To better understand the evolution of red-green color vision in vertebrates, we inferred the amino acid sequences of the ancestral pigments of 11 selected visual pigments: the LWS pigments of cave fish (Astyanax fasciatus), frog (Xenopus laevis), chicken (Gallus gallus), chameleon (Anolis carolinensis), goat (Capra hircus), and human (Homo sapiens); and the MWS pigments of cave fish, gecko (Gekko gekko), mouse (Mus musculus), squirrel (Sciurus carolinensis), and human. We constructed these ancestral pigments by introducing the necessary mutations into contemporary pigments and evaluated their absorption spectra using an in vitro assay. The results show that the common ancestor of vertebrates and most other ancestors had LWS pigments. Multiple regression analyses of ancestral and contemporary MWS and LWS pigments show that single mutations S180A, H197Y, Y277F, T285A, A308S, and double mutations S180A/H197Y shift the λmax of the pigments by −7, −28, −8, −15, −27, and 11 nm, respectively. It is most likely that this “five-sites” rule is the molecular basis of spectral tuning in the MWS and LWS pigments during vertebrate evolution.
HUMAN color vision is achieved through three types of photosensitive molecules: short wavelength- (or blue-) sensitive (SWS), middle wavelength- (or green-) sensitive (MWS), and long wavelength- (or red-) sensitive (LWS) visual pigments, which absorb light maximally (λmax) at ~420, ~530, and ~560 nm, respectively (Boynton 1979). However, having only SWS pigments and either MWS or LWS pigments, most mammals have dichromatic color vision. This condition is commonly known as “red-green color blindness.” Even in human, the “red-green color blindness” is relatively common, affecting ~8% of males (Kalmus 1965). When diverse species are surveyed, the λmax of most LWS pigments ranges from 550 to 560 nm, while the MWS pigments detect a wider range of light with λmax at ~510–540 nm (Yokoyama 2000a). All of these pigments belong to a specific evolutionary group, often referred to as the LWS/MWS group (Yokoyama and Yokoyama 1996; Yokoyama 1997, 1999, 2000a; Ebrey and Koutalos 2001).
Each visual pigment consists of a chromophore, 11-cis-retinal, and a transmembrane protein, opsin, which is encoded by a specific opsin gene. The SWS, MWS, and LWS opsin genes were cloned for the first time from human (Nathanset al. 1986). Since then, the mechanisms of spectral tuning in the MWS and LWS pigments of vertebrates have been studied extensively. It was shown that the spectral sensitivities of the MWS and LWS pigments in human and fish are determined mainly by A180/F277/A285 (amino acids A, F, and A at sites 180, 277, and 285, following the amino acid site numbers in the human LWS/MWS pigments, respectively) and S180/Y277/T285, respectively (Yokoyama and Yokoyama 1990; Neitzet al. 1991; Chanet al. 1992; Merbs and Nathans 1993; Asenjoet al. 1994). The spectral sensitivities of some MWS pigments from such species as mouse (Mus musculus), rat (Rattus norvegicus), and rabbit (Oryctolagus cuniculus), however, cannot be explained by this “three-sites” rule and require amino acid changes at two entirely new sites: 197 and 308 (Sunet al. 1997; Radlwimmer and Yokoyama 1998; Yokoyama and Radlwimmer 1998, 1999). It was suggested that S180A (amino acid replacement S → A at site the 180), H197Y, Y277F, T285A, and A308S shift the λmax of the LWS/MWS pigments toward green by 7, 28, 7, 15, and 16 nm, respectively, and the reverse changes toward red by the same amounts (Yokoyama and Radlwimmer 1999). This five-sites rule explains the λmax of virtually all LWS/MWS pigments of vertebrates. The MWS pigment of bottlenose dolphin (Tursiops truncatus) is an exception, where the observed λmax is 13 nm lower than the expected value from this five-sites rule (Yokoyama and Radlwimmer 1999). Thus, the details of the mechanisms of spectral tuning in the MWS and LWS pigments are still not fully understood.
The evolution of red-green color vision was also studied by inferring the amino acid sequences at the five critical sites of the ancestral pigments. Comparing the MWS and LWS pigments from human and fish (Astyanax fasciatus), it was suggested that the LWS pigments in these species evolved from the ancestral MWS pigment independently (Yokoyama and Yokoyama 1990). The evolution of the LWS pigment from the MWS pigment was also suggested by Jacobs (1993) and Winderickx et al. (1992), but the primate ancestor appears to have had the LWS pigment (Neiet al. 1997). It is also speculated that the mammalian ancestor had the MWS pigment (Yokoyama and Radlwimmer 1999). In these analyses, the absorption spectra of ancestral pigments were not evaluated. Without having the actual absorption spectra of the ancestral pigments, the evolutionary results of red-green color vision remain speculative. However, once the amino acid sequences and λmax of such ancestral visual pigments are determined, we can study not only the evolutionary processes of LWS/MWS pigments directly but also the molecular bases of redgreen color vision through time.
Here, we infer the amino acid sequences of ancestral pigments from contemporary sequences, construct the ancestral pigments, evaluate their λmax, and test the validity of the five-sites rule during vertebrate evolution. To accomplish the first three goals, we consider 11 pigments: the LWS pigments of human, goat (Capra hircus), American chameleon (Anolis carolinensis), chicken (Gallus gallus), clawed frog (Xenopus laevis), and cave fish (A. fasciatus) and the MWS pigments of human, squirrel (Sciurus carolinensis), mouse (M. musculus), gecko (Gekko gekko), and cave fish (A. fasciatus). The results suggest that virtually all ancestral pigments had amino acids SHYTA (amino acids S, H, Y, T, and A at sites 180, 197, 277, 285, and 308, respectively) and were red sensitive. Statistical and mutagenesis analyses show that the five-sites rule, now including the synergistic effect of amino acid sites 180 and 197, fully explains the variation in the spectral sensitivity of all ancestral and currently known LWS/MWS pigments in vertebrates, including the dolphin MWS pigment.
MATERIALS AND METHODS
cDNA cloning and DNA sequencing: Total retinal RNAs of chicken (Gallus gallus) and frog (Xenopus laevis) are gifts from Dr. Mary Pierce at the SUNY Upstate Medical University at Syracuse, and that of cave fish (A. fasciatus) was isolated using the procedure of Yokoyama et al. (1995). Figure 1 shows reverse transcription (RT)-PCR primers used to amplify the four types of full-length opsin cDNAs.
For each set of primers, cDNA was reverse transcribed at 42° for 1 hr and at 95° for 5 min, and then PCR was carried out for 30 cycles at 94° for 45 sec, 55° for 1.5 min, and 72° for 2 min. PCR products were gel isolated and subcloned into the T-tailed EcoRV-digested Bluescript plasmid vector with T-over-hang attached to 3′ ends (Hadjeb and Berkowitz 1996). Nucleotide sequences of the entire region of the cDNA clones were determined by cycle sequencing reactions using the Sequitherm Excell II long-read kits (Epicentre Technologies, Madison, WI) with dye-labeled M13 forward and reverse primers. Reactions were run on a LI-COR (Lincoln, NE) 4200LD automated DNA sequencer.
Expression and spectral analyses of pigments: The PCR-amplified opsin cDNAs were subcloned into the EcoRI and SalI restriction sites of the expression vector pMT5 (Khoranaet al. 1988). These plasmids were expressed in COS1 cells by transient transfection. The pigments were regenerated by incubating the opsins with 11-cis-retinal (Storm Eye Institute, Medical University of South Carolina) and purified using immobilized 1D4 (The Culture Center, Minneapolis, MN) in buffer W1 [50 mm N-(2-hydroxyethyl) piperazine-N′-2-ethane-sulfonic acid (HEPES; pH 6.6), 140 mm NaCl, 3 mm MgCl2, 20% (w/v) glycerol, and 0.1% dodecyl maltoside], as previously described (Yokoyama 2000b).
Mutants were generated by using the QuickChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). All DNA fragments that were subjected to mutagenesis were sequenced to rule out spurious mutations. UV-visible spectra were recorded at 20° using a Hitachi U-3000 dual beam spectrophotometer. Visual pigments were bleached for 3 min using a 60-W standard light bulb equipped with a Kodak Wratten #3 filter at a distance of 20 cm. Data were analyzed using Sigmaplot software (Jandel Scientific, San Rafael, CA).
Sequence data analyses: The sources of the amino acid sequences of the MWS and LWS pigments of different vertebrate species are given in Table 1, where both of their amino acid sequences and λmax are known. So far, with the exception of the chicken, frog, and cave fish pigments, their λmax's have been evaluated using in vitro assay.
We selected two fish, two frog, two reptile, and one bird LWS/MWS pigments (Table 1). The numbers of mammalian pigments with λmax at ~510, ~530, and ~560 nm are 4, 7, and 6, respectively (Table 1), from which we selected 1, 2, and 2 pigments, respectively. Thus, with respect to the phylogenetic and spectral diversities, these 11 pigments represent every segment of currently known LWS/MWS pigments in vertebrates. The phylogenetic relationship of fish, frog, reptiles, and mammals is well established (e.g., see Kumar and Hedges 1998), but that of human, goat, and rodents is unclear (Novacek 1992; De Jong 1998; Shoshani and McKenna 1998; Huchonet al. 1999; Liu and Miyamoto 1999; Waddell et al. 1999a,b; Madsenet al. 2001; Murphyet al. 2001). Thus, they are assumed to be equally distantly related. Accordingly, we consider the phylogenetic relationship [((((human (P560), human (P530)), goat (P553), (squirrel (P532), mouse (P508))), ((chameleon (P560), gecko (P527)), chicken (P561))), frog (P557)), (cave fish (P558), cave fish (P530))]. On the basis of this tree topology, we infer the ancestral sequences of the opsins by using a computer program, PAML, based on a likelihood-based Bayesian method (Yanget al. 1995; Yang 1997). In the inference, paralogous bovine RH1 (GenBank accession no. U49742), goldfish RH2 (L11865), chameleon SWS1 (AF134194), and chicken SWS2 (M92037) pigments were used as the outgroup (for the description of these classes of pigments, see Yokoyama and Yokoyama 1996; Yokoyama 1997, 1999, 2000a; Ebrey and Koutalos 2001).
To study the spectral sensitivities of ancestral mammalian pigments further, we consider yet another phylogenetic tree of LWS/MWS pigments: [(((human (P560), human (P530)), (marmoset (P562), marmoset (P553), marmoset (P540))), rabbit (P509), (goat (P553), deer (P531))), guinea pig (P516), (squirrel (P532), mouse (P508))]. This tree topology of the mammalian pigments is basically the same as that in Yokoyama and Radlwimmer (1999). However, it contains three trifercations because of unresolved phylogenetic relationships among the mammalian species.
Absorption spectra of the chicken, frog, and cave fish LWS/MWS pigments: To study the mechanisms of spectral tuning of visual pigments, we applied the in vitro assay (Yokoyama 2000a) to the LWS/MWS pigments of chicken, frog, and cave fish. The regenerated pigments show two absorption peaks, one at ~280 nm and another at 530–561 nm (Figure 2). When these pigments are exposed to light, the second peak shifts to ~380 nm (results not shown), indicating the cis-trans isomerization of the chromophore (Hubbard and Kropf 1958). This control experiment demonstrates that the lower peaks are entirely due to opsins covalently linked to 11-cis-retinal via a Schiff base bond. When measured in the dark, the LWS pigments of cave fish, frog, and chicken have λmax at 558 ± 2, 557 ± 2, and 561 ± 2 nm, respectively, while the MWS pigment of cave fish has a λmax at 530 ± 2 nm (Figure 2, Table 1). The respective dark-light difference spectra are given by 557, 557, 561, and 531 nm, all of which are also precise to within ±2 nm (Figure 2, insets) and are very close to the corresponding dark spectra.
Previously, using a 3-[(3-cholamidopropyl) dimethylammonio]-1-propanesulfonate-phosphatidylcholine (CHAPS-PC) mixture and microspectrophotometry (MSP) methods, the λmax of the chicken LWS pigment were estimated to be 571 (Okanoet al. 1989) and 569 nm (Bowmaker and Knowles 1977), respectively. Our estimate is ~10 nm lower than these values. The cause of this difference is not immediately clear. However, in the CHAPS-PC method, the sample used for the evaluation of the λmax is not pure and contains ~92% of the LWS pigments, while the MSP analyses are often subject to sampling errors. Compared with these methods, the visual pigments regenerated using an in vitro assay are identical and are expected to provide a more reliable λmax of each specific pigment.
The two previous estimates of the λmax of frog (P557) pigment are very different: 611 nm by MSP (Witkovskyet al. 1981) and 533 nm by the light-evoked cone contraction method (Besharse and Witkovsky 1992). At present, the cause of the two very different estimates is not clear. Our estimate of 557 nm falls in between these two λmax's. MSP estimates for the cave fish MWS and LWS pigments regenerated with 11-cis-retinal showed λmax's at 533 ± 3 and 563 ± 4 nm, respectively (F. I. Harosi and J. Kleinschmidt, personal communication; see also Kleinschmidt and Harosi 1992). The present results, 530 and 558 nm, respectively, agree well with the MSP results. In the following, we use the dark spectra obtained from the in vitro assay for the λmax of the chicken, frog, and cave fish pigments (Table 1).
Amino acid sequences of the ancestral LWS/MWS pigments: Given the tree topology of the vertebrate pigments in Figure 3, the amino acid sequences of the ancestral pigments at nodes a–i (pigments a–i, respectively) were inferred by using the JTT model of amino acid replacements (Joneset al. 1992; Figure 4). As we can see in Figure 4, the accuracy of most inferred amino acids is quite high and only relatively small numbers of them have posterior probabilities of <0.9. Note that the inferred amino acid sequences of pigments g and h are identical. When the Dayhoff model of amino acid replacements (Dayhoffet al. 1978) is used, very similar ancestral amino acid sequences are inferred. For pigments a–f, g (= h), and i, the two models predict only three, one, five, one, zero, one, three, and two amino acid differences, respectively. At these ambiguous sites, the posterior probabilities associated with the amino acid inference are always <0.9.
Using the contemporary pigments, we suggested that the spectral sensitivities of the LWS/MWS pigments are controlled mainly by the amino acids at 180, 197, 277, 285, and 308 (Yokoyama and Radlwimmer 1998, 1999). At these five critical sites, pigments a–e, g (= h), and i have SHYTA, while pigment f has SYYTA (Figure 3). Most of these inferred amino acids have posterior probabilities >0.9. For pigments d–f, however, amino acids S180 and A180 are predicted with probabilities of 0.7 and 0.3, respectively.
Absorption spectra of the ancestral pigments: Many inferred amino acids of the ancestral pigments are identical to those of the contemporary pigments (Figure 4), showing that pigments a–i can be constructed by introducing necessary mutations into some extant pigments. Indeed, pigments d–f were constructed by introducing six mutations into human (P560) pigment, nine mutations into pigment d, and nine mutations into squirrel (P532) pigment, respectively (Figure 5A). The amino acid sequences of pigments a–c, g (= h), and i are similar to that of chameleon (P560) pigment, which can be divided into segments I–IV by three restriction sites (Figures 4 and 5B). Thus, by introducing necessary mutations into these segments, we constructed I1–I4, II1–II4, III1–III4, and IV1–IV2, respectively (Figure 5C). The pigments a–c, g (= h), and i were then obtained by recombining these four segments at the three restriction sites (Figure 5C). In this way, we constructed one ancestral pigment at each node, each amino acid having the highest posterior probability.
When measured in the dark, the λmax of pigments a–f, g (= h), and i are 563 ± 2, 563 ± 2, 561 ± 2, 558 ± 2, 558 ± 2, 536 ± 1, 561 ± 2, and 564 ± 2 nm, respectively (Figure 6). The respective dark-light difference spectra are given by 564, 563, 561, 558, 558, 536, 561, and 564 nm, all of which are precise to within ±2 nm (Figure 6, insets). They are virtually identical to the corresponding dark spectra. Thus, pigment f with SYYTA at the five critical sites has a λmax at 536 nm, while the other ancestral pigments with SHYTA have λmax at 558–564 nm (Figures 3 and 6). These results suggest that the pigment of the common ancestor was red sensitive and the contemporary LWS/MWS pigments evolved from the LWS pigment. As noted earlier, for pigments d–f, S180 and A180 are predicted with probabilities 0.7 and 0.3, respectively. As we see below, changes S180A and A180S should shift the λmax for no more than 7 nm and, therefore, the uncertain inference of amino acids at site 180 does not change the red and green sensitivities of the ancestral pigments.
Now, what do the spectral sensitivity data in Figure 3 tell us about the mechanisms of spectral tuning in the LWS and MWS pigments? Following Yokoyama and Radlwimmer (1999), let θ180, θ197, θ277, θ285, θ308, and Z be the magnitudes of the λmax shifts caused by S180A, H197Y, Y277F, T285A, A308S, and the absorption spectrum of a pigment with SHYTA. Note that the amino acid compositions for all contemporary and ancestral pigments are either Y277/T285 (Y277 and T285) or F277/A285 (Figure 4) and, therefore, two parameters, θ277 and θ285, cannot be estimated separately. When multiple linear regression analysis is applied to the λmax of the 11 contemporary and 9 ancestral pigments, , , , , and Z = 560 ± 1 nm. When the regression analysis is applied to only the 11 contemporary pigments, , , , , and Z = 559 ± 1 nm. Thus, the two sets of estimates are practically identical. Using the former set of estimates, we can predict the expected λmax for the contemporary and ancestral pigments (Table 2). Table 2 shows that the observed λmax of frog (P557) and cave fish (P558) pigments and ancestral pigments a, b, d, e, and i fall outside of the 95% confidence interval. However, considering the standard errors associated with the estimation of the λmax, the expected and corresponding observed λmax's are in reasonably good agreement.
Table 2 clearly shows that the spectral tuning of the LWS/MWS pigments is determined mainly by the amino acids at 180, 197, 277, 285, and 308. It is also true that, despite having the same amino acid composition SHYTA at the five critical sites, pigments a–e, g (= h), and i have somewhat different λmax, ranging from 558 to 564 nm (Figures 3 and 6). Among these, the λmax of pigments d and e (both with λmax = 558 nm) and that of pigment i (λmax = 564 nm) are significantly different (P < 0.05). Where does this variation in the λmax values come from? As noted earlier, the spectral sensitivities of human (P530) and human (P560) pigments are determined mainly by the three-sites rule. However, additional amino acid differences at sites 116, 230, 233, and 309 are known to have some minor influence on the differentiation of the two pigments (Asenjoet al. 1994). Pigments d, e, and i have two identical amino acids I230/Y309, but pigments d and e have two amino acid replacements S116Y/A233S (Figure 4). Both amino acids Y116 and S233 are human (P530) pigment specific. Therefore, the difference in the λmax between pigments d and e and pigment i is caused most probably by the slightly decreased λmax in the former two pigments due to S116Y/A233S. It should be stressed, however, that the minor variation in λmax is found only in certain mammalian pigments. Thus, this minor adjustment in the mechanisms of spectral tuning seems to be a local phenomenon.
Five-sites rule: Using the contemporary and ancestral pigments in Figure 3, we could not infer θ277 and θ285 separately. However, if we consider the amino acid sequences of all 26 pigments in Table 1, we can evaluate θ277 and θ285 separately and also determine interactions between sites 180 and 197 (θ180×197); 180 and 285 (θ180×285); 197 and 285 (θ197×285); 197 and 308 (θ197×308); 277 and 285 (θ277×285); and 180, 277, and 285 (θ180×277×285; appendix).
When the interaction terms are neglected, Z = 559 nm, , , , , and (Table 3, model a). When the λmax for the 26 pigments are estimated by using these θ values, they agree reasonably well with the corresponding observed values (Table 4, column a). However, the λmax of dolphin (P524) and squirrel (P532) pigments deviate from the corresponding expected values by 5 and 4 nm, respectively (see also Yokoyama and Radlwimmer 1999). The 95% confidence intervals for the λmax values of dolphin (P524) and squirrel (P532) pigments are 526–532 nm and 533–539 nm, respectively. The corresponding observed λmax fall outside of these intervals. Thus, some improvement in the estimation procedure is desirable. It turns out that such an improvement can be made by considering interactions among the five critical sites. The improvement in the estimation procedures with and without a specific interaction term can be tested by where SSE1 and SSE2 indicate the SSE values (see appendix) for the models with and without a specific interaction term, respectively. When interaction terms θ180×197, θ180×285, θ197×285, θ197×308, θ277×285, and θ180×277×285 are considered separately, the F20, 19 values are 3.4 (P < 0.001), 1.2 (P = 0.25–0.5), 1.2 (P = 0.25–0.5), 1.6 (P = 0.1–0.25), 1.0 (P ≅ 0.5), and 1.0 (P ≅ 0.5), respectively. These tests show that only the estimation procedure with θ180×197 should significantly improve the estimates.
If we include an interaction term θ180×197, then Z = 559 nm, , , , , , and (Table 3, model b). Thus, Z, , , , and are similar to those of model a, but decreases and, at the same time, causes a significant red shift in the λmax. When the expected λmax of the 26 pigments are inferred using these estimates, the differences between the expected and observed λmax are significantly reduced (Table 4, column b). The 95% confidence intervals for the λmax of dolphin (P524) and squirrel (P532) pigments are now 523–528 nm and 529–534 nm, respectively, and contain the observed λmax.
These analyses show that the five-sites rule, now including the synergistic effect of sites 180 and 197, explains the λmax of all contemporary pigments. According to the five-sites rule, the ancestral pigments with SHYTA and SYYTA should have λmax at 559 and 536 nm, respectively. The corresponding observed values are 558–564 and 536 nm (Figure 3). Thus, this rule explains the spectral tunings of the LWS/MWS pigments during the entire history of vertebrate evolution. Sites 180, 277, 285, and 308 are located near the chromophore (Palczewskiet al. 2000), where interaction between the chromophore and an opsin usually takes place (Yokoyama 2000a,b; Ebrey and Koutalos 2001). The site 197 is located outside of the transmembrane regions, but H197 is known for its important function of chloride binding (Sunet al. 1997). Although we cannot offer a structural explanation for the λmax shift, it suggests an intimate relationship between the chloride binding site at 197 and the spectral tuning residue at 180.
Ancestral mammalian pigments: By applying the JTT model of amino acid replacements to another tree topology of the mammalian pigments (Figure 7), we also inferred the amino acid sequences of the common ancestors of the 11 mammalian pigments (pigment x), of human (P560) and human (P530) pigments (pigment y), and of squirrel (P532) and mouse (P508) pigments (pigment z). Pigments x–z roughly correspond to pigments d–f in Figure 3, respectively. Most amino acids inferred have posterior probabilities of >0.9 and pigments x–z have only four, two, and five sites, where the posterior probabilities are <0.9 (Figure 4).
Using site-directed mutagenesis and recombinant DNA techniques, we then reconstructed these three ancestral pigments. Again, using an in vitro assay, the λmax of pigments x–z are determined to be 533 ± 1, 553 ± 1, and 533 ± 2 nm, respectively (Figures 7 and 8). These results are consistent with those of Yokoyama and Radlwimmer (1999). However, we have also seen that the λmax of pigments d–f are given by 558, 558, and 536 nm, respectively (see Figure 3). Thus, the λmax of pigment x is 25 nm lower than that of pigment d.
Pigments x–z have AYYTA, AHYTA, and AYYTA at the five critical sites, respectively, whereas pigments d–f have SHYTA, SHYTA, and SYYTA, respectively. Thus, the difference between the λmax of pigments x and d must have been caused by the amino acid differences at sites 180 and 197. Slightly lower λmax of pigments y and z than those of pigments e and f can be explained by the amino acid differences at site 180. These results clearly show that the inference of the ancestral pigments depends not only on the phylogenetic relationship of extant pigments but also on the types of pigments used for constructing the phylogenetic tree. As noted earlier, however, the phylogenetic relationship of mammalian species has not been resolved. Thus, until the issue is resolved, it is premature to discuss whether the mammalian ancestor had MWS or LWS pigment.
Compared with the variability in the spectral sensitivities of those of the mammalian pigments, the λmax of LWS/MWS pigments of nonmammalian vertebrates are more uniform. The currently available data show that all fish, amphibian, reptile, and avian lineages have LWS pigments but not necessarily MWS pigments (Table 1). Similarly, the orthologous pigment in marine lamprey (Lamptera marinus) has P180, Y277, and T285, showing the LWS pigment-specific feature (H. Zhang and S. Yokoyama, unpublished data). It seems that the divergence between chameleon (P560) and gecko (P527) pigments (Figure 3, node h) and between cave fish (P558) and cave fish (P530) pigments (node i) occurred ~150–190 million years (MY) ago (Yokoyama and Blow 2001) and ~190–320 MY ago (Yokoyama and Yokoyama 1990), respectively. The divergence between the two human pigments is estimated to be ~30 MY ago (Yokoyama and Yokoyama 1989). Thus, the divergences of the LWS and MWS pigments seem to be relatively recent events in vertebrate evolution. These observations are consistent with our suggestion that the vertebrate ancestor had the LWS pigment rather than the MWS pigment.
It is highly likely that the vertebrate ancestor possessed the LWS pigment with SHYTA at the five critical amino acid sites and that the contemporary LWS/MWS pigments evolved from the ancestral LWS pigment. We have also seen that the five-sites rule, now including the synergistic effect between amino acid sites 180 and 197, fully explains the variation in the λmax of the LWS/MWS pigments during the entire history of vertebrate evolution. The first conclusion was based on a single amino acid sequence at each ancestral node, while the second conclusion was derived using a purely statistical argument. Here, we address these points in more detail.
Uncertain inference of amino acids and spectral sensitivity: As noted earlier, ancestral mammalian pigments x–z have 4, 2, and 5 amino acid sites where the posterior probabilities are <0.9. We replaced these amino acids by those with the second highest posterior probabilities and determined the λmax of the mutant pigments. The results clearly show that the λmax of the regenerated mutant pigments are virtually identical to those without such mutations, showing that the amino acid differences at these 11 sites have little effect on the λmax shift (Table 5).
Interestingly, amino acid changes A180S introduced into pigments x and z do not shift the λmax either. Considering of −7 nm (Table 3, model b), this observation may be surprising. These results need to be understood, however, by considering the interaction between sites 180 and 197. Note that squirrel (P532) and mole rat (P534) pigments have AYYTA and SYYTA at the five critical sites, respectively, but their λmax values are 532 and 534 nm and are virtually identical (Table 4). On the other hand, human (P552), marmoset (P553), goat (P553), and cat (P553) pigments have AHYTA, whereas human (P560), marmoset (P561), chicken (P561), pigeon (P560), zebra finch (P560), chameleon (P560), frog (P557), cave fish (P558), and goldfish (P559) pigments all have SHYTA (Table 4). The λmax of the two groups of pigments differ by 4–9 nm (Table 4). Thus, the λmax shift caused by A108S is very small, if there is any, for pigments with Y197, but it is ~7 nm for pigments with H197. From these observations, it is expected that since both pigments x and z have AYYTA, A180S should not cause any significant λmax shift.
More on the five-sites rule: The statistical inferences strongly suggest that the five-sites rule explains the variation in the λmax of all contemporary and ancestral LWS/MWS pigments. When the five critical sites are considered, the contemporary LWS/MWS pigments are classified into nine different groups (Table 6). To test the validity of the statistical result further, we reconstructed the nine types of pigments by modifying the amino acid compositions at the five critical sites of pigment x. Using the in vitro assay, the λmax of these mutant pigments are measured. Table 6 clearly shows that, as long as the amino acid compositions at the five critical sites are the same, the λmax of the mutants are virtually identical to those of the corresponding contemporary pigments.
It should be stressed that these mutant pigments with the background of pigment x and the corresponding contemporary LWS/MWS pigments can have very different amino acid compositions at sites other than the 5 critical sites. For example, reflecting dynamic evolutionary processes of amino acid replacements during vertebrate evolution, pigment x differs from human (P560) pigment at 27 amino acid sites, whereas it differs from cave fish (P558) pigment at 91 sites. Thus, the majority of these amino acid replacements have contributed very little to the spectral tuning in the LWS/MWS pigments and may be considered as “neutral” changes. The mutagenesis analyses using the mammalian ancestral pigment strengthen the argument that the “five-sites” rule has been the molecular mechanism of spectral tuning in the LWS/MWS pigments throughout vertebrate evolution.
Red-green color vision: To study the structure-function relationships of the LWS/MWS pigments, we considered only the visual pigments with 11-cis-retinal (also known as vitamin A1 aldehyde). In nature, however, the λmax of these visual pigments can be modified further at the visual pigment level as well as at the photoreceptor cell level. It turns out that many fishes, amphibians, and reptiles can use 11-cis-3, 4-dehydroretinal (or vitamin A2 aldehyde) as the chromophore, which causes a red shift in the λmax (Whitmore and Bowmaker 1989; Harosi 1994). For example, goldfish (P559) pigment uses mostly vitamin A2 aldehyde as the chromophore and actually achieves a λmax value at ~620 nm (Palacioset al. 1998; Yokoyama and Radlwimmer 1999). Cave fish (P558) pigment uses vitamin A1 and A2 aldehydes with equal frequencies as the chromophore (Kleinschmidt and Harosi 1992). Thus, this pigment can achieve a λmax at ~560–620 nm depending upon the composition of the chromophore. Cone photoreceptor cells in many amphibians, reptiles, and birds use different devices, colored oil droplets. Although their exact functions have not been fully elucidated, the oil droplets often contain a high concentration of carotenoids and are likely to serve as cutoff filters (Bowmaker 1991). Both chicken (P561) and pigeon (P559) pigments are expressed in the cone cells with the red-colored oil droplets with cutoff wavelengths at 560–580 nm (Bowmakeret al. 1997). These photoreceptor cells may actually achieve λmax at 600–620 nm (Bowmaker and Knowles 1977). Thus, the photoreceptor cells with the LWS pigments in many vertebrate species actually detect more reddish light than the LWS pigments regenerated using the in vitro assay. By the same token, when vitamin A2 aldehyde is used as the chromophore, the MWS pigments can also achieve λmax at ~560 nm (see Palacioset al. 1998).
When the actual λmax of visual pigments are red shifted, how do animals achieve λmax at ~530 nm? It turns out that many species use an evolutionarily distantly related group of RH2 pigments for their green color detection (Yokoyama 2000a,b; Ebrey and Koutalos 2001). For example, by replacing vitamin A1 aldehyde with vitamin A2 aldehyde, the two goldfish RH2 pigments shift their λmax from ~510 nm to 530–540 nm (Johnsonet al. 1993; Palacioset al. 1998). The chicken photoreceptor cell with the RH2 pigments with λmax at ~510 nm actually achieves a λmax at 533 nm due to the presence of a green oil droplet in the cell (Bowmaker and Knowles 1977). Like the MWS pigments, these RH2 pigments are often called “green pigments,” sometimes creating confusing terminology. Placental mammals use neither vitamin A2 aldehyde nor colored oil droplets and, therefore, their red-green color vision is determined directly by the MWS and LWS pigments.
Vitamin A1 aldehyde is used by both vertebrates and invertebrates, but a vitamin A2 aldehyde-based pigment has not been found in invertebrates (Smith and Goldsmith 1990). Thus, vitamin A2 aldehyde might have been an evolutionary device to generate dual functions or a new function of one visual pigment. Similarly, as suggested by Thompson (1991), colored oil droplets might have been an efficient way to generate a variability in the spectral sensitivity in photoreceptors that had not yet achieved the mechanisms for expressing a single type of visual pigment in one photoreceptor cell (however, see Walls 1942). While these functional changes are underway in the RH2 pigments, it is reasonable to assume that the ancestor of the LWS/MWS pigments or photoreceptor cells with such pigments must have been red sensitive. These observations are also consistent with the idea that the vertebrate ancestor had the LWS pigment.
Purifying selection and adaptive evolution: For visual pigments, the numbers of amino acid replacements (K) can differ considerably between comparable branches. In Figure 3, for example, compared with the K for branch b-frog (P557) pigment (29), those for b-gecko (P527), b-mouse (P508), b-squirrel (P532), b-goat (P553), b-human (P530), and b-human (P560) pigments are 64, 99, 72, 80, 75, and 70, respectively. These values are all ⪢29 (P < 0.01). This difference is largely due to the accelerated evolution in the mammalian ancestor, represented by branch c–d. Similarly, the K for branches f-mouse (P508) pigment and h-gecko (P527) pigment are significantly larger than those for f-squirrel (P532) pigment and h-chameleon (P560) pigment, respectively (both with P < 0.01; Figure 3).
These results seem to show that the accelerated evolutionary rate of amino acid replacement leads to the acquisition of new functions. More specifically, the K tend to be larger for branches where amino acid replacements occur at the functionally important five sites 180, 197, 277, 285, and 308 than those without such changes. Furthermore, the gecko pigment has pure rod-retinas (Crescitelli 1972; Crescitelliet al. 1977) and the cone-specific MWS pigment had to adapt to this unique cellular change. The accelerated evolutionary rate of amino acid replacement for branch c–d is not immediately clear, but it may reflect the λmax shift from 558 nm to either 536 nm in the ancestral rodent pigment or 553 nm of the goat pigment (Figure 3). If it turns out that the rodents are most distantly related to the other groups of mammals, then the K associated with H197Y in the ancestral rodent pigment will be >8. On the other hand, if the goat pigment is most distantly related, then the K associated with S180A can be >26. Thus, there is a positive correlation between the evolutionary rate of amino acid replacement and the level of the divergence in the λmax. Such correlation has also been detected for various evolutionary groups of visual pigments in the gecko (Yokoyama and Blow 2001) and avian species (Yokoyamaet al. 2000). All of these examples seem to show that the relaxation from purifying selection caused by environmental changes is the first important step toward the adaptation of organisms to new environments.
Comments by Ruth Yokoyama, Stephen Schaeffer, and two anonymous reviewers are greatly appreciated. This work was supported by National Institutes of Health grant GM-42379.
Consider the relationships where ei's (i = 1, 2, … , 26) denote random errors.
This is represented in a matrix form as where and
If we assume that the random term, e, has a normal distribution with mean 0 and σ2I, then the mean and standard error of θ′ = [θ180 θ197 θ277 θ285 θ308 θ180×197 θ180×285 θ197×308 θ277×285 θ180×277×285 Z] are estimated from where and n and p denote the number of samples and parameters, respectively (Searle 1971).
Communicating editor: S. W. Schaeffer
- Received March 19, 2001.
- Accepted May 29, 2001.
- Copyright © 2001 by the Genetics Society of America