Hox-mediated endodermal identity patterns pharyngeal muscle formation in the chordate pharynx

The pharynx, with gill slits and the endostyle (or thyroid gland), is a defining chordate feature, whereas the existence of gill slits can be extended to deuterostomes. The chordate pharynx comprises multiple tissues, such as nerves, muscles and endodermal epithelia, that are well organized along the embryonic anterior-posterior (AP) axis. Hox genes are thought to play important roles in establishing the regional identity of the pharynx along the AP axis (Couly et al., 1998; Gendron-Maguire et al., 1993; Hunt et al., 1991; Rijli et al., 1993). In the pharyngeal endoderm, Hox genes also exhibit coordinated expression during vertebrate development (Shone et al., 2016). In mice, loss of Hoxa1 and Hoxb1 functions results in absence of the third pharyngeal pouch derivatives, such as the thymus and parathyroid (Rossel and Capecchi, 1999). However, the tissue-specific role of Hox genes in the pharyngeal endoderm is not well understood, except for Hoxa3, which is involved in the development of the third pouch-derived organs (Chojnowski et al., 2014). A reason for the scarce understanding of the tissue-specific role of Hox genes is related to the multiplicity of their functions; Hox genes are expressed and function in various organs including pharyngeal endoderm. Therefore, little is known about how the regional identities are established within this tissue by Hox genes and how the patterning is coordinated with those in other tissues.

We have previously found that Hox1 is expressed in the posterior part of the endostyle in the ascidian Ciona intestinalis (Sasakura et al., 2012). The endostyle is a mucus-producing organ formed on the ventral midline of the pharyngeal endoderm in non-vertebrate chordates and in larval lampreys, and is thought to be the precursor of the vertebrate thyroid gland (Salvatore, 1969). The ascidian endostyle is a representative structure of the pharynx and extends from the anterior to the posterior ends of the pharynx (Fig. 1K). Although Hox1 expression in the posterior endostyle suggests involvement of this gene in AP patterning of the pharynx, its function is unclear. Here, we report an essential role of Hox1 for establishing the AP axial identity of the endostyle in Ciona. We also find that the AP axial identity of the endostyle is necessary for the patterning of muscles in the pharynx, indicating that endodermal patterning is a key event for establishing the organized pharynx in chordates.

To explore the role of Hox1 in the endoderm, we carried out tissue-specific knockout of Hox1 using transcription activator-like effector nucleases (TALENs) (Treen et al., 2014). A TALEN pair designed to target the Hox1 locus was expressed in the endodermal tissues under the control of Titf1 regulatory sequence, which expresses downstream genes in the endoderm and a small subset of neural cells (Ristoratore et al., 1999; Sasakura et al., 2012). The endoderm-specific disruption of Hox1 allowed embryos to develop to juveniles with normal atrial siphons and gill slits, both of which are lost in Hox1 mutant animals (Sasakura et al., 2012). We first examined the expression of Otx and Hox1, which respectively mark the anterior and posterior endostyle as reported in another tunicate species (Cañestro et al., 2008) (Fig. 1A,E). Because TALENs generally introduce small deletions and/or insertions, expression of target genes usually remains in knockout animals, thereby allowing us to examine whether target genes are necessary for their own expression. Expression of Hox1 in the posterior endostyle was not detected when Hox1-TALENs were expressed in the endodermal tissues (Fig. 1B). By contrast, Otx was ectopically expressed in the posterior endostyle in addition to the anterior endostyle (Fig. 1F). These results suggest that the identity of the posterior endostyle is transformed into the anterior one in the Hox1 knockout juveniles, and that Hox1 is required to repress the anterior identity in the posterior endostyle. Next, we tested whether Hox1 is capable of suppressing the anterior identity. Overexpression of Hox1 in the endodermal tissues was carried out using the Titf1 cis-element (Titf1>Hox1). Expression of Otx was not detected in the anterior endostyle of Titf1>Hox1-electroporated animals (Fig. 1I), confirming that Hox1 represses the anterior identity in the endostyle.

It has been reported that Hox1 is expressed in the posterior trunk endoderm giving rise to the caudal pharynx at larval stages (Ikuta et al., 2004). To distinguish the role of Hox1 in the endostyle from that in the larval endoderm, we expressed Hox1-TALENs using an enhancer that drives gene expression in the endostyle after metamorphosis (Awazu et al., 2004). The stage-limited knockout of Hox1 recapitulated the ectopic expression of Otx in the posterior endostyle (Fig. S1). We also found that knockout of Hox1 in the whole endoderm does not affect the expression of Otx at the larval stage (Fig. S1). These results indicate that the alteration of the posterior endostyle identity observed in Hox1-TALEN-electroporated juveniles is not a secondary effect of its disruption in the larval endoderm.

Next, we examined the role of Otx for establishing anterior identity in the endostyle. Knockout of Otx in the endoderm resulted in loss of Otx expression in the anterior endostyle (Fig. 1G). On the other hand, Hox1 expression was expanded throughout the endostyle of juveniles electroporated with Otx-TALENs (Fig. 1C). These results indicate that Otx is required for establishing the anterior identity of the endostyle and repressing the posterior one. Then we carried out overexpression of Otx in the endoderm. Expression of Hox1 in the posterior endostyle was not detected when Titf1>Otx was introduced (Fig. 1J). Taken together, these results suggest that Hox1 and Otx play opposing roles in the organization of AP identity in the pharyngeal endoderm: Hox1 establishes the posterior identity of the endostyle and represses Otx expression and subsequent anterior identity, and vice versa.

Because Otx is expressed in the endodermal cells of cleavage-stage embryos (Hudson and Lemaire, 2001), we examined whether the phenotypes seen in Otx knockout juveniles are the result of the disruption of this gene in early embryogenesis. First, we did not detect mutations in the Otx locus of early gastrula embryos electroporated with Otx-TALENs, but mutations were detectable at the larval stage (Fig. S1), suggesting that function of Otx in early development was not disrupted by TALENs expressed by the Titf1 driver. We also found that expression of Hox1 is not altered in the larvae electroporated with Otx-TALENs (Fig. S1). These results support the notion that expression of Otx in the anterior endostyle is required for establishing the anterior identity of the endostyle.

We also found that morphology of the endostyle depends on the molecular identity. The anterior tip of endostyle comprises large columnar cells and shows protruded morphology, whereas the posterior tip has no such structure. In the Hox1-TALEN-electroporated juveniles, the posterior tip of the endostyle had the anterior-like morphology as well as expressing Otx (Fig. 1B,F; Fig. S1). On the other hand, the anterior tip became the posterior-like shape upon knockout of Otx (Fig. 1C,G; Fig. S1). These results indicate that morphology of the endostyle is determined by Hox1 and Otx.

We next carried out simultaneous knockout of Hox1 and Otx in the endoderm. In the double-knockout animals, Hox1 expression was not detected whereas Otx was expressed in both the anterior and posterior endostyle (Fig. 1D,H). This result suggests that Hox1 function is indispensable for Hox1 expression itself, and the default identity of the posterior endostyle is the same as the anterior one expressing Otx.

The default of anterior identity in the endostyle suggests that posterior identity with Hox1 expression might require induction from other tissues. We investigated the mechanism that induces Hox1 expression in the posterior endostyle. Retinoic acid (RA) is a well-known regulator of Hox1 in Ciona (Kanda et al., 2009; Nagatomo and Fujiwara, 2003) as well as in vertebrates (Marshall et al., 1996). In Ciona, RA is involved in patterning of the central nervous system and posterior epidermis (Imai et al., 2009; Pasini et al., 2012). RA is also known to regulate endodermal patterning in tunicates (Hinman and Degnan, 1998, 2000). We carried out RA administration assay using the transgenic line EJ[MiTSAdTPOG]124 (hereafter referred to as EJ124), which is an enhancer trap line that expresses GFP under the control of a Hox1 enhancer (Sasakura et al., 2012). GFP expression was detected throughout the endostyle of EJ124 juveniles cultured with RA after metamorphosis, suggesting that Hox1 expression in the endostyle is activated by RA signaling (Fig. 2A,B). To determine whether RA signaling is necessary to induce Hox1 expression in the posterior endostyle, we carried out endoderm-specific knockout of the retinoic acid receptor, RAR (Nagatomo et al., 2003). Expression of GFP in the posterior endostyle became detectable from 2 days post-fertilization (dpf) in control EJ124 animals (Fig. 2D). By contrast, GFP expression in the posterior endostyle was undetectable in EJ124 animals electroporated with TALENs for RAR (Fig. 2C). This suggests that RAR is indispensable for induction of Hox1 expression in the posterior endostyle. We next analyzed expression of Raldh2, which encodes the RA-synthesizing enzyme (Zhao et al., 1996). Raldh2 is expressed in a subset of tail muscle cells during embryogenesis (Nagatomo and Fujiwara, 2003). In addition, we found that Raldh2 is expressed in the posterior trunk endoderm of swimming larvae (Fig. S2). To test whether RA synthesized in the larval tissues is required to induce Hox1 expression in the endostyle, we carried out knockout of Raldh2 in muscle and endodermal tissues simultaneously. Expression of Hox1 in the posterior endostyle was diminished in animals electroporated with Raldh2-TALENs (Fig. S2), suggesting that RA signaling from larval muscle and endoderm induces Hox1 expression in the endostyle.

To examine whether maintenance of the Hox1 expression in the posterior endostyle depends on RA, we analyzed expression of Raldh2 in juveniles. Raldh2 was expressed in the posterior end of the endostyle in juveniles (Fig. 2I). This localized expression became evident after tail absorption (Fig. S2). Then we tested whether this localized Raldh2 expression depends on the posterior identity of the endostyle. Expression of Raldh2 was undetectable both in the Hox1 knockout and in the Otx-overexpressing animals (Fig. 2J,K). On the other hand, overexpression of Hox1 or knockout of Otx resulted in duplicated expression of Raldh2 in both the anterior and posterior endostyle (Fig. 2L,M). These results suggest that expression of Raldh2 in the posterior endostyle depends on Hox1-mediated posterior identity.

In order to test whether Raldh2 in the endostyle is required for maintenance of Hox1 expression, we carried out endoderm-specific knockout of Raldh2 using EJ124 animals. GFP expression in the posterior endostyle was observed in Raldh2-TALEN-electroporated animals as well as in controls at 2 dpf (Fig. 2E,F), suggesting that Raldh2 function in the endoderm is not necessary for initiation of Hox1 expression in the endostyle. However, at 6 dpf, GFP expression in the posterior endostyle became hardly detectable in the Raldh2 knockout animals (Fig. 2G,H). This indicates that RA synthesis in the posterior endostyle is required to maintain Hox1 expression in the posterior endostyle. Taken together, these findings suggest that the RA-Hox1 positive-feedback loop is necessary to establish the posterior identity of the endostyle: RA signaling from surrounding tail muscles and the posterior trunk endoderm can initiate Hox1 expression in the posterior endostyle region; in turn Hox1 upregulates Raldh2 expression in the endostyle, and RA synthesized in the posterior endostyle maintains Hox1 expression by the positive feedback (Fig. 2N). These data also suggest that the broader expression of Hox1 throughout the endostyle observed in Otx knockout juveniles (Fig. 1C) is likely to be due to excess production of RA in both the anterior and posterior endostyle (Fig. 2M). We examined expression of Raldh2 in larvae electroporated with TALENs targeting Hox1, Otx or RAR. Expression of Raldh2 in the posterior trunk endoderm was unaffected by knockout of these genes (Fig. S2). This suggests that the RA-Hox1 and Otx genetic network establishes the AP identity of pharyngeal endoderm during post-larval development.

In order to address the significance of the regional identity of the endostyle for the development of the pharynx, we analyzed the formation of the body wall muscles (BWMs), which develop along the AP axis of the pharynx. Ascidian BWMs, as well as atrial siphon muscles, are specified through a shared genetic program with vertebrate pharyngeal muscles (Diogo et al., 2015; Stolfi et al., 2010). BWM precursor cells are located in the atrial siphon primordium before metamorphosis and BWMs elongate posteriorly towards the endostyle during metamorphosis (Stolfi et al., 2010) (Fig. 3A). Formation of BWMs was observed by expression of Myosin heavy chain 3 (Mhc3) gene, which marks the oral and atrial siphon muscles and BWMs (Stolfi et al., 2010). We found that BWMs failed to elongate towards the posterior endostyle when Hox1 was knocked out in the endoderm. In Hox1 knockout juveniles, approximately 45% of observed BWMs displayed misdirected elongation (Fig. 3B) and 29% of them did not show elongation (Fig. 3C). These results suggest that the posterior identity of the endostyle is required for the directional elongation of BWMs towards the posterior endostyle. Because the endostyle and the atrial siphon, where precursors of BWMs are situated, are not adjacent to each other, there must be a signaling factor that is downstream of Hox1 and promotes the directional elongation of BWMs. RA is a feasible candidate factor to promote the posterior elongation of BWMs. To test this possibility, we examined the BWM formation in the animals treated with an RA synthesis inhibitor, citral (Kanda et al., 2009; Marsh-Armstrong et al., 1994). In the citral-treated condition, BWMs seemed to start the posterior elongation process but failed to complete it (Fig. 3D,E). Then we observed BWM formation in juveniles electroporated with Raldh2 TALENs. Knockout of Raldh2 in the endodermal tissue disrupted the elongation of BWMs (Fig. 3F,G), suggesting that RA synthesized in the posterior endostyle is required for posterior elongation of BWMs. Next, we treated juveniles with RA and analyzed BWM formation. In RA-treated animals, elongated BWMs were not observed, suggesting that ubiquitous RA input inhibits the elongation or differentiation of BWMs (Fig. S3). In order to clarify whether BWMs are not formed or failed to elongate by RA treatment, we carried out live imaging of BWM formation by expressing Kaede fluorescent protein in the BWMs with a cis-regulatory region of Mhc3. This imaging analysis revealed that BWMs made ectopic protrusions in various directions but the muscles failed to elongate in the posterior direction upon RA administration (Fig. 3H,I). These results further support the idea that RA synthesized in the posterior endostyle is necessary for directional elongation of the BWMs towards the posterior endostyle. It is also possible that there is another factor that regulates directional elongation of BWMs, because citral treatment or Raldh2 knockout disrupted posterior elongation of BWMs but did not recapitulate the misdirected elongation observed in Hox1 knockout animals. To understand fully the mechanism underlining the directional elongation of BWMs, identification of the factor(s) downstream of Hox1 and/or Otx is necessary.

This study shows that in C. intestinalis, the anterior and posterior ends of the endostyle, a representative structure of pharyngeal endoderm, have distinct identities established by Otx and Hox1 transcription factors, respectively. In chordates, these factors are expressed in the pharyngeal endoderm during embryogenesis. In mouse development, Otx2 is expressed in the first arch endoderm (Ang et al., 1994), and Hoxa1 and Hoxb1 are expressed in the caudal pharynx in an RA- and Raldh2-dependent manner (Niederreither et al., 2003; Wendling et al., 2000). In the cephalochordate amphioxus, Hox1 is expressed in the endoderm under the control of RA and represses Otx expression, which determines the posterior limit of the pharynx (Schubert et al., 2005). Therefore, the AP patterning mechanism of the Ciona endostyle, involving Otx and RA-Hox1, might have utilized the shared genetic mechanism that patterns the AP axis of the early pharyngeal endoderm in chordate development.

In vertebrate pharyngeal development, patterning and differentiation of pharyngeal muscles are regulated by cranial neural crest cells (NCCs), which give rise to skeletal elements and tendons associated with muscles (Noden and Trainor, 2005; Rinon et al., 2007). Our present study demonstrates that proper formation of the BWM, a musculature surrounding the pharynx, is dependent on the regional identity of the endodermal organ. This suggests that in Ciona, which has no definitive NCC, it is primarily the endoderm that is responsible for organizing the pharynx along the AP axis.

Wild-type Ciona intestinalis was cultivated at Maizuru (Kyoto), Misaki (Kanagawa), Mukaishima (Hiroshima) and Usa (Kochi). The transgenic line of the Hox1 enhancer trap, EJ[MiTSAdTPOG]124, was maintained by an inland culture system (Joly et al., 2007).

TALENs were assembled using the four-module golden gate method (Sakuma et al., 2013). The activity of the constructed TALENs were estimated by previously described methods (Fig. S4) (Treen et al., 2014). Details of the construction of expression vectors and electroporation procedures are described in supplementary Materials and Methods.

Genomic DNAs were isolated from early gastrula embryos and swimming larvae developed from eggs, into which the Titf1>Otx-TALEN pair or the left side of Otx-TALEN (control) was electroporated, using the Wizard genomic DNA purification kit (Promega), and the genomic region including the target site of the Otx-TALEN pair was amplified by PCR. The PCR bands were analyzed by heteroduplex mobility shift assay with polyacrylamide gel electrophoresis (Ota et al., 2013) to examine their heterogeneity of the sequence that reflects the presence of the mutated gene.

Preparation of stock solutions of RA and citral was carried out as described previously (Kanda et al., 2009). Metamorphosing larvae were treated with 1 μM all-trans RA, 0.1% dimethylsulfoxide, 20 μM citral or 0.1% ethanol from 45 hours post-fertilization (hpf), which is prior to the onset of BWM elongation, to 62 hpf, after the posterior elongation of BWMs was complete.

Whole-mount in situ hybridization (WISH) was performed basically according to the previous study (Yoshida and Sasakura, 2012). Details of WISH and imaging are described in supplementary Materials and Methods.

We are grateful to the members of the Shimoda Marine Research Center at the University of Tsukuba for maintenance of transgenic animals and to Drs Shigeki Fujiwara, Kunifumi Tagawa, Nobuo Yamaguchi and all members of the Department of Zoology, Kyoto University, the Misaki Marine Biological Station, the University of Tokyo, and the Maizuru Fishery Research Station of Kyoto University for the cultivation and provision of wild-type Ciona adults. This study was further supported by the National BioResource Project, Japan.