Evolutionary History of Bilaterian FoxP Genes: Complex Ancestral Functions and Evolutionary Changes Spanning 2R-WGD in the Vertebrate Lineage

Abstract

Human and fly FoxP homologs are well-known for their roles in the development of cognitive abilities. These findings have led to the hypothesis that the ancestral function of FoxP was in the development of cognitive neural circuits. However, complex brains in human and fly evolved independently, and the similar cognitive function of FoxP in human and fly may thus be interpreted as a result of convergent evolution. In addition, the 4 gnathostome FoxP paralogs also possess diverse developmental functions unrelated to neurodevelopment, which might have been overlooked in comparative studies of invertebrate FoxP homologs. To resolve these uncertainties, we set out to improve the phylogenetic reconstruction of vertebrate FoxP homologs and broaden the taxonomic sampling of gene expression profiling to include an invertebrate chordate, ambulacrarian deuterostomes, and a spiralian protostome. Using phylogenetic analysis combined with synteny mapping, we elaborated the hypothesis that the 4 FoxP paralogs arose through the 2R-WGD events shared by all gnathostome species. Based on this evolutionary scenario, we examined the FoxP expression pattern in amphioxus development and concluded that FoxP already had complex developmental functions across all germ layers in the chordate ancestor. Moreover, in sea urchin, hemichordate, and catenulid flatworm, FoxP was expressed in the gut prominently, in addition to the anterior neurogenic ectoderm. This surprising similarity shared among these distantly related species implies that FoxP may have a significant function in gut development in addition to the neural development function in the last common ancestor of bilaterians.

Introduction

FoxP genes, a subfamily within the Forkhead box (Fox) transcription factor gene family (Weigel and Jäckle 1990; Mazet et al. 2003), are best known for their role in the development of cognitive functions in the human brain (Co et al. 2020). Among the 4 human FoxP paralogs, mutations in FoxP1 are associated with an assortment of neurodevelopmental disorders, including autism spectrum disorder and intellectual and motor skill disabilities (Meerschaut et al. 2017; Siper et al. 2017), while FoxP2 deficiencies primarily lead to defects in language and communication functions (Lai et al. 2001; Reuter et al. 2017). Similar neurodevelopmental functions related to cognitive abilities for FoxP1 and FoxP2 were also identified in birds and other mammals (Shu et al. 2005; Araujo et al. 2015, 2017; Bacon et al. 2015; Chen et al. 2016; Braccioli et al. 2017; Burkett et al. 2018). On the other hand, FoxP4 is specifically required for motor coordination, as it is essential for maintaining the dendritic arborization of Purkinje cells in the cerebellum (Tam et al. 2011). Consistent with their respective functions in the development of cognitive abilities and motor coordination, FoxP1, FoxP2, and FoxP4 are expressed in the developing brain (Co et al. 2020). In contrast, FoxP3 is expressed specifically in the immune system (Brunkow et al. 2001) and is required for the development of regulatory T cells (Fontenot et al. 2003, 2005; Hori et al. 2003; Gavin et al. 2007).

Functional studies of invertebrate FoxP genes have been carried out mainly on the fly. Strikingly similar to the mammalian FoxP1, FoxP2, and FoxP4, the single FoxP ortholog in Drosophila is also required for neurobehavioral traits, such as decision-making, learning, and motor coordination (DasGupta et al. 2014; Lawton et al. 2014; Mendoza et al. 2014; Groschner et al. 2018; Castells-Nobau et al. 2019; Kottler et al. 2019; Palazzo et al. 2020). At the cellular level, mammalian and fly FoxP genes have similar roles in circuit-building developmental processes, such as regulating neurite outgrowth and synaptogenesis (Vernes et al. 2011; Chen et al. 2016; Castells-Nobau et al. 2019). These similarities have led to a hypothesis that the roles of FoxP in the development of cognitive neural circuits are conserved among bilaterians (Lawton et al. 2014; Mendoza et al. 2014; Castells-Nobau et al. 2019).

Nevertheless, two lines of evidence suggest that the idea of urbilaterian FoxP functioning in forming the cognitive neural circuits might be a gross oversimplification. First, complex brains capable of deriving highly elaborated cognitive abilities have evolved independently in insects and vertebrates (Farris 2013; Roth 2015), and thus, the cognitive functions of FoxP in the fly and vertebrates may result from convergent evolution. Secondly, FoxP genes play various roles during vertebrate embryogenesis. With such an intense focus on neural development and the high complexity in the expression patterns of the duplicated vertebrate FoxP paralogs, some of the conserved functions might have been overlooked when investigators compared the expression patterns between vertebrate and invertebrate species. For example, FoxP was reported to be expressed in the gut of sea urchin larvae (Tu et al. 2006; Paganos et al. 2021). Interestingly, some mouse FoxP paralogs are also expressed in anterior endoderm-derived organs, such as the lung and esophagus, and are required for their development (Shu et al. 2007). Together, these data would hint at a possibility that the role of FoxP in gut development is conserved, at least among deuterostomes.

To reconstruct the evolutionary history of FoxP genes and decipher their possible conserved functions in bilaterians, we investigated FoxP homologs in selected invertebrate species. First, we demonstrated that the 4 vertebrate FoxP paralogs arose from the 2 rounds of whole-genome duplication (2R-WGD) events occurring near the base of the vertebrate tree by synteny and phylogenetic analyses. Next, we characterized the developmental expression pattern of the amphioxus FoxP gene and compared it with the vertebrate homologs to identify the plesiomorphies and apomorphies among the vertebrate FoxP functions. We then extended our analysis to representative ambulacrarian species to identify the shared expression patterns of FoxP among the deuterostomes. Finally, we characterized FoxP expression in a catenulid flatworm. Surprisingly, in ambulacrarian embryogenesis and catenulid asexual growth, FoxP homologs are both expressed in the foregut and anterior neurogenic ectoderm, suggesting that the ancestral function of FoxP among bilaterian species may be in the development of both gut and anterior neurogenic ectoderm.

Results and Discussion

2R-WGD. and the Origin of Gnathostome FoxP1-4 Paralogs

Members of the FoxP subfamily are structurally identified by the presence of FoxP-specific zinc finger (ZF) and leucine zipper (LZ) domains in an N-terminal region of their protein products and a C-terminal forkhead (FH) DNA binding domain (Li et al. 2004; Santos et al. 2011). The ZF-LZ domains mediate protein dimerization and are critical for their repressor function (Wang et al. 2003). FoxP arose well before the last common ancestor of bilaterians, as it is recovered from the genomes of nonbilaterian metazoans such as cnidarians and sponges (Adell and Müller 2004; Larroux et al. 2008; Yuan et al. 2022).

In most invertebrate genomes, FoxP exists as a single-copy gene, though sporadic lineage-specific duplications and losses were reported across the bilaterian tree (Mazet et al. 2003; Tu et al. 2006; Yu et al. 2008; Fritzenwanker et al. 2014; Pascual-Carreras et al. 2021; Schomburg et al. 2022; Seudre et al. 2022). Notably, 4 FoxP paralogs are typically found in nonteleost gnathostome genomes. The intron–exon organization of FoxP loci in the genomes of human and selected invertebrate species is largely conserved (supplementary fig. S1, Supplementary Material online), indicating that the vertebrate FoxP paralogs arose through chromosome segment duplication or whole-genome duplication, but not retroduplication.

Based on comparative genomic data, two rounds of whole-genome duplication (2R-WGD) were inferred to have occurred at the base of the vertebrate lineage, with the first round taking place in the stem vertebrate lineage and the second round in the stem gnathostome lineage, after its separation from the cyclostome lineage (Simakov et al. 2020). Shared microsynteny among the 4 human FoxP paralogs led to the hypothesis that the 4 gnathostome FoxPs (FoxP1-4) arose by 2R-WGD (Song et al. 2016). However, microsynteny could also result from serial duplications of a chromosomal segment containing the ancestral FoxP and its proximal neighbors, and previous phylogenetic analyses supported the serial-duplication hypothesis (Santos et al. 2011). To further characterize the origin of vertebrate FoxP genes, we set out to test (i) if vertebrate FoxP paralogs share micro- and macrosynteny with amphioxus FoxP and (ii) if the phylogenetic relationships among the vertebrate FoxP duplicates are consistent with the 2R-WGD hypothesis.

Synteny Informs Phylogenetic Relationships Between Amphioxus FoxP and Its Vertebrate Homologs

To identify local syntenies (microsyntenies) shared between the amphioxus and vertebrate FoxP loci, we mapped the genomes of selected vertebrate species for the orthologs of the contiguous genes near the FoxP loci in the amphioxus chromosomes 5. The lamprey Petromyzon marinus was selected to represent jawless vertebrates that diverged after the first round of WGD. For technical simplicity, the spotted gar Lepisosteus oculatus, a holostean, was chosen to represent the lineage that emerged after 2R-WGD, given that it did not undergo the third round of WGD specific to the teleosts (Taylor et al. 2003; Braasch et al. 2016). In both the lamprey and the spotted gar genomes, a significant number of syntenic genes was identified near the FoxP loci (supplementary fig. S2, Supplementary Material online).

Comparative genomic studies using chromosome-level genome assembly information from diverse metazoan phyla have enabled the identification of chromosome-level syntenies (macrosyntenies) and revealed the evolutionary process of chromosomes in various metazoan lineages (Simakov et al. 2020, 2022; Lin et al. 2024). Based on the available chromosome-level genome assemblies, we set out to delineate the duplication history of FoxP-bearing chromosomes in early vertebrate evolution. We located FoxP-bearing chromosomes in the genome assemblies of the scallop P. yessoensis, the sea urchin S. purpuratus, the amphioxus B. floridae, and the spotted gar L. oculatus (Fig. 1a). Notably, FoxP-bearing chromosomes exhibited remarkable macrosyntenic conservation among the 3 invertebrate species examined, while in the spotted gar genome, FoxP paralogs are distributed in 4 chromosomes showing detectable macrosyntenies with the amphioxus FoxP-bearing chromosome Bfl5 (Fig. 1a and supplementary fig. S3, Supplementary Material online). Specifically, spotted gar FoxP1 is located in chromosome Loc5, FoxP2 in Loc8, FoxP3 in Loc1, and FoxP4 in Loc3. We then used previously inferred metazoan ancestral linkage group information and confirmed that amphioxus FoxP-bearing chromosome Bfl5 and spotted gar chromosomes Loc5, Loc8, Loc1, and Loc3 are all descended from the ALG_E (Simakov et al. 2020, 2022). Simakov et al. (2020) suggested that after the first round of WGD, ALG_E duplicated into 2 chromosomes (designated as 1 and 2 in Fig. 1b); subsequently, these 2 chromosomes experienced a second round of WGD, resulting in 4 chromosomes, naming 1α (corresponding to the spotted gar chromosome Loc5) and 1β (Loc1) from the first duplicate, and 2α (Loc8) and 2β (Loc3) from the second. Our finding that each of the 4 spotted gar FoxP paralogs can be assigned to a paralogon suggested that the 4 gnathostome FoxP paralogs arose through 2R-WGD, not segmental duplications.

Fig. 1.

Macrosyntenic analysis between scallop, sea urchin, amphioxus, and spotted gar. a) Ribbon diagram showing conserved macrosyntenies among FoxP-bearing chromosomes of the 4 species. Horizontal bars represent chromosomes, and pink lines indicate the positions of FoxP orthologs/paralogs. Green arches connected orthologous gene pairs on the chromosome containing FoxP. Gray arches indicate orthologous gene pairs on other corresponding chromosomes among animals. Pairwise comparisons are presented in supplementary fig. S3, Supplementary Material online. b) Cladogram showing the evolutionary history leading to 4 spotted gar FoxP-bearing chromosomes (Loc1, Loc3, Loc5, and Loc8), starting from the proto-vertebrate ALG_E. The evolutionary history of vertebrate chromosomes is based on data from Simakov et al. (2020).

Analysis of gene retention fraction at the whole-genome scale has revealed that α segments retain over twice as many genes as β segments after the second round of WGD (Simakov et al. 2020). Consistently, we identified 194, 196, 91, and 93 homologous gene pairs between amphioxus Bfl5 and spotted gar Loc5 (FoxP1-bearing, 1α derivative), Loc8 (FoxP2-bearing, 2α derivative), Loc1 (FoxP3-bearing, 1β derivative), and Loc3 (FoxP4-bearing, 2β derivative), respectively. Following the evolutionary scenario of the vertebrate chromosomes and the chromosomal location of FoxP paralogs, we thus inferred that the first vertebrate WGD produced 2 sister genes, namely FoxP1/3 and FoxP2/4, before the split of jawed vertebrates (gnathostomes) and jawless vertebrates (cyclostomes); subsequently in the jawed-vertebrate lineage, the second WGD further resulted in the 4 extant FoxP paralogs (Fig. 1).

Phylogenetic Analysis of Vertebrate FoxP Genes

Next, we performed phylogenetic analyses using protein sequence information to test whether the pattern of FoxP molecular phylogeny is consistent with the inferred evolutionary process of vertebrate chromosomes. Under the scenario that 2R-WGD gave birth to the gnathostome FoxP1-4 orthologous groups, their phylogenetic tree is expected to exhibit symmetric bifurcations (hereinafter referred to as 2R-WGD topology; Fig. 2a), but this pattern was never attained in previous studies (Yu et al. 2008; Santos et al. 2011; Song et al. 2016; Seudre et al. 2022). Furthermore, cyclostomes and gnathostomes share the first, but not the second, round of WGD. Therefore, cyclostome FoxP paralogs should arise from the ancestral FoxP1/3 and FoxP2/4, independent of the second round of WGD that gave rise to the gnathostome FoxP1-4 ortholog groups (Fig. 2a).

Fig. 2.

Phylogenetic analysis of deuterostome FoxP-encoded proteins. a) A phylogenetic hypothesis for deuterostome FoxP genes based on the 2R-WGD scenario. The 4 gnathostome paralogs arose from the 2R-WGD events. Our JTT + C60 + G ML tree suggests that the cyclostome FoxP paralogs are likely FoxP2/4 descendants, while the foxp1/3 lineage was lost. b) Maximum-likelihood (ML) tree of a set of deuterostome FoxP proteins produced using the site-heterogeneous JTT + C60 + G model. c) ML tree of the same set of genes shown in b) produced by using the site-homogeneous JTT + F + I + G4 model. Note that the lamprey genes are affiliated with the gnathostome FoxP3 in this tree. The scale bars differ between b) and c); branch lengths are underestimated in the tree shown in panel c). For clarity, the bootstrap support values of the internal nodes in the FoxP1-4 clades are omitted. Only those with a bootstrap value under 80 are displayed for deeper nodes. Detailed depictions of tree topologies and bootstrap values are given in supplementary fig. S4, Supplementary Material online.

Given the unequal divergence rates among vertebrate FoxP paralogs (Santos et al. 2011), the potential impact of phylogenetic artifacts resulting from long-branch attraction (LBA; Felsenstein 1978; Bergsten 2005) should not be ignored. To reduce the effect of LBA, we first considered the outgroup composition. It has been shown that an excessive number of distant outgroups may aggravate the LBA artifact (Pisani et al. 2015), and this seems to be a prevalent problem in previous attempts at FoxP phylogeny since non-FoxP members and nondeuterostome FoxP genes were often included. Here, we used a small set of FoxP singletons from representative invertebrate deuterostomes as the outgroup taxa for our phylogenetic test of vertebrate FoxP genes. On the other hand, site-heterogeneous mixture models have been shown capable of counteracting the LBA effect in phylogenomic analysis (Lartillot et al. 2007; Rodríguez-Ezpeleta et al. 2007), though their superior performance is not guaranteed for short alignments (Baños et al. 2024). Since gene family phylogenies, especially those of transcription factors, are inferred from relatively short alignments, we took this opportunity to compare the performance of the site-heterogeneous model against the best-fit site-homogeneous model using our FoxP datasets.

In the Maximum-likelihood (ML) tree produced using the site-heterogeneous JTT + C60 + G model, the 2R-WGD topology was recovered, and each of the gnathostome FoxP1, FoxP2, FoxP3, and FoxP4 orthologous groups was recovered as a monophyletic group with a high bootstrap support (Fig. 2b, supplementary fig. S4a, Supplementary Material online). However, the bootstrap support for the deep nodes associated with 2R-WGD events was generally poor (21% to 49%), indicating phylogenetic instability. In contrast, the best-fit site-homogeneous model JTT + F + I + G4 gave rise to a different topology supporting the hypothesis that the gnathostome FoxP1-4 arose through serial duplications (Fig. 2c, supplementary fig. S4b, Supplementary Material online). Suppose the synteny-supported 2R-WGD topology is the true phylogenetic relationship among the gnathostome FoxP1-4 genes. In that case, our results suggest that, at least for this instance, site-heterogeneous models can indeed outperform site-homogeneous models in gene-level phylogenetic analysis.

In our JTT + C60 + G ML tree, the 5 lamprey FoxP genes formed a monophyletic group sister to the gnathostome FoxP2/4. This pattern is consistent with the 2R-WGD hypothesis in that the cyclostome FoxP paralogs arose after the first round of WGD, but not with the post-1R hexaploidization hypothesis, where a cyclostome-specific hexaploidization event took place after the first WGD shared between cyclostomes and gnathostomes (Nakatani et al. 2021; Marlétaz et al. 2024; Yu et al. 2024). Chromosome mapping of lamprey FoxP genes supports the notion that 1R and post-1R hexaploidization whole-genome duplication events were responsible for the birth of at least 4 lamprey FoxP genes, given that 6 lamprey chromosomes were inferred as ALG_E derivatives, with 4 containing a FoxP locus (supplementary fig. S5a, Supplementary Material online). However, the phylogenetic affinities of the 6 lamprey chromosomes to the 4 gnathostome ALG_E derivatives were not well resolved (Marlétaz et al. 2024), and we cannot reliably assign individual lamprey FoxP genes as being FoxP1/3- or FoxP2/4-derivative by using phylogenetic tests (supplementary fig. S5b, Supplementary Material online). In any case, given the instability of deep nodes, the results of our phylogenetic analysis should be interpreted with caution. Using sequence-based phylogenetic analysis alone, we could not have confidently concluded that the 4 gnathostome FoxP paralogs arose from the FoxP singleton through 2R-WGD. From a technical point of view, our results demonstrated the usefulness of synteny mapping in solving difficult gene-level phylogenetic problems.

Expression Patterns of FoxP During Amphioxus Development Inform the Functional Evolution of the Duplicated Vertebrate FoxP Genes

Duplicated genes are generally preserved through subfunctionalization (or specialization) and neofunctionalization (Force et al. 1999; Marlétaz et al. 2018). As such, gene duplicates generated by 2R-WGD may be associated with clade-defining traits arising in the early stage of vertebrate evolution; therefore, 2R-WGD has been considered an important source of genetic materials underlying the developmental and morphological innovations in the vertebrate lineage (Holland et al. 1994; Gil-Gálvez et al. 2022). Given that the 4 gnathostome FoxP genes are hypothesized to be the products of the 2R-WGD events, it might be possible to identify signatures of neofunctionalization and subfunctionalization/specialization by comparing FoxP expression patterns between gnathostome species and amphioxus, a representative of the pre-2R chordate lineages (Holland et al. 2008; Putnam et al. 2008).

To serve as the basis for comparison with gnathostome FoxP1-4 genes, we examined the expression patterns of amphioxus FoxP at various developmental stages by whole mount in situ hybridization (WMISH). Consistent with a developmental transcriptomic dataset of amphioxus (Hu et al. 2017), we found no maternal expression of FoxP (Fig. 3a to c). In early gastrulae (stages G0 to G4), the amphioxus FoxP expression was first detected ubiquitously but stronger anteriorly, and the expression was subsequently confined to the dorsal ectoderm (Fig. 3d to g). FoxP expression was then detected in the presumptive neural plate and the underlying dorsal axial mesendoderm in late gastrula (stage G6) (Fig. 3h and h”), and subsequently in the forming cerebral vesicle, the ventral part of the rostral neural tube, notochord, medial wall of the forming somites, and the gut wall in early-neurula (stage N1) (Fig. 3i, i’). At the late neurula stages (N4/5), FoxP expression was most significant in the cerebral vesicle and the rostral part of the notochord (Fig. 3j). The expression in the notochord is diminished gradually by the larva stage. Still, the expression is maintained in the cerebral vesicle, the neural tube, the ductal part of the club-shaped gland, Hatschek's pit, the pharyngeal endoderm, the posterior gut, and the tailbud (Fig. 3k to m). In the L1-stage larva, transient FoxP expression was detected in the rostral tip (Fig. 3l), which houses sensory neurons in the adult (Lacalli 2004). Based on these observations, we concluded that FoxP expression is dynamic and distributed across all 3 germ layers during amphioxus development.

Fig. 3.

Expression pattern of FoxP in amphioxus embryonic development. WMISH of FoxP in amphioxus B. floridae at different developmental stages, including unfertilized egg a), 64-cell stage b), blastula c), early gastrula d to g), late gastrula h), neurula i and j), and larva k to m). d to h, i, j to m) are viewed from the lateral side. Embryos in h′) and h″) are the blastopore view and dorsal view of the same embryo in h), respectively. i′) shows a cross-section of the plane indicated by the vertical line in i). Asterisks mark the anterior end of the embryo. Dorsal is to the top in d to m). The dark color of the pigment spot (ps) and the frontal eye (fe) in k to m) is endogenous pigmentation, not an in situ hybridization signal. csg, club-shaped gland; cv, cerebral vesicle; en, endoderm; fe, frontal eye; g, gut; gs, gill slit; lhc, left head cavity; me, mesendoderm; nc, notochord; pnp, presumptive neural plate; nt, neural tube; ps, pigment spot; rt, rostral tip; so, somite; tb, tail bud.

Among gnathostomes, expression of some FoxP genes at early developmental stages has been described in Xenopus and zebrafish. Zebrafish FoxP1a is prominently expressed in the presumptive neural plate in the gastrula and also in the notochord and somites immediately after gastrulation (Cheng et al. 2007). In Xenopus, FoxP2 is expressed in the forebrain and the pharyngeal arches in the neurula; FoxP4 is expressed in the anterior neural plate at similar stages (Schön et al. 2006). These expression sites are comparable to the amphioxus FoxP at the gastrula and neurula stages.

During organogenesis, gnathostome FoxP1, FoxP2, and FoxP4 are broadly expressed in all 3 germ layers; their expression in the brain, spinal cord, heart tube, pronephros, limb/fin bud, pharyngeal arches, and gut was reported from various gnathostome species (Lu et al. 2002; Ferland et al. 2003; Tamura et al. 2003; Teufel et al. 2003; Teramitsu et al. 2004; Bonkowsky and Chien 2005; Pohl et al. 2005; Schön et al. 2006; Shah et al. 2006; Cheng et al. 2007). The expression of amphioxus FoxP in the cerebral vesicle and the neural tube is comparable to the expression of gnathostome FoxP1, FoxP2, and FoxP4 in the brain and spinal cord, given that these structures are considered homologous (Holland and Holland 1999; Holland 2009). In the gnathostome brains, FoxP paralogs are expressed in different patterns but with overlapping domains (Ferland et al. 2003; Teramitsu et al. 2004), indicating divergent roles for these paralogs in distinct cell populations. Therefore, duplication of FoxP genes may contribute to the compartmentalization and diversification of brain regions in vertebrates.

FoxP1 expression in spinal motor neurons also specifies the region-specific motor neurite branching pattern in tetrapods and the skate (Dasen et al. 2008; Jung et al. 2018). However, we did not observe FoxP expression specific to the motor neurons or motor neuron precursors in amphioxus. Thus, FoxP1 might have acquired a new role in specifying regional identities of spinal motor neurons in the stem gnathostome or proto-vertebrate.

We detected broad and prominent expression of FoxP in the pharyngeal and gut endoderm in amphioxus, similar to endodermal FoxP expression during embryonic development of sea urchin and mouse (Shu et al. 2001, 2007; Tu et al. 2006; Paganos et al. 2021). This finding suggests that endoderm expression of FoxP is likely an ancestral trait conserved among the deuterostomes. Mouse FoxP1, FoxP2, and FoxP4 are all highly expressed in the endoderm and are cooperatively required for endoderm development (Shu et al. 2001, 2007; Lu et al. 2002), indicating that the gnathostome FoxP paralogs have maintained overlapping functions in the endoderm after duplication. This is distinct from their apparent post-duplication functional diversification in neural development.

Comparative Transcriptome Analysis Reveals Ancient Dosage Subfunctionalization Among Gnathostome FoxP Genes

To further characterize the functional evolution of FoxP across 2R-WGD, we compared the expression levels of the amphioxus FoxP and the 4 gnathostome paralogs by analyzing the publicly available tissue-specific transcriptome datasets (see Materials and Methods section for detail). The relative abundance of transcripts is not directly comparable between different tissue samples, but these expression profiles are nevertheless useful as a first approximation of the tissue specificity of gene expression. In amphioxus and mouse, the expression profiles extracted from transcriptome datasets are largely consistent with in situ hybridization data here and elsewhere (Fig. 4a and d). While in situ hybridization data are not available for the spotted gar and the shark, transcriptomic profiles of their FoxP1, FoxP2, and FoxP4 suggest these genes are also broadly expressed (Fig. 4b and c), as in mouse. Moreover, we also noted that the expression levels of FoxP1, FoxP2, and FoxP4 in mouse tissues are generally higher during embryonic development than in the adult stage (Fig. 4d). This is consistent with the proposed role of the FH-family transcription factors as the pioneering factors responsible for opening chromatin for accessibility during development (Barral and Zaret 2024).

Fig. 4.

Transcriptome profiling of FoxP expression in amphioxus and representative gnathostome species. Heatmaps show the relative abundance of transcripts in tissue-specific transcriptomes of the amphioxus (Branchiostoma lanceolatum) FoxP a), the spotted gar (Lepisosteus oculatus) FoxP1-4 b), the catshark (Scyliorhinus torazame) FoxP1, 2 and 4 c), and the mouse (Mus musculus) FoxP1-4 d). Expression levels are shown in transcript per million (TPM).

In the same tissue type, the expression levels of FoxP paralogs varied widely (Fig. 4b and d). Notably, between the most recent duplicate pairs, one is always expressed at a higher level than its sister in all tissue types. In mouse and gar, FoxP1 transcript is more abundant than FoxP3, even in the immune organs, whereas FoxP4 is higher than FoxP2 (Fig. 4b and d). However, in the shark, FoxP2 is expressed at a higher level than FoxP4 in general (Fig. 4c), while FoxP3 is lost from the shark genomes (Venkatesh et al. 2014).

The parallel divergence in gene expression level between sister genes may be explained by dosage subfunctionalization, in which a constraint on gene dosage drives a divergence in the expression levels of sister genes following the duplication event (Gout and Lynch 2015). Dosage subfunctionalization is prevalent among recent duplicate pairs (Gout and Lynch 2015; Song et al. 2020) and is instrumental in dictating their evolutionary fates (Makino and McLysaght 2010; Gout and Lynch 2015; Wilson and Liberles 2023). Remarkably, the signature of dosage subfunctionalization among the FoxP duplicates remains detectable after more than 420 million years since their birth (dos Reis et al. 2015); the sustained dosage constraint imposed on the co-expressed FoxP paralogs in certain tissue types, probably endodermal, may help to maintain dosage subfunctionalization, while novel paralog-specific functions evolved in other tissue types to prevent gene loss. The inversed directions of dosage divergence in the FoxP2-FoxP4 pair between Chondrichthyes, shark specifically (Fig. 4c), and Osteichthyes, spotted gar and mouse specifically (Fig. 4b and d), imply that the evolutionary divergence of gene functions among the FoxP duplicates occurred independently in these 2 lineages.

FoxP3 and the Evolution of Vertebrate Immune System

FoxP3 is unique among the 4 FoxP paralogs in vertebrates. Our comparison of gene structure (supplementary fig. S1, Supplementary Material online) and phylogenetic analysis (Fig. 2) shows that it has a relatively divergent sequence content and is structurally derived. Unlike the other 3 broadly expressed paralogs, FoxP3 is uniquely involved in lymphocyte-based adaptive immunity, specifically for the development of CD4⁺CD25⁺ Treg cells (Fontenot et al. 2003, 2005; Hori et al. 2003; Gavin et al. 2007), although the co-expressed FoxP1 is required for FoxP3 functionality in Treg cells (Konopacki et al. 2019). The loss of FoxP3 and other Treg-specific genes in the shark genome led to the conclusion that sharks do not have Treg cells (Venkatesh et al. 2014). In contrast, FoxP3 is required for zebrafish Treg development (Quintana et al. 2010; Sugimoto et al. 2017), suggesting the Treg function of FoxP3 was already in place in the stem Osteichthyes, which gave rise to ray-finned fish, lobe-finned fish, and tetrapods. At face value, the loss of FoxP3 in the shark would have led to the conclusion that Treg was secondarily lost in this lineage. However, under the dosage subfunctionalization hypothesis, the paralog expressing at the lowest levels, which is FoxP3 in gnathostomes, is destined for gene loss if it does not acquire a nonredundant function before becoming pseudogenized (Gout and Lynch 2015). Therefore, FoxP3 may be preserved among Osteichthyes species only because of the emergence of the Treg cells at the base of this lineage, and thus, the absence of Treg cells in the shark may represent an ancestral condition for gnathostomes.

Since the divergence from their last common ancestor, cyclostomes and gnathostomes have evolved dramatically different adaptive immune systems (Cooper and Alder 2006). Gnathostome T cells mature in the thymus, one of the pharyngeal organs derived from outpockets of the anterior endoderm (Gordon and Manley 2011). Cyclostomes lack a well-defined thymus (Amemiya et al. 2007), but lamprey orthologs of genes essential for gnathostome T cell development, such as Delta and FoxN1, the latter representing a thymopoietic marker gene, are expressed in the gill basket epithelium (Bajoghli et al. 2011). Additionally, amphioxus FoxN1 and Delta are expressed in the pharyngeal endoderm and the anterior endoderm-derived club-shaped gland (Bajoghli et al. 2009), suggesting that these pharyngeal structures share a common origin with the gnathostome thymus. Therefore, the observed FoxP expression in the pharyngeal endoderm and club-shaped gland of amphioxus might be homologous to the thymopoietic expression of FoxP3 in the osteichthyan gnathostomes. Nevertheless, these data do not suggest that amphioxus has lymphocytes. Instead, it is more likely that the role of FoxP-related genetic program in the development of pharyngeal endoderm is ancient in chordates, and the program could have first co-opted for thymopoietic function in the stem vertebrate and further elaborated into Treg development in the stem osteichthyans. Future work on the expression and function of FoxP in the cyclostome immune system will be useful for reconstructing the evolutionary history of cell-mediated adaptive immunity in the vertebrate lineage.

Comparison of FoxP Expression Patterns Between the Ambulacrarians and the Protostomes Revealed the Ancestral Function of FoxP in Bilateria

Our comparison of gene expression patterns in amphioxus and gnathostomes suggests that FoxP had already exhibited a complex expression pattern in the last common ancestor of extant chordates. To infer the ancestral function of FoxP in deuterostomes, we included the sister clade of Chordata, Ambulacraria, in our comparative gene expression analysis. Among the ambulacrarians, expression patterns of FoxP were previously described for the sea urchin Strongylocentrotus purpuratus (Tu et al. 2006; Paganos et al. 2021) and the hemichordate Saccoglossus kowalevskii (Fritzenwanker et al. 2014). Our independent WMISH analysis of FoxP expression in the sea urchin S. purpuratus was largely consistent with previously published results (Tu et al. 2006; Paganos et al. 2021). Briefly, FoxP is first expressed in the vegetal plate in the mesenchyme blastula and early gastrula (supplementary fig. S6a to c, Supplementary Material online); at the late gastrula stage, it is highly expressed in the foregut and the oral ectoderm (supplementary fig. S6d to f, Supplementary Material online). The expression domain in the endoderm expands to the midgut and hindgut at the prism stage and persists in the pluteus larva (supplementary fig. S6g to j, Supplementary Material online). In a published single-cell transcriptome dataset from S. purpuratus larva (Paganos et al. 2021), FoxP transcript was also found in a subset of neurons and some mesoderm derivatives, including the esophageal muscles. Notably, we detected a weak signal in the apical ectoderm of late gastrula and pluteus larva (supplementary fig. S6d to g; S6j, Supplementary Material online). The apical ectoderm is one of the major sites of neurogenesis in the sea urchin embryo (Garner et al. 2016). Therefore, the FoxP-expressing cells in the apical ectoderm may include some neurons or neuron precursors. Our whole mount preparation does not provide enough spatial resolution to distinguish closely associated esophageal muscle precursors from the darkly stained foregut. Together, these data show that the sea urchin FoxP is most prominently expressed in the gut, but it may also be involved in the development of anterior neurons and mesoderm derivatives.

It has been reported that FoxP is broadly expressed in the embryonic development of the directly developing hemichordate S. kowalevskii (Fritzenwanker et al. 2014). We characterized the expression pattern of FoxP in Ptychodera flava, an indirectly developing hemichordate that produces larvae morphologically comparable to those of echinoderms (Röttinger and Lowe 2012; Su et al. 2019). In P. flava, no maternal transcript of FoxP was detected in the unfertilized egg (Fig. 5a). The transcript of FoxP was first detected ubiquitously at the blastula stage and became more prominent in the endoderm during gastrulation (Fig. 5b to e). At the late gastrula stage, FoxP was strongly expressed in the oral ectoderm, foregut, hindgut, and mesoderm; lower levels of FoxP expression were also detected in the apical ectoderm (Fig. 5f and g). In the early tornaria larva, the expression of FoxP could still be detected in the foregut and oral ectoderm, but the transcript level was decreased in the apical ectoderm and mesoderm (Fig. 5h). These results are similar to those in the sea urchin, suggesting that the developmental functions of FoxP are largely conserved among these indirectly developing ambulacrarian species.

Fig. 5.

Expression pattern of FoxP in hemichordate P. flava embryos. WMISH of FoxP in hemichordate P. flava at different developmental stages, including unfertilized egg a), blastula b), gastrula c to g), and tornaria larva h). The animal pole is at the top in a to h) and f′–h′). The specimens were viewed from the lateral, oral, or apical side as indicated on the left f to h). The oral side is to the left in f to h) and f″ to h″). hpf, hours postfertilization; ape, apical ectoderm; fg, foregut; hg, hindgut; me, mesoderm; oe, oral ectoderm.

Since FoxP is prominently expressed in the gut and neural structures in both ambulacrarians and chordates, these sites may represent deuterostome plesiomorphy regarding FoxP expression. Gut development has not been previously inferred as an ancestral function for FoxP because its expression and developmental function are restricted to the neural tissues in panarthropods (Janssen et al. 2022) and the nematode C. elegans (https://wormbase.org/species/c_elegans/gene/WBGene00001439). Together, these data suggest that the ancestral function of FoxP is in neural development in the ecdysozoan lineages. However, the FoxP transcript was enriched in subsets of neurons and endodermal glandular cells in a single-cell transcriptome dataset of the sea anemone Nematostella vectensis (Sebé-Pedrós et al. 2018). This observation alludes to the possibility that the gut-associated expression of FoxP may predate the last common ancestor of bilaterians.

To determine whether the expression in the developing gut is ancestral to bilaterians, it is necessary to examine FoxP expression in the third bilaterian superphylum, Spiralia. The expression and function of FoxP have not been broadly characterized in spiralians. In the polychaete Owenia, FoxP transcript was detected in the apical ectoderm of the blastula but not in the endoderm of the gastrula and larva (Seudre et al. 2022). It remains to be determined whether FoxP expression is turned on in the gut at later developmental stages in Owenia (Seudre et al. 2022). On the other hand, the planarian represents a peculiar case in which the FoxP homolog is expressed in the differentiating pigment cells and is required for body pigmentation (Wang et al. 2016). This finding deviates from all hypotheses concerning the ancestral role of FoxP proposed here and elsewhere. Here, we report the expression pattern of FoxP in a catenulid worm Stenostomum, representing the sister clade of the remaining extant members of Phylum Platyhelminthes (Egger et al. 2015; Laumer et al. 2015).

A FoxP homolog, which encodes for the FoxP protein-specific ZF-LZ domains (Fig. 6a) and the conserved FH DNA binding domain, was identified in the transcriptome of asexually growing worms. WMISH experiments revealed that Stenostomum FoxP is strongly expressed in the dorsal posterior region of the pharynx wall, a foregut component, and a bilateral pair of cells on the ventral side of the pharynx-intestine (or foregut-midgut) junction (Fig. 6b, c, e, and f). The identity of this pair of cells is unknown. Stenostomum undergoes asexual reproduction by paratomic fission (Child 1902). Structures signifying the anterior body region, such as the brain, cephalic sensory pits, mouth, and pharynx, develop in the middle of the body and give rise to a zooid head. Once reaching maturation, the zooid separates from the mother worm at the fission plane. A high level of FoxP transcript was detected in an unpaired pouch-like structure on the ventral side of a fission plane (Fig. 6b, d, and g). This pouch, probably arising from the lining of the gut wall, with the body wall invaginating toward it, would eventually give rise to the mouth and pharynx after zooid separation. Therefore, the FoxP-expressing tissue on the ventral side of the fission plane is likely the developing zooid pharynx. On the other hand, lower levels of FoxP expression were detected in a bilateral pair of body wall invaginations on the dorsal side of the fission plane (Fig. 6b). These invaginations give rise to the cephalic sensory pits, which are enriched with sensory cells and closely associated with the anterior neural elements—including the brain (Kepner and Cash 1915; Palmberg and Reuter 1992; Reuter et al. 1993; Gąsiorowski et al. 2023). We found that cells expressing Elav, a conserved marker for postmitotic neuron precursors across metazoan phyla (Robinow and White 1988; Kim et al. 1996; Benito-Gutiérrez et al. 2005; Marlow et al. 2009; Meyer and Seaver 2009; Nomaksteinsky et al. 2009), are situated in the deepest part of these dorsal invaginations (Fig. 6h), suggesting that the zooid brain and other anterior neural elements might have arisen from these invaginations. Low levels of FoxP signal were also detected in the mature brain of the mother worm (Fig. 6b). Based on these observations, we conclude that Stenostomum FoxP is expressed in the foregut and the anterior neural structure. The tissue types expressing FoxP in the asexually developing Stenostomum are surprisingly reminiscent of ambulacrarian embryonic development. These similarities suggest that FoxP expression in the foregut and anterior neural structure may be plesiomorphic for bilaterians; the peculiar role of planarian FoxP in pigment development is likely a planarian-specific apomorphic trait.

Fig. 6.

Characterization of FoxP in the catenulid worm Stenostomum. a) Amino acid alignment of the region corresponding to the FoxP-specific zinc finger and LZ domains. b to g) Expression pattern of Stenostomum FoxP in a worm undergoing asexual growth. The FoxP-expressing cells in the dorsal posterior region of the pharynx wall and the ventral body wall invagination at the mature fission plane are indicated with red and black arrows, respectively. Expression of FoxP was not detected in the newly developed fission plane (open arrow). The FoxP transcript signals in a pair of unidentified cells at the junction between the pharynx and intestine are denoted by black arrowheads. Lower levels of FoxP expression at the base of cephalic sensory pits (red asterisks) and the bilaterally paired dorsal body wall invaginations at the fission plane (black asterisks) are indicated. b) and e) show the lateral and dorsal views, respectively, of the same specimen. c) and f) are higher magnifications of the pharynx and intestine junction area from the lateral and dorsal view, respectively. d) and g) are higher magnifications of the mature fission plane in the posterior body from the lateral and the ventral view, respectively, focusing on the FoxP-expressing area in the ventral pouch that would give rise to the pharynx. h) Expression pattern of the pan-neuronal marker Elav. ph., pharynx; int., intestine.

Evolutionary Implications

This study yields 2 major findings. The first concerns the evolution of FoxP duplicates in gnathostomes. Here, we have elaborated the hypothesis that the 4 gnathostome FoxP paralogs arose from the 2R-WGD events in the early stage of vertebrate evolution through comparative genomics and phylogenetic analyses. In the case that the functional evolution of a gene is mainly driven by cis-regulatory divergence, following the paradigm that duplicated genes are preserved through subfunctionalization/specialization and neofunctionalization, the gnathostome FoxP paralogs may have acquired new expression domains or subdivided the ancestral expression domains. However, contrary to this prediction, our data suggest that FoxP already had a complex and dynamic expression pattern in the last common ancestor of chordates. Furthermore, except for gnathostome FoxP3, which is specifically involved in lymphocyte development, each of the gnathostome FoxP paralogs is expressed in a comparable, though not identical, pattern to that of the amphioxus FoxP and each other. Therefore, cis-regulation diversification may only be partially involved in preserving duplicated gnathostome FoxP paralogs.

Based on the comparison of gene expression patterns alone, we could not precisely pinpoint the mechanism by which the duplicated FoxP paralogs were retained. As discussed earlier, we have identified a potential role of dosage subfunctionalization in preserving the gnathostome FoxP duplicates. Our data also allude to a possible role for coding sequence divergence. Under the 2R-WGD scenario, the birth time of the 4 paralogs is identical; therefore, the branch lengths of the gnathostome FoxP lineages directly reflect their substitution rates. The asymmetric branch lengths among the paralogous FoxP-encoded proteins of gnathostomes (Fig. 3b) would imply differing levels of selection acting upon the coding sequence of these paralogs, and thus gnathostome FoxP duplicates may be partly preserved through the evolutionary divergence of protein functions. The differences in the expression pattern of gnathostome FoxP paralogs might have evolved later as the gnathostome body plan became increasingly complicated. Together, these evolutionary drives may factor differentially at different stages in the evolutionary history of gnathostome FoxP genes.

The second major finding is that, in addition to neural development, the ancestral function of FoxP may include the development of the gut, especially the foregut. Based on the comparison of expression patterns and functional data from Drosophila and vertebrates, it has long been postulated that the ancestral function of FoxP is in brain development, particularly that associated with cognitive function (Lawton et al. 2014; Mendoza et al. 2014; Castells-Nobau et al. 2019). In this study, we have expanded the taxonomic sampling of gene expression analysis to several invertebrate deuterostomes and a spiralian protostome. We showed that FoxP expression patterns are strikingly similar among these invertebrates with simple nervous systems, as their FoxP homologs are all found to be expressed in the anterior neurogenic ectoderm and the gut components. In addition, FoxP is generally expressed at a significantly higher level in the gut-associated tissues than in the anterior neurogenic ectoderm, suggesting that FoxP may have a more pronounced role in gut development than in neural development in these species. Therefore, we propose that the ancestral sites of FoxP expression might include both the anterior neural structure and gut tissue in Urbilateria.

Invertebrate anterior neurogenic ectoderm shares a common ancestry with the vertebrate forebrain (Lowe et al. 2003; Pani et al. 2012), and therefore, FoxP may have a general role in brain development in urbilaterian. In contrast, FoxP deficiency impairs cognitive behaviors but not the basic function and gross anatomy of the brain in mouse and fly (Mendoza et al. 2014; Bacon et al. 2015; Castells-Nobau et al. 2019; Medvedeva et al. 2019), suggesting that FoxP genes have specific and restricted roles in these derived complex brains. In the vertebrate and insect lineages, FoxP might have been independently recruited to the later-stage developmental processes required for the cognitive functions as an “add-on” when the brain underwent an evolutionary expansion.

Although the remarkable similarity in FoxP expression pattern between the cetanulid worms and the ambulacrarian larvae suggests deep conservation, it would require further expanding the taxonomic sampling, especially in marine spiralians with a larva stage in their life cycle, to further elaborate our hypothesis. In addition, the exact roles of FoxP in neural and gut development in these invertebrates are still unclear. Future loss-of-function and gain-of-function studies of FoxP in these species would help to address these questions. As we have pointed out earlier, FoxP was shown to be expressed in both neurons and glandular cells in Nematostella (Sebé-Pedrós et al. 2018). We noted that a developmental trajectory analysis of Nematostella single-cell transcriptome suggests that neurons and endodermal glandular cells may have a common origin (Steger et al. 2022). Taken together, we speculate that FoxP might be expressed in this hypothetical ancestral cell type that would eventually give rise to neurons and endodermal glandular cells and play a role in its development. The study of FoxP in earlier branching metazoans, such as sponges, may allow us to reconstruct the origin of these metazoan cell types.

Materials and Methods

Gene Identification

The cDNA sequences of FoxP genes from the amphioxus Branchiostoma floridae and the hemichordate acorn worm Ptychodera flava were obtained from the transcriptome database and were conceptually translated to protein sequences (Chen et al. 2014; Hu et al. 2017). The cDNA sequence of Stenostomum FoxP was retrieved by tblastn search of Stenostomum transcriptomes (SRA accession numbers: SRX872404, SRX951992). Genomic and amino acid sequences from other species were retrieved from the GenBank and Ensembl databases. Conserved functional domains were identified using SMART (Letunic and Bork 2018).

Genomic Structure and Synteny Analysis

The genome sequences were retrieved from the Ensembl Genome databases. The cDNA sequences of FoxPs were aligned to their respective genome sequences to map individual exons. The translated amino acid residues on a splice site were considered part of the preintron–exon. The homologous exons were operationally defined by an amino acid sequence identity of over 20%.

The syntenic genes around the FoxP loci in amphioxus, lamprey, spotted gar, and human were mapped using MCscan (Tang et al. 2008). The gene models downloaded from Ensembl or NCBI were subjected to a search by MCscan and were aligned using LAST designed for large-genome comparison (Kielbasa et al. 2011). The C-score threshold was set to 0.7; the distance was set to 100.

Chromosome-level genome assemblies of the scallop (Wang et al. 2017), the sea urchin (Arshinoff et al. 2022), the amphioxus (Simakov et al. 2020), and the spotted gar (Braasch et al. 2016) were used to search for macrosyntenies of FoxP-bearing chromosomes. Syntenic dot plots between species were generated using MCscan (Python version) of JCVI (version 1.2.7) (Tang et al. 2008; Wang et al. 2012). The C-score threshold parameter was set to 0.99 to retrieve the reciprocal best hits between the scallop, the sea urchin, and the amphioxus. For aligning amphioxus to multiple spotted gar sequences due to the 2-round WGD, the C-score threshold was set to 0.7. The minimum number of gene pairs was set to 1, and the restricted distance in a cluster was eliminated. Fisher's exact test with Bonferroni correction was employed to identify the corresponding chromosome pairs with a cutoff of adjusted P-value <0.05. The macrosynteny visualization only depicted orthologs located on the corresponding chromosome pairs.

The lamprey FoxP genes were mapped to the chromosome-level germline genome assembly of the sea lamprey Petromyzon marinus (Smith et al. 2018). The macrosynteny-informed phylogenetic history of the lamprey FoxP genes was reconstructed using a paralogon phylogeny of ALG_E-derived chromosomes in the cyclostome lineage (Marlétaz et al. 2024).

Phylogenetic Analysis

For the phylogenetic analysis of deuterostome FoxPs genes, the amino acid sequences of FoxP-encoded proteins in the sea urchin Strongylocentrotus purpuratus (Spu-FoxP, ABB89487.1), the hemichordate Saccoglossus kowalevskii (Sko-FoxP, NP_001158441.1), the ascidan Ciona intestinalis (Cin-FoxP, NP_001071939.1), the sea lamprey Petromyzon marinus (Pma-FoxPa, XP_032825384.1; Pma-FoxPb, XP_032822620.1; Pma-FoxPc, XP_032810180.1; Pma-FoxPd XP_032819089.1; Pma-FoxPx, XP_032827285.1), the spotted gar Lepisosteus oculatus (Loc-FoxP1, XP_015202674.1; Loc-FoxP2, XP_015208104.1; Loc-FoxP3, XP_006625452.1; FoxP4, XP_015198071.1), the frog Xenopus laevis (Xla-FoxP1, NP_001089002.1; FoxP2, NP_001089138.1; FoxP3, NP_001121199.1; FoxP4, NP_001089084.1), the chick Gallus gallus (Gga-FoxP1, NP_001019998.1; Gga-FoxP2, NP_001305342.1; Gga-FoxP4, XP_004948470.1), the mouse Mus musculus (Mmu-FoxP1, NP_001334274.1; Mmu-FoxP2, NP_444472.2; Mmu-FoxP3, NP_473380.1; Mmu-FoxP4, NP_001104294.1), and the human Homo sapiens (Hsa-FoxP1, NP_116071.2; Hsa-FoxP2, NP_055306.1; Hsa-FoxP3, XP_006724596.2; Hsa-FoxP4, NP_001012426.1) were retrieved from the NCBI Protein database. The amino acid sequences of FoxP-encoded proteins in the hemichordate Ptychodera flava (Pfl-FoxP) and the amphioxus Branchiostoma floridae (Bfl-FoxP) were obtained by conceptually translating the cDNA sequences identified in the transcriptome databases.

The amino acid sequences were aligned using the MUSCLE algorithm (Edgar 2004). Maximum-likelihood tree searches were performed using IQ-TREE (Nguyen et al. 2015) with 1,000 X ultrafast bootstrap (Hoang et al. 2018). The best-fit site-homogeneous substitution model was determined to be JTT + F + I + G4 by ModelFinder (Kalyaanamoorthy et al. 2017), and we used the site-heterogeneous model JTT + C60 + G for comparative analysis of substitution models. The approximately unbiased (AU) test (Shimodaira 2002), implemented with the -au option in IQ-TREE, was performed to determine the P-values for the alternative macrosynteny-informed tree topologies.

Transcriptome Analysis

The previously published tissue-specific transcriptome datasets of amphioxus (Marlétaz et al. 2018), shark (Hara et al. 2018), and spotted gar (Braasch et al. 2016) were retrieved from the SRA database (see supplementary file, Supplementary Material online for the list of SRA numbers). Relative abundance in transcript per million (TPM) was calculated for each FoxP gene in each transcriptome. The TPM counts for the mouse FoxP genes were retrieved by querying the GXD database (the adult dataset: E-GEOD-74747; the embryonic dataset: E-MTAB-6798). The relative expression levels in different tissue types were displayed in the heatmap format.

Animals

Amphioxus (Branchiostoma floridae), hemichordate acorn worms (Ptychodera flava), and sea urchins (Strongylocentrotus purpuratus) were collected in Tampa Bay, Florida, United States, Chito Bay, Penghu Islands, Taiwan, and California, United States, respectively. The spawning and culturing of the embryos were performed as previously described (Luo and Su 2012; Lin et al. 2016; Yong et al. 2019). The staging of amphioxus embryos followed (Carvalho et al. 2021). The catenulid Stenostomum was obtained from a breeding laboratory colony established with worms collected from Keelung, Taiwan.

Whole-Mount In Situ Hybridization

Polymerase chain reaction was performed to amplify the cDNA fragments using the following primer pairs: Bfl-FoxP (5′-ACAGGAAACAGCGACGACTG-3′ and 5′-CATGACGAGTCCTGCGTGTA-3′); Pfl-FoxP (5′-ATGGCCCAACAGCTTCATCA-3′ and 5′-CGTTCCGCCTGAAATATGCG-3′); Spu-FoxP (5′-GGGTCAGGTCCTTAGCCAAC-3′ and 5′-TGAAGACTGAGGTTGTGCCG-3′); Stenostomum FoxP (5′-CTCACGCTCATGCCGAATTG-3′ and 5′-CATGTAGCTTCGTTGCGTCG-3′). The cDNA fragments were subcloned into plasmid to serve as templates for riboprobe synthesis. The digoxigenin-labeled antisense riboprobes were synthesized from the linearized plasmids containing templates using T7 or Sp6 RNA polymerase.

Embryos of acorn worms, amphioxus, and sea urchins were collected and fixed at 4 °C overnight. Stenostomum was fixed with methanol briefly and then transferred to 4% paraformaldehyde in PBS for further fixation at room temperature for 1 h. The fixed specimens were washed and stored in methanol. The specimens were washed in PBS before being subjected to whole-mount in situ hybridization (WMISH).

WMISH was performed as previously described for each species (Lu et al. 2012; Luo and Su 2012; Ikuta et al. 2013). There is no published protocol for WMISH for Stenostomum; given the similarity in size and tissue complexity, it was performed following the same protocol for late-stage leech embryos (Weisblat and Kuo 2009). The chromogenic reaction for amphioxus and Stenostomum specimens was performed by using nitro-blue tetrazolium and 5-bromo-4-chloro-3-indolyl phosphate as the alkaline phosphatase (AP) substrate, whereas BM purple was used as the AP substrate for hemichordate and sea urchin embryos. The N1 stage amphioxus embryos were embedded in Tissue-Tek Optimal Cutting Temperature (O.C.T.) compound and cryosectioned at 10 µm after color reaction.

Digital micrographic images were taken using a Zeiss AxioCam MRc CCD camera mounted on a Zeiss Axio Imager A1 microscope or an Olympus DP72 camera mounted on an Olympus BX-63 microscope. Micrographic images are cropped and adjusted for brightness and contrast in ImageJ and Adobe Photoshop CS6.

Supplementary Material

Supplementary material is available at Molecular Biology and Evolution online.

Funding

This study was supported by grants from National Science and Technology Council, Taiwan to Y.H.S. (110-2311-B-001-031-MY3; 112-2326-B-001-004), J.K.Y. (110-2621-B-001-001-MY3), and D.H.K. (109-2621-B-002-001-MY3), and grant AS-GC-111-L01 from Academia Sinica, Taiwan to Y.H.S. and J.K.Y.

Data Availability

The data underlying this article are available in the article and its online Supplementary material. The corresponding authors will share further data upon reasonable request.