Wnt5b integrates Fak1a to mediate gastrulation cell movements via Rac1 and Cdc42

Focal adhesion kinase (FAK) mediates vital cellular pathways during development. Despite its necessity, how FAK regulates and integrates with other signals during early embryogenesis remains poorly understood. We found that the loss of Fak1a impaired epiboly, convergent extension and hypoblast cell migration in zebrafish embryos. We also observed a clear disturbance in cortical actin at the blastoderm margin and distribution of yolk syncytial nuclei. In addition, we investigated a possible link between Fak1a and a well-known gastrulation regulator, Wnt5b, and revealed that the overexpression of fak1a or wnt5b could cross-rescue convergence defects induced by a wnt5b or fak1a antisense morpholino (MO), respectively. Wnt5b and Fak1a were shown to converge in regulating Rac1 and Cdc42, which could synergistically rescue wnt5b and fak1a morphant phenotypes. Furthermore, we generated several alleles of fak1a mutants using CRISPR/Cas9, but those mutants only revealed mild gastrulation defects. However, injection of a subthreshold level of the wnt5b MO induced severe gastrulation defects in fak1a mutants, which suggested that the upregulated expression of wnt5b might complement the loss of Fak1a. Collectively, we demonstrated that a functional interaction between Wnt and FAK signalling mediates gastrulation cell movements via the possible regulation of Rac1 and Cdc42 and subsequent actin dynamics.


Introduction
Vertebrate gastrulation is vital to establish germ layers and body axes by coordinated cell movements in zebrafish [1,2]. At mid-gastrulation, the prospective mesendoderm cells are internalized at the blastoderm margin to form the hypoblast [3,4]. Hypoblast cells migrate on the dorsal epiblast via tightly regulated cell-cell and cell-extracellular matrix (ECM) adhesions and move anteriorly to become mesendodermal layers. Meanwhile, ventral and lateral cells converge and extend anterior-dorsally to elongate and narrow the germ layers to establish the anterior-posterior body axis [5].
Focal adhesion kinase (FAK) is a non-receptor tyrosine kinase and is central to the regulation of cell movements and cell-ECM adhesions [6,7]. It is autophosphorylated upon activation by integrin, growth factor stimuli and/or FAK is evolutionarily conserved in mammals and lower eukaryotic organisms [12]. It is highly expressed during embryogenesis in zebrafish [13], African clawed frog [14], chicken [15] and mouse [16]. FAK has a vital role in embryogenesis [17]. FAK-null mice die at embryonic day 8.5 with defects in the axial mesoderm and cardiovascular system [18,19]. These abnormalities are similar to those of fibronectin-and integrinknockout mice that may be due to a disorganized cytoskeleton [20], impaired cellular migration and adhesion [21], or cessation of mesodermal cell proliferation during embryogenesis [22]. FAK is critical for epiboly, cell polarity and intercalation during embryogenesis in Xenopus laevis [23][24][25]. Other cell signalling pathways such as Wnt signalling are also involved in gastrulation [26]. How those different signals are coordinated during gastrulation remains unclear.
The non-canonical Wnt pathway, which mediates planar cell polarity (PCP) via Frizzled or the tyrosine kinase (Ryk)related receptor, is a permissive cue for cell migration during development [27,28]. Non-canonical Wnt, wnt11 and wnt5b mutants are defective in cell migration during gastrulation in zebrafish [29,30]. Knockdown of wnt5a expression reduces focal adhesion dynamics by affecting FAK phosphorylation in cellular assays [31]. However, zebrafish wnt5a messenger RNA (mRNA) gradually decreases within 4 h post-fertilization (hpf). By contrast, another isoform of wnt5, wnt5b, is highly expressed at the margin of the epiblast during epiboly when fak is also expressed [32,33]. This implies a possible interaction between Fak and Wnt5b during gastrulation.
Two FAK paralogous genes, fak1a and fak1b, are identified in the zebrafish genome [34]. A comparative study showed that the fak1b locus is a duplicate of fak1a locus and both of them share syntenies with the site of the human FAK locus. High conservation of many important protein domains and 69% identity through the peptide sequence indicates partially overlapping functions of fak1a and fak1b [34]. Zebrafish Fak1a and Fak1b are highly similar to mammalian FAK (electronic supplementary material, figure S1). Using antisense morpholino (MO) oligonucleotides, we observed Fak1a MO caused severe gastrulation phenotypes than that of fak1b MO in zebrafish embryos. Primary amino acid sequence and phylogenetic analyses also showed Fak1a rather than Fak1b is more related to human FAK (electronic supplementary material, figure S2); thus, here we primarily focus on the role of Fak1a and its interaction with Wnt5b during gastrulation.
In this study, we show that the loss of Fak1a impairs gastrulation cell movements via regulation of actin dynamics in zebrafish. Fak1a was found to collaborate with Wnt5b in controlling Rac1 and Cdc42 activities to mediate gastrulation cell movements. We generated several fak1a zebrafish mutant alleles; however, those alleles only exhibited mild gastrulation defects. Interestingly, a subthreshold level of the wnt5b MO could induce a more severe gastrulation phenotype in the fak1a mutants. This suggests Wnt5b may compensate the loss of Fak1a in the mutants.

Zebrafish fak cloning and expression vector construction
Coding sequences (CDSs) of fak1a and fak1b were amplified from total zebrafish complementary (c)DNAs by a reverse transcription-polymerase chain reaction (RT-PCR). Total RNAs of zebrafish were isolated by Trizol (Invitrogen) and reversetranscribed using MMLV reverse transcriptase (Promega, Madison, WI, USA). Primers were designed according to reference RNA sequences (NM_131796.1 and NM_198819.1). The PCR was carried out at 95°C for 5 min followed by amplification at 95°C for 30 s; 58°C for 30 s; and 72°C for 30 s for 30 cycles with the proper primer set (electronic supplementary material, table S1). PCR products were cloned into the pGEM-T vector (Promega), sequenced and analysed. The correct fak1a CDS was subcloned into pEGFP, pKH3 or pCS2+ vectors for the overexpression experiments.

Immunofluorescence cell staining
Cells grown on a sterile cover glass were fixed in 4% paraformaldehyde (PFA) for 15 min. The cover glass was thoroughly washed with phosphate-buffered saline (PBS), blocked with 10% FBS and 0.3% Triton X-100, and then incubated overnight at 4°C with different primary antibodies against FAK, human influenza hemagglutinin (HA) and enhanced green fluorescent protein (EGFP; Santa Cruz Biotechnology, Santa Cruz, CA, USA). Samples were incubated with their corresponding secondary antibody and then stained by Hoechst 33258 dye (Sigma-Aldrich, St Louis, MO, USA) to reveal nuclei. Fluorescence signals were recorded with an epifluorescence microscope (Model IX71, Olympus, Japan) and analysed using Image-Pro Plus software (Media Cybernetics, Silver Spring, MD, USA).

5-bromodeoxyuridine (BrdU) incorporation assays
Cells were serum-starved for 24 h in DMEM with 0.5% FBS to arrest them in G0 phase and then incubated for 24 h in DMEM containing 10% serum and 150 µM BrdU (Sigma). Cells were then washed with PBS, fixed in 4% PFA, permeabilized with 0.3% Triton X-100, digested with DNase I (New England Biolabs, Ipswich, MA, USA), blocked with 10% FBS, and processed for immunofluorescent staining with anti-BrdU (1 : 200, Sigma) and anti-GFP (1 : 200) antibodies, as described previously [35]. Cells counted in multiple fields were scored for BrdU-positive cells in three independent experiments.

Cell migration assay
Cell migration was measured in a 48-well chemotaxis Boyden chamber using 8 μm polyvinyl pyrrolidone-free polycarbonate membranes as described previously [18]. The membrane was placed over the bottom chamber, and cells were loaded in the upper well of a chamber. After 6 h, the membrane was stained with crystal violet. Finally, the number of stained cells was calculated from five randomly selected fields of each well under a microscope at 20× magnification.

Cell transplantation
For transplantation preparation, donor embryos were injected with 1% rhodamine dextran and 5 ng of standard control MO (StdMO) or fak1a tMO 1 , respectively. Cells from donor embryos were transplanted into untreated embryos and fak1a morphants separately at sphere stage as described previously [36]. Transplanted cells were further monitored and recorded by using a Leica DM5000 B microscope with a charge-coupled device (CCD) camera.

Zebrafish maintenance and embryo culture
AB zebrafish were maintained at 28.5°C on a 14 h light/10 h dark cycle. Embryos collected from natural mating were maintained and staged according to Kimmel et al. [37] as hours post-fertilization.

Microinjection
Reagents were diluted with Danieau's buffer and phenol red to desired concentrations. All reagents were injected at the 1-cell stage unless otherwise stated. Embryos were immobilized on a 1% agar plate, and an injection pipette was forced into the area adjacent to the blastomeres. A solution of the desired volume was injected using a Nanoliter injector (World Precision Instrument, Sarasota, FL, USA).

Rhodamine-phalloidin staining and confocal imaging analysis
Fixed embryos were washed with PBS, blocked with 10% FBS, treated with 0.3% Triton X-100 in PBS and then incubated with rhodamine-phalloidin (Invitrogen) overnight at 4°C. After washing with PBS, nuclei were labelled with 4 0 ,6-diamidino-2-phenylindole (DAPI, Sigma) and mounted. Fluorescence signals were recorded with the Zeiss LSM780 Confocal Microscope Imaging System (Carl Zeiss, Jena, Germany). The z-stack analysed a total of 132 µm thickness with a 1.5 µm interval.

Whole-mount in situ hybridization
Embryos were fixed with fresh 4% paraformaldehyde (PFA) overnight and manually dechorionated. WISH was performed using DIG-labelled, SP6-RNA polymerase-made riboprobes according to a previous study [38]. Stained embryos were mounted in cellulose and observed under a stereomicroscope system (Mz75, Leica Microsystems). Photographs were taken using a digital camera (Coolpix 995, Nikon, Melville, NY, USA).

Measurement of degrees in convergence and extension
Convergent extension during gastrulation was assessed by measuring changes in the angles formed by respective expression patterns of marker genes [39,40]. Briefly, WISH against cathepsin L 1b (ctsl1b) and distal-less homeobox 3 (dlx3) were used to label prechordal plate and ectoderm borders, respectively, which formed a V-shaped pattern. We measured the angle of the V-shape to estimate the level of convergence. We also used WISH against ctsl1b and no tail (ntl) to mark the prechordal plate and presumptive notochord, respectively, to evaluate the dorsal axis extension in the bud stage. We enclosed each embryo within a rectangle, with the crossing of two diagonals being set to the exact middle of an embryo. A line was drawn from the anterior front of the prechordal plate to the centre of the embryo and then its centre was connected to the posterior ends of the tail bud. The angle towards the dorsal side was measured to estimate the length of the dorsal body axis, which was an indication of dorsal extension.

Time-lapse DIC imaging and analysis
Dechorionated embryos were mounted with 1% low-meltingpoint agar, and cell migration in the prechordal site and lateral site were recorded by the DIC system with the Leica DM5000B system under a 63× water immersion objective during gastrulation. The migration process was recorded at 10 s intervals for 10 min and analysed with Simple PCI Imagine System software (Compix, Sewickley, PA, USA) to track migration characteristics (migrating rate, velocity, cell protrusion numbers, tortuosity and direction of cell migration).

RacI and Cdc42 activation assay
The levels of activated RacI and Cdc42 were determined using the RacI and Cdc42 G-LISA Activation Assay kit (Cytoskeleton, Denver, CO, USA) according to the manufacturer's instructions. Embryos were cultured to the bud stage and dechorionated by proteinase. In total, 60 embryos were collected in a 1.5 ml pre-chilled tube and lysis buffer added. Embryos were homogenized using a 24-gauge needle, briefly royalsocietypublishing.org/journal/rsob Open Biol. 10: 190273 centrifuged at 10 000g for 1 min, and then immediately immersed in liquid nitrogen. Protein lysates were applied to G-LISA plates using a RacI and Cdc42-specific antibody for detecting the captured active small G-protein according to the manufacturer's directions to determine the absorbance at 490 mm.

Hae III mutagenesis assay and sequencing
The genomic DNA from adult tail fins or embryos was extracted by TE buffer with 0.2 mg ml −1 proteinase K, incubated at 55°C for 2 h, and inactivated at 85°C for 10 min. A 513 bp DNA fragment was amplified by a PCR using primers as shown in electronic supplementary material, table S1. PCR products were cleaned using a PCR clean-up kit (Qiagen, Hilden, Germany) and digested with Hae III (New England Biolabs, Ipswich, MA, USA) for 2 h at 37°C. Digested products were examined by electrophoresis in a 1% agarose gel.

Statistical analysis
All datasets are presented as the mean ± s.e.m. Results were further analysed by a one-way analysis of variance and post-tested using Tukey's multiple-comparison tests or Student's t-test.

Zebrafish Fak1a is a functional conserved homologue of human FAK
FAK is a critical factor for cell proliferation, motility and migration [42][43][44][45]. To determine whether zebrafish Fak has conserved properties, we subcloned zebrafish fak1a and its dominant-negative form, frnk1a (see the FAK domain structure in electronic supplementary material, figure S3) into a human influenza hemagglutinin (HA)-tagged pKH3 or a pEGFP-C3 vector. We successfully overexpressed them independently in 293T cells as shown by western blotting against HA or GFP (figure 1a) and found that fak1a and frnk1a were co-localized with paxillin in focal contacts (figure 1b). Cell proliferation was significantly enhanced in fak1a-overexpressing cells, but not in frnk1a-overexpressing or control pEGFP-transfected cells (figure 1c; p < 0.001). Similarly, cell transmigration activity was higher in fak1a-overexpressing cells (figure 1d; p < 0.001).
These results suggest that zebrafish Fak1a is functionally conserved compared to other FAK orthologues.
To examine whether key phosphorylation sites are conserved in zebrafish Fak1a, the HA-tagged fak1a and frnk1a were overexpressed in FAK −/− cells, immunoprecipitated and immunoblotted. As expected, we found that Fak1a, but not Frnk1a, was autophosphorylated at Y397 (electronic supplementary material, figure S4A). The phosphorylation of Fak1a, but not Frnk1a, was notably enhanced at Y397, Y576/ 577, S732 and Y861 as it was in chicken orthologues, thereby demonstrating that these sites are biochemically conserved on Fak1a (electronic supplementary material, figure S4B).

Fak1a and fak1b are maternally and ubiquitously expressed in zebrafish embryos
To determine the bona fide expression pattern of fak1a during zebrafish embryogenesis, fak1a mRNA was detected by whole-mount in situ hybridization (WISH) analysis. Results showed that fak1a mRNA was maternally expressed during the 1-cell stage and ubiquitously expressed through blastulation and gastrulation (electronic supplementary material, figure S5A). In addition, fak1a mRNA and protein were detected by RT-PCR (electronic supplementary material, figure S5B) or immunoblotting (electronic supplementary material, figure S5C), respectively. We observed that fak1a mRNA and protein were constantly expressed in embryos from 30% epiboly to 17-somite stages. Furthermore, we also examined the expression patterns of another fak isoform, fak1b. fak1b was constantly expressed from 1-cell stage to 5 days post-fertilization by RT-PCR analysis (electronic supplementary material, figure S6A). Spatially, like fak1a, it was ubiquitously expressed during early embryogenesis and localized to the anterior head region at later development by WISH analysis (electronic supplementary material, figure S6B).

Loss of Fak1a causes gastrulation defects
To assess the developmental role of Fak1a, we microinjected 1-cell-stage zebrafish embryos with or without an antisense translation blocking MO (tMO 1 ) targeting fak1a. The Fak1a protein was significantly reduced to approximately 30% in royalsocietypublishing.org/journal/rsob Open Biol. 10: 190273 fak1a tMO 1 -injected embryos compared to untreated ones (figure 2b). MO-injected embryos are called morphants hereafter. In fak1a morphants, embryos displayed a cylinder-like shape compared to the normal ball-like form. The progression of epiboly in these morphants was notably slower during gastrulation. The progression of epiboly was either delayed and eventually reached 100% epiboly (epiboly delay), or arrested. The movements of the deep cell layer (DCL) and enveloping layer (EVL) were asynchronous (figure 2c). Overall, we found 47% and 18% of fak1a tMO 1 morphants had delayed and arrested epiboly, respectively (figure 2d). Among the arrested embryos, 87% of them did not reach 100% epiboly at 16 hpf. This was not due to the toxicity of the MO, since no epiboly defect was observed in embryos injected with 10 ng of the N-25 control MO (electronic supplementary material, figure S7). To further test the specificity of the phenotype, fak1a mRNA lacking the target site of the fak1a tMO 1 was prepared and injected with or without fak1a tMO 1 . Relatively lower percentages (approx. 30%) epiboly defective embryos were observed in embryos injected with 125 pg fak1a mRNA, but it significantly rescued the epiboly defect when co-injected with the fak1a tMO 1 (figure 2c, left). The injection of fak1a mRNA at a higher dose (250 pg) further reduced the epiboly defects observed in fak1a morphants (electronic supplementary material, figure S7). The rescued embryos appeared mostly normal compared to control embryos (electronic supplementary material, figure S8). In addition, we also found dose-dependent epiboly defects in embryos injected with dominant-negative frnk1a mRNA (figure 2c, right). Furthermore, we investigated how the loss of fak1a affecting its downstream effectors and found that the levels of phosphorylation of paxillin, p38 mitogen-activated protein kinase (p38) and Akt were all reduced in fak1a morphants (figure 2d).
To ensure the MO-induced epiboly defects are specific to the loss of Fak1a, we designed and tested a second translational blocking fak1a MO (tMO 2 ), which does not overlap the tMO 1 target site, and found the fak1a tMO 2 also caused 70% delayed and 5% arrested epiboly (electronic supplementary material, figure S9A, B). Effects of the fak1a tMO 2 were also dose-dependent and could be rescued by the co-injection of fak1a mRNA (electronic supplementary material, figure S9C). Lastly, the standard control MO (figure 2c) and a random control MO (N-25) (data not shown) were both tested at up to 10 ng per embryo with no notable effect on epiboly or gastrulation. Therefore, untreated embryos were used as controls hereafter unless otherwise specified. Furthermore, the fak1a tMO 1 was relatively more effective, thus we used it for the rest of the experiments. Paxillin

Loss of Fak1a causes reduction in cortical actin fibres and uneven distribution of YSL nuclei
Filamentous (F)-actin forms a ring-like structure at the leading edge of the EVL, where it links the EVL and the yolk cell to provide the pulling force for blastoderm during epiboly [46] and is pivotal for gastrulation [36,47]. Thus, we examined the organization of F-actin of 75%-epiboly-stage embryos by probing with rhodamine-phalloidin. In addition, we also counterstained cell nuclei by 4 0 ,6-diamidino-2-phenylindole (DAPI) and Fak1a by immunostaining. Immunostaining showed the ubiquitous presence of Fak1a in blastodisc and YSL in control embryos, but the Fak1a staining was dramatically reduced in fak1a tMO 1 morphants (figure 3a). This further confirmed the efficacy of knockdown effect by the fak1a tMO 1 . The staining also clearly marked the EVL and DCL as indicated by arrows. In control embryos, the EVL and the DCL migrated coherently with a 92.8 ± 14.3 µm gap in between. By contrast, the gap was significantly increased to 243.6 ± 33.3 µm in fak1a tMO 1 morphants (figure 3b). Only about 20% of embryos with a gap smaller than 100 µm that indicates epiboly was significantly disturbed in most tMO 1 morphants (figure 3c, p < 0.001). Furthermore, by examining the region flanking the margin of the EVL at a higher magnification, we found that F-actin fibres were not tightly distributed in fak1a tMO 1 morphants compared to control embryos. Rough and uneven blastoderm margins were observed in fak1a morphants. This suggests that the loose actin network may lead to un-synchronized movements of the EVL and DCL in fak1a tMO 1 morphants.
To reveal the formation of the junction between the blastoderm and actin cap, we further examined the structure of F-actin in 60%-epiboly-stage embryos. YSL nuclei were evenly scattered at the EVL margin in untreated embryos, but unevenly distributed in fak1a morphants (figure 3d, blue nuclei labelled by yellow asterisks). Uneven blastoderm margin and distribution of F-actin were also found in fak1a tMO 1 morphants. We also observed previously reported actin-based structures, called actin bundles, in the vegetal cortex of the yolk cell that connect the vegetal cortex with the external yolk syncytial nuclei (eYSN) in a vegetal-toanimal direction (indicated by asterisks in figure 3d; Li et al. [48]). A significant loss of actin bundles along with a loose connection of the actin ring to the cap was seen in fak1a royalsocietypublishing.org/journal/rsob Open Biol. 10: 190273 morphants. These data indicate that Fak1a is critical to the arrangement of cortical actin during epiboly.

Loss of Fak1a causes convergence and extension defect
We further analysed whether the loss of Fak1a causes convergent extension defects in the bud-stage embryos by WISH analysis. cathepsin L 1b (ctsl1b) and distal-less homeobox 3 (dlx3) were used to label the prechordal plate and ectoderm borders, respectively. Staining gave rise to a V-shaped pattern as shown in figure 4a (left column), and the angle of the V-shape was measured to estimate the level of convergence in all embryos. The average angles of convergence were dose-dependently increased from 77.8°in untreated to 104°in 0-5 ng fak1a tMO 1 -injected embryos (figure 4b, left panel). The dose-dependent convergence defect was also observed in tMO 2 morphants (electronic supplementary material, figure S9D). royalsocietypublishing.org/journal/rsob Open Biol. 10: 190273 To evaluate the dorsal axis extension, we used WISH against ctsl1b and no tail (ntl) to mark the prechordal plate and presumptive notochord, respectively. We drew a line from the anterior front of the prechordal plate to the centre of an embryo (see the determination of centre in Material and Methods), and then connected it to the posterior end of the tail bud (figure 4a, right column). The angle towards the dorsal was used to estimate the length of the dorsal body axis, which reflects the degree of dorsal extension. The average angles of extension were 239.2°-182.4°in 0-5 ng fak1a tMO 1injected embryos. fak1a tMO 1 morphants showed significantly reduced angles compared to those in control embryos, implying an inhibition of extension (figure 4b, middle graph; p < 0.01). A similar defect in extension was also observed in tMO 2 morphants (electronic supplementary material, figure S9D).
In control embryos, individual ctsl1b-expressing cells consolidated and coordinately moved anterior to the frontal edge of the neural plate (expressing dlx3) as the tissue advanced (a) Embryos injected with a designated amount of a fak1a translation blocking morpholino (tMO 1 ) were subjected to WISH against indicated genes at the bud stage. Representative WISH staining photographs against ctsl1b/dlx3 or ctsl1b/ntl are presented for untreated and fak1a tMO 1 -treated embryos ( fak1a tMO 1 ). ctsl1b, dix3 and ntl staining were used to label the prechordal plate ( pcp), ectodermal borders (ecb) and notochord (nt)/tail bud (tb), respectively. For simplicity, tissues are labelled in untreated embryos only. To reveal dorsal convergence, embryos were probed with ctsl1b/dlx3 as shown in the anterior view (left). A round prechordal plate is in the middle with two ectodermal borders forming a V-shape. Lines were drawn along the ectodermal borders that matched the prechordal plates. The V-shape formed an angle as indicated by A°and A 0°i n untreated and fak1a tMO 1 -treated embryos, respectively. To reveal the dorsal extension, embryos were probed with ctsl1b/ntl as shown in the lateral view with the dorsal to the right (right). Lines were drawn from the anterior front of the prechordal plate and the posterior end of the tail bud to the centre of the embryo forming an angle towards the dorsal as indicated by B°and B 0°i n untreated and fak1a tMO 1 -treated embryos, respectively. royalsocietypublishing.org/journal/rsob Open Biol. 10: 190273 beyond the animal pole. In fak1a morphants, the ctsl1bexpressing cells failed to form a consolidated group and abnormally dispersed anteriorly-posteriorly (figure 4a, right column). Mesendodermal cells did not acquire the coordinated crescent shape anterior to the neural plate by the tail bud stage and exhibited a more diffused dlx3 expression. These phenomena illustrate that Fak1a is required for anterior movement and consolidation of the polster along with the simultaneous convergence and extension of posterior tissues. We also observed an expanded somite width and a shortened body axis in fak1a morphants at the 6-somite stage. The tail rudiment of fak1a morphants was either less extended or stalled at the vegetal pole ( figure 4c,e, left). These phenotypes are typical defects of convergence and extension, so we fixed both untreated embryos and fak1a morphants for WISH against myogenic differentiation 1 (myod) and ctsl1b to reveal somites and the prechordal plate, respectively (figure 4c,e, right column). The percentage of embryos with a normal somite width and body axis dose-dependently decreased in fak1a morphants ( figure 4d,f ). This further confirms the deteriorating effect on the convergent extension in the absence of fak1a.

No synergy exists between fak1a and fak1b in regulating gastrulation
To examine the role of another zebrafish Fak, Fak1b, during gastrulation, we applied two MOs, fak1b tMO 1 and fak1b tMO 2 in zebrafish embryos. In general, both fak1b MOs caused dose-dependent epiboly delay and arrest. The inhibition of fak1b tMO 2 was much weaker compared to that of fak1b tMO 1 (electronic supplementary material, figure S10A, B). In addition, both fak1b MOs resulted in significant convergent extension defects (electronic supplementary material, figure S10C,D).
To investigate the synergy of Fak1a and Fak1b during gastrulation, we co-injected fak1a tMO 1 and fak1b tMO 1 at two different subthreshold dosages into embryos. The injections of both MOs at 1.25 ng and 2.5 ng only showed additive defects during gastrulation (electronic supplementary material, figure S11). These results suggested no synergy existed between fak1a and fak1b in regulating gastrulation.  The polarity of each cell was measured as described in Results, and the analysis showed a significant loss in the polarity of the prechordal and lateral cell migration in fak1a tMO 1 -treated embryos (n ≥ 3; n ≥ 36). (d ) The migration velocity, migration rate, tortuosity (route/distance) and protrusion persistence of each recording were analysed with Simple PCI software, and comparisons between groups are shown. n.s., not significant, *p < 0.05; **p < 0.01, ***p < 0.001. royalsocietypublishing.org/journal/rsob Open Biol. 10: 190273 curvilinear distance/time) of STD > UT cells was 0.038 ± 0.004 μm s −1 (n = 59), but was only 0.025 ± 0.003 µm s −1 (n = 48) for the STD > MO cells ( p < 0.05). The MO > UT cells travelled with a faster speed and a longer distance in Vcl (0.034 ± 0.005 µm s −1 , n = 55) compared to those MO > MO cells (0.024 ± 0.002 µm s −1 , n = 77). Moreover, the STD > UT and MO > UT cells migrated straighter as shown by a higher straight-line velocity (Vsl, straight-line distance/time, figure 6f ), which were 0.020 ± 0.002 µm s −1 (n = 59) and 0.010 ± 0.002 µm s −1 (n = 48), respectively. By contrast, STD > MO (n = 55) and MO > MO (n = 77) cells migrated less linearly with a Vsl of 0.0178 ± 0.002 µm s −1 and 0.010 ± 0.001 µm s −1 , respectively, compared to cells in the UT hosts. It appeared that zebrafish Fak1a mediates gastrulation cell migration non-cell autonomously. However, it was puzzling that the number of protrusions formed in the STD > MO cells was not different from that of STD > UT cells (figure 6e). We reasoned that the directionality of protrusion might be different between groups. To determine the directionality of protrusions, the centre of a cell was considered as to be the centroid of the rose diagram. A rose diagram was divided into eight equal pies with designated angles.

No synergy between Wnt5b and Fak1a signalling during gastrulation
Zebrafish wnt5b/pipetail ( ppt) mutants exhibit similar convergence and extension defects [29,49] as that observed in fak1a morphants. The knockdown of wnt5a expression reduced focal adhesion dynamics by affecting FAK phosphorylation in cellular assays [50]. We suspected that Fak1a may interact with the Wnt5b-mediated signalling pathway during gastrulation. Hence, we tested whether fak1a overexpression can rescue the convergence and extension defects of ppt mutants. About 25% of embryos from a standard cross of heterozygous ppt mutants showed convergence and extension defects with a shortened body length and a spade tail compared to normal larvae at 48 hpf (figure 7a). The phenotypes in ppt mutants injected with fak1a mRNA were reduced to 2% (figure 7b, left; p < 0.001). Rescued ppt mutants had significantly longer trunks (averaging 1.55 mm; p < 0.001) than uninjected siblings (1.33 mm), while WT embryos had an average trunk length of 1.63 mm (figure 7b, right; p < 0.001). Homozygous Wnt mutants are embryonic lethal and the use of heterozygous mutants only produce a quarter of homozygous mutant embryos which cannot be identified until the appearance of phenotypes that is not practical for intensive analyses. Thus, we knocked down wnt5b by a specific MO, which had been shown to induce gastrulation defects independent of p53 (electronic supplementary material, figure S12) [51]. We observed approximately 75% of embryos had epiboly arrest and delay phenotypes, which were not observed in ppt mutants and presumably due to the presence of maternal Wnt5b and the co-injection of fak1a mRNA significantly rescued epiboly defects in wnt5b morphants (figure 7c). The higher penetrance of wnt5b MO-induced phenotypes allowed us to further examine the interaction of FAK and Wnt5b during gastrulation. The co-injection of fak1a mRNA significantly rescued the epiboly defects in wnt5b morphants (figure 7c). To test the synergistic effect between Wnt5b and Fak1a, we co-injected subthreshold dosages of wnt5b and fak1a MOs, but found no synergistic royalsocietypublishing.org/journal/rsob Open Biol. 10: 190273 increase in convergent extension defects (electronic supplementary material, figure S13). This suggests that Wnt5b and Fak1a might not operate through the same genetic pathway to mediate convergent extension.
To further elucidate the plausible interaction between Wnt5b and Fak1a, we examined wnt5b and fak1a expressions in embryos injected with the fak1a or wnt5b MO. At the mRNA level, fak1a expression was significantly reduced in wnt5b morphants. By contrast, wnt5b expression was increased in fak1a morphants (figure 7d). Both immunoblotting (figure 7e) and IHC analyses (electronic supplementary material, figure S14) showed downregulation of the Fak1a protein in wnt5b morphants. Unfortunately, we were unable to examine the protein level of Wnt5b due to the lack of a specific antibody.
Furthermore, both fak1a and wnt5b mRNAs could reciprocally rescue the convergence defects of their respective morphants (figure 8) and wnt5b mRNA could partially rescue epiboly arrest of fak1a morphants (electronic supplementary material, figure S15). wnt5b morphants showed no significant extension defect at the dosage tested (7.5 ng MO per embryo). By contrast, fak1a morphants had extension defects, but they could not be rescued by co-injecting wnt5b mRNA.
Collectively, the lack of synergy between Fak1a and Wnt5b suggests that they might not work via the same genetic pathway. However, the cross-rescue effect between them implies that they might converge at downstream factors to mediate gastrulation.

Wnt5b integrates Fak1a to mediate gastrulation via modulating Rac1 and Cdc42
Both FAK and Wnt signalling pathways are well known for their roles in cytoskeleton regulation. RhoA, Cdc42 and Rac1 are key mediators for actin dynamics during gastrulation in zebrafish [39,[52][53][54][55]. Thus, we hypothesized that Wnt5b and Fak1a signalling may converge to modulate small GTPase activities during gastrulation. Due to the possible antagonism between RhoA and Rac1 in coordinating F-actin organization [49,50], here we only tested whether rac1 or cdc42 mRNA could rescue wnt5b and fak1a morphant phenotypes. rac1 and cdc42 independently rescued both fak1a and wnt5b morphant phenotypes at certain dosages (electronic supplementary material, figure S16). Co-injection of 2.5 pg per embryo rac1 or cdc42 mRNA restored the percentage of normal embryos to about 30% and 45% in fak morphants, respectively (electronic supplementary material, figure S16B,D). Similarly, coinjection of 10 pg rac1 or 2.5 pg cdc42 mRNA per embryo also increased the percentage of normal embryos to about 40% and 55% in wnt5b morphants, respectively (electronic supplementary material, figure S16A,C). However, the rescue was not complete nor dose-dependent. It indicated that a good balance of different small GTPase activities may be necessary for gastrulation. Therefore, we injected subthreshold levels of rac1 and cdc42 mRNA to examine a possible synergistic rescue effect in both fak1a and wnt5b morphants. The injection of at 1.25 pg rac1 or cdc42 mRNA by themselves had minimum effects on gastrulation in WT embryos (electronic supplementary material, figure S16E) and had limited rescue effects in fak1a or wnt5b morphants. However, we observed synergistic rescue with the combination of rac1 and cdc42 to restore the percentage of normal embryos of wnt5b morphants from 22 to about 50%, which was notably higher than those of rac1-injected (35%) or cdc42-injected (38%) embryos (figure 9a; see a group photo in electronic supplementary material, figure  S17A). An even more significant synergistic rescue was observed in fak1a morphants, from 7 to 45% of normal embryos royalsocietypublishing.org/journal/rsob Open Biol. 10: 190273 13 in double rescue compared to 11 and 14% in embryos coinjected with rac1 or cdc42 mRNA only (figure 9b; see a group photo in electronic supplementary material, figure  S17A). To better understand whether expression or activity level is perturbed in fak1a and wnt5b morphants, we performed qPCR and activity assays on both morphants. The expression level of cdc42 was significantly reduced in wnt5b morphants but not in fak1a morphants. However, expression levels of rac1 were reduced in both fak1a and wnt5b morphants (figure 9c). We next measured the activities of Cdc42 and Rac1 in fak1a and wnt5b morphants by detecting the GTP-bound form of small GTPase. The Cdc42 activity was significantly lower in both fak1a and wnt5b morphants. By contrast, although Rac1 activity was also lower in both fak1a and wnt5b morphants, the reduction in wnt5b morphants was not significant (figure 9d). Collectively, these results suggested that both Cdc42 and Rac1 act downstream of Wnt5b and Fak1a during gastrulation. Lastly, we tested whether rac1 and cdc42 mRNA could rescue the loss of F-actin bundles, and found that the co-injection of rac1 and cdc42 mRNA significantly restored the number of actin bundles in fak1a morphants (figure 9e). 3.11. CRISPR/Cas9-mediated fak1a knockout causes a compensatory rise in wnt5b expression To examine the function of fak1a in vivo, we generated fak1a mutants by a clustering of regularly interspaced short palindromic repeats (CRISPR)/Cas9-mediated technique [56]. We designed two gRNAs targeting exon 3 and one gRNA targeting exon 10 of fak1a (see sequences in electronic  supplementary material, table S2). Genotyping analysis revealed that the one targeting exon 10 ( fak1a E10-1) successfully caused the insertion or deletion (indels) in F0 embryos Unexpectedly, only a portion (25%) of these embryos showed mild gastrulation defects at the bud stage but eventually grew normally at later stages (electronic supplementary material, figure S18). Similar results could be seen with two other mutant alleles, fak1a Δ5b and fak1a Δ7 (electronic supplementary material, figure S18A). fak1a Δ5 mutant embryos tended to have divergent and larger convergent angles but shorter extension angles compared to WT embryos. fak1a tMO 1 morphants appeared to have larger convergent and shorter extension angles compared to fak1a Δ5 mutant embryos. However, the effects of fak1a tMO 1 were alleviated when injected into fak1a Δ5 mutant embryos (figure 10f ). These results suggest that fak1a Δ5 mutant embryos have notably reduced sensitivity to the fak1a tMO 1 . It further demonstrates the specificity of the fak1a MO. Moreover, the injection of cas9 mRNA and fak1a gRNA into F0 embryos also resulted in gastrulation defects in about 50% of embryos, which could be rescued by co-injecting fak1a mRNA (electronic supplementary material, figure S19). Together, these data suggest that gastrulation defects observed in different contexts are all due to the specific loss of fak1a.
The lack of severe gastrulation defects observed in fak1a mutants might have resulted from gene complementation and/or compensation. One of the possible candidate genes, which might exert a complementation effect on fak1a, is its close relative, fak1b [34]. Because of the possibility that the expression of fak1b might complement the loss of Fak1a for gastrulation cell movements, we injected the fak1b MO into fak1a Δ5 mutant embryos but observed no further deterioration of gastrulation defects compared to WT embryos injected with the fak1b MO (electronic supplementary material, figure S20). We further examined changes in the gene expression of fak1b in WT embryos, morphants and mutants and found no significant change until 9 hpf, in which fak1b expression increased in both fak1a tMO 1 morphants and fak1a Δ5 mutants (electronic supplementary material, figure S21A). This suggests that the increase in fak1b expression might not explain the phenotypic difference observed between fak1a morphants and mutants.
Lastly, we injected a subthreshold dosage of the wnt5b MO at 1.875 ng, which by itself caused mild gastrulation defects in a small percentage of WT embryos, into fak1a Δ5 mutants and found a significant increase in the severe convergent extension defect and even epiboly arrest. The defects were only slightly higher when injected with 3.75 ng of the wnt5b MO (figure 10g). This indicates that an increase in wnt5b expression might occur to compensate for the loss of fak1a in fak1a Δ5 mutants.

Discussion
Despite its necessity, how FAK functions during early embryogenesis remains poorly understood. Here, we demonstrate that zebrafish Fak1a is chemically conserved with mammalian FAK and functions cooperatively with Wnt5b to regulate Rac1 and Cdc42 activities for the modulation of actin dynamics during gastrulation cell movements.
Embryos devoid of Fak1a showed abnormal epibolic movement and convergent extension. It is known that the YSL is crucial for the initiation and process of epiboly in teleosts [57,58]. By coordinating through microtubule arrays, the YSL pulls the tightly associated EVL towards the vegetal pore during epiboly [59]. In addition, actin dynamics are critical for epibolic movement. A punctuate actin band within the YSL also contributes to the progression of epiboly [46,60]. The activation of FAK can trigger F-actin remodelling and repositioning during cell migration [61]. Here, we observed that the F-actin network is severely distorted in fak1a morphants by an abnormal, uneven distribution of yolk syncytial nuclei and epiboly defects. By contrast, we found no notable differences in the perinuclear microtubule network or longitudinal microtubule arrays within the yolk syncytial layer between control embryos and fak1a morphants (see electronic supplementary material, figure S22). This further suggests an essential role of Fak1a in mediating actin dynamics, organization and YSL behaviour for gastrulation cell movements.
The deep cells of fak1a morphants showed significant retarded cell migration and reduced numbers of protrusion that have also been reported in mammalian cells [8,62]. These abnormalities might be due to the disturbance of FAK downstream signalling. One of the FAK downstream signalling molecules is mitogen-activated protein kinase (MAPK). Zebrafish defective in p38 or its downstream effector, MAPK-activated protein kinase 2 (MAPKAPK2), also have a disorganized F-actin network and epiboly defects [63]. Biochemically, we provide evidence to show that the reduced phosphorylation of p38, paxillin and Akt occurs in fak1a royalsocietypublishing.org/journal/rsob Open Biol.
513 bp R 11 10 12 morphants, which are all pivotal factors controlling cell mobility and migration [39,64]. Collectively, Fak1a signalling is mechanistically conserved in mediating gastrulation cell migration in zebrafish. Our MO-based loss-of-function assay clearly demonstrated the necessity of fak1a for gastrulation cell movements. However, only mild gastrulation defects were observed in fak1a MZ mutants. This suggests that possible secondary effects might occur in fak1a morphants. We performed the necessary MO-specificity control experiments, including using two MOs targeting non-overlapping sites, RNA rescue and coinjection of a p53 MO (data not shown), and all data supported the specific activity of the fak1a MO. A partial gastrulation defect was seen by overexpressing the dominant-negative frnk. In addition, cas9 mRNA and fak1a gRNA induced gastrulation defects, which were also partially rescued by fak1a mRNA. The injection of fak1a mRNA also caused mild gastrulation defects. Furthermore, the convergent extension defects induced by the fak1a tMO 1 were notably mild in fak1a mutant embryos compared to WT embryos, further strengthening the specificity of tMO 1 -induced gastrulation defects. These results support that perturbation of fak1a indeed interferes with gastrulation.
The lack of a phenotype is often observed in chemicalinduced zebrafish mutants [65]. Gene compensation reported in CRISPR/Cas9-induced mutants suggests it may be a common feature in zebrafish gene knockouts [66][67][68][69]. By contrast, genetic compensation does not occur in morphants, so the corresponding phenotypes can be observed. If one only trusts the phenotypes observed in genetic mutants, the important and valid phenotypes seen in the MO-based experiments may be overlooked. To settle this issue, guidelines for the use of MO in zebrafish were reported in PLoS Genetics [70] authored by a group of leading zebrafish experts. Our work exactly followed the guidelines to observe the lack of response for tMO1 in a fak1a null MZ mutant and the identification of the compensatory factor Wnt5b. We found that the injection of wnt5b MO at a subthreshold level recapitulated gastrulation defects in fak1a Δ5 mutants, suggesting that a compensatory increase in wnt5b may mask gastrulation phenotypes in fak1a Δ5 mutants. This further supports a possible integration of Fak1a and Wnt5b signalling; however, we cannot exclude the involvement of other compensatory genes. A largescale transcriptomic analysis would be required to unravel the causative gene(s).
The non-canonical Wnt5b pathway is known to modulate cell movements during embryogenesis [29,31]. Wnt5b may mediate the binding of Dishevelled and adenomatous polyposis coli (APC) to focal contacts to activate FAK at the leading edge of cells during migration [31,71]. We observed that fak1a mRNA rescued the wnt5b morphant phenotype and the expression, protein level and phosphorylation of C-terminal tyrosine were all downregulated in wnt5b morphants, suggesting that Wnt5b may act upstream of Fak1a. Interestingly, we also discovered that wnt5b expression increased in fak1a morphants, suggesting that the loss of fak1a might induce the compensatory expression of wnt5b to replenish reduced downstream signalling. However, no synergistic effect on gastrulation was observed in fak1a and wnt5b double-knockdown morphants, revealing that Fak1a and Wnt5b may not work via the same signalling pathway as previously described. Instead, Wnt5b and Fak1a might act in parallel and converging at a common downstream signalling pathway, such as small GTPases.
The small GTPases Rac1 and Cdc42 are known downstream effectors of both FAK and Wnt5 signalling [72][73][74][75]. Wnt5b is involved in non-canonical PCP and the Wnt/calcium pathway. An activated receptor recruits dishevelled and forms a complex with Daam1. The complex then binds with Rac1 to regulate actin dynamics. Wnt5b also regulates calcium release from the endoplasmic reticulum, and the increased intracellular calcium and diacylglycerol can activate Cdc42 through protein kinase C [76]. On the other hand, FAK can directly bind to p130cas to coordinate the activation of RAC1 [77]. FAK can also influence the functions of Cdc42 through the binding and phosphorylation of the Cdc42 effector Wiskott-Aldrich syndrome protein, N-WASP [78]. wnt5b is expressed at the margin of the EVL [79], and fak1a is ubiquitously expressed in the early embryo stage. It is possible that Wnt5b may integrate with the FAK signal to mediate the actin cytoskeleton and promote cell migration in a co-localized region. We discovered lower expression levels and activities of Rac1 and Cdc42 in both wnt5b and fak1a morphants, suggesting that they may thus relay signals from both FAK and Wnt5b to modulate actin dynamics during gastrulation. This was evidenced by the synergistic rescue of both fak1a and wnt5b morphants by Rac1 and Cdc42. Herein, we demonstrate for the first time the missing functional link between Wnt and FAK signalling to mediate gastrulation cell movements (epiboly and convergence) via the precise regulation of Rac1 and Cdc42 activities and subsequent actin dynamics.
The convergence of Wnt and FAK signalling suggests that they may collaborate to regulate the same cellular processes  Figure 11. Wnt5b integrates with Fak1a to mediate cell movements during gastrulation. Gastrulation cell migration is controlled by the Wnt5b pathway to activate calcium and Dsh/Daam1. Herein, we identified that Wnt5b can also integrate with FAK signalling, and both Wnt5b and Fak1a then activate small GTPase Rac1 and Cdc42 to mediate actin dynamics during gastrulation. Convergent is mainly mediated by Fak1a-mediated cell migration and extension is exerted by cell intercalation, which requires both Fak1a and Wnt5b.
royalsocietypublishing.org/journal/rsob Open Biol. 10: 190273 such as convergence and extension during gastrulation. Convergence and extension were once thought to be bundled processes driven by cellular intercalation [4]. Analysis of different gastrulation mutants later revealed that convergence and extension can be linked or separately regulated. Convergence and extension are both affected in knypek/glypican6 [80] and trilobite/strabismus [81,82] mutants. By contrast, convergence and extension is inhibited in bmp [83] and has2 [84] mutants, respectively. Convergence relies on the active cell migration of individual or groups of cells without cell rearrangement, while extension depends on cell intercalation [84]. Our data support this notion by showing that FAK affected both convergence and extension, while Wnt5b influenced only convergence. This further strengthens the idea that Wnt5b only affects cell migration, while Fak1a affects both migration and intercalation, despite their regulation of common GTPase targets.
We propose a working model to show the interaction of Wnt5b and Fak1a during gastrulation ( figure 11). Conventionally, Wnt5b works via Dishevelled/Daam1 or calcium (not shown) to activate Rac1/Cdc42 and subsequent actindependent cell migration during gastrulation. Here, we further demonstrate that Wnt5b can work in parallel with Fak1a or activate Fak1a signalling to modulate Rac/Cadc42 and downstream signalling. During gastrulation, convergence is mainly regulated by Fak1a-mediated cell migration. On the other hand, extension, which requires cell intercalation, is mediated by Wnt5b and Fak1a signalling. In the Fak1a mutants, the inhibition of gastrulation should occur like that of fak1a morphants because of the reduction in Rac1 and Cdc42 activities.
However, only a mild gastrulation defect was observed in mutants due to compensatory increases in wnt5b expression and presumably elevated Rac1 and Cdc42 activities via a Fak1a-independent conventional pathway. Overall, we provide strong evidence to support that Wnt integrates with FAK to fine-tune Rac1 and Cdc42 activities for the differential regulation of convergence extension during gastrulation.
Ethics. All animal handling procedures were approved by the use of laboratory animal committee at National Taiwan University, Taipei, Taiwan (IACUC Approval ID: 97 Animal Use document No. 55).