Parallel Evolution of Ameloblastic scpp Genes in Bony and Cartilaginous Vertebrates.

In bony vertebrates, skeletal mineralization relies on the secretory calcium-binding phosphoproteins (Scpp) family whose members are acidic extracellular proteins posttranslationally regulated by the Fam20°C kinase. As scpp genes are absent from the elephant shark genome, they are currently thought to be specific to bony fishes (osteichthyans). Here, we report a scpp gene present in elasmobranchs (sharks and rays) that evolved from local tandem duplication of sparc-L 5' exons and show that both genes experienced recent gene conversion in sharks. The elasmobranch scpp is remarkably similar to the osteichthyan scpp members as they share syntenic and gene structure features, code for a conserved signal peptide, tyrosine-rich and aspartate/glutamate-rich regions, and harbor putative Fam20°C phosphorylation sites. In addition, the catshark scpp is coexpressed with sparc-L and fam20°C in tooth and scale ameloblasts, similarly to some osteichthyan scpp genes. Despite these strong similarities, molecular clock and phylogenetic data demonstrate that the elasmobranch scpp gene originated independently from the osteichthyan scpp gene family. Our study reveals convergent events at the sparc-L locus in the two sister clades of jawed vertebrates, leading to parallel diversification of the skeletal biomineralization toolkit. The molecular evolution of sparc-L and its coexpression with fam20°C in catshark ameloblasts provides a unifying genetic basis that suggests that all convergent scpp duplicates inherited similar features from their sparc-L precursor. This conclusion supports a single origin for the hypermineralized outer odontode layer as produced by an ancestral developmental process performed by Sparc-L, implying the homology of the enamel and enameloid tissues in all vertebrates.

Introduction

The secretory calcium-binding phosphoproteins (Scpp) family comprises extracellular proteins playing pivotal roles during bony vertebrate (osteichthyan) skeletal mineralization (Kawasaki and Weiss 2003; Kawasaki 2011). The scpp genes are organized into genomic clusters coding for acidic proteins involved in enamel, bone, and dentin matrix mineralization (Nikoloudaki 2021; Ustriyana et al. 2021). Bony fish scpp genes are grouped in a single family from their shared intron–exon structure rather than from shared sequence features as these have quickly diverged and are mostly nonalignable (Kawasaki and Weiss 2003). Notwithstanding, distinct Scpp members expressed in ameloblasts, osteoblasts, and odontoblasts share a common functional feature based on their posttranslational regulation by Fam20°C kinase phosphorylation (Tagliabracci et al. 2012; Wang et al. 2013; Liu et al. 2017; Zhang et al. 2018; Schytte et al. 2020; Shin et al. 2020).

The evolutionary history of the scpp gene family has been the focus of intense research with regard to the evolution of the vertebrate mineralized skeleton (Kawasaki and Weiss 2003; Kawasaki et al. 2005, 2017, 2021; Kawasaki 2009, 2011; Kawasaki and Amemiya 2014; Venkatesh, Lee, Ravi, Maurya, Korzh, et al. 2014; Venkatesh, Lee, Ravi, Maurya, Lian, et al. 2014; Qu et al. 2015; Braasch et al. 2016; Lin et al. 2016; Lv et al. 2017; Cheng et al. 2021). The two rounds of whole genome duplication in the vertebrate lineage produced the sparc and sparc-L paralogs (Kawasaki and Weiss 2003; Kawasaki et al. 2005; Kawasaki 2009; Bertrand et al. 2013; Enault et al. 2018). In the osteichthyan lineage, sparc-L locally duplicated into sparc-L1 and sparc-L2 (Bertrand et al. 2013; Enault et al. 2018), and multiple duplications and losses at this locus also produced a broad array of scpp members (Kawasaki et al. 2005; Kawasaki 2009, 2011). Some of these duplications occurred concomitantly with the origin of mammals and are associated with evolutionary innovation, as is the case for the milk Caseins and the saliva Muc7 gene (Kawasaki et al. 2011; Xu et al. 2016; Luis Villanueva-Cañas et al. 2017). Genome projects have identified scpp genes at the sparc-L1/sparc-L2 locus in all osteichthyan species examined to date, although with poor support of 1-to-1 orthology between teleosts and tetrapods, reflecting a high rate of sequence divergence and gene turnover (Kawasaki and Amemiya 2014; Qu et al. 2015; Braasch et al. 2016; Lin et al. 2016; Kawasaki et al. 2017; Lv et al. 2017; Cheng et al. 2021).

As careful examination of the elephant shark genome failed to identify any scpp gene around sparc-L (Venkatesh, Lee, Ravi, Maurya, Lian, et al. 2014), the scpp family was considered to be osteichthyan-specific (Venkatesh, Lee, Ravi, Maurya, Korzh, et al. 2014; Venkatesh, Lee, Ravi, Maurya, Lian, et al. 2014; Kawasaki et al. 2017, 2021), implying that chondrichthyans use a default mineralization genetic toolkit merely composed of Sparc and Sparc-L. This argument was also used as a support to the conclusion that chondrichthyan enameloid, the outer hypermineralized tissue of teeth and scales, is not homologous to the osteichthyan enamel (in tetrapods) nor ganoin (in actinopterygians, coelacanth, and lungfish) that critically rely on Scpp proteins for their mineralization (Qu et al. 2015; Kawasaki et al. 2017, 2021).

Chondrichthyans and stem jawed vertebrates, however, exhibit a rich diversity of mineralized skeletal tissues with similarity to those of bony fish (Ørvig 1951; Min and Janvier 1998; Donoghue and Sansom 2002; Donoghue et al. 2006; Johanson et al. 2012; Brazeau and Friedman 2015; Keating et al. 2018; Lemierre and Germain 2019; Brazeau et al. 2020; Berio et al. 2021), which seems at odds with the idea of a Sparc/Sparc-L default toolkit. To address this issue, it remains of utmost importance to better characterize the genetic basis of skeletal mineralization in chondrichthyans. By exploring high-quality genome assemblies, we show here for the first time that elasmobranchs (a chondrichthyan monophyletic group including sharks and rays) have independently evolved a scpp gene. Our data reveal that distinct mutational and evolutionary processes at the chondrichthyan sparc-L locus mimic many of the key features that are emblematic of osteichthyan scpp genes and illustrate how molecular parallelism has continuously enriched the vertebrate skeletal mineralization genetic repertoire. Taken together, our genomic and expression data provide a unifying genetic basis that supports the homology of developmental processes for enamel and enameloid in all jawed vertebrates.

Results

A Novel Elasmobranch scpp Gene

We identified a nonannotated gene at the sparc-L locus in high-quality genome assemblies of two elasmobranch species, the lesser spotted catshark Scyliorhinus canicula and the skate Amblyraja radiata (fig. 1). We named this gene scpp because it is genomic, structural, and expression features are remarkably similar to the osteichthyan scpp genes. The syntenic location of the elasmobranch scpp gene and its tail-to-tail orientation with respect to sparc-L are equivalent to the osteichthyan acidic scpp gene cluster (fig. 1; see Kawasaki 2009; Braasch et al. 2016). In addition, the intron/exon structure of the catshark and skate scpp genes is identical to the 5′ intron/exon structure of the nearby sparc-L gene, supporting its origin from partial local duplication of the sparc-L locus (fig. 1). Hence, the first three introns located in the open reading frame of the elasmobranch scpp are in phase 0 (fig. 1), which is the phase most commonly observed in osteichthyan scpp genes (fig. 1; see Kawasaki and Weiss 2003). A fourth intron is located in phase 1 in scpp and sparc-L (fig. 1). We searched genomic and transcriptomic data and identified a similar scpp gene in other elasmobranch species belonging to selachimorphs (the sharks Scyliorhinus torazame, Chiloscyllium punctatum, and Carcharodon carcharias) and batoids (the ray Raja clavata and the skate Okamejei kenojei, supplementary material S1, Supplementary Material online). In all cases, the encoded elasmobranch Scpp protein displays a signal peptide conserved with Sparc, Sparc-L/L1/L2, and the osteichthyan Scpp sequences, together with an enrichment of aspartate/glutamate (D/E) residues and the presence of conspicuous putative Fam20°C phosphorylation SxE sites (fig. 1 and table 1; see Kawasaki et al. 2005, 2007; Kawasaki 2009; Qu et al. 2015). Furthermore, similarly to the osteichthyan Scpps, the elasmobranch Scpp protein lacks the Kazal/calcium-binding domain located in the C-terminal region of Sparc and Sparc-L/L1/L2 (fig. 1 and table 1; see Kawasaki et al. 2005, 2007; Kawasaki 2009; Bertrand et al. 2013; Qu et al. 2015). The elasmobranch Scpp protein also includes a N-terminal tyrosine-rich (Y-rich) region absent from Sparc and Sparc-L/L1/L2 but present in a wide array of osteichthyan Scpps (fig. 1 and table 1; see Kawasaki and Amemiya 2014; Kawasaki et al. 2021). The prediction for any novel scpp gene is that it should be expressed at sites that are relevant for biomineralization during skeletal development. We, therefore, performed in situ hybridizations on developing catshark embryos and found that scpp is expressed in ameloblasts from both teeth and scales, specifically at the secretory stage (fig. 2, , , and ). By contrast, sparc-L and fam20°C are expressed in ameloblasts, both at the secretory and maturation stages, with an additional expression of fam20°C in odontoblasts (fig. 2, , , and ). The spatially restricted scpp expression to the catshark secretory ameloblasts is strikingly similar to the situation reported for a variety of osteichthyan acidic and P/Q rich scpp gene members (table 1; see MacDougall et al. 1998; Kawasaki 2009; Gasse and Sire 2015; Kawasaki et al. 2021). Taken together, genomic, structural, and expression features support the classification of this novel elasmobranch gene as a bona fide member of the scpp gene family, with similarities to both the acidic and the P/Q rich osteichthyan members (table 1).

Fig. 1.

Characterization of a novel scpp gene in shark and ray genomes. Synteny and intron–exon structure of the sparc-L and scpp loci for the indicated osteichthyan (Lepisosteus oculatus) and chondrichthyan species (the holocephalan Callorhinchus milii; the shark S. canicula; the ray A. radiata). For synteny, triangles show relative gene position and orientation. The osteichthyan scpp members are represented as whole P/Q-rich or acidic clusters and not individually. For each exon, a color code legend identifies the UTRs or the encoded protein domains. Numbers at the bottom right corner of each exon indicate the position of the last encoded amino-acid, and numbers between two consecutive coding exons indicate the translation phase. Putative Fam20°C phosphorylation sites (SxE) are shown along each sequence. All exons are drawn to scale, except for the last exon which is trimmed. The position of the sparc-L and scpp in situ hybridization probes is indicated.

Table 1.

Comparison of the Structure of the Scpp and Sparc-L/L1/L2 Proteins and of the Expression Patterns of Their Corresponding Genes.

	Osteichthyan acidic Scpps	Osteichthyan P/Q-rich Scpps	Elasmobranch Scpp	Sparc-L/-L1/-L2
				N-terminal Region	C-terminal Region
Signal peptide (A)	++	++	++	++	−
Presence of SxE sites (B)	++	++	++	++	−
Kazal and calcium-binding domains (C)	−	−	−	−	++
Y-rich domain (D)	+	+	++	−	−
D/E-rich domain (E)	++	−	++	++	−
P/Q-rich domain (F)	−	++	−	−	−
PDYXXPXX repeated motif and conserved internal domain (G)	−	−	++	+ (only in elasmobranch Sparc-L)	−
Ameloblastic expression (H)	+ (mouse Dpm1, zebrafish spp1)	++	++	+ (only for the elasmobranch sparc-L)
Odontoblastic expression (I)	++	+	−	−

note.—The table recapitulates the presence (++, in all or most examined members; and +, at least in some members, shown in green) or the absence (−) of structural and expression features for the indicated proteins and their corresponding genes. Sources: (A) this study and see Kawasaki et al. (2005, 2007); (B) this study and see Kawasaki et al. (2005, 2007); (C) this study and see Kawasaki et al. (2005, 2007), Kawasaki (2009), Bertrand et al. (2013); (D) this study and see Kawasaki and Amemiya (2014), Kawasaki et al. (2021); (E) this study and see Kawasaki et al. (2005, 2007), Kawasaki (2009), Qu et al. (2015); (F) Kawasaki and Amemiya (2014), Kawasaki et al. (2017); (G) this study; (H) In osteichthyans, P/Q-rich scpp members are typically expressed in ameloblasts (Kawasaki 2009). Similarly to the elasmobranch scpp gene, a specific expression at the early secretory stage has been reported for a variety of P/Q-rich scpps such as the lizard and salamander amtn gene (Gasse and Sire 2015) and the gar ambn, enam, and scpp5 genes (Kawasaki et al. 2021). Osteichthyan acidic scpps are typically associated with dentine and bone (Kawasaki 2009), but a specific expression in secretory ameloblasts has also been reported for zebrafish spp1 (Kawasaki 2009) and mouse Dmp1 (MacDougall et al. 1998). Although the mouse Sparc-L1 gene is not expressed in teeth, nor in any other skeletal tissue (Soderling et al. 1997), the elasmobranch sparc-L gene harbors a pan-ameloblastic expression (this study and see Enault et al. 2018); (I) this study and see Enault et al. (2018), Kawasaki et al. (2005, 2021), Kawasaki (2009), Yonekura et al. (2013).

Fig. 2.

The lesser spotted catshark scpp gene is coexpressed with sparc-L and fam20°C in tooth and scale ameloblasts. Sections were performed at the level of developing scales (A–D, 8 cm-long embryos) and teeth (E–H, 9.5 cm-long embryos). Hematoxylin-Eosin-Safran histological staining (A and E) shows cells in the epithelium (ameloblasts, am) and the mesenchyme (odontoblasts, od), as well as the mineralized extracellular matrix (m). White arrowheads indicate secretory ameloblasts (sec am) and black arrowheads indicate maturation-stage ameloblasts (mat am). In situ hybridization signal identifies cells expressing scpp (B and F), sparc-L (C and G), or fam20°C (D and H).

Scpp and Sparc-L Share Specific Sequence Features in Elasmobranchs

Although the signal peptide is the only similar domain between Sparc-L1/L2 and the Scpp proteins in osteichthyans (Kawasaki et al. 2005, 2007), the elasmobranch Sparc-L and Scpp sequences can be readily aligned and compared (fig. 1, supplementary materials S2 and S3, Supplementary Material online). Notably, elasmobranch Sparc-L and Scpp proteins both contain a variable 8 amino acid-long periodic motif whose consensus is PDYXXPXX and that is absent from the elephant shark Sparc-L (fig. 1 and table 1, supplementary materials S1, S3, and S4, Supplementary Material online) and from all the osteichthyan Sparc-L1/L2 proteins we examined (not shown). The elasmobranch Sparc-L and Scpp proteins also share a well-conserved internal region which is absent from the elephant shark Sparc-L and from the osteichthyan Sparc-L1/L2 homologues (fig. 1, supplementary material S3, Supplementary Material online). We envision three possible evolutionary scenarios for the evolution of the jawed vertebrate scpp genes, each of which is associated with a specific phylogenetic tree topology (fig. 3). According to Hypothesis 1, scpp genes in bony and cartilaginous fishes are orthologous and evolved from a single scpp duplication in all jawed vertebrates and a subsequent loss in holocephalans. Hypothesis 2 states that parallel duplication events generated scpp sequences in bony and cartilaginous fishes, followed by a scpp loss in holocephalans. Finally, according to Hypothesis 3, parallel events of duplication generated scpp sequences in bony and elasmobranch fishes. Alignments and phylogenetic analyses were performed with full-length Scpp and Sparc-L sequences from cartilaginous fishes and osteichthyan Sparc-L1/L2 sequences (supplementary materials S2 and S3, Supplementary Material online). The scpp sequences from two bony fishes (Lepisosteus oculatus and Erpetoichthys calabaricus) were too divergent to generate an informative alignment (supplementary material S5, Supplementary Material online). Our phylogenetic reconstruction shows that the elasmobranch Sparc-L and Scpp group together in a well-supported node with the elephant shark Sparc-L sequence as an outgroup (fig. 3). This result is consistent with an independent origin of the elasmobranch scpp gene by duplication of the sparc-L 5′ exons after the split from holocephalans, thereby supporting Hypothesis 3 (fig. 3).

Fig. 3.

Evolution of scpp from the sparc-L locus. (A) The cladograms represent three possible evolutionary scenarios: scpp genes in bony and cartilaginous fishes are orthologous and evolved from a single scpp duplication in all jawed vertebrates, followed by a scpp loss in holocephalans (Hypothesis 1); parallel duplication events generated scpp sequences in bony fishes and in cartilaginous fishes, followed by a scpp loss in holocephalans (Hypothesis 2); parallel duplication events generated scpp sequences in bony and elasmobranch fishes (Hypothesis 3). (B) Phylogenetic reconstruction of chondrichthyan Sparc-L and Scpp proteins. The phylogeny was performed with full-length Scpp and Sparc-L sequences and was rooted with the sea lamprey Sparc-L sequence (refer to supplementary material S14, Supplementary Material online for species names and accession numbers). The evolutionary model used was JTT + I + G4 and the alignment was 538 amino acid long (supplementary material S3, Supplementary Material online). Node support was evaluated with sh-lrt and 5000 ultra-fast bootstrap replicates, values are indicated on each node. Scpp proteins are shown in orange and red dots indicate bootstraps values superior to 95.

The Elasmobranch scpp Gene Experienced Nonallelic Gene Conversion with sparc-L Specifically in Selachimorphs

The phylogenetic analysis using full-length proteins shows that the Sparc-L and Scpp sequences consistently group together as species pairs in the shark clade (fig. 3), a phenomenon reminiscent of concerted evolution due to nonallelic gene conversion between paralogs (Zhou et al. 2019). From this tree topology, it is impossible to deduce if the scpp gene evolved once at the base of the elasmobranchs, or if it results from independent duplications in selachimorphs and batoids. Performing a stringent genomic alignment reveals that gene conversion has occurred in four discrete genomic regions shared between the catshark sparc-L and scpp genes that encompass the coding exons 1 and 2, and overlaps with the end of the coding exon 3 (fig. 4, supplementary material S6, Supplementary Material online). Accordingly, a phylogenetic reconstruction performed solely with the regions subjected to gene conversion (i.e., amino acid sequences encoded by coding exons 1 and 2 and the end of coding exon 3) groups sequences by shark species pairs, whereas the ray sequences form two distinct Sparc-L and Scpp clades (fig. 4, supplementary material S7, Supplementary Material online). By contrast, a phylogenetic analysis excluding the conversion regions the variable PDYXXPXX repeat together with the Kazal and calcium-binding domains separate the elasmobranch sequences into two well-supported Sparc-L and Scpp clades (fig. 4, supplementary material S8, Supplementary Material online), which is also recovered in a phylogenetic reconstruction with the full species sampling (supplementary materials S9 and S10, Supplementary Material online). These analyses support both: (1) recent gene conversion in the shark lineage between sparc-L and scpp and (2) a single evolutionary origin for a scpp gene in all elasmobranchs. We reasoned that using the Ks value as a proxy to the timing of divergence between the sparc-L and scpp duplicates would represent an unbiased way to further test Hypothesis 3 (fig. 3). Hence, we calculated a synonymous mutation rate (Ks) in the nucleotide sequence located outside of the conversion zone (supplementary materials S11 and S12, Supplementary Material online), and used the available median value for selachimorph neutral mutation rate (Ks ≈ 0.2/100 My, according to Hara et al. 2018). Ks calculations between pairs of paralogs from the same selachimorph species had mean values of ∼0.7 (see branch lengths in supplementary material S12, Supplementary Material online) corresponding to a timing of divergence between the sparc-L and scpp genes of about 350 My. These values, therefore, confirm an origin of scpp from a local duplication of sparc-L in an ancestor of all elasmobranch fishes (265–400 My; third hypothesis in fig. 3).

Fig. 4.

The elasmobranch scpp gene experienced nonallelic gene conversion with sparc-L specifically in selachimorphs. (A) Dot blots representing the position of highly conserved sequences between genomic regions containing sparc-L and scpp of S. canicula (left panel) and A. radiata (right panel). The shaded areas correspond to the conversion zones (aligned in supplementary material S7, Supplementary Material online). The 1 kb scale applies horizontally and vertically. (B) Unrooted phylogenetic tree performed with protein regions under gene conversion, under the JTT model (alignment from supplementary material S7, Supplementary Material online). (C) Unrooted phylogenetic tree performed with sequences excluding the conversion regions, the variable repeat and the Kazal/calcium-binding domains, under the FLU + F + G4 model (alignment from supplementary material S8, Supplementary Material online). In (B) and (C), Sparc-L and Scpp proteins are shown in grey and orange, respectively, bootstrap values are indicated on branches and red nodes indicate values superior to 95.

An Acidic N-terminal Domain and Fam20°C Phosphorylation Sites are Ancestral Features of Chondrichthyan Sparc-L Proteins

In elasmobranchs, exon–intron structure, protein alignment, and phylogeny show that the Scpp proteins evolved from the Sparc-L N-terminal region (i.e., excluding the Kazal/calcium-binding domain). As the Sparc-L N-terminal region from sharks and rays is enriched in acidic residues and contains SxE consensus sites (fig. 1 and table 1, supplementary material S1, Supplementary Material online), we speculate that both structural features were already present in the last common ancestor of all elasmobranchs and were inherited by Scpp upon its emergence in this lineage. Similarly, previous work had noted that the elephant shark Sparc-L N-terminal region is highly acidic in nature (Kawasaki et al. 2017). Here, we further show that elephant shark Sparc-L acidity is dramatically exacerbated by the amplification of a 24 amino acid-long motif (containing 39.3% of D/E residues, fig. 1, see supplementary materials S1 and S4, Supplementary Material online) that suffices to decrease the theoretical isoelectric point of the protein (4.14 with 11 repeats vs. 4.52 with one repeat). In addition, the region encompassed by the duplicated motif contains 26 putative Fam20°C phosphorylation sites (fig. 1, supplementary material S1 and S4, Supplementary Material online) generating in the N-terminal region of elephant shark Sparc-L the highest abundance of SxE motifs of all sequences examined in this study (5.1 sites per 100 a.a. vs. 1.7 and 1.8 sites per 100 a.a. on average for the elasmobranch Sparc-L N-terminal regions and Scpp proteins, respectively). Despite being poorly alignable, the elasmobranch and holocephalan Sparc-L N-terminal regions display both an enrichment in acidic residues and the conspicuous presence of putative Fam20°C phosphorylation sites, which can be inferred to represent ancestral characters of all chondrichthyan Sparc-L proteins, and to have been inherited by Scpp upon its emergence in elasmobranchs.

Discussion

We demonstrate that a scpp gene evolved by duplication of the sparc-L 5′exons of in elasmobranchs, similarly to their evolution through the duplication of the sparc-L1 and/or L2 in osteichthyans, revealing at least two independent origins of scpp genes in jawed vertebrates. Independent evolution of the bony fish and elasmobranch scpp genes is supported both by our phylogenetic reconstructions and by molecular clock analysis. Despite convergent origins of the elasmobranch and the osteichthyan scpp genes, we chose to name this duplicate accordingly, because the duplication events that generated scpp genes in osteichthyans are largely unknown (discussed in Kawasaki et al. 2017) and may also include parallel duplication events. Furthermore, we reveal a history of gene conversion between scpp and sparc-L in selachimorphs, which is comparable to other cases of lineage-specific gene conversion between tandem duplicates in vertebrate genomes, as reported for the protocadherin and beta-globin genes (Noonan et al. 2004; Hoffmann et al. 2018). Remarkably, the elasmobranch Scpp convergently evolved emblematic characteristics of the osteichthyan Scpp proteins, as it lacks the Sparc-L Kazal/calcium-binding domain (Kawasaki et al. 2004) and acquired an N-terminal tyrosine-rich domain (Kawasaki and Amemiya 2014; Kawasaki et al. 2021). An additional level of convergence results from the highly specific expression patterns of scpp genes in presecretory ameloblasts (fig. 2). Indeed, expression data in shark and ray showed that sparc-L was already a pan-ameloblastic gene in the last common ancestor of all elasmobranchs (Enault et al. 2018), implying that, upon duplication, scpp diverged to restrict its expression pattern to the secretory phase of amelogenesis, as has been observed for several osteichthyan scpp genes (table 1; see Gasse et al. 2015; Kawasaki et al. 2021). In this respect, the parallel evolution of osteichthyan and elasmobranch Scpp proteins involved coordinated changes at: (1) the genomic level (i.e., a lineage-specific duplication of 5′ exons of the ancestral sparc-L or sparc-L1 or -L2 that excludes the exons coding for the C-terminal Kazal/calcium-binding domain); (2) the structural level (i.e., the emergence of a Y-rich domain, the maintenance of acidic residues and SxE sites in spite of significant sequence turnover); and (3) the regulatory level (i.e., the restriction of expression to the secretory ameloblasts). Scpp parallel evolution, therefore, represents a unique scenario that improves our understanding of independently evolved genes. Indeed, our proposed model (fig. 5) differs from other cases of independently evolved proteins that have originated from parallel fixation of mutations in different paralogs belonging to relatively large preexisting gene families (Christin et al. 2007; Flores-Aldama et al. 2020; Barua et al. 2021), from the progressive accumulation of mutations to produce a similar motif (Kriener et al. 2000), or from the emergence of biased amino-acid domains in proteins of completely unrelated origins (Calatayud et al. 2021).

Fig. 5.

Evolutionary scenario for the turnover of the jawed vertebrate ameloblastic genetic toolkit. The composition of the enamel/enameloid mineralization toolkit is shown for the hypothetical stem jawed vertebrates (A), holocephalans (B), elasmobranchs (C), and osteichthyans (D), whose odontodes contain a dentine core (Dn) covered by an enamel/enameloid layer (En). Arrows indicate the phosphorylation (P) of Sparc-L (orange) and/or Scpp (red) proteins by Fam20°C (yellow). Sparc-L harbors a C-terminal Kazal/calcium-binding domain (Ca2+) absent from Scpp proteins. The holocephalan Sparc-L N-terminal domain contains 11 tandem duplications of an acidic motif (vertical lines) enriched in SxE sites. A cladogram shows for each jawed vertebrate lineage the ameloblastic expression of sparc-L and scpp genes (orange and red branches, respectively). Asterisks indicate parallel scpp duplication events. According to this model the ancestral enamel/enameloid mineralization toolkit was based on Sparc-L, a situation maintained in holocephalans. The elasmobranch toolkit includes Sparc-L and Scpp, whereas Sparc-L1 and Sparc-L2 were functionally replaced by the Scpp family in osteichthyans.

The specific expression of scpp in secretory ameloblasts during catshark tooth and scale development strongly suggests that the elasmobranch scpp is a functional gene involved in enameloid matrix secretion and/or maturation. This idea is further supported by the presence of acidic domains and conspicuous putative Fam20°C phosphorylation sites in the elasmobranch Scpp proteins, as well as by the expression of fam20°C in catshark ameloblasts. Hence, in elasmobranchs, sequence and expression data support the hypothesis that Scpp binds calcium ions and is phosphorylated by Fam20°C during matrix mineralization, as has been reported for osteichthyan Scpp proteins (Tagliabracci et al. 2012; Wang et al. 2013; Klaning et al. 2014; Liu et al. 2017; Zhang et al. 2018; Schytte et al. 2020). Intriguingly, in the elephant shark, the Sparc-L protein is the most acidic and most enriched in putative Fam20°C phosphorylation sites, whereas the enameloid cap is relatively thin (Gillis and Donoghue 2007; Jerve et al. 2014). This apparent paradox might be resolved by considering that enameloid reduction in Callorhinchus milii is a derived character, and that holocephalan stem group specimen displays a well-mineralized outer odontode layer (Gillis and Donoghue 2007; Jerve et al. 2014). In addition to the chondrichthyan Sparc-L, the osteichthyan Sparc-L1/L2 proteins also harbor an N-terminal domain enriched in acidic residues (Kawasaki and Weiss 2006; Kawasaki et al. 2017) and in SxE sites (1.9 sites per 100 a.a. on average, data not shown). As a result, we conclude that, in the jawed vertebrate lineage, an acidic Sparc-L protein was ancestrally phosphorylated by Fam20°C and was also binding calcium during enameloid formation, a situation that probably already existed in osteostracans and placoderms (fig. 5), and that remained largely unchanged in the holocephalan lineage (fig. 5). Elasmobranchs and osteichthyans convergently evolved Scpp proteins that respectively function redundantly with Sparc-L (fig. 5) or eventually took over the role ancestrally played by Sparc-L1/L2 (fig. 5).

Previously, the absence of scpp genes in holocephalans was thought to be representative of the whole chondrichthyan group, and has been used as an argument to propose that osteichthyan enamel and chondrichthyan enameloid are not homologous tissues (Qu et al. 2015; Kawasaki et al. 2017, 2021). Here, we reinforce arguments supporting the homology of some developmental processes that generate an outer hypermineralized layer in vertebrate odontodes (also discussed in Kawasaki et al. 2017). We consolidate the idea that an acidic Sparc-L subjected to Fam20°C posttranslational regulation represents a crucial molecular element of the ancestral ameloblastic genetic toolkit. Although Sparc-L and Scpp are probably still partially redundant in elasmobranchs, the dramatic expansion of the Scpp gene family in osteichthyans has clearly minimized, or even abolished, the skeletal expression and function of Sparc-L1/L2 in this group (Bertrand et al. 2013; Enault et al. 2018), explaining why enamel was previously thought to be fundamentally different from chondrichthyan enameloid. One of the earliest works on the scpp genes employed the term “phenogenetic drift” to illustrate how homologous enamel and ganoin extracellular matrices from distinct osteichthyan lineages have progressively modified the nature and composition of the Scpp proteins they contain (Kawasaki et al. 2005). Our study expands the concept of phenogenetic drift by showing that the ameloblastic cassette is subjected to a high rate of turnover in jawed vertebrates, leading to different sets of proteins being secreted to the mineralizing front of the outer odontode layer of holocephalans, elasmobranchs, and osteichthyans. In this context, the hypothesis of an ancestral regulation of Sparc-L by Fam20°C represents a unifying genetic basis that, if demonstrated, would prove the homology of developmental processes that evolved in the remarkably diverse enamel/ganoin/enameloid cap covering the teeth and scales of extant and extinct jawed vertebrates.

Materials and Methods

Animal care, staging, histology, and in situ hybridization procedures of embryonic lesser spotted catshark S. canicula were performed as previously described (Enault et al. 2018). The scpp and fam20°C sequences were amplified and cloned using lesser spotted catshark vertebrae cDNA and the following primers (5′–3′): scpp-Forward GATTTGGGCAGCAACAGTCA, scpp-Reverse TGTCTAACCCCGGTGTGAAA; fam20°C-Forward GGCTGCTGATCATCATGGTG, fam20°C-Reverse GGAAAGCAGCAATCTCCGAG, to be used for RNA probe synthesis. Refer to Enault et al. (2018) for sparc-L probe synthesis. All in situ hybridizations included two negative control slides (no probe and sense probe), which were devoid of signal (supplementary material S13, Supplementary Material online).

All accession numbers and genomic coordinates are available in supplementary material S14, Supplementary Material online. Amino acid enrichments were detected with CAST (Promponas et al. 2000), internal repeats were searched with XSTREAM (Newman and Cooper 2007), and repeat logos were generated with the GGSEQLOGO R package (Wagih 2017). Isoelectric point was computed using Expasy online tool (Gasteiger et al. 2003). Highly similar sequences between the Sparc-L and Scpp loci were searched using BLAST 2.2.19 with default parameters (Altschul et al. 1997).

Sequences used for phylogenetic reconstructions were aligned using MAFFT (Katoh et al. 2002; Katoh and Standley 2013), using standard parameters. Alignments were cleaned using HmmCleaner with standard parameters (Di Franco et al. 2019). Phylogenetic analyses were performed using Maximum Likelihood under IQ-TREE 1.6.12 (Nguyen et al. 2015). Node support was estimated by performing five thousand ultra-fast (UF) bootstrap replicates (Hoang et al. 2018) and single branch tests (SH-aLRT, see Guindon et al. 2010). The minimum correlation coefficient for UF bootstrap convergence criterion was set to 0.90 and evolutionary models were chosen by ModelFinder (Kalyaanamoorthy et al. 2017). CDS nucleotide sequences for chondrichthyan scpp and sparc-L were aligned using MACSE (Ranwez et al. 2018). The resulting alignments were trimmed to exclude the conversion zones and Kazal domains. The trimmed alignment (supplementary material S11, Supplementary Material online) and the topology of the concatenated phylogeny (fig 4) with C. milii as an outgroup were used to compute Ks values for each branch using PAML v4.9c (Yang 2007).

Supplementary Material

Supplementary data are available at Molecular Biology and Evolution online.

Click here for additional data file.