We had previously identified Ciona robusta Xbp1 (Cr-Xbp1, formerly Ci-XBPa; gene model: KH.C4.516) and showed that this gene is predominantly expressed in notochord and epidermis (Figure 1A and A’); we had also observed that Cr-Xbp1 is overexpressed in embryos ectopically expressing Ci-Bra and downregulated in transgenic embryos expressing a repressor form of Ci-Bra (Kugler et al., 2008). To directly verify the requirement of Ci-Bra for Cr-Xbp1 notochord expression, we performed whole-mount in situ hybridization (WMISH) on embryos carrying a null mutation in the Ci-Bra coding region (Chiba et al., 2009). Compared to stage-matched controls hybridized in parallel (Figure 1B), embryos lacking Ci-Bra function show normal Cr-Xbp1 expression in epidermal cells and in a small region of the anterior sensory vesicle, but lack Cr-Xbp1 expression in the notochord (Figure 1C).

To investigate the role of Cr-Xbp1 in notochord development, we generated different constructs aimed at interfering with its activity. First, we cloned the region encoding its first 188 amino acid residues downstream of the Ci-Bra cis-regulatory region (Corbo et al., 1997). This construct expresses in the notochord a truncated Xbp1 protein that retains the leucine-zipper DNA-binding domain (DBD) but lacks the transactivation domain, and is expected to bind its target sequences without activating transcription, according to what has been reported for mouse Xbp1 (Lee et al., 2003). The resulting plasmid, Bra>Xbp1^DBD::GFP, was electroporated into Ciona zygotes in parallel with a plasmid able to induce the formation of Xbp1 shRNA and with the neutral notochord marker Bra>GFP plasmid (Corbo et al., 1997) as a control (Figure 2A–C). The resulting transgenic embryos were cultured under the same conditions until they reached the late-tailbud stage, when notochord development was assessed using laser-scanning confocal microscopy. Compared to embryos electroporated with Bra>GFP (Figure 2A), embryos electroporated with Bra>Xbp1^DBD::GFP had markedly shorter tails, irregularly shaped notochord cells, and displayed aberrant notochord intercalation (Figure 2B and D, Figure 2—figure supplement 1). In embryos expressing Xbp1 shRNA, we observed a phenotype milder than the one caused by Bra>Xbp1^DBD::GFP, in a lower percentage of embryos (Figure 2C and D).

To determine the effects of the Xbp1 gain-of-function on embryonic development, we cloned its full-length (FL) cDNA downstream of the Foxa.a promoter region (Di Gregorio et al., 2001). Transgenic embryos carrying the Foxa.a>Xbp1^FL plasmid ectopically express Xbp1 in CNS and endoderm and overexpress it in the notochord. This causes the tail to be bent upward and the notochord cells to be slightly smaller than normal and often arranged into two or more rows throughout the tail (Figure 2E and F). Lastly, we generated a presumed hyperactive form of Xbp1 by fusing its DBD to the VP16 transactivation domain (Sadowski et al., 1988) (abbreviated as Bra>Xbp1^DBD::VP16::GFP). Compared to stage-matched control embryos (Figure 2D, G and G’), embryos carrying the Bra>Xbp1^DBD::VP16::GFP transgene exhibited an evident change in the shape and localization of their notochord cells, which was dependent upon the percentage of transgene incorporation; we also observed an abnormally high number of transgenic cells in the tails of embryos expressing the Xbp1^DBD::VP16::GFP fusion protein (Figure 2H, I, H', and I'), which might be caused either by the repositioning of mesenchymal cells from the trunk to the tail or by a loss in the control of cell division in some of the presumptive notochord cells.

After analyzing the notochord phenotypes caused by the gain- and loss-of-function experiments described above, we sought to identify the notochord genes that were causing them and that presumably act downstream of Xbp1. To this aim, we collected transgenic embryos from the same clutch, electroporated in parallel with the same amount of DNA (see Materials and methods), expressing either the Xbp1^DBD::GFP or the Xbp1^DBD::VP16::GFP fusion proteins, alongside GFP-expressing embryos that were used as controls. RNAs were extracted from each of these three different populations of transgenic embryos, individually labeled, and hybridized to a C. robusta microarray. Genes that displayed statistically relevant up- or downregulation in the Xbp1^DBD::GFP- and in the Xbp1^DBD::VP16::GFP-expressing embryos compared to the GFP-expressing embryos were selected for further analysis. After these microarray screens were carried out in triplicate, 109 individual putative target genes of Cr-Xbp1 were identified. Expression patterns for 39 of these 109 genes had been previously published (Brozovic et al., 2018; Brozovic et al., 2016; Calfon et al., 2002; Christiaen et al., 2008; Fujiwara et al., 2002; Harder et al., 2019; Hudson et al., 2011; Hudson and Yasuo, 2005; Miwata et al., 2006; Noiret et al., 2012; Ogasawara et al., 2006; Parsons et al., 2002; Razy-Krajka et al., 2018; Shimozono et al., 2010; Tetsukawa et al., 2010), and the expression patterns of 61 of the remaining genes are first described in this study; 9 genes were either not analyzed or provided unclear results (Supplementary files 1 and 2). Five genes, Ci-fibrinogen-like, DnaJc7, Vps35l, Akr1b10, and KH.C8.749 (Supplementary file 1), were identified in both Xbp1^DBD::GFP- and Xbp1^DBD::VP16::GFP-expressing embryos; the limited overlap between the two datasets is likely due to the effect of the VP16 domain on target sequence selection and was observed in previous microarray screens as well (Butz et al., 2004; José-Edwards et al., 2013). We also noticed that, similarly to Xbp1^DBD, Xbp1^DBD::VP16 repressed the expression of several genes, likely through the activation of one or more repressor genes/pathways.

In total, 71 of the 100 expression patterns (71%) include notochord cells, their precursors, and/or broader territories encompassing the notochord (Supplementary file 1, Figure 3, Figure 3—figure supplement 1). The remaining 29 genes are predominantly expressed in mesenchyme (six genes, 6%), epidermis (three genes, 3%), sensory vesicle (five genes, 5%), or any combination of these patterns, often including trunk endoderm, while a few patterns could not be assigned to any specific tissue because their in situ hybridizations produced weak unlocalized staining (Supplementary file 2, Figure 3—figure supplement 2). The expression of Cr-Xbp1-downstream genes in trunk endoderm and mesenchyme reflects the late expression of Cr-Xbp1 reported in these tissues at the larva stage (Kusakabe et al., 2002). According to previous functional studies, morpholino-mediated inactivation of one of the genes downstream of Cr-Xbp1, KH.C12.323, which encodes for an aquaporin channel, causes the disorganization of the body plan (KH.C12.323, cicl027n09; Hamada et al., 2007). The Cr-Xbp1 target genes expressed in epidermis include Pitx, a well-characterized homeobox gene mainly expressed in the anterior sensory vesicle, epidermis, and oral siphon primordium (Christiaen et al., 2002; Christiaen et al., 2005), and Nodal, the activator responsible for Pitx asymmetric expression in the epidermis of the left side of the embryo (Yoshida and Saiga, 2008). These findings are consistent with the expression of Cr-Xbp1 in the epidermis of both trunk and tail (Figure 1A and A’) and in the anterior sensory vesicle (Figure 1B and C). In addition to being expressed during embryogenesis, Cr-Xbp1 is also detected after metamorphosis, in the newly formed endostyle, a structure homologous to the vertebrate thyroid gland (e.g., Sasaki et al., 2003), in circulating hemocytes, and in part of the digestive tract (Ogasawara et al., 2002); accordingly, some of its target genes are expressed in these post-metamorphic structures as well (Tables S2 and S2). In particular, the serpin gene that we have detected in the nerve cord (Figure 3—figure supplement 3) is expressed after metamorphosis in the endostyle and hemocyte-containing pharyngeal gills, in a pattern that matches the post-metamorphic expression of Cr-Xbp1 (Ogasawara et al., 2002).

The 71 presumptive notochord target genes of Cr-Xbp1 include 18 ‘Ciona-/ascidian-specific’ genes, which appear to be either specific to Ciona (4/18, 22.2%) or to be present in Ciona and other ascidians, but currently lack identifiable counterparts in the genomes of organisms from other divisions; nevertheless, five of these ascidian-specific genes (5/18, ~27.8%) contain recognizable protein domains (Supplementary file 1, Figure 3—figure supplement 2).

We have grouped the 53 Ciona notochord genes with vertebrate counterparts into different categories on the basis of their gene ontologies and the functions of their closest orthologs in other organisms. One of these categories includes seven proteins able to bind nucleic acids (7/53, 13.2%; Figure 3A–C, Supplementary file 1); among them are the notochord transcription factors Ci-Bra and Lmx-like (Figure 3A and B; Corbo et al., 1997; José-Edwards et al., 2011), the ubiquitously expressed NRF1/2/6 (Ishibashi et al., 2003), and an RNA-binding protein of the PTBP family (Figure 3C). Six Cr-Xbp1 targets encode transmembrane transporters and channels (6/53, 11.3%) and include five members of the solute carrier family (SLC; Figure 3D–F) and KCNMB3, a potassium channel (Satou et al., 2001; Supplementary file 1). Most of the transmembrane transporters downstream of Cr-Xbp1 belong to the SLC superfamily, which includes Na⁺-dependent transporters of anionic molecules (Pizzagalli et al., 2021). One of these transporters, KH.C11.596, is equally related to human di- and tricarboxylate transporters SLC13A2, SLC13A3, and SLC13A5, another, KH.C12.115, is closer to SLC23A1, SLC23A2, and SLC23A3, all of which transport L-ascorbic acid; KH.C8.466 is equally related to all three members of the small SLC43 subfamily, which includes specialized transporters of neutral amino acids (Pizzagalli et al., 2021). Interestingly, the Cr-Xbp1-downstream effectors also include Ciona Slc26, an extensively characterized anion transporter necessary for the formation of the central lumen of the notochord during tubulogenesis, the last step of notochord morphogenesis in Ciona and other ascidians (Dong et al., 2009; Deng et al., 2013).

A significant fraction of Cr-Xbp1 targets encode for ECM proteins (8/53, 15%; Figure 3G–I) and include a hemicentin previously reported as notochord-specific, which we detected also in trunk endoderm, in addition to the notochord (Figure 3H, Supplementary file 1), and an extracellular protein equally related to collagen and selectin (Figure 3I). Eleven genes encode for signaling molecules (11/53, 20.7%; Figure 3J–M), among which, in particular, transforming growth factor beta (TGF-β) and various components of its signaling pathway, such as Rb1cc1, a mediator of autophagy (Yao et al., 2021), olfactomedin2, an ER-localized downstream target of TGF-β (Shi et al., 2014), and bone morphogenetic protein BMP4/Lefty1/2, another member of the TGF-β superfamily signaling ligands (Supplementary file 1 and references therein). Fibrillin, an ECM coordinator of elastic fibers assembly, is responsible for sequestering TGF-β in the ECM in a latent state, thus regulating its bioavailability (Dallas et al., 2005; Godwin et al., 2019; Robertson and Rifkin, 2016). The signaling molecules controlled by Cr-Xbp1 also include Gnai1/2/3 (G-protein subunit alpha I), a presumed mediator of cell migration expressed in notochord and bipolar tail neurons (Kim et al., 2020; Figure 3J). Other Gene Ontology (GO) categories include genes presumably involved in various steps of the UPR, such as proteolysis (Figure 3N and O), autophagy (Figure 3P), cell-cycle regulation (Figure 3Q), ubiquitin-associated processes (Figure 3R and S), cell communication (Figure 3T and U), and trafficking (Figure 3V–X). The breakdown of the full complement of 53 Cr-Xbp1 notochord targets into different gene ontologies is provided in Figure 3Y and Supplementary file 1.

Among the previously published notochord genes targeted by Cr-Xbp1 is Noto4/PID1, which was first identified as a downstream target of Ci-Bra (Takahashi et al., 1999; Hotta et al., 2000), and later on demonstrated to be required for notochord intercalation (Yamada et al., 2011). This gene is a direct target of Ci-Bra (Katikala et al., 2013) as well as a target of Tbx2/3, which, like Ci-Bra, is a member of the T-box family of transcription factors (José-Edwards et al., 2013). The Noto4/PID1 notochord cis-regulatory module (CRM) relies upon a single T-box binding site that is likely targeted by Ci-Bra and/or Tbx2/3 (Katikala et al., 2013). Another well-characterized gene that was originally identified as a Ci-Bra target is fibrinogen-like (Takahashi et al., 1999; Hotta et al., 2000), which encodes for a secreted peptide required for the proper positioning of neurons along the developing nerve cord and for axon guidance (Yamada et al., 2009). Noto8, another direct Ci-Bra target (Katikala et al., 2013), encodes for a calmodulin-like protein whose closest counterparts in other organisms act as modulators of motility and ion channel function (Bennett et al., 2007; Inanobe et al., 2015).

In conclusion, the notochord genes controlled by Cr-Xbp1 can be tentatively grouped into three broad categories: UPR, TGF-β signaling pathway, and morphogenetic effectors sensu stricto; the latter category includes genes directly involved in notochord intercalation and tubulogenesis (Figure 3Z). Expression of Cr-Xbp1 is activated in the notochord by Ci-Bra; in turn, Cr-Xbp1 regulates, either directly or indirectly, the expression of its numerous notochord targets and generates a positive feedback loop on the expression of Ci-Bra itself. In addition to sharing some of its target genes with Ci-Bra, Cr-Xbp1 shares part of these notochord genes with Tbx2/3 (José-Edwards et al., 2013), and ChIP-chip experiments indicate that the genomic loci of some of the Cr-Xbp1 notochord targets are occupied by Foxa.a in early embryos (Kubo et al., 2010; Supplementary file 1). No overlap was found between the notochord genes downstream of Cr-Xbp1 and those controlled by another node of the Ciona notochord GRN, the ascidian-specific transcription factor Bhlh-tun1 (Kugler et al., 2019).

In control Ciona embryos carrying the developmentally neutral Bra>GFP transgene, the notochord develops normally and Ci-Bra mRNA is detected in all of its 40 cells (Figure 4A–C). At the mid-tailbud stage, the notochord cells are columnar in shape and exhibit a ‘stack of coins’ arrangement (Figure 4C, Figure 4—video 1). In embryos carrying the Bra>Xbp1^DBD::GFP construct (Figure 4D–F), the notochord is mostly composed of irregularly shaped cells, and only the cells that have not received this transgene express Ci-Bra (Figure 4E). The lack of overlap between the cells carrying the Bra>Xbp1^DBD::GFP construct (Figure 4D) and the cells expressing Ci-Bra (Figure 4E) indicates that the repressor form of Cr-Xbp1 is able to repress Ci-Bra expression (Figure 4F, Figure 4—video 2). This phenotype is reminiscent of the disruption of notochord development observed in Ci-Bra mutants (Chiba et al., 2009; Figure 1C).

To verify the effect of alterations in the function of Cr-Xbp1 on components of the TGF-β signaling pathway, we analyzed the expression of fibrillin in embryos carrying the Bra>Xbp1^DBD::VP16::GFP transgene (Figure 4G and H). As predicted by the results of the microarray screens (Supplementary file 1), fibrillin is uniformly expressed in the notochord in control embryos (Figure 4G), while in Bra>Xbp1^DBD::VP16::GFP transgenic embryos expression of this gene is limited to the notochord cells that did not incorporate the transgene (Figure 4H). This repressive activity of the Xbp1^DBD::VP16 fusion is likely indirect and could be due to the activation of a repressor of fibrillin expression; alternatively, the Xbp1^DBD::VP16 transgene might be solely occupying the Xbp1 binding sites in the regulatory regions of this and other Cr-Xbp1-downstream genes without being able to activate their expression, thus preventing the endogenous Cr-Xbp1 and/or other activators from binding.

To extend the results obtained in Ciona to higher chordates, we tested whether the positive feedback of Xbp1 on Brachyury was conserved in embryos of the amphibian Xenopus laevis, a vertebrate in which notochord expression of both these genes had been previously demonstrated (Smith et al., 1991; Zhao et al., 2003). In Xenopus, Xbra/Tbxt is expressed throughout the mesoderm during gastrulation and in the prospective notochord at neurula stages (Smith et al., 1991). Xbp1 is first detected in the dorsal lip of the blastopore at the early gastrula stage and persists in the involuting dorsal mesoderm as gastrulation proceeds (Zhao et al., 2003). To evaluate the regulation of Xbp1 by Xbra, we used an antisense morpholino oligonucleotide (MO; Shi et al., 2011) to specifically knock down expression of Xbra in the dorsal mesoderm. We found that nearly all Xbra morphant embryos showed reduced Xbp1 expression at the gastrula stage compared to their stage-matched controls (Figure 5A, B, B’ and G). We also performed the complementary experiment by targeting an Xbp1-specific MO (Yuan et al., 2008; Tanegashima et al., 2009) to the dorsal mesoderm and analyzing the consequences on Xbra expression. Our results show that Xbp1 is implicated in the regulation of Xbra expression in the dorsal mesoderm, as approximately 65% of the Xbp1 morphant embryos displayed reduced Xbra expression at gastrula stages (Figure 5C, D, D’ and H). To confirm the specificity of the knockdown phenotypes, we used a second set of MOs (XbraMOS and Xbp1MOS) that specifically interfere with Xbra and Xbp1 pre-mRNA splicing, resulting in the production of shorter transcripts due to exon 6 and exon 3 exclusion, respectively (Figure 5—figure supplement 1). The phenotype of XbraMOS- (Figure 5E, E’ and G) and Xbp1MOS-injected embryos (Figure 5F, F’ and H) was identical to the phenotype generated by the injection of their respective translation-blocking MO. Later in development, unlike stage-matched controls (Figure 6), Xbra and Xbp1 morphant embryos displayed axis elongation defects (Figure 6B and C) and posterior truncations (Figure 6B’ and C’). The phenotypes were classified into either ‘mild’ or ‘severe,’ and their respective proportions are reported in Figure 6D. Overall, these malformations are similar to those observed upon expression of a dominant-interfering Xbra::Engrailed fusion in previous studies (Xbra-En^R; Conlon et al., 1996). To investigate the evolutionary conservation of the Xbp1-downstream genes identified in Ciona, we searched the Xenbase database (Xenbase.org; Bowes et al., 2010) and available literature for the expression patterns of Xenopus orthologs of Ciona notochord genes. We could not find any information on the expression of more than half (27/52; ~52%) of the Xenopus putative orthologs of Cr-Xbp1-downstream notochord genes; however, 6 of the remaining genes are reportedly expressed in the notochord, while for the remaining 19 genes expression in this structure has not been reported (Supplementary file 3).

Lastly, as a read-out of notochord differentiation, we assessed the expression of the well-characterized notochord marker Sonic hedgehog (Shh). WMISH of Xenopus embryos at NF stage 35/36 with a Shh probe indicates that, compared to sibling controls, Shh expression is discontinuous in the notochord of Xbra and Xbp1 morphant embryos, and often confined to its posterior regions (Figure 6—figure supplement 1). These results suggest that notochord differentiation is impaired by alterations in the Xbra-Xbp1 subcircuit. Altogether, these results indicate that Xbp1 and Xbra regulate each other’s expression as part of a regulatory loop controlling not only formation of dorsal mesoderm, but also notochord development and differentiation.

In Ciona embryos lacking the function of Ci-Bra (Chiba et al., 2009), expression of Cr-Xbp1 in notochord cells is lost, while its expression in the epidermis and in the anterior region of the sensory vesicle remains unperturbed. These results indicate that Ci-Bra is required for notochord expression of Cr-Xbp1. Embryos carrying transgenes that express either a passive repressor form of Cr-Xbp1 (Xbp1^DBD) (Lee et al., 2003) or shRNA able to interfere with the function of this transcription factor are characterized by a block in the intercalation and differentiation of the notochord cells. These abnormalities resemble those observed in embryos homozygous for a mutation in the Ci-Bra coding region (Chiba et al., 2009). On the other hand, transgenic embryos in which all notochord cells have incorporated a presumed hyperactive form of Cr-Xbp1 (Xbp1^DBD::VP16) show numerous round, poorly differentiated transgenic cells dispersed throughout their tails.

To identify the Cr-Xbp1-downstream genes responsible for these phenotypes and to shed light on the molecular mechanisms employed by this transcription factor to regulate notochord development, we performed a transcriptomic analysis of embryos expressing either Xbp1^DBD::VP16 or Xbp1^DBD,using embryos expressing GFP as controls. This approach identified 109 genes whose expression is influenced either directly or indirectly by Cr-Xbp1, 71 of which were found expressed in the notochord, either through this study or through the analysis of previously published in situ hybridization data. Eighteen of these genes currently lack clear orthologs in organisms other than Ciona and other ascidians. The presence of recognizable protein domains and sequence homology with proteins identified in different organisms allowed the tentative classification of the remaining 53 genes into 15 GO categories. The most ample of these categories include genes whose predicted products either participate in signaling pathways or in the formation of the abundant ECM secreted by the developing notochord cells. 36 of the 71 Cr-Xbp1-downstream notochord genes (50.7%) had been previously reported as targets of Ci-Bra by microarray screens (Takahashi et al., 1999; Hotta et al., 1999; Hotta et al., 2000), genomic occupancy studies (Kubo et al., 2010), and/or RNA-Seq and related transcriptomic experiments (Reeves et al., 2017). The extensive overlap between the genes controlled by Cr-Xbp1 and those controlled by Ci-Bra suggests that Cr-Xbp1 might act as a transcriptional intermediary of Ci-Bra during the late stages of notochord development.

At least 50% of the Cr-Xbp1-downstream genes expressed in the Ciona notochord have been reported to participate in the UPR in other model organisms. In addition to participating in the response to ER stress, genes that mediate UPR are also activated under physiological conditions in cells specialized in secretion, such as hepatocytes (Reimold et al., 2000), plasma cells (Iwakoshi et al., 2003), pancreatic acinar cells (Lee et al., 2005), cells of the salivary glands (Lee et al., 2005), cells of the hatching gland in fish (Bennett et al., 2007), and the notochord cells of zebrafish and Xenopus (Bennett et al., 2007; Tanegashima et al., 2009). We have determined that Cr-Xbp1 controls evolutionarily conserved UPR effectors, such as proteases, chaperones, and other mediators of intracellular trafficking and protein recycling, including Vps35l and Clvs1 (Singla et al., 2019; Katoh et al., 2009), and the serine/threonine-protein kinase Dapk1, which acts as a UPR sensor and as a mediator of apoptosis and autophagy (Singh et al., 2016). Autophagy allows cells to selectively degrade misfolded proteins and defective organelles in response to stress signals (Kania et al., 2015); remarkably, this process is also activated in response to hyperosmotic stress in the notochord cells that compose the nuclei pulposi of the intervertebral discs (Jiang et al., 2015). The present study has identified both inducers and effectors of autophagy expressed in the Ciona notochord and regulated by Cr-Xbp1. Ubiquitin ligases, three of which are included among the Cr-Xbp1 targets, are among the main effectors of autophagy and act by tagging unfolded proteins and deteriorating organelles with ubiquitin chains, thus triggering their removal (Grumati and Dikic, 2018); accordingly, histone HIST1C/H1.2, another regulator of autophagy (Wang et al., 2017), is among the notochord genes controlled by Cr-Xbp1. In addition to autophagy, the UPR can also induce compensatory changes in mitochondrial function (Senft and Ronai, 2015), and the presence of mitochondrial metabolic enzymes among the Cr-Xbp1 target genes suggests that this physiological strategy might be present in Ciona as well.

Our investigation of the notochord targets of Cr-Xbp1 also connects this transcription factor to the TGF-β signaling pathway. Treatment with TGF-β has been reported to activate expression of Xbp1, and consequently UPR, in mouse and human fibroblasts (Baek et al., 2012). Our study indicates that in addition to modulating transcription of TGF-β and BMP4, Cr-Xbp1 regulates the expression of olfactomedin 2, which in smooth muscle acts downstream of TGF-β to activate a number of tissue-specific markers (Shi et al., 2014). We have previously characterized a notochord enhancer region associated with olfactomedin 2 and found that it relies on Fox, homeodomain, and AP1 binding sites for its activity (José-Edwards et al., 2015); this indicates that the regulation of this gene by Cr-Xbp1 is likely indirect.

A relevant fraction of the structural genes identified in this study are ECM components. The ECM components include, among others, hemicentins, which are known regulators of cell adhesion (Xu et al., 2013), and fibronectin-1, which coordinates ECM assembly and convergent extension (Halper and Kjaer, 2014; Davidson et al., 2006) and modulates TGF-β signaling together with fibrillin (Zilberberg et al., 2012). Of note, the product of another effector of Cr-Xbp1, fibrinogen-like (Hotta et al., 2000), is secreted in the ECM by the notochord cells and controls the positioning of neurons along the nerve cord through the Notch signaling pathway (Yamada et al., 2009). Together with these findings, the notochord phenotypes induced by the transgenes employed in this study suggest that Cr-Xbp1 is involved in ECM secretion, a crucial step of notochord morphogenesis. The notochord intercalation defect caused by the repressor form of Cr-Xbp1, Xbp1^DBD, is comparable to the phenotype observed in Ci-Bra mutants (Chiba et al., 2009) and is consistent with the downregulation of Ci-Bra that is observed in embryos carrying this transgene. The expression of the Xbp1^DBD::VP16 fusion causes scattering of notochord cells throughout the tail. These considerable defects in notochord formation can be explained by the altered expression of Noto4/PID1, which is required for notochord intercalation (Yamada et al., 2011), and of several ECM components. In addition to the defective synthesis and secretion of ECM components, which allows the notochord cells to disperse away from the midline, another candidate regulator of this process that is affected by the hyperactive form of Cr-Xbp1 is Gnai1/2/3; its corresponding protein has been recently shown to be preferentially localized to the plasma membrane and to the Golgi apparatus, where it might control trafficking of secretory vesicles (Kim et al., 2020 and references therein). Nek1/3/5, a member of a family of protein kinases involved in the regulation of the cell cycle and mitotic progression (e.g., Fry et al., 2012), is upregulated in Xbp1^DBD::VP16 transgenic embryos, which might explain the abnormal number of cells that is occasionally observed in these embryos.

Additionally, Cr-Xbp1 participates in the last steps of notochord morphogenesis, lumen formation and tubulogenesis, by regulating the expression of Slc26, the transmembrane transporter necessary for the expansion of the central lumen of the Ciona notochord (Deng et al., 2013) and of four additional genes encoding related solute carriers.

Among lower vertebrates, Xbp1 is expressed in the zebrafish notochord (Liang et al., 2001) and studies in medaka fish have shown that the vacuolization of the notochord requires the activity of effectors and transducers of the UPR (Ishikawa et al., 2017).

In Xenopus, morpholino-mediated knockdown of Xbp1 causes the formation of a smaller than normal notochord (Tanegashima et al., 2009). Xenopus Xbp1 regulates expression of the chaperone proteins Hsp5A/Bip, DNAJ9B, and HSP90B1 (Tanegashima et al., 2009), forms a regulatory loop with BMP-4 in the control of mesoderm and neural differentiation (Zhao et al., 2003; Cao et al., 2006), and is required for pancreas development (Yang et al., 2020). The present study has determined that the regulation of BMP-4 by Xbp1 is present in the Ciona notochord as well and has identified, among others, the chaperone protein DnaJc7 as a notochord target of Cr-Xbp1. Most importantly, the results of this analysis have uncovered a regulatory connection between Brachyury and Xbp1 that is maintained in Xenopus and is required for the proper development of the notochord in this vertebrate. The information gathered using Ciona on the genes influenced by Xbp1 will guide future studies on the notochord genes controlled by Xenopus Xbp1.

In mammals, the role of Xbp1 in notochord development remains to be explored. However, a ChIP-on-chip study carried out on chromatin purified from mouse plasma cells, pancreatic beta cells, and skeletal myotubes, both wild-type and subjected to ER stress, has uncovered 545 transcriptional targets of XBP1, most of which form a common core of XBP1-downstream UPR genes that are expressed by all cell types analyzed and are involved in the maintenance of ER homeostasis and control of secretion (Acosta-Alvear et al., 2007). At least half of the Cr-Xbp1-downstream genes with traceable vertebrate orthologs identified in this study can be predicted to be participating in the UPR as their gene ontologies correlate with those of XBP1 targets (Acosta-Alvear et al., 2007) and cover different facets of this complex process, such as protein folding, proteolysis, trafficking, autophagy, transmembrane transport, and ubiquitination. In mouse embryos, Xbp1 is expressed in osteoblasts and chondroblasts of several skeletal structures (Clauss et al., 1993), and a recent single-embryo, single-cell RNA-Seq study has detected the expression of Xbp1 in the node/notochord cell population (Mittnenzweig et al., 2021). The Xbp1^-/- mutation is embryonic lethal beginning at day E12.5 (Reimold et al., 2000), which leaves open the possibility that Xbp1 mutant mice might have defects in notochord formation as well. Together with the published expression of Xbp1 in the chick notochord (Bell et al., 2004; Darnell et al., 2007), all these findings and the present study suggest that Xbp1 and UPR genes have been incorporated into notochord formation early during chordate evolution and have been retained in vertebrates as components of the essential notochord developmental program.

Our research on the Ciona notochord GRN had previously elucidated the positive feed-forward regulatory input between Brachyury and Tbx2/3 (José-Edwards et al., 2013) and the synergistic control of notochord gene expression by the Bra/Foxa.a subcircuit (Passamaneck et al., 2009; José-Edwards et al., 2015; Kugler et al., 2019; Di Gregorio, 2020). Based on these results, on the outcome of morpholino-mediated knockdowns (Imai et al., 2006) and on additional evidence, the Ciona notochord GRN has been described as being mainly reliant on positive feed-forward interactions (Reeves et al., 2021); our results have provided the first report of a positive feedback loop within the Ciona notochord GRN and have uncovered a new regulatory subcircuit that links the UPR to notochord development. Furthermore, we have provided evidence that the cross-regulatory interaction between Brachyury and Xbp1 is conserved in the dorsal mesoderm and notochord of Xenopus. The regulatory relationship between Brachyury and Xbp1 identified through this research is far-reaching as both transcription factors play crucial roles in a variety of processes that extend far beyond notochord development and include immune response and tumorigenesis.

Adult C. robusta (formerly Ciona intestinalis type A; Pennati et al., 2015) were purchased from M-REP (Carlsbad, CA). C. robusta Brachyury mutant embryos (originally published as Ci-Bra mutants; Chiba et al., 2009) were kindly provided by Drs. Shota Chiba and William Smith (U.C. Santa Barbara, CA). Culturing and electroporations were performed as previously described (Oda-Ishii and Di Gregorio, 2007). A fraction of the embryos selected for imaging were counterstained with 1U rhodamine-phalloidin (Invitrogen, Carlsbad, CA) in 1X PBS/0.2% Triton X-100, for 3 hr at room temperature. All embryos selected for confocal imaging were mounted using VECTASHIELD with DAPI (Vector Laboratories, USA).

Ciona embryos were fixed at stages ranging from gastrula to late tailbud, hybridized, and stained essentially as previously described (José-Edwards et al., 2011; José-Edwards et al., 2013); whenever necessary, the experiments were repeated at different hybridization temperatures to increase the specificity of the hybridization signal. Fluorescent in situ hybridization and immunostaining were carried out as previously published (Wagner and Levine, 2012; José-Edwards et al., 2013). Anti-digoxigenin-POD (Roche, IN) and rabbit anti-GFP (Novus Biologicals, CO) antibodies were diluted 1:500 and 1:1000, respectively. In vitro synthesized antisense DIG-labeled RNA probes were visualized using the TSA (Tyramide Signal Amplification) Plus tetramethyl-rhodamine working solution (PerkinElmer, MA) for 5–20 min at room temperature, blocked for 1 hr in TNBS (100 mM Tris pH 7.5, 150 mM NaCl, 0.5% Roche blocking reagent, 2% normal goat serum), and incubated at 4°C overnight in the presence of goat anti-rabbit IgG Alexa Fluor 488 secondary antibody (Invitrogen), diluted 1:500. For most genes in this study, antisense RNA probes were synthesized using as templates ESTs from the Ciona Gene Collection release 1 (Satou et al., 2001) and/or the Ciona Unigene cDNA collection (Gilchrist et al., 2015; Tables S1 and S2).

The Bra>Xbp1^DBD::GFP construct was generated by digesting the pFBΔSP6 plasmid (Oda-Ishii and Di Gregorio, 2007) with XbaI and EcoRI to remove the Ci-Foxa.a basal promoter and the LacZ reporter gene, which were replaced by a linker sequence containing restriction enzyme sites for XbaI, SacI, KpnI, BsrGI, and EcoRI. This newly created multiple cloning site was digested with XbaI/SacI and ligated with the 3.5 kb C. robusta Brachyury enhancer/promoter region, lacking the Ci-Bra coding sequence (pBraLinker; Dunn and Di Gregorio, 2009). Subsequently, a 564bp region encoding for the N-terminal portion of Cr-Xbp1 (aa 1–188), which includes the predicted DBD, was amplified by RT-PCR from RNA extracted from early-tailbud embryos using the QIAGEN RNA miniprep kit (Valencia, CA) as previously described (Oda-Ishii and Di Gregorio, 2007) using the primers:

(restriction sites are indicated in lowercase), and was cloned downstream of the Ci-Bra enhancer/promoter as a SacI-KpnI fragment.

The Bra>Xbp1^DBD::VP16::GFP construct was generated by fusing the DBD of Xbp1 to the VP16 transactivation domain, as previously described (José-Edwards et al., 2013; Sadowski et al., 1988). The Cr-Xbp1 shRNA construct was prepared using primers matching nt 9–30 of the Cr-Xbp1 ORF according to the method reported in Nishiyama and Fujiwara, 2008. Primers were annealed, phosphorylated, and ligated into the EcoRI-EcoRV sites of pSP-U6RV (Nishiyama and Fujiwara, 2008).

The Foxa.a>Xbp1^FL construct was generated by cloning the complete Cr-Xbp1 ORF downstream of a 2.5 kb fragment of the Foxa.a promoter region (formerly Fkh/HNF-3beta; Di Gregorio et al., 2001) using the following primers:

(restriction sites are indicated in lowercase).

Approximately 100–300 C. robusta embryos were electroporated with 50 µg of either Bra>Xbp1^DBD::GFP, or Bra>Xbp1^DBD::VP16::GFP, or Bra>GFP plasmid. Fluorescent transgenic embryos from the same clutch were manually selected in comparable amounts from all three experimental samples using a Zeiss SteReo Discovery V12 epifluorescence microscope at ~6.25 hpf at 21°C, corresponding to initial tailbud I/II according to the standardized developmental table at 18°C (Hotta et al., 2007). This time point matches the approximate onset of endogenous Cr-Xbp1 expression (Kugler et al., 2008). Total RNAs were extracted from three biological replicates for each transgenic population using the RNeasy Micro Kit (QIAGEN), and, after being amplified and labeled using the Ambion MessageAmp Premier RNA Amplification Kit (Thermo Fisher Scientific, Waltham, MA), they were hybridized to the Ciona Affymetrix GeneChip CINT06a520380F by the Weill Cornell Genomics Resources Core Facility using standard Affymetrix protocols. RNAs extracted from embryos electroporated with the developmentally neutral plasmid Bra>GFP were regarded as experimental controls for development after the electroporation procedure and used to determine the extent and significance of up- or downregulation of genes in the Xbp1^DBD and XBP1^DBD::VP16 samples.

Results were RMA summarized from raw data and quantile normalized using GeneSpring GX 11 software by the Weill Cornell Epigenomics Core Facility. Only probe sets with p-values ≤ 0.05 with an absolute fold-change (FC) cut-off of 2.0 were further considered. Changed mRNA levels are expressed as ‘up’ or ‘down’ regulated in Tables S1 and S2. The complete dataset has been deposited into the NCBI Gene Expression Omnibus (GEO) under accession number GSE46751.

Xenopus Xbp1.L was amplified by PCR from stage 12 cDNA using the primers:

based on the published sequence (Zhao et al., 2003). A 1190 bp product was recovered, cloned into the pGEMT vector (Promega, Madison, WI), sequenced, and used to generate in situ hybridization probes. This construct is referred to as pGEMT-Xbp1.

The procedures were performed in accordance with the recommendations of the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health, approved by New York University Institutional Animal Care and Use Committee, under animal protocol #150201. X. laevis embryos were staged according to Nieuwkoop and Faber, 1967 (NF) and raised in 0.1× Normal Amphibian Medium (NAM; Slack and Forman, 1980). Antisense MOs were purchased from GeneTools (Philomath, OR). We used translation-blocking MOs targeting Xbra/Tbxt and Xbp1. XbraMO (5′-GCGCAGCTCTCGGTCGCACTCATTC-3′) targets both the short (Xbra.S) and the long (Xbra.L) forms of Xbra. For Xbp1 (Xbp1MO), we used a mix (1:1) of two MOs targeting the Xbp1-S (5′-GACATCTGGGCCTGCTCCTGCTGCA-3′) and Xbp1-L (5′-GCCCAACAAGAGATCAGACTCAGAG-3′). All three translation-blocking MOs have been previously validated (Shi et al., 2011; Yuan et al., 2008; Tanegashima et al., 2009). To further confirm the Xbra and Xbp1 morphant phenotypes, we used a second set of MOs interfering with Xbra and Xbp1 pre-mRNA splicing. Xbp1MOS (5′-TCTGGAAGAGATCAAACACATGACA-3′) targeting the intron 2/exon 3 junction of both forms of Xbp1, and a mix (1:1) of two MOs:

targeting the intron 5/exon 6 junction of the short and long forms of Xbra, respectively (XbraMOS) (Figure 5—figure supplement 1). In each experiment, approximately 30 ng of MOs were co-injected with a lineage tracer (LacZ mRNA; 0.5 ng) in the equatorial region of both dorsal blastomeres at the four-cell stage (NF stage 3), and the embryos were analyzed by in situ hybridization at NF stage 12 or stage 35/36. Each injection was performed on at least three independent batches of embryos.

Xenopus embryos at the appropriate developmental stages were fixed in MEMFA (0.1 M 3-N-morpholino-propanesulfonic acid pH 7.4, 2 mM EGTA, 1 mM MgSO₄, and 3.7% formaldehyde), stained for Red-Gal (Research Organics; Cleveland, OH) to visualize the lineage tracer (LacZ mRNA), and processed for in situ hybridization. Antisense digoxygenin-labeled probes (Genius kit; Roche, IN) were synthesized using as templates cDNAs encoding Xbra/Tbxt (Smith et al., 1991), Xbp1 (pGEMT-Xbp1), and Shh (Stolow and Shi, 1995). WMISH was performed as described (Harland, 1991; Saint-Jeannet, 2017).

For MO validation by RT-PCR, total RNAs from control and injected embryos were extracted with RNeasy Micro Kit (QIAGEN) and reverse-transcribed using SuperScript IV VILO Master Mix (Thermo Fisher Scientific) according to the manufacturer’s instructions, and used for PCR with Illustra PuReTaq Ready-To-Go PCR beads (GE Healthcare, Chicago, IL). The following primer sets spanning the entire coding sequence were used: