yap1b, a divergent Yap/Taz family member, cooperates with yap1 in survival and morphogenesis via common transcriptional targets

The Yap/Taz family of transcriptional co-activators plays a key role during embryogenesis, acting as master regulators of growth, cell fate specification and survival, as well as in adult organs, where they are crucial for tissue repair and cancer progression (Piccolo et al., 2014; Varelas, 2014; Zanconato et al., 2016). The Yes-associated protein 1 (Yap1 or Yap) was the first member of the family identified on the basis of its interaction with members of the Src/Yes protein tyrosine kinase family (Sudol, 1994). A few years later, a second paralog of the family, Taz (transcriptional co-activator with PDZ-binding domain; also known as WW domain-containing transcription regulator protein 1 or Wwtr1), was identified through the interaction with 14-3-3 (Kanai et al., 2000). Yap and Taz share a very similar domains structure, including N-terminal Tead- and 14-3-3-binding domains, one or two WW domains, and a C-terminal transcriptional activation domain that ends in a short PDZ-domain recognition sequence (Fig. 1). Yap/Taz, as well as their Drosophila ortholog Yorkie (Huang et al., 2005), do not contain a DNA-binding domain and hence need to be associated to a transcription factor to exert their regulatory function. Pioneer work revealed the interaction between the Yap N terminus and the four members of the Tead family of transcription factors in vertebrates (Vassilev et al., 2001). Furthermore, the study determined that the C-terminal acidic domain in Yap is in fact the transcriptional activation domain for Tead proteins (Vassilev et al., 2001). This interaction was further confirmed for the corresponding orthologs in Drosophila: Yorkie and Scalloped (Bandura and Edgar, 2008; Zhang et al., 2008). Although Yap/Taz activity can be modulated by a number of transcription factors, including members of the AP-1, Runx, β-catenin, Tbx, p53 and Smad families (Piccolo et al., 2014), only its association to Tead has been proved to be essential for chromatin-binding and transcriptional effects (Galli et al., 2015; Li et al., 2010; Stein et al., 2015; Zanconato et al., 2015; Zhao et al., 2008).

One of the most relevant features of the Yap/Taz-Tead module is its ability to integrate upstream information from signaling pathways and structural cell properties, such as cell adhesion, density, polarity, contractility and shape, into a contextual transcriptional response. This is achieved through a number of regulatory inputs, with the Hippo pathway being the best understood of them. The core components of the Hippo pathway were identified due to the striking tissue overgrowth observed in Drosophila mutants (Harvey et al., 2003; Tapon et al., 2002; Wu et al., 2003). This evolutionary conserved signaling cascade is regulated by two sequential kinases, Mst1/2 and Lats1/2 (Hippo and Warts respectively in Drosophila), which converge in the phosphorylation of Yap/Taz and result in its sequestration and/or degradation in the cytoplasm. When the inhibitory pathway is not active, Yap/Taz are free to shuttle from the cytoplasm to the nucleus and to exert their role as transcriptional co-activators (Meng et al., 2016). Apart from the Hippo pathway, the activity of the Yap/Taz is controlled by many other factors, most notably mechanical cues. Yap/Taz translocation to the nucleus is regulated by cell geometry, being active in cells that have undergone cell spreading and inactive in round and compact cells (Dupont et al., 2011). The ability of Yap/Taz to sense mechanical tension depends on myosin contractility, actin-capping and -severing proteins, and the integrin-talin mechanosensitive clutch (Aragona et al., 2013; Dupont et al., 2011; Elosegui-Artola et al., 2016). Moreover, recent work has shown that mechanical coupling between the ECM and the nuclear envelope is responsible for Yap/Taz nuclear translocation across nuclear pores, a process that does not depend on the activity of the Hippo pathway (Elosegui-Artola et al., 2017).

A second Hippo-independent regulatory input on Yap/Taz comes from members of the Src family of tyrosine kinases (SFKs). As already mentioned, Yap was initially identified as a Yes- and Src-interacting protein (Sudol, 1994). A number of SFKs, including c-Abl, Src and Yes, have been shown to phosphorylate Yap in a specific tyrosine residue (Y357 in humans) located at the C-terminal transcriptional activation domain (Levy et al., 2008; Rosenbluh et al., 2012; Vlahov et al., 2015; Zaidi et al., 2004). More recently, Src-mediated phosphorylation of the three conserved tyrosine residues (including Y357) of the Yap C-terminal domain has been reported in keratinocytes (Li et al., 2016). In contrast to the inhibitory role of the Hippo pathway, SFK-mediated phosphorylation of Yap results in the transcriptional activation of Yap/Tead (Calvo et al., 2013; Li et al., 2016; Tamm et al., 2011), Yap/Runx2 (Zaidi et al., 2004), Yap/p73 (Levy et al., 2008) or Yap/β-Catenin/Tbx5 complexes (Rosenbluh et al., 2012; Vlahov et al., 2015). Importantly, It has been suggested that Yap phosphorylation by Src may mediate the oncogenic activity of the kinase (Li et al., 2016), which is in agreement with the pervasive activation of Yap in cancerous cells (Zanconato et al., 2016).

Phylogenetic analysis has shown an ancient evolutionary history for the Yap-Tead module. Bona fide orthologs of Tead are already present in yeast, whereas Yap-related proteins can be traced back to basal metazoans such as Nematostella (Hilman and Gat, 2011). Despite this general conservation, significant differences have also been reported. Importantly for this work, the conserved tyrosine residues required for SFKs phosphorylation appear to be a vertebrate innovation in the C-terminal region of Yap/Taz (Elbediwy and Thompson, 2018; Hilman and Gat, 2011). The vertebrate paralogs Yap and Taz emerged after the two rounds of whole-genome duplication that occurred in the stem of the lineage (Dehal and Boore, 2005). Data from different vertebrate models indicate that Yap and Taz have partially overlapping roles during development. Individual mutations of Yap and Taz in mice result in very different phenotypes: from defects in elongation, impaired yolk sac vasculogenesis and lethality at day 8.5 in Yap^−/− embryos, to polycystic kidneys and premature death in Taz^−/− adults (Morin-Kensicki et al., 2006; Xin et al., 2013). However, the strong phenotype observed in both double mutant embryos (i.e. lethality prior to the morula stage) (Nishioka et al., 2009) and in tissue-specific double knockouts (Deng et al., 2017; Lee et al., 2016; Singh et al., 2016) indicate functional redundancy for Yap and Taz. The same paralogs, yap and taz, have been reported in teleosts. Similarly to mammalian genes, zebrafish paralogs seem to play redundant roles, severely impairing early development when they are inactivated simultaneously (Kimelman et al., 2017; Miesfeld et al., 2015).

Here, we analyze the distribution of Yap family proteins in teleosts and describe Yap1b as a new paralog with an atypical transcriptional activation domain. We use DamID-seq to characterize, early during development, the DNA sites targeted by Yap1 and Yap1b in the medaka genome, as well as by Yap1 and Taz in the zebrafish genome. We show that CRISPR/Cas9-induced mutation of yap1b in medaka neither results in apparent embryonic defects nor affects adult fish survival and fertility. However, simultaneous mutation of yap1 and yap1b leads to enhanced apoptosis and deeper morphogenetic defects, indicating that both paralogs cooperate during early embryogenesis in medaka.

Medaka embryos carrying a nonsense mutation for yap1 (i.e. a premature stop codon at the WW1 domain), named hirame (hir), display deep phenotypic malformations, including optic cup flattening, lens misalignment, delayed blastopore closure and increased cell death early during development (Porazinski et al., 2015). We confirmed these results by generating two additional yap1 mutant alleles using CRISPR/Cas9 in medaka. The chromosomal lesions consist of 7 and 11 bp deletions in exon 2 that result in frameshift mutations and premature stop codons before the WW1 domain of the protein (Fig. S1A,B). Both alleles showed a phenotype very similar to the previously described mutation hir, with the exception that the blastopore closure defects observed are very mild (Fig. S1C-H). The strong phenotype displayed in yap1 medaka mutants was much more severe compared with the milder defects observed in zebrafish upon yap1 mutation, and rather resembled the malformations observed in double yap1/taz mutants in zebrafish (Agarwala et al., 2015; Kimelman et al., 2017; Miesfeld et al., 2015).

To investigate whether the phenotypic discrepancy between medaka and zebrafish mutants could be due to a different contribution from paralogs, we carried out a phylogenetic analysis of the Yap protein family in teleosts. A maximum likelihood phylogenetic tree was constructed using multiple sequence alignment of proteins from seven representative teleost species. Human and mouse sequences were also included as an external reference. In spite of previous reports describing two paralogs, Yap and Taz (Wwtr1), in all vertebrates, our phylogenetic analysis unambiguously identified three distinct groups of proteins: a first group clustering all proteins annotated as Yap1; a second including Taz proteins; and a third here termed Yap1b, as it showed higher sequence homology to Yap1 than to Taz (Fig. 1A). We could identify the presence of genes encoding Yap1b proteins in all species examined from the Euteleosteomorpha clade (i.e. cod, medaka, tilapia and stickleback), the largest group within teleosts (Betancur et al., 2017). This suggests the possibility of yap1b emerging after the teleost-specific genome duplication (Amores et al., 1998). To test this hypothesis, we examined the syntenic arrangement of the genomic region containing Yap family genes in teleosts. Conserved synteny indicates that the loci containing yap1 and yap1b derived from a common ancestor after the teleost genome duplication, whereas Taz genes show an independent genome arrangement (Fig. S2). This result confirms that, in contrast to previous assumptions (Porazinski et al., 2015; Webb et al., 2011), the second paralog present in the medaka genome is not orthologous to taz but rather to yap1b.

We then examined the structure of Yap1b proteins. Multiple sequence alignments showed relatively well-conserved N-terminal Tead and WW domains for Yap1b. In particular, the Tead binding domain of the protein is more similar to that of Yap1 than to the Taz domain, showing average identities of 65.5% and 40.3%, respectively (Fig. S3). However, the transcriptional activation domain of Yap1b seems less conserved when compared with Yap1 and Taz equivalent regions (i.e. average identities 44.8% and 23.4%, respectively). Remarkably, all conserved tyrosine residues that can be phosphorylated by kinases of the Src family, and even the structurally related phenylalanine residues, are substituted by other conserved residues or eliminated in the transcriptional activation domain of Yap1b (Fig. 1B,C; Fig. S3D). These modifications are predicted to have a significant impact on the regulatory properties of the protein, as SFK-mediated phosphorylation plays a key role in the transcriptional activation of Yap/Tead complexes (Li et al., 2016).

To gain insight into the evolutionary history of Yap family proteins, particularly into the evolution of Yap1b, we scanned the genomes of four additional species that belong to key branches of the teleost phylogenetic tree (Fig. S4): Japanese eel (Anguilla japonica; Elopomorpha); elephantfish (Paramormyrops kingsleyae; Osteoglossomorpha); herring (Clupea harengus; Otomorpha: Cupleiformes); and northern pike (Esox lucius; Euteleosteomorpha: Protacanthopterygii). Multiple sequence analyses of the transcriptional activation domain N-terminal region of Yap-related proteins revealed the presence in basal teleosts (i.e. japanese eel and elephantfish) of four paralogs, likely arising by duplication of the original Yap and Taz (Fig. S4; Table S1). Based on the overall sequence conservation and on key reference residues, we could identify genes encoding both Yap1 and Yap1b in the basal teleost genomes. Nevertheless, the transcriptional activation domain of Yap1b seems less divergent in these species, and tyrosine and phenylalanine residues are partially conserved (Fig. S4A). Our analyses are consistent with the evolutionary loss of yap1b in the entire Otomorpha lineage, which comprises more than 8000 living species (including herring, mexican tetra and zebrafish). They are also consistent with the evolutionary divergence of Yap1b transcriptional activation domain early in the Euteleosteomorpha lineage (including pike, cod, medaka, tilapia and stickleback). Finally, although the evolutionary history of Taz paralogs seems complex and may require further investigation, it appears that both copies were lost sequentially during teleost evolution. Thus, only one copy is present in the Otomorpha group and none in Percomorpha (i.e. ≈15,000 species, one out of four vertebrate species) (Fig. S4B). In summary, our phylogenetic analyses reveal a complex evolutionary history for Yap family proteins in teleosts, and identify the emergence in Euteleosteomorpha of Yap1b, a paralog with a divergent transcriptional activation domain.

To understand the role of yap1b during early embryogenesis in medaka, we focused our expression studies on the developmental window extending from late gastrulation (stage 15-16) to optic cup stages (stage 23-24). This developmental period corresponds to the stages when morphogenetic defects first become apparent both in yap1^−/− medaka embryos (Porazinski et al., 2015) and yap^−/−;taz^−/− zebrafish embryos (Kimelman et al., 2017). Previous expression studies in zebrafish reported a yap1 broad tissue distribution, with an enriched expression at the neural plate and axial mesoderm (Kimelman et al., 2017; Thisse et al., 2001). To extend these observations to medaka embryos, we investigated yap1 and yap1b expression by in situ hybridization and qPCR (Fig. 2). In situ hybridization analysis at late gastrula stages revealed that medaka yap1 and yap1b display a similar expression pattern to their zebrafish ortholog yap1, being broadly expressed, yet enriched, at the axial mesoderm (Fig. 2A-B′). We then examined yap1 and yap1b relative transcript levels at three developmental stages by qPCR: before the onset of zygotic transcription (stage 7), at late gastrula (stage 16) and at optic cup stages (stage 24). This analysis revealed that, whereas yap1 levels are relatively constant, yap1b shows a high maternal contribution and is expressed at low levels once the zygotic transcription starts (Fig. 2C,D). For comparative purposes, similar quantitative analyses were performed in zebrafish for yap1 and taz at equivalent developmental stages. This analysis revealed that, in contrast to taz, yap1 has a strong maternal contribution (Fig. 2E,F). To achieve an accurate measurement of RNA levels, we performed a digital droplet PCR (ddPCR) study (Pinheiro et al., 2012). This approach confirmed our qPCR observations and allowed us to estimate the relative proportion of each paralog in medaka and zebrafish (Fig. 2G,H). Our data indicate that zebrafish yap1 and taz are expressed at similar levels after the maternal-to-zygotic transition (MZT). In contrast, and despite being strongly contributed by the mother, medaka yap1b is expressed only at low levels after MZT. This observation alone could explain the strong phenotype observed upon yap1 mutation in medaka.

To investigate the regulatory properties of Yap family proteins in teleosts, particularly those of Yap1b, we characterized their target sites in the genome. As ChIP-grade antibodies are not available for Yap proteins in teleosts, we used DamID-seq to profile their binding. The DNA adenine methyltransferase identification (DamID) method was initially developed in Drosophila (van Steensel et al., 2001), and it is based on the expression of transcription factors and chromatin modifiers tethered to the bacterial DAM methylase. Although extensively used in Drosophila (Aughey and Southall, 2016), the technique has only recently been adapted for in vivo studies in vertebrates (Gutierrez-Triana et al., 2016; Tosti et al., 2018). In this work, we followed the iDamID-seq/iDEAR pipeline, first established in medaka (Gutierrez-Triana et al., 2016). We generated mycDAM-tagged constructs for medaka DAM:yap1 and DAM:yap1b, and injected their corresponding mRNAs, as well as the control DAM:GFP mRNA, into one-cell stage medaka embryos. Proper shuttling of the fusion proteins to the nucleus was confirmed by following either DAM:GFP signal (Fig. S5A) or Myc-associated immunoreactivity (Fig. S5B). Embryos (∼40 per experimental condition) were harvested at 80% epiboly (stage 16) to capture DNA occupancy throughout gastrulation. For comparative purposes, DamID-seq experiments were carried out in zebrafish, also using 80% epiboly embryos (8 hpf), injected with DAM:yap1, DAM:taz and DAM:GFP mRNAs at the one-cell stage. After isolation and amplification of Dam methylated regions in three independent biological replicates (Fig. S6), samples were submitted for deep sequencing and retrieved reads were mapped on the medaka and zebrafish reference genomes. The correlation of mapped reads was high among replicas (i.e. above 70%), whereas it was very low when compared with the corresponding DAM:GFP controls (Fig. S7).

We used the iDEAR pipeline (Gutierrez-Triana et al., 2016) to identify differentially methylated peaks for each Yap family regulator. A representative genome browser example of peaks identified at the locus of amotl2a, a known Yap1 transcriptional target, is shown both for medaka and zebrafish (Fig. 3A,B); additional medaka examples for the known targets ctgfa, cyr61, lats2, ccne1, boka and yap1 itself are provided in Fig. S8.

A total of 4212 and 3792 DamID peaks were identified in the medaka genome for Yap1 and Yap1b, respectively, and 7553 and 8608 in the zebrafish genome for Yap1 and Taz, respectively (Table 1, Table S5). In medaka, we found a large overlap between Yap1 and Yap1b peaks (i.e. ≈65% of the Yap1b peak was also positive for Yap1), whereas only 28.6% of the Yap1 peaks are also targeted by Taz in zebrafish (Fig. 3C, Fig. S9A,B). The overlap between the identified DamID peaks and 80% epiboly ATAC-seq marks, which are available in medaka and zebrafish (Marlétaz et al., 2018), was also calculated. Approximately 42% and 45% of the Yap1 and Yap1b peaks, respectively, overlap with open chromatin regions in medaka (Fig. 3D,E); and 49% and 25% of the Yap1 and Taz peaks, respectively, extend over ATAC regions in zebrafish. A comparative analysis of peak distribution according to the distance to their closest transcription start site (TSS) revealed an enrichment of Yap-binding sites at promoters and proximal enhancers with respect to the Dam:GFP controls (Fig. S9C-F).

We used DREME (Bailey, 2011) to identify significantly enriched motifs in the collection of peaks targeted by the different Yap family proteins (Fig. 4, Fig. S10). Among the most enriched motifs for medaka Yap1 and Yap1b, the top hits corresponded to the consensus DNA sequence of TEAD proteins: DREME E-values 5.3×10⁻¹⁴³ and 5.1×10⁻⁹⁸, respectively (Fig. 4A,B). Further analyses revealed that a significantly large proportion of the Yap1 and Yap1b peaks contain the TEAD motif, particularly those regions with higher score in the iDEAR pipeline (Fig. 4C). Besides TEAD factors, a number of known motifs showed less significant enrichment in peaks targeted only by Yap1 (e.g. RREB1, TBX and ZNF143), by Yap1b (e.g. GATA2, OTX2 and TP73) or by both (e.g. POU and SOX) (Fig. 4A,B). In contrast to medaka paralogs, the search for enriched motifs using DREME returned different results when zebrafish Yap1 and Taz peaks were compared. Zebrafish Yap1 peaks showed enrichment in a series of known motifs similar to that of the medaka paralogs, including TEAD (as the second top hit: DREME E-value 7.4×10⁻¹⁴⁷), and also FOX, SOX, TP63 and POU (Fig. S10A). However, the collection of DNA-binding motifs associated with zebrafish Taz peaks seemed more divergent, with TEAD being only the 22nd best ranked motif (DREME E-value 9.5×10⁻³) (Fig. S10B). This observation may simply suggest divergent cis-regulatory properties for zebrafish Taz. Alternatively, the disparities observed could be attributed to an incomplete detection of TEAD motifs by DREME, owing to small differences in the consensus sequence. To explore this possibility, we scanned the Yap family protein peaks imposing a consensus TEAD motif using AME. Using this approach, we could detect a very significant enrichment of the TEAD motif in the sets of peaks associated with all Yap family proteins, including zebrafish Taz (Fig. 4D).

To further characterize the chromatin regions targeted by Yap family proteins, we examined the co-occurrence of Yap-associated motifs identified in previous enrichment analyses (Table S2) using FIMO (Grant et al., 2011). Significant interactions among motifs were represented as a Cytoscape network (Shannon et al., 2003). The comparative analysis of these interaction networks revealed a very similar topology (i.e. connectivity between nodes) for motifs associated with medaka Yap1 and Yap1b, with TEAD1, RUNX1, AP1, SMAD, SOX10, FOXD3 and POU5F1 being among the most interconnected nodes (Fig. S11A,B). In contrast, zebrafish Yap1 and Taz showed a more divergent network topology, but some of the nodes, such as POU5F1, FOXD3 and AP1, still displayed a dominant weight in both networks (Fig. S11C,D). This is in agreement with the observation that the overlap between medaka Yap1 and Yap1b peaks is larger than that of zebrafish Yap1 and Taz (Fig. S9A,B).

To examine the collection of genes potentially targeted by Yap family proteins, we associated each peak to the closest TSS. As previously detected for the DamID peaks themselves, we observed that a large fraction of the Yap1b target genes (71%) were also associated with Yap1 in medaka. However, the proportion of genes potentially co-regulated by Taz and Yap1 was smaller (43%) in zebrafish (Table 1, Fig. S9G,H). Our observations are in line with the conservation of the Tead-binding domain between Yap1 and Yap1b (Fig. S3), and indicate that, despite the Yap1b divergent transactivation domain, both medaka paralogs target a largely overlapping set of genes. To gain insight into the attributes of these genes, we used Panther (Mi et al., 2017) to investigate statistically overrepresented gene ontology (GO) terms. When GO terms from the categories ‘molecular function’ and ‘protein class’ were examined, a similar set of enriched terms was obtained for genes targeted by Yap1, by Yap1b or by both, including ‘transmembrane receptor tyrosine kinase activity’ and ‘transcription factor activity’ as the most significant terms (Table 2, Table S3). Regarding GO terms linked to the ‘Biological process’ category, this list included terms such as: ‘negative regulation of apoptotic process’, ‘embryo development’, ‘cell differentiation’ and ‘regulation of cell cycle’, which are consistent with the reported roles for Yap family members in differentiation, proliferation and cell survival (Piccolo et al., 2014). This finding, together with the observation of Yap1 and Yap1b binding to well-known targets of the Hippo pathway in mammals (Fig. 3A, Fig. S8) suggests shared features in the transcriptional program controlled by Yap proteins across vertebrates. To investigate the extent to which this program is conserved, we compared the overlap between the lists of genes we identified as Yap family targets by DamID-seq, with genes identified as targets in previous ChIP-seq studies using mammalian cells (Estarás et al., 2017; Lian et al., 2010; Zanconato et al., 2015). The overlap between these gene lists is statistically very significant in all cases (Fig. S12A). Thus, ∼25% of the genes identified in each our four DamID-seq experiments are also present in at least three out of the six ChIP-seq datasets considered. Furthermore, a few hundred genes, ∼10% of those identified by DamID-seq, are included in five or six of the ChIP-seq datasets (Fig. S12B, Table S6). This group comprises many known transcriptional targets of Yap/Tead complexes, such as ajuba, amotl2a, arhgef7a, ccnd2a, ctgf, cyr61, lats2, smad6b or tead1b (Dataset S2). This observation indicates that, despite the different cell types, developmental stages, paralogs and even species considered, a substantial part of the gene battery controlled by Yap family proteins is conserved across vertebrates.

Our DamID-seq analysis indicated that the identified paralog Yap1b targeted a similar collection of genes through a largely overlapping set of cis-regulatory elements. This may suggest that, in spite of its divergent C-terminal activation domain, Yap1b plays a cooperative role, together with Yap1, in transcriptional regulation. Alternatively, Yap1b could act as a dominant-negative form of Yap1 if the atypical transactivation domain was unable to recruit the Pol II complex. To distinguish between these possibilities and to further understand the role of Yap1b, we generated CRISPR-mediated targeted mutations at the exon2 of medaka yap1b. Two different alleles, yap1bΔ136pb and yap1bΔ84pb, were identified and maintained as stable lines. These harbor chromosomal lesions that consist of exon 2 deletions of 136 and 84 bp (Fig. 5A). The first, yap1bΔ136pb, results in a frameshift mutation and a premature stop codon within the Tead-binding domain of the protein; the second, yap1bΔ84pb, in a 28 amino acid deletion of the Tead-binding domain (Fig. 5B). Despite both alleles being predicted to be null mutations, yap1b^−/− embryos developed normally and could not be distinguished from their wild-type siblings (Fig. 5C-H). In fact, fish survival and fertility were not affected in adult yap1b^−/− mutants. This result excluded the possibility of yap1b playing an obvious non-redundant role during embryogenesis and tissue maintenance. To investigate whether the gene may have a cooperative role with its paralog yap1b, we crossed double heterozygous fish, yap1^+/−; yap1b^+/−, and examined the phenotype and genotype of their progeny. These analyses showed a negative epistatic interaction between both genes. Thus, the phenotypic malformations observed in yap1^−/− mutants (i.e. optic cup flattening, delayed blastopore closure and increased cell death) appeared enhanced in yap1^−/−; yap1b^+/− embryos, and were even more evident in yap1^−/−; yap1b^−/− double mutants (Fig. 6A-D). Quantitative measurements in wild type, single and double mutants at stage 20 confirmed an increased optic cup deformity, tissue flattening, embryo axis shortening and delayed blastopore closure in double mutants (Fig. S13). Bright-field microscope examination of double mutants also revealed signs of increased cell death from stage 20-21 onwards. To explore this possibility, we analyzed caspase 3 levels in stage 20 embryos collected from a yap1^+/−;yap1b^+/− cross (Fig. 6E-L). These stainings confirmed a significant and progressive increase in the number of apoptotic cells in yap1^−/− embryos when additional copies of yap1b are mutated (Fig. 6Q). In contrast, similar levels of apoptosis were observed in embryos harboring at least one wild-type allele of yap1 (Fig. S14). In fact, no morphological malformations were observed in these embryos, suggesting that a single copy of yap1 is sufficient for a complete phenotypic rescue. Finally, we examined the impact of yap1 and yap1b mutations on the mitotic index, as determined by pH3 staining in stage 20 embryos (Fig. 6M-P). In spite of the known role of Yap proteins in the control of growth and cell proliferation (Huang et al., 2005), we could not detect any significant difference in the number of mitotic cells per area (Fig. 6R). This finding indicates that, at least in the context of early developing embryos, Yap paralogs do not play a significant role in the control of cell proliferation.

Taken together, our results suggest that, although yap1 and yap1b cooperate in regulating cell survival and tissue morphogenesis, yap1 has a dominant role due to the reduced zygotic levels of yap1b. To confirm this dominance, and validate our DamID-seq analysis, we examined using qPCR the relative levels of a number of genes directly targeted by Yap family proteins (including yap1 and yap1b) in wild-type and single mutant embryos at stage 16 (Fig. 7A). We found significantly reduced levels for all targeted genes tested in yap1 mutants, and milder reductions in yap1b^−/− embryos (Fig. 7A). To further confirm our dominance hypothesis, we injected yap1b mRNA into one-cell stage yap1^−/− mutant embryos. Remarkably, the severe malformations exhibited by yap1^−/− mutants were partially (30%) or completely (70%) rescued by yap1b injection (Fig. 7B-E). Thus, despite the divergent architecture of the Yap1b transactivation domain, the protein seems to be transcriptionally active and can replace Yap1 function, at least in a developmental context.

Yap family proteins and Tead factors have been shown to work as a functional tandem, as Yap proteins cannot interact directly with the genome, and the function of Tead regulators depends on the ability of the Yap/Taz C-terminal activation domain to recruit the Pol II complex (Galli et al., 2015; Stein et al., 2015; Vassilev et al., 2001; Zanconato et al., 2015). The bipartite nature of the Yap/Tead transcriptional couple is well illustrated by classical experiments overexpressing Yap in lymphocytic cells, which resulted in a 300-fold induction of Tead activity (Vassilev et al., 2001). The identification in this study of Yap1b, a novel Yap paralog with a transcriptional activation domain very divergent from those previously described for Yap1 (Sudol, 1994) and Taz/Wwtr1 (Kanai et al., 2000), raised questions about its evolutionary origin, chromatin-binding properties and biological function.

Our phylogenetic analyses have outlined an evolutionary sequence for the emergence and radiation of Yap family paralogs in teleosts. Previous studies had shown that Yap emerged in metazoans as a novel Tead partner, and duplicated to give rise to the paralogs Yap and Taz in the vertebrate lineage (Hilman and Gat, 2011). Both paralogs share a common domain architecture and a particular feature, which appears to be a vertebrate-specific acquisition: the capacity of being activated by kinases of the Src family at their C-terminal activation domain (Elbediwy and Thompson, 2018; Sudol, 1994). Here, we show that, during the teleost-specific whole-genome duplication event, the two ancient paralog genes duplicated again to generate four functional copies, yap1, yap1b, taza and tazb. In fact, we show that these four paralogs are present in extant species of basal teleost groups, such us Elopomorpha (eel) and Osteoglosomorpha (elephantfish). Regarding taz paralogs, further evolution in the teleost lineage resulted in the subsequent elimination of one copy before the Otomorpha radiation, and of the second copy in Percomorpha. This finding implies that one in every four vertebrate species (i.e. the whole Percomorpha clade) does not harbor any functional copy of taz in their genomes. The evolutionary history of yap paralogs was quite different. First, yap1 diverged very little from the ancestral gene and became the functionally dominant paralog in teleosts (Kimelman et al., 2017; Miesfeld et al., 2015; Porazinski et al., 2015). In contrast, yap1b was lost in Otomorpha (e.g. zebrafish) and its transcriptional activation domain was substantially modified in euteleosts (e.g. medaka), eliminating the characteristic tyrosine residues that can be phosphorylated by Src. The elimination of these residues should affect a major Hippo-independent regulatory input on the protein, as SFK-mediated phosphorylation has been proved to be essential for the transcriptional activity of Yap/Tead complexes (Calvo et al., 2013; Li et al., 2016). In fact, the modification of Yap1b C-terminal domain is deep enough to raise the question of whether its transactivation properties are impaired. Provided that the Yap1b Tead-binding domain is well conserved and seems functional according to our DamID data, a transcriptionally inactive C-terminal domain could be compatible with a dominant-negative function for Yap1b. Our genetic data strongly argue against this possibility, as yap1b loss of function does not alleviate but rather intensifies the phenotypic consequences of yap1 mutation. The most likely hypothesis therefore is that Yap1b transcriptional activation domain, in contrast to that of other vertebrate paralogs, does not depend on SFK-mediated activation. This was further confirmed by the ability of yap1b to rescue yap1 mutation when injected as mRNA. Yap1b regulatory properties would then be similar to those of invertebrate Yap proteins that are not regulated by kinases of the Src family, as is the case for Yorkie in Drosophila (Elbediwy and Thompson, 2018).

The recruitment of the vertebrate Yap or Taz, or their co-factor Tead, to the chromatin has been previously investigated in genome-wide ChIP-seq studies performed either in mammalian ESCs (Estarás et al., 2017; Lian et al., 2010), cancer cell lines (Galli et al., 2015; Nardone et al., 2017; Stein et al., 2015; Zanconato et al., 2015) or embryonic tissues (Cebola et al., 2015; Lin et al., 2017). Although there are some discrepancies regarding the number of targeted regions (i.e. ranging from 1000 to 7000 peaks), most if not all of these works converged in the identification of a large overlap between Tead and Yap/Taz binding, and succeeded in revealing Tead as the most enriched motif in Yap/Taz peaks. Furthermore, these studies consistently identify a number of genes, including Amotl2, Ctgf, Cyr61 or Ajuba, as direct transcriptional targets. Here, we have used DamID-seq to characterize the binding of Yap paralogs to the genome in zebrafish and medaka gastrulating embryos. This approach allows us to bypass two important limitations of ChIP-seq studies: the dependency on ChIP-grade antibodies (which may not recognize the teleost proteins) and the need for a large number of sample cells. The latter requirement, in particular, may have prevented chromatin occupation analyses at early stages of development in vertebrates. DamID-seq has been used as a standardized protocol for in vivo chromatin profiling in Drosophila (Marshall and Brand, 2017; Marshall et al., 2016; Southall et al., 2013) and, more recently, once toxicity issues were experimentally resolved, in medaka (Gutierrez-Triana et al., 2016) and mammals (Cheetham et al., 2018; Tosti et al., 2018). Several lines of evidence indicate that our DamID-seq approach has succeeded in the identification of bona-fide binding sites for Yap paralogs during early embryogenesis. Our triplicated analyses show high reproducibility, and in all cases we could retrieve the Tead motif: either directly as the most-enriched motif using DREME (i.e. in medaka Yap1 and Yap1b DamID-seq experiments) or after imposing a consensus TEAD motif using AME (i.e. in zebrafish experiments for Yap1 and Taz). In addition, besides TEAD, we could identify a significant enrichment for some known motifs associated with transcriptional regulators reported to interact with Yap proteins, such as those of p63, p73 and Tbx5 (Piccolo et al., 2014). Furthermore, the identified marks, which partially overlap with open chromatin regions, can be found in the vicinity, if not directly in the promoter, of ‘beacon’ genes, such as ctgf, cyr61, amotl2, lats2, ajuba or ccne1, which have been proved to be direct transcriptional targets of Yap/Tead complexes in this and previous studies (Lin et al., 2017; Stein et al., 2015; Zanconato et al., 2015).

Our DamID-seq studies have allowed us to characterize the DNA-binding properties of the newly identified paralog Yap1b. As already anticipated by the sequence conservation of its Tead-binding domain, Yap1b targets a set of chromatin regions that largely overlaps those occupied by Yap1. Moreover, our comparative analyses provide a reference resource to investigate the genetic program controlled by the different co-regulators of the Yap family during early embryogenesis in vertebrates. The comparison between our results and previous ChIP-seq analyses suggests that a substantial part of the transcriptional program controlled by Yap proteins is conserved in vertebrates, regardless of the cell type and developmental stage considered. Further analyses will be required to systematically investigate gene regulatory commonalities versus tissue- and stage-specific adaptations.

One of the main motivations of this work arose from the phenotypic discrepancies observed between yap1 mutants in medaka and zebrafish. The mutation of yap1 in medaka results in strong malformations, which affect the development of many organ rudiments, including the heart, eye, brain and somites at segmentation stages (Porazinski et al., 2015). In contrast, the equivalent yap1 mutation in zebrafish only causes very mild defects. Strong embryonic malformations, similar to those observed in medaka yap1 mutants, are only obtained after the simultaneously mutation of yap1 and taz in zebrafish (Kimelman et al., 2017; Miesfeld et al., 2015). The analysis of these phenotypic discrepancies lead us to the finding that the second paralog present in the medaka genome, and in other euteleosts, was not taz, as previously assumed (Porazinski et al., 2015), but yap1b: a newly identified paralog of the yap family. Comparative analysis of the expression profiles of yap paralogs in zebrafish and medaka provides an explanation of the phenotypic differences observed upon yap1 mutation. Whereas yap1 and taz are expressed at similar levels during gastrulation and somitogenesis in zebrafish, low yap1b levels are detected during this window in medaka. Our observations suggest that paralog compensation by taz mitigates yap1 loss of function in zebrafish but not in medaka, where yap1 plays a more dominating role. Interestingly, the maternal contribution of yap1b is three times more than that of yap1. As Yap has been shown to play a key role in the activation of the early zygotic genome in mice (Yu et al., 2016), this result- may suggest an important role for Yap1b in maternal-to-zygotic transition in medaka. However, loss of maternal yap1b has no apparent phenotypic consequences, as medaka mutants are fertile and produce a normal offspring. Further studies need to be conducted to determine whether blocking the maternal contribution of yap1 and yap1b simultaneously will have any consequence on zygotic activation in medaka embryos.

Despite the absence of overt phenotypes in yap1b mutant embryos, the phenotypic analysis of yap1^−/−;yap1b^−/− double mutants in medaka revealed important roles for Yap1b in the regulation of tissue morphogenesis and cell survival during early embryogenesis. These functions are consistent with our GO-terms analysis for Yap1 and Yap1b target genes, which yielded terms such as ‘negative regulation of apoptotic process’ and ‘embryo development’, and with previously reported roles for Yap paralogs during embryogenesis in teleost (Kimelman et al., 2017; Porazinski et al., 2015). Interestingly, yap1b cooperative role does not depend on SFK-mediated activation. The physiological role of Yap phosphorylation at its transcriptional activation domain is still not well understood, either during embryogenesis or in adult tissues. However, SFK-mediated activation of Yap has been shown to be crucial for liver and intestinal mucosa regeneration (Taniguchi et al., 2015). The particular paralogs composition in euteleosts, with a single SFK-dependent protein, provides now a unique opportunity to further investigate the role of this regulatory input in developing and adult tissues.

The phylogenetic analysis of the Yap family in teleosts uncovered Yap1b, a new paralog protein with an atypical transcriptional activation domain, which cannot be phosphorylated by Src-related kinases. By applying a DamID-seq approach to gastrulating embryos, we could characterize the binding properties not only of Yap1b but also of all the other paralogs present in medaka and zebrafish. These analyses constitute a reference resource to investigate chromatin occupation by Yap family proteins during early embryogenesis in vertebrates, and specifically indicate that medaka Yap1 and Yap1b share similar chromatin-binding properties. Despite the divergent transcriptional activation domain of Yap1b, the phenotypic analysis of yap1 and yap1b medaka mutants reveals functional overlap of both paralogs. Our data show that yap1 and yap1b cooperatively regulate cell death and tissue morphogenesis during early embryogenesis.

The zebrafish (Danio rerio) AB/Tübingen (AB/TU) and medaka (Oryzias latipes) iCab wild-type strains were maintained under previously described experimental conditions (Iwamatsu, 2004; Kimmel et al., 1995). All animal experiments were carried out according to the guidelines of our Institutional Animal Ethics Committee.

Proteins of the Yap family were aligned using MAFFT as implemented in T-Rex (Boc et al., 2012). Maximum likelihood trees were inferred with PHYML software under a WAG +G model using 100 bootstrap replicates (Boc et al., 2012). To retrieve protein sequences and to reconstruct gene arrangement in the vicinity of Yap family members, we used genome annotations available in NCBI (www.ncbi.nlm.nih.gov/genome/), UCSC (genome.ucsc.edu/) and Ensembl (www.ensembl.org/index.html). The following genome assembly versions were used: Anguilla japonica (Japanese eel) Ajaponica_ver_D2; Astyanax mexicanus (Pachón cavefish) 2.0; Clupea harengus (Atlantic herring) ASM96633v1; Danio rerio (Zebrafish) Zv9; Esox lucius (Northern pike) EsoLuc1.0; Gadus morhua (Atlantic cod) gadMor1; Gasterosteus aculeatus (Stickleback) BROAD_S1; Homo sapiens (Human) hg38; Lepisosteus oculatus (Spotted gar) LepOcu1; Mus musculus (Mouse) m38; Oreochromis niloticus (Tilapia) Orenil1.0; Oryzias latipes (Medaka) v1.0; and Paramormyrops kingsleyae (Elephantfish) Pkings_v0.1.

Three sgRNAs were designed against exon 2 of the medaka yap1 gene (ENSORLG00000002708) using the CCTop online tool (Stemmer et al., 2015). Designed primers (Table S4) were hybridized to generate dimers that were inserted into pDR274 plasmids (Addgene, #42250). Resulting vectors were linearized with HindIII and purified by column (GFX PCR DNA and Gel Band purification Kit, G&E). sgRNA molecules without cap and poly(A) tail were generated using the T7 MAXIscript kit (Ambion). To produce Cas9 mRNA, pCS2-nCas9n plasmid (Addgene, #47929) was linearized using NotI, column purified and transcribed (SP6 mMessage mMachine Kit, Ambion). Transcribed sgRNAs were treated with DNAseI to remove the DNA template, precipitated by adding lithium chloride and ethanol, cleaned up (RNA cleanup Kit, Qiagen) and resuspended in nuclease-free water. An ∼5 nl of a mix of Cas9 mRNA at 300 ng/µl and the three sgRNAs produced each were injected at 40 ng/µl into one-cell stage medaka embryos. Founders were identified by genotyping fertilized F1 embryos using a T7 endonuclease assay (Table S4). To identify the specific mutant allele of each F1 adult fish, the affected genomic region was amplified using the Advantage 2 PCR Kit (Clontech), cloned into pSpark vectors (Canvax Biotech) and sequenced. yap1Δ7pb and yap1Δ11pb mutant embryos can be genotyped through PCR and digestion with MspI and TfiI restriction enzymes, respectively.

Two sgRNAs were designed against medaka yap1b (ENSORLG00000005573) exon 2 using CRISPRScan (Moreno-Mateos et al., 2015). For each target, a 52 nt oligo including a T7 promoter followed by the CRISPR-binding site without the PAM sequence (Table S4) were annealed against a universal reverse 80 nt oligo to generate a 117 bp DNA duplex by PCR. Each DNA duplex was purified by column and transcribed to generate sgRNAs using the MAXIscript T7 Transcription kit (Ambion). sgRNAs were injected and mutant alleles identified as described in the previous section. yap1bΔ136pb and yap1bΔ84pb mutants can be genotyped by PCR (Table S4).

To rescue yap1 mutation, yap1b mRNA was synthesized using the SP6 mMessage mMachine kit (Ambion) and injected into one-cell stage medaka at 25 ng/µl.

For iDamID-seq experiments and posterior bioinformatics analysis, we followed the protocol reported for medaka (Gutierrez-Triana et al., 2016) (see also Fig. S5). Zebrafish yap1 (ENSDARG00000068401) and taz/wwtr1 (ENSDARG00000067719), as well as medaka yap1- and yap1b-coding sequences were cloned into pCS2+::cMyc-oDam-FL-GFP-NLS, by replacing the GFP-NLS sequence (Table S4). mRNAs for each DAM-TF cassette were synthesized using the SP6 mMessage mMachine kit (Ambion) and injected into one-cell stage medaka at 10 ng/µl and into zebrafish embryos at 50 ng/µl. Forty embryos were used in each independent DamID replicate. After injection, gDNA was extracted, either from 80%-epiboly zebrafish (8 hpf) embryos or stage 16 medaka embryos, and fragmented using DpnII (NEB). DNA was subsequently dephosphorylated with alkaline phosphatase (NEB) and digested with DpnI to cleave specifically GATC methylated sites. Adapters, produced after hybridizing two oligos (Table S4), were ligated to both ends for each fragment to build up DNA libraries through a 25-cycle PCR amplification step using a phosphorothioate primer (Table S4). Each PCR product was run on a 1% agarose gel to check for the presence of a smear in DpnI-treated samples. Each library was then amplified five times before deep sequencing (Illumina HiSeq 2500 sequencing). Remaining DNA templates were removed with T7 exonuclease before sequencing. Three replicates for each sample were used for the iDamID-seq analysis.

Raw sequencing data was pre-processed using the Cutadapt tool to discard short reads (-m 34 parameter) and the ones corresponding to the adaptors used for deep sequencing. Reads from zebrafish and medaka proteins were mapped against the GRCz10/danRer10 and the NIG/UT MEDAKA1/oryLat2 genome versions of each breed, respectively, using Bowtie2. Duplicated reads were removed using Samtools rmdup. The identification of enriched regions was carried out using the R package iDEAR, available at bitbucket.org/juanlmateo/idear (branch v0.2). Default parameters were applied to the analysis.

We assigned each DamID peak to its nearest gene using bedtools closest. The position of the transcription start site (TSS) of each gene was obtained from ENSEMBL database (Version 89) using the online tool Biomart (www.ensembl.org/biomart). To examine the percentage of peaks overlapping with open chromatin regions, their coordinates were compared with those from ATAC-seq peaks using bedtools intersect. As the zebrafish ATAC-seq data were mapped to an older genome version (danRer7), they had to be converted to the current version using LiftOver.

DREME (Bailey, 2011) was used to identify predefined motifs in the complete set of specific peaks for the four examined proteins (-p parameter) with respect to their corresponding DAM-GFP control peaks (-n parameter). To ensure the statistical significance of this analysis, the set of control peaks were processed to have the same length distribution. To check for the presence of the TEAD motif from the JASPAR Database (MA0090.1) we used AME (McLeay and Bailey, 2010), applying a Mann–Whitney non-parametric test.

FIMO (Grant et al., 2011) was used to scan the peak sequences for motifs either found in our enrichment analysis or previously associated with YAP/TAZ (Table S2). The motif hits found by FIMO were filtered to avoid motif overlap, discarding the motif with lowest significance if the overlap was for more than three bases. After this filtering, a PCA analysis was performed to identify the motifs in the four peaks sets that contribute the most. Next, for each pair of selected motifs, the proportion of peaks with at least one match for both motifs was calculated and this value was used a proxy for the degree of cooperation of both factors. Cytoscape (Shannon et al., 2003) was used to build networks representing the level of interaction between factors.

To obtain lists of genes associated with Yap1, Taz, or Tead binding in mammalian cells, we used the coordinates of the peaks available from different publications (Lian et al., 2010, Tables S1 and S2; Estarás et al., 2017, Table S2; Zanconato et al., 2015, YAP.narrowPeak and TAZ.narrowPeak). These coordinates were first converted to the assembly GRGh38 or GRGm38 using the Assembly Converter tool from www.ensembl.org. Using Biomart and the ENSEMBL database (Version 89), a protein-coding gene was assigned to each peak that was either overlapping its gene body or the closest to its TSS, within a limit of 50 kb for medaka, 100 kb for zebrafish or 200 kb for mouse or human, to account for the differences in genome length. The hypergeometric test was used to compute the significance of the overlap between the medaka- or zebrafish-associated genes and the genes associated with the ChIP-seq peaks.

Gene ontology analyses were performed using PANTHER (Mi et al., 2013). Only the GO terms identified with a False Discovery Rate lower than 0.05 using the PANTHER GO-Slim algorithm were taken into account.

To check yap1, taz (wwtr1) and yap1b expression levels during pre-MBT, epiboly and phylotypic stages in teleosts, mRNA from medaka embryos at stages 7, 16 and 24, and from zebrafish embryos of 3, 8 and 24 hpf was obtained using TRIzol (Ambion). cDNA samples were obtained through retro-transcription using the iScript cDNA Synthesis kit (BioRad) and their concentration was measured in a Qubit fluorometer.

cDNA from medaka embryos at stage 24 was used to amplify part of the coding sequence of medaka yap1 and yap1b genes. PCR products were cloned into pSC-A-amp/kan Strataclone plasmids (Agilent) to generate probes for whole-mount in situ hybridization experiments (Table S4). Probes were synthesized using digoxigenin-11-UTP nucleotides (Roche) and the T3 or the T7 polymerase (Roche) depending on the insert orientation. In situ hybridization was performed on fixed embryos at stage 15-16 following a previous protocol (Thisse and Thisse, 2008). Embryos were permeabilized using proteinase K for 4 min. Transcribed probes were used at a final concentration of 1 ng/µl. Embryos were washed with 0.05× SSC/PBS in decreasing salt concentrations.

The expression levels of medaka yap1 and yap1b, and zebrafish genes yap1 and taz was quantified through RT-qPCR (CFX96 Touch Real-Time PCR Detection System) and normalized to the expression level of the housekeeping gene ef1a (Table S4). qPCR results were confirmed using the QX200 Droplet Digital PCR system (BioRad). Duplicated PCR reactions for each one of the tested genes were prepared using 15 ng of cDNA per reaction and following the manufacturer specifications. Every PCR reaction was atomized into a set of ∼22,000 1 nl drops using the QX200 Droplet Generator system. These droplets were transferred to 96-well plates (#12001925, Bio-Rad). After the PCR, the relative amount of every amplified product was quantified using a QX200 Droplet Reader examining the fluorescence rate of each droplet. Reads were normalized to the levels of the housekeeping gene ef1a.

For qPCR experiments in mutant embryos, single embryos were homogenized in 1.0 ml of TRIzol (Invitrogen). DNA and RNA were extracted at the same time following the manufacturer's instructions. DNA was used for embryo genotyping. Total RNA was retro-transcribed with SuperScript IV VILO Master Mix (Thermo Fisher Scientific) and qPCR reactions were performed in triplicate using SsoAdvanced in a CFX98 real-time C1000 thermal cycler (Bio-Rad). Results were normalized with ef1a (for primer sequences, see Table S4).

For the detection of apoptotic and mitotic cells, embryos collected from yap1^+/−;yap1b^+/− adult fishes were fixed at stage 20 (5 somites) using PFA 4%. For YAP nuclear localization analysis, wild-type embryos were injected with the DamID-Yap1 and Yap1b cassette mRNA at the one-cell stage as described in the ‘iDamID-seq’ section, and fixed at stage 10 using 4% PFA. Fixed embryos were dechorionated with forceps, extensively washed with PBS-Tween 0.2% (PBT), treated with blocking solution (prepared with 10% fetal bovine serum) at room temperature for 2 h and incubated overnight at 4°C with the corresponding primary antibody diluted 1:500. Embryos were incubated with purified rabbit anti-active caspase-3 antibodies (BD Biosciences, 559565), rabbit anti-phospho-histone H3 antibodies (Merck Millipore, 06-570) or rabbit anti-c-Myc antibodies (Sigma-Aldrich, PLA0001). In the case of the c-Myc antibodies, 1% DMSO and 0.25% BSA were added to the blocking solution. Processed embryos were repeatedly washed with PBT and incubated overnight at 4°C in the dark with the Alexa Fluor 555 goat anti-rabbit antibody (diluted 1:500 (Thermo Fisher, A21429). Nuclei counterstaining was performed using DAPI (Sigma-Aldrich) diluted 1:1000 and phalloidin using phalloidin-Alexa488 (Invitrogen) diluted 1:20 and incubated in PBT overnight at 4°C.

Selected embryos were mounted in 35 mm FluoroDish plates (WPI, FD3510-100) using 1% low melting agarose. These embryos were examined using confocal microscopy (Zeiss LSM 880) with 25× or 40× multi-immersion objectives. All pictures were processed using FIJI/ImageJ (Version 1.50i). For each image and channel (red and blue), a maximum projection was generated and the background noise was completely minimized (Subtract Background set at 200). For quantification of apoptotic and mitotic cells, masks were applied for both channels. The mask generated for the red channel was segmented using the Watershed algorithm. Only the regions marked with the primary antibody that also corresponded to nuclei with a 6-200 µm² area were considered to avoid debris. Apoptotic and pH3-positive cells were counted on the embryo surface and extrapolated to a section of 1 mm². Morphometric analysis of wild-type and mutant embryo structures at stage 20 was performed also using FIJI/ImageJ (Version 1.50i).

Quantitative data were evaluated using Prism 6.0 (GraphPad). The statistical analyses applied, n values and significance levels are indicated in figure legends.

We thank Nacho Maeso, Joaquin Letelier, Cristian Undurraga, Juan Tena and Rocío Polvillo for their critical input and/or their excellent technical assistance. We are grateful to Bryan Link and Joel Miesfeld for sharing zebrafish yap and taz mutants.