Larval tunic formation in Phallusia mammillata

Previous studies on the molecular control of larval tunic formation have been performed on Ciona robusta (Ciona intestinalis type A) [4, 7, 13]. We thus first described larval tunic formation during P. mammillata development. Since tunic is transparent, we have used three dyes, calcofluor white, DirectRed23, and CBM3a-GFP (see “Methods”) to visualize the tunic. Although all dyes may not be specific to cellulose since they might bind other polysaccharides [6], they were useful to follow the course of tunic formation (Fig. 1; Additional file 2: Fig. S2). Tunic was first seen at late tailbud stages (St. 24) as a thin layer surrounding the embryo (Fig. 1A). It progressively thickened with a small caudal fin visible at St. 25 (Fig. 1B). At hatching (St. 26), extended median and caudal fins were readily visible (Fig. 1C, D).

Fig. 1

Tunic and fin blades development in Phallusia mammillata. A–C Embryos at St. 24 (A) and St. 25 (B) that developed in their chorions, and a hatching larva (C) were stained with calcofluor white (blue) and Sytox Green (green). Maximum intensity projections from confocal z-stacks (the corresponding 3D visualizations can be found in Additional file 3: Movie S1, Additional file 4: Movie S2 and Additional file 5: Movie S3). D 3D surface rendering from a confocal z-stack of a larva stained with CBM3a-GFP (green) and DAPI (blue) (the corresponding 3D visualization can be found as Additional file 6: Movie S4). E–H Effects of temperature during development on the formation of the tunic (stained with CBM3a-GFP in green) of larvae that arose from dechorionated eggs. Although the tunic of larvae that developed at 18 °C was devoid of median and caudal fins most of the time (E), the tunic of larvae that developed at 22 °C had more frequently median fins with a normal appearance and caudal fins with a reduced size (G). Chorionated embryos systematically gave rise to larvae with well-formed median and caudal fins (F,H). I Graph representing the proportions of larvae presenting median (blue) and caudal (orange) fins at the different culture temperatures (average values with error bars denoting standard deviation). With chorion: n = 78 (3 experiments: one at 16 °C, one at 18 °C, and one at 24 °C). 18 °C without chorion: n = 141 (3 independent experiments). 24 °C without chorion: n = 123 (3 independent experiments). 22 °C without chorion: n = 195 (4 independent experiments). 16 °C, then 22 °C without chorion: n = 48 from a single experiment. Individual data values can be found in Additional file 7: Table S1. Note that although the tunic was uniformly stained by CBM3a-GFP in dechorionated larvae, the tunic staining of chorionated larvae was always less intense in the middle part. Embryos in A and B have developed in their chorions, consequently they appear rolled up. Larvae in C–H are shown in lateral views with anterior to the left and dorsal to the top. Scale bars: 50 µm

In order to manipulate ascidian embryos, the protective chorion is usually removed before or just after fertilization. It has been known for long time that embryos deriving from such treatment are not fully normal. In particular, left/right asymmetry and tunic formation are disrupted [21, 24, 25]. When we observed the tunic of larvae that derived from eggs whose chorion was removed before fertilization (using our routine protocol [26]), the tunic was stained using cellulose dyes, but median and caudal fins were extremely reduced in size, abnormal or absent (Fig. 1E, I) when compared with larvae that developed within their chorions (Fig. 1F, I). We fortuitously found that increasing the temperature of embryonic development partly compensated for the absence of chorion. Dechorionated larvae that resulted from a development at 22°C or 24°C possessed fins more frequently (this frequency being quite variable from batch to batch), albeit usually reduced in size (Fig. 1G, I). Interestingly, increasing the temperature before tunic and fin formation at early tailbud stages (St. 20/21) was not sufficient to recover fin formation (Fig. 1I). This observation suggests that early events (possibly gene regulation), rather than a biophysical effect of temperature on tunic production, are the targets of temperature.

Tail epidermis patterning regulates median fin morphogenesis

Median fin blades elongate above a specific epidermal cell population, the tail neurogenic midlines (Fig. 2A–C). We first inhibited the BMP signaling pathway that is necessary for ventral tail midline fate acquisition [20, 23, 27]. Strikingly, treated larvae had no ventral fin (DMH1: 71% of larvae with median fin had only a dorsal fin, n = 188; DMSO/BSA-control: 100% of larvae with median fin had both dorsal and ventral fins, n = 181; results from four independent experiments; Fig. 2D, E). A reciprocal experiment, treatment with recombinant BMP2 protein, leads to ectopic ventral midline fate in the entire tail epidermis [22, 23]. Contrary to our expectations, we did not observe a radial median fin all around the larval tail following such a treatment (Fig. 2F). A median fin was not recognizable anymore and the tunic was somehow inflated making bulges all around the tail (this phenotype was observed in 99% of the embryos, n = 272; results from three independent experiments). This observation is consistent with an excess of tunic production. We performed the same treatments on C. intestinalis embryos (Additional file 8: Fig. S3). Although DMH1 treatment also led to a loss of ventral median fin, BMP2 treatment was different: the median fin was readily visible but numerous isolated fibers were seen protruding out of the tunic. While we are not yet able to explain the differences in phenotype for both species, the results argue for an excess of tunic production following BMP pathway activation and epidermis midline fate expansion.

Fig. 2

Molecular patterning of the tail epidermis regulates median fin formation. A–C Schematic representation of an early tailbud (A), a larva in lateral view (B), and a cross-section of the larval tail (C) with the embryo/larva itself in gray surrounded by the tunic. The tail epidermal neurogenic midlines that lie where median fins are positioned are highlighted in light blue. D–F Effects of BMP pathway modulation on median fin formation: control (BSA- and DMSO-treated larva) (Dii), DMH1-treated larva did not develop a ventral fin (Eii), and BMP2-treated larva with an excess of tunic making bulges (Fii). G–I Effects of CRISPR/Cas9 gene inactivation on tunic formation: Tyr-CRISPR larva formed a normal tunic with fins but lacked pigment cells (Gii), Msx-CRISPR larva lacked both ventral and dorsal median fins and had normal pigment cells (Hii), and Klf1/2/4/17-CRISPR larva presented indentations in the median fin that had an undulated shape (Iii). CBM3a-GFP staining of larvae in D–I (green). Insets in Gii and Hii: transmitted light picture of the trunk. Inset in Iii: dorsal view focused on the median fin. Larvae are shown in lateral view with anterior to the left and dorsal to the top. On the left panels (i): schematic representation of the distribution of the midline fate (light blue) in the different conditions. White arrows highlight the absence of median fin. Individual data values can be found in Additional file 7: Table S1. Scale bar: 100 µm

We next aimed at targeting downstream genes that belong to the tail PNS GRN using CRISPR/Cas9 gene editing. We used microinjection-mediated introduction of the ribonucleoprotein complex that has been recently described for Phallusia [28]. As a control, we targeted Tyrosinase that is essential for melanin pigmentation of sensory organs found in the larval brain, the otolith, and the ocellus. Absence of pigment cells was observed in 41% of larvae on average (59% with 2 pigment cells, 6% with 1 pigment cell, and 35% with no pigment cell; n = 101 from three independent experiments; Fig. 2G). We first examined the effects of mutating Msx, the gene coding for a homeodomain-containing transcription factor that lies at the top of the GRN [22, 29]. As anticipated, median fin formation was inhibited (Tyr-CRISPR: 79% of larvae with a median fin, n = 78; Msx-CRISPR: 35% of larvae with a median fin, n = 91; results from two independent experiments; Fig. 2H). The penetrance of this phenotype was not complete, but was in agreement with the fact that we observed that 6/10 larvae were mutated at the Msx locus in a separate experiment (Additional file 9: File S1). We then mutated Klf1/2/4/17 that codes for a Zn-finger transcription factor acting downstream of Msx [29]. Surprisingly, we did not phenocopy Msx-CRISPR larvae. The frequency of fin formation was similar to control Tyr-CRISPR larvae (Tyr-CRISPR: 87% median fin, 43% caudal fin, n = 64; Klf1/2/4/17-CRISPR: 82% median fin, 46% caudal fin, n = 52; results of two independent experiments). However, median fins were strongly malformed with a seemingly random local reduction in size leading to a wavy edge of the median fins (F ig. 2I). This phenotype that was present in 77% of the larvae with median fins was never observed in Tyr-CRISPR, non-injected or non-dechorionated larvae. In addition, the median fin had an undulated shape, (Fig. 2Iii inset). Genotyping indicated that 7/8 larvae had a mutated Klf1/2/4/17 locus in an independent experiment (Additional file 9: File S1).

The results of this section demonstrate that acquisition of midline fate in the epidermis is essential for median fin formation.

Phylogenetic distribution, embryonic expression, and transcriptional regulation of cellulose-related HGT-acquired genes in ascidians

We hypothesized that elongation of the larval tunic into fin blades could be the result of specific regulation of cellulose production. We thus examined the phylogenetic distribution and embryonic expression of the two cellulose-related HGT-acquired genes CesA and Gh6 in ascidians.

Pan epidermal expression of CesA

When searching for the ortholog of CesA in P. mammillata genome, we found two hits located on different scaffolds. Although predicted proteins differed in size, they had similar structures with seven transmembrane domains, a glycosyl transferase 2 (GT2)/cellulose synthase domain and a glycosyl hydrolase 6 (GH6)/cellulase domain (Additional file 10: Fig. S4). Both genes had similar temporal dynamics (expressed during late embryonic development), but their levels of expression were highly different, CesA.b being expressed at very low levels compared to CesA. To verify that the existence of these two genes was not an artifact of the Phallusia genome assembly, we surveyed tunicate genomic and transcriptomic data for CesA genes. As previously reported, we found a single CesA for most tunicates except Oikopleura dioica for which a lineage-specific duplication occurred [6, 30]. However, in addition to P. mammillata, we found three species that had 2 CesA genes (Additional file 11: Fig. S5; Additional file 17: Table S3): Phallusia fumigata, Ascidia mentula, and Corella inflata. Phylogenetic analysis demonstrated the existence of three clades (Additional file 11: Fig. S5): one that includes most previously described CesA, one with CesA proteins from Oikopleura, and one that includes the second copy present in these four species (we named these proteins CesA.b). The chromosomal assembly of A. mentula indicates that CesA and CesA.b are located on different chromosomes (Additional file 17: Table S3), suggesting that recent local tandem duplication is not the source of CesA.b emergence.

We next determined the expression of CesA in three species: P. mammillata, A. mentula, and Molgula appendiculata by in situ hybridization (Fig. 3A–C; Additional file 12: Fig. S6). For all species, CesA had a pan-epidermal expression during embryonic development like in Ciona [4, 5, 13]. However, in M. appendiculata, the expression was initiated much earlier (early gastrula stages) than in the other three species (late neurula stages). Despite CesA.b was expressed at much lower levels (Additional file 10: Fig. S4), we managed to detect its expression pattern during embryogenesis by in situ hybridization in P. mammillata (Fig. 3D; Additional file 12: Fig. S6). CesA.b was broadly expressed in the epidermis of late tailbud embryos (St. 24) with weaker expression in the tail ventral and dorsal midlines.

Fig. 3

CesA expression and regulation in different ascidian species. A–C In situ hybridization for CesA during embryonic development of Phallusia mammillata (A), Ascidia mentula (B), and Molgula appendiculata (C). In P. mammillata, the expression was first detected in the caudal part of late neurulae (Aii). Expression was found throughout the epidermis with weaker staining in the palp forming region (Aiv). A cross-section of the tail shows that the staining was limited to the epidermis and not found in internal tissues (Av). A similar pattern was found in A. mentula (B) and M. appendiculata (C) but with a much earlier onset of expression in the latter species (early gastrula: Ci). D In situ hybridization for CesA.b during embryonic development of P. mammillata. Transcripts were found throughout the epidermis from late tailbud stages (St. 24) with weaker expression in dorsal and ventral tail midlines (Diii–Dv). Embryos at the following stages are shown: gastrula (St. 10–11) (Ai,Bi,Ci), neurula (St. 14–16) (Aii,Bii,Cii), early/mid tailbud (St. 20–22) (Aiii-Av,Biii,Ciii,Di,Dii), and late tailbud (St. 24–25) (Avi,Biv,Civ,Diii-Dv). E–G Transcriptional regulation of Phmamm.CesA in the epidermis. A region of around 1 kb, Phmamm.CesA.up1, situated immediately upstream of CesA (E) was PCR-amplified and placed upstream of the LacZ reporter in inverted orientation. When tested in P. mammillata embryos (F), it was inactive at early tailbud stages (St. 19/20, no embryo stained in the epidermis, n = 185, a single experiment) and active in the epidermis at late tailbud stages (St. 23/24, 60% of the embryos with epidermis staining, n = 703 from 3 independent experiments). This region contains 2 putative Tfap2 binding sites (yellow bars in G). A synthetic version of Phmamm.CesA.up1, Phmamm.CesA.up, and 3 variants with Tfap2 mutations were placed in “correct” orientation upstream of LacZ, and their activity tested at mid-tailbud stages (St. 22) is summarized on the graph on the right in G (Phmamm.CesA.up, n = 296; Phmamm.CesA.up_mut1, n = 282; Phmamm.CesA.up_mut2, n = 363; Phmamm.CesA.up_mut12, n = 268; results from three independent experiments). H Swap assay of putative CesA promoters from three ascidian species. Each region was tested in both P. mammillata (Phmamm, blue) and C. intestinalis (Ciinte, orange), and the activity at mid/late tailbud stages in epidermis is displayed on the graph on the right (Phmamm.CesA.up1: in Phmamm n = 703 from 3 experiments, in Ciinte n = 348 from 2 experiments; Asment.CesA.up1: in Phmamm n = 899 from 4 experiments, in Ciinte n = 896 from 3 experiments; Moappe.CesA.up1: in Phmamm n = 767 from 3 experiments, in Ciinte n = 712 from 3 experiments). All pictures of embryos are lateral views with dorsal to the top and anterior to the left, except: animal views (Ai,Bi), vegetal view (Ci), neural plate views (Bii,Cii) with anterior to the left, frontal view (Aiv), dorsal view (Div), ventral view (Dv), and cross-section of the tail (Av). Individual data values can be found in Additional file 7: Table S1

Shared and divergent mechanisms regulating CesA expression in the epidermis

We isolated Phmamm.CesA.up1, a 1-kb genomic fragment immediately upstream of Phmamm.CesA, that proves sufficient to drive epidermal activity (Fig. 3E, F). Since Cirobu.CesA is regulated by the transcription factor Tfap2-r.b [31], we searched for Tfap2 binding sites and identified two putative sites (Fig. 3G). Mutagenesis followed by in vivo transcriptional assay indicated that the proximal site was dispensable (this site was actually predicted only by the GCCN3/4GGC motif, and not by the Jaspar matrices) and the distal site participated in CesA expression (Fig. 3G). In fact, binding site mutation only diminished reporter activity whereas mutation of the single Tfap2 site fully abolished reporter activity in the Ciona experiment [31]. To further probe potential differences between Ciona and Phallusia, we tested the transcriptional activity of Phmamm.CesA.up1 in C. intestinalis. Surprisingly, it was inactive (Fig. 3H). We isolated Asment.CesA.up1, a 1.6-kb long putative promoter from A. mentula that contains three predicted Tfap2 binding sites. Similarly to Phmamm.CesA.up1, this region was active in Phallusia epidermis but not in Ciona epidermis (Fig. 3H). We next tested a candidate region from a distantly related species M. appendiculata. This 0.6-kb region containing two putative Tfap2 sites was equally active in both Phallusia and Ciona albeit at weaker levels than previous genomic fragments.

Patterned epidermal expression of Gh6

We had identified the presence, in tunicate genomes, of another HGT gene originating from bacteria that codes for a most likely functional cellulase containing a transmembrane domain and an extracellular glycosyl hydrolase family 6 domain [32] that we named Gh6. Before the present study was completed, this finding was published by colleagues [14]. Very recently, Gh6 was described in C. robusta as being expressed in the epidermis during embryogenesis and as regulating larval tunic formation and metamorphosis [13]. We confirmed the published phylogeny of tunicate Gh6 proteins (Additional file 13: Fig. S7). We identified a single gene for each species, except for the Thaliacean Salpa thompsoni that presented a lineage-specific duplication (Additional file 13: Fig. S7) [14].

We examined the expression pattern of Gh6 during the embryonic development of four ascidian species. As expected, Ciinte.Gh6 was expressed like Cirobu.Gh6 (Fig. 4A) [13] with first detectable expression in the epidermis of the tip of the tail at late neurula/early tailbud stages (Fig. 4Aii). Novel domains of expression in the epidermis appeared at late tailbud stages: palp region, dorsal epidermis at the trunk/tail junction, and at the dorsal and ventral aspects of the tail (Fig. 4Aiv). We found similar expression patterns in P. mammillata and A. mentula (Fig. 4B, C; Additional file 12: Fig. S6). In Phallusia, we identified tail expressing cells as the four medio-lateral longitudinal rows of cells (two ventral and two dorsal rows) [20] (Fig. 4Biv-Bvi, Bviii). In the palp region, Gh6 transcripts were depleted from future papillary protrusions (Fig. 4Bvii). By contrast, Moappe.Gh6 had a very different pattern (Fig. 4D). Although it was expressed in the epidermis, the expression was detected earlier (neurula stages) and very broadly (entire epidermis with a depletion from the posterior tail as development proceeds).

Fig. 4

Gh6 expression and regulation in different ascidian species. A–D In situ hybridization for Gh6 during embryonic development of Ciona intestinalis (A), Phallusia mammillata (B), Ascidia mentula (C), and Molgula appendiculata (D). For the three phlebobranch species, Gh6 was first detected at early tailbud stages in the epidermis at the tip of the tail. Then expression started anteriorly in the palp region and in the tail epidermis at late tailbud stages. Tail epidermis expression was detected in the four medio-lateral rows of cells (Bv,Bvi,Bviii). Expression in the palp region was likely absent from the future protruding papillae (Bvii). By contrast, Moappe.Gh6 was expressed broadly in the epidermis starting at neurula stages (Dii–Div). Embryos at the following stages are shown: gastrula (St. 10–11) (Bi,Ci,Di), neurula (St. 14–16) (Ai,Bii,Cii,Dii), early/mid tailbud (St. 20–22) (Aii,Aiii,Biii,Ciii,Diii), and late tailbud (St. 24–25) (Aiv,Biv–Bviii,Civ,Div). E Transcriptional regulation of Phmamm.Gh6 in the epidermis. A 3.6-kb region starting at the beginning of the scaffold 1127 immediately upstream of Gh6, Phmamm.Gh6.up1, was PCR-amplified, placed upstream of the LacZ reporter, and tested in vivo. It was active only in a part of the Phmamm.Gh6 expression domain, the tail epidermal medio-lateral cells at late tailbud stages (St. 19–22: 3% of embryos with staining in this expression domain, n = 401; St. 24–25: 75% of embryos with staining in this expression domain, n = 521; results from four experiments). Phmamm.Gh6.up2, a 1.5-kb region embedded in Phmamm.Gh6.up1 containing conserved segments, had a similar activity (St. 19–22: 0% of embryos with this expression domain, n = 670 from three experiments; St. 24–25: 56% of embryos with this expression domain, n = 556 from four experiments). F Transcriptional regulation of Cirobu.Gh6 in the epidermis. Cirobu.Gh6.up1, a 2.7-kb region that almost corresponds to the entire upstream intergenic region, fully recapitulated Gh6 expression with tail tip activity (arrows) detected in early tailbuds, and palp (arrows) and tail epidermis activity in late tailbuds (St. 19–22: 45% of embryos with expression in endogenous territories, n = 300; St. 24–25: 73% of embryos with expression in endogenous territories, n = 300; results from two experiments). G Phmamm.Gh6.up1 was not active in C. intestinalis embryos (n = 367 from two experiments). H Cirobu.Gh6.up1 was active in Gh6 expression domains in P. mammillata embryos (St. 21–22: 36% of embryos with expression in endogenous territories, n = 448 from three experiments; St. 23–24: 72% of embryos with expression in endogenous territories, n = 210 from two experiments). All pictures of embryos are lateral views with dorsal to the top and anterior to the left, except: animal views (Bi,Ci), vegetal view (Di), neural plate views (Ai,Cii,Dii) with anterior to the left, frontal view (Bvii), dorsal view (Bv), ventral view (Bvi), and cross-section of the tail (Bviii). Individual data values can be found in Additional file 7: Table S1

Shared and divergent mechanisms regulating Gh6 expression in the epidermis

To apprehend transcriptional regulation of Gh6, we isolated candidate cis-regulatory regions in both Phallusia and Ciona. In Phallusia, the largest region we could test (the Gh6 locus lies at the edge of the currently available P. mammillata genome assembly MTP2014) reproduced late expression in the medio-lateral domains of the tail epidermis but not the palp and tail tip regions (Fig. 4E). A smaller fragment with DNA sequences conserved between P. mammillata and P. fumigata had a similar activity. In Ciona, the intergenic region upstream of Gh6, with no detectable DNA conservation between C. robusta and C. savignyi, recapitulated both the early tail tip expression and the later palp and medio-lateral tail epidermis expression (Fig. 4F). Surprisingly, although the regulatory elements isolated from Phallusia were inactive in Ciona, the elements from Ciona were active in Phallusia (Fig. 4G, H).

Altogether, the results from this section indicated an unexpected variability of CesA and Gh6 at all level examined: phylogenetic distribution in tunicates, expression domains, and possibly expression regulation. Nevertheless, these HGT-acquired genes are clearly deployed in the epidermis during late embryogenesis.

Gh6 regulates caudal fin formation

As described previously (Fig. 2), modifying tail epidermis patterning impacted fin blade formation. Since CesA.b and Gh6 displayed a patterned expression in the epidermis, we determined their expression following BMP pathway modulation (Additional file 14: Fig. S8). Their expression patterns were modified in agreement with their site of expression: loss of ventral medio-lateral Gh6 expression and ectopic CesA.b expression at the ventral midline when BMP was inhibited. We thus functionally evaluated the role of Gh6 in tunic and fin blade formation. We selected this gene for two reasons. First, Gh6 is expressed closely to where median and caudal fins emerge (Fig. 4B). Second, by analogy with data from plants, we postulated that Gh6 may contribute to fin blade elongation through local increase of cellulose production [33]. We first overexpressed Gh6 using pFog, an early pan-ectodermal driver that we have extensively used in Ciona and that is also active in Phallusia [20, 22, 23, 29]. We were surprised that median and caudal fins formed with the same frequency as in electroporation controls (Additional file 15: Fig. S9). We overexpressed Gh6 using two additional drivers (pSoxB2 active in the tail epidermis from the beginning of gastrulation [29] and Phmamm.CesA.up1 (Fig. 3)). We did not detect any effect on fin formation (Additional file 15: Fig. S9). To test the above model for Gh6 function, we nevertheless turned to gene inactivation using CRISPR/Cas9 (Fig. 5). Contrary to our expectations, Gh6 inactivation did not abolish fin elongation but rather promoted an opposite fin phenotype. While the median fin was present similarly in both control-CRISPR larvae (in 73% of larvae, n = 136 from two experiments) and Gh6-CRISPR larvae (69%, n = 155 from two experiments), the frequency of caudal fin presence increased from 48% in control-CRISPR larvae to 75% in Gh6-CRISPR larvae. Mutations of the Gh6 locus by CRISPR/Cas9 were observed in 7/10 larvae that were genotyped independently (Additional file 9: File S1). Overall, our functional analysis indicates that Gh6 is not the major effector of fin morphogenesis downstream of the epidermis patterning developmental network or that it plays a minor role possibly in conjunction with other yet unidentified actors.

Fig. 5

Gh6 regulates caudal fin formation. Larvae resulting from microinjection of the CRISPR/Cas9 components targeting the Brlanc.Ascl1/2.1 gene, a sequence absent from the P. mammillata genome (Control-CRISPR, A–C) or the Phmamm.Gh6 gene (Gh6-CRISPR, D–F) were stained with CBM3a-GFP to visualize the tunic and fin blades. The caudal fin of the control larva ended where the tail ends (B,C). In contrast, the caudal fin of the Gh6-CRISPR larva was well developed. Scale bar: 50 µm. Individual data values can be found in Additional file 7: Table S1