To determine the spatiotemporal gene expression of dact1 and dact2 during embryogenesis we performed wholemount RNA in situ hybridization (ISH) across key time points (Figure 1). Given that the described craniofacial phenotypes of the dact2 morphant and the wnt11f2 mutant are similar (Heisenberg and Nüsslein-Volhard, 1997; Waxman et al., 2004), we also performed wnt11f2 ISH to compare to dact1 and dact2 expression patterns.

During gastrulation at 8 hours post-fertilization (hpf; 75% epiboly), some regions of dact1 and dact2 gene expression were shared and some areas are distinct to each dact gene (Figure 1A). Further, dact gene expression was distinct from wnt11f2 in that wnt11f2 expression was not detected in the presumptive dorsal mesoderm. Transcripts of dact1, dact2, and wnt11f2 were all detected in the blastoderm margin, as previously described (Makita et al., 1998; Heisenberg et al., 2000; Gillhouse et al., 2004). Transcripts of dact2, and to a lesser extent dact1, were also detected in the prechordal plate and chordamesoderm (Figure 1A). Additionally, dact2 gene expression was concentrated in the shield and presumptive organizer or Nieuwkoop center along with wnt11f2. This finding is consistent with previously described expression patterns in zebrafish and supports a role for dact1 and dact2 in mesoderm induction and dact2 in embryo dorsalization (Thisse et al., 2001; Gillhouse et al., 2004; Muyskens and Kimmel, 2007; Oteiza et al., 2010). At the end of gastrulation and during somitogenesis the differences in the domains of dact1 and dact2 gene expressions became more distinct (Figure 1B, C). At tailbud stage, dact1 transcripts were detected in the neuroectoderm and the posterior paraxial mesoderm, whereas dact2 transcripts were detected in the anterior neural plate, notochord, and tailbud. Anterior notochord and tailbud expression overlapped with wnt11f2 gene expression (Makita et al., 1998; Heisenberg et al., 2000). The expression of dact2 was unique in that its expression demarcated the anterior border of the neural plate. As dact2 morphants exhibited a craniofacial defect with medially displaced eyes and midfacial hypoplasia (Waxman et al., 2004), we examined dact1 and dact2 expression in the orofacial tissues. At 24 hpf we found some overlap but predominantly distinct expression patterns of dact1 and dact2 with dact1 being more highly expressed in the pharyngeal arches and dact2 being expressed in the midbrain/hindbrain boundary. Both dact1 and dact2 appeared to be expressed in the developing oral cavity. At 48 hpf dact1 expression is consistent with expression in the developing craniofacial cartilage elements, while dact2 expression appears within the developing mouth. The distinct cellular expression profiles of dact1 and dact2 were more clear in histological sections through the craniofacial region at 72 hpf. Utilizing RNAscope ISH, we found that dact2 and the epithelial gene irf6 were co-expressed in the surface and oral epithelium that surround the cartilaginous structures (Figure 1F). This is in contrast to dact1 which was expressed in the developing cartilage of the anterior neurocranium (ANC)/EP and palatoquadrate of the zebrafish larvae (Figure 1F).

We examined the overall expression patterns of dact1, dact2, gpc4, and wnt11f2 using Daniocell single-cell sequencing data (Farrell et al., 2018). In general, we found dact1 spatiotemporal gene expression to be more similar to gpc4 while dact2 gene expression was more similar to wnt11f2 (Figure 1—figure supplement 1). These results of shared but also distinct domains of spatiotemporal gene expression of dact1 and dact2 suggest that the dact paralogs may have some overlapping developmental functions while other roles are paralog-specific. dact1 and dact2 contribute to axis extension and dact1/2 compound mutants exhibit a CE defect dact1 and dact2 are known to interact with disheveled and regulate noncanonical Wnt signaling (Gloy et al., 2002; Waxman et al., 2004; Gao et al., 2008; Wen et al., 2010; Ma et al., 2015; Lee et al., 2018) and we have previously described the craniofacial anomalies of several zebrafish Wnt mutants (Dougherty et al., 2012; Kamel et al., 2013; Rochard et al., 2016; Ling et al., 2017; Alhazmi et al., 2021). Previous work investigated the effect of dact1 and dact2 disruption during zebrafish embryogenesis using morpholinos and reported morphant phenotypes in embryonic axis extension and eye fusion (Waxman et al., 2004). However, the limitations associated with morpholino-induced gene disruption (Kok et al., 2015; Morcos et al., 2015; Rossi et al., 2015) and the fact that craniofacial morphogenesis was not detailed for the dact1 and dact2 morphants, warranted the generation of mutant germline alleles (Figure 2—figure supplement 1). We created a dact1 mutant allele (22 bp deletion, hereafter dact1−/−) and a dact2 mutant allele (7 bp deletion, hereafter dact2−/−), both resulting in a premature stop codon and presumed protein truncation (Figure 2—figure supplement 1A, B). Gene expression of dact1 and dact2 was measured in pooled dact1−/−, dact2−/−, and dact1/2−/− embryos (Figure 2—figure supplement 1C). We found a decrease in dact1 mRNA and an increase in dact2 mRNA levels in the respective CRISPR single mutants. We hypothesize that dact2 mRNA levels are maintained or elevated in the dact2−/− mutant due to the relative 3′ position of the deletion. In the dact2−/− embryos we found a slight increase in dact1 mRNA levels, suggesting a possible compensatory effect of dact2 disruption. The specificity of the gene disruption was demonstrated by phenotypic rescue of the rod-like EP with the injection of dact1 or dact2 mRNA. Injection of dact1 mRNA or dact2 mRNA or in combination decreased the percentage of rod-like EP phenotype from near the expected 25% (35% actual) to 2–7% (Figure 2—figure supplement 1D and E).

Analysis of compound dact1 and dact2 heterozygote and homozygote alleles during late gastrulation and early segmentation time points identified embryonic axis extension anomalies (Figure 2A, B). dact1−/− or dact2−/− homozygotes develop to be phenotypically normal and viable. However, at 12 hpf, dact2−/− single mutants have a significantly shorter body axis relative to wildtype. In contrast, body length shortening phenotype was not observed in dact1−/− homozygotes. Compound heterozygotes of dact1+/-; dact2+/- also developed normally but exhibited shorter body axis relative to wildtype. The most significant axis shortening occurred in dact1−/−; dact2−/− double homozygotes with a less severe truncation phenotype in the compound heterozygote dact1+/-; dact2−/− (Figure 2A, B). Interestingly, these changes in body axis extension do not preclude the compound heterozygous larvae from reaching adulthood, except in the dact1−/−; dact2−/− double homozygotes which did not survive from larval to juvenile stages.

Body axis truncation has been attributed to impaired CE during gastrulation (Tada and Heisenberg, 2012). To delineate CE hallmarks in the dact1−/−;dact2−/− mutants, we performed wholemount RNA ISH detecting genes that are expressed in key domains along the body axis. At bud stage, dact1−/−; dact2−/− embryos demonstrate bifurcated expression of pax2a and decreased anterior extension of gsc expression, suggesting impaired midline convergence and anterior extension of the mesoderm (Figure 2C). At the 1–2 somite stage, zic1, pax2a, and tbx6 are expressed in neural plate, prospective midbrain and the tailbud, respectively, in both the wildtype and dact1−/−; dact2−/− embryos. However, the spacing of these genes clearly revealed the shortening of the antero-posterior body axis in the dact1−/−; dact2−/− embryos. Midline convergence is decreased and the anterior border of the neural plate (marked by zic1 expression) was narrower in the dact1−/−; dact2−/− embryos (Figure 2D). At the 10-somite stage (ss), dact1−/−; dact2−/− embryos demonstrated decreased spacing between ctslb1 and pax2a gene expression, suggesting impaired lengthening of the anterior portion of the embryo. Detection of muscle marker myo1d in the dact1−/−; dact2−/− embryos delineated impaired posterior lengthening as well as reduced somitogenesis, evidenced by the decreased number of somites (Figure 2E). These data point to impaired CE of the mesoderm in dact1/2−/− double mutants, which resulted in a shorter body axis. The aberrant CE and axis extension in the dact1/2−/− phenotypes were similar to findings in other Wnt mutants, such as slb and kyn (Heisenberg et al., 2000; Topczewski et al., 2001) in that the body axis is truncated upon segmentation.

Given the defective converge phenotype and shortened axis in the dact mutants during gastrulation, we examined the fish at 4 dpf for axis defects and for evidence of defective morphogenesis in the craniofacial cartilages. Craniofacial morphology is an excellent model for studying CE morphogenesis as many craniofacial cartilage elements develop through this cellular mechanism (Kamel et al., 2013; Mork and Crump, 2015; Sisson et al., 2015; Rochard et al., 2016). No craniofacial phenotype was observed in dact1 or dact2 single mutants (data not shown). However, in-crossing to generate dact1/2−/− compound homozygotes resulted in dramatic craniofacial malformation (Figure 3). Specificity of this phenotype to dact1/2 was confirmed via rescue with dact1 or dact2 mRNA injection (Figure 2—figure supplement 1D, E). The dact1/2−/− mutant embryos exhibited fully penetrant midfacial hypoplasia (Figure 3A); however, the degree of eye field convergence in the midline varied between individuals. The forebrain protruded dorsally and the mouth opening and ventral cartilage structures were displaced ventrally (Figure 3A). Alcian blue staining of cartilage elements revealed severe narrowing of the EP into a rod-like structure in 100% of double mutants, while the ventral cartilage elements were largely unaffected (Figure 3B). Notably the trabeculae extending posteriorly from the EP and the rest of the posterior neurocranium exhibit wildtype morphology in dact1/2. This dact1/2−/− double mutant phenotype is highly similar to that described for wnt11f2 (slb) mutants, a key regulator of noncanonical Wnt signaling and CE (Kimmel et al., 2001).

As axis lengthening was found to be affected by loss of dact1 and dact2 (Figure 2A) we measured body length in 5 dpf dact1/2 compound mutants. Using the length of the vertebral spine as a measure of body length we found a trend (p = 0.06) toward an effect of dact1 and dact2 on shortening of the body length. Similar to axis length during gastrulation/segmentation, the shortening was most pronounced in dact1/2−/− double homozygous mutants versus wildtype clutch-mates (Figure 3C, Figure 2—figure supplement 2A, B).

As wnt11f2 signals via disheveled and since dact proteins are known to interact with disheveled (Wong et al., 2003; Zhang et al., 2006; Kivimäe et al., 2011), it is suspected that dact has a role in wnt11f2 signaling. Combinatorial gene disruption with morpholinos showed that dact2 morpholino exasperated the wnt11 morpholino midfacial/eye fusion defect (Waxman et al., 2004). We hypothesized that the shared phenotypes between wnt11f2 and dact1/2 mutants point to these genes acting in the same signaling pathway. To test for genetic epistasis between wnt11f2, dact1, and dact2 genes we generated wnt11f2/dact1/2−/− triple homozygous mutants. If wnt11f2 and dact1/2 had independent developmental requirements, the wnt11f2/dact1/2−/− mutant may exhibit a phenotype distinct from wnt11f2−/− or dact1/2−/− mutants. We found that the wnt11f2/dact1/2−/− triple homozygous mutant phenotype of the linear rod-like EP was the same as the wnt11f2−/− mutant or dact1/2−/− double mutant, without exhibiting additional or neo-phenotypes in the craniofacial cartilages or body axis (Figure 3D, E). This result supports dact1 and dact2 acting downstream of wnt11f2 signaling during ANC morphogenesis, where loss of dact1/2 function recapitulates a loss of wnt11f2 signaling.

The EP forms from the convergence of a central frontal prominence-derived structure with bilateral maxillary prominence-derived elements (Wada et al., 2005; Swartz et al., 2011; Dougherty et al., 2012; Mork and Crump, 2015; Rochard et al., 2016).The stereotypic convergent migration of cranial neural crest cells (NCCs) and their derivatives presents an excellent model to examine CE movements and their effects on tissue morphogenesis. The zebrafish EP is formed from the joining of a midline frontal prominence derived from the anteromost cranial NCC population that migrate over the eyes and turn caudally, to join paired lateral maxillary prominences derived from the second stream of cranial NCC population that migrate rostrally (Kimmel et al., 2001; Wada et al., 2005; Schilling and Le Pabic, 2009; Dougherty et al., 2012; Mork and Crump, 2015). The EP that forms is a planar fan-shaped structure where we and others have shown that the morphology is governed by Wnt signaling (Kimmel et al., 2001; Rochard et al., 2016).

Given the rod-like EP we observed in the dact1/2−/− double mutants, we hypothesized that the dysmorphology could be due to aberrant migration of the anteromost midline stream of cranial NCCs resulting in fusion of the lateral maxillary components. Conversely, an abrogated contribution from the second paired stream of maxillary NCCs could lead to an EP composed entirely of the medial component. To distinguish between these possibilities, we carried out lineage tracing of the cranial NCC populations in wildtype and dact1/2 mutants. The dact1/2 compound mutants were bred onto a sox10:kaede transgenic background, where we and others have shown that the sox10 reporter is a reliable driver of cranial neural crest labeling (Wada et al., 2005; Dutton et al., 2008; Swartz et al., 2011; Dougherty et al., 2012; Kague et al., 2012; Mork and Crump, 2015). Cranial NCC populations in wildtype and dact1/2 mutants were targeted at 19 hpf to photoconvert Kaede reporter protein in either the anterior cranial NCCs that contribute to the frontal prominence, or the second stream of NCCs that contribute to the maxillary prominence, where the labeled cells were followed longitudinally over 4.5 days of development (Figure 4). We found that the anterior NCCs of wildtype embryos migrated antero-dorsally to the eye and populated the medial EP. To our surprise, the anterior cranial NCC also migrated to contribute to the median element of the rod-like EP, suggesting the complex anterior then caudal migration of the anterior NCC is not disrupted by dact1/2 mutation (Figure 4A, arrows). This finding is in contrast to lineage tracing in another midline mutant with a similarly shaped rod-like EP, the syu (sonic hedgehog null) mutant, where the anterior NCCs failed to populate the ANC (Wada et al., 2005).

Next, the second stream of NCC population that contribute to the maxillary prominence was labeled, where they migrate and contribute to the lateral element of the EP as expected in the wildtype (Figure 4B). When the second stream of cranial NCCs were labeled and followed in the dact1/2 mutants, the cells were found to migrate normally up to 36 hpf, but did not ultimately populate the EP in the mutant (arrows). These results suggest that NCC migration itself is not regulated by dact1/2 but that loss of dact1/2 hinders the second stream of NCCs’ ability to populate the ANC by an alternative means. Further, we have found that a rod-like EP can be formed from two different NCC origins, where in the dact1/2 mutants the EP is contributed by the anteromost frontonasal NCCs, in contrast to the similar rod-shaped EP of the syu mutants that is formed from the more posterior stream of maxillary NCCs (Wada et al., 2005).

Given the role of Dact/dapper as modifiers of Wnt signaling, we hypothesized that genetic interaction of dact1/2 with wls and gpc4 will modify facial morphology. Gpc4 is a glycoprotein that binds Wnt ligands and modulates Wnt signaling. gpc4 zebrafish mutants have impaired CE which leads to a shortened body axis (Topczewski et al., 2001). Wls is a posttranslational modifier of Wnt ligands which promotes their secretion (Bänziger et al., 2006; Bartscherer et al., 2006). We previously described that these components of the Wnt/PCP pathway (gpc4 receptor, wls intracellular ligand chaperon, and Wnt ligands wnt9a and wnt5b) are required for craniofacial morphogenesis, where each gene affects particular morphologic aspects of chondrocytes arrangement in the cardinal axis of the ANC and Meckel’s cartilage (Rochard et al., 2016; Ling et al., 2017). Using the EP as a morphologic readout, we examined the genetic interaction of dact1 and dact2 with wls and gpc4. Compound mutants of dact1, dact2, gpc4, or wls were generated by breeding the single alleles. Compared to wildtype ANC morphology, abrogation of gpc4 led to increased width in the transverse axis, but shorter in the antero-posterior axis (Rochard et al., 2016). Disruption of wls leads to ANC morphology that is also wider in the transverse dimension, but to a lesser degree than observed in gpc4. Additionally, in the wls mutant, chondrocytes stack in greater layers in the sagittal axis (Rochard et al., 2016).

Disruption of gpc4 or wls in addition to dact1/2 generated EP morphology that contained phenotypic attributes from each single mutant, so that the resultant ANC morphology represented a novel ANC form. The EP of a triple homozygous gpc4/dact1/2−/− mutant was triangular, wider in the transverse axis and shorter in the antero-posterior axis compared to the rod-like ANC observed in the dact1/2−/− double mutant (Figure 5A, B). Similarly, the ANC of a triple homozygous wls/dact1/2−/− mutant was in the shape of a rod, shorter in the antero-posterior axis and thicker in the sagittal axis compared to the dact1/2−/− double mutant, reflecting attributes of the wls mutant (Figure 5C, D). In addition to the EP phenotypes, the triple homozygous gpc4/dact1/2−/− mutant also had a short body axis and truncated tail similar but more severe than the gpc4 mutant (Figure 5A). Since compound disruption of dact1, dact2, and gpc4 or wls resulted in a new phenotype we conclude that these genes function in different components of Wnt signaling during craniofacial development.

As we analyzed the subsequent genotypes of our dact1/dact2/gpc4 triple heterozygote in-cross we gleaned more functional information about dact1 and dact2. We found that dact1 heterozygosity in the context of dact2−/−; gpc4−/− was sufficient to replicate the triple dact1/dact2/gpc4 homozygous phenotype (Figure 5E). In contrast, dact2 heterozygosity in the context of dact1−/−; gpc4−/− double mutant produced ANC in the opposite phenotypic spectrum of ANC morphology, appearing similar to the gpc4−/− mutant phenotype (Figure 5E). These results show that dact1 and dact2 do not have redundant function during craniofacial morphogenesis, and that dact2 function is more indispensable than dact1. These results also suggest that dact1 and gpc4 may have overlapping roles in craniofacial development.

Our analyses of axis extension and the hallmarks of a CE defect (namely decreased length and increased width between early tissues) demonstrate that dact1 and/or dact2 are required for CE and anterior–posterior axis lengthening during gastrulation (Figure 3). An axis lengthening and CE defect has also been described in gpc4 (aka kny) mutants (Topczewski et al., 2001). We also observe a defect in axis lengthening in gpc4−/− in our hands (representative image Figure 6A) that is grossly similar to the dact1/2−/− mutants. Interestingly, the midfacial hypoplasia of the wnt11f2 (slb) mutant has been attributed to a defect in axis extension and anterior neural plate patterning (Heisenberg and Nüsslein-Volhard, 1997), whereas defective axis extension does not lead to midfacial hypoplasia in the gpc4−/− mutant. Therefore, we hypothesized that by comparing and contrasting the gene expression changes in dact1/2 versus gpc4 mutants during axis extension we could identify cell programs specifically responsible for the anterior axis defect and subsequent midfacial hypoplasia. We performed single-cell transcriptional analysis to compare dact1/2 mutants, gpc4 mutants, and wildtype embryos during the segmentation stage. Single-cell encapsulation and barcoded cDNA libraries were prepared from individual dissociated 4 ss wildtype, dact1/2−/− compound mutant and gpc4−/− mutant embryos using the 10X Genomics Chromium platform and Illumina next-generation sequencing. Genotyping of the embryos was not possible but quality control analysis by considering the top 2000 most variable genes across the dataset showed good clustering by genotype, indicating the reproducibility of individuals in each group. Twenty clusters were identified using Louvain clustering and identity was assigned by reviewing cluster-specific markers in light of published expression data (Farrell et al., 2018; Farnsworth et al., 2020; Bradford et al., 2022; Figure 6B, C). Qualitatively, we did not observe any significant difference in cluster abundance between genotype groups (Figure 6—figure supplement 1). We found that dact1, dact2, and gpc4 were detected at various levels across clusters, though dact1 expression was lower than dact2 (Figure 6D), consistent with what we observed in RNA wholemount ISH analysis (Figure 1).

To assess the relative differences in gene expression between genotype groups, we merged clusters into broader cell lineages: ectoderm, axial mesoderm, and paraxial mesoderm (Figure 7A). We focused on these cell types because they contribute significantly to CE processes and axis establishment. For each of these cell lineages, we performed independent pseudo-bulk differential expression analyses (DEA) of wildtype versus dact1/2−/− mutant and wildtype vs. gpc4−/− mutant (Figure 7A). In all three cases, we found differentially expressed genes (DEGs) that were commonly in dact1−/−;dact2−/− and gpc4−/− mutant relative to wildtype (Figure 7B). To address the hypothesis that dact1 and dact2 regulate molecular pathways distinct from those regulated by gpc4 we also identified genes that were differentially expressed only in dact1/2−/− mutants or only in gpc4−/− mutants (Figure 7B). Functional analysis of these DEGs found unique enrichment of intermediate filament genes in gpc4−/− whereas dact1/2−/− mutants had enrichment for pathways associated with proteolysis (Figure 7C, Figure 7—figure supplement 1). Enrichment for pathways associated with calcium-binding were found in both gpc4−/− and dact1/2−/−, although the specific DEGs were distinct (Figure 7—figure supplement 1). We performed functional analyses specifically for genes that were differentially expressed in dact1/2−/− mutants, but not in gpc4−/− mutants, and found enrichment in pathways associated with proteolysis (Figure 7C) suggesting a novel role for Dact in embryogenesis.

Interrogation of dact1/2−/− mutant-specific DEGs found that the calcium-dependent cysteine protease calpain 8 (capn8) was significantly overexpressed in dact1/2−/− mutants in paraxial mesoderm (103-fold), axial mesoderm (33-fold), and in ectoderm (3-fold; Figure 7A). We also found that loss of dact1/2 causes significant changes to capn8 expression pattern (Figure 8A). Whereas capn8 gene expression is principally restricted to the epidermis of wildtype embryos, loss of dact1/2 leads to significant expansion of ectopic capn8 gene expression in broader cell types such as in mesodermal tissues (Figure 8A). We corroborated this finding with wholemount RNA ISH for capn8 expression in wildtype versus dact1/2−/− 12 hpf embryos (Figure 8B). The expression of smad1 was found to be decreased uniquely in the ectoderm of dact1/2−/− embryos relative to wildtype (Figure 7A), however this finding was not investigated further.

Capn8 is considered a ‘classical’ calpain, with domain homology similar to Capn1 and Capn2 (Macqueen and Wilcox, 2014). In adult human and mouse tissue, Capn8 expression is largely restricted to the gastrointestinal tract (Sorimachi et al., 1993; Macqueen and Wilcox, 2014), however embryonic expression in mammals has not been characterized. Proteolytic targets of Capn8 have not been identified, however, other classical calpains have been implicated in Wnt and cell–cell/ECM signaling (Konze et al., 2014), including in Wnt/Ca⁺² regulation of CE in Xenopus (Zanardelli et al., 2013). To determine whether the dact1/2−/− mutant craniofacial phenotype could be attributed to capn8 overexpression, we performed injection of capn8 or gfp control mRNA into 1 cell-stage zebrafish embryos. In wildtype zebrafish, exogenous capn8 mRNA caused the distinct dact1/2−/− craniofacial phenotype including a rod-like ANC at a very low frequency (1 in 142 injected embryos). This craniofacial phenotype was not observed in wildtype larvae, or when wildtype embryos were injected with an equal concentration of gfp mRNA (0 in 192 injected embryos) (data not shown). When mRNA was injected into 1 cell-stage embryos generated from dact1/2+/- interbreeding, capn8 caused a significant increase in the number of larvae with the mutant craniofacial phenotype when on a dact1/2+/- genetic background (Figure 8D, 0.0% vs. 7.5%). We did not find an effect of exogenous capn8 on any other genotype, including dact1−/−,dact2+/- which we suspect to be due to the smaller number of those individuals in our experimental population. These findings suggest a new contribution of capn8 to embryonic development as well as anterior neural plate patterning and craniofacial development. Further, the regulation of capn8 by dact may be required for normal embryogenesis and craniofacial morphogenesis.

All animal husbandry and experiments were performed in accordance with and approval from the Massachusetts General Hospital Institutional Animal Care and Use Committee (protocol number 2010N000106) and the Children’s Hospital of Philadelphia Animal Care and Use Committee (protocol number IAC22001475). Zebrafish (Danio rerio) embryos and adults were maintained in accordance with institutional protocols. Embryos were raised at 28.5°C in E3 medium (Carroll et al., 2020) and staged visually and according to standardized developmental time points (Westerfield, 1993). All zebrafish lines used for experiments and gene editing were generated from the Tubingen or AB strain. The wnt11f2 mutant line and gpc4−/− mutant line were obtained from Zebrafish International Resource Center (wnt11f2^tx226/+ and gpc4^hi1688Tg/+, respectively). The wls−/− mutant line was originally gifted to the lab and independently generated, as previously described (Rochard et al., 2016). The sox10:kaede transgenic line was previously generated and described by our lab (Dougherty et al., 2012).

CRISPR sgRNA guides were designed using computational programs ZiFiT Targeter v4.2 (https://zifit.partners.org/ZiFit) (Sander et al., 2007), crispr.mit.edu (https://zlab.bio/guide-design-resouces) (Ran et al., 2013), and ChopChop (https://chopchop.cbu.uib.no) (Montague et al., 2014) with traditional sequence constraints. Guides were chosen that were predicted to give high efficiency and specificity. Guides best meeting these parameters were selected in exon 2 of dact1 and exon 4 of dact2. No suitable gRNA with sufficient efficiency were identified for dact2 5′ of exon 4 and the resulting phenotype was reassuring compared to previous morpholino published results. Guides for dact1 and dact2 and Cas9 protein were prepared and microinjected into 1 cell-stage zebrafish embryos and founders were identified as previously described (Carroll et al., 2020). Primers flanking the sgRNA guide site were designed for genotyping and fragment analysis was performed on genomic DNA to detect base pair insertion/deletion. Sanger sequencing was performed to verify targeted gene mutation and confirm the inclusion of a premature stop codon. dact1 forward primer: TACAGAAGCTGCTGAAGTACCG, dact1 reverse primer: CCCTCTCTCAAAGTGTTTTGGT, dact2 forward primer: TGAAGAGCTCCACTCCCCTGT, dact2 reverse primer: GCAGTTGAGGTCCATTCAGC.

Pooled wildtype, dact1−/−, dact2−/−, and dact1/2−/− fish were collected and RNA extractions were performed using RNeasy Mini Kit (QIAGEN). cDNA was generated using High Capacity cDNA Reverse Transcription Kit (ThermoFisher). Quantitative PCR was performed using dact1 (Dr03152516_m1) and dact2 (Dr03426298_s1) TaqMan Gene Expression Assays. Expression was normalized to 18S rRNA expression (Hs03003631_g1). TaqMan Fast Advanced master mix (Thermo Fisher) and a StepOnePlus Real-Time PCR system (Applied Biosystems) were used to measure relative mRNA levels, which were calculated using the ddCT method.

Template DNA for in vivo mRNA transcription was generated by PCR amplification of the gene of interest from a zebrafish embryo cDNA library and cloning into pCS2+8 destination plasmid. mRNA for injection was generated using an in vitro transcription kit (Invitrogen mMessage mMachine). One-cell-stage zebrafish embryos were injected with 2 nl of mRNA in solution. To test genetic knockout specificity, 150 or 300 pg or dact1, dact2, or dact1 and dact2 mRNA was injected. For capn8 overexpression analysis 200 pg or GFP or capn8 mRNA was injected. Following phenotyping analysis, genotypes were determined by fragment analysis of dact1 and dact2 genotyping PCR products.

Wholemount ISH was performed as previously described (Carroll et al., 2020). Zebrafish embryonic cDNA was used as a template to generate riboprobes. Primers were designed to PCR amplify the specific riboprobe sequence with a T7 promoter sequence linked to the reverse primer. In vitro transcription was performed using a T7 polymerase and DIG-labeled nucleotides (Roche). Probe primer sequences are; dact1 fwd; 5′AGCGCGATTCTCAGATGCAC3′, rev 5′gaaatTAATACGACTCACTATAggCCTGCTCGGGTTTCTGTTCA3′, dact2 fwd: 5′CAGTCGCATAGCGGATCTCAT3′, rev: 5′gaaatTAATACGACTCACTATAggGTGGACTGGGGTAACGGTAA3′, wnt11f2 fwd: 5′TCCGTGGTGTATCTTGACCG3′, rev: 5′gaaatTAATACGACTCACTATAggCTTGGTGGCCGACAGGTATT3′, pax2a fwd: 5′CCAAACCAAAAGTGGCGACG3′, rev: 5′gaaatTAATACGACTCACTATAggGTTGCTGAACCGCCAAGC3′, gsc fwd: 5′CCAGCGCCGAACTTACAATC3′, rev: 5′gaaatTAATACGACTCACTATAggTCTTCAGCTACAGCCCATTCC3′, zic1 fwd: 5′TAGGGGATCGGAGTTTGCCT3′, rev: 5′gaaatTAATACGACTCACTATAggTTCGTCAGCTGCTCTGGTTC3′, tbx6 fwd: 5′ACAGAGATCGAGATGTGCCG3′, rev: 5′gaaatTAATACGACTCACTATAggTGGAAGGGCGGTGTTCATAA3′, myo1d fwd: 5′TCTACGACGACCCTTGCTTC3′, rev: 5′gaaatTAATACGACTCACTATAggGGTCGGATTCGCCTTTTTCTG3′, ctslb1 fwd: 5′AGACCGCCTCTATGTTCGGA3′, rev: 5′gaaatTAATACGACTCACTATAggAGCGACATTAAAACGGGGGT3′, capn8 fwd: 5′AAGGGCTGGGGACAAATGAG3′, rev: 5′gaaatTAATACGACTCACTATAggCACTAGGAATGTGCAGCCGT3′.

RNAscope was performed on sectioned zebrafish larvae as previously described (Carroll et al., 2020). Probes were designed by and purchased from ACD Bio. Hybridization and detection were performed according to the manufacturer’s protocol. Sections were imaged using a confocal microscope (Leica SP8) and z-stack maximum projections were generated using Fiji software.

Alcian blue staining and imaging were performed at 4 dpf as previously described (Carroll et al., 2020). Briefly, larvae were fixed in 4% vol/vol formaldehyde overnight at 4°C. Larvae were dehydrated in 50% vol/vol ethanol and stained with Alcian blue as described (Walker and Kimmel, 2007). Whole and dissected larvae were imaged in 3% wt/vol methylcellulose using a Nikon Eclipse 80i compound microscope with a Nikon DS Ri1 camera. Z-stacked images were taken and extended depth of field was calculated using NIS Element BR 3.2 software. Images were processed using Fiji software. After image capture embryos were genotyped by PCR and fragment analysis.

Compound dact1+/-; dact2+/- zebrafish were in-crossed and progeny were collected from two separate clutches and fixed in 4% formaldehyde at approximately 8 ss. Embryos were individually imaged using a Zeiss Axiozoom stereoscope and processed for DNA extraction and genotyping (Westerfield, 1993). Images were analyzed using Fiji. A circle was drawn to overlay the yolk and the geometric center was determined using the function on Fiji. Using the Fiji angle tool, lines were drawn from the center point to the anterior-most point of the embryo and from the center to the posterior-most point of the embryo. The resulting inner angle of these lines was determined. Each angle measurement was then calculated as a ratio to the average angle of the wildtype embryos. All measurements were performed on images blinded for genotype. ANOVA was performed to determine statistical significance, p < 0.05.

Live embryos were mounted in 1% wt/vol low melt agarose and covered with E3 medium containing 0.013% wt/vol tricaine. Wildtype control and dact1/2−/− compound mutants on a Tg(soxI0:kaede) background were imaged on a Leica DMi8 confocal microscope and photoconverted using the UV laser (404 nm) until the green kaede fluorescence disappeared. For each embryo, one side was photoconverted and the contralateral side served as an internal control. After photoconversion, embryos were removed from the agarose and raised in E3 medium at 28.5°C until the required developmental time point, at which time they were similarly re-mounted and re-imaged. Z-stacked images were processed as maximum-intensity projections using Fiji software.

Single-cell transcriptomic analyses were performed on 10 zebrafish embryos, including 4 wildtype, 3 dact1−/−;dact2−/− compound mutant, and 3 gpc4−/− mutant embryos. Embryos were collected at 4 ss with dact1−/−,dact2−/− and gpc4−/− mutants being identified by their truncated body axis. Embryos were dechorionated with a short (approximately 10 min) incubation in 1 mg/ml Pronase and then washed 3× in embryo medium. Cell dissociation was performed with modifications as previously described (Farrell et al., 2018). Each embryo was transferred to 50 μl DMEM/F12 media on ice. To dissociate cells, media was replaced with 200 μl Dulbecco’s phosphate-buffered saline (DPBS) (without Ca²⁺ and Mg²⁺) with 0.1% wt/vol bovine serum albumin (BSA). Embryos were disrupted by pipetting 10× with a P200 pipette tip. 500 μl of DPBS + 0.1% BSA was added and cells were centrifuged at 300 × g for 1 min. Cell pellets were resuspended in 200 μl DPBS + 0.1% BSA and kept on ice. Just prior to encapsulation, cells were passed through a 40 μm cell strainer, and cell counts and viability were measured. After droplet encapsulation, barcoding, and library preparation using the 10X Genomics Chromium Single Cell 3′ kit (version 3), data were sequenced on an Illumina NovaSeq 6000 sequencer.

FASTQ files were demultiplexed and aligned to the GRCz11 build of the zebrafish genome using Cellranger (version 6.1.0) (Zheng et al., 2017). Raw Cellranger count matrices were imported into R (version 4.1.2) using Seurat (version 4.1.0) (Hao et al., 2021). First, we reviewed data for quality and excluded any droplet that did not meet all of the following criteria: (1) have at least 1500 unique molecular identifiers (UMIs); (2) covering at least 750 distinct genes; (3) have <5% of genes mapping to the mitochondrial genome; and (4) have a log10 of detected genes per UMI >80%. After quality control, the dataset was also filtered to exclude genes with a detection rate below 1 in 3000 cells, leaving a total of 20,078 distinct genes expressed across 19,457 cells for analysis.

The quality-controlled count data were normalized using Pearson’s residuals from the regularized negative binomial regression model, as implemented in Seurat::SCTransfrom (Hafemeister and Satija, 2019). When computing the SCT model, the effect of the total number of UMIs and number of detected genes per cell were regressed out. After normalization, the top 3000 most variably expressed genes were used to calculate principal components. Data were then integrated by source sample using Harmony (version 0.1.0) (Korsunsky et al., 2019). A two-dimensional uniform manifold approximation and projection (Becht et al., 2018) was then derived from the first 40 Harmony embeddings for visualization.

Using the integrated Harmony embeddings, clusters were defined with the Louvain clustering method, as implemented within Seurat. A resolution of 0.3 was used for cluster definition. Cluster identities were assigned by manually reviewing the results of Seurat::FindAllMarkers, searching for genes associated to known developmental lineages. Gene expression data for key markers that guided cluster identity assignment were visualized using Seurat::DotPlot.

Following this detailed annotation, some clusters were grouped to focus downstream analyses on three major lineages: ventral mesoderm (grouping cells from the pronephros, vasculogenic/myeloid precursors, hematopoietic cells, heart primordium, and cephalic mesoderm clusters), dorsal mesoderm (adaxial cells, segmental plate, and paraxial mesoderm), and ectoderm (CNS, mid/hindbrain boundary, spinal cord, and neural crest). For those three lineages, single-cell level data were aggregated per sample and cluster to perform pseudobulk DEA contrasting genotypes. Independent pairwise comparisons of dact1−/−;dact2−/− versus wildtype and gpc4−/− versus wildtype were performed using DESeq2 (v1.34.0) (Love et al., 2014). p-values were corrected for multiple testing using the default Benjamini–Hochberg method; log2 fold change values were corrected using the apeglm shrinkage estimator (Zhu et al., 2019). Significance was defined as an adjusted p-value <0.1 and log2 fold change >0.58 in absolute value. Heatmaps of the top most significant DEGs were generated from the regularized log transformed data using pheatmap (version 1.0.12). Overlap in significant genes across pairwise comparisons were determined and visualized in Venn diagrams. Over-representation analyses against the Gene Ontology database were ran using clusterProfiler (version 4.2.2) (Wu et al., 2021), using as input the set of genes found to be differentially expressed in the comparison of dact1−/−;dact−/− versus wildtype but not gpc4−/− versus wildtype.

All sample sizes represent biological replicates. Analyses were performed using Prism Software (GraphPad) unless otherwise specified. An unpaired Student’s t test or one-way ANOVA with multiple comparisons was used as indicated and a p-value <0.05 was considered significant. Graphs represent the mean ± SEM and n represents biological replicates. For categorical data (normal vs. mutant phenotype) a Fisher exact test was performed between gfp and capn8 injected embryos and the odds ratio was determined. The confidence interval was determined by the Baptista–Pike method.