Mutational analysis of dishevelled genes in zebrafish reveals distinct functions in embryonic patterning and gastrulation cell movements

Wnt signaling plays critical roles in dorsoventral fate specification and anteroposterior patterning, as well as in morphogenetic cell movements. Dishevelled proteins, or Dvls, mediate the activation of Wnt/ß-catenin and Wnt/planar cell polarity pathways. There are at least three highly conserved Dvl proteins in vertebrates, but the implication of each Dvl in key early developmental processes remains poorly understood. In this study, we use genome-editing approach to generate different combinations of maternal and zygotic dvl mutants in zebrafish, and examine their functions during early development. Maternal transcripts for dvl2 and dvl3a are most abundantly expressed, whereas the transcript levels of other dvl genes are negligible. Phenotypic and molecular analyses show that early dorsal fate specification is not affected in maternal and zygotic dvl2 and dvl3a double mutants, suggesting that the two proteins may be dispensable for the activation of maternal Wnt/ß-catenin signaling. Interestingly, convergence and extension movements and anteroposterior patterning require both maternal and the zygotic functions of Dvl2 and Dvl3a, but these processes are more sensitive to Dvl2 dosage. Zygotic dvl2 and dvl3a double mutants display mild axis extension defect with correct anteroposterior patterning. However, maternal and zygotic double mutants exhibit most strongly impaired convergence and extension movements, severe trunk and posterior deficiencies, and frequent occurrence of cyclopia and craniofacial defects. Our results suggest that Dvl2 and Dvl3a products are required for the activation of zygotic Wnt/ß-catenin signaling and Wnt/planar cell polarity pathway, and regulate zygotic developmental processes in a dosage-dependent manner. This work provides insight into the mechanisms of Dvl-mediated Wnt signaling pathways during early vertebrate development.


Introduction
The specification of the dorsoventral (DV) axis is tightly linked to, and ultimately determines, the processes that establish the anteroposterior (AP) pattern in vertebrate embryos. A large body of work has elucidated the induction and patterning processes underlying DV and AP polarities [1,2]. It is well established that, in Xenopus and zebrafish, maternal canonical Wnt/ßcatenin signaling is activated on the future dorsal side following fertilization, and ß-catenin is absolutely required for the establishment of the primary DV asymmetry through cooperation with other factors, such as members of the Nodal family [1][2][3][4][5]. During gastrulation, dorsolateral cells converge toward the dorsal midline, while dorsal midline cells undergo extension along the AP axis. These movements, called convergence and extension (CE), not only provide the driving force for gastrulation, but also make an important contribution to the elongation of AP axis. The conserved non-canonical Wnt/PCP (planar cell polarity) signaling plays a key role in CE movements in all vertebrates [6][7][8][9][10][11][12][13]. Thus, disruption of Wnt/ß-catenin and Wnt/PCP signaling pathways can result in severe defects in the formation of embryonic axes.
Dishevelled (Dvl) is a key intracellular signaling molecule that mediates the activation of both Wnt/ß-catenin and Wnt/PCP pathways during early development [6,8,[14][15][16][17][18]. Functional analyses in Xenopus suggest that Dvl2 (also called Xdsh) exhibits dorsalizing and neuralizing activity [14], and controls polarized cell behaviors that are required for CE movements during gastrulation [15]. However, its requirement for the activation of maternal Wnt/ß-catenin signaling in dorsal axis specification has not been clearly established, and remains largely enigmatic [17]. Simultaneous depletion of maternally expressed dvl2 and dvl3 from Xenopus oocytes using morpholino antisense oligonucleotides has no obvious effect on the expression of maternal Wnt/ß-catenin target genes and the specification of dorsal axis [19]. This negative result may be due to the presence of stored punctae of Dvl proteins in the oocyte cortex, which translocate to the dorsal region soon after fertilization [20], or due to the insufficiency of maternal dvl mRNA depletion. Therefore, appropriate genetic approaches will be necessary to determine whether Dvls function upstream of ß-catenin in early dorsal fate specification.
There are three dvl genes (dvl1, dvl2 and dvl3) in human, mouse, and Xenopus, and at least five in zebrafish [21,22]. They are highly conserved and broadly expressed throughout early development [21,23]. Extensive analyses of mutant phenotypes in mice have uncovered both unique and redundant functions among the three Dvl genes [24][25][26][27]. Although single mutants for these mouse Dvl genes generally survived to adulthood, abnormal neural tube closure and defective organogenesis were observed in different combinations of double mutants, which also cause embryonic lethality [25][26][27]. This, combined with the development in utero, make it less convenient to assay Dvl implication in early axis patterning and morphogenetic processes. In Xenopus, triple knockdown of dvl1, dvl2 and dvl3 led to CE defects [21], which were similar as those resulted from overexpression of Xdd1, a Dvl2 (Xdsh) mutant lacking the PDZ domain [28]. Interestingly, Dvl1 and Dvl2 were found to be involved in neural crest specification and somite segmentation, while Dvl3 was required to maintain muscle gene expression [21]. These observations reveal a distinct requirement of Dvl proteins for Wnt signaling in regulating the expression of developmental genes. However, at present, the relative contribution of individual Dvl protein in CE movements that are dependent on the Wnt/PCP pathway has not been clearly determined, making this important question quite open for further investigation.
One of the critical functions of Wnt signaling during early development is the requirement of zygotic Wnt/ß-catenin pathway for ventroposterior development in Xenopus and zebrafish embryos. In contrast to maternal Wnt/ß-catenin signaling that specifies dorsal fate, zygotic Wnt/ß-catenin signaling inhibits anterior development by activating the expression of different target genes that specify ventral and posterior tissues [1][2][3][4][5]. How different dvl genes are implicated in this process also remains poorly understood. Another difficulty in studying Dvl function is the presence of abundant maternal dvl transcripts, as shown both in Xenopus [14,19], and in zebrafish [29,30]. These maternal products play an important role to support early developmental processes, and when translated into proteins and stored during oogenesis, could not be targeted by knockdown approaches. Also, zygotic homozygous mutants often do not survive to fertile adulthood to produce maternal and zygotic (MZ) mutant embryos for the analysis of maternal gene function. Taken together, it is clear that the maternal and zygotic contributions of dvl genes in different patterning and morphogenetic processes remain elusive, and merits further investigation.
In the present study, we take advantage of the genome-editing approach to generate different combinations of maternal and zygotic dvl mutants in zebrafish. Transcriptomic analysis revealed that, among the five dvl genes, only dvl2 and dvl3a are maternally and abundantly expressed, whereas the transcript levels of the other three dvl genes (dvl1a, dvl1b, dvl3b) are negligible [29]. By creating targeted mutations, we find that MZdvl2 mutants display most severe CE and craniofacial defects, but with correct AP patterning. In dvl2 and dvl3a double mutants, dvl3a dosage exerts a permissive effect on the loss of dvl2 in CE movements and posterior development. By further targeting the wild-type (WT) allele in triallelic mutant embryos to generate mosaic germline transmissible double homozygous adults, we obtained MZdvl2; MZdvl3a offspring. These mutants show correct dorsal fate specification, but display severe CE defects and develop trunk and posterior deficiencies, as well as cyclopia. These observations indicate that Dvl2 and Dvl3a may be not required for maternal Wnt/ß-catenin pathway activation. Instead, Dvl2 plays a major role in Wnt/PCP and zygotic Wnt/ß-catenin signaling. Our findings thus help to better understand the function of Dvl proteins in different patterning and morphogenetic processes.

MZdvl2 mutants display axis extension and craniofacial defects
We used transcription activator-like effector nucleases (TALENs) genome-editing approach to generate mutant lines for the five zebrafish dvl genes (dvl1a, dvl1b, dvl2, dvl3a, dvl3b). All indel mutations led to premature stop codons in the transcripts, and resulted in proteins truncated either at the DIX or the PDZ domain (S1-S5 Figs). Analysis of the phenotypes at different stages indicated that, except for dvl2, maternal and zygotic mutants for the other dvl genes developed normally, and could survive to adulthood and were fertile (Fig 1 and S6 Fig). Because dvl2 and dvl3a represent the most abundantly expressed maternal transcripts (S7 Fig), we focused our analyses on these two mutant lines. Compared with WT embryos at different stages ( Fig 1A-1C), Zdvl2 mutants showed weakly reduced AP axis extension at 11.5 hpf (hours post-fertilization), as judged by the degree of the angle between the anterior end and posterior end, with vertex at the geometric center of the embryo (Fig 1M). This defect was largely recovered at 30 hpf (Fig 1E). At 5 dpf (days post-fertilization), Zdvl2 mutants displayed essentially a normal AP axis, except for the presence of a smaller gas-filled swim bladder ( Fig  1F). However, only about half of these Zdvl2 mutants could survive to adulthood, and only about two-third of the survived adult female fish could spawn, whereas all male Zdvl2 mutants were not fertile due to the absence of courtship behavior.
To obtain MZdvl2 embryos, we crossed female dvl2 -/fish with male dvl2 +/fish. At 11.5 hpf, about half of the resulting embryos showed more severe axis extension defect (Fig 1G and  1M). At 30 hpf, they exhibited a shortened AP axis, associated with a reduced yolk extension, which are characteristics of defective CE movements (Fig 1H). At this stage, MZdvl2 mutants also displayed an obvious cyclopic phenotype (compare insets in Fig 1B and 1H). At 5 dpf, although MZdvl2 mutants had a similar length of AP axis as WT embryos, abnormalities in the head region were clearly apparent. These include severe craniofacial defects, and cyclopia or fused eyes (compare insets in Fig 1C and 1I). In particular, pharyngeal arches were not correctly positioned, and eventually protruded outward (Fig 1I). Alcian blue staining of larval head cartilages indicated that, among other abnormalities, the pair of trabeculae was fused and the ethmoid plate was absent ( Fig 1N). All these phenotypes are reminiscent of impaired Wnt/ PCP signaling and defective extension of axial tissues, which are frequently observed in other Wnt/PCP-specific mutants, such as trilobite/vangl2, and slb/wnt11 [31][32][33]. Most strikingly, the protrusion outward of pharyngeal cartilages is much similar as the "bulldog" facial phenotype described in slb/wnt11 mutant [32][33][34]. A more detailed analysis of these late phenotypes is beyond the scope of this study, but it will be interesting for future work. Due to these defects, MZdvl2 mutant embryos could not survive beyond 5 dpf. Genotyping by allele-specific PCR (S1 Table and S8 Fig) of large numbers of severely affected embryos derived from three independent fish pairs confirmed that they were indeed MZdvl2 mutants (Fig 1O and 1P). In contrast to MZdvl2 mutants, the late phenotype of MZdvl3a mutants was indistinguishable from that of WT embryos (Fig 1J-1L), although statistical analysis revealed a weakly reduced AP axis extension at 11.5 hpf (Fig 1M). These results show that both maternal and zygotic Dvl2 make an important contribution to CE movements. They also suggest that deficiency of maternal and zygotic Dvl2 or Dvl3a is not sufficient to affect DV and AP patterning.
By RT-PCR analysis, we found that maternal dvl2 mutant transcripts were subjected to nonsense-mediated mRNA decay (NMD) at cleavage stages, which was further confirmed by in situ hybridization. Maternal dvl3a mutant transcripts also underwent NMD, but to a lesser extent. We then checked whether there was a mutual compensation between dvl2 and dvl3a. No significant change in the level of maternal dvl2 transcripts was found in MZdvl3a mutants, whereas the level of maternal dvl3a transcripts showed a weak decrease in Mdvl2 mutants (S9 Fig). We also attempted to verify if Dvl2 or Dvl3a protein was absent in the mutants by western blotting using available antibodies, but failed to detect any specific signal either in WT or in mutant embryos. Alternatively, to examine whether dvl2 and dvl3a mutant transcripts may be translated by using an alternative ATG or by a stop codon bypass mechanism, we cloned WT and mutant dvl2 and dvl3a coding sequences upstream of myc sequences, and injected the corresponding mRNA (100 pg) into zebrafish embryos. Analysis by western blotting did not detect any signal in embryos injected with these mutant mRNAs (S10A and S10B Fig). This suggests that dvl2 and dvl3a mutant transcripts should not be translated into proteins. mutants, obtained by crosses between heterozygous dvl2 +/carriers, display weak axis extension defect at 11.5 hpf. They are normal at 30 hpf, and show reduced swim bladder (sb) at 5 hpf. (G-I) MZdvl2 mutants, obtained from crosses between female dvl2 -/fish and male dvl2 +/fish, display a reduced AP axis at 11.5 hpf and 30 hpf. They present fused eyes at 30 hpf (inset, ventral view; arrowhead indicates fused lenses), and develop craniofacial defects and cyclopia (inset, dorsal view), with pharyngeal cartilages protruding outward (arrow) at 5 dpf. (J-L) MZdvl3a mutants from crosses between dvl3a -/carriers display weak axis extension defect, but are indistinguishable from WT embryos at 30 hpf and 5 dpf. (M) Statistical analysis of the extent of axis extension delay. The embryos were imaged at 11.5 hpf followed by genotyping. Those embryos with expected genotypes were used to measure the angle between the anterior end and posterior end, with vertex at the geometric center of the embryo (inset). Bars represent the mean ± s.d. from indicated numbers of embryos ( ÃÃÃÃ , P<0.0001). (N) Alcian blue staining of head cartilages at 5 dpf. Cartilage structures of the basicranium are outlined in grey, showing the fusion of trabeculae and the absence of ethmoid plate (ep) in MZdvl2 mutants. (O) Quantitative analysis of MZdvl2 mutant phenotypes at 5 dpf in offspring from three independent female dvl2 -/and male dvl2 +/fish pairs. Control embryos were obtained from crosses between female WT fish and male dvl2 +/fish. Numbers on the top of each column indicate total embryos analyzed. (P) Genotyping of dvl2 mutants with normal and defective phenotypes. All embryos with a normal phenotype are dvl2 +/mutants, whereas all defective embryos are MZdvl2 mutants. Numbers on the top of each column indicate total embryos genotyped from three independent fish pairs. Scale bars: (A, D, G, J) 400 μm; (B, E, H, K) 400 μm; (C, F, I, L) 400 μm; (N) 100 μm. https://doi.org/10.1371/journal.pgen.1007551.g001

Zygotic Dvl2 and Dvl3a cooperate to regulate axis extension
To analyze the function of Dvl2 and Dvl3a dosages in AP axis extension and patterning, we first generated double heterozygous dvl2 +/-;dvl3a +/mutants by crosses between female Zdvl2 and male Zdvl3a. Compared to WT embryos (Fig 2A-2C), dvl2 +/-;dvl3a +/mutants showed weak, but significant delay in axis extension at 11.5 hpf (Fig 2D and 2P). They were completely normal at 30 hpf and 5 dpf (Fig 2E and 2F), and developed to fertile adults, indicating that Wnt/ß-catenin signaling was not affected, and Wnt/PCP signaling was weakly affected during gastrulation. We then intercrossed dvl2 +/-;dvl3a +/fish to analyze the phenotypes of triallelic mutants. Similar as dvl2 +/-;dvl3a +/mutants, dvl2 +/-;Zdvl3a mutants displayed weak axis extension defect at 11.5 hpf (Fig 2G and 2P), and slightly shortened AP axis at 30 hpf ( Fig 2H). These mutants also recovered to a nearly normal phenotype at 5 dpf ( Fig 2I). However, the extent of axis extension delay was more pronounced in Zdvl2;dvl3a +/mutant embryos at 11.5 hpf (Fig 2J and 2P), which further developed a shortened AP axis at 30 hpf (Fig 2K), and a compressed head at 5 dpf ( Fig 2L). These embryos, which were grouped as type I, presented a more severely affected phenotype than Zdvl2 mutants and did not survive to adulthood. This suggests that removal of one dvl3a allele could enhance CE defects in Zdvl2 mutants.
Since dvl2 +/-;Zdvl3a mutants could survive to adulthood and were fertile, the male fish could be used for generating Zdvl2;Zdvl3a double mutants, after crosses with female dvl2 +/-; dvl3a +/fish. We found that double Zdvl2;Zdvl3a mutants displayed more strongly affected axis extension at 11.5 hpf (Fig 2M and 2P), and developed a shortened AP axis, with a reduced yolk extension at 30 hpf ( Fig 2N). The distance between the two eyes was also reduced (compare insets in Fig 2B and 2N). All these phenotypes are suggestive of impaired CE movements. These embryos, grouped as type II, further developed bent axis and craniofacial defects at 5 dpf ( Fig 2O). From three independent fish pairs, we found that the occurrence of the defective axis extension phenotype in the resulting offspring was quite reproducible because similar proportions of type I (13%) and type II (15%) mutant embryos have been obtained (Fig 2Q and  2R), which could be expected from the crosses between dvl2 +/-;dvl3a +/and dvl2 +/-;Zdvl3a fish. Since both head and tail were present in zygotic double homozygous mutants, this indicates that half of the maternal Dvl2 and Dvl3a products should be largely sufficient to support AP patterning.

Maternal contribution of Dvl2 and Dvl3a in axis extension and AP patterning
Since AP patterning does not seem to be affected in Zdvl2;Zdvl3a mutants, we examined the maternal contribution of Dvl2 and Dvl3a in this process. Intercrosses between dvl2 +/-;Zdvl3a triallelic fish could generate three types of mutant offspring, including MZdvl3a, dvl2 +/-; MZdvl3a, and Zdvl2;MZdvl3a. Analysis of the phenotypes followed by genotyping indicated that, compared to WT embryos (Fig 3A-3C), dvl2 +/-;MZdvl3a mutants displayed mild axis extension defect at 11.5 hpf (Fig 3D). At 30 hpf, these embryos were short in length, but AP patterning was not affected because head and tail regions were correctly formed (Fig 3E). At 5 dpf, they completely recovered to a normal phenotype (Fig 3F), and eventually developed to fertile adult. This indicates that, in the absence of Dvl3a, maternal and zygotic Dvl2 product derived from one allele is sufficient for AP patterning, although Wnt/PCP pathway activation is reduced at early stages.
Strikingly, Zdvl2;MZdvl3a mutants displayed strong axis extension defect at 11.5 hpf ( Fig  3G), indicating severely impaired Wnt/PCP signaling. At 30 hpf and 5 dpf, axis extension and AP patterning defects became particularly prominent. These mutants displayed a reduced head size with cyclopia or fused eyes (Fig 3H and 3I). This phenotype is likely caused by Double heterozygous dvl2 +/-;dvl3a +/mutants, from crosses between female dvl2 -/fish and male dvl3a -/fish, show weakly delayed axis extension at 11.5 hpf, but are phenotypically normal at later stages. (G-I) Triallelic dvl2 +/-;Zdvl3a mutants,from crosses between dvl2 +/-; dvl3a +/carriers, display weak axis extension defect at 11.5 hpf and 30 hpf, and recover at 5 dpf. (J-L) Triallelic Zdvl2;dvl3a +/mutants, from crosses between dvl2 +/-;dvl3a +/carriers, exhibit more obvious axis extension defect at 11.5 hpf and 30 hpf, and develop shortened AP axis, compressed head, and reduced swim bladder at 5 dpf, which are referred as type I phenotype. (M-O) Zdvl2;Zdvl3a mutants, from crosses between female dvl2 +/-;dvl3a +/fish and male dvl2 +/-;Zdvl3a fish, show more strong axis extension defect at 11.5 hpf and 30 hpf, and present a severe type II phenotype, with shortened and wavy axis, craniofacial defects, and complete disappearance of swim bladder (sb) at 5 dpf. (P) Statistical analysis of the extent of axis extension delay, after genotyping of imaged embryos at 11.5 hpf. Capital letters of the abscissa correspond to the images at 11.5 hpf. Bars represent the mean ± s.d. from indicated numbers of embryos, and asterisks above the bars refer to comparison with WT embryos ( ÃÃÃ , P<0.001; ÃÃÃÃ , P<0.0001; NS, not significant). (Q) Quantitative analysis of type I and type II phenotypes at 5 dpf from three independent female dvl2 +/-;dvl3a +/and male dvl2 +/-;dvl3a -/fish pairs. Control embryos were from crosses between female WT fish and male dvl2 +/-;dvl3a -/fish. (R) Genotypes of type I and type perturbed midline development, and is a fish-specific consequence independent of the affected genetic pathway [35]. Cardiac edema was also evident at 5 dpf ( Fig 3I). Importantly, Zdvl2; MZdvl3a mutants exhibited a severely reduced body length, with posterior truncation (Fig 3H  and 3I). We then intercrossed three independent dvl2 +/-;Zdvl3a fish pairs, and found that the appearance of severe axis extension defect associated with posterior deficiency was quite reproducible in the resulting offspring, with an average of 24.6% that corresponds to the Mendel  Dishevelled mutants in zebrafish inheritance ( Fig 3J). Genotyping of these severely affected embryos confirmed that they were indeed Zdvl2;MZdvl3a mutants ( Fig 3K). This analysis indicates that, in the MZdvl3a background, the reduction of maternal Dvl2 dosage along with the absence of zygotic Dvl2 products strongly impairs Wnt/PCP-dependent axis extension, and leads to defective posterior patterning.

Generation of MZdvl2;MZdvl3a double mutants
Zdvl2;Zdvl3a double mutants could not survive to adulthood, this prevented us from obtaining maternal and zygotic double mutants to fully address the maternal contribution of Dvl2 and Dvl3a in DV and AP patterning. To circumvent this obstacle, we used a strategy to generate mosaic dvl2 -/-;dvl3a -/fertile adult fish, by disrupting the remaining dvl2 WT allele in dvl2 +/-; dvl3a -/embryos derived from intercrosses between triallelic dvl2 +/-;dvl3a -/fish ( Fig 4A). To obtain viable mosaic adult fish, low amounts of the original dvl2 TALENs mRNAs (100 pg each) were injected in these triallelic mutant embryos at 1-cell stage. The resulting adult female fish were crossed with male dvl2 +/-;dvl3a -/fish, and genotyping of the offspring by allelic-specific PCR followed by sequencing was performed to screen female fish carrying a new germline transmissible dvl2 mutant allele ( Fig 4B). Those positive female fish were designated as mdvl2 +(-)/-;dvl3a -/-, for mosaic dvl2 homozygous genotype. By this approach, if mutations of the remaining dvl2 WT allele occur in some germ cells, MZdvl2;MZdvl3a offspring could be obtained through crosses between female mdvl2 +(-)/-;dvl3a -/fish and male dvl2 +/-;dvl3a -/fish (S11 Fig). Indeed, this strategy has allowed us to generate rather efficiently mdvl2 +(-)/-;dvl3a -/fish. Among a total of 45 female adults tested, 7 female fish produced offspring with a varied proportion of extremely severe phenotype. At 11.5 hpf, these embryos displayed most strongly reduced axis extension associated with sharply widened paraxial mesoderm (Fig 4C-4F). At 30 hpf, they exhibited severe trunk and posterior deficiencies, but the head region was still present (Fig 4G and 4H). At 5 dpf, all these embryos developed cyclopia or fused eyes (S12 Fig). Other siblings either presented posterior truncation, which were genotyped as Zdvl2;MZdvl3a mutants ( Fig 4I), or developed relatively normally, which were genotyped as dvl2 +/-;MZdvl3a triallelic mutants or MZdvl3a mutants. These embryos could be expected from the crosses, since a high proportion of germ cells with one dvl2 WT allele should be still present in mdvl2 +(-)/-;dvl3a -/fish.
We then genotyped all the most severely affected embryos to confirm the presence of separate mutations in the dvl2 alleles. As expected, analysis of the sequencing chromatograms revealed that, in addition to the original mutated allele that has a deletion of 5 nucleotides, different new indels were also detected in these embryos ( Fig 4J). It is unlikely that the re-injected dvl2 TALENs could target the original mutant allele in dvl2 +/-;dvl3a -/embryos, since we verified that they had no effect in dvl2 -/embryos, and it has been shown the TALEN pair had no gene modification activity when separated by 11 nucleotides or less [36], which is the case for the original dvl2 mutant allele. Thus, we can conclude that the extremely severe defects were specific to MZdvl2;MZdvl3a mutants, and were caused by the deficiency of maternal and zygotic Dvl2 and Dvl3a products. When the phenotypes of offspring from three independent crosses between a fixed pair of female mdvl2 +(-)/-;dvl3a -/and male dvl2 +/-;dvl3a -/fish were analyzed, we reproducibly obtained most severely affected embryos, although the proportion varied among fish pairs, or between crosses from a fixed fish pair ( Fig 4K). Some female mdvl2 +(-)/-;dvl3a -/fish (#2, #25, #39) produced a relatively high proportion of MZdvl2; MZdvl3a mutant embryos, ranging from 10% to 24%, depending on the crosses. Thus, by generating mosaic double mutants, we revealed a maternal requirement for Dvl2 and Dvl3a in AP patterning, and in axis extension.

Dorsal fate specification is not affected in MZdvl2;MZdvl3a mutants
An implication of Dvl in activating maternal Wnt/ß-catenin signaling for dorsal fate specification has been unclear in Xenopus [17,19,20,28], and this issue has not been addressed in zebrafish. Unlike the maternal effect mutant, ichabod, that reduces ß-catenin2 transcripts and results in severe dorsal and anterior deficiencies during early development [37,38], anterior structures such as eyes were present in MZdvl2;MZdvl3a mutants, although they were fused or cyclopic. This implies that dorsal and anterior fate specification should not be disrupted in the absence of maternal and zygotic Dvl2 and Dvl3a products. To further test this possibility, we examined the expression of dorsal mesoderm genes, goosecoid and chordin, and the pan-mesoderm gene, tbxta (ntla), at dome stage by in situ hybridization. Since only a small proportion of MZdvl2;MZdvl3a embryos was present in the offspring, and no phenotype difference could be observed among siblings at early stages, all the offspring from the crosses between female mdvl2 +(-)/-;dvl3a -/and male dvl2 +/-;dvl3a -/fish were collected, and divided into three parts for hybridization with each probe. Following in situ hybridization, the embryos were individually imaged and genotyped by sequencing ( Fig 5A). The experiment was performed using three independent fish pairs and the results did not reveal any obvious difference in the expression patterns of goosecoid, chordin, and tbxta between WT embryos and MZdvl2;MZdvl3a mutants ( Fig 5B-5F, 5H-5L and 5N-5R). By contrast, in the embryos injected with 2 ng ß-catenin2 morpholino (ß-cat2MO), goosecoid expression was strongly reduced, chordin expression was absent, but tbxta expression was not affected (Fig 5G, 5M and 5S), confirming that maternal ß-catenin2 is required for dorsal fate specification. Together with phenotype analysis, this result suggests that maternal Dvl2 and Dvl3a may be not required for dorsal axis specification in zebrafish.
The negative result could be interpreted by the presence of residual amounts of other Dvl proteins. To test this possibility, we injected 2 ng of each dvl1a, dvl1b, and dvl3b morpholino oligonucleotides (referred to as dvl1a/1b/3bMOs) in 1-cell stage offspring derived from the crosses between female mdvl2 +(-)/-;dvl3a -/fish and male dvl2 +/-;dvl3a -/fish. In situ hybridization analysis of the expression of dorsal organizer genes goosecoid and chordin was performed Dishevelled mutants in zebrafish at dome stage. The results showed that simultaneous knockdown of dvl1a, dvl1b, and dvl3 in WT or MZdvl2;MZdvl3a embryos did not change the expression of goosecoid and chordin, compared to the embryos injected with 2.5 ng control morpholino (Fig 6A-6D and 6F-6I). Consistent with this observation, analysis by immunocytochemistry indicated that the nuclear accumulation of endogenous ß-catenin in dorsal marginal cells was comparable between WT and MZdvl2;MZdvl3a embryos at high stage (compare Fig 6E and 6J). Similarly, injection of dvl1a/1b/3bMOs had no effect on axis extension in WT embryos (Fig 6K, 6L, 6P, 6Q and 6T), and did not aggravate the defective phenotype of MZdvl2;MZdvl3a mutants ( Fig 6M-6O and 6R-6T). Moreover, quantitative RT-PCR analysis at different stages, and analysis by RNA sequencing at 12 hpf all showed that there was no upregulation of dvl1a, dvl1b and dvl3b transcripts in MZdvl2;MZdvl3a mutants (S13 Fig). These results suggest that Dvl1a, Dvl1b, and Dvl3b could not compensate for the loss of Dvl2 and Dvl3a in early dorsal fate specification and in axis extension. To further confirm the absence of Dvl activity in MZdvl2;MZdvl3a early embryos, we injected synthetic wnt8a mRNA (50 pg) in 1-cell stage offspring obtained as above. In situ hybridization analysis of goosecoid and chordin ectopic expression clearly showed that overexpression of Wnt8a was able to strongly induce ectopic expression of these genes in WT embryos (Figs 7A, 7B, 7G and 6H). However, reducing Dvl dosage progressively decreased the activity of Wnt8a to activate ectopic organizer gene expression (Fig 7C, 7D, 7I and 7J). In particular, Wnt8a had no effect in Zdvl2;MZdvl3a or MZdvl2;MZdvl3a embryos, which showed similar goosecoid and chordin expression patterns as in WT embryos (Fig 7E, 7F, 7L and 7M). The fact that Wnt8a failed to activate maternal Wnt/ß-catenin signaling Knockdown of other dvl genes has no effect in MZdvl2;MZdvl3a mutants. The embryos were injected with coMO (2.5 ng) or a mixture of equal amount of dvl1a/1b/3bMOs (6 ng in total) at 1-cell stage, except for immunostaining. All mutant embryos were imaged after different analyses, followed by genotyping. Dishevelled mutants in zebrafish leading to ectopic organizer gene expression suggests that Dvl activity should be absent in these mutants.

Maternal Dvl2 and Dvl3a regulate AP patterning
Inhibition of zygotic Wnt/ß-catenin signaling, in particular Wnt8, is known to cause dorsalization and anteriorization [1,3,5]. However, this does not seem to occur in MZdvl2;MZdvl3a embryos, since the expression domains of goosecoid and chordin at 60% epiboly did not show obvious lateral expansion, although lateral and ventral expression of axin2 was inhibited (S14 Fig). This suggests that downregulation of Wnt8 and Dvl activity exerts distinct effects to restrict the organizer domain.
To analyze how AP patterning is affected in MZdvl2;MZdvl3a mutants, we first performed in situ hybridization using two well characterized posterior markers, sp5l (spr2) and tbx16l (tbx6l), which mediate zygotic Wnt/ß-catenin signaling in posterior patterning [39][40][41]. At 12 hpf, the phenotypes specific to Zdvl2;MZdvl3a or MZdvl2;MZdvl3a mutants were easily discernible, which allowed us to select sufficient mutant embryos from different crosses. Confirmed by genotyping after in situ hybridization, Zdvl2;MZdvl3a mutants showed no obvious or only weak alternation in the expression of sp5l (Fig 8A, 8A', 8C and 8C'), but they displayed a markedly reduced expression of tbx16l (Fig 8B, 8B', 8D and 8D'). In MZdvl2;MZdvl3a mutants, however, the expression of sp5l was strongly decreased (Fig 8E and 8E'), and the expression of tbx16l was reduced to residual level (Fig 8F and 8F'). Consistently, TOPFlash luciferase assay revealed that there was approximately a 30% decrease of Wnt/ß-catenin transcriptional activity in Zdvl2;MZdvl3a mutants at 12 hpf, and about a 75% decrease in MZdvl2;MZdvl3a mutants ( Fig 8G). This decrease of reporter activity correlated well with a reduction of endogenous ßcatenin nuclear accumulation in ventral marginal cells at shield stage, when zygotic transcription has already started, as assayed by immunofluorescence staining (Fig 8H and 8I). These observations strongly suggest that maternal Dvl2 and Dv3a play an important role in the activation of zygotic Wnt/ß-catenin signaling, and that Dvl2 may exert a predominant role.
We further examined the expression of a panel of AP genes in MZdvl2;MZdvl3a mutants at 12 hpf by in situ hybridization. This indicated that the expression of otx1 (forebrain and  midbrain), otx2 (forebrain, midbrain and midbrain-hindbrain boundary), and pax2a (midbrain-hindbrain boundary, otic placode and pronepheric mesoderm) was reduced, and widened mediolaterally due to impaired axis extension (Fig 8J-8L). The expression of egr2b (krox20) in rhombomeres 3 and 5, and hoxb1b in the notochord and paraxial mesoderm was compressed and widened (Fig 8M and 8N). Notably, the expression of hoxb1b in the tailbud was absent (Fig 8N). The expression of axin2 in the neural plate, and of cdx4 in the posterior paraxial mesoderm was strongly inhibited (Fig 8O and 8P). Analysis by RNA sequencing confirmed that axin2, tbx16l, sp5l and cdx4, which are zygotic Wnt/ß-catenin target genes in AP patterning [3,[39][40][41], as well as hoxb1b, were all downregulated (Fig 8Q-8U). These indicate that AP patterning, and in particular posterior development, are strongly affected in MZdvl2; MZdvl3a mutants.
To determine whether AP patterning defect in MZdvl2;MZdvl3a mutants was due to a decreased zygotic Wnt/ß-catenin signaling, offspring at 5 hpf derived from the crosses between female mdvl2 +(-)/-;dvl3a -/fish and male dvl2 +/-;dvl3a -/fish were treated with LiCl, which activates Wnt/ß-catenin signaling downstream of Dvl. In situ hybridization was performed to examine the expression of axin2 and tbx16l at 12 hpf. Following genotyping, we found that MZdvl2;MZdvl3a mutants treated with LiCl displayed increased expression of axin2 and tbx16l, compared to untreated mutants (Fig 8V and 8W). Injection of a low dose of synthetic mRNA (50 pg) encoding the constitutively active ß-catenin into 1-cell stage embryos also rescued the expression of axin2 and tbx16l, but to a lesser extent (S15A-S15F Fig), likely due to the mosaic distribution of injected mRNA. However, phenotypic examination indicated that this injection could effectively rescue tail development in Zdvl2;MZdvl3a mutants (S15G-S15I Fig). These observations further demonstrate that Dvl2 and Dvl3a deficiency causes defective AP patterning by preventing zygotic Wnt/ß-catenin signaling.

Maternal Dvl2 and Dvl3a contribute to CE movements
The most severely impaired axis elongation in MZdvl2;MZdvl3a mutants clearly suggests an important maternal contribution of Dvl2 and Dvl3a for CE movements. To further clarify this aspect and to determine Dvl dosages in Wnt/PCP signaling, we first compared the extent of axis extension defect between different mutants by phenotype analysis, and by simultaneous in situ hybridization using tbxta as a marker of the notochord, dlx3 as a marker of the neural plate borders, and ctslb (hgg1) as a marker of the prechordal plate mesoderm. At 11.5 hpf, Zdvl3a, Zdvl2 or MZdvl3a mutants displayed weak, but obvious delay in neural plate convergence and notochord elongation (Fig 9A-9D, 9I-9L and 9Q). However, MZdvl2 mutants showed more severe defect (Fig 9E, 9M and 9Q). Thus, it is clear that either MZdvl2 or MZdvl3a mutants present more severely affected CE phenotypes than the respective zygotic mutants. The same situation was also observed in double mutants. MZdvl2;MZdvl3a mutants displayed most severely impaired neural plate convergence and axis extension, compared with Zdvl2;Zdvl3a and Zdvl2;MZdvl3a mutants ( Fig 9F-9H and 9N-9Q), indicating that maternal Dvl dosage is important in Wnt/PCP signaling.
Dvl was shown to regulate cytoskeletal architecture and cell polarity upstream of Rac and Jun N-terminal kinase (JNK) both in Xenopus and zebrafish [42,43]. When the AP1 luciferase reporter was used to monitor JNK activation [44], we found a 40% decrease of AP1 reporter activity in Zdvl2;MZdvl3a mutants, and a 60% decrease in MZdvl2;MZdvl3a mutants ( Fig 9R). Consistent with this result, analysis of cell polarity indicated that WT midline cells at 12 hpf were elongated mediolaterally, while MZdvl2;MZdvl3a mutant cells were rounded with a strongly reduced length to width ratio (Fig 9S and 9T), which was much similar as dvl2 and dvl3a knockdown cells [43]. However, injection of synthetic mRNA (200 pg) encoding a constitutively active JNK, along with synthetic mRNAs encoding Histone2B and membrane GFP in one blastomere at 64-cell stage, significantly rescued mediolateral elongation of hpf. Lateral view, with anterior region on the top. (I-P) Dorsal view of indicated embryos simultaneously hybridized with ctslb, dlx3, and tbxta probes to reflect the position of prechordal plate mesoderm, neural plate borders, and notochord, respectively. MZdvl3a mutants were from intercrosses between dvl3a -/carriers; MZdvl2 mutants were from crosses between female dvl2 -/fish and male dvl2 +/fish; Zdvl2;Zdvl3a mutants were from crosses between female dvl2 +/-;dvl3a -/fish and male dvl2 +/-;dvl3a +/fish; Zdvl2;MZdvl3a mutants were from intercrosses between dvl2 +/-;dvl3a -/carriers; MZdvl2;MZdvl3a mutants were from crosses between female mdvl2 +(-)/-;dvl3a -/fish and male dvl2 +/-;dvl3a -/fish. (Q) Statistical analysis shows that progressive reduction of Dvl2 and Dvl3a dosages increasingly aggravates axis extension defect. The embryos were imaged at 11.5 hpf, and genotyped before measuring the angle (inset). Bars represent the mean ± s.d. from indicated numbers of embryos collected from three independent experiments, and asterisks above the bars indicate significance with respect to WT embryos (  Dishevelled mutants in zebrafish fluorescently-labeled descendent MZdvl2;MZdvl3a mutant cells (Fig 9S and 9T). This suggests that deficiency of Dvl2 and Dvl3a affects cell polarity by preventing at least partially JNK activation. Furthermore, we performed live time-lapse analysis of cell movement behaviors in Zdvl2;MZdvl3a and MZdvl2;MZdvl3a mutants, in comparison with the PCP mutant trilobite/ vangl2 [45]. Since the AP axis was strongly shortened in these mutants, and the notochord became severely irregular, we examined the dorsal convergence behaviors of lateral cells. In WT embryos, these cells moved toward the notochord with regular trajectories (Fig 9U and S1 Movie). In zygotic trilobite/vang2 mutants, lateral cells displaced along irregular trajectories and moved in more posterior direction (Fig 9V and S2 Movie). Lateral cells in Zdvl2;MZdvl3a mutants moved similarly as in zygotic trilobite/vang2 mutants (Fig 9W and S3 Movie). Strikingly, the directional movement of lateral cells in MZdvl2;MZdvl3a mutants was most severely affected. These cells moved along zigzagging trajectories, and displayed forward and back movements (Fig 9X and S4 Movie), with a more strong increase in the ratio of trajectory distance relative to net distance toward the notochord (Fig 9Y). These analyses strongly indicate an important maternal contribution of Dvl2 and Dv3a in Wnt/PCP signaling and CE movements. The fact that lateral cells in these mutants tend to move in more vegetal direction is likely due to an impaired AP patterning. Thus, it could not be excluded that the severely affected CE movements may result from the combined effects of defective Wnt/PCP and zygotic Wnt/ß-catenin signaling.
We further demonstrated the maternal contribution of Dvl2 and Dvl3a in axis extension using maternal double mutants. By crossing female mdvl2 +(-)/-;dvl3a -/fish with male WT fish, the resulting double heterozygous embryos that carry a new mutation in one dvl2 allele are maternal double mutants because of the absence of maternal Dvl2 and Dvl3a products (see S11 Fig). These Mdvl2;Mdvl3a mutants displayed strongly impaired axis extension at 11.5 hpf. However, they only presented a slightly shortened AP axis at 30 hpf (S16A-S16D and S16I Fig), suggesting that zygotic Dvl2 and Dvl3a can rescue axis extension at late stages. By contrast, Mdvl3a mutants or dvl2 +/-;Mdvl3a mutants (lacking half of the maternal Dvl2 products), obtained from the crosses between female dvl2 +/-;dvl3a -/fish and male WT fish, were either normal or displayed weak axis extension defect at 11.5 hpf, and were completely normal at 30 hpf (S16E-S16I Fig). These observations indicate a predominant contribution of maternal Dvl2 to embryonic axis elongation, and Dvl3a exerts a permissive effect. Altogether, our results show that both maternal and zygotic Dvl2 and Dvl3a cooperate to orchestrate CE movements and AP patterning.

Discussion
Dvl proteins play key roles in both Wnt/ß-catenin and Wnt/PCP signaling pathways, and they are highly conserved and broadly expressed during early development in all vertebrates. However, many aspects of their involvement in early developmental processes remain elusive and enigmatic. In this study, we resolved some of the unanswered issues through comprehensive mutational analyses. Our results demonstrate that the two most abundantly expressed Dvl proteins, Dvl2 and Dv3a, are not required for early dorsal fate specification, which is dependent on the activation of maternal Wnt/ß-catenin signaling. Instead, maternal and zygotic Dvl2 and Dvl3a, in particular Dvl2, are important in CE movements, which are regulated by Wnt/PCP signaling, and in AP patterning. These findings help to clarify the implication of Dvl proteins in Wnt-regulated developmental events, and provide insight into the mechanisms underlying embryonic axis formation.
The early development in many vertebrates and invertebrates is supported by maternal products accumulated during oogenesis, zygotic transcription does not occur until the start of mid-blastula transition [1][2][3][4][5]. Inactivation of key genes implicated in early developmental processes frequently leads to embryonic lethality, or unproductive adults, preventing the analysis of maternal gene function. The situation becomes more complex when multiple gene paralogs are expressed. This is particularly true for functional analyses of Dvl paralogs during early development. Before this work, no dvl mutant has been reported in zebrafish, and the maternal and zygotic functions of Dvl proteins are not clear. We have used TALENs genome-editing technology to generate single mutants for all five zebrafish dvl paralogs, as well as a panel of dvl2 and dvl3a triallelic and double homozygous mutants, and examined the maternal and zygotic contributions of Dvl2 and Dvl3a in embryonic patterning and morphogenetic movements. These mutants represent a valuable resource for the study of important developmental processes, which are dependent on the activation of the Wnt pathways.
A significant finding is the absence of implication for maternal Dvl2 and Dvl3a in early dorsal fate specification. This suggests that they are not required for the activation of maternal Wnt/ß-catenin signaling. Indeed, the expression of the dorsal organizer genes, goosecoid and chordin, is not affected in these mutants, whereas it is strongly decreased in ß-catenin2 morphants. Moreover, the phenotypes of MZdvl2;MZdvl3a or Mdvl2;Mdvl3a mutants completely differ from those of zebrafish ichabod mutants, which display an absence of anterior structures caused by a reduced ß-catenin2 activity [37,38]. This further argues against a requirement of Dvl2 and Dvl3a in dorsal axis formation. Our results from mutational analyses are consistent with the observations showing that simultaneous knockdown of dvl2 and dvl3a in zebrafish does not apparently affect head formation [43,46,47], but mostly affects AP axis elongation and tail development [43]. They are supported by functional studies in Xenopus, which show that depletion of maternally expressed Dvl2 and Dvl3 from oocytes also has no effect on dorsal fate specification [19]. There is a possibility that these negative outcomes could be due to the inefficiency of the approaches to inhibit endogenous Dvl function [20]. However, our genetic evidence now suggests that the activation of maternal Wnt/ß-catenin signaling is independent of Dvl2 and Dvl3a. Although it was extremely difficult to assay maternal Wnt/ß-catenin activity in MZdvl2;MZdvl3a mutants at blastula stages, the correct expression of dorsal organizer genes is suggestive of an unaffected maternal Wnt/ß-catenin signaling.
Zebrafish genome contains at least five dvl genes, but transcriptomic analysis has revealed that dvl2 and dvl3a represent approximately 98% of total dvl transcripts from fertilization until before the end of gastrulation [29], indicating that they are the major dvl genes expressed in the early embryo. It is unlikely that the loss-of-function of Dvl2 and Dvl3a could be compensated by other Dvl proteins, since they are expressed at an extremely low level, and their maternal and zygotic mutants do not result in any discernable phenotype at all stages examined. Moreover, our quantitative analyses indicate that the maternal and zygotic expression of dvl1a, dvl1b and dvl3b is not upregulated in MZdvl2;MZdvl3a mutants, and that simultaneous knockdown of these genes has no effect. A recent study indicates that Wnt/ß-catenin signaling is not affected in mouse ependymal cells lacking 5 of the 6 Dvl alleles [48], there is thus a possibility that trace amounts of Dvl protein could be sufficient for dorsal fate specification. However, several studies in zebrafish suggest a dose-dependent function of maternal Wnt/ß-catenin activity in organizer formation. In ichabod embryos with reduced ß-catenin2 level, dorsal and anterior deficiencies occur with variable expressivity [37], and knockdown of ß-catenin2 increases the severity of ichabod phenotypes [38]. By contrast, MZdvl2;MZdvl3a embryos displayed correct organizer gene expression at the onset of zygotic transcription, suggesting that maternal Wnt/ß-catenin signaling should not be affected by the deficiency of Dvl activity. Thus, our present results support the model that early dorsal axis formation is a consequence of dorsal accumulation of ß-catenin caused by asymmetrical translocation of vegetally localized dorsal determinants [1][2][3][4][5]. They suggest that Dvls may be dispensable for the activation of maternal Wnt/ß-catenin signaling in dorsal fate specification. Nevertheless, there may be a possibility that maternal Wnt/ß-catenin signaling is activated by other mechanisms that are independent of maternal Wnt ligand/receptor signaling.
In Xenopus, Wnt11 was shown to be required for the activation of maternal Wnt/ß-catenin signaling in dorsal axis formation [49]. In zebrafish, vegetally localized maternal wnt8a is transported to the dorsal region and has been thought to play a role in specifying dorsal fate [50]. In this regard, it would be interesting to analyze the maternal effect following removal of the bicistronic wnt8 locus [51]. While this work was in progress, it was reported that maternal mutants for the two wnt8a open reading frames did not show axis formation defect [52]. Thus, our present results are consistent with this observation, and in particular, the trunk and posterior deficiencies in MZwnt8a mutants are much similar as those observed in our MZdvl2; MZdvl3a mutants. However, it is worth to mention that MZwnt8a mutants show dorsalized phenotype [52], whereas MZdvl2;MZdvl3a mutants exhibit CE defects with strongly reduced ventroposterior gene expression, without obvious dorsalizing effect. There are at least two credible explanations that may account for these differences. First, zygotic Wnt8a only activates Wnt/ß-catenin signaling and participates in ventral and posterior tissue formation, but not in CE movements. Second, extracellular Wnt8a ligand should also function to antagonize organizer-secreted Wnt inhibitors during gastrulation, and its absence leads to an expansion of organizer activity [5]. However, the absence of maternal and zygotic Dvl2 and Dvl3a should not disturb this functional antagonism, thus keeping early dorsal fate unaffected. Also as a consequence, MZdvl2;MZdvl3a mutants do not show anteriorization at late stages.
It is well established that Dvl plays a critical role in mediating the activation of Wnt/PCP signaling in CE movements during gastrulation [17]. Several studies in mice suggest that functional redundancy exists between Dvl proteins, and that the Wnt/PCP pathway during neurulation is more readily affected following reduction of Dvl dosage [25][26][27]. However, it is still unclear how Dvl dosage influences CE movements during gastrulation, and how Dvl proteins functionally interact in these processes in zebrafish. Our analyses by using single and double mutants clearly indicate that Dvl2 plays a predominant role in embryonic axis extension. There is only a partial functional redundancy between Dvl2 and Dvl3a, because MZdvl2 mutants display obvious CE defects, whereas MZdvl3a mutants are phenotypically normal. Nevertheless, reducing Dvl3a dosage in zygotic dvl2 mutants sensibly aggravates the defective CE phenotypes, indicating that Dvl3a exerts a permissive effect on Dvl2 in Wnt/PCP signaling. In mice, Dvl2 -/-;Dvl3 +/triallelic mutants exhibit more severely shortened AP axis than Dvl3 -/or Dvl2 +/-;Dvl3 -/mutants [26]. Taken together, our present results suggest that Dvl2 plays an important role in Wnt/PCP signaling during CE movements, which may be conserved in vertebrates.
Interestingly, mutational analyses indicate that the absence of zygotic Dvl2 and Dvl3a function only results in moderate CE defects. However, removal of both maternal and zygotic Dvl2 and Dvl3a produces the most severely affected CE phenotypes, indicating clearly an important maternal contribution of these proteins in Wnt/PCP-mediated CE movements. This is further confirmed by analyzing the phenotypes of maternal dvl2 and dvl3a double mutants, which show severe axis extension defect during gastrulation. Thus, by comparison of the extent of axis extension defect between Zdvl2;Zdvl3a, Mdvl2;Mdvl3a, and MZdvl2;MZdvl3a at different stages, it can be concluded that, to a large extent, maternal Dvl2 and Dvl3a may be sufficient to activate Wnt/PCP signaling in CE movements during gastrulation, whereas zygotic Dvl2 and Dvl3a are implicated to a lesser extent. Consistently, MZdvl2;MZdvl3a mutants display strongly reduced ability to activate the AP1 reporter, which monitors JNK activation [44], and the disrupted cell polarity can be rescued by a constitutively active JNK. However, the activity of the AP1 reporter was not completely blocked in MZdvl2;MZdvl3a mutants at late gastrula stages. This may be due to an independent activation of JNK signaling by other proteins such as the paraxial protocadherin that regulates morphogenesis and signals through the small GTPases RhoA and Rac1 to JNK [53,54]. Altogether, our analyses indicate that Wnt/PCPmediated CE movements are particularly sensitive to both maternal and zygotic Dvl dosages.
Another striking observation is the requirement for maternal Dvl function in AP patterning that is dependent on zygotic Wnt/ß-catenin signaling to activate region-specific gene expression. It is well established that an endogenous Wnt/ß-catenin signaling gradient, with highest activity in the posterior region, is important for AP patterning [3,5]. At present, there is only limited evidence implicating Dvl in AP axis specification. In Xenopus, overexpression study shows that graded amounts of Dvl elicit AP fates in the prospective ectoderm [55], suggesting that Dvl dosage may be important to differentially activate zygotic gene expression along the AP axis. This is supported by the present observation showing an implication of Dvl2 and Dvl3a in this process in a dosage-dependent manner. We find that progressive reduction of Dvl dosage gradually elicits AP patterning defect, ranging from posterior deficiency to a complete lack of trunk and tail. The maternal contribution of Dvl2 and Dvl3a is clearly evident. While Zdvl2;Zdvl3a mutants display a relatively normal AP axis, Zdvl2;MZdvl3a mutants begin to show caudal truncation. The most severe defect of AP patterning in MZdvl2;MZdvl3a mutants is clearly caused by a strongly impaired Wnt/ß-catenin signaling, which results in a severely decreased expression of target genes.
Maternal and zygotic Wnt/ß-catenin signaling has opposite functions in the specification of embryonic axes. While maternal Wnt/ß-catenin signaling specifies dorsal fate, zygotic Wnt/ßcatenin signaling induces ventroposterior mesoderm and inhibits anterior development [5]. This apparently contradicts with the requirement of maternal Dvl in AP axis specification, but it could be explained by the region-specific expression and regulation of other components of the Wnt/ß-catenin pathway. Following zygotic gene activation, the ventral region expresses several Wnt ligands. Thus, maternal Dvl proteins may serve to relay extracellular Wnt signals for region-specific activation of the pathway. Another possibility is that maternal Dvl may play a role in the specification of AP cell fates in the prospective ectoderm, as observed in Xenopus embryo [19,55]. In overexpression experiments, high levels of Xdsh (Dvl2) activate posterior neural markers, whereas low levels induce the expression of anterior neural genes. Accordingly, in our zebrafish dvl mutants, deficiency of posterior and trunk tissues could be obtained by substantially reducing maternal Dvl dosage.
In the present study, we have revealed a predominant role for Dvl2 dosage both in CE movements and in AP patterning, however, it is clear that Dvl3a also cooperates with Dvl2 in these processes. Our results are consistent with previous studies indicating that Dvl proteins differentially activate the Wnt pathways and regulate distinct developmental processes. Indeed, the three mammalian DVL proteins differentially mediate the activation of Wnt/ß-catenin signaling in cultured cells [56]. In Xenopus, knockdown of dvl1 and dvl2 causes severe neural crest defects, while knockdown of dvl3 affects muscle gene expression and sclerotome development [21]. In this regard, it is of interest to note that MZdvl2 mutants develop craniofacial defects that at least partially result from fusion or absence of neural crest-derived cartilages. In addition, the heart abnormality in dvl2 and dvl3a double mutants is consistent with previous observations showing that Dvl2 mutant mice display defects in cardiac neural crest development [25]. At present, it is still intriguing that why Dvl2 plays a major role both in Wnt/ß-catenin and Wnt/PCP signaling, and how it distinguishes these two pathways? Our previous structural and functional analyses have provided some clues as to how Dvl2 activity in Wnt/ PCP signaling is regulated by its C-terminus [57,58]. The present observations further indicate that the activaty of Wnt/ß-catenin or Wnt/PCP signaling during development may be regulated by Dvl dosages.
In summary, our findings have uncovered, to a significant extent, the manner in which these Dvl proteins are implicated in regulating the activation of different Wnt signaling pathways. In particular, we clarified that they are not required for dorsal fate specification, and demonstrated that maternal and zygotic Dvl2 dosages, in cooperation with Dvl3a, play a predominant role in regulating important zygotic events, such as AP patterning and morphogenetic movements.

Ethical statement
All experiments using zebrafish adults and embryos were performed according to the ARRIVE guidelines and approved by the Ethics Committee for Animal Research of Life Science of Shandong University (permit number SYDWLL-2018-05).

Zebrafish and microinjections
Zebrafish adult of the AB strain were maintained at 28.5˚C. The embryos were staged as described [59], and for most experiments, were injected at 1-cell stage in the yolk using a PLI-100A Picoliter microinjector (Harvard Apparatus).

Targeted gene mutations and generation of MZdvl2;MZdvl3a mutants
The two TALEN mRNAs were mixed at equal amounts (200 pg each) and injected into 1-cell stage embryos. The targeting efficiency was determined by Sanger sequencing of PCR products amplified from genomic DNA extracted from 15 randomly selected F0 embryos at 24 hpf. When the result indicates a positive targeting effect, other embryos were reared to adulthood for outcross to screen F1 heterozygotes using genomic DNA extracted from the tail fin.
To generate MZdvl2;MZdvl3a mutant lines, we first used germline replacement approach by transplanting blastoderm cells from Zdvl2 donors at dome stage into dvl2 +/-;dvl3a -/hosts. Since this only generated 21 male chimera fish, we next used a strategy to target the remaining dvl2 WT allele in dvl2 +/-;dvl3a -/embryos. The offspring obtained from crosses between dvl2 +/-;dvl3a -/carriers were injected at 1-cell stage with dvl2 TALEN mRNAs (100 pg each) and raised to adulthood. Due to the mosaic distribution of TALEN mRNAs and incomplete targeting efficiency, mosaic fish with mixed dvl2 +/-;dvl3a -/and dvl2 -/-;dvl3a -/genotypes could be obtained, and denoted as mdvl2 +(-)/-;dvl3a -/-, for mosaic homozygous dvl2 mutations. Female mdvl2 +(-)/-;dvl3a -/fish were then crossed with male dvl2 +/-;dvl3a -/-, and the resulting offspring were screened by PCR using allele-specific primers (S1 Table), followed by sequencing to detect de novo mutations in the dvl2 allele. The offspring that contained a new indel along with the original indel were MZdvl2;MZdvl3a mutants, and the parental fish were selected for further experiments. To obtain Mdvl2;Mdvl3a mutants, female mdvl2 +(-)/-;dvl3a -/fish were crossed with male WT fish. The resulting dvl2 and dvl3a heterozygous offspring that carry either the original indel or a new indel in one dvl2 allele were devoid of maternal Dvl2 and Dvl3a gene products.

Cartilage staining
Larvae at 5 dpf were fixed in 4% paraformaldehyde at 4˚C overnight, then washed twice in PBS for 10 minutes. The larvae were incubated in alcian blue solution (0.37% HCl, 70% ethanol, 0.1% alcian blue) for 6 hours, and washed in destaining solution (1% HCl, 70% ethanol). Following dehydration in ethanol, the larvae were cleared in benzyl benzoate, and imaged using an upright microscope (Leica DM2500).

Live time-lapse imaging
Embryos at 90% epiboly stage were mounted in a cavity microscope slide in 1% low-melting agarose as described [43]. Cell movements were recorded using an upright light microscope (Leica, LM2500) equipped with a CCD digital camera (Leica, IC180), under differential interference contrast. The embryos were imaged every 30 seconds for a period of 60 minutes, and mutant embryos were then subjected to genotyping by Sanger sequencing. Time-lapse movies were generated using ImageJ software (NIH Image).

Analysis of cell polarity
At 64-cell stage, a single marginal cell was injected with a mixture of Histone2B-RFP (50 pg) and mGFP (100 pg) mRNAs, with or without caJNK mRNA (200 pg). At 12 hpf, mosaically labeled embryos were dechorionated and placed on a microscope slide in a drop of Ringer's solution. The yolk was removed, and the embryos were flat mounted with neuroectoderm facing upward. Following image acquisition using an upright fluorescence microscope (Leica LM2500), the embryos were subjected to genotyping.

Semi-quantitative and quantitative RT-PCR, and RNA sequencing
Total RNA was reverse transcribed using M-MLV reverse transcriptase (Invitrogen). Semiquantitative PCR was performed using gene-specific primers, with ß-actin as a loading control (S1 Table). The intensity of PCR products was analyzed using the Lane 1D software (Sagecreation). Quantitative PCR was performed using Quant qRT-PCR Kit (Tiangen) with gene-specific primers (S1 Table). RNA sequencing was performed on Illumina HiSeq 2000, using 12 hpf mRNA libraries constructed by TruSeq RNA Library Preparation Kit. The data were aligned and analyzed as described [65].

PCR-based genotyping
Single embryo was placed in an Ependorf tube containing 40 μl of lysis buffer (10 mM Tris-HCl, pH 8.0, 2 mM EDTA, 0.2% Triton X-100, 100 μg/ml proteinase K), and homogenized by pipetting. The tube was heated at 50˚C for 2 hours, then at 94˚C for 10 minutes. After a brief centrifugation, 1 μl of the solution was used for PCR reaction.

Luciferase assays
WT and mutant embryos at 1-cell stage were injected with 50 pg TOPFlash or AP1 reporter DNA, along with 5 pg pRL-TK DNA as an internal control. Fifteen to twenty embryos at 12 hpf were manually dechorionated and lysed in 60 μl lysis buffer (Promega). The lysate was clarified by centrifugation and luciferase activities were measured using the Dual-Luciferase1 Reporter Assay System (Promega), according to the manufacturer's instruction. The values were normalized with respect to Renilla luciferase activities, and the value in control condition was set as 1, and expressed as relative luciferase activity.

Statistical analysis
All data were obtained from at least three independent experiments, and analyzed using paired Student's t test. dvl3a -/fish, when the germline of the female fish contains a novel mutation for dvl2, the gametes that it produces will give rise to an MZdvl2;MZdvl3a offspring when fertilized by a male gamete with the original mutation, and an Mdvl2;MZdvl3a offspring when fertilized by a male gamete with WT dvl2 allele. The same female gametes will give rise to an Mdvl2;Mdvl3a offspring when fertilized by a male gamete from WT fish. (B) When a female mdvl2 +(-)/-;dvl3a -/fish is crossed with a male dvl2 +/-;dvl3a -/fish, if the germline of the female fish only contains the original mutation, the genotypes of offspring should include dvl2 +/-;MZdvl3a -/-, MZdvl3a, or Zdvl2;Mdvl3a zygotes, depending on the genotype of the male gamete. These female gametes will give rise to Mdvl3a offspring when fertilized by a male gamete from WT fish. In all these cases, the early embryos should lack half of the dvl2 gene product. Only the dvl2 alleles are indicated in the schema. (JPG) S12 Fig. Development of cyclopia in dvl2 and dvl3a mutants. (A-E) Ventral view of representative images of normal and different degrees of eye phenotypes at 3 dpf. (F) Quantitative analysis of different degrees of eye phenotypes in indicated mutants. Except for WT embryos, all mutants were analyzed from three independent crosses using the same fish pair (indicated below the horizontal line). Numbers on the top of each column indicate total embryos carrying the indicated genotypes (above the horizontal line). Scale bar: (A-E) 400 μm. (JPG) S13 Fig. Analysis of dvl1a, dvl1b, and dvl3b  The embryos were imaged at 11.5 hpf and 30 hpf, followed by genotyping, and the extent of axis extension defect was reflected by the angle between the anterior end and the posterior end at 11.5 hpf. (A, B) WT embryos. (C, D) Mdvl2;Mdvl3a mutants from crosses between female mdvl2 +(-)/-;dvl3a -/fish and male WT fish were genotyped for the presence of a novel indel (red) along with a WT allele in the dvl2 locus. They have dvl2 and dvl3a heterozygous mutations. (E, H) The offspring with the two possible genotypes (dvl2 +/+ ;dvl3a +/and dvl2 +/-;dvl3a +/-; the effects of these mutations are indicated in parenthesis), derived from a cross between female dvl2 +/-;dvl3a -/fish and male WT fish, are maternal mutants for dvl3a, with a reduced dosage of maternal dvl2 in both cases, despite of the genotype. (I) Statistical analysis of the extent of axis extension delay in three types of maternal mutants. Bars represent the mean ± s.d. from indicated numbers of embryos, and asterisks above the bars show significance with respect to WT embryos ( ÃÃÃ , P<0.001; ÃÃÃÃ , P<0.0001). (J) Quantitative analysis of defective axis extension at 11.5 hpf. Each type of cross was done using three independent fish pairs, and total numbers of embryos ana-