Natural Selection of the Rhodopsin Gene During the Adaptive Radiation of East African Great Lakes Cichlid Fishes

The Great Lakes of the East African Rift Valley, Lakes Victoria, Malawi, and Tanganyika, harbor approximately 200, 400, and 170 endemic species of cichlid fishes, respectively (Fryer and Iles 1972 , pp. 6–590; Greenwood 1991 ). These fishes have fascinated evolutionary biologists as spectacular examples of explosive adaptive radiation among living vertebrates because the fishes have adapted to a variety of niches in the lakes (Fryer and Iles 1972 ; Greenwood 1984 ). Accordingly, they are extremely diverse, both ecologically and morphologically, despite having evolved over a very short period (Meyer et al. 1990 ; Sturmbauer and Meyer 1992 ; Johnson et al. 1996 ; Takahashi et al. 2001 ). The visual systems of these fishes are of particular interest because they are important for feeding (Fryer and Iles 1972 ) as well as for mate choice (Seehausen, van Alphen, and Witte 1997 ; Seehausen and van Alphen 1998 ), and cichlids have adapted to different photic conditions that vary with depth and time of day (Fryer and Iles 1972 ; Coulter 1991 ). Furthermore, differences in visual capabilities among coexisting cichlid species may reduce the competition for resources (van der Meer and Bowmaker 1995 ). Therefore, it is reasonable to postulate that genes that are essential for visual acuity must have evolved adaptively during adaptive radiation of the cichlids in these African Great Lakes. We attempted to identify such genes by focusing on cichlid opsin genes. Visual pigments are composed of a light-absorbing chromophore, typically 11-cis-retinal, that is bound to the protein opsin (Wald 1968 ). The absorption spectrum of the chromophore can be altered by amino acid substitutions within opsin, and key substitution sites are located in the seven transmembrane α-helices that lie close to the retinal-binding pocket (Yokoyama R. and Yokoyama S. 1990 ; Yokoyama S. and Yokoyama R. 1996 ; Kochendoerfer et al. 1999 ). Cichlids have six known opsin genes whose products are sensitive to different wavelengths of visible-ultraviolet light, namely SWS-1 (ultraviolet), SWS-2A and SWS-2B (short wavelength), RH2 (midwavelength), LWS (long wavelength) (Carleton, Harosi, and Kocher 2000 ; Carleton and Kocher 2001 ), and rhodopsin (GenBank accession number AF315354). Rhodopsin is localized to rod cells that enable black and white images to be seen in dim light (Yokoyama R. and Yokoyama S. 1990 ; Yokoyama S. and Yokoyama R. 1996 ). The others are localized within cone cells that mediate color vision in bright light (Yokoyama S. and Yokoyama R. 1996 ; Yokoyama 2000 ). The evolution of visual pigments is a prime example of molecular adaptation in vertebrates (Yokoyama S. and Yokoyama R. 1996 ; Yokoyama 2000 ), and several such cases have been reported in diverse species such as primates (Shyue et al. 1995 ), birds (Yokoyama, Radlwimmer, and Blow 2000 ), and fish (Hunt et al. 1996 ; Yokoyama et al. 1999 ). As a first step toward understanding the molecular basis for the adaptation of opsins in cichlids, we cloned and characterized a cichlid rhodopsin gene. To identify possible adaptive changes in the rhodopsin gene during cichlid adaptive radiation, we sequenced the coding region (924 bp) from 19 species of the major cichlid lineages from the African Great Lakes and from five species of riverine cichlids (fig. 1a ). Southern blot analysis showed that the cichlid rhodopsin gene is encoded in a single gene at this level of analysis (data not shown). The gene was amplified using a pair of polymerase chain reaction (PCR) primers designated RhodF1 (5′-AGCCAGAAGAAACACCTCTGAAG-3′) and RhodR1 (5′-TTGGAGGCAGTAGAAGATGCT-3′) on the basis of the rhodopsin sequence of Astatotilapia burtoni (GenBank accession number AF315354). Amplifications were carried out in a PTC-100 Programmable Thermal Controller (MJ Research). Genomic DNA (50 ng) was added to the reaction mixture containing 1× ExTaq PCR buffer, 0.2 mM each of the four deoxynucleoside triphosphates, 0.1 μM primers, and 1 U ExTaq polymerase (Takara, Shiga, Japan). The PCR program consisted of a denaturation step for 3 min at 94°C, followed by 30 cycles consisting of 30 s denaturation at 94°C, 30 s annealing at 55°C, and 2 min extension at 72°C. Direct sequencing of the amplified gene was performed using a DNA sequencing kit (Applied Biosystems) and the Applied Biosystems Automated 3100 Sequencer. The nucleotide sequences were deposited in GenBank under accession numbers AB084924–AB084947. All pairwise values of nonsynonymous substitutions per nonsynonymous site (Dn) and synonymous substitutions per synonymous site (Ds) were calculated using version 2.1 of the program MEGA (Kumar et al. 2001 ). The ratio Dn/Ds provides an estimate of the evolutionary rate of amino acid substitutions as well as standard errors by bootstrap resampling (Miyata and Yasunaga 1980 ; Felsenstein 1985 ). To elucidate the evolution of a particular gene, in general, it is necessary to have prior knowledge of the phylogeny of the species in question. On the basis of the currently accepted phylogeny of cichlid species (fig. 1a, see legend), African cichlid fishes are divided into two groups (fig. 1a ). The riverine group includes a small number of species and thus may serve as an out-group for comparison purposes (Greenwood 1991 ; Ribbink 1991 ; Mayer, Tichy, and Klein 1998 ). The Great Lakes group includes members of the tribes in Lake Tanganyika and the East African riverine Haplochromine, as well as those of the Lake Victoria and Lake Malawi flocks (Mayer, Tichy, and Klein 1998 ). For comparison, average Dn/Ds values were calculated for both riverine and Great Lakes cichlid rhodopsin genes. We postulated that the rhodopsin gene has evolved along with the adaptation of each of the cichlid species in the lakes. If so, then amino acid substitutions should have occurred at an accelerated rate. Such a hypothesis would imply that average Dn/Ds values for the Great Lakes group should be higher than those for the riverine group because the adaptation for each niche occurred more frequently in the Great Lakes lineage. But if the rhodopsin gene did not play a role in adaptation in the Great Lakes cichlids, then the Dn/Ds values should be similar between the two groups. Figure 1a presents the cichlid phylogeny used in this study, as well as average Dn/Ds values within each group. Average values for the rhodopsin gene were estimated as 0.176 ± 0.028 for the riverine group and 0.915 ± 0.058 for the Great Lakes group, indicating that the average value for the Great Lakes group is fivefold higher than that for the riverine group. The confidence intervals of the estimates for the two groups did not overlap, and the difference between the estimated Dn/Ds values for the two groups was statistically significant, as estimated by bootstrap resampling (Felsenstein 1985 ). The higher average Dn/Ds value for the Great Lakes group suggests that the amino acids encoded by the rhodopsin gene underwent substitutions at a relatively accelerated rate in this group. Thus, there appears to be a correlation between the adaptation to each niche in the Great Lakes and the high rate of change in amino acids encoded by the rhodopsin gene. We also calculated average Dn/Ds values for the rhodopsin gene within a subset of the Great Lakes group, rapid speciated group, consisting of the tribes of Tropheini and Haplochromini (Poll 1986 ) in Lake Tanganyika as well as flocks in Lake Victoria and Lake Malawi (designated the THMV group; see fig. 1a ). The average Dn/Ds value for the THMV group was only slightly higher than that for the Great Lakes group, and the difference was not significant (fig. 1a ). The average Dn/Ds value for the THMV group was expected to be significantly higher; however, there were no synonymous substitutions in the four sequence pairs within the THMV group and thus the Dn/Ds values for these sequence pairs could not be included in the analysis. There are two reasonable explanations for the accelerated evolution of the rhodopsin gene, namely, positive selection or relaxation of negative selection. To discern these two alternatives, we performed a one-tailed Z-test (n = 1,000 replicates) and determined the level of significance of positive selection (Dn/Ds > 1 for each pairwise value in the Great Lakes group) using version 2.1 of the program MEGA (Kumar et al. 2001 ). Among the pairwise Dn/Ds values from the Great Lakes group, 25% were >1 (data not shown). Table 1 shows that 16 of the pairwise Dn/Ds values were >2, and seven of these were statistically significant (P < 0.05). These results suggest that the amino acids encoded by the cichlid rhodopsin gene have evolved at an accelerated rate by positive selection, at least in several lineages, during the adaptive radiation of cichlids in the Great Lakes. The fact that there is evolutionary convergence of amino acids encoded by the rhodopsin gene represents further evidence that positive selection occurred. At 19 amino acid positions (fig. 1b, shaded), substitutions of the same amino acid occurred in separate lineages, independently and repeatedly (for example, at position 292 in fig. 1a; see below). These substitutions were assumed to result from functional convergence. These findings raise the question of how variation relates to adaptation among Great Lakes cichlid fishes. Molecular adaptation of the rhodopsin gene has been reported in certain fish (Hunt et al. 1996 ; Yokoyama et al. 1999 ). The Comoran coelacanth (living fossil) lives at depths of ∼200 m and perceives only a narrow range of color around 480 nm. Two amino acid substitutions in the Comoran coelacanth proteins RH1 (rhodopsin) and RH2 (RH1-like opsin) elicit a shift in peak wavelength absorbance (λmax) of ∼20 nm toward the blue portion of the spectrum, relative to corresponding orthologous pigments (λmax ∼ 500 nm; Yokoyama et al. 1999 ). Another example is the Cottoid fishes that live at various depths in Lake Baikal, the deepest lake in eastern Siberia. The λmax of rhodopsin in Cottoid fish is shifted in a stepwise manner from 516 nm in species that live in shallow water to about 484 nm in species living in deep water (Hunt et al. 1996 ). This shifting of λmax is mediated by amino acid substitutions at four positions (Hunt et al. 1996 ). Among the amino acid substitutions in cichlid rhodopsins (fig. 1b ), it is presumed that position 292 might shift the absorption spectra. This position corresponds to position 308 in human red and mouse green pigments, and the amino acid change from alanine to serine (A308S) shifts the λmax of each pigment toward the blue portion of the spectrum by ∼18 nm (Sun, Macke, and Nathans 1997 ; Yokoyama 2000 ). Therefore, it is reasonable to expect that the cichlid rhodopsin A292S substitution would produce a similar shift (Yokoyama and Radlwimmer 1998 ). This substitution occurs in Trematocara macrostoma, Cyphotilapia frontosa, and Benthochromis tricoti (table 1 and fig. 1a ). The Trematocara species live in clear deepwater habitats during the daytime, and the fact that they typically have large eyes suggests adaptation under low illumination so as to improve visual acuity (Coulter 1991 ). Benthochromis tricoti has been caught more commonly near the deeper end of its range, and C. frontosa prefers a rocky bottom habitat in deep water, being caught most abundantly between 60 and 120 m in Lake Tanganyika (Coulter 1991 ). Therefore, the evolutionary convergence of the A292S substitution in cichlid rhodopsin might be due to adaptation for clear deepwater habitat (not muddy) via a blueshift of the λmax of the rod pigment. Differences in photic conditions with respect to water color may also have influenced the extent of adaptive evolution of cichlid rhodopsin. In contrast to the waters of Lake Malawi and Lake Tanganyika, those of Lake Victoria are generally quite turbid and, therefore, largely impenetrable to shorter-wavelength light. Color discrimination under these conditions therefore shift to longer wavelengths of the visible spectrum (van der Meer and Bowmaker 1995 ). As shown in figure 1b, amino acid substitutions at five positions were specific to Lake Victoria. These substitutions may be a consequence of adaptation to water color. The significance of these five substitutions as well as most of the others observed in cichlid rhodopsins remains unknown at present, and an analysis of potential substitution-dependent shifts in absorption in these rhodopsins is now in progress. The adaptive changes that are apparent in cichlid rhodopsin may help to explain the adaptation of cichlid fishes to the various niches in the East African Great Lakes. Analysis of the genes that play an important role in the cichlid visual system provides insight into the mechanisms of species diversification at the molecular level. Dan Graur, Reviewing Editor Keywords: cichlid adaptive radiation visual pigment gene rhodopsin positive selection Address for correspondence and reprints: Norihiro Okada, Graduate School of Bioscience and Biotechnology, Tokyo Institute of Technology, 4259 Nagatsuta-cho, Midori-ku, Yokohama 226-8501, Japan. nokada@bio.titech.ac.jp Table 1 Positive Selection of Rhodopsin Table 1 Positive Selection of Rhodopsin Fig. 1.—(a) Phylogeny of the cichlid species used in this study, and the average values of Dn/Ds within each group. The consensus phylogenetic tree of cichlids is based on molecular data from several sources (Nishida 1991 ; Kocher et al. 1993 ; Sturmbauer and Meyer 1993 ; Mayer, Tichy, and Klein 1998 ; Takahashi et al. 2001 ). The A292S above certain branches indicates that this amino acid substitution occurred in the lineage (if C. frontosa and B. tricoti form a sister-group, then alanine was substituted by serine in a common ancestor of these species). (b) The informative positions of amino acid substitutions among rhodopsin genes from African cichlids. Read vertically; the top three lines specify the amino acid positions in the consensus sequence of cichlid rhodopsin. A dot indicates identity with the consensus sequence. Shaded columns indicate that identical amino acid substitutions occurred independently in separate lineages We thank Drs. Y. Shichida and H. Imai (Graduate School of Science, Kyoto University, Kyoto, Japan) for their expertise with opsins and for helpful discussions and Dr. T. Sato (Enviromental Education WWF Coral Reef Conservation and Research Centre) for helpful discussions. This work was supported by the Ministry of Education, Culture, Sports, Science, and Technology of Japan.