Two distinct motifs for Zic-r.a drive specific gene expression in two cell lineages

In ascidians, Zic-r.a (also called Macho-1), which encodes a Zic transcription factor, is expressed maternally and is essential for specification of muscle cells (Nishida and Sawada, 2001). Zic-r.a mRNA is localized in the most posterior cells, which have the potential to adopt a germ cell fate and are transcriptionally quiescent (magenta cells in Fig. 1A) (Kumano et al., 2011; Shirae-Kurabayashi et al., 2011). Therefore, Zic-r.a does not act in these most posterior cells, but instead acts in their sister cells (B5.1 and B6.4 in Fig. 1A), which contribute to muscle cells and are transcriptionally active. In these cells, Zic-r.a activates Tbx6-r.b, which encodes a key transcription factor for muscle cell specification (Yagi et al., 2005; Yu et al., 2019) at the 16-cell and 32-cell stages (Oda-Ishii et al., 2016; Yagi et al., 2004a).

At the 16-cell stage, Tbx6-r.b is expressed in B5.1 (brown cells in Fig. 1A) through the cooperative action of Zic-r.a, β-catenin and Tcf7 (Oda-Ishii et al., 2016). At the 32-cell stage, Tbx6-r.b is hardly detectable with in situ hybridization in daughter cells of B5.1, and it is again activated in B6.4 (green cells in Fig. 1A). Although Zic-r.a is similarly required for Tbx6-r.b expression in B6.4 (Yagi et al., 2004a), it has been suggested that β-catenin does not activate its target in these cells (Hudson et al., 2013). Thus, collectively, these studies suggest that Tbx6-r.b is regulated differently in these two muscle lineages. In the present study, by addressing the issue of this differential regulation, we found that Tbx6-r.b is activated in B6.4 of the 32-cell embryo through Zic-r.a binding to sites with non-canonical motifs, the nucleotide sequences of which are dissimilar to the canonical motif we previously determined using an in vitro selection assay (Yagi et al., 2004a).

Many studies have shown that low-affinity binding sites for transcription factors are important for driving temporally or spatially controlled gene expression. These include a study of a temporal control by Pha-4 in nematodes (Gaudet and Mango, 2002), spatial control of Otx in ascidians (Farley et al., 2015), and temporal and spatial control of Svb in flies (Crocker et al., 2015; Fuqua et al., 2020). In these studies, each binding site that transcription factors recognize is represented by a single motif, and low-affinity sites are regarded as its divergent versions. On the other hand, it has been postulated that some transcription factors recognize multiple motifs (Badis et al., 2009; Franco-Zorrilla et al., 2014), although the biological importance of this property has not been well clarified. For example, it was reported that mammalian Zic3 transcription factor recognizes two motifs, one of which resembles the canonical motif of ascidian Zic-r.a (Badis et al., 2009). In the present study, to clarify the biological importance of this property, we show that Zic-r.a has two recognition motifs that are not divergent versions of a single motif, and that these two distinct binding motifs are used for driving different sets of genes in early and late embryos.

Tbx6-r.b is activated by Zic-r.a in concert with β-catenin and Tcf7 in the B5.1 lineage at the 16-cell stage (Oda-Ishii et al., 2016). Therefore, we first examined whether Tbx6-r.b was similarly regulated in the B6.4 lineage (B6.4) (Fig. 1B-D). Previously, we showed that Tbx6-r.b expression is lost in the B6.4 lineage of 32-cell embryos injected with a morpholino antisense oligonucleotide (MO) against Zic-r.a (Zic-r.a morphants) (Yagi et al., 2004a) (Fig. 1C). On the other hand, in the present study, we discovered that Tbx6-r.b expression was not lost in B6.4 of β-catenin morphants at the 32-cell stage (Fig. 1D). Therefore, in contrast to Tbx6-r.b expression in the B5.1 lineage of 16-cell embryos, Tbx6-r.b expression in the B6.4 lineage of 32-cell embryos is regulated independently of β-catenin.

This result was further confirmed using reporter constructs. A construct containing the 189 bp upstream sequence of Tbx6-r.b (Fig. S1A) was expressed in the B5.1 lineage at the 16-cell stage, and a construct with mutations in two binding sites for Tcf7, which forms a complex with β-catenin, was not expressed, as reported previously (Oda-Ishii et al., 2016) (Fig. 1E,F). However, in the B6.4 lineage of 32-cell embryos, both of these constructs were expressed (Fig. 1G,H). Thus, Tbx6-r.b expression in the B6.4 lineage is regulated independently of β-catenin/Tcf7 at the 32-cell stage.

Previously, we suggested that direct binding of Zic-r.a to DNA might be unnecessary for Tbx6 expression in the B5.1 lineage at the 16-cell stage, because Zic-r.a can bind to DNA indirectly by interacting with Tcf7 (Oda-Ishii et al., 2016). However, our reporter analysis in the present study showed that Tcf7 binding sites are not necessary for Tbx6-r.b expression in the B6.4 lineage at the 32-cell stage. Therefore, we considered the possibility that Zic-r.a directly binds to the upstream region of Tbx6-r.b in the B6.4 lineage at the 32-cell stage.

Although Zic-r.a preferentially binds to sequences similar to 5′-GCAGCGGGGGG-3′ (Fig. 2A) and potential binding sites are identified between −1095 and −1547 nucleotide positions from the transcription start site of Tbx6-r.b (Kugler et al., 2010; Yagi et al., 2004a), no similar sequences are found in the 189 bp upstream sequence of Tbx6-r.b, using the computer program Patser (Hertz and Stormo, 1999) and the Ciona position weight matrix (Yagi et al., 2004a). Meanwhile, a study of mouse Zic proteins has shown that Zic proteins bind to two different sequences, with consensus sequences of ‘CCCCnnGGGG’ and ‘CnCAGCAGG’ (Fig. 2A) (Badis et al., 2009). Using the same computer program, we found no significant hits in the Ciona Tbx6-r.b upstream region with the matrix for the mouse primary canonical motif, which is more similar to the Ciona motif. However, it identified one significant hit with the matrix for the mouse secondary, non-canonical motif (Fig. 2B). Additionally, another sequence similar to the non-canonical motif was identified in a nearby region, although this was not significant (Fig. 2B). Hereafter, we call the distal and proximal sites Zic-d and Zic-p, respectively.

We examined whether these two sites could bind Zic-r.a in vitro using a gel-shift assay. We observed shifted bands, indicating binding of Zic-r.a to these two sites (Fig. 2C). Next, to compare Zic-r.a binding activity between canonical and non-canonical motifs, we added them to the reaction as competitors. The shifted band was weakened by incubation with a 20 or 50 molar excess of the non-canonical motif competitor, and almost completely disappeared when the non-canonical motif competitor was increased to a 100 or 200 molar excess (Fig. 2D). On the other hand, the shifted band disappeared in the presence of 5 molar excess of the canonical motif competitor, which we have used in our previous study (Yagi et al., 2004a) (Fig. 2D). These results show that Zic-r.a does bind to the non-canonical motif, but that its affinity for this motif is much weaker than for the canonical motif.

Finally, we introduced mutations into the Zic-r.a binding sites, Zic-d and Zic-p, in the reporter construct containing the 189 bp upstream sequence of Tbx6-r.b. These mutations reduced reporter expression in the B6.4 lineage of 32-cell embryos (Fig. 2E), although 35% of embryos still expressed the reporter; therefore, the reporter may contain additional unrecognized Zic-r.a-binding sites. Our observation indicates that Zic-r.a binds to the Zic-d- and Zic-p-binding sites to activate Tbx6-r.b in the B6.4 lineage at the 32-cell stage. In addition, these mutations also reduced reporter expression in B5.1 of 16-cell embryos (Fig. S2), so these sites contribute to expression in B5.1 (see Discussion).

We have previously demonstrated that Zic-r.a protein is detected in nuclei of B5.1 and B5.2 and in the posterior pole of B5.2 at the 16-cell stage, using immunostaining with an anti–Zic-r.a antibody (Oda-Ishii et al., 2016). Immunostaining of 32-cell embryos with the same antibody showed that the Zic-r.a protein was barely detectable in the descendants of B5.1 (B6.1 and B6.2), and that it was detected in nuclei of B6.4 and B6.3, and in the posterior pole (Fig. 3A). This observation was confirmed by quantification of fluorescent signals in nuclei (Fig. 3B). Specifically, Zic-r.a is expressed strongly in cells in which Zic-r.a mRNA is localized (B5.2 in 16-cell embryos and B6.3 in 32-cell embryos) and in their sister cells (B5.1 in 16-cell embryos and B6.4 in 32-cell embryos). Although signal strengths in B6.4 were almost the same as in B6.3, where Zic-r.a mRNA is localized, they were significantly stronger than those in B5.1 of 16-cell embryos. These observations support the hypothesis that Tbx6-r.b is activated through direct binding of Zic-r.a to the Zic-p and Zic-d sites in the B6.4 lineage of 32-cell embryos, because a higher concentration of Zic-r.a is apparently necessary for sufficient binding to low-affinity sites. On the other hand, the absence of Tbx6-r.b mRNA in B6.3 is consistent with the earlier finding that the B6.3 cell pair is transcriptionally silent (Kumano et al., 2011; Shirae-Kurabayashi et al., 2011).

Overexpression of Zic-r.a evoked Tbx6-r.b ectopic expression at the 16-cell stage, even in the animal hemisphere, where no nuclear β-catenin is expected (Fig. 3C,C′). Our result was consistent with the hypothesis that a high concentration of Zic-r.a can evoke Tbx6-r.b expression independently of β-catenin, although we cannot completely rule out the possibility that Zic-r.a overexpression induced nuclear translocation of β-catenin ectopically in the animal hemisphere.

To further test this hypothesis, we next used modified Tbx6-r.b reporter constructs, because it is expected that more binding sites recruit more Zic-r.a (Fig. S1BC). In the first construct (3xZic-p+ΔTCF>Gfp), the two Tcf-binding sites and the Zic-d site were replaced with Zic-p sites (three Zic-p sites in total). The second construct (5xZic-p+ΔTCF>Gfp) contains two additional Zic-p sites (five Zic-p sites in total). In spite of the lack of Tcf7 sites, 3xZic-p+ΔTCF>Gfp was expressed in B5.1 at the 16-cell stage (Fig. 3D,F; compare it with Fig. 1F), and 5xZic-p+ΔTCF>Gfp drove the reporter more frequently than did 3xZic-p+ΔTCF>Gfp (Fig. 3E,F) (Fisher's exact test P=1.8e-08). These constructs indicate that the enhancer drives gene expression independently of β-catenin/Tcf7 when it binds enough Zic-r.a. Indeed, injection of the β-catenin MO rarely affected reporter expression (Fig. 3F).

Taken together, our data indicate that Zic-r.a binding to the Tbx6-r.b enhancer in normal 16-cell embryos is insufficient to activate Tbx6-r.b. This explains why β-catenin/Tcf7 is required at the 16-cell stage. Our data also indicate that the enhancer binds Zic-r.b sufficiently to activate Tbx6-r.b independently of β-catenin/Tcf7 in B6.4 cells of 32-cell embryos, but not in any other cells of 16-cell or 32-cell embryos.

To understand functional differences between the canonical and non-canonical-motif sites, we changed the Zic-d and Zic-p site sequences to the primary canonical motif sequence (2xcano>Gfp, in which ‘ATCAGCAGGAGAG’ (Zic-p) and ‘CTCAGTGTGACGC’ (Zic-d) were changed to ‘ATCAGCGGGGGGC’ and ‘CTCAGCGGGGGGC’) (Fig. S1D). Although the control non-mutated upstream sequence specifically drove reporter expression only in B5.1 at the 16-cell stage (wt>Gfp), the mutated upstream sequence drove the reporter ectopically (Fig. 4A,B). Ectopic expression was found mainly in b-line cells, which are derived from cells where Zic-r.a mRNA is localized at the four-cell stage (see Fig. 1A). Among the cells of 16-cell embryos, these cells ranked third and fourth, after the germline cells (where transcription is quiescent) and B5.1, in the amount of Zic-r.a they contained (Fig. 3B).

In addition to maternal expression, Zic-r.a is zygotically expressed in neural cells at the tailbud stage (Satou et al., 2002). Judging from its mRNA expression pattern (Satou et al., 2002) and its protein distribution pattern (Oda-Ishii et al., 2016), it is likely that these neural cells are derived from the a-, b- and A-lineages. Concomitant with this zygotic expression, the above mutant construct was expressed ectopically in neural cells, in which Zic-r.a is expressed. However, it was also expressed ectopically at the gastrula stage in nerve cord cells in which Zic-r.a is not expressed. It is possible that a paralog Zic-r.b, which is expressed in nerve cord cells (Imai et al., 2002) and binds to a sequence similar to the canonical site (Yagi et al., 2004b), activates the reporter. Therefore, we used the 3.4 kbp upstream sequence, which was used by a previous study (Kugler et al., 2010). The construct containing the intact 3.4 kb upstream region was expressed specifically in B5.1 at the 16-cell stage (Fig. S3A) and was expressed only in muscle cell lineages at the tailbud stage (Fig. 4C), indicating that the 189 bp upstream sequence lacked elements responsible for repression at the gastrula stage. On the other hand, the mutant construct, which contained two canonical-motif sites instead of the non-canonical-motif sites, induced ectopic expression in neural cells at the tailbud stage (Fig. 4D), as well as ectopic expression in the animal hemisphere at the 16-cell stage (Fig. S3B). Thus, canonical-motif sites drove reporter expression in neural cells at the tailbud stage.

The above observation raised the possibility that Zic-r.a, which is derived from zygotically expressed mRNA, regulates gene expression through canonical binding-motif sites in neural cells at the tailbud stage. To identify such genes, we obtained candidates from embryos in which neural fate was induced by upregulation of Zic-r.b activity and downregulation of activities of Foxa.a, Foxd, Neurog and Erk signaling (Kobayashi et al., 2018). Specifically, we compared transcriptomes of such embryos and embryos into which the Zic-r.a MO was also injected (Fig. 5A). Among greatly downregulated genes, we chose 20 genes (NOIseqsim P>0.99; reads per million values >50), and then examined whether they were expressed in cells with Zic-r.a expression, using published single-cell transcriptome data for middle and late tailbud embryos (Cao et al., 2019). We found that two genes were expressed in Zic-r.a-expressing cells at the middle and late tailbud stages. One (KY.Chr10.963) encodes Claudin,; the other (KY.Chr7.686) did not show a significant similarity to known proteins. Specifically, Claudin was expressed in a subset of cells with Zic-r.a expression, whereas the latter was expressed not only in cells with Zic-r.a expression, but also in other cell populations (Fig. 5B; Fig. S4).

To confirm whether zygotically expressed Zic-r.a controls these genes, we mutated Zic-r.a by injecting a pair of TALEN mRNAs whose products were designed to target the third zinc-finger domain. Because this injection cannot affect maternal Zic-r.a mRNA, we were able to examine effects of zygotically expressed Zic-r.a. To confirm that injection of these TALEN mRNAs successfully mutated Zic-r.a, we extracted genomic DNA and amplified this genomic site by PCR. PCR products were cloned into a plasmid vector and nucleotide sequences of 24 randomly selected clones were determined. All clones contained an insertion or deletion in this site, suggesting high efficiency of these TALEN mRNAs (Fig. S5). We next measured expression of the putative targets by reverse transcription, followed by quantitative PCR (RT-qPCR). We found that mRNA levels of Claudin and KY.Chr7.686 were significantly decreased, while mRNA levels of ubiquitously expressed EF1α or a gene specifically expressed in notochord (Noto1) were not significantly changed (Fig. 5C). The effect upon the expression level of KY.Chr7.686 was weaker than that of Claudin. This is probably because KY.Chr7.686 is expressed not only in cells with zygotic Zic-r.a expression, but also in other cell populations (Fig. 5B; Fig. S4).

In the 1 kb upstream regions of Claudin and KY.Chr7.686, two and one canonical-motif sites were recognized by Patser (Hertz and Stormo, 1999) and the Ciona position weight matrix (Yagi et al., 2004a) (Fig. S6), although no non-canonical-motif sites were recognized with the position weight matrix for the mouse secondary motif. A gel-shift assay showed that these sites competed more strongly than non-canonical-motif sites, which indicates that these sites act as high-affinity sites for Zic-r.a (Fig. 5D). These data support the hypothesis that Claudin and KY.Chr.7.686 are regulated through Zic-r.a binding to primary canonical-motif sites in their upstream regions.

To understand how Zic-r.a recognized canonical and non-canonical motifs, we examined whether Zic-r.a proteins with mutations in one of its five zinc fingers could drive gene expression. We changed two residues that are crucial for DNA recognition (Wolfe et al., 2000) to glycine in each zinc-finger domain (Fig. 6A). Injection of mRNA encoding intact Zic-r.a protein induced ectopic expression of the reporter genes, wt>Gfp and 2xcano>Gfp, in a and A lineages of cells (Fig. 6B,C). The mutation introduced into the first zinc-finger domain (μZF1) reduced the proportion of embryos with ectopic expression of 2xcano>Gfp, but not of wt>Gfp, while the mutation introduced into the third zinc-finger domain reduced the proportion of embryos with ectopic expression of wt>Gfp, but not of 2xcano>Gfp (Fig. 6B,C). In addition, the mutation introduced into the fourth zinc-finger domain slightly reduced the proportion of embryos with ectopic expression of 2xcano>Gfp. Thus, zinc-finger domains that are primarily responsible for binding to canonical- and non-canonical-motif sites appear to differ (Fig. 6D,E). This observation indicates that the non-canonical motif is not a divergent version of the canonical motif. In other words, these two motifs are recognized in different ways by Zic-r.a.

In the present study, we showed that Tbx6-r.b expression is regulated differently in B6.4 of 32-cell embryos from B5.1 of 16-cell embryos (Fig. 7). While maternal Zic-r.a activates Tbx6-r.b cooperatively with β-catenin and Tcf7 in the B5.1 lineage of 16-cell embryos (Oda-Ishii et al., 2016), Zic-r.a alone activates Tbx6-r.b in the B6.4 lineage of 32-cell embryos. In B6.4, the limited number of weak non-canonical-motif sites (Zic-d and Zic-p) activate Tbx6-r.b in response to a high Zic-r.a concentration. The B5.1 lineage contains less Zic-r.a than does the B6.4 lineage; therefore, it is likely that the Zic-d- and Zic-p-binding sites cannot respond sufficiently at the 16-cell stage. On the other hand, because the Zic-d site abuts the Tcf7-binding site (Fig. S1), and because Zic-r.a and Tcf7 can interact (Oda-Ishii et al., 2016), Zic-r.a may bind to these non-canonical sites with the help of Tcf7 in 16-cell embryos. Thus, the weak non-canonical motif is used by reducing and tuning the activity of the enhancer, and it induces Tbx6-r.b expression in the B6.4 lineage without the help of β-catenin/Tcf7.

In ascidian embryos, zygotic transcription begins between the 8-cell and 16-cell stages in most somatic cells (Oda-Ishii et al., 2018). However, the most posterior cells, which have the potential to give rise to germ cells, are transcriptionally silent (Kumano et al., 2011; Shirae-Kurabayashi et al., 2011), and B6.4 cells in the 32-cell embryo are somatic daughters of the most posterior cell pair of the 16-cell embryo. Therefore, transcription in the B6.4 lineage begins at the 32-cell stage. Because nuclear β-catenin is unavailable in this lineage (Hudson et al., 2013), the regulatory mechanism that activates Tbx6-r.b at the 16-cell stage does not act in the B6.4 lineage at the 32-cell stage. This is probably the reason that the ascidian embryo has evolved a distinct regulatory mechanism activating Tbx6-r.b in the B6.4 lineage.

Similarly, another Zic gene, Zic-r.b, and Snail are expressed in the B5.1 and B6.4 lineages at the 32-cell stage, and the regulatory mechanisms regulating these genes differ in these two lineages: Zic-r.b is activated in the daughter cells of B5.1 by maternal Gata.a, while the expression in B6.4 is not lost in Gata.a morphants (Imai et al., 2016); Snail is activated in the daughter cells of B5.1 by Tbx6-r.b, and in B6.4 by constitutively active form of Raf (Tokuoka et al., 2018). Intriguingly, the regulatory mechanisms activating these three genes in the B6.4 lineage also differ, suggesting that the ascidian embryo has independently evolved these mechanisms under a common selective pressure.

The present and previous studies have established that Zic-r.a acts in three modes (Fig. 7). First, at the 16-cell stage, it acts by interacting with a complex of Tcf7 and β-catenin to activate Tbx6-r.b (Oda-Ishii et al., 2016). Second, in B6.4 of the 32-cell embryo, a high level of Zic-r.a expression allows it to bind to non-canonical-motif sites in the upstream regulatory region of Tbx6-r.b. Third, in late embryos, Zic-r.a, which is zygotically expressed in the nervous system, activates neural genes through canonical-motif sites.

One unsolved problem is why neural genes that are controlled by Zic-r.a through canonical sites are not expressed in early embryos. It is possible that a higher order of transcriptional regulation, including epigenetic regulation, is responsible, although this remains to be tested. It is also possible that Zic-r.a activates neural genes combinatorially with additional factors specific to neural cells. On the other hand, our experiments indicate the possibility that Tbx6-r.b is not activated in neural cells of late embryos because the amount of Zic-r.a is insufficient. This possibility has not been tested either, because comparisons of Zic-r.a concentrations in early and late embryos are technically difficult.

We showed that the first and third zinc-finger domains are important to drive gene expression through canonical and noncanonical binding sites. This observation indicates that these two sites (motifs) are recognized differently by Zic-r.a. That is, the non-canonical-binding motif is not a variant of the canonical motif. Therefore, Zic-r.a indeed uses two sites with different motifs to induce expression of specific genes in early and late embryos, respectively.

On the other hand, the non-canonical-motif sites are low-affinity sites. Low-affinity sites are often used to control gene expression temporally and spatially (Crocker et al., 2015; Farley et al., 2015; Fuqua et al., 2020; Gaudet and Mango, 2002). Although most low-affinity sites are thought to be variants of high-affinity sites, the non-canonical motif recognized by Zic-r.a is not. Several studies have suggested that some transcription factors can recognize two or more motifs (Badis et al., 2009; Franco-Zorrilla et al., 2014), although this remains controversial (Jolma et al., 2013; Morris et al., 2011; Zhao and Stormo, 2011). In ascidian embryos, Zic-r.a recognizes two motifs with different affinities, and activates its target genes in a specific cell pair through non-canonical-motif sites. Zic-r.a is a transcription factor with five zinc-finger domains. Our data indicate that this modular nature enables this transcription factor to recognize multiple motifs.

Zic-r.a is composed of three exons and the five zinc-finger domains are encoded by these three exons. While expressed sequence tags have been obtained from 34 cDNAs of seven cDNA libraries (Satou et al., 2005; Tassy et al., 2010), none of them indicates alternative splicing. Therefore, it is not likely that different isoforms recognize different motifs, but it is more likely that a single isoform recognizes the canonical and non-canonical motif sites.

Gene regulatory networks for embryonic fate specification have been studied extensively. Knockout, knockdown and overexpression assays to identify network connections have been complemented by reporter assays and chromatin immunoprecipitation assays to identify direct interactions. However, it is common that not all peaks identified by chromatin immunoprecipitation assays contain recognizable binding sites. Non-canonical sites may be present within such peak regions.

In summary, the non-canonical motif of Zic-r.a is functional and important for ascidian embryogenesis. Multiple motifs that some transcription factors recognize may be used for tuning enhancer activity to evoke expression of specific targets.

Adult specimens of Ciona intestinalis (type A; also called Ciona robusta) were obtained from the National BioResource Project for Ciona. cDNA clones were obtained from our EST clone collection (Satou et al., 2005). Whole-mount in situ hybridization was performed as described previously (Satou et al., 1995). Identifiers for genes examined in this study are: KY.Chr11.468-470 for Tbx6-r.b, KY.Chr1.1698 for Zic-r.a, KY.C9.48 for β-catenin, KY.Chr11.1167 for Foxa.a, KY.Chr8.660/661 for Foxd, KY.Chr6.427 for Neurog, KY.Chr10.963 for Claudin, KY.Chr4.359 for Pou2, KY.Chr6.606 for Noto1 and KY.Chr14.194 for Ef1a.

MOs for Zic-r.a, β-catenin, Foxa.a, Foxd and Neurog, which block translation, have been used previously and their specificity has been evaluated (Imai et al., 2004; Kobayashi et al., 2018; Yagi et al., 2004a). These MOs were microinjected under a microscope. For overexpression, the coding sequence of Zic-r.a was cloned into pBluscript RN3 (Lemaire et al., 1995). Injected RNAs were transcribed using a mMESSAGE mMACHINE T3 Transcription Kit (ThermoFisher Scientific).

To mutate Zic-r.a, we used TALEN (transcription activator-like effector nuclease) technology. N- and C-terminal domains of TALE and FokI nuclease domain were taken from the Platinum Gate TALEN kit (Sakuma et al., 2013). Protein products derived from a pair of synthetic mRNAs were designed to cleave the first exon of Zic-r.a (recognition sequences 5′-CTTTGCTCGAAGCGAAAACC-3′ and 5′-GTCTCTTACCAGTGTGAGTT-3′). These mRNAs were introduced by microinjection into fertilized eggs. To confirm successful cleavage, we performed PCR using the following primers: 5′-GATCTATTTCTGGTTTTACCTTG-3′ and 5′-CTGGTACGATGGGATATGAATC-3′ .

Reporter constructs were introduced into fertilized eggs by electroporation or microinjection. Chromosomal positions of upstream sequences for reporter constructs and mutated sequences are indicated in Fig. S1. We randomly chose embryos with introduced reporter constructs to examine reporter construct expression using in situ hybridization. These experiments were performed at least twice with different batches of embryos.

The patser program was run with the following parameters: ‘-c -A a:t 65 c:g 35 -li’.

Recombinant Zic-r.a protein was produced as a His-tag-fusion protein in E. coli and purified with Ni-NTA agarose (QIAGEN). After annealing two complementary oligonucleotides (Zic-p, 5′-aaaACACGAATCAGCAGGAGAGTTC-3′ and 5′-aaaGAACTCTCCTGCTGATTCGTGT-3′; Zic-d, 5′-aaaATGCGTCACACTGAGTTTTGGA-3′ and 5′-aaaTCCAAAACTCAGTGTGACGCAT-3′; cano, 5′-aaaACTAGTGCCCCCCGCTGCGCGG-3′ and 5′-aaaCCGCGCAGCGGGGGGCACTAGT-3′), the protruding ends of the double-stranded oligonucleotides were filled with biotin-11-dUTP. The resultant biotin-labeled oligonucleotides were used as probes. Unlabeled double-stranded oligonucleotides with the same sequences were used as specific competitors. Other competitors were similarly made from two oligonucleotides (Claudin up, 5′-aaaTGTCTTCCCCCACCCTACTATACAT-3′ and 5′-aaaATGTATAGTAGGGTGGGGGAAGACA-3′; Claudin down, 5′-aaaATAGTAGGGTGGGGGGAGATGGAAC-3′ and 5′-aaaGTTCCATCTCCCCCCACCCTACTAT-3′; KY.Chr7.686, 5′-aaaTTGCGTCGCAGGGGGCTGACACGGC-3′ and 5′-aaaGCCGTGTCAGCCCCCTGCGACGCAA-3′). Mutant competitors were produced from two complementary oligonucleotides (μZic-p, 5′-aaaACACGAATTCATTGGAGAGTTC-3′ and 5′-aaaGAACTCTCCAATGAATTCGTGT-3′; μZic-d, 5′-aaaATGCGTCATGTCAAGTTTTGGA-3′ and 5′-aaaTCCAAAACTTGACATGACGCAT-3′). Proteins and biotin-labeled probes were mixed in 25 mM Hepes (pH 7.5), 100 mM KCl, 1 mM DTT, 50 ng/μl poly(dIdC), 2.5% glycerol, 0.05% NP40 and 50 μM ZnSO₄ with or without competitor double-stranded DNAs. Protein concentrations were empirically determined. Protein-DNA complexes were detected using a Chemiluminescent Nucleic Acid Detection Module Kit (ThermoFisher Scientific). Bands were quantified as arbitrary units using an imager (ChemiDoc XRS, BioRad) and Quantity-One software (BioRad).

Immunostaining with anti-Zic-r.a antibody was performed as described previously (Oda-Ishii et al., 2016). ImageJ was used to quantify fluorescence intensity. All photographs were taken in the same conditions and DAPI signal intensity was used for reference.

To induce neural fate in single-cell syncytium embryos, we injected Zic-r.b mRNA and MOs for Foxa.a, Foxd and Neurog, and incubated injected embryos in sea water containing the MEK inhibitor U0126 and cytochalasin B, as described previously (Kobayashi et al., 2018). We also prepared embryos injected with the Zic-r.a MO. We next performed RNA-sequencing for these two specimens using Ion Total RNA-Seq kit ver 2 (ThermoFisher Scientific) and an Ion PGM instrument (ThermoFisher Scientific), as described previously (Tokuhiro et al., 2017) (DRA accession number DRA011314). We did not take duplicates, because we used this experiment for screening; the obtained result was confirmed with other methods. NOISeq software (Tarazona et al., 2011) was used to identify differentially expressed genes. Single-cell transcriptome data (Cao et al., 2019) were mapped to a gene model set recently developed (Satou et al., 2019) and analyzed using Cell Ranger (10×Genomics).

For RT-qPCR, we extracted RNA and converted RNA to cDNA using a Cells-to-Ct kit (ThermoFisher Scientific). cDNA samples were analyzed by quantitative PCR with the SYBRG method. Primers used were: Claudin, 5′-CGATGTCTTACAGGGTGGTTCT-3′ and 5′-CCGATGGTATAACCAAGTCCAG-3′; KY.Chr7.686, 5′-TGTACCAGACATGAGAGCGAAA-3′ and 5′-CGTAAGCCGCCTTCAGTTC-3′; Pou2, 5'-AAGATGGTTGCTGGATGCTAATAAT-3′ and 5'-TTGGATTGGAGTGGGAATAACAA-3′; Ef1α, 5'-CTCCCGGTCACAGAGATTTCA-3′ and 5′-CAATAAGCACGGCACAATCG-3′; and Noto1, 5′-GGCTTGCCTGCGAATGG-3′ and 5′-GAGCACACGACTGCATCGTAA-3′.

We thank Reiko Yoshida (Kyoto University), Manabu Yoshida (University of Tokyo) and other members working under the National BioResource Project for Ciona (MEXT, Japan) at Kyoto University and the University of Tokyo for providing experimental animals. We thank Steven D. Aird for editing a draft of the manuscript.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199538