Extraordinarily low evolutionary rates of short wavelength-sensitive opsin pseudogenes

Abstract

Aquatic organisms such as cichlids, coelacanths, seals, and cetaceans are active in UV-blue color environments, but many of them mysteriously lost their abilities to detect these colors. The loss of these functions is a consequence of the pseudogenization of their short wavelength-sensitive (SWS1) opsin genes without gene duplication. We show that the SWS1 gene (Bden_S1ψ) of the deep-sea fish, pearleye (Benthalbella dentata), became a pseudogene in a similar fashion about 130 million years ago (Mya) yet it is still transcribed. The rates of nucleotide substitution (~1.4 × 10⁻⁹ /site/year) of the pseudogenes of these aquatic species as well as some prosimian and bat species are much smaller than the previous estimates for the globin and immunoglobulin pseudogenes.

1. Introduction

The high level of DNA sequence variations found in nature has been explained by both neutral mutations (Nei 2005; Nei et al. 2010) and by adaptive mutations (Arbiza et al. 2006; Bakewell et al. 2007; Kosiol et al. 2008; Studer et al. 2008; Lindblad-Toh et al. 2011). The extremely high rates of nucleotide substitution (~5–13 × 10⁻⁹ /site/year) of pseudogenes have been used as strong supportive evidence for the neutral theory of molecular evolution (Li et al. 1981; Miyata and Yasunaga 1981; Nei 2005; Nei et al. 2010). However, recent molecular analyses of certain pseudogenes reveal that their presumed non-functionality is equivocal (Balakirev and Ayala 2003; Podlaha and Zhang 2010). The ENCODE team goes further by saying that more than two thirds of non-coding DNA sequences in the human genome are transcribed and have biochemical functions (The ENCODE Project Consortium 2012; Djebali et al. 2012). Yet, such pseudogenes are still subjected to much less selective constraint than protein-coding genes (Pei et al. 2012). Under these circumstances, it is of interest to re-evaluate the evolutionary rates of pseudogenes.

Many aquatic animals such as cichlids (e. g., Neolamprologus brichardi and N. mondabu) (O’Quin et al. 2010), coelacanths (Latimeria chalumnae and L. menadoensis) (Yokoyama et al. 1999; Yokoyama and Tada 2000), seals (Phoca groenlandica and P. vitulina), dolphin (Tursiops truncatus) (Newman and Robinson 2005), and whales (Globicephala melas, Mesoplodon densirostris, and Megaptera novaeangliae) (Newman and Robinson 2005; Koito et al. 2010) are active in UV-blue color environments, but they lost their abilities to make the UV/blue-sensitive or short wavelength-sensitive (SWS1) visual pigments. Their opsin genes have not been isolated, but many shark species have also lost the ability to make the SWS1 pigments (Hart et al. 2011).

Here, we isolated the SWS1 opsin gene from the deep-sea fish, pearleye (Benthalbella dentata). This gene (Bden_S1ψ) contains premature stop codons, the typical characteristic of a pseudogene, but it is transcribed in the retina. Comparing the SWS1 pseudogenes of the pearleye, cichlids, coelacanths, cetaceans, seals, prosimians (Galago senegalensis and Nycticebus coucang) (Kawamura and Kubotera 2004), and bats (Rhinolophus affinis and Rhinolophus ferrumequinum) (Zhao et al. 2009) with opsin-coding SWS1 genes, we study the evolutionary rates of nucleotide substitution before and after pseudogenization.

2. Materials and Methods

2.1. Molecular cloning of the pearleye SWS1 gene

High molecular weight DNAs of the pearleye (B. dentate) was isolated from body tissues using a standard phenol-chloroform extraction procedure (e.g. Yokoyama et al. 1999). Bden_S1ψ was cloned by polymerase chain reaction (PCR) first using a set of degenerate primers (F: 5′-GCNTCNACNCARAARGCNGA-3′ and R: 5′-ACRTANATNAYNGGRTTRTA-3′) and, then, by inverse PCR using another set of primers (F: 5′-GTGCACTTCTGAAGG-3′ and R: 5′-GGAGCCCACCGTCATCACG-3′). The PCR was performed by 30 cycles at 92°C for 45 sec, 55°C for 60 sec, and 72°C for 90 sec. At each cycle, the duration of the extension reaction was progressively extended by 3 sec.

The total retinal RNAs from the pearleye retina was also isolated as described previously (Yokoyama et al. 1995). To clone the SWS1 opsin cDNA of the pearleye, the internal sequence was cloned first by RT-PCR using the degenerate primers used for obtaining the first genomic sequence. To determine the rest of the cDNA sequences, we constructed additional gene specific primers (GSPs) and performed 5′ and 3′ rapid amplification of cDNA ends (RACE) analyses (e.g. Yokoyama et al. 1995). For the 3′ RACE, the first strand cDNA was made using the oligo (dT)-containing adaptor primer (AP), provided by the manufacturer (Gibco BRL, Gaitherburg, MD) and the original mRNA was degraded by RNase H. Then, two sequential PCR amplifications were performed applying two sets of GSPs and universal adapter primer (UAPs) to these cDNAs, first using (GSP1: 5′-CCGACGAGAACAAAGACTACCG-3′ and GSP2: 5′-CCATTCCAGCATTCTTCTCC-3′) with abridged UAPs supplied by the manufacturer. For the 5′ RACE, the cDNAs were first synthesized from total RNA using GSP1 (5′-CGGTAGTCTTTGTTCTCGTCGG-3′). The entire coding region was obtained by two sequential PCR amplifications, using two sets of nested GSPs (GSP2: 5′-AAGTACAACGCTGCGATGGC-3′ and GSP3: 5′-GGAGCCCACCGTCATCACG-3′) and abridged UAPs. Using these primers, cDNAs were reverse transcribed at 42°C for 1 hr, 95°C for 5 min and then PCR amplification was carried out for 30 cycles at 94°C for 45 sec, 55°C for 1.5 min, and 72°C for 2 min.

Nucleotide sequences of these cDNA clones were determined by cycle sequencing reactions using the Sequitherm Excel 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.

2.2. Inferences on phylogenetic trees and positive selection

In the analyses, we have aligned the nucleotide sequences of a total of 33 SWS1 opsin genes. The codons that are not shared by the functional SWS1 genes and pseudogenes were excluded (see Fig. S1). Then, the numbers of nucleotide substitutions per site (d) for pairwise comparisons were estimated by d = − (3/4) ln [1 − (4/3)p], where p is the proportion of different nucleotide per site (Jukes and Cantor 1969). The branch lengths of the composite phylogenetic tree of the 33 representative SWS1 opsin genes were inferred by PAML (Yang 2007) using the evolutionarily distantly-related RH1 gene of bovine (M21606) as well as RH2 (AB087805) and SWS2 (AB087809) genes of zebrafish as the outgroup.

A rooted phylogenetic tree of the SWS1 pseudogene (Bden_S1ψ) and the RH1 (Sana_RH1A and Sana_RH1B) and RH2 (Sana_RH2) genes of another pearleye species (Scopelarchus analis) (Table S1) was constructed applying the neighbor-joining (NJ) method (Saitou and Nei 1987) to their DNA sequences and those of the LWS opsin genes of zebrafish (AB087803) and goldfish (L11867) as the outgroup.

To search for positively selected amino acid sites, we studied the ratio of non-synonymous and synonymous nucleotide substitutions using the codon-based maximum likelihood (ML) based Bayesian method (Yang 2007). We considered the naive empirical Bayes (NEB) method without accounting sampling errors and Bayes empirical Bayes (BEB) method with accounting sampling errors (Yang et al. 2005). These Bayesian methods were applied to the SWS1 pseudogenes of 29 representative cetacean and 2 seal species (Table S1 and Fig. S2) using two initial ω values (0.4 and 3.4).

2.3. An evolutionary model for the SWS1 opsin genes

Using the two closely related SWS1 pseudogenes of coelacanths, seals, cetaceans, prosimians, and bats (sequences A and B) with known divergence times and two orthologous functional genes (sequences C and D), the times since pseudogenization (T_n) can be evaluated (Fig. 1A). In Fig. 1A, a and a′ describe the evolutionary rates of nucleotide substitution of historically younger and older groups of functional genes, respectively, while the parameter b denotes the evolutionary rate of the pseudogenes. For each data set of this four-sequence model, the numbers of nucleotide substitutions per site (d_XY) for pairwise comparisons were estimated by d_XY = − (3/4) ln [1 − (4/3) p_XY], where p_XY is again the proportion of different nucleotide per site (X, Y = A, B, C, D). Then, the relationships d_AB = 2bT₃, d_AC = d_BC = a(2T₂ − T_n) + bTn, d_AD = d_BD = a′(2T₁ − T₂) + aT₂ + (b − a)Tn, and d_CD = a′(2T₁ − T₂) + aT₂ hold. Hence, parameters a, a′, b, and T_n can be evaluated by

Fig. 1.

Plausible phylogenetic tree for pseudogenes and orthologous functional genes. (A) A tree consisting of two pseudogenes (sequences A and B) and two functional genes (sequences C and D). T₁, T₂, and T₃ denote divergence times between sequence D and others, between sequence C and sequence A (or B), and between sequences A and B, respectively, and T_n the time of pseudogenization of sequences A and B. (B) A tree consisting of one pseudogene (sequence A) and two functional genes (sequences C and D).

$$a=(d_{AC}+d_{BC}+2d_{CD}-d_{AD}-d_{BD})/(4T_2)$$ (1a)

$$a^{\prime }=(d_{AD}+d_{BD}+2d_{CD}-d_{AC}-d_{BC})/[4(2T_1-T_2)]$$ (1b)

$$b=d_{AB}/(2T_3)$$ (1c)

$$T_n=(d_{AC}+d_{BC}-4{aT}_2)/[2(b-a)].$$ (1d)

When another closely related pseudogene is not known (pearleye) or the divergence time of the two known pseudogenes has not been determined (cichlids), T_n can be evaluated using the formula considering the three sequence model (Fig. 1B): T_n = [(y₁ + y₂)/2 − y₃)/[a₃ − (a₁ + a₂)/2], where y_i = d_ADi − d_CDi (or d_BDi − d_CDi) and a_i is the rate of change at the i^th position of a codon (i = 1, 2, 3) (Li et al. 1981).

2.4. Method for estimating the variance of T_n

A resampling method based on using model generated population values was used to estimate the variance of T_n. Each resample used randomized branch lengths taken from the binomial distribution to calculate T_n values. For the three-sequence model there are three independent branch lengths (l_i, m_i and n_i in Li et al. 1981). In the four-sequence model there are five independent branches. Each of these lengths was used as the p parameter of a binomial distributed variate [k~B(N,p)] where N is the number of nucleotides compared and p (or d) is the proportion of changes in that branch. Replicate (1000) sets of randomized branch lengths (k/N) were used to generate distances (d_AB, d_AC, d_AD, d_BC, d_BD, d_CD) that were then used to calculate values of T_n according to equation (1d) in the four sequence model, or (6) in Li et al. (1981). The replicated values of T_n were used to estimate the variance of T_n. This procedure is analogous to the commonly used bootstrapping procedure where branch lengths are randomized by resampling the original sequence data (non-parametric bootstrapping), however, in the present case we used parametric bootstrapping by assuming the binomial is the appropriate distribution for the branch length observations.

3. Results

3.1. The genomic and cDNA sequences of the Bden_S1ψ

The pearleye SWS1 gene (Bden_S1ψ) is characterized by the deletions of two stretches of its DNA sequence: compared with a typical functional SWS1 gene (Sleu_S1) of the lampfish (Table S1), 1) the segments between codons 203 and 220 and the entire intron 4 are missing from Bden_S1ψ and 2) the initiation codon ATG was replaced by ATA and a single nucleotide insertion and deletion can be found between codons 47 and 48 and at codon 224, respectively (Fig. 2A). These structural changes introduce several premature stop codons into Bden_S1ψ (Fig. 2B).

Fig. 2.

The SWS1 opsin gene of the pearleye (Bden_S1ψ). (A) The nucleotide sequences of the opsin cDNA (without introns) and genomic DNA (with introns). The deleted nucleotides in Bden_S1ψ are shown by dashes (−). The codon sites follow those of the closely related functional SWS1 opsin gene of the lampfish (Sleu_S1). (B) The amino acid sequence deduced from the Bden_S1ψ. Stars (*) indicate stop codons.

3.2. Branch lengths of the composite phylogenetic tree of the SWS1 opsin genes

For a total of 33 species, consisting of 15 species with the SWS1 pseudogenes and 18 representative species with the orthologous opsin-coding genes (Fig. 3A and Table S1), we first established a tree topology using “TimeTree of Life” web (www.timetree.org) server. This was done to avoid obtaining an erroneous tree topology that may be caused by the possibility of long branches of pseudogenes. This composite evolutionary tree shows that the SWS1 genes have become pseudogenes independently along seven separate lineages (pearleye, cichlids, coelacanths, bats, prosimians, cetaceans, and seals). Based on this tree topology and considering all positions of codons common to the 33 SWS1 genes, we evaluated the lengths of various branches (Fig. 3A).

Fig. 3.

Phylogenetic tree of SWS1 opsin genes. (A) The ML tree of 33 representative genes constructed using the program PAML. (For the names of specific species and the source of the sequence data, see table S1.) (B) The NJ tree of SWS1 Bden_S1ψ, Bden_RH1A, Bden_RH1B, and Bden_RH2 of the pearleye was constructed by using the NJ method, in which the time scale was based on the a value for the coelacanth data set.

Much to our surprise, we could not find extraordinarily long branches leading to the pseudogenes, which was unexpected from the previous results on the evolutionary rates of the globin and immunoglobulin pseudogenes (Li et al. 1981; Miyata and Hayashida 1981; Miyata and Yasunaga 1981). In particular, the branch leading to the coelacanth pseudogenes is the shortest among the 33 SWS1 genes. This ML tree has a striking resemblance to the phylogenetic tree based on the 251 concatenated protein-coding genes of various vertebrates (Amemiya et al. 2013), in which the coelacanth has the shortest branch length. In the coelacanth, therefore, both the SWS1 pseudogene and protein-coding genes have evolved very slowly. Similarly, the branch lengths of the pseudogenes of pearleye, cichlids, cetaceans, and seals are similar to those of the closely-related opsin-coding genes in other species. On the other hand, the pseudogenes of some bat species and prosimians (galago and loris) have much longer branch lengths than those of closely related opsin-coding genes; however, when they are compared to the orhtologous human gene, the branch length differences become less prominent.

3.3. Divergence times and evolutionary rates

Considering the seven sets of pseudogenes with orthologous opsin-coding SWS1 genes separately, we estimated the evolutionary rates of nucleotide substitution before (a) and after pseudogenizations (b) as well as the time of pseudogenization (T_n) by considering four sequence and three sequence models separately (see Section 2.3).

For the coelacanth, seal, cetacean, prosimian, and bat pseudogene data, divergence times (T₁, T₂, and T₃) have been estimated by using the “TimeTree of Life” web server (www.timetree.org). Using these divergence times, T_n values vary from 16 Mya of the seal genes to 138 Mya of the coelacanth pseudogenes; for the cichlid and pearleye pseudogenes, T_n values are given by 16 and 134 Mya, respectively (Table 1).

Table 1.

Estimated times of pseudogenization (T_n).

Group	Sequences	Divergence times (Mya)
		T1a	T2a	T3a	Tn
coelacanths	A: LchaS1ψ, B: LmenS1ψ, C: XlaeS1; AcarS1; ClivS1; HsapS1, D: DrerS1; SsalS1	455	430	5.5	138 ± 119d
seals	A: PvitS1ψ, B: PgroS1ψ, C: CfamS1, D: BtauS1; SscrS1; HsapS1	89	44	7.8	16 ± 7d
cetaceans	A: MnovS1ψ, B: TtruS1ψ; GmelS1ψ; MdenS1ψ, C: BtauS1; SscrS1, D: HsapS1	91	61	32.3	61e
prosimians	A: GsenS1ψ NcouS1ψ, C: DmadS1; EfulS1, D: BtauS1; SscrS1; CfamS1	97	62	34.2	62e
bats	A: RferS1ψ RaffS1ψ, C: TausS1; MfulS1, D: CfamS1; BtauS1; SscrS1	83	62	17	50f
cichlids	A: NmonS1ψ; NbriS1ψ OnilS1, D: OlatS1; PretS1	104b	36b	?	16g
pearleye	A: BdenS1ψ, C: SleuS1; OnilS1, D: DrerS1; SsalS1	284c	160c	?	134 ± 30h

^aThe divergence times have been evaluated from TimeTree of Life (www.timetree.org).

^bThe divergence times are taken from Genner et al. (2007).

^cThe average divergence time between pearleye and lampfish (307 Mya) and between pearleye and tilapia (264 Mya).

^dThe standard errors were estimated by a parametric Monte Carlo resampling (or the parametric bootstrap) method.

^eWe were unable to compute proper standard errors because T_n was assumed to be equal to T₂.

^fWe were unable to compute proper standard errors because T_n was assumed to be equal to the divergence time (50 Mya) between the pseudogenes from Rhinolophus affinis (and Rhinolophus ferrumequinum) and the closely-related orthologous functional gene from Megaderma spasma (www.timetree.org)).

^gThe data set used were d_AC1 = 0.045, d_AD1 = 0.125, d_CD1 = 0.115, d_AC2 = 0.022, d_AD2 = 0.087, d_CD2 = 0.067, d_AC3 = 0.120, d_AD3 = 0.505, and d_CD3 = 0.481.

^hThe data set used were d_AC1 = 0.137, d_AD1 = 0.221, d_CD1 = 0.178, d_AC2 = 0.090, d_AD2 = 0.127, d_CD2 = 0.094, d_AC3 = 0.584, d_AD3 = 0.677, and d_CD3 = 0.795.

?: unknown.

The four-sequence analyses show that a values vary between 0.40 and 0.77 × 10⁻⁹ /site/year, while b varies between 0.84–1.65 × 10⁻⁹ /site/year, respectively (Table 2, the last column). The reliabilities of formulae (1a), (1c), and (1d) based on the four-sequence model can be tested by using the three-sequence model with the T_n values estimated and relationships a_i = (d_ACi − d_ADi + d_CDi)/(2T₂) and b_i = [d_ACi − a_i (2T₂ − T_n)]/T_n (i = 1, 2, and 3). The results show that the average values (a) of a₁, a₂, and a₃ vary between 0.47 and 0.89 × 10⁻⁹ /site/year with the overall average of 0.65 × 10⁻⁹ /site/year; similarly, the average values (b) of b₁, b₂, and b₃ vary between 0.80 and 2.0 × 10⁻⁹ /site/year with the overall average of 1.37 × 10⁻⁹ /site/year (Table 2). These a and b values are very similar to the corresponding a and b values estimated using formulae (1a) and (1c), respectively, justifying the use of these formulae. In addition, despite significant differences between T₃ and T_n (Table 1), the evolutionary rates of the pseudogenes estimated using the three- and four-sequence models (b vs b) are similar, again justifying the use of formulae (1c) and (1d) in estimating b and T_n, respectively.

Table 2.

Evolutionary rates of nucleotide substitution of SWS1 opsin genes.

Group	Evolutionary rates (× 10−9)				All positions (× 10−9)
	a1	a2	a3	a1+2	a	a
coelacanths	0.27 ± 0.05	0.15 ± 0.04**	0.99 ± 0.13**,††	0.21 ± 0.03††	0.47 ± 0.04	0.40
seals	0.05 ± 0.06**	0.24 ± 0.13	1.46 ± 0.33**,††	0.15 ± 0.07††	0.58 ± 0.12	0.56
cetaceans	0.31 ± 0.13	0.16 ± 0.09**	1.41 ± 0.28**,††	0.24 ± 0.08††	0.63 ± 0.11	0.60
prosimians	0.44 ± 0.15	0.24 ± 0.11**	0.93 ± 0.22**,†	0.34 ± 0.09†	0.54 ± 0.10	0.52
bats	0.53 ± 0.20	0.25 ± 0.13**	1.61 ± 0.35**,††	0.39 ± 0.12††	0.80 ± 0.24	0.77
cichlids	0.48 ± 0.20	0.02 ± 0.04**	1.34 ± 0.34**,††	0.25 ± 0.10††	0.61 ± 0.13	ND
pearleye	0.29 ± 0.08	0.18 ± 0.06**	2.19 ± 0.26**,††	0.23 ± 0.05††	0.89 ± 0.08	ND
Average	0.34 ± 0.04	0.18 ± 0.03**	1.42 ± 0.19**,††	0.25 ± 0.02††	0.65 ± 0.07	0.57
	b1	b2	b3	b1+2	b	b
coelacanths	1.13 ± 0.20	0.52 ± 0.13**	1.25 ± 0.22**	0.83 ± 0.12	0.97 ± 0.11	0.94
seals	2.57 ± 0.71*	1.06 ± 0.45*	1.20 ± 0.47	1.81 ± 0.42	1.61 ± 0.32	1.65
cetaceans	0.65 ± 0.19	0.75 ± 0.21	1.30 ± 0.28†	0.68 ± 0.14†	0.90 ± 0.13	0.84
prosimians	1.88 ± 0.33	1.12 ± 0.25	1.65 ± 0.30	1.50 ± 0.20	1.55 ± 0.17	1.26
bats	1.22 ± 0.34*	2.76 ± 0.53*	2.03 ± 0.44	1.99 ± 0.31	2.00 ± 0.44	1.61
cichlids	1.12 ± 0.47	1.29 ± 0.50	2.82 ± 0.75	1.22 ± 0.34	1.74 ± 0.34	ND
pearleye	0.56 ± 0.11	0.39 ± 0.09**	1.46 ± 0.20**,†	0.52 ± 0.08†	0.80 ± 0.08	ND
Average	1.31 ± 0.16	1.13 ± 0.13	1.65 ± 0.25	1.22 ± 0.10	1.37 ± 0.17	1.26

The largest difference among a_i or b_i (i = 1, 2, 3) is significantly different at the 5% (*) and 1% level (**). The difference between a₁₊₂ and a₃ or between b₁₊₂ and b₃ is significantly different at the 5% (^†) and 1% level (^††). ND: not determined.

In these analyses, the numbers of nucleotide substitutions per site (d) were estimated using the Jukes and Cantor (JC) method (Section 2.2), which underestimates the d value as the divergence time of a pair of sequences increases. However, the d values for the SWS1 gene pairs are usually much smaller than 1.0 and the JC method gives reasonably accurate T_n, a, and b values. For example, Tamura and Nei’s (1993) method corrects the underestimation very effectively (see Fig. 3.1 in Nei and Kumar 2000). For the coelacanth data, for example, T_n (138 Mya), a (0.40 × 10⁻⁹) and b (0.94 × 10⁻⁹) values obtained using Tamura and Nei method are virtually identical to the corresponding estimates of 138 Mya, 0.40 × 10⁻⁹, and 0.94 × 10⁻⁹ under the JC model.

For the opsin-coding genes, a₃ is generally the largest, followed by a₁ and a₂, in that order, showing the functional constraint on the nucleotide changes expected for the protein-coding genes (Kimura 1983). As may be expected from the relationship a₃ > a₁ > a₂ for the opsin-coding genes, the evolutionary rate at the first and second positions of a codon (a₁₊₂) is always much smaller than a₃ (Table 2).

For the pseudogenes, the b₁, b₂, and b₃ values cannot be ordered in a uniform fashion and any one of them can be largest. The overall average values of b₁, b₂, and b₃ for the seven sets of pseudogenes are given by 1.31 × 10⁻⁹, 1.13 × 10⁻⁹, and 1.65 × 10⁻⁹ /site/year, respectively. Compared with those of opsin-coding SWS1 genes, the corresponding b₁₊₂ and b₃ values do not differ significantly, for five out of the seven sets of pseudogenes, revealing dramatically different patterns of nucleotide substitution before and after pseudogenization (Table 2). This supports the basic assumption that the rates of nucleotide substitution at the three positions of codons are uniform for the pseudogenes (Li et al. 1981; Miyata and Hayashida 1981; Miyata and Yasunaga 1981). The conservative nature of the SWS1 pseudogene evolution is suspected (Fig. 3A), but it is still surprising to see that this evolutionary rate is 4–10 times smaller than those of the globin and immunoglobulin pseudogenes.

When all positions of a codon are considered, the opsin-coding SWS1 genes and SWS1 pseudogenes have evolved at rates of ~0.6 × 10⁻⁹ and ~1.4 × 10⁻⁹ /site/year, respectively. Since, the evolutionary rates of the opsin-coding rhodopsin (RH1) and middle and long wavelength-sensitive (M/LWS) genes are 0.3–0.6 /site/year (Yokoyama and Yokoyama 1990a, b), the evolutionary rates of the opsin-coding SWS1 genes are similar to those of the other opsin genes and the SWS1 pseudogenes have evolved 2–3 times faster than the opsin-coding genes.

4. Discussion

The SWS1 pseudogene of pearleye (B. dentata) lost its opsin-coding ability about 130 Mya (Table 1). From another genus of pearleye (Scopelarchus analis), three functional opsin genes have been sequenced: two RH1 (Sana_RH1A and Sana_RH1B) genes and one RH2 gene (Sana_RH2) (Pointer et al. 2007). The NJ tree shows that Sana_RH1A and Sana_RH1B are most closely related, their ancestor diverged from the ancestor of Sana_RH2, and their common ancestor diverged from the ancestor of Bden_S1ψ before that (Fig. 3B). This was expected from the phylogenetic relationship of the five paralogous opsin genes (Yokoyama 2000). If we take the a value as 0.5 × 10⁻⁹ /site/year, then the divergence time between the RH1 and RH2 genes is 540 Mya (Fig. 3B), which is also consistent with the previous observation that the vertebrate ancestor already possessed all five groups of evolutionarily distant visual pigments (Yokoyama and Yokoyama 1996). However, the totally unexpected feature of the NJ tree is that despite the old pseudogenization event of Bden_S1ψ, the pseudogene and the paralogous opsin-coding genes have maintained similar evolutionary rates for the last 800 My.

Recent ENCODE analyses suggest that a significant portion of non-coding DNA sequences, including over a 17,000 DNA stretches of pseudogenes (www.pseudogene.org), is transcribed and is used for gene regulation (Djebali et al. 2012; however see Graur et al. 2013; Doolittle 2013). From this survey, the expression of Bden_S1ψ may not be surprising. In animals, short antisense RNAs are used to inhibit translation or to degrade cytoplasmic mRNA post-transcriptionally, which not only protect against viral infection, prevent transposon mobilization, and regulate the expression of endogenous genes but also maintain genome integrity by preventing transposon mobilization and double-stranded break repair (Castel and Martienssen 2013; Leslie 2013). One way to determine whether or not these SWS1 pseudogenes have such biochemical functions may be to try to isolate double-stranded RNAs using methods applied to the functional analyses of pseudogenes in mice (Tam et al. 2008; Watanabe et al. 2008). However, since the pearleye does not have any evolutionarily closely related opsin-coding SWS1 gene, the isolation of double stranded RNAs may not be a fruitful approach. The problem is further complicated by the fact that no mRNAs of the SWS1 pseudogenes of the cichlid, coelacanth, prosimian, bat, seal, and cetacean species has been identified, strongly suggesting that they are not transcribed. For these reasons, we cannot offer any plausible explanation for the slow evolution of the SWS1 pseudogenes.

In analysing various DNA sequence data, molecular evolutionists and molecular biologists often claim the existence of adaptive evolution by showing that the number of nonsynonymous substitutions per nonsynonymous site (dn) is greater than that of synonymous substitutions per synonymous site (ds) (Hughes and Nei 1988; Nei and Kumar 2000). The condition of “dn > ds (or ω = dn/ds > 1),” however, is an untested assumption and these statistical results contain significant proportions of false positives and false negatives (Yokoyama et al. 2008; Nozawa et al. 2009a). Recently, theoretical basis and reliabilities of these statistical methods have been debated intensely among statistical evolutionary geneticists (Nozawa et al. 2009a; Nozawa et al. 2009b; Yang et al. 2009; Yang and dos Reis 2011; Nei 2013). These authors, however, seem to agree that the final answer of adaptive evolution can be obtained only by subjecting the statistical results to some form of experimental test.

SWS1 pseudogenes do not encode functional opsins and the concept of codon is irrelevant. Consequently, the positively selected sites inferred by any statistical methods based on the condition of “dn > ds” are false positives and, therefore, the pseudogenes can be used as a negative control in evaluating the reliabilities of such statistical methods. Hence, a total of 88 codons that are common to the 29 representative cetacean and two seal SWS1 pseudogenes were analysed by the NEB and BEB approaches of Bayesian method. We found false-positives at 3 codon sites (69, 107, and 118) and 2 sites (107 and 118) using the NEB and BEB models, respectively (Table S2). If these sequences could encode amino acids, then 4, 8, and 3 amino acid changes would have occurred at sites 69, 107, and 118, respectively (Fig. 4). At site 107, a total of 8 nucleotide substitutions occurred only at the first and second positions of codons: A → C (once), G → A (twice), G → T (once), C → A (once), and C → T (three times). These nucleotide changes agree well with the mutation profiles of pseudogenes observed, where the mutations C → T (66%) and G → A (62%) are particularly high (Li et al. 1984). It is expected that 5 and 3 changes should occur at the first two positions and at the third position of a codon, respectively, but the chance that all 8 mutations occur at the first two positions is still 0.1. Similarly, the four hypothetical nonsynonymous substitutions each at sites 69 and 118 exhibit the pseudogene-characteristic nucleotide substitutions. Therefore, we do not have to invoke any positive selection for the biased nucleotide substitutions in the cetacean pseudogenes. These observations again warn the danger of the blind use of the untested assumption of dn > ds (or ω > 1) in inferring positive selection and show the necessity of experimental tests of such statistical predictions.

Fig. 4.

Positively selected codon sites with hypothetical amino acid replacements predicted by the NEB and BEB approaches of the Bayesian method. The composite phylogenetic tree is based on the three sources (Price et al. 2005; Koito et al. 2010; McGowen 2011) and the sources of their nucleotide sequences are given in Table S1. *significant at the 5% level. ** significant at the 1% level.

Highlights.

• We cloned the SWS1 opsin pseudogene of the deep-sea fish pearleye

• SWS1 genes in several vertebrate lineages have lost their abilities to make opsins

• The pseudogenization events took place separately 15 - 140 × 10⁶ years ago

• The evolutionary rates of these pseudogenes ranged between 0.9–2.0 × 10⁻⁹/site/year