The Evolution of Unusually Small Amelogenin Genes in Cetaceans; Pseudogenization, X–Y Gene Conversion, and Feeding Strategy

Abstract

Among extant cetaceans, mysticetes are filter feeders that do not possess teeth and use their baleen for feeding, while most odontocetes are considered suction feeders, which capture prey by suction without biting or chewing with teeth. In the present study, we address the functionality of amelogenin (AMEL) genes in cetaceans. AMEL encodes a protein that is specifically involved in dental enamel formation and is located on the sex chromosomes in eutherians. The X-copy AMELX is functional in enamel-bearing eutherians, whereas the Y-copy AMELY appears to have undergone decay and was completely lost in some species. Consistent with these premises, we detected various deleterious mutations and/or non-canonical splice junctions in AMELX of mysticetes and four suction feeding odontocetes, Delphinapterus leucas, Monodon monoceros, Kogia breviceps, and Physeter macrocephalus, and in AMELY of mysticetes and odontocetes. Regardless of the functionality, both AMELX and AMELY are equally and unusually small in cetaceans, and even their functional AMELX genes presumably encode a degenerate core region, which is thought to be essential for enamel matrix assembly and enamel crystal growth. Furthermore, our results suggest that the most recent common ancestors of extant cetaceans had functional AMELX and AMELY, both of which are similar to AMELX of Platanista minor. Similar small AMELX and AMELY in archaic cetaceans can be explained by gene conversion between AMELX and AMELY. We speculate that common ancestors of modern cetaceans employed a degenerate AMELX, transferred from a decaying AMELY by gene conversion, at an early stage of their transition to suction feeders.

Introduction

Cetacea comprises Mysticeti (mysticete) and Odontoceti (odontocete) in the crown group (Fordyce 2018). The total group Cetacea split from the lineage that led to Hippopotamidae ~ 53 Mya and adapted to the life in fully aquatic environments ~ 40 Mya (Fig. 1) (Gatesy et al. 2013). It was immediately after this transition that Mysticeti and Odontoceti were separated ~ 36 Mya (McGowen et al. 2009). Cetacea and Hippopotamidae are phylogenetically close to Ruminantia, as well as Suina and Tylopoda, and all these clades constitute Cetartiodactyla (Fig. 1) (Nikaido et al. 1999; Geisler 2018). In modern Cetacea, Mysticeti are filter feeders that do not possess teeth (teeth may develop but degrade before eruption in the fetus) and use the baleen for feeding (Ishikawa and Amasaki 1995; Peredo et al. 2017; Berta and Deméré 2018), while most Odontoceti are considered suction feeders, which capture prey by suction and swallow the prey whole without biting or chewing with teeth (Werth 2000, 2006). Changes in their feeding strategy presumably led to relaxation of purifying selection on various genes that are specifically involved in dental enamel formation (Deméré et al. 2008; Meredith et al. 2009, 2011; Kawasaki et al. 2014; Springer et al. 2016).

Fig. 1.

Phylogeny of modern Cetartiodactyla. A scale for the divergence times is shown at the bottom (Mya, million years ago). Cetaceans are thought to be adapted to life in fully aquatic environments ~ 40 Mya (dark background). The phylogeny and divergence times are based on previous studies (Montgelard et al. 1997; Benton and Donoghue 2007; Gatesy 2009; McGowen et al. 2009; Gatesy et al. 2013). The divergence time of N. a. asiaeorientalis and N. phocaenoides is not to scale (shown with dashed branches). D, P, and M represent Delphinidae, Phocoenidae, and Monodontidae, respectively. For the Tursiops, Sousa, and Kogia genera, T. truncatus, S. sinensis, and K. sima are shown. See Table S1 for common names

Enamel forms in a highly specialized organic matrix, in which mineral grows as thin ribbons (Hu et al. 2016; Smith et al. 2016). The enamel ribbons extend from the surface of dentin to the apical membrane of ameloblasts that secrete the matrix. In humans, and in most other mammals, each ameloblast raises a Tomes’ process on the apical surface, and enamel ribbons produced in the distal and proximal portions of Tomes’ processes are organized into a characteristic structure (Simmer et al. 2012; Nanci 2017), known as the enamel prism that is thought to resist abrasion of teeth (Sander 2000). Enamel grows as the ameloblasts retreat from the dentin-enamel junction. When the enamel matrix reaches near final thickness, ameloblasts retract their Tomes’ processes. The organic materials are subsequently removed from the developing enamel, and the vacant space is filled with mineral. Enamel thus matures into a hypermineralized inorganic tissue (Simmer et al. 2012; Nanci 2017).

The organic enamel matrix principally consists of amelogenin (AMEL), ameloblastin (AMBN), and enamelin (ENAM) (Moradian-Oldak 2012; Bartlett 2013), all of which are members of the secretory calcium-binding phosphoprotein (SCPP) family (Kawasaki and Weiss 2003). The SCPP family also includes proteins involved in enamel maturation in bony vertebrates (Kawasaki 2013) and milk caseins and salivary proteins in mammals (Kawasaki et al. 2011). All genes encoding SCPPs arose by tandem duplication and form two clusters on a single autosome in most eutherians (Kawasaki et al. 2009). The only exception is AMEL that is located on the X and Y chromosomes (AMELX and AMELY, respectively) (Lahn and Page 1999) or only on the X chromosome, if AMELY was secondarily lost (Lau et al. 1989). Other than eutherians, AMEL is located on an autosome but has never been found clustered with any other SCPP genes (Kawasaki and Amemiya 2014). AMEL encodes the most abundant enamel matrix protein that accounts for > 90% of the total proteins in the matrix (Fincham et al. 1999). In humans, various mutations in AMELX have been associated with enamel defects, known as amelogenesis imperfecta (OMIM 301200) (Smith et al. 2017). Furthermore, AMELX-null mice initially generate mineral ribbons, but these mice fail to form the Tomes’ process (Hu et al. 2016; Smith et al. 2016) and develop only a thin prismless enamel (Gibson et al. 2001). These studies suggest that AMEL is necessary for thickening the enamel layer (Hu et al. 2016).

In eutherians, recombination between the X and Y chromosomes is restricted to the pseudoautosomal region (PAR), and the male-specific region of the Y chromosome (MSY) does not undergo recombination (Skaletsky et al. 2003; Graves 2006). Because of the lack of X–Y recombination, most genes located in the MSY accumulated deleterious mutations or were completely lost from the chromosome (Ross et al. 2005). AMELY is thought to be among such genes that undergo decay and loss (Graves 1995; Wilson and Makova 2009). The MSY appears to have expanded occasionally through a series of intrachromosomal inversions (Lahn and Page 1999). It is thought that AMEL was translocated from an autosomal region to the PAR in eutherians, and that AMELY subsequently became part of the MSY, as the MSY expanded (Graves 1995; Lahn and Page 1999). The exact boundary and the timing of the expansion event that led to differentiation into AMELX and AMELY are not well resolved, however (Lahn and Page 1999; Iwase et al. 2003; Lemaitre et al. 2009; Pandey et al. 2013; Bellott et al. 2014; Cortez et al. 2014). Arguably, the low resolution is partly due to gene conversion (unidirectional genetic exchange) between AMELX and AMELY, because gene conversion erases the sequence divergence, accumulated since the suppression of recombination (Wyckoff et al. 2002; Marais and Galtier 2003; Bellott et al. 2014).

In the present study, we addressed the functionality of AMELX and AMELY in cetaceans, because teeth are not essential for feeding in mysticetes and in various suction feeding odontocetes, and because AMELY is thought to be decaying. As expected, various nonsense mutations, frameshift mutations, and non-canonical splice junctions were identified in various AMELX and AMELY genes. Furthermore, in cetaceans, both AMELX and AMELY are unusually small, and even their functional genes presumably encode a degenerate protein. The unusually small cetacean AMELX and AMELY genes can be explained by a single gene conversion event between these two genes, which occurred in a common ancestor of modern cetaceans. We speculate that this event occurred during the transition of their feeding strategy.

Materials and Methods

Nucleotide Sequences of AMEL Genes

We amplified intron 5 and exon 6 of AMEL sequences by PCR using genomic DNA of various cetaceans as templates and determined the nucleotide sequences of these PCR products (see Table S1 for GenBank accession numbers). Exon 6 is the largest protein-coding exon (encoding 121 out of 171 residues in most cetacean AMEL genes) and encodes the entirety of the variable core region (see below). Intron 5 has a large interstitial insertion (Macé and Crouau-Roy 2008), and an inquiry about the presence or absence of this insertion led us to investigate X–Y gene conversion in cetacean AMEL genes. For Phocoenoides dalli, cDNA of AMELX was cloned and sequenced. Information about these samples, PCR primers, and detailed procedures is summarized in Supplementary Methods. For Tursiops aduncus, Sousa chinensis, Phocoena phocoena, Monodon monoceros, Lipotes vexillifer, Pontoporia blainvillei, Ziphius cavirostris, Mesoplodon bidens, Platanista minor, Kogia breviceps, and various artiodactyl species, including Capra hircus, we identified AMEL genes in the assembled genomic sequences (Table S1) by blastn searches at the NCBI website (https://www.ncbi.nlm.nih.gov/) using already known AMEL sequences from other species as queries. Although the sex of the L. vexillifer, M. bidens, and K. breviceps samples used for genome sequencing is unknown, we considered these samples to be female, because the dataset does not contain SRY and because only one AMEL sequence was identified. We distinguished AMELX from AMELY in Bos taurus using their cDNA sequences (Gibson et al. 1992). Because the latest version of the B. taurus genome sequence (NKLS02) does not contain AMELY, we searched Y chromosome sequence contigs for AMELY using blastn (Table S1). For the other cetacean species and Hippopotamus amphibius (Table S1), we identified AMEL sequences in the SRA database available at the NCBI website by blastn searches using already known AMEL sequences from other species as queries.

During our study, we identified many misassembled genomic sequences for regions containing AMELX and AMELY, mostly chimeric sequences of these two genes. Because nucleotide sequences of cetacean AMELX and AMELY genes show high similarities, we used assembled genome sequences of male cetacean samples, only when their raw SRA data are available. When the raw SRA data are available, we assembled these data (Table S1) using the CAP3 software (https://doua.prabi.fr/software/cap3) (Huang and Madan 1999) and manually corrected assembled sequences. When the SRA data are available for both female and male samples, we first determined the nucleotide sequence of AMELX using the female data and then AMELY using the male data. Although reading lengths of SRA data are small, we were able to detect a sequence difference in every 100 or 150 nucleotides for most datasets and distinguished two closely related AMELX and AMELY sequences of cetacean samples, including Tursiops truncatus, Orcinus orca, Neophocaena asiaeorientalis asiaeorientalis, and Eschrichtius robustus (Fig. S1). For AMELX and AMELY that show an extremely high sequence identity, we determined their nucleotide sequences using PCR products (e.g., in Z. cavirostris, AMELX differs from AMELY by only one nucleotide; see below).

Bioinformatic Analysis

The nucleotide sequences were processed for multiple alignment software, TCOFFEE (https://www.tcoffee.org/) (Notredame et al. 2000) and used for various analyses. Phylogenetic trees were constructed using MEGA7.0.21 by the maximum likelihood method employing the best-fit nucleotide substitution model (the Tamura 3-parameter model for the LTRup tree and the Kimura 2-parameter model for the LTR and exon 6 trees) (Nei and Kumar 2000; Kumar et al. 2016). To construct phylogenetic trees, all gaps in the sequence alignment were excluded from our analyses and treated all sequences as an equal length. The reliability of the interior branches in these phylogenetic trees was calculated by 500 bootstrap samples. We also used MEGA7.0.21 for estimation of the number of synonymous (dS) and nonsynonymous (dN) differences per site using the modified Nei-Gojobori method with the Jukes-Cantor correction (Nei and Kumar 2000). MEGA7.0.21 was also used to reconstruct the exon 6 sequence in the most recent common ancestors (MRCA) of odontocetes and mysticetes. In this process, all frameshift mutations were restored by adding deleted nucleotides or removing inserted nucleotides based on sequences of closely related species. The average disorder prediction score was estimated using the VLXT predictor of the PONDR software (https://www.pondr.com/) (Romero et al. 1997, 2001; Li et al. 1999).

Results

Mutations in Cetacean AMELX and AMELY Genes

First, we investigated the entire protein-coding nucleotide sequence of AMELX in six mysticete species and AMELY in two mysticete species (Table S1). Among these genes, one or more deleterious (nonsense or frameshift) mutations were detected in AMELX in five species (Megaptera novaeangliae, Balaenoptera physalus, E. robustus, Balaenoptera bonaerensis, and Balaenoptera acutorostrata scammoni) and AMELY in both species (M. novaeangliae and E. robustus). These deleterious mutations include the same nonsense mutation in exon 5 of AMELX in all five species and the same frameshift mutation in exon 3 of AMELY in both species (Fig. S1). In Balaena mysticetus AMELX, no deleterious mutation was detected, but the splice donor of exon 2 was substituted by AT (Fig. S1). We also examined the nucleotide sequence of exon 6 (the largest protein-coding exon; Fig. S1) of both AMELX and AMELY (hereafter, exon 6X and exon 6Y, respectively) in Balaenoptera musculus and Eubalaena japonica, and detected a different frameshift mutation in exon 6Y in each species but no apparently deleterious mutation in exon 6X (Fig. S1). Other exons need to be analyzed to determine the presence or absence of deleterious mutations in AMELX in these two species.

We then investigated the entire protein-coding sequence of AMELX in 18 odontocete species and AMELY in three odontocete species (Table S1; exon 7 unidentified for AMELY). In AMELX, no apparently deleterious mutation was detected in 17 species (T. truncatus, T. aduncus, Sousa sahulensis, S. chinensis, Lagenorhynchus obliquidens, O. orca, P. dalli, P. phocoena, N. a. asiaeorientalis, Delphinapterus leucas, M. monoceros, L. vexillifer, P. blainvillei, Z. cavirostris, M. bidens, P. minor, and Physeter macrocephalus). The splice donor of exon 2 was, however, substituted by AT in all four X chromosomes (two females) of D. leucas we studied (Fig. S2) and in the assembled genome sequence of two M. monoceros females (Table S1). Furthermore, the splice acceptor of exon 6 was substituted by AA in one of 24 X chromosomes (12 females) of P. macrocephalus (Fig. S3). In K. breviceps AMELX, a frameshift mutation and a loss (substitution) of the initiation codon were detected in exon 2 (Fig. S1). For P. dalli, expression of AMELX was detected in a tooth germ by RT-PCR. In AMELY, one or more deleterious mutations were detected in all three species (T. truncatus, O. orca, and N. a. asiaeorientalis). We also examined the nucleotide sequence of exon 6X in Grampus griseus, Berardius bairdii, and Kogia sima, exon 6Y in P. dalli, Z. cavirostris, B. bairdii, K. sima, and P. macrocephalus, and exon 6 of AMEL (unidentified chromosomal origin) in Neophocaena phocaenoides. As a result, the same nonsense mutation was detected in P. dalli exon 6Y and N. phocaenoides exon 6, and a different nonsense mutation in Z. cavirostris exon 6Y (Figs. S1 and S4). Because N. phocaenoides has enamel (Ishiyama 1987), and because the same nonsense mutation was detected in exon 6Y of P. dalli (and also N. a. asiaeorientalis), we considered this N. phocaenoides exon to be exon 6Y. Although no other deleterious mutation was detected in exon 6X or exon 6Y, other exons may have deleterious mutations.

In exon 4 of AMELX, the same frameshift mutation was identified in ten species of Delphinidae, Phocoenidae, and Monodontidae (Figs. 1 and S1). We, however, do not consider this a deleterious mutation, because exon 4 of AMELX is not functional in some mammals, including B. taurus and Sus scrofa (Yuan et al. 1996; Hu et al. 2002). In the present analysis, we detected no deleterious mutations common to both odontocetes and mysticetes in either AMELX or AMELY. Furthermore, we did not identify apparently functional AMELY in mysticetes or odontocetes. These results suggest that both AMELX and AMELY were functional in common ancestors of extant cetaceans, and that AMELY became a pseudogene independently in mysticetes and odontocetes.

Unusually Small Core Region of Cetacean AMELX and AMELY Proteins

The mature AMEL protein consists of the N-terminal Tyrrich region, the Pro/Gln-rich hydrophobic core region, and the C-terminal hydrophilic region (Fincham and Simmer 1997). In mammals, both the N-terminal and the C-terminal regions are relatively well conserved in size and sequence, whereas the core region varies considerably, especially in the PXY repeat region (repeats of a Pro residue in every three residues; Fig. S5) (Toyosawa et al. 1998; Delgado et al. 2005, 2008; Jin et al. 2009). For this reason, we investigated sequence variations of the core region, which is entirely encoded by exon 6 (Fig. S5). In artiodactyls, the size of the core region ranges from 111 residues (C. hircus AMELY; Fig. 2a) to 143 residues (Giraffa camelopardalis and Okapia johnstoni AMELX proteins; Table S2). In cetaceans, by contrast, the size of the core region is invariable and unusually small (95 residues; genes containing frameshift mutations were excluded from this analysis, while a premature termination codon was counted as one amino acid), with the only exception being O. orca AMELX (94 residues; Table S2), which has the smallest core region in all mammals studied to date (Delgado et al. 2007; Jin et al. 2009; Bai et al. 2016). The largest core region of cetacean AMELX proteins is, therefore, 16 residues smaller than the smallest core region of artiodactyl AMELX and AMELY proteins (Fig. 2a).

Fig. 2.

Characteristics of the core region encoded by AMELX and AMELY. The X axis represents the size (the number of amino acid residues; AA in a), the average disorder prediction (PONDR) score (b), the number of amino acid residues predicted to adopt disordered conformations (AA in c), and the disorder index (DI in d). The Y axis shows the percentage of AMELX genes (%AMELX) in cetaceans (a) or odontocetes (b–d; see the text) and in artiodactyls (a–d) above the X axis, and the percentage of AMELY genes (%AMELY) in cetaceans (a) or odontocetes (b–d) and in artiodactyls (a–d) below the X axis. The percentage of AMELX and AMELY genes in cetaceans (or odontocetes) and 10 different families of artiodactyls are shown by bars in different colors, and the ranges of these characteristics in cetaceans (odontocetes) and artiodactyls are shown by dotted brackets. The largest value of each characteristic in cetaceans (odontocetes) and the smallest value of each characteristic in artiodactyls are indicated. The number of samples used to calculate these characteristics is also shown (e.g., N = 53 for artiodactyl AMELX genes). See Table S2 for the raw data and the correspondence between colors of the bars and names of the families

A large portion of the Pro/Gln-rich core region was shown to adopt intrinsically disordered conformations equilibrated with transient polyproline type-II (PPII) structures (Delak et al. 2009; Moradian-Oldak and Lakshminarayanan 2010). Intrinsically disordered regions lack globular tertiary folding but may fold upon binding to target molecules (Uversky 2002), and such labile conformations of the core region are thought to be important for enamel matrix assembly and enamel crystal growth (Delak et al. 2009; Moradian-Oldak and Lakshminarayanan 2010; Kalmar et al. 2012). We therefore calculated the propensity to disordered conformation of the core region, provided by the average disorder prediction (PONDR) score (Romero et al. 2001). In the following analysis, we used only odontocete AMELX genes for cetaceans, because mutations occurred after pseudogenization are irrelevant to the conformation of encoded proteins. For the same reason, AMELX of K. breviceps (deleterious mutations in exon 2) was removed from the analysis. In odontocetes, the average PONDR score in AMELX proteins ranges from 0.73 to 0.81 (Fig. 2b and Table S2). In artiodactyls, by contrast, the average PONDR score in AMELX proteins ranges from 0.81 to 0.91, and that in AMELY proteins from 0.76 to 0.91. With the notable exception of H. amphibius AMELY (0.76; Fig. 2b), the average PONDR score in artiodactyl AMELX and AMELY proteins is higher than or equal to the largest average PONDR score in odontocete AMELX proteins.

Since both the total size and the average PONDR score of the core region are smaller in odontocete AMELX proteins than in most artiodactyl AMELX and AMELY proteins (three exceptions in the average PONDR score; Fig. 2b), the total number of residues that are predicted to adopt disordered conformations is also smaller in odontocete AMELX proteins than in artiodactyl AMELX and AMELY proteins (Fig. 2c). The total number of disordered residues was calculated as the number of residues that showed the PONDR score greater than the threshold of 0.50. We assumed, however, that the product of the total length and the average PONDR score (disorder index, DI) may more accurately correlate with the degree of disordered conformation of the core region, because the DI value more directly reflects the disorder propensity without using the threshold. The DI values are lower in odontocete AMELX proteins (77.13 or lower) than in artiodactyl AMELX and AMELY proteins (87.95 for H. amphibius AMELY or higher; Fig. 2d). In our study, DI values were obtained from both AMELX and AMELY for six artiodactyl families. In all these families, the highest DI value in AMELY proteins is lower than or equal to the lowest DI value in AMELX proteins (Fig. 2d; Table S2). Moreover, in Bovidae and Cervidae, for which the DI value was obtained from both AMELX and AMELY for 16 and seven species, respectively, the highest DI value in AMELY proteins (106.27 for Bovidae and 100.40 for Cervidae) is considerably lower than the lowest DI value in AMELX proteins (117.60 for Bovidae and 111.64 for Cervidae). The finding of the lower DI values in AMELY proteins than in AMELX proteins in both Bovidae and Cervidae is consistent with the idea that, whereas AMELX is maintained by purifying selection, AMELY might be decaying (Wilson and Makova 2009).

In most cetaceans, the core region is composed of 95 residues in both AMELX and AMELY (Fig. 2a). This unusually small core region common to AMELX and AMELY suggests that exons 6X and 6Y, encoding the entire core region, underwent genetic exchange, most likely gene conversion (see below), in common ancestors of extant cetaceans. Since the core region, which is the most variable region within the AMEL protein, is similar in cetacean AMELX and AMELY proteins, gene conversion between exons 6X and 6Y appears to have considerably impacted upon the evolution of AMELX and AMELY in cetaceans.

Gene Conversion Between AMELX and AMELY in Intron 5 and Exon 6

It was reported that an interstitial insertion is present within intron 5 of AMELY in both ruminants and cetaceans, but not in other mammals (Macé and Crouau-Roy 2008). We identified this insertion as a long terminal repeat (LTR) present in AMELY, but not in AMELX, of B. taurus, C. hircus, H. amphibius, and all four mysticete species we studied (Fig. S6). In odontocetes, however, the exact same LTR was detected in all 24 AMEL genes, including 18 AMELX genes, that we studied (Fig. S6). Based on this finding, we hypothesized that the LTR and its flanking regions were unidirectionally transferred from AMELY to AMELX by gene conversion in a common ancestor of odontocetes after their divergence from mysticetes (Fig. 3a).

Fig. 3.

Gene conversion of the LTR sequence from AMELY to AMELX (a) and phylogenetic trees of the nucleotide sequences of the LTR region (b), the LTRup region (c), and exon 6 (d) of AMEL genes. (a) An LTR sequence was inserted in intron 5 of AMELY in a common ancestor of Ruminantia, Hippopotamidae, and Cetacea, and this LTR was presumably transferred from AMELY to AMELX by gene conversion in a common ancestor of odontocetes. (b–d) Nodes supported by 95% or higher bootstrap values are shown with a closed circle and 75–95% with an open circle. A scale is shown at the bottom of each tree. (c) The node for the two clusters (1) one composed of Mysticeti, Hippopotamidae, and Ruminantia AMELX sequences and (2) the other composed of Mysticeti, Hippopotamidae, and Ruminantia AMELY sequences and Odontoceti AMELX and AMELY sequences, is supported by a bootstrap value of 57%. (d) DPX represents the cluster consisting of Delphinidae, Phocoenidae, M. monoceros, D. leucas, L. vexillifer, and P. blainvillei exon 6X sequences, while DPY represents the cluster comprising Delphinidae and Phocoenidae exon 6Y sequences. The LTRdown (see Fig. 4) sequences are too small and were not used in our phylogenetic analysis

To test this hypothesis, we investigated sequences of this LTR and its flanking regions (LTRup and exon 6; Fig. 3a) by constructing phylogenetic trees. As a result, the LTR sequences on Y chromosomes (LTR-Y) reproduced the phylogeny of ruminants, hippopotamuses, mysticetes, and odontocetes (Fig. 3b). Furthermore, all LTR sequences on X chromosomes (LTR-X) of odontocetes were incorporated in the odontocete LTR-Y cluster. This result supports our scenario of gene conversion; an LTR was inserted into the Y chromosome in a common ancestor of ruminants, hippopotamuses, and cetaceans, and this LTR was subsequently transferred to the X chromosome in a common ancestor of odontocetes (Fig. 3a).

In our tree of the LTRup region, all ruminant, hippopotamus, and mysticete X-chromosome (LTRup-X) sequences and all their Y-chromosome (LTRup-Y) sequences formed two separate clusters (Fig. 3c). While all odontocete LTRup-Y sequences were clustered with all the other LTRup-Y sequences, all odontocete LTRup-X sequences were also clustered with LTRup-Y sequences. This topology suggests that LTRup-X and LTRup-Y initiated differentiation in a common ancestor of ruminants, hippopotamuses, and cetaceans. Subsequently, in odontocetes, a considerable portion of the LTRup sequence was transferred from the Y chromosome to the X by gene conversion probably along with the LTR sequence. In B. bairdii, sequences of LTRup- X and LTRup-Y, as well as LTR-X and LTR-Y, are closely related to each other (Fig. 3b and c), which can be explained by more recent gene conversion, probably involving both the LTRup and LTR regions.

In our tree of exon 6, all cetacean sequences and all ruminant and hippopotamus sequences formed two separate clusters (Fig. 3d). This topology implies that exon 6X and 6Y sequences were similar to each other in the MRCA of extant cetaceans, which can be explained by gene conversion between exons 6X and 6Y in common ancestors of extant cetaceans. Based on this assumption, we reconstructed the nucleotide sequence of the putative exon 6, common to exons 6X and 6Y, in the MRCA of extant cetaceans (Fig. S7). The encoded amino acid sequence differs from the sequence encoded by exon 6X of P. minor by only one residue (an Ilu residue substituted by a Val residue in P. minor; Fig. S7). Because the core region contains the most variable sequence in the AMEL proteins, this result implies that amino acid sequences encoded by AMELX and AMELY have been largely unchanged from the MRCA of odontocetes and mysticetes to their extant descendants.

In all four mysticete species (B. musculus, M. novaeangliae, E. robustus, and E. japonica) and in all four odontocete species (B. bairdii, Z. cavirostris, K. sima, and P. macrocephalus), for which sequences of both exons 6X and 6Y are available, exon 6X sequences are closely related to the exon 6Y sequence of the same species (Fig. 3d). Notably, Z. cavirostris exon 6X and 6Y sequences differ by only one nucleotide (Figs. S1 and S4). These findings imply that exons 6X and 6Y underwent recent gene conversion, even though purifying selection on both AMELX and AMELY must have been relaxed in mysticetes. In Delphinidae and Phocoenidae (Fig. 1), exon 6X and 6Y sequences formed separate clusters (DPX and DPY in Fig. 3d), and the DPX cluster included D. leucas, M. monoceros, L. vexillifer, and P. blainvillei exon 6X sequences. This topology can be explained by gene conversion between exons 6X and 6Y in a common ancestor of Delphinidae, Phocoenidae, Monodontidae, L. vexillifer, and P. blainvillei (Fig. 1).

Frequent Gene Conversion Events in Intron 5 and Exon 6

Gene conversion progresses as mismatch repair processes, involving double-strand break and heteroduplex formation of the repaired strand with the template strand, which results in unidirectional genetic exchange (Duret and Galtier 2009) and differs from mutations that occur during replication. In exon 6, we detected 45 sites (50 substitutions) that potentially underwent gene conversion between AMELX and AMELY (Fig. 4). For example, the nucleotide at position 006 of both exons 6X and 6Y (the third column below exon 6 in Fig. 4) is G in all cetaceans, but T in artiodactyls. This substitution pattern can be explained by a T to G mutation in either exon 6X or 6Y and a subsequent transfer of this G to its gametolog by gene conversion in common ancestors of cetaceans. Similarly, the nucleotide at position 333 of both exons 6X and 6Y is C in odontocetes, but A in mysticetes and artiodactyls (the last column below exon 6 in Fig. 4). This substitution pattern can be explained by an A to C mutation and a following transfer of this C to its gametolog by gene conversion in common ancestors of odontocetes. Although these substitutions can be also explained by two independent mutations at the same position, it is unlikely that the substitution at all these 45 sites are results of coincidental recurrent mutations. Among these 45 sites, 26 were detected in E. japonica, which is consistent with an especially close relationship between exon 6X and 6Y sequences of E. japonica (Fig. 3d).

Fig. 4.

Nucleotides potentially underwent gene conversion in cetaceans. The four regions used in this analysis (LTRup, LTR, LTRdown, and exon 6) and their lengths (nucleotides) are illustrated on the top. Each column below shows the position (number) within the region and the nucleotides (A, G, C, or T; a hyphen represents a gap in the sequence alignment) at the position in various cetacean and artiodactyl AMELX and AMELY genes. Substitutions possibly involved in gene conversion are highlighted in yellow. Deletions extending positions 183–184 and 216–245 in the LTR region (highlighted in sky blue) were considered to be two discrete sites, because each site probably underwent a single gene conversion event. The timing of each putative gene conversion event (Timing) can be assigned to three most likely periods: (1) in common ancestors of odontocetes and mysticetes (2) in the odontocete lineage, and (3) in the mysticete lineage, or a combination of two events at different timings (2/3 or 1/2). For each position in exon 6, whether the putative gene conversion is associated with a synonymous substitution (+) or a non-synonymous substitution (−) is shown in the “AA change” row. Complete multiple nucleotide sequence alignments are provided in Fig. S1 for exon 6 and Fig. S6 for the LTRup, LTR, and LTRdown regions. Substitutions from A or T to G or C, and those from G or C to A or T are shown by “X” in the “A/T to G/C” and “G/C to A/T” rows, respectively, while substitutions between G and C and between A and T are shown with “X” in the “Btw G&C or A&T” row

It has been proposed that gene conversion increases GC contents, similar to recombination (Marais and Galtier 2003; Duret and Galtier 2009). In accordance with this, we detected an increase in the GC content: 28 substitutions from A/T to G/C, compared to 14 substitutions from G/C to A/T and eight substitutions between G and C or between A and T (among 50 substitutions in Fig. 4). This observation agrees with the idea that a large portion of these substitutions is explained by gene conversion (see below for the reason that excludes recombination).

The presence of the LTR in intron 5 of odontocete AMELX genes can be explained by a single gene conversion event. After this event, 17 sites in the LTR region potentially underwent further gene conversion in various odontocete lineages (Fig. 4). Among these 17 sites, twelve were detected only in B. bairdii. In the LTRup region, we detected 15 sites potentially involved in gene conversion, of which nine were found only in B. bairdii (Fig. 4). In fact, 22 putative gene conversion sites, detected exclusively in B. bairdii, are concentrated in the 668-nucleotide region, extending from position 007 of the LTRup region to position 011 of the LTRdown region (between the LTR and exon 6 in Fig. 4). This finding suggests that a relatively large gene conversion occurred in a recent ancestor of B. bairdii, which is in line with the close relationship between the X and Y sequences of the LTR and LTRup regions in B. bairdii (Figs. 3b and 3c). In the entire intron 5, we detected 32 substitutions that potentially underwent gene conversion, and these substitutions include 25 A/T to G/C substitutions (Fig. 4). This finding is also consistent with GC-biased gene conversion (Duret and Galtier 2009).

Discussion

Mutations in Odontocete AMELX Genes

In this study, various deleterious mutations were detected in mysticete AMELX and AMELY genes and odontocete AMELY genes (Table S1). This result is expected, because mysticetes do not possess teeth, and because AMELY is thought to be decaying. Among odontocete AMELX genes, K. breviceps AMELX is presumably non-functional, because of the frameshift mutation and the loss of the initiation codon detected in exon 2. In K. breviceps, two frameshift mutations were also detected in a different enamel matrix protein gene, ENAM (Meredith et al. 2009) (Fig. S8a). The finding of these mutations in K. breviceps AMELX and ENAM is in line with the observation that adult K. breviceps do not possess enamel on teeth, although some young individuals may grow enamel (Plön 2004).

In D. leucas and M. monoceros, the two extant species of Monodontidae, no apparently deleterious mutation was detected. In fact, the amino acid sequences encoded by D. leucas and M. monoceros AMELX genes retain all 34 residues that are highly conserved in 80 amniote species (Fig. S9) (Delgado et al. 2007). The splice donor of exon 2 was, however, substituted by AT (Fig. S2) in both genes. The same substitution was also detected in B. mysticetus AMELX but presumably occurred independently, because this mutation was identified only in AMELX genes of two phylogenetically separated taxa (Monodontidae and B. mysticetus in Fig. 1). In human genes, non-canonical AT-AG splice junctions are uncommon (26 sites) and account for only 0.0117% of the entire splice junctions (Parada et al. 2014). Among these 26 AT-AG splice junctions, six were classified as U2-like and 20 as U12-like splice sites, which are processed by distinct spliceosome machineries and have different adjacent consensus sequences (Sibley et al. 2016). Because the regions adjacent to the AT-splice donors of D. leucas, M. monoceros, and B. mysticetus AMELX genes retain the original sequences (Fig. S1), we considered these sites to be U2-like splice sites. This situation is similar to an artificially introduced AT-splice donor in the β-globin gene, which drastically reduced correct splice products and increased products that lack the upstream exon (Aebi et al. 1986). It is thus likely that, in D. leucas, M. monoceros, and B. mysticetus, the levels of correct transcripts of AMELX genes, and hence correct AMELX proteins, are much lower than those in their ancestor that had the canonical splice junction.

Whereas this non-canonical AT-AG splice junction does not affect B. mysticetus that does not possess teeth (Thewissen et al. 2017), it apparently does affect D. leucas that grows teeth covered with enamel. This enamel is, however, prismless and unusually thin, 7–10 μm in thickness, and is worn out soon after eruption (Ishiyama 1987). In fact, AMELX-knockout mice develop prismless hypoplastic enamel, a phenotype common to amelogenesis imperfecta (Gibson et al. 2001). Notably, this hypoplastic enamel is 20.3 ± 3.3 μm in thickness (Hu et al. 2016), which is thicker than the enamel found in D. leucas. Furthermore, in D. leucas, the other two enamel matrix genes, AMBN and ENAM, are apparently devoid of defects in their protein-coding sequences and splice junctions (Fig. S8b). We thus consider the D. leucas phenotype to be similar to amelogenesis imperfecta, resulting from the absence or a low level of correct AMELX transcripts. Unlike D. leucas, male M. monoceros erupts a single tooth (tusk), on which no enamel was detected (Ishiyama 1987). Consistent with this observation, a nonsense mutation was identified in both AMBN and ENAM in M. monoceros (Fig. S8c). This finding implies that these nonsense mutations occurred in the M. monoceros lineage after the mutation in the splice junction of AMELX occurred in a common ancestor of D. leucas and M. monoceros. M. monoceros does not grow enamel probably because of the mutation in AMBN and/or ENAM.

We also found that the splice junction between exon 5 and exon 6 of P. macrocephalus AMELX is non-canonical (GT-AA) in one of 24 X chromosomes (Fig. S3). In human genes, only nine GT-AA splice junctions have been identified (Parada et al. 2014). Furthermore, both AMBN and ENAM are apparently functional in P. macrocephalus (Fig. S8d). Enamel of P. macrocephalus is prismless and varies in the thickness of individual teeth, implying degeneration of enamel (Ishiyama 1987). We assessed sequence changes of exon 6X during the evolution of P. macrocephalus from the MRCA of odontocetes and mysticetes. Results showed that the number of synonymous differences per site (dS) is smaller than that of non-synonymous differences per site (dN; dS-dN = − 0.017 ± 0.008; Fig. S7), which we interpret as a relaxed purifying selection on exon 6X. It seems likely that the polymorphic GT-AA splice junction reflects a relaxed purifying selection on AMELX, and that a small population of P. macrocephalus shows an amelogenesis imperfecta-like phenotype.

Evolution of AMELX and AMELY and the Impact of Gene Conversion

While cetartiodactyl AMELY genes reside in the MSY that does not undergo recombination (Raudsepp and Chowdhary 2015), whether AMELY was located in the MSY or the PAR in their MRCA remains unclear (Bellott et al. 2014; Cortez et al. 2014). It is, however, unlikely that the LTR in intron 5 was inserted in the PAR, because, if so, this LTR must have been maintained as a polymorphism until this region became the MSY in ruminants and also in the lineage that led to hippopotamuses and cetaceans so that this LTR was independently fixed on the Y but not on the X chromosome. Instead, this LTR was presumably inserted in the MSY in their common ancestor. Subsequently, in odontocetes, this LTR was transferred to the X chromosome by gene conversion (Fig. 3a). This scenario is consistent with our results in the LTRup region; the topology of the LTRup tree (Fig. 3c) suggests that the LTRup-Y sequence was located in the MSY and initiated differentiation from the LTRup-X sequence in common ancestors of ruminants, hippopotamuses, and cetaceans, and that a considerable portion of this region was transferred from the Y to the X chromosome by gene conversion in odontocetes.

We also identified 15 sites that potentially underwent gene conversion in the LTRup region (Fig. 4). Gene conversion reduces the evolutionary distance between the two sequences involved and obscures the evidence of early differentiation (Marais and Galtier 2003). In our phylogenetic analysis, the divergence of the LTRup-X and LTRup-Y sequences in common ancestors of ruminants, hippopotamuses, and cetaceans is supported by a bootstrap value of 57% (Fig. 3c). We suspect that this relatively low statistical support is largely due to gene conversion. Indeed, our analysis of exon 7 significantly supported the divergence of AMELX and AMELY sequences in common ancestors of ruminants, hippopotamuses, and cetaceans (99% bootstrap value; Fig. S10). Because both upstream and downstream regions of exon 6 initiated differentiation presumably much earlier than the divergence of odontocetes and mysticetes, the close relationship between exon 6X and 6Y sequences in cetaceans (Fig. 3d) implies a significant impact of gene conversion upon the evolution of their exon 6.

In exon 6 of cetacean AMEL genes, we detected 45 sites that potentially underwent gene conversion, among which 28 are associated with non-synonymous substitutions (Fig. 4). However, the largest impact of gene conversion is presumably upon the unusually small size of the core region, which could be regarded as degeneration of AMEL (see below). A previous study also showed evidence of gene conversion in this region of AMELX and AMELY in S. scrofa (Marais and Galtier 2003), but the biological consequence of gene conversion is unclear in this case. In ruminants and H. amphibius, by contrast, exon 6X and 6Y sequences differ considerably (Table S2), which suggests that gene conversion was less important for the evolution of their core region.

Evolution of the AMEL Gene and Feeding Strategy in Cetaceans

Phylogenetic differences of AMEL sequences largely reside in the core region, especially in the number of PXY repeats. While the PXY repeat has been associated with disordered conformations and transient PPII structures, these conformations or structures are not exclusively attributable to PXY repeat units (Delak et al. 2009; Moradian-Oldak and Lakshminarayanan 2010). We therefore considered that this region could be better characterized by the DI value, which reflects both the size and the disorder propensity, rather than the number of PXY repeats. It was shown that an increase in the number of PXY repeats (the DI value) augments the radius of the self-assembled AMEL nanosphere and the length of enamel crystals (Jin et al. 2009; Lacruz et al. 2011). In fact, a low number of PXY repeats is known from the Japanese striped snake (DI = 66.49; Table S2), which grows a prismless enamel in 1–2 μm thickness, extremely thin among reptiles (Ishiyama et al. 1998). Furthermore, the number of PXY repeats is low in amphibians, which possess prismless enamel (Jin et al. 2009). It is thus possible that AMEL with an unusually low DI value exhibits a low assembling efficiency of the enamel matrix and/or a slow growth of enamel crystals, which may not be able to develop a thick enamel, and we could consider AMEL proteins showing the DI value decreased from the ancestral state to be degenerated.

In our analysis, the highest DI value in odontocete AMELX proteins (DI = 77.13 in B. bairdii) is considerably lower than the lowest DI value in artiodactyl AMELX and AMELY proteins (DI = 87.95 in H. amphibius AMELY; Fig. 2d). In fact, the DI value in H. amphibius AMELY is markedly low among artiodactyl AMELX and AMELY proteins we studied (the second smallest DI = 93.60 in Cephalophus harveyi AMELY), while the DI value in H. amphibius AMELX (DI = 105.33) is not significantly low among the DI value in artiodactyl AMELX proteins (Table S2). Furthermore, DI values in extant odontocete AMELX proteins do not differ from the DI value in putative AMELX and AMELY in the MRCA of extant cetaceans (DI = 74.81; Table S2). It is thus conceivable that the DI value initiated to decrease in AMELY in common ancestors of H. amphibius and cetaceans and that unusually low DI values emerged in both AMELX and AMELY in common ancestors of extant cetaceans. Subsequently, the DI value has been largely unchanged in functional AMELX proteins in the odontocete lineage.

The secondary loss of teeth in mysticetes has been associated with pseudogenization of various genes primarily involved in enamel formation, including AMBN, ENAM, AMTN, KLK4, and MMP20 (Deméré et al. 2008; Meredith et al. 2009, 2011; Kawasaki et al. 2014; Springer et al. 2016). Pseudogenization of mysticete AMEL genes, however, has not been well understood, probably because it is difficult to attribute deleterious mutations to either AMELX or AMELY. Analysis of odontocete AMEL genes may be even more complicated, because most odontocetes maintain teeth, but capture prey by suction and swallow the prey whole without biting or chewing with teeth (Werth 2000, 2006). The fact that suction feeders do not use teeth predominantly for feeding could result in reduction or loss of enamel in some odontocetes (Werth et al. 2019). In fact, D. leucas, M. monoceros, K. breviceps, and P. macrocephalus are known as such suction feeders (Werth 2000, 2004), which appears to explain the frameshift mutation and the loss of the initiation codon in K. breviceps AMELX and the presence of non-canonical splice junctions in D. leucas, M. monoceros, and P. macrocephalus AMELX genes. Functional AMEL genes are presumably not essential for these suction feeders.

Although not all odontocetes are suction feeders, the origin of suction feeding may be traced back to archaeocetes (stem cetaceans) (Johnston and Berta 2011; Marx et al. 2016). Archaeocetes possessed sharp and multicusped teeth and are considered raptorial feeders that capture and masticate prey with teeth (Hocking et al. 2017; Uhen 2018). Even for raptorial feeders, however, suction would be, to some extent, necessary underwater for transport of prey deep into the oral cavity for swallowing (Werth 2000, 2006). It was argued that such suction-assisted raptorial feeding evolved in cetaceans soon after their initial transition to an aquatic environment (Marx et al. 2015, 2016), and that this ancestral feeding strategy evolved into capture suction feeding that uses suction to capture prey likely by the emergence of the MRCA of odontocetes and mysticetes (Johnston and Berta 2011). Although the emergence of suction feeding in stem cetaceans remains controversial (Lindberg and Pyenson 2007; Fahlke et al. 2013; Peredo et al. 2017), an early transition from an intensive raptorial feeding to a higher degree of suction feeding is consistent with the finding of presumed suction feeders in both stem odontocetes and stem mysticetes (Fordyce 2002; Marx et al. 2015; Boessenecker et al. 2017; Fordyce and Marx 2018). Progressive suction feeding in archaeocetes and then in both stem odontocetes and stem mysticetes is also suggested by decreases of mastication, detected by gradual loss of their precise occlusion and evolution of pointed teeth (Werth 2000; Thewissen et al. 2011; Armfield et al. 2013; Peredo et al. 2018).

The decrease in the DI value in extant cetaceans could be explained if some degree of suction feeding evolved before the divergence of odontocetes and mysticetes. A slightly relaxed selective pressure upon tooth wearing by prey capture and/or mastication in suction-assisted raptorial feeders could have compromised the function of enamel to some extent and allowed the use of AMEL with a decreased DI value. Gene conversion can transfer accumulated multiple mutations on a decaying AMELY to an AMELX that undergoes purifying selection. We speculate that common ancestors of modern cetaceans employed a degenerated AMELX, transferred from a decaying AMELY by gene conversion, at an early stage of their transition to suction feeders.