Transcriptome analysis and phylogenetic analysis
To reveal the repertoire of the SPILE genes that are expressed in the early developmental stages of bivalves, we conducted transcriptome analysis of multiple early developmental stages (mixture of cleavage, gastrula and trochophore stages) of the mussel M. virgata. In the de novo transcriptome assembly, we identified 24 TALE class genes using a BLAST-based search (Additional file 2: Table S1 and S2). To extract SPILE genes, we performed phylogenetic analysis with TALE genes from bilaterian animal lineages and found that 13 genes were classified as SPILE genes (Additional file 1: Fig. S1).
To elucidate the evolutionary history of SPILE genes, we constructed a phylogenetic tree using annotated SPILE gene sequences from two bivalves Pinctada fucata and Crassostrea gigas, and two equal-cleaving gastropod species, the limpet N. fuscoviridis and Lottia gigantea. (Fig. 2). Based on their phylogenetic relationships, we divided SPILE genes into four clades (B-, D-, E- and A/C-clades; the name of each clade was given based on the NfSPILE genes in that clade). Although the support value for each clade was not high, these values were improved by the removal of two fast-evolving MvSPILE genes—MvSPILE-11 and MvSPILE-12 (Additional file 1: Fig. S2; Fig. 2). In addition, genes belonging to the same clade shared sequence features outside homeodomains. Each B-clade SPILE gene, as reported previously [29, 30], contains two homeodomains (Additional file 1: Fig. S3). Although a previous study has classified the D-clade SPILE genes of bivalves and limpets into separate clades, (, Additional file 2: Table S2), we found that the sequences outside the homeodomain were also similar among the D-clade genes, (Additional file 1: Fig. S4), supporting the close relationships between D-clade SPILE genes. SPILE genes in the A/C-clade have been classified into various gene clades in a previous study (, Additional file 2: Table S2). On the other hand, most of these genes reportedly have a PADRE domain, a conserved region of approximately 90 amino acids that lies upstream of the homeodomain ([27, 30] Additional file 2: Table S2). Actually, most of the A/C-clade SPILE genes, including mussel genes, have a region similar to the PADRE domain upstream of the homeodomain (Additional file 1: Fig. S5; Additional file 2: Table S2). Concerning TALE genes in molluscs, the existence of a PADRE domain outside the A/C-clade was reported only for NfSPILE-D (; Additional file 2: Table S2). However, we found that all other D-clade genes and most of the B-clade genes have a region similar to the PADRE domain upstream of the homeodomain, although the corresponding regions are slightly longer than the PADRE domain in the A/C-clade genes (Additional file 1: Fig. S6). We did not find conserved regions that corresponded to the PADRE domain in the E-clade SPILE genes (Additional file 1: Fig. S7).
Based on this phylogenetic framework, the results showed that single SPILE-D and E clade genes were present in the examined species (Table 1). Conversely, in the B-clade, one SPILE gene was found in each limpet, whereas two to three genes were identified in the bivalves, suggesting that duplication events had occurred in the bivalve lineage. The number of SPILE A/C-type genes found in the limpet Lottia gigantea and bivalve species was similar (Table 1). However, eight homeodomain sequences in the limpet (LgTALEHD 21, 22, and 29–34) were nearly identical and located tandemly in the genome (; Fig. 2; Additional file 2: Table S2), suggesting recent and lineage-specific duplications of these homeobox sequences. Conversely, seven or eight SPILE genes in the A/C clade have been predicted in each bivalve species, and many of these genes are not in tandem in the genome (; Additional file 2: Table S2).
To gain further insight into the timing of duplication of the A/C-clade SPILE genes, we extracted SPILE sequences from additional molluscan species, including one chiton species (Acanthopleura granulata ), three gastropod species (abalone Haliotis discus hannai , deep-sea hydrothermal vent snails Gigantopelta aegis  and Chrysomallon squamiferum ), and the four bivalve species (hard-shelled mussel Mytilus coruscus , the eastern oyster Crassostrea virginica , the Japanese scallop Mizuhopecten yessoensis , and the king scallop Pecten maximus . Among the three gastropods, abalone is an equal-cleaving mollusc ; however, no information on the embryonic development of the other two species is available. Although no information is available on the cleavage pattern of the chiton species examined (A. granulata), several chiton species show equal cleavage in the first two divisions [41,42,43]. The A/C-clade homeobox sequences from these species were extracted by combining BLAST and phylogenetic analysis (Additional file 1: Fig. S8–10; Additional file 2: Table S2; see the “Methods” section for details). Furthermore, to increase the resolution of phylogenetic relationships in the A/C-clade SPILE genes, phylogenetic analysis without B/E-clade SPILE genes (D-clade SPILE genes were used as an outgroup) was performed (Additional file 1: Fig. S11, using sequences of molluscs; Additional file 1: Fig. S12, using sequences of bivalves). Although many clades composed of SPILE genes from the same or closely related species were recovered with bootstrap values ≥ 50, most clades containing SPILE genes from distantly related species were not recovered with bootstrap values ≥ 50 (Additional file 1: Fig. S9–12). Only the clade containing MvSPILE-13 and CgTALE-6 (TALE-V clade; ) was supported with high bootstrap values as a clade composed of SPILE genes from distant bivalve species (Additional file 1: Fig. S9–12). The number of A/C-clade genes (including homeodomain sequences which are not predicted as genes), position on the genome, and clades with bootstrap values ≥ 50 are summarised in Fig. 3.
Although the SPILE-A/C gene duplications were also found in chitons and gastropods, most duplications likely occurred in each lineage independently (highlighted in yellow in Fig. 3). For example, although seven SPILE-A/C clade genes were identified in chiton, all of them were located on the same scaffold and had nearly identical homeobox sequences, suggesting that one type of homeobox gene was duplicated in the chiton lineage (Fig. 3). In abalone, eight A/C-clade genes were identified, and four genes were closely located on the two scaffolds (Additional file 2: Table S2; Fig. 3, HDSC07248 and HDSC00981). Two and three SPILE-A/C genes were identified in the two deep-sea gastropods, respectively, and the two genes in C. squamiferum located in the same scaffold and homeodomain sequence were identical, indicating a recent duplication event had occurred. These results suggest that most of the duplications observed in the A/C-clade SPILE genes in gastropods and chiton occurred independently in each lineage and that the ancestors of molluscs may have 1–2 types of SPILE genes in the A/C-clade.
In bivalves, although the number of A/C-clade SPILE genes varied widely from 4 to 12 and there were potential lineage-specific duplications (Fig. 3; highlighted in yellow), diverse types of homeodomains were found in all species examined. For example, in the scallop (M. yessoensis) that had the smallest number (four) of A/C-clade SPILE genes among the examined bivalve species, three types of distinct homeobox sequences were found (Additional file 1: Fig. S9–12; Fig. 3; e.g. My_NW_018408528.1_504212-504400, My_XP_021369004.1, and My_XP_021357151.1/My_NW_018406777.1_389495-389683). Similarly, more than three types of homeobox sequences were identified in the other bivalves examined (Additional file 1: Fig. S9–12; Fig. 3). These results suggest that duplications and subsequent sequence divergence of the SPILE A/C-clade genes in the bivalve lineages had occurred.
Expression patterns of MvSPILE genes
To explore the roles of SPILE genes in the early development of bivalves, we examined the expression patterns of 12 SPILE genes in the mussel, with the exception of MvSPILE-13, for which only a short fragment of the gene was recovered (Additional file 3: Dataset S1).
MvSPILE genes in D- and E- clades
MvSPILE-1 and MvSPILE-2, in the D and E clades are expressed maternally (Fig. 4a, g). The counterparts of these two genes in limpets, NfSPILE-D and NfSPILE-E, were also reported to exhibit maternal expression patterns . In limpets, the NfSPILE-D gene is expressed in all blastomeres until the eight-cell stage, but expression is restricted to the four blastomeres at the animal pole at the 16- and 32-cell stages. Similarly, MvSPILE-1 was expressed uniformly in all blastomeres until the 16-cell stage, but was expressed predominantly in the two animal-most blastomeres, 1c11 and 1d11, at the 32-cell stage (Fig. 4b–f; Additional file 1: Fig. S13) By contrast, unlike limpet NfSPILE-D, expression was predominantly restricted to the C and D blastomeres among 1q11 lineages (Fig. 4f; Additional file 1: Fig. S13). MvSPILE-2 was expressed in all blastomeres until the 16-cell stage (Fig. 4g–k; Additional file 1: Fig. S13), similar to its limpet counterpart gene .
MvSPILE genes in B-clade
Although only one gene belonging to the SPILE-B clade was found in limpets (Fig. 2; ), three SPILE-B genes were found in the mussel transcriptome. All three genes, MvSPILE-3, -4, and -5, showed expression in 2d blastomeres at the 16-cell stage (Fig. 5a, b, d, e, f, h, i, j, l). In limpets, NfSPILE-B expression was detected in the second quartet at the 16-cell stage, but its expression was not restricted to the D lineage , which differed from the findings in bivalves. All B-clade SPILE genes were initially expressed in the 2d blastomere at the 16-cell stage, but their expression patterns diverged later at the 32-cell stage (Fig. 5c, g, k). MvSPILE-3, the signal was restricted to the 2d2 blastomere (Fig. 5c), whereas the signal of MvSPILE-4 was present in both 2d daughter cells (2d1 and 2d2; Fig. 5g), and no specific expression of MvSPILE-5 was detected at the 32-cell stage (Fig. 5k). These expression patterns imply that functional differentiation has occurred between MvSPILE-3, -4, and -5.
MvSPILE genes in A/C-clade
Transcriptome analysis of the early developmental stage identified two SPILE genes (NfSPILE-A and -C) in limpets , whereas eight SPILE genes were identified in mussels. One of those, MvSPILE-6 was expressed in all blastomeres at the 4- and 8-cell stage, but exhibited macromere-specific expression at the 16-cell stages (Fig. 6a–i). Similarly, MvSPILE-10 expression was first detected in all blastomeres at 4-cell stage (Fig. 6j-k, p). However, at the 8-cell stage, signals of the MvSPILE-10 gene appeared to be specifically excluded from the 1D blastomere (Fig. 6m, q). Nevertheless, at the 16-cell stage, MvSPILE-10 signals were again evident in the 2D cell along with the other macromeres, similar to MvSPILE-6 (Fig. 6n, o, r).
In contrast, three of the other A/C-clade MvSPILE genes exhibited quadrant-specific expression patterns at the 4-cell stage. For example, MvSPILE-7 and MvSPILE-8 were expressed in the D blastomere at the 4-cell stage (Fig. 6s–u and y; bb-dd and hh). In addition, MvSPILE-9 appeared to be expressed in the C and D blastomeres at the 4-cell stage (Fig. 6 kk-mm and rr). These SPILE genes were expressed in all daughter cells of D-quadrant (MvSPILE-7 and -8) or C- and D-quadrants (MvSPILE-9) at the 8- and 16-cell stages (Fig. 6 v-x, z and aa; ee-gg, ii and jj; nn-qq, ss, and tt).
Two of the remaining genes were expressed/not expressed only in certain blastomeres. The expression of MvSPILE-11 was first detected in the 1D blastomere at the eight-cell stage, and the signal was restricted to the 2D cell at the 16-cell stage (Fig. 7a–e). The expression of MvSPILE-12 was detected in all blastomeres at the eight-cell stage, but it was no longer expressed in the 2d and 2D cells at the 16-cell stage, while remaining in all other cells (Fig. 7f–k). We summarised the expression patterns and phylogenetic relationships of SPILE genes in limpets and mussels (Fig. 8).