Specification and survival of post-metamorphic branchiomeric neurons in a non-vertebrate chordate

The simple embryos of the non-vertebrate chordate Ciona and related tunicates constitute a highly tractable system in which to study the regulation of cellular processes in development (Farley et al., 2015; Bernadskaya and Christiaen, 2016; Cao et al., 2019; Wang et al., 2019; Guignard et al., 2020). Their classification as tunicates, the sister group to the vertebrates (Delsuc et al., 2006; Putnam et al., 2008), means they share with vertebrates many chordate-specific gene families, cell types, organs and anatomical structures (Christiaen et al., 2002; Dufour et al., 2006; Razy-Krajka et al., 2014; Abitua et al., 2015; Stolfi et al., 2015; Di Gregorio, 2020; Fodor et al., 2021; Lemaire et al., 2021; Papadogiannis et al., 2022). The simplicity of their embryos overshadows the fact that the larva is but one part of a biphasic life cycle alternating between a free-swimming larval phase and a sessile adult phase (Fig. 1A). During metamorphosis, most larval structures degenerate and are replaced by adult tissues (Chiba et al., 2004; Nakayama-Ishimura et al., 2009; Sasakura and Hozumi, 2018). This metamorphosis is not entirely catastrophic, as some adult structures begin to form during the larval phase (Nakazawa et al., 2013; Razy-Krajka et al., 2014), which is pronounced in colonial species (Berrill, 1935). Although the invariant cell lineages and gene regulatory networks specifying many larval cell types of Ciona have been investigated in detail (Imai et al., 2006; Cao et al., 2019), metamorphosis and adult development are still poorly understood. We know that various undifferentiated progenitor cells set aside in the larva as discrete compartments can give rise to adult structures (Dufour et al., 2006; Horie et al., 2011; Razy-Krajka et al., 2014; Sasakura and Hozumi, 2018). These adult progenitor cells are specified in an invariant manner alongside differentiated larval cells, but only differentiate after the larva settles and undergoes metamorphosis. The transition from invariant (stereotyped) embryogenesis to variable (plastic) post-metamorphic development is unique among the chordates, and thus of potential interest for revealing previously unknown mechanisms for precise spatiotemporal regulation of cellular quiescence, survival, proliferation and differentiation.

In this work, we focused on a compartment of Ciona adult neural precursor cells in the larva called the ‘neck’ (Fig. 1B,C). The neck has been proposed to be homologous to parts of the hindbrain of vertebrates based on shared expression of Pax2/5/8.a, Phox2 and Hox1 orthologs, specifically regions giving rise to branchiomeric motor neurons (Fig. 1C,D) (Dufour et al., 2006; Imai et al., 2009; Hudson and Yasuo, 2021). Although it is difficult to draw one-to-one correspondences between specific tunicate and vertebrate central nervous system (CNS) compartments (Ikuta and Saiga, 2007), most recent evolutionary models propose an anterior hindbrain-like identity for the neck (Hudson and Yasuo, 2021). Although the neck is specified during embryonic development and shares a close developmental lineage with differentiated larval brain neurons (Fig. 1B-D), most of it remains in an undifferentiated state until after metamorphosis (Dufour et al., 2006; Imai et al., 2009). With some rare exceptions, most differentiated neurons of the larval CNS degenerate during metamorphosis and undifferentiated neural progenitors in the neck and in other compartments survive to form the adult nervous system (Fig. 1D) (Horie et al., 2011; Hozumi et al., 2015). The signaling pathways and regulatory networks that direct neck specification, survival and differentiation remain uncharacterized.

Here, we investigate the molecular mechanisms underlying the specification and patterning of the neck. We show that the neck cells continue to divide throughout embryogenesis and the larval phase, with some derivatives differentiating into larval- and adult-specific neurons. We show that Pax2/5/8.a and Phox2 are required for the specification of neck-derived adult neurons, and that FGF signaling regulates the balance of differentiation and proliferation/survival in the neck during metamorphosis. Thus, we reveal key molecular mechanisms underlying the ability of Ciona to generate an entirely separate adult nervous system even as its larval nervous system is largely eliminated.

Presently, the Ciona robusta neck has been described as six ependymal cells per side, thought to be quiescent in nature, and a bilateral pair of differentiated neurons, the so-called ‘neck neurons’ (NNs) (Ryan et al., 2016). To establish how the neck lineage develops between early tailbud and hatched larval stage, we examined neck cell number, cell type, and morphology using Pax2/5/8.a (Ryan et al., 2017; Oonuma et al., 2021), Phox2 (Dufour et al., 2006), and a novel GRIK fluorescent reporter at hourly intervals, in embryos raised at 20°C (Fig. 2A-K).

At the early tailbud stage [Hotta stage (St.) 19, ∼9 hours post-fertilization (hpf)], two Pax2/5/8.a>GFP⁺ cells on either the left or the right side of the embryo could be observed, corresponding to A11.61 and A11.62 (Fig. 2A) (Ikuta and Saiga, 2007; Imai et al., 2009). These two cells continued to divide, forming a single line of four cells on either side by the mid-tailbud stage (St. 20-22, ∼9-10 hpf; Fig. 2B). By this stage, the four cells started to express Phox2>GFP, which was only detectable by anti-GFP immunofluorescence likely due to the low levels of mature GFP protein at this stage (Fig. 2C). Starting at St. 23 (∼11 hpf), the posterior-most cell (A12.121) divided first (Fig. 2D), followed by the anterior-most cell (A12.124) dividing around St. 24 (∼12 hpf; Fig. 2E). This brought the total number of Pax2/5/8.a⁺ cells on either side to six. These six neck cells had not yet differentiated and maintained a typical ependymal cell morphology within the neural tube epithelium. Although the cells appeared to divide in a stereotypical order, the exact timing of their divisions was slightly variable from embryo to embryo and even between left and right sides in the same individual (Fig. S1A), as has been previously reported in the Ciona neural tube (Cole and Meinertzhagen, 2004).

During the late tailbud stage (St. 25, ∼14 hpf), we observed bright Phox2>GFP expression mostly in the middle to anterior part of the neck (Fig. 2F). Notably, Phox2>GFP was not visible in the posterior cells of the lineage, which appeared to delaminate from the neural tube epithelium, separating from the Phox2⁺ cells (Fig. 2F,G). Beyond stage 26 (larval stages, ∼16 hpf onwards), additional cell divisions were seen among the Phox2>GFP⁺ cells, as well as the anterior Phox2-negative cells. This increased the number of Pax2/5/8.a⁺ cells to an average of 12 on each side, but the order of divisions and lineage could not be easily traced (Fig. 2G, Fig. S2B). Furthermore, Phox2>GFP expression expanded anteriorly in the lineage during the larval phase, but was still not detectable in the posterior cells that had delaminated from the neural tube epithelium (Fig. 2G). This Phox2>GFP expression in the anterior cells may have resulted from uneven proliferation of the Phox2⁺ cells descended from A12.123/A12.122, or from earlier transient expression in the anterior cells, which might vary based on mosaicism. The average number of Pax2/5/8.a⁺ neck cells was quantified at each stage (Fig. S2B). We conclude that neck cells continue to divide into the larval stage, suggesting they are not quiescent.

At the larval stage, we first observed the differentiation of the posterior-most cell of the lineage on either side at larval stages, likely the single left/right pair of NNs previously described (Ryan et al., 2016, 2017). Shortly thereafter, we observed ascending Pax2/5/8.a reporter-expressing neurites extending from three cells situated just anterior to the NNs (Fig. 2H,I, Fig. S2). To examine further these different Neck-derived neurons, we generated a reporter using an intronic cis-regulatory element from the Glutamate receptor ionotropic kainate (GRIK) gene, expression of which had previously been detected by in situ hybridization in four neck cells on either side (Borba et al., 2023 preprint). In early larvae (St. 27, ∼17 hpf and St. 30, 24 hpf) we frequently observed GRIK>GFP signal in four cells on either side, including the NN and the three more anterior ascending neurons (Fig. 2H,I). In animals undergoing body-axis rotation (St. 36, 48 hpf), GRIK>GFP⁺ cells appeared to continue extending their axons into the body cavity (Fig. 2J). In early post-metamorphic juveniles (St. 36, 38, ∼72-95 hpf), we observed neurons labeled by Pax2/5/8.a and Phox2 reporters innervating the gill slits (Fig. S1C,D, Movie 1), similar to those reported previously (Dufour et al., 2006), and that presumably correspond to the cholinergic ciliomotor neurons (CMNs), which control branchial ciliary flow (Hozumi et al., 2015; Jokura et al., 2020). In post-metamorphic juveniles, Phox2>mSc3⁺ neurons innervating the gill slits were also GRIK>GFP⁺, supporting our model that some GRIK>GFP⁺ cells observed in the larva might correspond to post-metamorphic CMNs (Fig. 2K). This interpretation was further supported by photoconversion of Kaede, a photoconvertible fluorescent reporter (Ando et al., 2002). Kaede was expressed in posterior larval neck cells using the GRIK driver, and photoconverted before metamorphosis (Fig. 3A). Photoconverted Kaede⁺ cells were observed in the cerebral ganglion after metamorphosis, but none was seen in post-metamorphic juveniles that were not subjected to photoconverting illumination (Fig. 3B,C). Together, these observations suggest that the adult nervous system begins to differentiate before the larval phase is over. This was surprising given the traditional assumption that larval neurons are eliminated during metamorphosis, whereas adult neurons arise from set-aside, undifferentiated progenitors. Our results suggest a less clear contrast between pre- and post-metamorphic neurogenesis in tunicates. We propose an updated model of the neck cell lineage in Fig. 2L.

Having established that the neck gives rise to gill slit-innervating CMNs, we next tested the roles of Pax2/5/8.a and Phox2 in their specification. More specifically, we used tissue-specific CRISPR/Cas9-mediated mutagenesis to disrupt either gene in neural progenitors prior to hatching and metamorphosis. Embryos were electroporated with ‘negative control’ or gene-specific single-chain guide RNA (sgRNA) expression plasmids, Sox1/2/3>Cas9::Geminin^N-ter plasmid to drive expression of Cas9 in early neural progenitors, and a Phox2>GFP reporter to label CMNs. The Sox1/2/3 (also known as SoxB1) promoter drives Cas9 expression in the ectoderm and is pan-neuronal (Stolfi et al., 2014b). The sgRNAs were validated by Illumina sequencing of amplicons from targeted embryos (see Materials and Methods). Animals were grown to St. 37 (∼72 hpf) and imaged to assay the presence of properly differentiated CMNs. In negative control animals, Phox2>GFP⁺ CMN axons were observed innervating gill slits in a majority of animals (Fig. 4A,D). In contrast, disrupting either Pax2/5/8.a or Phox2 by CRISPR resulted in fewer gill-innervating axons, suggesting loss or incomplete differentiation of CMNs (Fig. 4B-D). Consistent with this, Pax2/5/8.a CRISPR also caused loss of GRIK>GFP⁺ neck cells in larvae (Fig. S1E,F). This suggests that Pax2/5/8.a and Phox2 are necessary for the formation of post-metamorphic CMNs.

To understand how overexpression of Pax2/5/8.a might impose a neck cell identity, we expressed Pax2/5/8.a (transcript variant 1, or tv1) under control of the pan-neuronal Nut promoter (Nut>Pax2/5/8.a) (Shimai et al., 2010). In Nut>lacZ negative controls, Phox2>GFP⁺ CMNs innervate the early gill slits as expected (Fig. 4E). In animals overexpressing Pax2/5/8.a, we observed an expansion of Phox2⁺ cells throughout the juvenile brain region, but we did not observe any obvious excess of gill slit-innervating Phox2⁺ axons (Fig. 4F). Together, this suggests that Pax2/5/8.a is sufficient to activate Phox2 expression in the post-metamorphic CNS, but might not be sufficient to impart specifically a CMN-like identity.

In order to confirm the role of Pax2/5/8.a as a regulator of neck identity and transcriptional programs, we used bulk RNA sequencing (RNAseq) to measure transcriptome changes elicited by ectopic Pax2/5/8.a overexpression. We used the Nut promoter (Shimai et al., 2010) to overexpress both isoforms of Pax2/5/8.a (transcript variants ‘tv1’ and ‘tv2’) throughout the neural tube, and compared whole-embryo transcriptomes of Pax2/5/8.a-overexpression and negative control (no overexpression) embryos at 10 hpf. Differential gene expression analysis revealed the enrichment, or depletion, of transcripts upon overexpression of either Pax2/5/8.a variant (Fig. 5A, Table S1). Previously known neck markers and Pax2/5/8.a targets (e.g. Phox2, Gli, Eph.c, FGF9/16/20) were among the top 600 upregulated genes (out of ∼16,000), with some being considerably higher in ranking (e.g. Phox2). In contrast, several known markers of the brain and motor ganglion (MG) were among the top 50 genes most downregulated by Pax2/5/8.a (e.g. Pax6, Nk6, Neurogenin, BCO). Correlation between differential gene expression values elicited by the two isoforms was modestly high (Pearson correlation r=0.70) (Fig. S3), suggesting high reproducibility and specificity of the downstream effects of Pax2/5/8.a. Interestingly, despite the neck-specific expression of the GRIK>GFP reporter and its loss in Pax2/5/8.a CRISPR larvae, GRIK was not significantly upregulated by Pax2/5/8.a in our RNAseq experiment, suggesting that Pax2/5/8.a might be necessary but not sufficient for GRIK expression, or a mechanism exists that limits GRIK activation prior to hatching. Similarly, Hox1 expression in the neck appeared to be independent of Pax2/5/8.a, consistent with previous studies that revealed its activation by retinoic acid signaling instead (Imai et al., 2009).

One gene previously known to be downstream of Pax2/5/8.a, Gli (Imai et al., 2009), was among the top 31 upregulated genes using either Pax2/5/8.a isoform (Fig. 5A). Indeed, CRISPR/Cas9-mediated knockout of Pax2/5/8.a largely abolished Gli>GFP reporter expression in the neck, confirming its regulation by Pax2/5/8.a (Fig. 5B-D). Following the same approach, we identified a previously unknown target of Pax2/5/8.a, the gene Vanabin4 (Van4, gene IDs KH.C3.88/KY21.Chr3.256). Van4 encodes a tunicate-specific vanadium-binding protein and was the second most upregulated gene by both Pax2/5/8.a isoforms (Fig. 5A). In situ mRNA hybridization revealed expression of Van4 specifically in the neck (Fig. 5E,E′). Van4 expression was lost upon Pax2/5/8.a CRISPR knockout (Fig. 5F,F′) and expanded throughout the neural tube upon overexpression of Pax2/5/8.a (Fig. 5G,G′). Van4>GFP reporter expression in the neck was similarly lost upon Pax2/5/8.a CRISPR knockout (Fig. 5H-J) and reporter expression was expanded across the neural tube upon overexpression of Pax2/5/8.a (Fig. 5K,L). Taken together, these data reveal a transcriptional program for neck specification downstream of Pax2/5/8.a.

During analysis of RNAseq data, we observed upregulation of Phox2 (KY21.Chr14.160) and the neighboring gene KY21.Chr14.159 by Pax2/5/8.a overexpression. On closer inspection, we found that RNAseq reads spanned exons covering both gene models, and further identified two cryptic exons (exon 1 and exon 5) (Fig. S4A,B). Our corrected gene model (Fig. S4C) was supported by cloning and sequencing a full-length Phox2 cDNA spanning all seven exons (Fig. S4D), and by protein sequence alignment with predicted Phox2 sequences from other tunicate species (Fig. S4E). Finally, mRNA in situ hybridization using probes designed for both Chr14.160 and Chr14.159 gene models revealed identical expression patterns, further supporting the finding that these are the same gene (Fig. S4F,G). This could explain why enrichment of Phox2 might go undetected in certain RNAseq data when using 3′-biased sequencing. The function of the extended C terminus (which is divergent relative to that of vertebrate Phox2 family members) is unknown but will likely be important for future studies on transcriptional regulation by Phox2.

To characterize further the molecular profiles of neck cells, we re-analyzed published whole-embryo single-cell RNAseq (scRNAseq) data obtained at the mid-tailbud II stage (Cao et al., 2019). This revealed a single cluster enriched for Pax2/5/8.a reads (cluster 25; Figs S5A, S6A,B, Table S2). Re-clustering performed only on cluster 25 cells revealed two distinct subclusters, with Pax2/5/8.a enriched in one subcluster (subcluster 0) and depleted in the other (subcluster 1) (Figs S5B, S6C). Double in situ hybridization using probes for Pax2/5/8.a and a top subcluster 1 marker, Crls1 (KH.C11.724), confirmed that subcluster 0 represents the neck, whereas subcluster 1 appears to represent larval brain neurons and photoreceptors just anterior to the neck (Fig. S5C). Other markers enriched in subcluster 0 cells further confirmed their identity as neck cells, showing substantial overlap with the top genes upregulated by Pax2/5/8.a overexpression as measured by our bulk RNAseq experiment above (Fig. S5D,E, Table S3). Correlation between enrichment/depletion in the neck by scRNAseq and average upregulation/downregulation by Pax2/5/8.a was modest (Pearson correlation=0.43; Spearman's rank correlation=0.46) (Table S4). We observed notable exceptions, such as Hox1, which is highly expressed in the neck but was not upregulated by Pax2/5/8.a (Fig. S5D,E). This suggests that Pax2/5/8.a is not sufficient to activate the entirety of the transcriptional program of the neck, and that some important neck regulators might be expressed in parallel to, not downstream of, Pax2/5/8.a. In contrast, the top markers enriched in the ‘brain’ subcluster (subcluster 1) were among those transcripts most highly depleted by Pax2/5/8.a overexpression, suggesting that Pax2/5/8.a is also instrumental for repressing larval brain neuron/photoreceptor identity.

As expected, Phox2 expression (taken as the expression of the more 3′ KH.C14.119/KY21.Chr14.159 gene model due to the 3′ bias of 10x Genomics system) was unevenly distributed within subcluster 0. Phox2 was expressed more highly in one half of the subcluster than in the other half (Fig. S5F), recapitulating its expression in a subset of neck cells as seen by Phox2>GFP expression. Additional genes showed a similar distribution in the neck subcluster (e.g. Rspo3, Tesk; Figs S5F, S6D), further hinting at a distinction between proliferating/undifferentiated cells in the anterior and differentiating neurons in the posterior.

FGF/MAPK signaling plays numerous roles in the development of the Ciona larval nervous system (Davidson et al., 2006; Picco et al., 2007; Stolfi et al., 2011; Haupaix et al., 2013; Razy-Krajka et al., 2018). In fact, FGF8/17/18 from the neighboring A9.30 cell lineage is required for expression of Pax2/5/8.a, which in turn activates expression of FGF9/16/20 in the neck (Fig. 6A,A′) (Imai et al., 2002, 2009). Furthermore, Ephrin/Eph signaling antagonizes FGF/MAPK signaling as a spatial cue for FGF/MAPK-dependent cell fate choices in the nervous system, including the MG, in which downregulation of FGF/MEK promotes cell cycle exit and neuronal differentiation (Stolfi et al., 2011; Haupaix et al., 2013). By in situ hybridization, we confirmed the expression of FGF9/16/20 and Eph.c (formerly Eph3) in the neck (Fig. 6A-B′), and EphrinA.b and EphrinA.d (Fig. 6C-D′) in the cells of MG just posterior to the neck (Imai et al., 2009). Immunostaining for doubly phosphorylated ERK (dpERK) showed that anterior Pax2/5/8.a>H2b::mCh⁺ cells were dpERK⁺, whereas the posterior-most fourth cell in contact with Ephrin⁺ MG cells was not labeled with dpERK antibody (Fig. 6E,E′, Fig. S7). As expected, dpERK signal was abolished in the neck by forced expression of a dominant-negative FGF receptor (dnFGFR) (Fig. 6F,F′). Taken together, these data suggested that FGF and Ephrin/Eph signaling might be key to regulating cell fate and neuronal differentiation in the neck as well.

To test the role of FGF and Ephrin signaling in the neck, we used the Pax2/5/8.a promoter to overexpress dnFGFR, dominant negative Eph.c receptor (dnEph.c), or a constitutively active form of the MAPK kinase MEK (MEK^S220E,S216D, also called caMEK) (Davidson et al., 2006; Picco et al., 2007; Shi and Levine, 2008; Razy-Krajka et al., 2018). We examined the neck at the late swimming larval stage (St. 29), when there are approximately eight anterior putatively undifferentiated, neuroepithelial cells and approximately four posterior, differentiating neurons in control larvae (Fig. 6G,K,L). In larvae expressing the dnFGFR, we observed fewer cells overall in the neck, and undifferentiated cells appeared to be replaced by supernumerary neurons, as evidenced by loss of epithelial organization and excess axon outgrowth (Fig. 6H,K,L). This contrasted starkly with larvae expressing the dnEph.c, in which we observed a near-complete loss of differentiated neurons, most clearly evidenced by a distinct absence of the posterior NN axon extending towards the tail (Fig. 6I,K,L). Instead, the entire neck took on the morphology of undifferentiated neural precursors in a tightly packed neuroepithelium. This was phenocopied by caMEK, which constitutively activates the FGF/MAPK pathway in all cells (Fig. 6J-L). We observed these phenotypes consistently across two biological replicates (Fig. 6K,L). We also examined the expression of Phox2>GFP in the neck, which is normally expressed by the ependymal-like cells in the middle part of the neck (Fig. 2C, Fig. S5F). As expected, we observed a loss of Phox2⁺ neck cells in larvae electroporated with Pax2/5/8.a>dnFGFR (Fig. S8A,B,E). However, neither Pax2/5/8.a>dnEph.c nor Pax2/5/8.a>caMEK abolished Phox2>GFP expression but rather made its spatial distribution within the neck inconsistent (Fig. S8C-E). This suggests that sustained FGF/MEK signaling is permissive but not sufficient for Phox2 activation in the neck. Taken together, these results suggest that Ephrin/Eph-mediated suppression of FGF/MAPK signaling is sufficient and necessary for neuronal differentiation in the neck, and that sustained FGF/MAPK signaling promotes an undifferentiated, neuroepithelial state instead.

Because our data had revealed upregulation of genes encoding the BMP antagonist Noggin and the Hedgehog (Hh) pathway effector Gli, downstream of Pax2/5/8.a (Fig. 4, Table S4), we also tested the effects of perturbing these pathways. We did not observe any obvious effects on neck morphogenesis upon overexpression of dominant-negative or constitutive BMP receptors (Fig. S9A-D) or CRISPR/Cas9-mediated knockout of Gli (Fig. S9E-G). Although these results do not rule out a role for BMP and/ Hh signaling pathways in the development of the neck or neck-derived neurons, we were unable to find any evidence that they impact the balance of differentiation and proliferation like Ephrin and FGF do.

Finally, because disruption of Ephrin/FGF/MAPK signaling significantly impacted cell proliferation and differentiation in larvae, we examined how these perturbations affected the development of CMNs in post-metamorphic juveniles. We observed a loss of Pax2/5/8.a>GFP⁺ CMN axons in juveniles that had expressed the dnFGFR receptor in the neck at the larval stage (Fig. 7A,B,E), but not in animals that expressed dnEph.c or caMEK (Fig. 7C-E). We conclude that the supernumerary, differentiating neurons seen in the larvae generated by Pax2/5/8.a>dnFGFR do not survive metamorphosis, failing to give rise to fully differentiated CMNs in the juvenile. In contrast, undifferentiated neuroepithelial cells elicited by dnEph.c or caMEK overexpression still retain the potential to differentiate into CMNs after metamorphosis. Taken together, these data suggest that the balance between neuronal differentiation in the larva and survival of neuronal precursors set aside for the adult CNS likely depends on a careful balance of FGF signaling prior to metamorphosis (Fig. 7F,G).

We have described the development of the Ciona ‘neck’ through metamorphosis and post-metamorphic neuronal differentiation. Despite reports of neck cells being quiescent in the larva, we observe ongoing proliferation and precocious adult neuron differentiation throughout the larval stage. We also show that the transcription factors Pax2/5/8.a and Phox2 are essential for the formation of adult (post-metamorphic) CNS neurons. Through RNAseq we have identified the neck transcriptional program downstream of Pax2/5/8.a, revealing candidate targets of Pax2/5/8.a that might be important for post-metamorphic neurodevelopment. Finally, although we did not identify an obvious role for either BMP or Hedgehog pathways, we establish that Ephrin and FGF signaling are responsible for patterning the neck into differentiated and undifferentiated compartments, with serious consequences for adult CNS formation if disturbed.

Across many species, Pax2/5/8 homologs are midbrain/hindbrain regulators, suggesting they are part of an ancestral gene regulatory network (Bridi et al., 2020; Schuster and Hirth, 2023). In vertebrates, Phox2 proteins regulate the development of various motor neurons of the head, which are also cholinergic (Pattyn et al., 2000; Mazzoni et al., 2013; Curto et al., 2015). For instance, Phox2a (along with Pax2/5) establishes oculomotor and trochlear cranial motor neurons at the isthmic organizer boundary between the midbrain and hindbrain (Deng et al., 2011; Fritzsch, 2024), and loss of Phox2b ablates cranial branchial neurons and visceral motor neurons of the hindbrain in mouse (Pattyn et al., 1999, 2000). Also in mouse, Phox2a promotes cranial motor neuron fate over spinal cord motor neuron fate (Mazzoni et al., 2013). In Ciona, Pax2/5/8.a⁺/Phox2⁺ neck cells give rise to post-metamorphic neurons that innervate pharyngeal gill slits and possibly branchiomeric siphon muscles of the adult later on (Wada et al., 1998; Dufour et al., 2006; Hozumi et al., 2015). Here, we show that Pax2/5/8.a and Phox2 are required for the formation of a subset of these neck-derived cholinergic CMNs. In contrast, Ciona Phox2 is not expressed in larval motor neurons, which innervate (paraxial) muscles of the tail, like vertebrate spinal cord motor neurons. It is unclear whether the neck and/or Pax2/5/8⁺/Phox2⁺ cells also give rise to additional neuron types of the juvenile/adult. Taken together, these findings suggest that a conserved transcriptional program for cholinergic branchiomeric efferent specification and differentiation might have evolved in the last common ancestor of tunicates and vertebrates, predating the origin of cranial nerves and jaw muscles.

Like many tunicates, Ciona have a biphasic lifecycle; a motile, non-feeding larval stage and a sessile, filter-feeding juvenile/adult stage. It has been traditionally assumed that the development of the larval and adult nervous systems of Ciona are neatly separated by metamorphosis. However, our observations suggest that the larval–adult boundary of the CNS might not be as sharp. Instead, we observe delamination of multiple GRIK⁺ cells in the posterior neck that project axons ventrally and anteriorly during the late larval stage (Fig. 2I), several hours before tail absorption and body axis rotation (Hotta et al., 2020). By following cells using a photoconvertible Kaede reporter, we find evidence that these larval GRIK⁺ cells survive metamorphosis, suggesting they might represent the earliest-differentiating CMNs of the juvenile. In juveniles/adults, CMNs have been shown to use cholinergic neurotransmission to regulate ciliary flow in the branchial epithelium, which generates the water flow that brings food particles into the mouth (Petersen et al., 1999; Jokura et al., 2020). Therefore, the early, pre-metamorphic differentiation of these neurons may be key for the rapid onset of feeding and further growth after metamorphosis. This is consistent with the ‘anticipatory development’ of post-metamorphic structures seen in diverse tunicate larvae (Jacobs et al., 2008). Unfortunately, without an axon-targeted Kaede fusion available we were unable to verify the morphology of photoconverted cells to confirm their identity as CMNs. Highly stable axon-targeted Kaede fusion proteins will have to be developed and adapted to Ciona to fully address this question.

Although Phox2 and GRIK show mutually exclusive expression domains at embryonic and larval stages, their reporters are co-expressed in post-metamorphic CMNs. Based on our interpretations, we believe Phox2 is activated de novo in previously Phox2-negative, GRIK⁺ larval cells (descended from A12.121) that give rise to early-differentiating CMNs. However, we cannot rule out that additional CMNs derive from Phox2⁺, GRIK-negative larval cells well after metamorphosis. As the branchial basket continues to grow, it will likely require additional innervation by Phox2⁺/GRIK⁺ CMNs. Whether this requires ongoing neurogenesis of neck-derived Phox2⁺/GRIK⁺ cells is not known, but is reasonable to assume that this may be the case. Furthermore, it is possible that there is developmental plasticity depending on the timing of larval settlement and/or rate of post-settlement growth. For instance, early-differentiating CMNs might be readily incorporated into the juvenile CNS when settlement and metamorphosis proceed rapidly, but they might not survive in cases of delayed metamorphosis. In the latter case, a pool of undifferentiated CMN progenitors might be relied upon to form CMNs when post-metamorphic growth resumes.

In addition to post-metamorphic CMNs, the neck also gives rise to the narrowly defined NNs, which here we show is the posterior-most cell of the neck lineage on either side of the larva, forming a bilaterally symmetric left-right pair of neurons. The NNs have been previously described at the early larval stage in the C. intestinalis whole-larva connectome studies (Ryan et al., 2016; Ryan and Meinertzhagen, 2019). The function of the NNs remains unknown, but the connectome described them as receiving synaptic inputs from ascending MG interneurons (AMG neurons) and synapsing primarily onto the basement membrane (Ryan et al., 2016). However, NNs might not be fully differentiated in early larvae, and it is possible they form connections later in larval development. Indeed, we noticed that NN axons continue to extend posteriorly during the larval phase and appear to exit the trunk and enter the proximal portion of the tail prior to larval settlement. Alternatively, the NNs might not serve a function other than to prevent nascent CMN axons from projecting posteriorly during the larval phase. This ‘roadblock’ effect is observed when MG neurons are ‘twinned’ under certain genetic perturbation conditions, in which ectopic neurons project anteriorly instead of posteriorly into the tail (Stolfi et al., 2011). Regardless of function, the NNs likely do not survive metamorphosis as we rarely observed them in juveniles. Interestingly, Pax2/5/8.a>dnFGFR converted the entire neck lineage into neurons that did not appear to survive metamorphosis (Fig. 7B). Because of this, we believe that the supernumerary neurons generated by dnFGFR represent NN-like neurons that do not persist to the adult stage. Although the supernumerary axons in the dnFGFR condition did not project towards the tail like the NNs, this may be due to disrupted neuronal polarity mechanisms. Consistent with this hypothesis, we have previously observed supernumerary MG neurons aberrantly project away from the tail (Stolfi et al., 2011). Future studies aim to identify regulatory differences between CMNs and NNs and to investigate what factors determine whether they survive or perish despite their shared lineage history. For instance, it is not yet clear whether sustained FGF signaling in neural progenitors is inherently pro-survival, or if FGF downregulation in the neck is simply a molecular switch for the specification of NNs, which may be pre-programmed to degenerate during metamorphosis.

It is possible that cell survival factors act downstream of Pax2/5/8.a, including those we identified here using RNAseq. One interesting candidate is Vanabin4 (Van4), which encodes a vanadium-binding protein and was the top Pax2/5/8.a target by RNAseq. The biological function of vanadium remains elusive, and it is unclear what the role of Van4 might be in the neck (Ueki et al., 2015). Many tunicates accumulate high levels of vanadium, but not all species have vanabins and their purpose remains unknown (Ueki et al., 2003). One theory is that adults accumulate vanadium to deter predation (Odate and Pawlik, 2007). It is possible that vanadium provides a protective effect against oxidative stress (Tripathi et al., 2018), which would promote survival during metamorphosis. Future studies will examine the role of vanabins in Ciona and other marine organisms that accumulate vanadium (Thompson et al., 2018; Mendonca et al., 2024).

Adult Ciona robusta (intestinalis Type A) were collected from San Diego, CA (M-REP). Dechorionated zygotes were generated and electroporated as previously described (Christiaen et al., 2009a,b). Embryos were raised in artificial sea water at 20°C. Animals raised beyond 24 hpf were moved to new agarose-coated plates with fresh artificial sea water containing 1.0% Penicillin-Streptomycin (Gibco) daily. Post-metamorphic animals were paralyzed with L-menthol prior to fixation, as previously described (Osugi et al., 2020). Staging is based on the TUNICANATO database (Hotta et al., 2020). Cell lineage nomenclature is based on Conklin (Conklin, 1905; Nicol and Meinertzhagen, 1991). All sequences of plasmids, probes and sgRNAs not previously published can be found in the supplementary Materials and Methods.

Sample processing for fluorescence and immunostaining were performed as previously described (Beh et al., 2007; Ikuta and Saiga, 2007; Stolfi et al., 2011). Unc-76-tagged GFP and mCherry (Dynes and Ngai, 1998) were used to improve labeling of cell bodies and axons, instead of GFP/mCherry alone. An Unc-76-tagged StayGold green fluorescent protein was also designed and used for its improved signal longevity (Hirano et al., 2022). Immunolabeling of Cas9 protein by monoclonal mouse anti-Cas9 antibody at 1:500 (4G10; Diagenode) and GFP protein by mouse anti-GFP antibody at 1:500 (Supply Solutions Roche, clones 7.1 and 13.1) was performed with blocking in PBS Super Block (37580; Thermo Fisher Scientific) and visualized with goat anti-mouse Alexa Fluor 568 or goat anti-mouse Alexa Fluor 488 (Thermo Fisher Scientific). Detection of MAP kinase pathway activity with dpERK antibody (1:500; Sigma-Aldrich, M9692) was performed as previously described (Stolfi et al., 2011, 2013). Juvenile animals were stained with phalloidin-Alexa Fluor 405 or 647 (Thermo Fisher Scientific; A30104, A22287) to visualize F-actin.

In situ hybridization coupled to immunostaining was performed as previously described (Beh et al., 2007), using TSA Plus amplification kits (Akoya Biosciences) and mouse anti-β-galactosidase (Promega, Z378; 1:1000) or rabbit anti-mCherry (BioVision, accession number ACY24904; 1:500) primary antibodies. Two-color (fluorescein and Cy3) double in situ hybridization was performed using TSA Plus amplification kits as previously described (Ikuta and Saiga, 2007; Stolfi et al., 2011). Probe template sequences can be found in the supplementary Materials and Methods.

In Kaede (Ando et al., 2002) photoconversion experiments, animals were electroporated with a plasmid using the GRIK cis-regulatory sequences to drive expression of Kaede fused to a nuclear localization sequence (Kaede::NLS) (Razy-Krajka et al., 2014). At 22 hpf, animals were transferred to an uncoated 10 cm plastic dish in artificial sea water (Instant Ocean Sea Salt). Half of all animals immediately underwent the photoconversion process and the other half were returned to the incubator. The Kaede construct was photoconverted using 400 nM light from an optoWELL (Opto Biolabs) at 100% intensity for 35 min. After photoconversion, animals were replaced in the incubator until post-metamorphic fixation at 44 hpf. Juveniles were fixed as previously described.

All standard images were captured on a Leica epifluorescence compound microscope. Confocal images were captured on a Nikon AX R with NSPARC and processed in ImageJ (1.54d). Phenotypes were quantified on a Leica DMi8 or DMIL LED inverted epifluorescence microscope. Biological replicates are presented in side-by-side graphs generated in Prism (9.5.1). Cell count data were analyzed by one-way ANOVA and Dunnett's test of multiple comparisons. Phenotypic data were analyzed by Fisher's exact test comparing each experimental group to the control. Significance was reported when appropriate with a minimum threshold of P<0.05. Fluorescence profiles and region of interest intensity of anti-dpERK-stained embryos was quantified in ImageJ and analyzed in Prism.

sgRNAs were designed using CRISPOR (http://crispor.tefor.net/) and vectors were constructed as previously described (Haeussler et al., 2016; Gandhi et al., 2018) or custom synthesized and cloned (Twist Bioscience). Individual sgRNA vectors were validated in vivo as previously described (Johnson et al., 2024) with a ubiquitous Eef1a −1955/-1>Cas9 or Eef1a −1955/-1>Cas9::Geminin (Stolfi et al., 2014a; Johnson et al., 2024). Genomic DNA was isolated using QiaAMP Micro extraction kit (QIAGEN), targeted regions amplified by PCR using AccuPrime Pfx (Thermo Fisher Scientific), and PCR products purified using the QiaQuick PCR Purification kit (QIAGEN) following the published protocol (Johnson et al., 2024). Samples were sequenced using commercial Illumina-based amplicon sequencing (Amplicon-EZ; Azenta). Sequence data were analyzed by Azenta, and sgRNA efficiency was determined by the percentage of sequences containing insertions and/or deletions (indels) in and around the sgRNA target site. Data on sgRNA efficiency is provided in Fig. S10. Details of the electroporation mixes can be found in the supplementary Material and Methods. Sequences for sgRNAs and PCR primers are provided in the Table S5; gene model nomenclature can be found in Table S6.

Fertilized zygotes were prepared and electroporated as described above. Zygotes were electroporated with DNA plasmids using a pan-neuronal driver Nut −1155/-1 (Shimai et al., 2010 ;Johnson et al., 2024): Nut-1>LacZ Control, Nut-1>Pax2/5/8.a transcript variant 1 (‘tv1’, transcript model KY21.Chr6.690.v2.SL1-1), or Nut-1>Pax2/5/8.a transcript variant 2 (‘tv2’, transcript model KY21.Chr6.690.v2.SL1-1). The experiment was repeated in duplicate. RNA was isolated from 10 hpf embryos using a Monarch total RNA miniprep kit (New England BioLabs Inc.) RNA was stored at −80°C until sample analysis and sequencing by the Molecular Evolution Core at Georgia Tech as previously reported (Johnson et al., 2024). Briefly, total RNA integrity levels were using by the Agilent Bioanalyzer RNA 6000 Nano kit and all samples had RINs above 9. Enrichment for mRNAs was performed using the NEBNext Poly(A) mRNA isolation module and Illumina libraries were prepared by the NEBNext Ultra II RNA directional library preparation kit. Libraries were pooled and sequenced on the NovaSeq 6000 with an SP Flow Cell, to obtain PE100 bp reads. Pax2/5/8.a transcriptional variant fold change Pearson correlation analysis and graphs generated in Prism (9.5.1).

Data quality control and analysis were performed in Galaxy (usegalaxy.org) (Galaxy Community, 2022). Raw reads were quality controlled with FastQC and Cutadapt. Reads were mapped to the HT_KY21 C. robusta genome with RNA STAR and checked using the Integrative Genomics Viewer (IGV; Version 2.14.1) (Satou et al., 2005, 2022). The number of reads per annotated genes were counted using featureCounts and DESeq2 was then used on the read counts to normalize them to the controls. Datasets were then annotated with the HT_KY21 genome (Satou et al., 2022). KY21 gene models were linked to KH gene models using the Ciona Gene Model Converter https://github.com/katarzynampiekarz/ciona_gene_model_converter. For each step, quality reports were aggregated using MultiQC. Raw FASTQ files can be found in the SRA database under BioProject accession number PRJNA981160. Analyzed data are provided in Table S1.

Re-processed scRNAseq data from Cao et al. were analyzed (Cao et al., 2019; Johnson et al., 2024). Data from Ciona embryos in the Mid-tailbud II stage (roughly 10 hpf) were analyzed using the Seurat v3 package in R to identify cell type clusters based on unique gene expression markers (Fig. S6A) (Satija et al., 2015; Stuart et al., 2019). Replicates were integrated and pre-processing and clustering were performed using the SCtransform and FindMarker functions (Hafemeister and Satija, 2019). Cluster 25 was identified as containing neck cells based on high enrichment for Pax2/5/8.a reads (Fig. S6B, Table S2). Cluster 25 cells were re-clustered based on differential gene expression, resulting in two sub-clusters (see ‘Results’ section). The FeaturePlot function was used to visualize feature expression in low-dimensional space, and to use colors to map out the relative expression levels for each gene of interest in the neck and brain clusters. Additionally, VlnPlot and RidgePlot functions were applied to the expression distributions within the clusters, allowing for the heterogeneity of the neck cluster to be further examined and the potential for additional sub-populations noted. R objects and code can be accessed on OSF at https://osf.io/uc32x/.

RNA was isolated from 17 hpf C. robusta larvae using the Monarch Total RNA miniprep kit (New England BioLabs). cDNA was produced using the Omniscript reverse transcription kit (QIAGEN). The resulting cDNA was diluted in nuclease-free water and stored at −20°C. Based on the RNA model discovered during RNA sequence analysis, forward (5′-CATAACGATGGACTACCCTGC) and reverse (5′-CAGACATGTCGTGGTAGGATAGG) primers targeting the new exon 1 of Phox2 and stop site of KY21.CH14.159, respectively, were used to perform PCR on the single-strand cDNA with OneTaq 2X Master Mix (NEB, M0482S) in 50 μl reactions with a touchdown PCR program. PCR products were purified (QIAquick PCR purification kit; QIAGEN) and TOPO-cloned into a dual promoter empty vector (Thermo Fisher Scientific, 450640). White colonies grown on LB plates with 100 mg/ml ampicillin and coated with X-Gal were selected and tested for an insert of correct size by restriction enzyme digest. Plasmids of the appropriate size were sequenced by Sanger sequencing with M13 forward and reverse primers (Eurofins Genomics).

We thank Shohon Rafique, Tanner Shearer and Lindsey Cohen for technical assistance. We thank members of the lab for feedback and suggestions. We thank the Optical Microscopy Core for assistance with confocal imaging. We thank Shweta Biliya of the Molecular Evolution Core facility at Georgia Tech for processing bulk RNAseq samples. We thank Bernd Fritzsch for helpful feedback on the manuscript.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202719.reviewer-comments.pdf