Contrary to previous observations of E. stoutii, we did not observe a trough-like structure during any of the developmental stages of E. burgeri (Fig. 2 and Additional file 1: Fig. S2E-G). Instead, we found a putative thyroid anlage in the form of a thickened epithelium on the pharyngeal floor juxtaposed to the aortic sac during the early pharyngula stage [17, 23] (at stage 45; Fig. 2A–C’), which is considerably earlier than the stage wherein the trough-like structure was observed by Stockard (at stage 53) [15]. Since the ventral pharynx of the hagfish embryo is rostroventrally everted in early development, the position of the hagfish thyroid anlage is difficult to identify based on the level of the pharyngeal arches [10]. Nevertheless, this thyroid anlage arose from a focal region of the pharyngeal floor slightly posterior to the root of the lower lip anlage and did not extend throughout the entire gill region as previously reported [15] (Fig. 2B, B’, C). The collection of hagfish embryos prepared by Bashford Dean, which was subsequently used in Stockard’s histological studies, showed a frequent distortion of embryonic tissues [24, 25]. The trough-like organ described by Stockard was not a consistent observation; thus, it was likely an artifact due to flawed fixation.
The position of the thyroid primordium in the hagfish embryo from stages 45 to 51 was continuously associated with a bifurcation site of the ventral aorta probably including the precardiac mesoderm (Fig. 2C’, F’, I’). In gnathostomes, the thyroid develops from an endodermal thickening of the pharyngeal floor in close vicinity to the aortic sac [3, 19], as was observed in the cloudy catshark (Scyliorhinus torazame; Fig. 3X; Additional file 1: Fig. S3A’-E’). Furthermore, in mice, frogs, and zebrafish, extrinsic factors, such as FGF2 and BMP4, presumably secreted from the precardiac mesoderm, have been shown to induce thyroid fate [19, 26]. Unlike the initiation of the thyroid development in hagfish and gnathostomes, evagination of the lamprey endostyle primordium from the midline pharyngeal endoderm occurs from the 2nd to 4th pharyngeal arches, and only the posterior end of the endostyle is in contact with the bifurcation site of the aortic sac [2] (Fig. 3W). Thus, the comparable topological relationships of the developing thyroid and the precardiac mesoderm, a possible specifier of thyroid origins, between hagfish and gnathostomes implies that the direct development of the thyroid gland may represent an ancestral condition in the LCA of modern vertebrates.
In hagfish, we found that the pharyngeal region co-expressing thyroid marker genes (Nkx2-1/2-4, Pax2/5/8, and Hhex) was confined to the thyroid anlage in stage 51 and 53 embryos (Fig. 3B, C, E, F, H, I), which is similar to the expression patterns observed in gnathostomes [19, 20] (Additional file 1: Fig. S3F-J). Co-expression of Nkx2-1/2-4, Pax2/5/8, and Hhex in hagfish and gnathostomes suggests that the conserved GRN for thyroid follicle formation was present in the LCA of modern vertebrates. However, unlike gnathostomes, in which Nkx2-1, Pax8, and Hhex are co-expressed in the earliest thyroid anlage [19], only Nkx2-1/2-4 was detected in the early primordia of the hagfish thyroid (stage 45; Fig. 3A’, D’, G’). Moreover, its expression domain extended rostrally towards the endodermal epithelium, up to the level of the oropharyngeal membrane, and was not confined to the prospective thyroid anlage (Fig. 3A, A’, V). In L. camtschaticum at Tahara’s stage 24 [27], the endostyle diverticulum was distinguishable by the broad medial expression of Nkx2-1/2-4A and Nkx2-1/2-4C genes in the pharynx (Fig. 3K, K’ and Additional file 1: Fig. S4J, J’), as previously reported [1, 2, 25]. Similar to the expression patterns in the hagfish, only the hybridized signal of Nkx2-1/2-4 paralogues was detected in the ventral pharynx of the lamprey before the initiation of endostyle formation (Tahara’s stage 23; Fig. 3J and Additional file 1: Fig. S4I). The spatiotemporal expression patterns of Nkx2-1/2-4 shared by hagfish and lamprey are similar to those of orthologous genes during the endostyle development in non-vertebrate chordates [13]. Furthermore, even in the hemichordate embryo which does not have an endostyle, the expression of Nkx2-1/2-4 is evident in the pharyngeal endoderm [28]. Nkx2-1 is focally expressed in the laryngotracheal diverticulum, in addition to the thyroid primordium; the laryngotracheal diverticulum is situated in the most caudal part of the pharynx during early development in mammals and birds [29, 30]. However, the lung and its homolog (the swim bladder) are presumed to be a synapomorphy of Osteichthyes (bony vertebrates), and thus, the expression domain of Nkx2-1 in these respiratory organs was likely acquired after Osteichthyes and Chondrichthyes split from each other and likely does not reflect an ancestral condition of gnathostomes. Thus, our findings suggest that the broad expression of Nkx2-1/2-4 orthologs in the ventral pharyngeal endoderm during the earliest stages of thyroid/endostyle development represents a plesiomorphic trait of chordates, which has been retained in cyclostomes, and may have been secondarily modified in crown gnathostomes where the Nkx2-1/2-4 expression domain was not broad and regionally confined to the thyroid primordium [19].
We identified an additional Pax2/5/8 paralogue in cyclostomes, termed Pax2/5/8B (Additional file 1: Fig. S4, S5). Expression signals of the lamprey Pax2/5/8A and Pax2/5/8B were not detected in the endostyle before stage 25 but were evident at stage 26 when the endostyle was almost differentiated (Fig. 3N–Q’ and Additional file 1: Fig. S4M-P). At stage 24, Hhex was expressed bilaterally at both ends of the diverticulum (Fig. 3S’). All thyroid marker genes were expressed in the endostyle of lamprey larvae only after stage 26 (Fig. 3M, Q, and U and Additional file 1: Fig. S4H, L, P), which were consistent with the previous studies [25, 31], but the expression of Pax2/5/8A and Hhex did not extend to the entire endostyle anlage in cross-sectional view (Fig. 3Q’, U’). The distributions of thyroid-related genes including Pax2/5/8 and Nkx2-1/2-4 orthologs in adult amphioxi and tunicates similarly do not extend to the entire endostyle in cross-sectional view. For example, the Nkx2-1/2-4 ortholog is not expressed in the thyroid-equivalent elements in the ascidian Ciona intestinalis, and the Pax2/5/8 ortholog is not expressed in part of the supporting elements in the cephalochordate Branchiostoma belcheri [12]. As co-expression of Nkx2-1/2-4, Pax2/5/8, and Hhex and their interactions are essential for the development of the thyroid gland in gnathostomes [19], the heterochronic, and heterogeneous expression of Nkx2-1/2-4, Pax2/5/8, and Hhex in the lamprey endostyle suggests that the core set of thyroid transcription factors, except for Nkx2-1/2-4, is not required for the endostyle development with regard to their organogenesis.
The expression of Nkx2-1/2-4, Pax2/5/8, and Hhex [32] orthologs in the endostyle of chordates, including the modern lampreys, may be responsible for thyroidal functions but not for its development. In the mammalian thyroid, Nkx2-1 and Pax8 genes are known to be involved both in organogenesis and physiological function [19, 21]. Thyroglobulin (Tg) and thyroid peroxidase (TPO) play critical roles in thyroid hormone synthesis, and the gene expression of Tg and TPO are coordinately regulated by Nkx2-1 and Pax8 in the thyroid [21, 33]. Likewise, TPO orthologs are co-expressed with Nkx2-1 and Pax8 in the endostyle of amphioxus and with Pax8 in that of tunicates, suggesting that these regulatory genes for thyroidal function are conserved among modern chordates [12, 13]. Thus, the regionally confined expression patterns of Nkx2-1/2-4, Pax2/5/8, and Hhex in the endostyles of the lamprey and non-vertebrate chordates imply that these transcription factors do not orchestrate a regulatory network that forms either the endostyle or thyroid follicle. Indeed, the thyroid follicles never occur prior to metamorphosis in the modern lampreys and throughout life in non-vertebrate chordates. In other words, our findings suggest that the developmental GRNs for the endostyle and thyroid organogenesis have an independent and distinct modular nature.
The life histories of hagfish and lamprey differ conspicuously. While all modern lampreys go through a larval phase for approximately 2–8 years and reach maturity only after metamorphosis [34], modern hagfishes develop into the adult morphology directly after hatching [18]. Recently, Miyashita et al. reported that fossil stem lampreys probably lacked the filter-feeding larval phase, since fossils of yolk-sac lampreys were found with prominent eyes and a circumoral feeding apparatus [35]. The fossil evidence suggested that they may have directly developed into macrophagous predators following the yolk-dependent lecithotrophic period. Although the presence of a rudimentary endostyle in these stem lampreys cannot be ruled out, these observations cast doubt on the prevailing view that the ammocoete larva best resembles the earliest vertebrates [36, 37]. In the context of thyroid evolution, the lack of a larval phase in stem lampreys suggests an alternative hypothesis, namely, that the primitive endostyle (which is retained in modern non-vertebrate chordates) was lost in the LCA of vertebrates, and the endostyle present in modern lampreys represents secondary acquisition.
In the present study, we found that both gnathostomes and hagfish directly developed a thyroid. There is currently no evidence that ancestral species possessed an endostyle in the branches leading to these taxa. Considering the direct development (without metamorphosis) of stem lampreys [35], our findings further raise the possibility that the endostyle was acquired secondarily in crown lampreys. Thus, we propose the following alternative evolutionary scenario for the vertebrate thyroid and endostyle (Fig. 4). The broad expression of Nkx2-1/2-4 in the pharyngeal endoderm shared among extant cyclostomes, non-vertebrate chordates, and hemichordates represents a primitive trait of deuterostomes, and the expression domain was restricted to the ventral side of the pharynx in early chordates. A GRN involving unidentified transcription factors, instead of Pax2/5/8 and Hhex, which induces the formation of the primitive endostyle, may have first occurred downstream of Nkx2-1/2-4 (Fig. 4A). Before the acquisition of the follicular thyroid, Pax2/5/8 and Hhex orthologs were probably not coupled with Nkx2-1/2-4 and thus not involved in the endostyle or thyroid GRN, as observed in extant non-vertebrate chordates [12] (Fig. 4A). Subsequently, Nkx2-1/2-4, Pax2/5/8, and Hhex were organized in a colocalized expression domain, thereby establishing an interactive GRN to produce thyroid follicles in the earliest vertebrates (Fig. 4B). The acquisition of the thyroid and the subsequent loss of primitive endostyle occurred prior to the divergence of gnathostome and cyclostome lineages (Fig. 4C). An endostyle may have evolved secondarily at the base of crown lampreys in association with the filter-feeding larval stage. The cyclostome-specific (probably ancestral) broad and early expression domain of Nkx2-1/2-4 in the pharyngeal floor (Fig. 4D) may have facilitated the co-option of an ancestral endostyle GRN.
Stem groups of vertebrates, such as Myllokunmingia, are considered to be microphagous suspension-feeders like modern non-vertebrate chordates [38], suggesting that the shared developmental mechanisms of the primitive endostyle were retained in these animals. Meanwhile, according to the alternative hypothesis, the fossil lamprey (Mesomyzon mengae) from the early Cretaceous has an ammocoete larval phase [39], suggesting that the endostyle in the modern lamprey lineage was likely acquired after the Palaeozoic.
The genetic basis of developmental mechanisms underlying endostyle formation remains largely unknown, and comparative single-cell transcriptomic analyses of the developing endostyles of non-vertebrate chordates and larval lampreys may elucidate similarities and differences in the GRN involved in morphogenesis. Furthermore, gene expression profiles in the endostyle of metamorphosing lampreys may also provide insights into the mechanisms of how the thyroid GRN is induced during the development. The novel hypothesis that the crown lamprey secondarily acquired the endostyle thus warrants further testing via future studies. According to the classical scenario, the independent nature of the GRN of the primitive endostyle from the thyroid gland may have enabled the loss of its morphology without influencing thyroid development in the respective ancestors of gnathostomes, hagfishes, and possibly some stem lampreys (Fig. 4C’). Alternatively, the LCA of modern vertebrates may have lost the primitive endostyle following the acquisition of thyroid follicles (Fig. 4C).