Regulation of anterior neurectoderm specification and differentiation by BMP signaling in ascidians

Ascidians (or sea squirts) belong to a group of marine invertebrates, the tunicates, that is the sister group of vertebrates (Delsuc et al., 2006). This phylogenetic position, together with a stereotyped embryonic development with few cells, puts ascidians as interesting models for developmental biology and comparative approaches to address questions regarding chordate evolution and the emergence of vertebrates. Ascidians have a biphasic life cycle: following external development, the embryo gives rise to a swimming tadpole-like larva with typical chordate features (notochord, dorsal neural tube) that then attaches to a substrate before metamorphosing into a sessile adult ascidian with a radically different body plan, a ‘bag’ with two siphons. Metamorphosis is controlled by a specific organ, the palps (also referred to as the adhesive organ or the adhesive papillae), which is located at the anterior end of the larva. The palps are a specialized part of the ectoderm that has adhesive and sensory properties (Cloney, 1977; Imai and Meinertzhagen, 2007; Satoh, 1994). They enable the larva to select a suitable substrate for metamorphosis, via a chemo- and/or mechano-sensory function, and to attach to it through the secretion of adhesive materials (reviewed by Pennati and Rothbächer, 2015). The palps contain at least four cells types for which the specification and function have not yet been deciphered in detail (Johnson et al., 2020; Zeng et al., 2019). Three cell types – the ciliated sensory neurons, the collocytes (containing vesicles filled with adhesive material) and the axial columnar cells (ACCs) (myoepithelial cells controlling palp retraction following adhesion) – are elongated cells forming a protrusion. Three protrusions or papillae (two dorsal papillae that are bilaterally symmetrical, and a ventral papilla located at the midline; Fig. 1C) make up the palps and are separated by the fourth cell type, the non-elongated inter-papillae cells.

Palps belong to the peripheral nervous system and studies of palps have been instrumental in the proposal of evolutionary scenarios relating to the nervous system in chordates (Cao et al., 2019; Horie et al., 2018; Poncelet and Shimeld, 2020; Thawani and Groves, 2020). In the ascidian Ciona intestinalis, palp cell lineage and topology, together with gene expression data and functional studies, have shown to have similarities to anterior derivatives of the vertebrate nervous system, the olfactory placodes and the telencephalon (Cao et al., 2019; Horie et al., 2018; Hudson et al., 2003; Liu and Satou, 2019; Poncelet and Shimeld, 2020; Thawani and Groves, 2020; Wagner and Levine, 2012; Wagner et al., 2014). Palps originate from precursors that are located at the anterior edge of the neural plate during gastrulation, which we will refer to as the anterior neural border (ANB) (Fig. 1W,X) (Horie et al., 2018; Liu and Satou, 2019). Although the ANB is not part of the central nervous system (CNS), it originates from the same lineage specified by fibroblast growth factor (FGF)-mediated neural induction at the 32-cell stage and expresses neural markers such as Celf3/4/5/6 (also known as Etr and Celf3.a) and Otx (Horie et al., 2018; Hudson, 2016; Hudson et al., 2003; Nishida, 1987). The separation between these two lineages is regulated by FGF/Erk signaling at gastrula/neurula stages, FGF being active in the CNS precursors (Hudson, 2016; Hudson et al., 2003; Wagner and Levine, 2012). FGF signaling thus regulates positively and negatively two separate phases of palp specification. The ANB also expresses Dmrt and Foxc, coding for transcription factors that are essential for palp formation (Imai et al., 2006; Wagner and Levine, 2012). From neurulation and through differentiation, palps express genes such as Dlx.c, Foxg, Isl or Sp6/7/8/9 (also known as Zf220 and Btd) orthologs of which specify anterior neural territories and placodes in vertebrates (Cao et al., 2019; Liu and Satou, 2019; Wagner et al., 2014). In particular, Foxg and Isl are essential for palp formation (Liu and Satou, 2019; Wagner et al., 2014). The ANB thus shares similarities with vertebrate anterior cranial placodes, and the palps share similarities with derivatives of the vertebrate telencephalon, such as the olfactory bulb, and of the anterior placodes. It has been proposed that co-option of ANB/palp gene network to the anterior CNS led to the emergence of the vertebrate telencephalon (Cao et al., 2019).

Although knowledge of transcription factor functions and interactions in palp formation has been elucidated in some detail (Horie et al., 2018; Liu and Satou, 2019; Wagner et al., 2014), the role of cell–cell communication is scarce except for the involvement of FGF/Erk pathway (Hudson et al., 2003; Wagner and Levine, 2012). We have previously shown that inhibition of canonical Wnt pathway is essential for ANB specification (Feinberg et al., 2019), but an understanding of the role of Wnt signaling later in development is still lacking. In a distantly related ascidian species, Halocynthia roretzi, morphological data indicate that activating the BMP pathway abolishes palp formation whereas BMP inhibition results in palps made of a single protrusion instead of three (Darras and Nishida, 2001), but the lack of molecular analysis prevents us from precisely determining the function of BMP signaling.

We have directly addressed the function of BMP signaling pathway in palp formation during embryogenesis of the ascidian C. intestinalis. We show that BMP is involved in two consecutive phases. Up to neurulation, ANB specification is incompatible with active BMP signaling, and the ANB forms in a region devoid of active BMP (as revealed by phospho-Smad1/5/8 immunostaining). Consequently, early activation of BMP prevents palp formation through the inhibition of ANB precursor formation. Following gastrulation, BMP participates in the differentiation of the palps through the specification of the ventral papilla and the regulation of the papillae versus inter-papillae fate decision. In particular, BMP-inhibited larvae harbor a single, large protrusion made of elongated cells, the Cyrano phenotype, with an increased number of sensory neurons and ACCs. We propose that the competence to become a papilla is regulated by BMP through the transcription factor-coding genes Foxg and Sp6/7/8/9. Interestingly, we show that modulating BMP pathway in the ascidian Phallusia mammillata (275 My of divergence time) produces the same phenotypes as in C. intestinalis. This allowed us to use previously published RNA-seq data (Chowdhury et al., 2022) to identify a number of genes expressed in the ANB and the palps. Altogether, our work points to a role for signaling pathway inhibition in ANB specification, similar to early anterior neurectoderm formation in vertebrates. Moreover, we provide a significant contribution to knowledge of the palp gene network, an essential requisite to probing its conservation with the networks regulating cranial placodes and telencephalon formation in vertebrates.

When we activated the BMP signaling pathway by overexpressing, in the ectoderm, the BMP ligand Admp by electroporation using the pFog driver (active from the 16-cell stage) (Pasini et al., 2006; Rothbacher et al., 2007), we observed an absence of protrusions that are obvious features of the adhesive palps. The anterior end of the larvae were smooth, and epidermal cells were flat and did not display the typical elongated shape (Fig. 1A-D). This morphological evidence was accompanied by repression of the expression at mid-tailbud stages of all the genes expressed in the palps that we examined: Sp6/7/8/9, Isl, Foxg, Celf3/4/5/6, Pou4 and Emx (Fig. 1E-P). Palps derive from the median ANB at gastrula stages (Fig. 1W,X) and can be tracked by the expression of genes essential for palp formation, Foxg at neurula stages (Fig. 1Q,R) and Foxc at gastrula stages (Fig. 1U,V) (Liu and Satou, 2019; Wagner and Levine, 2012). Both genes were repressed by BMP activation, and this repression is sufficient to explain the later lack of palp gene expression and differentiation. Interestingly, palps were not converted into general epidermis as the epidermal marker Sox14/15/21 (also known as SoxB2) was normally expressed and did not show ectopic expression in the palp area (Fig. 1S,T).

The above results suggest that active BMP signaling is incompatible with palp formation. Active BMP signaling can be determined by examining the phosphorylated (active) form of the BMP transducer Smad1/5/8. It has been previously shown that BMP is active from late gastrula to early tailbud stages in the ventral epidermis midline of Ciona robusta embryos (Waki et al., 2015). We obtained similar results in C. intestinalis using a different antibody (Fig. 2). More specifically, up to gastrulation, we did not detect significant levels for P-Smad1/5/8 except in a few posterior endomesodermal cells (Fig. 2A). At mid-gastrula, P-Smad1/5/8 was present in the nuclei of the posterior (b-line) ventral midline epidermis (Fig. 2B). Consequently, at the onset of Foxc expression in palp precursors (St. 10), BMP was not active in the palp-forming region (Fig. 2L). Shortly after, at early neurula stages, P-Smad1/5/8 extended into the anterior (a-line) ventral midline epidermis (Fig. 2C). During neurulation, the posterior limit of P-Smad1/5/8 gradually shifted anteriorly, in agreement with the dynamic posterior-to-anterior expression of candidate target genes (Roure and Darras, 2016); by mid-tailbud stages, P-Smad1/5/8 was restricted to the trunk ventral epidermis (Fig. 2D,E,J). In addition, active Smad1/5/8 was also detected in the endoderm underlying the ventral epidermis midline with a similar temporal dynamic (Fig. 2E,F,H,J,K), and in a group of cells of the anterior sensory vesicle at mid-tailbud stages (Fig. 2E,J,K). We validated the specificity of these results by modulating the BMP pathway: when embryos were treated with BMP2 protein, P-Smad1/5/8 was ectopically detected in the entire epidermis at early neurula stages, whereas no staining was observed following inhibition using the pharmacological inhibitor DMH1 (Fig. S1).

To relate precisely the location of active signaling in the ectoderm and palp precursors, we performed double staining: P-Smad1/5/8 immunostaining and in situ hybridization for the early palp markers Foxc and Foxg (Fig. 2F-J). At late gastrula stages, P-Smad1/5/8 abutted Foxc expression domain, confirming that palp precursors were specified in a BMP-negative domain (Fig. 2F,G,L). Later, we observed P-Smad1/5/8 in the median Foxc expression domain at mid neurula stages (not shown) and in the median part of the U-shaped Foxg expression domain in late neurulae (Fig. 2H,I,L), corresponding presumably to the future ventral palp. This was confirmed by co-expression of P-Smad1/5/8 and Foxg in ventral cells of early tailbuds (Fig. 2J-L). In summary (Fig. 2K,L), BMP signaling was absent from the ANB (marked by Foxc) until the early neurula stages; active BMP signaling was detected from mid neurula stages in the future protrusion (Foxg⁺ cells) of the ventral palp.

We next tested whether BMP inhibition was sufficient to induce an ANB fate. When BMP signaling was blocked either by overexpression of the secreted inhibitor Noggin or by treatment with the BMP receptor inhibitor DMH1, Foxc expression at late gastrula stages was unchanged (Fig. 3A-C). The fact that Foxc was not ectopically expressed following BMP inhibition could be explained by an incomplete BMP blockade. However, DMH1 treatment led to undetectable P-Smad1/5/8 levels (Fig. S1). Alternatively, it could be that the number of cells that are competent to become ANB in response to BMP inhibition could be restricted to the cells already expressing Foxc. Foxc expression and palp fate are regulated by FGF signaling following neural induction and cell fate segregation (Wagner and Levine, 2012). We thus aimed at increasing the number of cells competent to form ANB by early activation of FGF signaling using treatment with recombinant bFGF protein, and testing the effects of BMP pathway modulations in this context. As expected, bFGF treatment from the 8-cell stage neuralized the entire ectoderm, as revealed by the ectopic expression of the neural markers Otx and Celf3/4/5/6 and the downregulation of the epidermal marker Tfap2-r.b (also known as Ap2-like2) at late gastrula stages [(Fig. 3Div-Fiv,Hiv) when compared with control embryos (Fig. 3Di-Fi,Hi)]. The effects of FGF were not modified by activation (Fig. 3Dv-Fv,Hv) or inhibition (Fig. 3Dvi-Fvi,Hvi). Foxc behaved somewhat unexpectedly: it was either ectopically expressed in a fraction of the embryos (53%, n=69; Fig. 3Giv) or repressed in the others (38%, n=69; not shown) when compared with ANB expression in controls (Fig. 3Gi). The repression of Foxc might be explained by the fact that FGF/Erk is downregulated in the palp lineage during gastrulation (Wagner and Levine, 2012), hence our continuous treatment might inhibit Foxc expression. Nevertheless, when the BMP pathway was inhibited on top of FGF activation, Foxc was strongly expressed, in all embryos, as a cup covering the anterior end including the ventral epidermis (Fig. 3Gvi). In addition, FGF activation did not overcome BMP-induced loss of Foxc expression (Fig. 3Gv). Our observations demonstrate that Foxc expression and ANB formation can only occur in a domain devoid of active BMP signaling.

Importantly, the loss of Foxc following BMP activation using recombinant BMP2 protein treatment was specific to this gene and did not result from neural tissue inhibition, as in vertebrates, because the neural markers Otx and Celf3/4/5/6 were still expressed in the CNS but downregulated in the ANB (Fig. 3Dii,Eii,Hii). Reciprocally, BMP inhibition was not sufficient to lead to ectopic neural tissue formation (Fig. 3Diii-Hiii). This is in agreement with similar data produced in the distantly related ascidian H. roretzi (Darras and Nishida, 2001). Interestingly, although Foxc and Tfap2-r.b were co-expressed in the ANB, only Foxc was repressed by BMP (Fig. 3Fi,Fii,Gi,Gii and Fig. S2). However, because the ANB does not convert into epidermis (Fig. 1), it is likely that, although inhibiting palp fate, BMP signaling does not abolish all the effects of the induction by FGF.

We examined whether modulating BMP had any impact on palp formation besides ANB specification. We thus performed whole-embryo treatments starting at progressively later stages of embryonic development and examined the early marker Foxc and the late marker Isl (Fig. 4).

Activating BMP at early gastrula stages (St. 10) partially repressed Foxc whereas mid-gastrula (St. 12) treatment had no effect (Fig. 4). The time-dependent effects of BMP activation coincided with the dynamics of Foxc expression: before it was expressed (8-cell stage), the repression was complete (Fig. 3Gii); at the onset of expression (St. 10), the repression was milder; and once Foxc was robustly expressed (St. 12), there was no repression (Fig. 4). This is further supported by the fact that BMP2 treatment led to fast P-Smad1/5/8 nuclear accumulation (the shortest treatment we tested was 30 min; Fig. S1). The ventral spot of Isl expression was lost for treatments starting at St. 10 and St. 12, whereas later treatments did not change Isl expression (Fig. 4).

Inhibiting BMP had no effect on Foxc expression, similarly to the earliest treatment (Fig. 3Giii). Isl, which is normally expressed as three spots, had a U-shaped expression (Fig. 4). The same effect was observed by overexpressing Noggin using electroporation (Fig. S3). This phenotype was much less frequent when the DMH1 treatment started at late neurula stages (St. 16).

The changes in Isl expression prompted us to determine how palps differentiate when BMP is modulated from gastrula stages.

In DMH1-treated embryos, whereas Isl was expressed in the shape of a large U at mid-tailbud stages (St. 23) (Fig. 4), it was concentrated in a protruding structure at the anterior tip at late tailbud stages (St. 25) (Fig. 5E) when it was expressed in each of the three protrusions of the control embryos (Fig. 5A). In larvae, this single, large protrusion was made of elongated cells as visualized by phalloidin staining (Fig. 5F-H), similar to what was observed in each of the three papillae of control larvae (Fig. 5B-D). We coined this phenotype Cyrano (in memory of the famous character depicted by Edmond Rostand). In Ciona, it is thought that palps contain a fixed number of the different cell types (Zeng et al., 2019). We performed fluorescence in situ hybridization at the late-tailbud stage (St. 23) using four genes and made 3D reconstruction of the z-stacks acquired by confocal microscopy (see Materials and Methods) in order to determine the differentiation of the palps in the Cyrano embryos. In agreement with previous reports, we found that the ACC marker Isl was expressed in 12 cells in control embryos (four cells per papilla) (Fig. 5I,J,Q). By contrast, we found that the sensory neuron marker Pou4 was expressed in ten cells instead of 12 (Fig. 5I,J,Q). Interestingly, both dorsal palps had four cells that surrounded the Isl-positive cells whereas the ventral palp contained two Pou4-positive cells located dorsally to the Isl-positive cells. In DMH1-treated embryos, the number of neurons increased to 18 on average and the number of ACCs to 18 (Fig. 5Q). The increase of Celf3/4/5/6 cells was not statistically significant. Interestingly, the number of cells expressing Sp6/7/8/9, which has been described as an inter-papillae marker (Wagner et al., 2014), was decreased in DMH1-treated embryos (Fig. 5Q). These data are in agreement with the interpretation that the number of cells with papilla fate has increased, but their physical proximity likely leads to the formation of a single protrusion.

In DMH1-treated embryos, Pou4 was expressed all around the Isl cells, like in the dorsal palps (Fig. 5M,N). This suggested that the Cyrano protrusion may have a dorsal identity. In support of this interpretation, we found that expression of the homeobox transcription factor Msx, which we found transiently expressed in the future ventral palp at the onset of Isl expression (Fig. 5K,L), was lost following BMP inhibition (Fig. 5O,P). Although we have performed a detailed analysis of the Cyrano phenotype only on embryos generated by DMH1 treatment starting at gastrula stages, a similar phenotype was observed upon Noggin overexpression (a single protrusion visualized by phalloidin staining in Fig. 5F; U-shape/ectopic expression of Isl and Celf3/4/5/6 in Fig. S3). In Cyrano embryos, the ventral palp is missing, the number of inter-palp cells is reduced, and there is an excess of dorsal protruding cells. This suggests that BMP is required to specify the ventral palp and inter-palp cells. When this pathway was inhibited, cells that have lost these fates would adopt a ‘default’ dorsal palp fate.

As expected from the Isl profile (Fig. 4), only two protruding papillae made of elongated cells developed dorsally in BMP2-treated embryos (Fig. 6G,H). They were similar to the papillae of control larvae (Fig. 6A,B). This morphological absence of ventral palp was only partially confirmed at the molecular level. Similarly to Isl, Msx expression was abolished in treated embryos, but the expression of Pou4 revealed that one or two neurons were still present (Fig. 6C,D,I,J). Sp6/7/8/9 was repressed in its most ventral expression domain (Fig. 6E,K). In conclusion, although the ventral protrusion is absent, palp identity is not completely suppressed. Accordingly, we did not detect ectopic expression of the epidermal gene Sox14/15/21 (Fig. 6L) as its expression was similar in control embryos (Fig. 6F).

Although BMP is required to define ventral palp fate (Fig. 5), BMP activation neither leads to ectopic ventral palp formation nor ventralizes the dorsal palps (we did not detect obvious defects in dorsal palp differentiation, and we identified four Pou4⁺ cells surrounding Isl⁺ cells as in controls; Fig. 6C,I). Given that BMP suppresses ventral palp, it is likely that excessive or precocious BMP signaling levels are responsible for this phenotype.

The U-shape pattern of Isl in DMH1 embryos reminded us of endogenous Foxg expression (Liu and Satou, 2019): at neurula stages, Foxg was expressed in the future palps following a U-shape that gradually converted into a three-spot pattern at tailbud stages that prefigures the three papillae protrusions (Fig. 7A-D). It thus seems that inhibiting BMP prevented the refinement of Foxg expression. Accordingly, Foxg expression was U-shaped following DMH1 treatment (Fig. 7E). Interestingly, knockdown of the zinc finger transcription factor-coding gene Sp6/7/8/9 leads to a U-shaped Foxg expression (Liu and Satou, 2019). Because Sp6/7/8/9 and Foxg are initially partially co-expressed before showing exclusive patterns (Fig. 2L), it has been proposed that Foxg restriction to the future protrusions is the result of repression by Sp6/7/8/9. We thus determined Sp6/7/8/9 and Foxg expression following BMP modulation from early gastrula stages (Fig. 7). Although we confirmed initial co-expression using double-fluorescence in situ hybridization, we failed to obtain robust simultaneous expression allowing analysis the effects of the treatments (not shown). At St. 15/16, the onset of Sp6/7/8/9 expression, Sp6/7/8/9 was barely detectable (Fig. 7F), and at St. 18/19, when Sp6/7/8/9 expression was strong (Fig. 7G), Foxg showed a transient downregulation (Fig. 7B). We thus analyzed each gene at different stages [Foxg at late neurula stages (St. 16) (Fig. 7I-K) and Sp6/7/8/9 at initial tailbud stages (St. 18) (Fig. 7L-N)]. When embryos were treated with BMP2 protein, the ventral expression of Foxg in the U-shape was missing (Fig. 7J) and Sp6/7/8/9 was ectopically expressed at this location (Fig. 7M). Reciprocally, following DMH1 treatment, Foxg was unchanged (Fig. 7K) and ventral Sp6/7/8/9 expression was shifted to a median position (Fig. 7N). Hence, Sp6/7/8/9 was no more expressed in the ventral part of the U-shaped, palp-forming row of cells (Fig. 7T). We have summarized our understanding of these results in embryo schematics (Fig. 7O-T). Interestingly, DMH1 treatment had limited effects on Isl expression at St. 16 (Fig. 4), a timing that coincides with the onset of Sp6/7/8/9 expression (Figs 2N and 7F).

Our results indicate that Sp6/7/8/9 is positively regulated by BMP signaling. However, because it is not expressed in P-Smad1/5/8⁺ cells (Fig. 2L), we propose that an intermediate, yet unidentified, factor activates Sp6/7/8/9 downstream of BMP in the neighboring cells (Fig. 7U,V). This hypothetical model of gene interaction is sufficient to explain Sp6/7/8/9 and Foxg expression patterns and final phenotypes (Fig. 7U-X). Importantly, our data show that the dorsal-most cells of the Foxg U-shape and dorsal palps develop independently of BMP signaling. In conclusion, we propose that the ventral papilla and the papilla versus inter-papilla fate choice is controlled by BMP signaling through the indirect regulation of Sp6/7/8/9 expression.

We investigated the conservation of the role of BMP in palp formation by examining embryos of the ascidian P. mammillata, which belongs to the same family as Ciona, the Phlebobranchia, but with a significant divergence time (275 My) (Fig. 8A) (Delsuc et al., 2018). First, we determined that BMP signaling was active in the ventral part of the embryo with a similar dynamic to Ciona as revealed by P-Smad1/5/8 immunostaining (Fig. S4). Next, we identified single orthologs for Celf3/4/5/6, Pou4 and Isl genes, which were all expressed in the palps (Fig. 8B,E,I,M) (Chowdhury et al., 2022; Coulcher et al., 2020; Dardaillon et al., 2020). Treatment with recombinant BMP2 protein from the 8-cell stage abolished expression of all three markers in the palps, like in Ciona (Fig. 8C,F,J,N). Following DMH1 treatment from the 8-cell stage, both Celf3/4/5/6 and Isl were expressed in the palp territory following a U-shape pattern like in Ciona, but not in all cases. For a large fraction of embryos, the pattern appeared as two bars of intense staining resembling the U-shape but without the ventral part (Fig. 8D,G,H,K,L). This phenotype, which we did not observe in Ciona, might reveal some differences in the role of BMP in the two species.

Given the overall similar effects on palp formation after alterations of BMP signaling, we sought to identify novel palp molecular markers using a dataset previously generated in P. mammillata (Chowdhury et al., 2022). In this previous study, we generated, at several developmental stages, RNA-seq data for whole embryos treated with recombinant BMP4 protein and/or DAPT, a pharmacological Notch inhibitor. We identified 1098 genes repressed by BMP signaling at least at one developmental stage (Table S1). In this list, we found the orthologs for 11 well-defined Ciona palp markers, and four of them (Otx, Isl, Atoh1/7 and Celf3/4/5/6) have been described as expressed in the palp lineage in Phallusia (Coulcher et al., 2020; Dardaillon et al., 2020). Using Gene Ontology analysis, we selected a list of 53 genes encoding developmental regulators (transcription factors and signaling molecules) or involved in neural tissue formation, and examined their expression patterns (Table S2). Within this list, the expression patterns of 26 genes were previously determined [from the Aniseed database (Dardaillon et al., 2020) and from our previous data (Chowdhury et al., 2022; Coulcher et al., 2020)] and 12 of them were expressed in the palps. We performed in situ hybridization for the remaining 27 genes, and discovered eight novel palp markers (Tp53inp, Fzd9/10, Plg, Mucin, Hes.b, Barhl, Fbn and Wscd), expression of which is shown in Fig. 9.

Surprisingly, by examining the expression data generated previously, we found that some genes with palp expression were upregulated by BMP in our dataset, such as Chrdl and Nos (Table S2, Fig. 9A,K). To have a broader view of the potential effects of BMP signaling on gene regulation in the palps, we gathered, from previous publications (Chen et al., 2011; Chowdhury et al., 2022; Coulcher et al., 2020; Joyce Tang et al., 2013; Kusakabe et al., 2012; Liu and Satou, 2019; Pasini et al., 2006; Roure and Darras, 2016; Shimeld et al., 2005; Wagner and Levine, 2012; Wagner et al., 2014), from the Aniseed database (Dardaillon et al., 2020) and from the present study, a list of 68 genes with expression in the palp lineage in Ciona and/or Phallusia (Table S3). We plotted the results of our Phallusia RNA-seq data, and found that 70% of the genes were regulated by BMP signaling. Most of them were repressed by BMP, but 20 genes were activated by BMP, and a smaller fraction was repressed or activated depending on the stage. Consequently, the precise function of BMP, which is likely to be dynamic in the course of palp differentiation, needs to be investigated in further detail. Interestingly, Notch is likely to play a role in the specification of the different cell types that compose the palps. For instance, it has been shown that activating Notch represses palp neuronal markers in H. roretzi (Akanuma et al., 2002). We found 30 genes regulated by Notch in our dataset.

We have shown that BMP signaling regulates two distinct steps of palp formation in C. intestinalis: ANB specification, and ventral papilla versus inter-papilla specification. Moreover, we have shown conservation of gene expression and regulation by BMP in P. mammillata.

ANB specification is regulated by inputs from several signaling pathways: FGF, Wnt and BMP. Although FGF is positively required early on, at the time of neural induction (32-cell stage), all three pathways are inactive at the time of ANB fate acquisition as revealed by the expression of Foxc (mid-gastrula). This situation is reminiscent of data from vertebrates wherein anterior neural fate is determined by the triple inhibition of BMP, Nodal and Wnt pathways (Andoniadou and Martinez-Barbera, 2013; Niehrs et al., 2003; Wilson and Houart, 2004). It would thus be interesting to test the function of Nodal inhibition in ANB specification given that we have already shown that it is involved in posterior neural fate determination in Ciona (Roure et al., 2014). Although it appears that active FGF, Wnt or BMP signaling is incompatible with ANB determination, the specific function of each pathway seems different. FGF appears to regulate the anterior CNS versus ANB fate decision along the antero-posterior axis (Wagner and Levine, 2012). Wnt seems to regulate Foxc⁺ ANB fate versus Foxc⁻ ANB fate along the mediolateral/dorsoventral axis (Feinberg et al., 2019). Finally, BMP might participate in the segregation between ANB and immediately anterior/ventral epidermal fates. Finer details on the function of these pathways in ANB fate determination and on their likely cross-talk should be an exciting line of research in this simple and geometric model system.

Our results of late inhibition of BMP signaling (from gastrula stages) indicate that Foxg, expressed in a single row of cells with a U-shape, delineates cells competent to become papilla. A network of gene interactions has previously been identified that regulates the transition of Foxg from a U-shape to a three-spot pattern eventually forming protruding papillae (Liu and Satou, 2019). BMP is an input to this network, presumably through the indirect regulation of ventral Sp6/7/8/9 expression. We hypothesized the involvement of a signaling molecule that would be a direct target of BMP (Fig. 7). Interestingly, MAPK inhibition during neurulation results in a U-shaped expression of Isl (Wagner et al., 2014), similar to what we observed by inhibiting BMP. It is tempting to propose an FGF ligand to be the factor downstream of BMP; however, none has been described with a discrete pattern in the palps (Imai et al., 2004). Again, studying epistatic relationships and cross-talk between these signaling pathways is a future line of research.

Importantly, we have shown that BMP signaling is active and required in the median palp-forming region, most likely corresponding to the future ventral palp, before the onset of Foxg and Sp6/7/8/9 expression. However, activating BMP at this stage does not result in ectopic palp formation but rather to an absence of the ventral palp. This discrepancy might be better understood by more fine control of the levels of BMP signaling, but also its timing and the cells that receive it, through optogenetics for example. Nevertheless, our observations point to differences between the two symmetrically bilateral dorsal palps and the single median ventral palp. Although we are not aware of ventral palp-specific marker, we have shown that Msx is transiently expressed only in the ventral palp (Fig. 5); this may also be the case for Hes.a (Chowdhury et al., 2022). In addition, the Pou4⁺ sensory neurons are located dorsally in the ventral palp, whereas they are located around Isl⁺ cells in the dorsal palps. Specific dorsal and ventral genetic sub-networks would thus be interesting to uncover.

The specification of the three cell types (ACCs, neurons and collocytes) that make the papillae and their relationships (e.g. lineage, alternative cell fate) are still poorly understood, but will most likely be the subject of future research (Zeng et al., 2019). For example, we are not aware of a collocyte-specific gene marker; however, these cells can be distinguished from other palp cells with peanut agglutinin (Cao et al., 2019; Sato and Morisawa, 1999; Zeng et al., 2019). The collocytes are involved in the secretion of adhesive materials for the larva to attach to a substrate before metamorphosis. The palps thus constitute an adhesive organ homology of which with adhesive organs that exist in the larvae of some vertebrates (e.g. the cement gland of Xenopus) has been previously proposed (Yoshida et al., 2012). Our present work adds to the similarities observed between the frog cement gland and ascidian palps: they are ectodermal derivatives specialized in adhesion, they are located at the anterior-most part of the larva, they share expression of the transcription factor-coding genes Otx and Pitx, and their formation is regulated by BMP signaling (Gammill and Sive, 2000; Jin and Weinstein, 2018; Yoshida et al., 2012). Further detailed comparison of the shared, but also divergent, parts of the developmental networks regulating adhesive organ/sensory organ formation in ascidians and vertebrates should be of great interest.

The ascidian larval PNS, palps included, originates from the neural plate border with the exception of the ventral tail PNS, which originates from a region at the opposite end of the embryo, the ventral epidermis. Signaling pathways are pleiotropic and are consequently poor indicators of possible evolutionary conservation. Nevertheless, it is striking that FGF, Wnt and BMP are deployed in ascidians to regulate neural plate border specification and differentiation of its derivatives, most likely with changing dynamic requirements at diverse developmental stages. This is reminiscent of the mechanisms regulating the neural plate border and its derivatives, the cranial placodes and the neural crest (Martik and Bronner, 2021; Pla and Monsoro-Burq, 2018; Stundl et al., 2021). The similarities extend beyond signaling pathways as a suite of genes have conserved expression between ascidians and vertebrates, and have led to the proposal of several evolutionary scenarios (Cao et al., 2019; Horie et al., 2018; Pasini et al., 2006; Poncelet and Shimeld, 2020). Our present study adds further useful information to gene network level comparisons.

The ascidian PNS comprises epidermal sensory neurons that have different morphologies, connectivity and sensory capacities depending on their location (Abitua et al., 2015; Imai and Meinertzhagen, 2007; Ryan et al., 2018). However, they share a number of genes marking the presumptive domains or differentiating neurons. Yet, what regulates their specific identities is still incompletely understood (Chacha et al., 2022). For example, a number of genes expressed in the tail PNS are also expressed in the palps, and these expression domains are conserved in species that diverged almost 400 My ago (Table S3) (Akanuma et al., 2002; Coulcher et al., 2020; Joyce Tang et al., 2013; Pasini et al., 2006; Roure and Darras, 2016). Comparative approaches of PNS formation between divergent ascidian species and across chordates (vertebrates and cephalochordates) promise to yield insights into PNS evolution and the flexibility of developmental mechanisms.

Adults from Ciona intestinalis (formerly referred to Ciona intestinalis type B; Brunetti et al., 2015) were provided by the Centre de Ressources Biologiques Marines in Roscoff (EMBRC-France). Adults of Phallusia mammillata were provided by the Centre de Ressources Biologiques Marines in Banyuls-sur-Mer (EMBRC-France) following diving or by professional fishermen following trawling in the Banyuls-sur-Mer (France) area. Gamete collection, in vitro fertilization, dechorionation and electroporation were performed as previously described (Coulcher et al., 2020; Darras, 2021), and staging of embryos was performed according to the developmental table of Ciona robusta (Hotta et al., 2007).

Electroporation constructs used in this study have been previously described (Pasini et al., 2006). Embryos were treated with 150 ng/ml of recombinant mouse BMP2 protein [355-BEC, R&D Systems; 100 µg/ml stock solution in 4 mM HCl+0.1% bovine serum albumin (BSA)], 100 ng/ml of recombinant human bFGF (F0291, Sigma-Aldrich; 50 µg/ml stock solution in 20 mM Tris pH 7.5+0.1% BSA) complemented with 0.1% BSA, or 2.5 µM of the BMP receptor inhibitor DMH1 (S7146, Euromedex; 10 mM stock solution in DMSO) at the stages indicated in the text and figures. These concentrations were determined following pilot experiments. Control embryos were incubated with sea water containing 0.1% BSA and/or 0.025% DMSO.

For all labeling experiments, embryos were fixed in 0.5 M NaCl, 100 mM MOPS pH 7.5 and 3.7% formaldehyde. Whole-mount chromogenic in situ hybridization was performed using plasmid cDNA or synthetic DNA (eBlocks Gene Fragment, IDT) as templates for probe synthesis (Tables S2 and S4) as described previously (Chowdhury et al., 2022). Gene models and identifiers correspond to the genome assemblies KH2012 for Ciona robusta (Satou et al., 2008) and MTP2014 for Phallusia mammillata, which were retrieved from the Aniseed database (Dardaillon et al., 2020). Images were acquired using an AxioCam ERc5 s digital camera mounted on a stereomicroscope (Discovery V20, Zeiss). The numbers of experiments and embryos for phenotypic effects by gene expression analysis are shown in the figures and their legends.

The fluorescence in situ hybridization protocol was adapted from that described by Racioppi et al. (2014). Briefly, digoxigenin-labeled probes were recognized using an anti-DIG antibody coupled to peroxidase (11207733910, Roche), and fluorescein-labeled probes were recognized using an anti-FLUO antibody coupled to peroxidase (11426346910, Roche). Fluorescence signal was produced using the TSA plus kit (NEL753001KT, PerkinElmer) following the manufacturer's recommendations with Cy3 and fluorescein for DIG- and FLUO-probes, respectively. Active BMP signaling was visualized by immunostaining using a rabbit monoclonal antibody against mammal Smad1, Smad5 and Smad8 phosphorylated at two serine residues at the C-terminal end (clone 41D10, #9516, Cell Signaling Technology) diluted at 1:200. The epitope is present in the single ortholog Smad1/5/8 of both Ciona intestinalis and Phallusia mammillata. Anti-rabbit antibodies coupled to Alexa Fluor 568 (A11011, Invitrogen) were used at 1:400 for visualization. Similar data were obtained using another antibody (clone D5B10, #13820, Cell Signaling Technology) (data not shown). Membranes were stained using Alexa Fluor 594 phalloidin (A12381, Invitrogen) used at 1:1000. Nuclei were stained using DAPI. Image acquisition was performed using confocal microscopy (Leica SP8-X, BioPiC platform, Banyuls-sur-Mer). Confocal z-stacks were visualized and analyzed in 3D using the Imaris 8.3 software (Bitplane). In particular, this software was used to count the number of cells expressing a gene of interest. In brief, fluorescence signals were converted as 3D objects: in situ hybridization signals as surface objects, and DAPI-labeled nuclei as spots. The number of spots within a given surface was used as a proxy for the number of cells expressing a gene. Snapshots of such analyses and 3D renderings are shown in Figs 1, 2, 5, 6 and Figs S1, S2. Maximum intensity projections of Fig. S4 were obtained using ImageJ.

Image panels and figures were constructed with Affinity Photo and Affinity Designer.

We thank G. Diaz (Port-Vendres) and staff (M. Fuentes, divers and boat crew) at the marine stations of Banyuls-sur-Mer and Roscoff (French node of the European research infrastructure EMBRC) for providing animals. We would like to acknowledge the BioPiC imaging facility (Sorbonne Université/CNRS, Banyuls-sur-mer) for access to the confocal microscope, R. Dumollard and H. Yasuo for sharing plasmids, and V. Thomé for advice on fluorescence in situ hybridization and immunostaining.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201575.reviewer-comments.pdf