was designated Ancestor 3 (A3). Amino acid sequences of these ancestral opsins were estimated using the likelihood-based Bayesian statistics with the JTT and Dayhoff models of aa replacements on the basis of the phylogenetic tree topology given in figure 2. Figure 3 shows the aa sequences of Ancestors 1–3 inferred with the JTT model and those of RH2-1–RH2-4 (the residue numbers of the sequences coincide with the bovine rod opsin, except for 11 aa in the C-terminal). Average posterior probability over the entire aa sites was 0.98 in all the three ancestral sequences with both JTT and Dayhoff models. Amino acid sites given less than 0.9 posterior probabilities by either JTT or Dayhoff model are listed in table 1. Including these lower probability sites, estimated sequences by the two models were identical except at one site (residue 88) of Ancestor 2 where phenylalanine and isoleucine were inferred by the JTT and the Dayhoff models, respectively (table 1).
Table 1 Amino Acids Given Less Than 0.9 Posterior Probabilities by Either JTT or Dayhoff Model in Ancestral RH2 aa Sequences
The Ancestor 1 and Ancestor 3 pigments were reconstituted by introducing 13 and 9 aa changes, respectively, into the RH2-4 pigment, while the Ancestor 2 pigment was reconstituted by introducing 5 aa changes into the RH2-2 pigment according to the aa sequences inferred with the JTT model (fig. 3). Absorption spectra of the reconstituted ancestral pigments are shown in figure 4. The max values of Ancestors 1, 2, and 3 were measured to be 506, 474, and 506 nm, respectively (table 2). All pigments were shown to be photoreactive by exhibiting 380-nm max of all-trans retinal upon light exposure (dark-light difference spectra are shown in fig. 4 insets). To check whether the aa discrepancy at residue 88 of Ancestor 2 between JTT and Dayhoff models had any influence on the estimation of its ancestral absorption spectrum, the Dayhoff version of the Ancestor 2 pigment (A2_F88I) was reconstituted. The resulting max was measured to be 476 nm (table 2) and was shifted by 2 nm from the JTT-based one (474 nm). These results indicate that a major short-wave shift of the absorption spectra occurred in three evolutionary lineages, Ancestors 1 to 2 (branch A) by 30–32 nm, Ancestor 2 to RH2-1 (branch B) by 7–9 nm, and Ancestor 3 to RH2-3 (branch C) by 18 nm (fig. 2). Hereafter, JTT-based Ancestor 2 sequence was used for a template of mutagenesis analyses.
FIG. 4.— Absorption spectra of ancestral RH2 pigments of zebrafish measured in the dark. Insets: dark-light difference spectra.
Table 2 max of Zebrafish RH2 Visual Pigments and Their Reconstituted Ancestral Pigments Measured In Vitro
Effect of E122Q
From the study of the coelacanth RH2 pigment by Yokoyama et al. (1999), it is known that aa changes E122Q and L207M could cause a large spectral shift of RH2 absorption spectra toward a shorter wavelength by approximately 13 and 6 nm, respectively. In the case of zebrafish RH2 opsins, E122Q was inferred at two branches, A and C, while residue 207 was invariable (all methionine) (fig. 3). The site 122 was variable among the vertebrate RH2 sequences, either E or Q, and the posterior probabilities for E at the site in Ancestors 1 and 3 were relatively low (0.766 and 0.863, respectively; see table 1). However, the Dayhoff model also supported the results (0.780 and 0.879, respectively). To test the spectral effect of residue 122 on the zebrafish RH2 pigments, we introduced reverse mutation Q122E into the Ancestor 2 and RH2-3 pigments (resulting pigments denoted A2_Q122E and RH2-3_Q122E, respectively). The max of the A2_Q122E was 489 nm (table 3) and was 15 nm longer than that of Ancestor 2 (474 nm), and the max of the RH2-3_Q122E was 502 nm (table 4) and was 14 nm longer than that of RH2-3 (488 nm) (fig. 5). Therefore, a single aa replacement E122Q can explain nearly half (15/32) of the spectral difference between Ancestors 1 and 2 and a majority (14/18) of the spectral difference between Ancestor 3 and RH2-3.
Table 3 Spectral Effects of Branch A Mutations: Introduction of Ancestor 1 Sequences to Ancestor 2
Table 4 Spectral Effects of Branch B and Branch C Mutations
FIG. 5.— Spectral effect of the Q122E mutation to the Ancestor 2 and RH2-3 pigments. Absorbance is normalized to set the peak value as 1 for all pigments. Absorbance curves are depicted with solid and broken lines for mutant and intact pigments, respectively.
Mutations in Branch A
There are 32 aa that are different between Ancestors 1 and 2, other than the E/Q difference at the residue 122 (fig. 3). To explore aa replacements that may explain the remaining 17-nm difference between Ancestors 1 and 2, we first constructed chimeric pigments where a segment of Ancestor 2 was replaced by that of Ancestor 1. Amino acids 1–99, 100–234, and 235–349 of the Ancestor 2 pigment were replaced using restriction sites for AseI and PstI (fig. 3) (resulting chimeric pigments designated A1(99)A2, A2(99)A1(234)A2, and A2(234)A1, respectively). The max of the three pigments were 480, 495, and 478 nm, being 6, 21, and 4 nm longer than that of Ancestor 2, respectively (table 3). The sum of these effects is 31 nm, which approximates the difference (32 nm) between the max values of Ancestors 1 and 2. Because the middle segment contains residue 122, the net effect of the middle segment excluding that of E122Q was expected to be approximately 6 nm.
We then introduced point mutations into Ancestor 2 or A2_Q122E to further locate the residues responsible for spectral differentiation. Among the 32 aa that are different, 23 are located in the putative TM regions and 1 is in the second extracellular (E2) loop (fig. 3). These are the domains which contain aa positions that together constitute the retinal-binding pocket (Palczewski et al. 2000). All aa replacements involved in spectral tuning have been identified in the TM and E2 regions (Yokoyama 2002; Takahashi and Ebrey 2003), and these regions were subjected to mutagenesis (fig. 3).
In region 1–99, there are 9 aa differences in the TM domains between Ancestors 1 and 2 (see fig. 3). All of these mutations were introduced into Ancestor 2 singly or doubly when the two sites were in close proximity (table 3). Maximum max shift by these mutations was only +3 nm (by F88C). Double mutations C97S/S98A resulted in no recognizable absorbance peak. When these mutations were introduced to the A2_F88C, the max shift from A2_F88C was negligible (–1 nm). Simple addition of these single- or double-replacement effects is only +1 nm (or 0 nm if effect of C97S/S98A is counted as –1 nm) (table 3). This is at odds with the results of the segment swapping of region 1–99, which showed a +6 nm max shift. Amino acid difference (P/T at residue 27; fig. 3) was not considered for the mutagenesis, is located in the N-terminal tail, and is unlikely to have any spectral effect. These results suggest that individual spectral effects of aa replacements in this region are not additive but rather synergistically cooperate together.
In the region 100–234, there are 9 aa differences in the TM domains other than the E/Q difference at residue 122 and one difference in the E2 region between Ancestors 1 and 2 (see fig. 3). All of these mutations were introduced into A2_Q122E. Maximum max shift from A2_Q122E was +4 nm (by N151S), followed by +3 nm (by S209I) and +2 nm (by T218I) (table 3). A mutation in the E2 region, T185C, resulted in an opposite spectral shift, –4 nm, though the pigment's absorption peak was low and not clearly discernable (table 3). Simple addition of these effects and other minor ±1-nm shifts count +5 nm, which is close to the +6-nm difference between A2_Q122E and A2(99)A1(234)A2. However, when the mutations causing +2- to +4-nm shifts were introduced together (A2_Q122E/N151S/S209I/T218I), combined effect for the max shift from A2_Q122E was only 2 nm (table 3). These results suggest that individual spectral effects in this region are neither additive nor synergistic but are partly regressive. The net effect of them could vary depending on the physicochemical background of the residues making up the region.
In the region 235–349, there are 5 aa differences in the TM domains between Ancestors 1 and 2 (see fig. 3), all of which were tested. Individual mutation effects were all minor, +1 or +2 nm (table 3). Simple addition of the effects adds up to +9 nm, which is greater than the +4-nm difference between the max values of Ancestor 2 and A2(234)A1. When mutations with +2-nm effect were introduced together (V266T/F271V/A287F/A297S), the spectral shift was only +3 nm and the combined effect appeared to be regressive. This total effect is, however, close to the segment-swapping effect of 4 nm, and the regressive spectral effects can explain the spectral shift by this segment from these sites.
Mutations in Branch B
To explore aa replacements that explain the 7-nm max difference between RH2-1 and Ancestor 2, we first constructed two chimeric pigments, where aa 1–144 and 145–349 of Ancestor 2 were replaced with those of RH2-1 using HindIII restriction site (fig. 3) (the resulting chimeric pigments designated RH2-1(144)A2 and A2(144)RH2-1, respectively). Whereas RH2-1(144)A2 pigment was successfully reconstituted, functional A2(144)RH2-1 pigment with a recognizable absorption peak was not formed. However, the max value of RH2-1(144)A2 was identical to that of RH2-1 (467 nm) (table 4), suggesting that aa replacements responsible for the max difference between RH2-1 and Ancestor 2 pigments lie in the 1–144 region. We therefore introduced point mutations to the TM domains in the 1–144 region.
There are 11 aa differences in the TM domains in the region 1–144 between Ancestor 2 and RH2-1 (see fig. 3). All of these mutations were introduced into RH2-1, except at residue 88 for which Ancestor 2 was used as a template (A2_F88I) (table 2). Unexpectedly, many mutations resulted in opposite spectral shifts (K36Q, L46F, I49C, V60L, L108T, and M112I causing a short-wave shift and F88I to Ancestor 2 a long-wave shift) (table 4). S94T had no spectral effect, and V111A showed only a 1-nm long-wave shift. A double mutation L99I/V100N resulted in no recognizable peak. Simple addition of the spectral effects by these single point mutations counts as a –16-nm shift from RH2-1 (with I88F assumed to cause –2-nm shift from RH2-1), contrasting with the +7-nm shift caused by the segment swapping of the entire 1–144 region. When some mutations were introduced together (RH2-1_V111A/M112I, RH2-1_S94T/V111A/M112I and RH2-1_L46F/I49C/S94T/V111A/M112I; see table 4), the amount of spectral shift markedly differed from their sum, e.g., +4-nm shift in RH2-1_L46F/I49C/S94T/V111A/M112I, while the sum of the individual effect was –7-nm. It has been reported that residues 46 and 49 are part of the spectral tuning sites of SWS1 opsins which show individually unrecognizable shifts, but large spectral effects occur when combined together (Shi, Radlwimmer, and Yokoyama 2001). These results indicate that the spectral effect of these aa replacements can vary depending on their physicochemical environment as shown for mutations in branch A.
Mutations in Branch C
In the evolutionary branch C between Ancestor 3 and RH2-3, there are 18 aa replacements in the TM and E2 domains (see fig. 3). However, a majority of the spectral difference (14/18) was explained by a single aa replacement E122Q (fig. 5, right panel). E122Q was estimated to have occurred independently in branches A and C (fig. 3). Such a parallel substitution was also inferred at residues 50 (L50F in branches A and C), 60 (L60V in B and C), 151 (S151N in A and C), 162 (V162F in B and V162I in C), 165 (M165S in A, S165C in B, and M165L in C), 209 (I209S in A and I209C in C), 214 (V214I in B and L214F in C), 218 (I218T in A and C), 270 (T270S in A and C), and 300 (L300V in B and L300I in C) (fig. 3). Among these, 2 nm or more of spectral shift was experimentally verified with an expected direction in branch A mutations N151S, S209I, and T218I introduced to Ancestor 2 (table 3). We then tested the spectral effect of a combined mutation Q122E/N151S/C209I/T218I introduced to RH2-3. The observed spectral shift from RH2-3 was +16 nm (table 4) and was close to the actual spectral difference of Ancestor 3 from RH2-3 (+18 nm), although contributions from the other 14 aa substitutions cannot be ruled out.
Discussion
We have previously shown that zebrafish have four tandemly duplicated RH2 opsin genes in the genome and that the max of their 11-cis retinal–based photopigments differ from one another in the range of 505 nm (RH2-4) to 467 nm (RH2-1) with approximately 10 to 20-nm intervals (fig. 1) (Chinen et al. 2003). In this study, we estimated ancestral aa sequences of the four RH2 opsins using the likelihood-based Bayesian statistics with the JTT and Dayhoff models of aa replacements and reconstituted the ancestral photopigments with 11-cis retinal. We showed that the 505-nm max of RH2-4, a typical value among vertebrate RH2 photopigments with 11-cis retinal, is the ancestral max value of the four pigments, and that of the other three opsins achieved short-wave shifts which occurred after every gene duplication in only one of the two sister genes with 30-, 20-, or 10-nm magnitude (fig. 2).
Figure 6 summarizes the evolutionary changes that occurred in the phylogenetic clade derived from Ancestor 1. In branch A, 25 aa changes occurred in the TM + E2 regions, and we tested all of them for their spectral effects (table 3). The E122Q accounted for about half (15 nm) of the spectral difference (32 nm) between A1 and A2 pigments. The 32-nm total difference was nearly perfectly explained by linear addition of the spectral effects of three segmental replacements (table 3). However, within the segments, such additivity no longer held for individual aa mutations (table 3). There are 9, 11, and 5 mutations in TM + E2 regions in segments 1–99, 100–234, and 235–349, respectively (fig. 3). These effects were individually small and not additive and may have been even partly regressive. To clarify the aa replacements, other than E122Q, that explain the spectral shift, it is necessary to examine all the combinations of aa mutations in each segment. The same holds true for mutations in branch B. We verified that segment 1–144 is responsible for the spectral difference of 7 nm between A2 and RH2-1 (table 4). There are 11 mutations in the TM regions in the segment (fig. 3). We tested all of their spectral effects and showed that their individual effects were small and many of them were oriented to the opposite direction (table 4). It is necessary to explore the combinations of the mutations that account for the spectral difference between A2 and RH2-1.
FIG. 6.— The summary of evolutionary changes that occurred in the phylogenetic clade derived from the Ancestor 1 (A1) opsin. Branches associated with max shifts are depicted with thick lines. The numbers of aa substitutions in the TM + E2 regions (219-aa length) and those in the other regions (130-aa length) are indicated along each branch. Their proportions in the designated regions are indicated in parentheses. The ancestral opsin of the two goldfish opsins (GFgr-1 and GFgr-2) is designated GF and that of GF and Ancestor 3 is designated ZG. The measured (for A1, A2, and A3) or inferred (for ZG and GF) max values of the ancestral opsins are indicated at their respective nodes. The spectral ZF represents zebrafish shifts and associated aa replacements are indicated for the branches A, B, and C.
It is noted that the evolutionary branches associated with spectral shifts (i.e., branches A, B, and C and GF-GFgr-1) have a larger proportion of aa substitution in TM + E2 regions than in the other regions, whereas the other branches have similar rates of aa substitutions between the two regions (fig. 6). This is consistent with our mutagenesis results that not only a single aa substitution, E122Q with a large spectral effect, but also multiple aa substitutions with individually minor spectral effects are involved in the spectral tuning of the zebrafish RH2. In this regard, our mutagenesis conducted for branch C may not be sufficient, where 18 aa changes occurred in TM + E2 regions (fig. 3) but only four of them including E122Q were tested (table 4), although E122Q explains 14-nm out of 18-nm difference between A3 and RH2-3.
Figure 6 also indicates that spectral differentiation of zebrafish RH2 opsins occurred independently of that of goldfish RH2 opsins. The ancestral opsin of the two goldfish opsins (GFgr-1 and GFgr-2; Johnson et al. 1993) was designated GF and that of GF and Ancestor 3 was designated ZG. The ZG was inferred to have an identical aa sequence with Ancestor 3 and must have a max of 506 nm as well. The max of GF would also be 506 nm because those of ZG and GFgr-2 are both 506 nm and because the 6 aa changes in the TM + E2 regions in the ZG-GF branch are all conservative and outside the retinal-surrounding sites (sequences not shown).
The recent X-ray diffraction analysis of bovine rod opsin revealed its three-dimensional crystal structure at 2.8-? resolution (Palczewski et al. 2000). Residue 122 is one of the 27 aa forming the retinal-binding pocket which are located within 4.5 ? from 11-cis retinal (residues 43, 44, 47, 94, 113, 117, 118, 120, 121, 122, 178, 181, 186, 187, 188, 189, 191, 207, 211, 212, 261, 265, 268, 269, 293, 295, and 296) (Palczewski et al. 2000). In addition, a total of 38 aa sites, including the 27 sites, were identified as surrounding 11-cis retinal (Menon, Han, and Sakmar 2001; Takahashi and Ebrey 2003) (the sites indicated in red in fig. 3). Residue 122 interacts directly with ?-ionone ring of 11-cis retinal, and its replacement has been verified to result in significant spectral shifts in various visual pigments (Takahashi and Ebrey 2003). In general, there are two factors controlling the opsin shift: (1) electrostatic interaction between the glutamic acid counterion at residue 113 and the protonated Schiff base of retinal at lysine residue 296 and (2) electrostatic interaction between the retinal body and charged or polar groups of aa surrounding it (Takahashi and Ebrey 2003). In both cases, aa replacements surrounding retinal can directly influence absorption spectra of visual pigments, and hence their spectral effects tend to be additive as manifested in the cases of M/LWS pigments (Yokoyama and Radlwimmer 2001) and coelacanth RH2 and RH1 pigments (Yokoyama et al. 1999). The 5 aa sites 164, 181, 261, 269, and 292, which are known to exert major contributions to tuning of the M/LWS opsins, the two sites 122 and 207 in the coelacanth RH2 opsin, and the two sites 122 and 292 in the coelacanth RH1 opsin are all in the retinal-surrounding sites (see fig. 3).
On the other hand, spectral effects of aa replacements outside the retinal-surrounding sites would be generally indirect, small, nonadditive, and dependent on the molecular environment of the aa sites concerned. These replacements possibly work through alteration of conformation or hydrophobicity of TM domains. The spectral tuning mechanism of the violet/uv (SWS1) pigments appears to be a typical example of such an indirect interaction mechanism, where combinations of 10 aa sites (46, 49, 52, 86, 90, 93, 97, 114, 116, and 118) are involved in the spectral differentiation and seven of them (sites other than 90, 114, and 118) are located outside the retinal-surrounding sites (Wilkie et al. 2000; Yokoyama, Radlwimmer, and Blow 2000; Shi, Radlwimmer, and Yokoyama 2001; Fasick, Applebury, and Oprian 2002; Shi and Yokoyama 2003; Parry et al. 2004). The tuning sites possibly affect trafficking of water molecules at the retinal-binding pocket by altering net hydrophobicity of TM domains with other residues and control protonation and deprotonation status of the Schiff base (Shi, Radlwimmer, and Yokoyama 2001). Whereas spectral effects of the aa replacements in the retinal-surrounding region are likely to be universal beyond differences of opsin groups or animal species because of their directness of the interaction with retinal, spectral effects of the aa replacements outside the region found in a system may be of little general implication to another system because of their indirectness of the interaction with retinal.
All the mutations, other than E122Q tested in this study, fall outside the retinal-surrounding region (fig. 3) (site 94 is in the region but see the following explanations). These mutations resulted in only minor spectral effects with 0 to 4-nm shift (tables 3 and 4). S94A is known to cause a 14-nm blueshift in newt SWS2 pigment (Takahashi and Ebrey 2003), but the substitution in our case was between the same polar residues with a hydroxyl group, S94T, and resulted in no recognizable spectral shift (table 4). Characteristically and consistently with their locations relative to retinal, combined effects of these minor sites were not additive and were, in some combinations, even opposite in direction of spectral shift from the shift direction in individual mutations. However, this nonlinear nature of cumulative effects from multiple aa sites with individual minor effects does play an important role in the spectral differentiation of the zebrafish RH2 opsins.
In zebrafish RH2 opsins, spectral diversification is achieved by direct and indirect interaction mechanisms with retinal, the former by residue 122 in the retinal-surrounding region with a large spectral effect and the latter by multiple residues outside it with individually minor and collectively nonadditive effects. This pattern of spectral differentiation could represent a general paradigm to account for the spectral diversity of vertebrate RH2 pigments, considering that a range of 470–510 nm is covered by the four zebrafish pigments.
Acknowledgements
This study was supported by Grants-in-Aid for Scientific Research (B) (12440243) and Exploratory Research (13874105) from the Japan Society for the Promotion of Science (JSPS) to S.K. and a Grant-in-Aid for JSPS Fellows (14-08073) to A.C. The manuscript was proofread by BioMed Proofreading Service.
References
Carleton, K. L., and T. D. Kocher. 2001. Cone opsin genes of African cichlid fishes: tuning spectral sensitivity by differential gene expression. Mol. Biol. Evol. 18:1540–1550.
Chinen, A., T. Hamaoka, Y. Yamada, and S. Kawamura. 2003. Gene duplication and spectral diversification of cone visual pigments of zebrafish. Genetics 163:663–675.
Dayhoff, M. O., R. M. Schwartz, and B. C. Orcutt. 1978. A model of evolutionary change in proteins. Pp. 345–352 in M. O. Dayhoff, ed. Atlas of protein sequence and structure. National Biomedical Research Foundation, Silver Spring, Md.
Ebrey, T., and Y. Koutalos. 2001. Vertebrate photoreceptors. Prog. Retin. Eye Res. 20:49–94.
Fasick, J. I., M. L. Applebury, and D. D. Oprian. 2002. Spectral tuning in the mammalian short-wavelength sensitive cone pigments. Biochemistry 41:6860–6865.
Gorry, P. A. 1990. General least-squares smoothing and differentiation by the convolution (Savitzky-Golay) method. Anal. Chem. 62:570–573.
Johnson, R. L., K. B. Grant, T. C. Zankel, M. F. Boehm, S. L. Merbs, J. Nathans, and K. Nakanishi. 1993. Cloning and expression of goldfish opsin sequences. Biochemistry 32:208–214.
Jones, D. T., W. R. Taylor, and J. M. Thornton. 1992. The rapid generation of mutation data matrices from protein sequences. Comput. Appl. Biosci. 8:275–282.
Kawamura, S., and S. Yokoyama. 1998. Functional characterization of visual and nonvisual pigments of American chameleon (Anolis carolinensis). Vision Res. 38:37–44.
Khorana, H. G., B. E. Knox, E. Nasi, R. Swanson, and D. A. Thompson. 1988. Expression of a bovine rhodopsin gene in Xenopus oocytes: demonstration of light-dependent ionic currents. Proc. Natl. Acad. Sci. USA 85:7917–7921.
Kito, Y., T. Suzuki, M. Azuma, and Y. Sekoguti. 1968. Absorption spectrum of rhodopsin denatured with acid. Nature 218:955–957.
Menon, S. T., M. Han, and T. P. Sakmar. 2001. Rhodopsin: structural basis of molecular physiology. Physiol. Rev. 81:1659–1688.
Molday, R. S., and D. MacKenzie. 1983. Monoclonal antibodies to rhodopsin: characterization, cross-reactivity, and application as structural probes. Biochemistry 22:653–660.
Palczewski, K., T. Kumasaka, T. Hori et al. (12 co-authors). 2000. Crystal structure of rhodopsin: a G protein-coupled receptor. Science 289:739–745.
Parry, J. W., S. Poopalasundaram, J. K. Bowmaker, and D. M. Hunt. 2004. A novel amino acid substitution is responsible for spectral tuning in a rodent violet-sensitive visual pigment. Biochemistry 43:8014–8020.
Shi, Y., F. B. Radlwimmer, and S. Yokoyama. 2001. Molecular genetics and the evolution of ultraviolet vision in vertebrates. Proc. Natl. Acad. Sci. USA 98:11731–11736.
Shi, Y., and S. Yokoyama. 2003. Molecular analysis of the evolutionary significance of ultraviolet vision in vertebrates. Proc. Natl. Acad. Sci. USA 100:8308–8313.
Takahashi, Y., and T. G. Ebrey. 2003. Molecular basis of spectral tuning in the newt short wavelength sensitive visual pigment. Biochemistry 42:6025–6034.
Wilkie, S. E., P. R. Robinson, T. W. Cronin, S. Poopalasundaram, J. K. Bowmaker, and D. M. Hunt. 2000. Spectral tuning of avian violet- and ultraviolet-sensitive visual pigments. Biochemistry 39:7895–7901.
Yang, Z. 1997. PAML: a program package for phylogenetic analysis by maximum likelihood. Comput. Appl. Biosci. 13:555–556.
Yang, Z., S. Kumar, and M. Nei. 1995. A new method of inference of ancestral nucleotide and amino acid sequences. Genetics 141:1641–1650.
Yokoyama, S. 2000. Molecular evolution of vertebrate visual pigments. Prog. Retin. Eye Res. 19:385–419.
———. 2002. Molecular evolution of color vision in vertebrates. Gene 300:69–78.
Yokoyama, S., and F. B. Radlwimmer. 1998. The "five-sites" rule and the evolution of red and green color vision in mammals. Mol. Biol. Evol. 15:560–567.
———. 1999. The molecular genetics of red and green color vision in mammals. Genetics 153:919–932.
———. 2001. The molecular genetics and evolution of red and green color vision in vertebrates. Genetics 158:1697–1710.
Yokoyama, S., F. B. Radlwimmer, and N. S. Blow. 2000. Ultraviolet pigments in birds evolved from violet pigments by a single amino acid change. Proc. Natl. Acad. Sci. USA 97:7366–7371.
Yokoyama, S., H. Zhang, F. B. Radlwimmer, and N. S. Blow. 1999. Adaptive evolution of color vision of the Comoran coelacanth (Latimeria chalumnae). Proc. Natl. Acad. Sci. USA 96:6279–6284.(Akito Chinen, Yoshifumi M)
Correspondence: E-mail: kawamura@k.u-tokyo.ac.jp.
Abstract
We previously reported that zebrafish have four tandemly duplicated green (RH2) opsin genes (RH2-1, RH2-2, RH2-3, and RH2-4). Absorption spectra vary widely among the four photopigments reconstituted with 11-cis retinal, with their peak absorption spectra (max) being 467, 476, 488, and 505 nm, respectively. In this study, we inferred the ancestral amino acid (aa) sequences of the zebrafish RH2 opsins by likelihood-based Bayesian statistics and reconstituted the ancestral opsins by site-directed mutagenesis. The ancestral pigment (A1) to the four zebrafish RH2 pigments and that (A3) to RH2-3 and RH2-4 showed max at 506 nm, while that (A2) to RH2-1 and RH2-2 showed a max at 474 nm, indicating that a spectral shift had occurred toward the shorter wavelength on the evolutionary lineages A1 to A2 by 32 nm, A2 to RH2-1 by 7 nm, and A3 to RH2-3 by 18 nm. Pigment chimeras and site-directed mutagenesis revealed a large contribution (15 nm) of glutamic acid to glutamine substitution at residue 122 (E122Q) to the A1 to A2 and A3 to RH2-3 spectral shifts. However, the remaining spectral differences appeared to result from complex interactive effects of a number of aa replacements, each of which has only a minor spectral contribution (1–3 nm). The four zebrafish RH2 pigments cover nearly an entire range of max distribution among vertebrate RH2 pigments and provide an excellent model to study spectral tuning mechanisms of RH2 in vertebrates.
Key Words: zebrafish ? RH2 opsins ? visual pigments ? gene duplication ? spectral differentiation ? ancestral sequence
Introduction
Vertebrate visual pigments consist of a protein moiety, an opsin, and a chromophore, either 11-cis retinal or 11-cis 3,4-dehydroretinal. The pigments reside in rod (for dim light vision) and cone (for daylight and color vision) photoreceptor cells in the retina. Opsin is a member of the G-protein–coupled receptor family with seven transmembrane (TM) domains. The retinal chromophore is bound to the opsin by a Schiff base linkage at lysine residue 296 located in the seventh TM domain (residue numbers hereafter follow those in bovine rod opsin [rhodopsin]). Free 11-cis retinal with protonated Schiff base linkage absorbs light maximally at 440 nm in organic solvents (Kito et al. 1968). Opsin perturbs the electric environment of the retinal chromophore and hence tunes its absorption spectra over a wide range of wavelengths, with the peak absorption spectra (max) ranging from 360 to 560 nm (Yokoyama 2000; Ebrey and Koutalos 2001). This spectral effect of opsins on the retinal chromophore is called "opsin shift."
Vertebrate visual opsins can be classified into five phylogenetic groups: RH1 (rod opsin or rhodopsin), RH2 (RH1-like, or green, cone opsin), SWS1 (short wavelength–sensitive type 1, or UV-blue, cone opsin), SWS2 (short wavelength–sensitive type 2, or blue, cone opsin), and M/LWS (middle to long wavelength–sensitive, or red-green, cone opsin) (Yokoyama 2000). RH1 has been a major subject of study for biophysical and biochemical properties of visual pigments. Approximately 20 amino acid (aa) residues responsible for RH1 spectral tuning have been identified (Takahashi and Ebrey 2003). M/LWS and SWS1 pigments have also been well investigated for spectral tuning mechanisms by site-directed mutagenesis.
The max values of the M/LWS pigments are mostly determined by 5 aa residues at positions 164, 181, 261, 269, and 292 (Yokoyama and Radlwimmer 1998, 1999, 2001). In typical "red" opsins, these sites are occupied by serine, histidine, tyrosine, threonine, and alanine, respectively. The aa replacements from serine to alanine at 164 (denoted S164A), H181Y, Y261F, T269A, and A292S shift the max values toward the shorter wavelength by 7, 28, 8, 15, and 27 nm, respectively (Yokoyama and Radlwimmer 2001). Effects of these aa replacements on the max shift can be detected separately and are nearly additive.
As for SWS1, spectral differentiation between the UV (max 360 nm) and violet (max 410 nm) pigments is achieved by combinations of 10 aa replacements, but individual spectral effects appear to be negligible and are largely dependent on background aa sequences, with combined effects being nonadditive and synergistic (Wilkie et al. 2000; Yokoyama, Radlwimmer, and Blow 2000; Shi, Radlwimmer, and Yokoyama 2001; Fasick, Applebury, and Oprian 2002; Shi and Yokoyama 2003; Parry et al. 2004). For example, when 7 of the 10 aa residues of the mouse UV pigment (max 358 nm) are individually replaced to the corresponding aa residues of the orthologous human blue pigment (max 414 nm) (F46T, F49L, T52F, F86L, T93P, A114G, and S118T), they do not cause any max shift, but when all mutations are introduced simultaneously, the max shifts by 50 nm (Shi, Radlwimmer, and Yokoyama 2001).
Compared to RH1, M/LWS, and SWS1 pigments, much less is known about the spectral tuning mechanisms of RH2 pigments. Two aa replacements, E122Q and M207L, result in 13- and 6-nm blueshift, respectively, to the coelacanth RH2 pigment, and the two effects are nearly additive (Yokoyama et al. 1999). This is the only report where the effect of aa replacements on the absorption spectra was experimentally tested in RH2 pigments. We previously reported that zebrafish (Danio rerio) have four tandemly duplicated RH2 opsin genes (RH2-1, RH2-2, RH2-3, and RH2-4) (Chinen et al. 2003). The reconstituted photopigments for the four genes with 11-cis retinal showed a wide variety of absorption spectra, with the max being 467, 476, 488, and 505 nm, respectively (fig. 1). Gene expression patterns differ among the four zebrafish RH2 opsin genes in quantity, temporal order, and locality in the retina. In adult retina, expression of RH2-2 is most abundant among the four RH2 opsin genes (Chinen et al. 2003). RH2-1 and RH2-2 are expressed in the central to dorsal area of the adult retina, whereas RH2-3 and RH2-4 are expressed in the peripheral area, especially in the ventral side, and in the larval retina RH2-1 is the first to be expressed (M. Takechi and S. Kawamura, unpublished data). In the present study, we aimed to clarify the evolutionary process of the max differentiation among the four zebrafish RH2 pigments and to identify the relevant aa substitutions to the spectral shifts by inferring ancestral aa sequences of the zebrafish RH2 opsins with likelihood-based Bayesian statistics and by applying site-directed mutagenesis to the reconstituted photopigments.
FIG. 1.— Normalized absorption spectra of the zebrafish RH2-1, RH2-2, RH2-3, and RH2-4 photopigments reconstituted in vitro with 11-cis retinal by Chinen et al. (2003).
Materials and Methods
Estimation of Ancestral Sequences
All fish RH2 gene sequences currently known were considered in this study: cichlid (Dimidiochromis compressiceps, GenBank accession number AF247121) (orthologous sequences from two other cichlid species, Metriaclima zebra and Labeotropheus fuelleborni, are known but were omitted in our analyses because of their close evolutionary relatedness to D. compressiceps [Carleton and Kocher 2001]), tilapia (Oreochromis niloticus, GenBank accession number AF247124), fugu (Takifugu rubripes, GenBank accession number AF226989), halibut (Hippoglossus hippoglossus, GenBank accession number AF156263), mullet (Mullus surmuletus, GenBank accession number Y18680), sand goby (Pomatoschistus minutus, GenBank accession number Y18679), medaka (Oryzias latipes, GenBank accession number AB001603), zebrafish RH2-1 (GenBank accession number AB087805), zebrafish RH2-2 (GenBank accession number AB087806), zebrafish RH2-3 (GenBank accession number AB087807), zebrafish RH2-4 (GenBank accession number AB087808), goldfish GFgr-1 (Carassius auratus, GenBank accession number L11865), goldfish GFgr-2 (GenBank accession number L11866), and Mexican cavefish (Astyanax fasciatus, GenBank accession number AH004622). Additionally, RH2 genes of coelacanth (Latimeria chalumnae, GenBank accession number AH007713), two species of lizards, American chameleon (Anolis carolinensis, GenBank accession number AH007735) and gecko (Gekko gekko, GenBank accession number M92035), and two species of birds, chicken (Gallus gallus, GenBank accession number M88178) and pigeon (Columba livia, GenBank accession number AH007731), were included in the analysis as representative out-groups. For zebrafish RH2-2 and RH2-3, cDNA sequences used for our previous spectral analysis differ from their corresponding genomic sequences registered in GenBank at 2 aa sites (phenylalanine in the genome and tyrosine in the cDNA at residue 198 [denoted F198Y] and E332D in RH2-2 and S166A and V173F in RH2-3, Chinen et al. 2003); therefore, the cDNA sequences were used in the study. The cDNA/genome difference did not affect the reconstructed phylogenetic tree topology and estimation of ancestral sequences.
Given the tree topology in figure 2 (determined in Chinen et al. 2003), the ancestral opsin sequences at each phylogenetic node were inferred by using PAML computer program with a likelihood-based Bayesian method (http://abacus.gene.ucl.ac.uk/software/paml.html) (Yang, Kumar, and Nei 1995; Yang 1997). In the computation, the empirical substitution matrix of Jones, Taylor, and Thornton (1992) (JTT model) and that of Dayhoff, Schwartz, and Orcutt (1978) (Dayhoff model) were used as mathematical models of a substitution. The certainty of each of the inferred aa was expressed as a posterior probability.
FIG. 2.— The phylogenetic tree topology of the vertebrate RH2 opsins used for the estimation of their ancestral sequences. The max values (in nm) are indicated in parentheses for the opsin photopigments of which absorption spectra have been measured in vitro. The max values of the ancestral opsins at nodes A1 (Ancestor 1), A2 (Ancestor 2), and A3 (Ancestor 3) based on the JTT model are also indicated. The branches A, B, and C denote the branches A1-A2, A2–zebrafish RH2-1, and A3–zebrafish RH2-3, respectively, and are emphasized with thick lines. The max shifts that occurred along the branches A, B, and C are indicated.
Reconstitution of Ancestral and Mutant Pigments
Point mutations were introduced into the cDNAs of the zebrafish four RH2 opsins (Chinen et al. 2003) cloned into the pMT5 expression vector (Khorana et al. 1988) by using QuikChange site-directed mutation kit (Stratagene, La Jolla, Calif.). Various chimeric pigments were constructed by ligating the restriction fragments from the cDNA clones using restriction sites for AseI, HindIII, and PstI (fig. 3). All mutagenized cDNAs were sequenced to confirm that no spurious mutation was incorporated. Sequencing was carried out by using Thermo Sequence cycle sequencing kit (Amersham, Piscataway, N.J.) with dye-labeled primers and the LI-COR 4200L-1 automated DNA sequencer.
FIG. 3.— Amino acid sequences of the ancestral RH2 opsins estimated with the JTT model of the aa substitution and the contemporary zebrafish RH2 opsins. For phylogenetic positions of the ancestral opsins (Ancestors 1–3) see figure 2. The seven TM regions and the second extracellular loop (E2) region (Palczewski et al. 2000) are indicated with solid and dotted lines, respectively. Amino acid pairs considered for site-directed mutagenesis are boxed. Amino acid sites surrounding the 11-cis retinal are depicted in red. Position 122 is highlighted with boldface letters. Positions of AseI, HindIII, and PstI restriction sites are indicated with arrows.
The pMT5 expression vector contains the last 15 aa of the bovine rod opsin necessary for immunoaffinity purification by 1D4 monoclonal antibody (Molday and MacKenzie 1983). Each pMT5-cDNA clone was transfected into cultured COS-1 cells (RIKEN Cell Bank, Tsukuba, Japan); the resuspended cells were incubated with 5 μM 11-cis retinal (Storm Eye Institute, Medical University of South Carolina, Charleston, S.C.) and solubilized with 1% dodecyl maltoside. The resulting pigments were purified using the immobilized 1D4 (The Cell Culture Center, Minneapolis, Minn.) as in Kawamura and Yokoyama (1998). The UV-visible absorption spectrum of each pigment was recorded from 250 to 650 nm by 1-nm intervals using the Hitachi U3010 dual beam spectrometer at 20°C for a minimum of five times in the dark and for a minimum of five more times after 3 min of light exposure (with a <440-nm cutoff filter) as in Kawamura and Yokoyama (1998). The Savitzky-Golay's least squares smoothing (Gorry 1990) was carried out for each absorbance curve with 100 repetitions to eliminate spurious spikes. All max values in this study were taken from the dark spectra.
Results
Ancestral Pigments of Zebrafish RH2
We previously analyzed phylogenetic positions of the zebrafish RH2-1, RH2-2, RH2-3, and RH2-4 opsin genes among teleost RH2 opsin genes (Chinen et al. 2003), and figure 2 summarizes the tree topology among them. The ancestral opsin of the four zebrafish RH2 opsins was designated Ancestor 1 (A1), that of RH2-1 and RH2-2 opsins was designated Ancestor 2 (A2), and that of RH2-3 and RH2-4 opsins was designated Ancestor 3 (A3). Amino acid sequences of these ancestral opsins were estimated using the likelihood-based Bayesian statistics with the JTT and Dayhoff models of aa replacements on the basis of the phylogenetic tree topology given in figure 2. Figure 3 shows the aa sequences of Ancestors 1–3 inferred with the JTT model and those of RH2-1–RH2-4 (the residue numbers of the sequences coincide with the bovine rod opsin, except for 11 aa in the C-terminal). Average posterior probability over the entire aa sites was 0.98 in all the three ancestral sequences with both JTT and Dayhoff models. Amino acid sites given less than 0.9 posterior probabilities by either JTT or Dayhoff model are listed in table 1. Including these lower probability sites, estimated sequences by the two models were identical except at one site (residue 88) of Ancestor 2 where phenylalanine and isoleucine were inferred by the JTT and the Dayhoff models, respectively (table 1).
Table 1 Amino Acids Given Less Than 0.9 Posterior Probabilities by Either JTT or Dayhoff Model in Ancestral RH2 aa Sequences
The Ancestor 1 and Ancestor 3 pigments were reconstituted by introducing 13 and 9 aa changes, respectively, into the RH2-4 pigment, while the Ancestor 2 pigment was reconstituted by introducing 5 aa changes into the RH2-2 pigment according to the aa sequences inferred with the JTT model (fig. 3). Absorption spectra of the reconstituted ancestral pigments are shown in figure 4. The max values of Ancestors 1, 2, and 3 were measured to be 506, 474, and 506 nm, respectively (table 2). All pigments were shown to be photoreactive by exhibiting 380-nm max of all-trans retinal upon light exposure (dark-light difference spectra are shown in fig. 4 insets). To check whether the aa discrepancy at residue 88 of Ancestor 2 between JTT and Dayhoff models had any influence on the estimation of its ancestral absorption spectrum, the Dayhoff version of the Ancestor 2 pigment (A2_F88I) was reconstituted. The resulting max was measured to be 476 nm (table 2) and was shifted by 2 nm from the JTT-based one (474 nm). These results indicate that a major short-wave shift of the absorption spectra occurred in three evolutionary lineages, Ancestors 1 to 2 (branch A) by 30–32 nm, Ancestor 2 to RH2-1 (branch B) by 7–9 nm, and Ancestor 3 to RH2-3 (branch C) by 18 nm (fig. 2). Hereafter, JTT-based Ancestor 2 sequence was used for a template of mutagenesis analyses.
FIG. 4.— Absorption spectra of ancestral RH2 pigments of zebrafish measured in the dark. Insets: dark-light difference spectra.
Table 2 max of Zebrafish RH2 Visual Pigments and Their Reconstituted Ancestral Pigments Measured In Vitro
Effect of E122Q
From the study of the coelacanth RH2 pigment by Yokoyama et al. (1999), it is known that aa changes E122Q and L207M could cause a large spectral shift of RH2 absorption spectra toward a shorter wavelength by approximately 13 and 6 nm, respectively. In the case of zebrafish RH2 opsins, E122Q was inferred at two branches, A and C, while residue 207 was invariable (all methionine) (fig. 3). The site 122 was variable among the vertebrate RH2 sequences, either E or Q, and the posterior probabilities for E at the site in Ancestors 1 and 3 were relatively low (0.766 and 0.863, respectively; see table 1). However, the Dayhoff model also supported the results (0.780 and 0.879, respectively). To test the spectral effect of residue 122 on the zebrafish RH2 pigments, we introduced reverse mutation Q122E into the Ancestor 2 and RH2-3 pigments (resulting pigments denoted A2_Q122E and RH2-3_Q122E, respectively). The max of the A2_Q122E was 489 nm (table 3) and was 15 nm longer than that of Ancestor 2 (474 nm), and the max of the RH2-3_Q122E was 502 nm (table 4) and was 14 nm longer than that of RH2-3 (488 nm) (fig. 5). Therefore, a single aa replacement E122Q can explain nearly half (15/32) of the spectral difference between Ancestors 1 and 2 and a majority (14/18) of the spectral difference between Ancestor 3 and RH2-3.
Table 3 Spectral Effects of Branch A Mutations: Introduction of Ancestor 1 Sequences to Ancestor 2
Table 4 Spectral Effects of Branch B and Branch C Mutations
FIG. 5.— Spectral effect of the Q122E mutation to the Ancestor 2 and RH2-3 pigments. Absorbance is normalized to set the peak value as 1 for all pigments. Absorbance curves are depicted with solid and broken lines for mutant and intact pigments, respectively.
Mutations in Branch A
There are 32 aa that are different between Ancestors 1 and 2, other than the E/Q difference at the residue 122 (fig. 3). To explore aa replacements that may explain the remaining 17-nm difference between Ancestors 1 and 2, we first constructed chimeric pigments where a segment of Ancestor 2 was replaced by that of Ancestor 1. Amino acids 1–99, 100–234, and 235–349 of the Ancestor 2 pigment were replaced using restriction sites for AseI and PstI (fig. 3) (resulting chimeric pigments designated A1(99)A2, A2(99)A1(234)A2, and A2(234)A1, respectively). The max of the three pigments were 480, 495, and 478 nm, being 6, 21, and 4 nm longer than that of Ancestor 2, respectively (table 3). The sum of these effects is 31 nm, which approximates the difference (32 nm) between the max values of Ancestors 1 and 2. Because the middle segment contains residue 122, the net effect of the middle segment excluding that of E122Q was expected to be approximately 6 nm.
We then introduced point mutations into Ancestor 2 or A2_Q122E to further locate the residues responsible for spectral differentiation. Among the 32 aa that are different, 23 are located in the putative TM regions and 1 is in the second extracellular (E2) loop (fig. 3). These are the domains which contain aa positions that together constitute the retinal-binding pocket (Palczewski et al. 2000). All aa replacements involved in spectral tuning have been identified in the TM and E2 regions (Yokoyama 2002; Takahashi and Ebrey 2003), and these regions were subjected to mutagenesis (fig. 3).
In region 1–99, there are 9 aa differences in the TM domains between Ancestors 1 and 2 (see fig. 3). All of these mutations were introduced into Ancestor 2 singly or doubly when the two sites were in close proximity (table 3). Maximum max shift by these mutations was only +3 nm (by F88C). Double mutations C97S/S98A resulted in no recognizable absorbance peak. When these mutations were introduced to the A2_F88C, the max shift from A2_F88C was negligible (–1 nm). Simple addition of these single- or double-replacement effects is only +1 nm (or 0 nm if effect of C97S/S98A is counted as –1 nm) (table 3). This is at odds with the results of the segment swapping of region 1–99, which showed a +6 nm max shift. Amino acid difference (P/T at residue 27; fig. 3) was not considered for the mutagenesis, is located in the N-terminal tail, and is unlikely to have any spectral effect. These results suggest that individual spectral effects of aa replacements in this region are not additive but rather synergistically cooperate together.
In the region 100–234, there are 9 aa differences in the TM domains other than the E/Q difference at residue 122 and one difference in the E2 region between Ancestors 1 and 2 (see fig. 3). All of these mutations were introduced into A2_Q122E. Maximum max shift from A2_Q122E was +4 nm (by N151S), followed by +3 nm (by S209I) and +2 nm (by T218I) (table 3). A mutation in the E2 region, T185C, resulted in an opposite spectral shift, –4 nm, though the pigment's absorption peak was low and not clearly discernable (table 3). Simple addition of these effects and other minor ±1-nm shifts count +5 nm, which is close to the +6-nm difference between A2_Q122E and A2(99)A1(234)A2. However, when the mutations causing +2- to +4-nm shifts were introduced together (A2_Q122E/N151S/S209I/T218I), combined effect for the max shift from A2_Q122E was only 2 nm (table 3). These results suggest that individual spectral effects in this region are neither additive nor synergistic but are partly regressive. The net effect of them could vary depending on the physicochemical background of the residues making up the region.
In the region 235–349, there are 5 aa differences in the TM domains between Ancestors 1 and 2 (see fig. 3), all of which were tested. Individual mutation effects were all minor, +1 or +2 nm (table 3). Simple addition of the effects adds up to +9 nm, which is greater than the +4-nm difference between the max values of Ancestor 2 and A2(234)A1. When mutations with +2-nm effect were introduced together (V266T/F271V/A287F/A297S), the spectral shift was only +3 nm and the combined effect appeared to be regressive. This total effect is, however, close to the segment-swapping effect of 4 nm, and the regressive spectral effects can explain the spectral shift by this segment from these sites.
Mutations in Branch B
To explore aa replacements that explain the 7-nm max difference between RH2-1 and Ancestor 2, we first constructed two chimeric pigments, where aa 1–144 and 145–349 of Ancestor 2 were replaced with those of RH2-1 using HindIII restriction site (fig. 3) (the resulting chimeric pigments designated RH2-1(144)A2 and A2(144)RH2-1, respectively). Whereas RH2-1(144)A2 pigment was successfully reconstituted, functional A2(144)RH2-1 pigment with a recognizable absorption peak was not formed. However, the max value of RH2-1(144)A2 was identical to that of RH2-1 (467 nm) (table 4), suggesting that aa replacements responsible for the max difference between RH2-1 and Ancestor 2 pigments lie in the 1–144 region. We therefore introduced point mutations to the TM domains in the 1–144 region.
There are 11 aa differences in the TM domains in the region 1–144 between Ancestor 2 and RH2-1 (see fig. 3). All of these mutations were introduced into RH2-1, except at residue 88 for which Ancestor 2 was used as a template (A2_F88I) (table 2). Unexpectedly, many mutations resulted in opposite spectral shifts (K36Q, L46F, I49C, V60L, L108T, and M112I causing a short-wave shift and F88I to Ancestor 2 a long-wave shift) (table 4). S94T had no spectral effect, and V111A showed only a 1-nm long-wave shift. A double mutation L99I/V100N resulted in no recognizable peak. Simple addition of the spectral effects by these single point mutations counts as a –16-nm shift from RH2-1 (with I88F assumed to cause –2-nm shift from RH2-1), contrasting with the +7-nm shift caused by the segment swapping of the entire 1–144 region. When some mutations were introduced together (RH2-1_V111A/M112I, RH2-1_S94T/V111A/M112I and RH2-1_L46F/I49C/S94T/V111A/M112I; see table 4), the amount of spectral shift markedly differed from their sum, e.g., +4-nm shift in RH2-1_L46F/I49C/S94T/V111A/M112I, while the sum of the individual effect was –7-nm. It has been reported that residues 46 and 49 are part of the spectral tuning sites of SWS1 opsins which show individually unrecognizable shifts, but large spectral effects occur when combined together (Shi, Radlwimmer, and Yokoyama 2001). These results indicate that the spectral effect of these aa replacements can vary depending on their physicochemical environment as shown for mutations in branch A.
Mutations in Branch C
In the evolutionary branch C between Ancestor 3 and RH2-3, there are 18 aa replacements in the TM and E2 domains (see fig. 3). However, a majority of the spectral difference (14/18) was explained by a single aa replacement E122Q (fig. 5, right panel). E122Q was estimated to have occurred independently in branches A and C (fig. 3). Such a parallel substitution was also inferred at residues 50 (L50F in branches A and C), 60 (L60V in B and C), 151 (S151N in A and C), 162 (V162F in B and V162I in C), 165 (M165S in A, S165C in B, and M165L in C), 209 (I209S in A and I209C in C), 214 (V214I in B and L214F in C), 218 (I218T in A and C), 270 (T270S in A and C), and 300 (L300V in B and L300I in C) (fig. 3). Among these, 2 nm or more of spectral shift was experimentally verified with an expected direction in branch A mutations N151S, S209I, and T218I introduced to Ancestor 2 (table 3). We then tested the spectral effect of a combined mutation Q122E/N151S/C209I/T218I introduced to RH2-3. The observed spectral shift from RH2-3 was +16 nm (table 4) and was close to the actual spectral difference of Ancestor 3 from RH2-3 (+18 nm), although contributions from the other 14 aa substitutions cannot be ruled out.
Discussion
We have previously shown that zebrafish have four tandemly duplicated RH2 opsin genes in the genome and that the max of their 11-cis retinal–based photopigments differ from one another in the range of 505 nm (RH2-4) to 467 nm (RH2-1) with approximately 10 to 20-nm intervals (fig. 1) (Chinen et al. 2003). In this study, we estimated ancestral aa sequences of the four RH2 opsins using the likelihood-based Bayesian statistics with the JTT and Dayhoff models of aa replacements and reconstituted the ancestral photopigments with 11-cis retinal. We showed that the 505-nm max of RH2-4, a typical value among vertebrate RH2 photopigments with 11-cis retinal, is the ancestral max value of the four pigments, and that of the other three opsins achieved short-wave shifts which occurred after every gene duplication in only one of the two sister genes with 30-, 20-, or 10-nm magnitude (fig. 2).
Figure 6 summarizes the evolutionary changes that occurred in the phylogenetic clade derived from Ancestor 1. In branch A, 25 aa changes occurred in the TM + E2 regions, and we tested all of them for their spectral effects (table 3). The E122Q accounted for about half (15 nm) of the spectral difference (32 nm) between A1 and A2 pigments. The 32-nm total difference was nearly perfectly explained by linear addition of the spectral effects of three segmental replacements (table 3). However, within the segments, such additivity no longer held for individual aa mutations (table 3). There are 9, 11, and 5 mutations in TM + E2 regions in segments 1–99, 100–234, and 235–349, respectively (fig. 3). These effects were individually small and not additive and may have been even partly regressive. To clarify the aa replacements, other than E122Q, that explain the spectral shift, it is necessary to examine all the combinations of aa mutations in each segment. The same holds true for mutations in branch B. We verified that segment 1–144 is responsible for the spectral difference of 7 nm between A2 and RH2-1 (table 4). There are 11 mutations in the TM regions in the segment (fig. 3). We tested all of their spectral effects and showed that their individual effects were small and many of them were oriented to the opposite direction (table 4). It is necessary to explore the combinations of the mutations that account for the spectral difference between A2 and RH2-1.
FIG. 6.— The summary of evolutionary changes that occurred in the phylogenetic clade derived from the Ancestor 1 (A1) opsin. Branches associated with max shifts are depicted with thick lines. The numbers of aa substitutions in the TM + E2 regions (219-aa length) and those in the other regions (130-aa length) are indicated along each branch. Their proportions in the designated regions are indicated in parentheses. The ancestral opsin of the two goldfish opsins (GFgr-1 and GFgr-2) is designated GF and that of GF and Ancestor 3 is designated ZG. The measured (for A1, A2, and A3) or inferred (for ZG and GF) max values of the ancestral opsins are indicated at their respective nodes. The spectral ZF represents zebrafish shifts and associated aa replacements are indicated for the branches A, B, and C.
It is noted that the evolutionary branches associated with spectral shifts (i.e., branches A, B, and C and GF-GFgr-1) have a larger proportion of aa substitution in TM + E2 regions than in the other regions, whereas the other branches have similar rates of aa substitutions between the two regions (fig. 6). This is consistent with our mutagenesis results that not only a single aa substitution, E122Q with a large spectral effect, but also multiple aa substitutions with individually minor spectral effects are involved in the spectral tuning of the zebrafish RH2. In this regard, our mutagenesis conducted for branch C may not be sufficient, where 18 aa changes occurred in TM + E2 regions (fig. 3) but only four of them including E122Q were tested (table 4), although E122Q explains 14-nm out of 18-nm difference between A3 and RH2-3.
Figure 6 also indicates that spectral differentiation of zebrafish RH2 opsins occurred independently of that of goldfish RH2 opsins. The ancestral opsin of the two goldfish opsins (GFgr-1 and GFgr-2; Johnson et al. 1993) was designated GF and that of GF and Ancestor 3 was designated ZG. The ZG was inferred to have an identical aa sequence with Ancestor 3 and must have a max of 506 nm as well. The max of GF would also be 506 nm because those of ZG and GFgr-2 are both 506 nm and because the 6 aa changes in the TM + E2 regions in the ZG-GF branch are all conservative and outside the retinal-surrounding sites (sequences not shown).
The recent X-ray diffraction analysis of bovine rod opsin revealed its three-dimensional crystal structure at 2.8-? resolution (Palczewski et al. 2000). Residue 122 is one of the 27 aa forming the retinal-binding pocket which are located within 4.5 ? from 11-cis retinal (residues 43, 44, 47, 94, 113, 117, 118, 120, 121, 122, 178, 181, 186, 187, 188, 189, 191, 207, 211, 212, 261, 265, 268, 269, 293, 295, and 296) (Palczewski et al. 2000). In addition, a total of 38 aa sites, including the 27 sites, were identified as surrounding 11-cis retinal (Menon, Han, and Sakmar 2001; Takahashi and Ebrey 2003) (the sites indicated in red in fig. 3). Residue 122 interacts directly with ?-ionone ring of 11-cis retinal, and its replacement has been verified to result in significant spectral shifts in various visual pigments (Takahashi and Ebrey 2003). In general, there are two factors controlling the opsin shift: (1) electrostatic interaction between the glutamic acid counterion at residue 113 and the protonated Schiff base of retinal at lysine residue 296 and (2) electrostatic interaction between the retinal body and charged or polar groups of aa surrounding it (Takahashi and Ebrey 2003). In both cases, aa replacements surrounding retinal can directly influence absorption spectra of visual pigments, and hence their spectral effects tend to be additive as manifested in the cases of M/LWS pigments (Yokoyama and Radlwimmer 2001) and coelacanth RH2 and RH1 pigments (Yokoyama et al. 1999). The 5 aa sites 164, 181, 261, 269, and 292, which are known to exert major contributions to tuning of the M/LWS opsins, the two sites 122 and 207 in the coelacanth RH2 opsin, and the two sites 122 and 292 in the coelacanth RH1 opsin are all in the retinal-surrounding sites (see fig. 3).
On the other hand, spectral effects of aa replacements outside the retinal-surrounding sites would be generally indirect, small, nonadditive, and dependent on the molecular environment of the aa sites concerned. These replacements possibly work through alteration of conformation or hydrophobicity of TM domains. The spectral tuning mechanism of the violet/uv (SWS1) pigments appears to be a typical example of such an indirect interaction mechanism, where combinations of 10 aa sites (46, 49, 52, 86, 90, 93, 97, 114, 116, and 118) are involved in the spectral differentiation and seven of them (sites other than 90, 114, and 118) are located outside the retinal-surrounding sites (Wilkie et al. 2000; Yokoyama, Radlwimmer, and Blow 2000; Shi, Radlwimmer, and Yokoyama 2001; Fasick, Applebury, and Oprian 2002; Shi and Yokoyama 2003; Parry et al. 2004). The tuning sites possibly affect trafficking of water molecules at the retinal-binding pocket by altering net hydrophobicity of TM domains with other residues and control protonation and deprotonation status of the Schiff base (Shi, Radlwimmer, and Yokoyama 2001). Whereas spectral effects of the aa replacements in the retinal-surrounding region are likely to be universal beyond differences of opsin groups or animal species because of their directness of the interaction with retinal, spectral effects of the aa replacements outside the region found in a system may be of little general implication to another system because of their indirectness of the interaction with retinal.
All the mutations, other than E122Q tested in this study, fall outside the retinal-surrounding region (fig. 3) (site 94 is in the region but see the following explanations). These mutations resulted in only minor spectral effects with 0 to 4-nm shift (tables 3 and 4). S94A is known to cause a 14-nm blueshift in newt SWS2 pigment (Takahashi and Ebrey 2003), but the substitution in our case was between the same polar residues with a hydroxyl group, S94T, and resulted in no recognizable spectral shift (table 4). Characteristically and consistently with their locations relative to retinal, combined effects of these minor sites were not additive and were, in some combinations, even opposite in direction of spectral shift from the shift direction in individual mutations. However, this nonlinear nature of cumulative effects from multiple aa sites with individual minor effects does play an important role in the spectral differentiation of the zebrafish RH2 opsins.
In zebrafish RH2 opsins, spectral diversification is achieved by direct and indirect interaction mechanisms with retinal, the former by residue 122 in the retinal-surrounding region with a large spectral effect and the latter by multiple residues outside it with individually minor and collectively nonadditive effects. This pattern of spectral differentiation could represent a general paradigm to account for the spectral diversity of vertebrate RH2 pigments, considering that a range of 470–510 nm is covered by the four zebrafish pigments.
Acknowledgements
This study was supported by Grants-in-Aid for Scientific Research (B) (12440243) and Exploratory Research (13874105) from the Japan Society for the Promotion of Science (JSPS) to S.K. and a Grant-in-Aid for JSPS Fellows (14-08073) to A.C. The manuscript was proofread by BioMed Proofreading Service.
References
Carleton, K. L., and T. D. Kocher. 2001. Cone opsin genes of African cichlid fishes: tuning spectral sensitivity by differential gene expression. Mol. Biol. Evol. 18:1540–1550.
Chinen, A., T. Hamaoka, Y. Yamada, and S. Kawamura. 2003. Gene duplication and spectral diversification of cone visual pigments of zebrafish. Genetics 163:663–675.
Dayhoff, M. O., R. M. Schwartz, and B. C. Orcutt. 1978. A model of evolutionary change in proteins. Pp. 345–352 in M. O. Dayhoff, ed. Atlas of protein sequence and structure. National Biomedical Research Foundation, Silver Spring, Md.
Ebrey, T., and Y. Koutalos. 2001. Vertebrate photoreceptors. Prog. Retin. Eye Res. 20:49–94.
Fasick, J. I., M. L. Applebury, and D. D. Oprian. 2002. Spectral tuning in the mammalian short-wavelength sensitive cone pigments. Biochemistry 41:6860–6865.
Gorry, P. A. 1990. General least-squares smoothing and differentiation by the convolution (Savitzky-Golay) method. Anal. Chem. 62:570–573.
Johnson, R. L., K. B. Grant, T. C. Zankel, M. F. Boehm, S. L. Merbs, J. Nathans, and K. Nakanishi. 1993. Cloning and expression of goldfish opsin sequences. Biochemistry 32:208–214.
Jones, D. T., W. R. Taylor, and J. M. Thornton. 1992. The rapid generation of mutation data matrices from protein sequences. Comput. Appl. Biosci. 8:275–282.
Kawamura, S., and S. Yokoyama. 1998. Functional characterization of visual and nonvisual pigments of American chameleon (Anolis carolinensis). Vision Res. 38:37–44.
Khorana, H. G., B. E. Knox, E. Nasi, R. Swanson, and D. A. Thompson. 1988. Expression of a bovine rhodopsin gene in Xenopus oocytes: demonstration of light-dependent ionic currents. Proc. Natl. Acad. Sci. USA 85:7917–7921.
Kito, Y., T. Suzuki, M. Azuma, and Y. Sekoguti. 1968. Absorption spectrum of rhodopsin denatured with acid. Nature 218:955–957.
Menon, S. T., M. Han, and T. P. Sakmar. 2001. Rhodopsin: structural basis of molecular physiology. Physiol. Rev. 81:1659–1688.
Molday, R. S., and D. MacKenzie. 1983. Monoclonal antibodies to rhodopsin: characterization, cross-reactivity, and application as structural probes. Biochemistry 22:653–660.
Palczewski, K., T. Kumasaka, T. Hori et al. (12 co-authors). 2000. Crystal structure of rhodopsin: a G protein-coupled receptor. Science 289:739–745.
Parry, J. W., S. Poopalasundaram, J. K. Bowmaker, and D. M. Hunt. 2004. A novel amino acid substitution is responsible for spectral tuning in a rodent violet-sensitive visual pigment. Biochemistry 43:8014–8020.
Shi, Y., F. B. Radlwimmer, and S. Yokoyama. 2001. Molecular genetics and the evolution of ultraviolet vision in vertebrates. Proc. Natl. Acad. Sci. USA 98:11731–11736.
Shi, Y., and S. Yokoyama. 2003. Molecular analysis of the evolutionary significance of ultraviolet vision in vertebrates. Proc. Natl. Acad. Sci. USA 100:8308–8313.
Takahashi, Y., and T. G. Ebrey. 2003. Molecular basis of spectral tuning in the newt short wavelength sensitive visual pigment. Biochemistry 42:6025–6034.
Wilkie, S. E., P. R. Robinson, T. W. Cronin, S. Poopalasundaram, J. K. Bowmaker, and D. M. Hunt. 2000. Spectral tuning of avian violet- and ultraviolet-sensitive visual pigments. Biochemistry 39:7895–7901.
Yang, Z. 1997. PAML: a program package for phylogenetic analysis by maximum likelihood. Comput. Appl. Biosci. 13:555–556.
Yang, Z., S. Kumar, and M. Nei. 1995. A new method of inference of ancestral nucleotide and amino acid sequences. Genetics 141:1641–1650.
Yokoyama, S. 2000. Molecular evolution of vertebrate visual pigments. Prog. Retin. Eye Res. 19:385–419.
———. 2002. Molecular evolution of color vision in vertebrates. Gene 300:69–78.
Yokoyama, S., and F. B. Radlwimmer. 1998. The "five-sites" rule and the evolution of red and green color vision in mammals. Mol. Biol. Evol. 15:560–567.
———. 1999. The molecular genetics of red and green color vision in mammals. Genetics 153:919–932.
———. 2001. The molecular genetics and evolution of red and green color vision in vertebrates. Genetics 158:1697–1710.
Yokoyama, S., F. B. Radlwimmer, and N. S. Blow. 2000. Ultraviolet pigments in birds evolved from violet pigments by a single amino acid change. Proc. Natl. Acad. Sci. USA 97:7366–7371.
Yokoyama, S., H. Zhang, F. B. Radlwimmer, and N. S. Blow. 1999. Adaptive evolution of color vision of the Comoran coelacanth (Latimeria chalumnae). Proc. Natl. Acad. Sci. USA 96:6279–6284.(Akito Chinen, Yoshifumi M)