Eukaryotic initiation factor 4A3 inhibits Wnt/β-catenin signaling and regulates axis formation in zebrafish embryos

Wnt/β-catenin signaling plays pivotal roles in regulating embryogenesis in multicellular organisms, adult tissue homeostasis, regeneration, and stem cell pluripotency (Clevers and Nusse, 2012; MacDonald et al., 2009; Nusse and Clevers, 2017). Dysregulation of Wnt/β-catenin activity is associated with many human diseases, including congenital malformations, tumorigenesis, and various skeletal diseases (Clevers and Nusse, 2012; MacDonald et al., 2009; Nusse and Clevers, 2017; Regard et al., 2012). Wnt ligands bind to LRP5/6 and members of the frizzled family of transmembrane receptors, triggering a signaling cascade that inhibits GSK3 and leads to the stabilization of β-catenin. Stabilized β-catenin is translocated into the nucleus, where it forms a transcription activation complex with Tcf/Lefs to activate the transcription of Wnt target genes (MacDonald and He, 2012; MacDonald et al., 2009; Niehrs, 2012). Thus, in the Wnt/β-catenin pathway, the binding between β-catenin and Tcf/Lefs is a key step to form the transcription activation complex and stimulate transcription of Wnt target genes.

The developmental role of Wnt/β-catenin signaling in the formation of the dorsoventral and anteroposterior axes in multiple animal species has been well documented (Hikasa and Sokol, 2013; Langdon and Mullins, 2011; Petersen and Reddien, 2009). For example, in zebrafish embryos, maternally provided β-catenin has an important role in the establishment of the dorsal organizer before gastrulation, while zygotic Wnt/β-catenin signaling promotes the ventrolateral mesodermal development to limit the organizer after the onset of gastrulation (Bellipanni et al., 2006; Lekven et al., 2001; Yan et al., 2018). In addition, zygotic Wnt/β-catenin signaling directs anteroposterior neural development (Baker et al., 2010; Erter et al., 2001; Lekven et al., 2001; Ramel et al., 2005).

Eukaryotic initiation factor 4A3 (EIF4A3; also known as DDX48) is a core component of the exon junction complexes (EJCs). The mammalian EJC contains four core subunits: EIF4A3, MAGOH, Y14 (also known as RBM8A) and MLN51 (also known as CASC3 or BTZ) (Le Hir et al., 2016; Linder and Jankowsky, 2011). All four proteins can shuttle between the nucleus and cytoplasm. EIF4A3, MAGOH and Y14 are located in the nucleus, while MLN51 is mainly cytoplasmic (Le Hir et al., 2016). Within this complex, EIF4A3 acts as an ATP-dependent DEAD-box RNA helicase and serves as a platform for the EJC assembly. EIF4A3 is highly conserved in invertebrates and vertebrates, and possesses all conserved motifs defining the DEAD-box family of RNA helicases (Le Hir et al., 2016; Linder and Jankowsky, 2011; Shibuya et al., 2006). The EJC plays important roles in the post-transcriptional fate of mRNA, including pre-mRNA splicing (Michelle et al., 2012), export (Le Hir et al., 2001), stability (Gehring et al., 2005; Palacios et al., 2004), translation (Chazal et al., 2013; Nott et al., 2004), localization and nonsense-mediated decay (Hachet and Ephrussi, 2004; Palacios et al., 2004). The EJC is involved in a variety of biological processes in vivo. For example, in Drosophila, it is essential for the localization of oskar mRNA to the posterior pole of the oocyte and is required for the differentiation of the photoreceptor by maintaining MAP Kinase splicing (Ashton-Beaucage et al., 2010; Hachet and Ephrussi, 2004; Mohr et al., 2001; Newmark and Boswell, 1994; Palacios et al., 2004; Roignant and Treisman, 2010; van Eeden et al., 2001). Additionally, the Drosophila EJC regulates the splicing of the cell polarity gene dlg1 to control Wnt signaling (Liu et al., 2016). The depletion of Eif4a3 in Xenopus leads to full-body paralysis and defects in sensory neurons, pigment cells and cardiac development. Knockdown of magoh or Y14 resulted in a phenotype virtually identical to the Eif4a3 morphant phenotype (Haremaki et al., 2010; Haremaki and Weinstein, 2012). Mice with haploinsufficiency for EJC subunits exhibit altered embryonic neurogenesis and microcephaly, while the genetic suppression of p53 (tp53) significantly rescues microcephaly in these mice (Mao et al., 2016, 2015; McMahon et al., 2014; Silver et al., 2010). Together, these observations from different organisms suggest that the in vivo effect of EIF4A3 is due to its function within the EJC. Recently, several studies have indicated that EIF4A3 is associated with Richieri-Costa-Pereira syndrome (RCPS) which is autosomal-recessive acrofacial dysostosis (Favaro et al., 2014, 2011; Hsia et al., 2018). In this disease, genetic alterations cause a partial loss of function of EIF4A3; however, the underlying molecular mechanism is unknown (Favaro et al., 2014; Miller et al., 2017). Other than the cellular function of EIF4A3 in RNA processing and the developmental roles of EIF4A3 as a core subunit of the EJC complex, the physiological effects of EIF4A3 and its underlying mechanisms have, thus far, remained unclear.

In this study, we found that EIF4A3 functions as a novel binding partner of the β-catenin and Tcf as well as inhibits Wnt/β-catenin signaling. Mechanistic studies demonstrated that EIF4A3 interferes with the β-catenin and Tcf/Lefs transcription activation complex and suppresses the Wnt signaling activity. Wnt stimulation leads to accumulated β-catenin displacing EIF4A3 from a transcriptional complex with Tcf/Lef, allowing the active transcription complex to facilitate the expression of target genes. Additionally, we found that Eif4a3 attenuates Wnt activity to promote dorsal development and anterior neuroectoderm pattern formation in zebrafish embryos. Thus, EIF4A3/Eif4a3 functions not only as a Wnt/β-catenin signaling repressor but also as a regulator of axis formation during zebrafish embryogenesis.

To identify Wnt/β-catenin regulators in zebrafish, we performed a yeast two-hybrid assay with the β-catenin armadillo (ARM) region as bait to search a zebrafish one-cell stage cDNA library. We identified Eif4a3 as a candidate β-catenin-binding partner. Zebrafish Eif4a3 shows a high degree of sequence identity with its human homolog EIF4A3 (Fig. S1A) (Shibuya et al., 2006). Using zebrafish embryos and cultured human cells, we first tested whether EIF4A3/Eif4a3 can regulate endogenous Wnt/β-catenin transcriptional activity both in vivo and in vitro. Forced expression of zebrafish eif4a3 mRNA decreased the basal levels of Wnt reporter (TOPFlash) activity, rather than Bmp reporter (BRE) activity or Nodal/Tgfβ reporter (ARE) activity in zebrafish embryos in a dose-dependent manner (Fig. 1A). Likewise, human EIF4A3 exhibited a similar effect on Wnt reporter activity in HEK293T cells (Fig. 1B). Therefore, we speculated that EIF4A3/Eif4a3 affects the Wnt/β-catenin pathway. Subsequently, we investigated the genetic interaction between EIF4A3/Eif4a3 and Wnt/β-catenin signals in zebrafish embryos and HEK293T cells. Wnt/β-catenin signal transduction is a multi-step process involving several molecular components, including Wnt ligands, β-catenin and Tcf/Lef. Consistent with previously reported results, injection of wnt3a, constitutively active β-catenin (β-catΔN, β-catenin that lacks the first 45 residues in the N-terminal region and translocates into the nucleus to interact with the transcription factor to activate target gene expression) or constitutively active Tcf7l1 (vp16-tcf7l1ΔN, a β-catenin-independent VP16-Tcf7l1 fusion protein that lacks the β-catenin-binding site) mRNAs in zebrafish embryos enhanced Wnt reporter activity at 6 hpf (hours post-fertilization) (Fig. 1C-E) (Feng et al., 2012; Rong et al., 2014, 2017). Co-injection of eif4a3 and wnt3a, β-catΔN or vp16-tcf7l1ΔN mRNA diminished Wnt reporter activity induced by Wnt3a and β-CatΔN, but not by VP16-Tcf7l1ΔN in zebrafish embryos (Fig. 1C-E). Likewise, overexpression of EIF4A3 inhibited Wnt3a- and β-CatΔN- but not VP16-Tcf7l1ΔN-induced Wnt activity in HEK293T cells (Fig. 1F-H). Similarly, injection of wnt3a, β-catΔN or vp16-tcf7l1ΔN mRNA into zebrafish embryos resulted in a dorsalized phenotype at 12.5 hpf (Fig. 1I,J). Co-injection of eif4a3 mRNA with wnt3a, β-catΔN or vp16-tcf7l1ΔN mRNA blocked Wnt3a- and β-CatΔN- but not VP16-Tcf7l1ΔN-induced dorsalizing activities in zebrafish embryos (Fig. 1I,J). Collectively, these data suggest that EIF4A3/Eif4a3 inhibits Wnt/β-catenin signaling and may act at the β-catenin level.

To determine the molecular mechanism underlying Wnt inhibition by EIF4A3/Eif4a3, we first assessed whether EIF4A3 and β-catenin interact with each other by co-immunoprecipitation (co-IP) assays. Both exogenous and endogenous EIF4A3 and β-catenin were observed in a same complex in HEK293T cells (Fig. 2A, upper and middle panels). Likewise, the two endogenous proteins bound to each other under physiological conditions in developing zebrafish embryos (Fig. 2A, lower panel). A series of human EIF4A3 deletion mutants were generated based on the known conserved motifs (Fig. 2B, Fig. S1A) (Le Hir et al., 2016; Shibuya et al., 2006). A region comprising amino acids 57-254 of EIF4A3, which corresponds to the EIF4A3 DEAD domain 1, was required for its interaction with β-catenin (Fig. 2C). We next mapped the binding regions required for the interaction between EIF4A3 and β-catenin using a series of previously generated β-catenin truncation mutants (Lu et al., 2015). We observed that EIF4A3 bound to amino acids 1-318 on the truncated form of β-catenin rather than to amino acids 1-274 or any other shorter N-terminal truncations. Additionally, EIF4A3 bound to both amino acid 319-484 and amino acid 485-781 mutants (Fig. 2D,E). Hence, the binding region for EIF4A3 is within amino acids 275-781, which includes 4-12 armadillo (Arm) repeats as well as the transactivation activity domain of β-catenin. Thus, we concluded that β-catenin and EIF4A3/Eif4a3 bind to each other both in cultured human cells and zebrafish embryos.

Given that the binding site for EIF4A3 on β-catenin largely overlaps with that for TCF/LEF (Polakis, 2012; Valenta et al., 2012), we speculated that EIF4A3 may compete with TCF/LEF for the binding site on β-catenin to prevent the interaction between β-catenin and TCF/LEF, and interfere with the β-catenin/TCF complex. We first confirmed that EIF4A3 indeed localized in the nucleus of cultured HeLa cells (Fig. 3A). We next employed a pull-down assay to test whether EIF4A3 can directly interact with β-catenin or TCF. As shown in Fig. 3B, purified GST-tagged EIF4A3 binds to β-catenin and to TCF7L2, suggesting that EIF4A3 directly interacts with β-catenin and TCF7L2. To further confirm the interaction between TCF7L2 and EIF4A3, we used a series of domain deletion mutants of TCF7L2 to map the region of TCF7L2 that is responsible for EIF4A3 interaction by co-immunoprecipitation assays (Fig. 3C). When the HMG box was deleted, the resulting mutated TCF7L2 did not bind to EIF4A3 (Fig. 3D). Thus, the results suggested that the HMG box of TCF7L2 is likely important for its association with EIF4A3. Moreover, interaction of EIF4A3 with β-catenin and with TCF7L2 at endogenous levels was further confirmed by co-immunoprecipitation assays in HEK293T cells (Fig. 3I-K, right panel, lane 5). Therefore, we conclude from these experiments that EIF4A3 associates with β-catenin and TCF7L2. In addition, endogenous EIF4A3 bound to the promoters of the Wnt target genes, such as AXIN2, CCND1, DKK1, LEF1 and SP5L, but not to GAPDH or α-satellite in HEK293T cells, as indicated by chromatin immunoprecipitation (ChIP)-PCR analysis (Fig. 3E).

As EIF4A3 can interact with β-catenin and TCF7L2, as well as associate with the promoters of Wnt target genes, we hypothesized that EIF4A3 functions as an inhibitor protein that interferes with the formation of β-catenin/TCF complex. To test this hypothesis, we first examined the effect of EIF4A3 on transcriptional activity of the β-catenin/TCF complex in HCT116 cells, which harbor a mutant β-catenin allele and exhibit overactivated Wnt activity. When β-CatΔN and TCF7L2 were co-transfected into HCT116 cells along with different amounts of EIF4A3, β-CatΔN/TCF7L2-induced Wnt reporter activity was inhibited by EIF4A3 in a dose-dependent manner (Fig. 3F). We next performed a co-immunoprecipitation assay using HCT116 cells with co-transfected Flag-tagged β-catenin and Myc-tagged Tcf7l2 along with various doses of HA-tagged EIF4A3. The amounts of Tcf7l2 that co-immunoprecipitated with β-catenin decreased as the amounts of HA-tagged EIF4A3 that associated with β-catenin increased, along with increasing doses of EIF4A3 (Fig. 3G, left panel). Similarly, the amounts of endogenous TCF7L2 that co-precipitated with endogenous β-catenin also decreased as the amounts of HA-tagged EIF4A3 that associated with endogenous β-catenin increased, along with increasing doses of EIF4A3 (Fig. 3G, right panel). As EIF4A3 associates with the promoters of the Wnt target genes in HEK293T cells, which were regarded as a Wnt-off cell line, these results suggested that EIF4A3 likely competes with β-catenin for binding to TCF7L2 and interferes in the formation of β-catenin/TCF7L2 transcriptional activator complex.

To test this possibility, HEK293T cells were treated with BIO, a small molecular compound that activates Wnt signaling by inhibiting GSK3β activity (Sato et al., 2004). Activated Wnt increased the ChIP signals of β-catenin on the promoters of Wnt target genes, including AXIN2 and DKK1, to facilitate their expression in HEK293T cells (Choi et al., 2013; Wu et al., 2013). However, activated Wnt led to reduction of the ChIP signals of EIF4A3 on the promoters of AXIN2 and DKK1 genes (Fig. 3H), suggesting that EIF4A3 is released from the promoters of these genes upon Wnt stimulation. In addition, treating cells with BIO activates the Wnt signaling pathway, as indicated by the increased levels of non-phospho (active) β-catenin (non-p-β-catenin); it does not, however, alter the expression level of EIF4A3 (Fig. 3I-K, left panels, compare lane 1 with lane 2). This result implied that activated Wnt has little effect on the stability of EIF4A3. Given that accumulated β-catenin binds to Lef/Tcf to form the transcription activation complex and activated Wnt signaling stimulates the release of EIF4A3 from the promoters of Wnt target genes, we speculated that accumulated β-catenin may displace EIF4A3 from the TCF7L2 complex at the promoters of Wnt target genes upon Wnt stimulation. To test this, we evaluated the binding among EIF4A3, β-catenin and TCF7L2 under different Wnt activity by reciprocal co-immunoprecipitation assays. In agreement with previous results, Wnt activation increased the association of β-catenin with TCF7L2 (Fig. 3I,J, right panels, compare lane 5 with lane 8). Conversely, the associations of EIF4A3 with β-catenin and with TCF7L2 reduced simultaneously upon BIO treatment (Fig. 3I-K, right panels, compare lane 5 with lane 8). These findings suggest that accumulated β-catenin likely competes with EIF4A3 for binding to TCF7L2 and displaces EIF4A3 from a transcriptional complex with TCF7L2 under increased Wnt stimulation. Taken together, we propose a role for EIF4A3 in Wnt/β-catenin signaling: EIF4A3 binds to β-catenin or to TCF7L2 and also associates with the promoters of Wnt target genes; β-catenin and EIF4A3 are mutually exclusive in their binding to TCF7L2; and EIF4A3 would be displaced from a transcriptional complex with TCF7L2 by accumulated β-catenin upon Wnt stimulation, which allows the transcription activation complex to stimulate transcription of Wnt target genes (Fig. 3L).

We next determined the physiological relevance of eif4a3 in zebrafish embryos. Forced expression of eif4a3 in zebrafish embryos led to typically ventralized phenotypes at 24 hpf with a reduced head structure and enlarged ventral structure (Fig. 4A). The ventralizing effect of Eif4a3 was dose dependent (Fig. 4B, upper panel). Human EIF4A3 has a comparable ventralizing effect to that of zebrafish eif4a3 (Fig. 4B, lower panel). The ventralizing effects of EIF4A3/Eif4a3 were investigated by performing whole-mount in situ hybridization (WISH) to examine the expression patterns of an anterior neuroectoderm marker, otx2, at the bud stage and a hematopoietic marker, gata1, in the blood island at 24 hpf. Overexpression of EIF4A3/Eif4a3 reduced the expression domain of otx2 and enlarged the expression domain of gata1 (Fig. 4C-F). Maternal β-catenin promotes the development of the dorsal organizer before gastrulation (Bellipanni et al., 2006; Yan et al., 2018). We, therefore, postulated that the ventralized phenotypes caused by the overexpression of EIF4A3/Eif4a3 might result from the inhibition of maternal β-catenin. Hence, we performed WISH to examine the expression of the organizer-specific markers chordin (chd) and goosecoid (gsc) at 4.3 hpf to assess the development of the dorsal organizer because both markers are main targets of maternal Wnt at this stage. As shown in Fig. 4G,H, embryos injected with eif4a3 mRNA showed reduced expression areas for chd and gsc, indicating that Eif4a3 has a ventralizing effect before gastrulation and overexpression of Eif4a3 inhibits maternal Wnt signaling.

We next defined the functional domain(s) of EIF4A3/Eif4a3 involved in ventralizing zebrafish embryos. To achieve this, the mRNAs for various human EIF4A3 mutants (Fig. 2B) were injected into zebrafish embryos to examine the effect on the expression patterns of chd and gsc at 4.3 hpf and otx2 at the bud stage. As shown in Fig. S2A-F, the EIF4A3ΔD1 mutant completely abolished the ventralizing action, suggesting that DEAD domain 1 of EIF4A3 is required for maternal Wnt inhibiting action. Consistently, zebrafish Eif4a3ΔD1 mutant also lost the inhibitory effect on Wnt3a-induced Wnt reporter activity in zebrafish embryos (Fig. S2G). Thus, we concluded that DEAD domain 1 of EIF4A3/Eif4a3 was required to inhibit Wnt/β-catenin signaling.

The role of EIF4A3/Eif4a3 in RNA processing is ATPase dependent (Le Hir et al., 2016; Linder and Jankowsky, 2011). To determine whether the function of EIF4A3/Eif4a3 within the RNA processing is required to inhibit Wnt/β-catenin signaling, we generated two mutants of human EIF4A3: K88N and E188Q. Both mutants abolished ATP binding capability according to previously published biochemical studies (Gehring et al., 2009; Shibuya et al., 2006). The two mutants exhibited inhibitory effects on the formation of the dorsal organizer at 4.3 hpf (Fig. 4I,J). Overexpression of β-CatΔN strongly induced maternal Wnt activity in zebrafish, leading to expanded expression areas of chd and gsc at 4.3 hpf (Fig. 4K,L). Like wild-type EIF4A3, both point mutants inhibited the ectopic Wnt activity induced by β-catΔN mRNA injection (Fig. 4K,L). Likewise, in HCT116 cells, EIF4A3 and each point mutant significantly inhibited Wnt reporter activity induced by β-CatΔN (Fig. 4M). The data suggest that overexpression of EIF4A3/Eif4a3 inhibits Wnt/β-catenin signaling and impairs the formation of the dorsal organizer in an RNA processing-independent manner.

The spatiotemporal expression pattern of eif4a3 in zebrafish embryos suggested that eif4a3 was maternally deposited and ubiquitously expressed before 24 hpf, and then enriched in the head region at 36 hpf and 48 hpf (Fig. S1B,C). The eif4a3-null mutant was generated by using the TALEN-based knockout system to explore the physiological role of Eif4a3 (Fig. S3A). We obtained a line with an 8-bp deletion in the third exon of eif4a3, which resulted in the early termination of translation (Fig. S3B-D). Western blot analysis showed that Eif4a3 protein levels were gradually reduced in eif4a3 mutants both at 15 hpf and at 24 hpf (Fig. S3E). We noted that Eif4a3 protein levels were markedly reduced but not completely depleted, suggesting abundant maternally deposited Eif4a3 protein. The zygotic mutant fish exhibited a ventralized phenotype, characterized by a small head, shortened and bent body axes, and a reduced dorsal tail fin, along with apoptotic cells in the whole body at 24 hpf (Fig. S3F-H). These phenotypes could be rescued by the injection of 50 pg of eif4a3 mRNA (Fig. S3F,G), suggesting that the eif4a3 knockout is specific.

Increased apoptotic cells and microcephaly were previously observed in mice haploinsufficient for Eif4a3, and the genetic ablation of p53 was sufficient to rescue this phenotype (Mao et al., 2016). Apoptotic cells in eif4a3-null mutant embryos may impinge the observation on the effects of Wnt activation. The use of p53-null zebrafish could potentially decrease the collateral tissue damage, thus allowing us to observe the important phenotypes masked by p53-induced cell death (Robu et al., 2007). We noted that the ventralized phenotype, produced in response to depletion of Eif4a3, is retained in the p53 mutant background, apart from the lack of apoptotic cells at 24 hpf (Fig. 5A). Additionally, we observed a larger expression domain of gata1 in the blood island and a smaller expression domain of retina marker rx1 in the double mutant embryos at 24 hpf (Fig. 5B,C). These results implied that the zebrafish eif4a3 knockout caused a significant increase in p53-induced apoptosis, which is similar to what occurs in mice; nevertheless, the ventralized phenotype is almost negligibly affected by p53-mediated cell apoptosis.

Two non-overlapping translation-blocking morpholinos (MOs) targeting eif4a3 mRNA were used as an independent and complementary approach to perform eif4a3 knockdown analysis (Fig. S4A). The MOs effectively blocked the translation of an eif4a3 5′-UTR-GFP reporter (Fig. S4B). Injection of MOs reduced Eif4a3 protein levels in zebrafish embryos at 6 hpf rather than at 4.3 hpf (Fig. S4C). These results suggest that the MOs blocked the translation of eif4a3 effectively but did not decrease the maternal levels of the Eif4a3 protein. The injection of eif4a3 MO1 and MO2 resulted in a consistent phenotype at 24 hpf that was similar to those of eif4a3 mutants (Fig. S4D). This phenotype could also be neutralized by co-injection of 50 pg eif4a3 mRNA (Fig. S4E). Additionally, ventralized phenotypes were also observed in eif4a3 morphants with the p53^−/− genetic background (Fig. S4F). Hereafter, we evaluated effects of depletion of eif4a3 in zebrafish embryos with the p53^−/− background.

The above ventralized phenotypes prompted us to investigate whether or not the dorsal organizer formation was impaired after the onset of gastrulation. We, therefore, examined the expression of the dorsoventral marker genes at the shield stage. The eif4a3-depleted embryos showed reduced expressions of chd and gsc, but laterally and dorsally expanded expression of the ventral marker gene even-skipped-like1 (eve1) (Fig. 5D,E, Fig. S5A,B). We next investigated anteroposterior neural pattern formation after depletion of eif4a3. The expression of six3b (the indicative marker of forebrain), pax2a (the indicative marker of midbrain-hindbrain boundary) and krox20 (the indicative marker of hindbrain), were used simultaneously to assess anteroposterior neural patterning. The expression levels of six3b and pax2a were reduced in eif4a3 mutants and morphants at 12.5 hpf; however, no apparent alterations in krox20 levels were observed, suggesting that Eif4a3 promotes anterior neural development (Fig. 5F and Fig. S5C). Taken together, these results indicated that the depletion of zygotic Eif4a3 inhibited dorsal and anterior neuroectoderm development. We then investigated the effect of the loss of Eif4a3 on Wnt/β-catenin signals. Knockdown of eif4a3 significantly increased Wnt reporter activity (Fig. S5D). Additionally, we observed that depletion of Eif4a3 expanded the expression areas of cdx4 and sp5l, two direct target genes of zygotic Wnt, at the 80% epiboly stage, as accessed by WISH (Fig. 5G and Fig. S5E). Similarly, the expression levels of the direct zygotic Wnt targets, cdx4, sp5l, vent, axin2 and ccnd1, increased in eif4a3 mutants at the 80% epiboly stage, as indicated by qRT-PCR (Fig. 5H). In addition, EIF4A3 was depleted in HCT116 cells with a p53^−/− genetic background via a shRNA-mediated knockdown approach (Fig. S5F). Again, the Wnt reporter activity and the transcriptional levels of Wnt target genes were measured. Similarly, knockdown of EIF4A3 in HCT116 cells also increased Wnt reporter activity and transcriptional levels of the direct Wnt target genes, AXIN2 and DKK1 (Fig. S5G,H). Therefore, we concluded that depletion of EIF4A3/Eif4a3 increases Wnt/β-catenin signaling in both zebrafish embryos and cultured cells.

As Bmp, Nodal and Fgf signaling pathways are also involved in axial patterning in zebrafish (Langdon and Mullins, 2011), it is not clear whether these pathways are affected during gastrulation when Eif4a3 is depleted. Indeed, Bmp, Nodal, and Fgf signaling pathways appear unaffected, as indicated by the expression of a series of marker genes (Fig. S6A-C). Collectively, these data strongly suggest that zebrafish eif4a3 promotes dorsal development and anterior neuroectoderm pattern formation and depletion of EIF4A3/Eif4a3 increases Wnt/β-catenin signaling.

We wondered whether Eif4a3 promotes dorsal development and anterior neuroectoderm pattern formation by inhibiting Wnt/β-catenin signaling. To prove this idea, a Wnt/β-catenin signaling inhibitor, human TCF7L2ΔN, which lacks the β-catenin-binding domain and acts as a constitutive repressor of Wnt/β-catenin signaling, was introduced into eif4a3-depleted embryos. First, we evaluated the inhibitory effect of TCF7L2ΔN on the increased Wnt activity by depletion of eif4a3. Injection of TCF7L2ΔN mRNA rendered the expansion of cdx4 and sp5l expression areas in eif4a3 mutants and morphants at the 80% epiboly stage, suggesting that TCF7L2ΔN antagonizes eif4a3 depletion-induced Wnt activity (Fig. 6A and Fig. S7A). We then assessed the antagonistic effect of TCF7L2ΔN on dorsoventral and anterior neuroectoderm development in eif4a3-depleted embryos. When comparing with gfp mRNA-injected eif4a3 mutants and morphants, injection of TCF7L2ΔN mRNA into eif4a3 mutants and morphants alleviated the eif4a3 depletion-induced reduction in chd and gsc, as well as the expansion of eve1 at the shield stage (Fig. 6B,C and Fig. S7B,C). Notably, when TCF7L2ΔN mRNA was injected into eif4a3 mutants and morphants, a small proportion of embryos exhibited further reduction of expression of chd and gsc, as well as more expanded expression of eve1 (Fig. 6C and Fig. S7C). The tendencies observed here are likely due to the inhibition of maternal Wnt signaling by TCF7L2ΔN.

Previous studies have indicated that hyperactive Wnt signaling leads to the reduction of the anterior neuroectoderm markers, six3b and opl (Kim et al., 2002; Liu et al., 2013). The well-established zebrafish apc mutant exhibited reduced expression domain of the opl by constitutively increasing Wnt/β-catenin activity (Liu et al., 2013). We then examined whether Eif4a3 could rescue this effect by acting as a Wnt inhibitor. Consistent with a previous report, not only was the ratio of strongly reduced to reduced to normal of opl close to 1:2:1, but the opl expression pattern and the embryonic genotypes were also correlated very well in the progenies of apc^+/−×apc^+/− at bud stage (Fig. 6D,E). Injection of TCF7L2ΔN mRNA expanded the expression areas of opl in embryos with all three genetic backgrounds, in comparison with the gfp mRNA-injected group (Fig. 6D,E). Likewise, injection of eif4a3 mRNA resulted in a similar effect to that observed with TCF7L2ΔN mRNA injection (Fig. 6D,E), albeit the effect of eif4a3 was not as great as that of TCF7L2ΔN. Additionally, we also investigated the reciprocal action between Wnt signaling and Eif4a3 on the development of the anterior neuroectoderm. Introduction of TCF7L2ΔN mRNA into eif4a3-depleted embryos rescued the expression of six3b and pax2a at 12.5 hpf (Fig. 6F and Fig. S7D). These results suggest that Eif4a3 promotes dorsal development and anterior neuroectoderm pattern formation by inhibiting Wnt/β-catenin signaling.

As a member of the DEAD-box protein family, EIF4A3 is a core component of the EJC. Within the EJC, EIF4A3 exhibits RNA-dependent ATPase activity and ATP-dependent RNA helicase activity. In the present study, we have identified a mechanism by which EIF4A3/Eif4a3 suppresses the β-catenin/Tcf transcription complex activity, thereby limiting the output of Wnt signaling. We found that EIF4A3 binds to either β-catenin or to TCF7L2 and associates with the promoters of Wnt target genes. EIF4A3 likely competes with β-catenin for binding to TCF7L2. EIF4A3 dissociated from a transcriptional complex and from promoters of Wnt target genes, which, along with the increasing Wnt signaling, suggests that EIF4A3 acts as a Wnt responsible inhibitor to control β-catenin/Tcf transcriptional activity. In zebrafish embryos, Eif4a3 acts as a Wnt inhibitor to promote dorsal development and anterior neuroectoderm pattern formation.

A key finding of this study is that EIF4A3/Eif4a3 specifically inhibits Wnt/β-catenin signaling. We observed that EIF4A3/Eif4a3 inhibited Wnt activity induced by Wnt3a and β-CatΔN but not by VP16-Tcf7l1ΔN both in vitro and in vivo. Importantly, injection of TCF7L2ΔN rendered the broad expansion of Wnt target genes expression in eif4a3-depleted zebrafish embryos, suggesting that EIF4A3/Eif4a3 acts at the level of β-catenin. Moreover, Wnt reporter activity induced by β-CatΔN/TCF7L2 decreased, along with the increasing amounts of EIF4A3. A series of co-immunoprecipitation assays indicated that EIF4A3 associates with β-catenin or with TCF7L2, and a ChIP assay suggested that EIF4A3 binds to the promoters of Wnt target genes under physiological conditions. When EIF4A3 was overexpressed, the association of β-catenin with TCF7L2 was attenuated as the association of β-catenin with EIF4A3 increased in a dose-dependent manner. However, upon Wnt stimulation, the association of EIF4A3 with β-catenin and with TCF7L2 decreased as the association of β-catenin with TCF7L2 increased, and EIF4A3 was released from the promoter of Wnt target genes. Our results suggest that EIF4A3 and β-catenin have a mutually exclusive interaction with TCF7L2, and that accumulated β-catenin drives TCF7L2 into a complex with β-catenin, displacing EIF4A3 by competition in response to increasing Wnt levels, which leads to release of EIF4A3 from the promoter of Wnt target genes. We do not know where EIF4A3 would go after dissociation from the transcription complex under Wnt stimulation. It has been reported that EIF4A3 can shuttle between the nucleus and cytoplasm (Le Hir et al., 2016). Future studies will be needed to determine the distribution of EIF4A3 responsible for Wnt stimulation. EIF4A3 acts as an inhibitor of the Wnt/β-catenin signaling pathway and creates a relatively more complicated regulatory mechanism than appears necessary in this pathway. One advantageous possibility is that EIF4A3 likely exists in a pool and disassociates from a transcriptional complex under the stimulation of Wnt signaling to help to shape expression levels of Wnt gradient-induced target genes.

The inhibitory effects of EIF4A3 on the Wnt pathway in cell lines were recapitulated in zebrafish embryos. Eif4a3 regulates dorsoventral pattern formation and anterior neuroectoderm development by inhibiting the Wnt/β-catenin pathway. A previous study indeed showed that misexpression of Eif4a3 inhibited dorsal development of Xenopus ectoderm (Weinstein et al., 1997). In zebrafish, overexpression of EIF4A3/Eif4a3 caused ventralized phenotypes at 24 hpf. This ventralizing effect was likely due to the inhibition of maternal Wnt signaling, which was supported by the reduced expression of dorsal organizer markers in late blastula stage. Although abundant maternally deposited Eif4a3 mRNA and protein may obscure and alleviate the forced expression effect, we conclude that the ventralizing effect caused by overexpression of Eif4a3 occurs because maternal Wnt activity is inhibited. Depletion of Eif4a3 by either knockout or knockdown caused a weakly ventralized phenotype and reduced the anterior neuroectoderm structure in zebrafish embryos with a p53^−/− background. It is conceivable that the contribution of the maternal mRNA and protein may compromise the loss-of-function effect in both zygotic knockout embryos and translation-blocking MO knockdown embryos. In support of this view, the deposited maternal protein was detected even in zygotic mutant embryos at 24 hpf, although we observed a dramatic reduction in the level of Eif4a3 protein from 15 hpf to 24 hpf in eif4a3 mutant embryos. At first, we expected the translation-blocking MOs to produce dorsalized embryos but we did not observe this. Indeed, the protein levels were reduced at the beginning of gastrulation but were not apparently altered at the late blastula stage, when the knockdown was performed. Therefore, the morphant embryos exhibit a ventralized phenotype similar to that of zygotic mutant embryos due to depletion of the Eif4a3 protein in zygotic Wnt functional stages.

The Wnt/β-catenin signaling pathway plays a key role in anteroposterior neuroectoderm formation in a concentration-dependent manner in multiple animal species (Dorsky et al., 2003; Kiecker and Niehrs, 2001; Rhinn et al., 2005). In zebrafish, overactivation of Wnt signaling leads to neuroectoderm posteriorization (Green et al., 2015; Petersen and Reddien, 2009). We observed that expression levels decreased but the position of the anterior neuroectoderm markers six3b and pax2a in eif4a3 mutants and morphants did not shift and the expression pattern of hindbrain marker krox20 was unchanged. As discussed earlier, our results demonstrate that Eif4a3 inhibits Wnt signaling while the association between Eif4a3 and transcriptional complex becomes attenuated upon Wnt stimulation. Such a molecular mechanism implies that Eif4a3 may have an inhibitory role that differs in degree along with the gradient Wnt activity from the anterior to the posterior. This molecular mechanism of Eif4a3 in Wnt signaling suggests that depletion of Eif4a3 resulted in the activity of Wnt increased at different degrees and exhibited distinct action based on the local level of Wnt signaling during anteroposterior neuroectoderm patterning. The zygotic Wnt pathway has relative less activity in the anterior compared with that in the posterior. This may explain the anterior neuroectoderm defect in Eif4a3-depleted embryos we observed.

Two major phenotypes were observed in zygotic mutant embryos and morphant embryos. First, the depletion of Eif4a3 in zebrafish embryos leads to the apoptotic phenotype, while the loss of p53 significantly rescues this phenotype. Similar effects have been observed previously in mice, thus supporting this view. In mice, the conditional haploinsufficiency of Eif4a3, Y14 (Rbm8a) or Magoh induces apoptosis and p53 ablation rescues apoptosis in the three EJC mutants (Mao et al., 2016). In addition, loss of Y14 (rbma8) or magoh in zebrafish also leads to the apoptotic phenotype as eif4a3 mutant embryos and morphant embryos (Gangras et al., 2020). Therefore, these data suggested that the apoptosis is likely due to the impairment of the EJC complex. Second, zygotic eif4a3 mutants and eif4a3 morphants with a p53^−/− background exhibited a ventralized phenotype and reduced anterior neuroectoderm development with increased Wnt activity. The ventralized phenotype and the reduction in anterior neuroectoderm development are likely due to the enhancement of the Wnt/β-catenin signaling pathway. The evidence includes the following two observations: (1) injection of TCF7L2ΔN mRNA counteracting the ventralizing effect in Eif4a3-depleted embryos at shield stage; and (2) the genetic interaction between Eif4a3 and the components of the Wnt/β-catenin pathway, Apc and Tcf7l2, occurring in the formation of the anterior neuroectoderm in zebrafish embryos.

It has been reported that the EJC complex has various biological functions, and EJC deficiency triggers p53-induced apoptosis. In Xenopus, the depletion of Eif4a3 leads to full-body paralysis by altering the splicing of the ryanodine receptor, with defects in sensory neurons, pigment cells and cardiac development. Similar phenotypes were observed in magoh or Y14 MO knockdown embryos (Haremaki et al., 2010; Haremaki and Weinstein, 2012). Likewise, the conditional haploinsufficiency of Eif4a3, Y14 and Magoh in mice results in highly similar embryonic neurogenesis defect and microcephaly as does activation of p53. Furthermore, the genetic ablation of p53 significantly rescues microcephaly in mice with the conditional haploinsufficiency of all three EJC subunits (Mao et al., 2016, 2015; McMahon et al., 2014; Silver et al., 2010). In Drosophila, Eif4a3, Magoh and Y14 maintained MAPK transcript levels, and depletion of Eif4a3, Magoh or Y14 led to photoreceptor differentiation defects and extensive apoptosis (Roignant and Treisman, 2010). Intriguingly, the depletion of Eif4a3, Magoh or Y14 in Drosophila results in phenotypes associated with reduced Wnt signaling, suggesting that EJC positively regulates the Wnt/β-catenin signaling pathway. Mechanistically, the EJC controls the splicing of dlg1, and Dlg1 interacts directly with Dsh to protect it from lysosomal degradation (Liu et al., 2016). Overall, the effects of Eif4a3 observed in various organisms can likely be explained by its function as a core subunit within the EJC, as the depletion of other EJC subunits result in similar effects. These phenotypes are different from those observed in zebrafish eif4a3 mutants. Considering the broad role of the EJCs in post-transcriptional processes, it is possible that the depletion of eif4a3 impedes other pathways directly or indirectly, and this may contribute to this observation. However, our data demonstrate that Bmp, Nodal and Fgf signaling pathways, which are all involved in axial patterning, are not affected by the depletion of Eif4a3 in zebrafish embryos during gastrulation.

In humans, the biallelic expansion of a complex repeat motif in the 5′ untranslated region of EIF4A3 reduces the level of EIF4A3 transcript and causes Richieri-Costa-Pereira syndrome (RCPS), which is mainly characterized by craniofacial and limb malformations (Favaro et al., 2014, 2011; Hsia et al., 2018). Alterations in the physiological functions of EIF4A3 might contribute to RCPS (Mao et al., 2016). Wnt/β-catenin signaling plays important roles in cranial skeletogenesis and limb formation (Delgado and Torres, 2017; Regard et al., 2012). For example, the genetic ablation of the Wnt inhibitors Axin2, Nkd1 and Nkd2, in mice alters cranial bone morphology (Liu et al., 2007; Yu et al., 2005; Zhang et al., 2007). Our study reveals that EIF4A3 functions as an inhibitor of canonical Wnt signaling. These findings improve our understanding of the molecular mechanisms underlying EIF4A3/Eif4a3 function at the cellular level and during embryonic development. Future studies of EIF4A3 as a Wnt modifier will shed light on the specificity of the molecular regulation of the Wnt pathway and may contribute to understand the mechanisms of RCPS.

Restriction enzymes were purchased from New England BioLabs. Oligo(dT)₁₈ was purchased from Sangon Biotech. DIG-UTP and anti-digoxigenin-AP were purchased from Roche. M-MLV Reverse Transcriptase and the Dual-Glo Luciferase Assay System were purchased from Promega. The mMESSAGE mMACHINE mRNA Synthesis Kit was purchased from Ambion. KOD DNA polymerase was purchased from Toyobo. Dulbecco's Modified Eagle's Medium (DMEM) was purchased from Hyclone. Fetal bovine serum (FBS) was purchased from PAN. Morpholino oligonucleotides were purchased from Gene Tools. Protein A/G Plus-agarose was purchased from Santa Cruz Biotechnology. The following antibodies were used in this study: mouse anti-EIF4A3 (1:1000 for western blotting, 1:200 for immunocytochemistry, 2 µg for co-immunoprecipitation and ChIP assays, 05-1527; Millipore), rabbit anti-EIF4A3 (1:1000 for western blotting, sc-67369; Santa Cruz Biotechnology), rabbit anti-EIF4A3 (1:1000 for western blotting, ab32485; Abcam), rabbit anti-TCF7L2 (1:1000 for western blotting and 2 µg for co-immunoprecipitation assays, 2569; Cell Signaling), mouse anti-β-catenin (1:1000 for western blotting and 2 µg for co-immunoprecipitation assays, M24002; Abmart), rabbit anti-non-phospho-β-catenin (1:1000 for western blotting, 8814; Cell Signaling), rabbit anti-GAPDH (1:1000 for western blotting, D110016; BBI Solutions), mouse anti-Myc (1:1000 for western blotting and 2 µg for co-immunoprecipitation assays, sc-40; Santa Cruz Biotechnology), murine anti-Flag (1:1000 for western blotting and 2 µg for co-immunoprecipitation assays, F1804; Sigma), rabbit anti-HA (1:1000 for western blotting and 2 µg for co-immunoprecipitation assays, 3724; Cell Signaling) and rabbit anti-Histone H3 (1:1000 for western blotting, P30266M; Abmart).

The wild-type zebrafish (Danio rerio) strain Tübingen and p53-defective mutant strain tp53^M214K were maintained on a 14 h light/10 h dark cycle at 28.5°C and fed twice daily. The eif4a3-knockout mutant strain was constructed using the TALEN system. The eif4a3 and p53 double mutant strain was obtained by natural crossing. Embryos obtained by natural crosses were kept in embryo-rearing solution in an incubator at 28.5°C. Embryos were strictly staged according to standard methods (Kimmel et al., 1995).

All experimental protocols were approved by and conducted in accordance with the Ethical Committee of Experimental Animal Care, Ocean University of China.

The sequences of human EIF4A3 and zebrafish eif4a3 were retrieved from NCBI (GenBank Accession Number: NC_000017.11 and NC_007122.7). The full-length cDNA of zebrafish eif4a3 and human EIF4A3 with/without the 3′-UTR were amplified and cloned into the pCS2+ expression vector. EIF4A3 domain-deleted mutant sequences (domain 1, domain 2, C terminus, N terminus and the middle section were deleted) were amplified and subcloned into the pCS2+ expression vector. All primers used in this study are listed in Table S1. The amino acid sequence alignment was performed using ClustalX and GeneDoc.

Total RNA was isolated from zebrafish embryos using TRIzol reagent (Invitrogen). The cDNAs were reverse-transcribed using Oligo(dT)₁₈ and M-MLV into first-strand cDNA according to the manufacturer's instructions. RT-PCR was performed using Taq DNA polymerase. Quantitative real-time RT-PCR (qRT-PCR) was performed using an iCycler iQ Multicolor Real-time PCR Detection System (Bio-Rad Laboratories). Samples from three independent experiments were collected and each sample was measured in duplicate. The mRNA levels of the genes of interest were calculated using the 2^−ΔΔCt method and normalized to β-actin or GAPDH.

Whole-mount in situ hybridization using a DIG-labeled RNA riboprobe was performed as previously described (Feng et al., 2012; Rong et al., 2014, 2017). The plasmid DNA containing partial ORF and the 3′ untranslated region (UTR) from zebrafish eif4a3 was used to generate sense and antisense riboprobes. The specificity of the riboprobes was verified by a dot-blot analysis. Images were obtained using a dissecting stereo microscope. For detecting the apoptotic cells, acridine orange (AO) (Sigma, A1121) staining was performed as previously described (Liu et al., 2014).

Capped mRNA was synthesized using the mMESSAGE mMACHINE Kit. To knock down eif4a3, two translation-blocking morpholino oligonucleotides (MOs) targeting eif4a3 were designed and purchased from Gene Tools according the targeting guideline. The sequences of the MOs are listed in Fig. S4A. A standard control MO (5′-CCTCTTACCTCAGTTACAATTTATA-3′) from Gene Tools was used as the control. A GFP reporter plasmid containing the 5′-UTR and partial ORF (−106 to −564 bp) of zebrafish eif4a3 was constructed and used to examine the efficiency of the MOs. All MOs were diluted to the desired concentration. Diluted MO and/or mRNA were injected into one-cell stage zebrafish embryos. After injection, the embryos were placed in embryo-rearing medium and maintained at 28.5°C.

HEK293T and HCT116 cells were purchased from the ATCC. Cells were cultured in DMEM medium with 10% FBS and 1% PS (penicillin and streptomycin) in a humid 37°C incubator containing 5% CO₂. Cells were seeded into 12-well plates to reach 70%-80% confluence at the time of transfection. Transfections were performed in dishes by using Polyethylenimine (Polysciences, 23966-2). Plasmids were co-transfected with 240 ng of TOPFlash DNA and 40 ng of Renilla DNA. The empty pCS2+ vector was used both as a control and to adjust the DNA amount to 1.0-1.5 µg/well. After 24 h, a luciferase reporter assay was performed using a Dual-Luciferase Assay Kit. TOPFlash luciferase activity was normalized by the luciferase activity of Renilla. The in vivo luciferase assay was performed as previously described (Feng et al., 2012; Rong et al., 2014, 2017). Briefly, one- to two-cell stage embryos were injected with MOs and/or mRNA plus 100 pg of TOPFlash DNA and 20 pg of Renilla plasmid DNA, and then reared to the shield stage. Two independent groups of embryos (each with more than 20 embryos) were lysed and measured each time.

The stable EIF4A3 knockdown cell line was established by lentiviral delivery of shRNA in the HCT116 cell line with a p53^−/− genetic background (a gift from Dr Jun Chen, Zhejiang University, China). EIF4A3 shRNA-1 sequence 5′-GGATATTCAGGTTCGTGAA-3′ and EIF4A3 shRNA-2 sequence 5′-GCTCTCGGTGACTACATGA-3′ were used (Alexandrov et al., 2011). The lentiviral pLKO.1-GFP+Puromycin vector was selected as a shuttle vector. The shEIF4A3 plasmid and two packaging plasmids were co-transfected into HEK293T cells. After 48 h, supernatant was collected and filtered through a 0.22 µm filter. Viral supernatant containing 8 µg/ml polybrene was added to 30-40% confluent HCT116 cells for infection. After 48 h of incubation, the puromycin was added to select puromycin-resistant cells. The knockdown efficiency was detected by immunoblotting.

Cells were transfected with various constructs at 70%-80% confluence and harvested 24 h after transfection. Protein complexes were precipitated from whole-cell lysate or indicated zebrafish embryo deyolked lysates with indicated immunoprecipitation antibodies and enriched by protein A/G PLUS-Agarose. Immunoprecipitates were eluted by boiling the beads in loading buffer, followed by immunoblotting.

Cell lysates, embryonic extracts or immunoprecipitated protein eluates were separated by SDS-PAGE and blotted onto PVDF membranes for immunoblotting. GAPDH and Histone H3 were detected as the loading control when needed. All results are from three independent experiments. The figures shown are representative results.

HeLa cells were fixed with 4% paraformaldehyde for 10 min at room temperature, followed by 0.2% Triton X-100 treatment for 5 min and blocking with 20% BSA (bovine serum albumin). The cells were then incubated with corresponding primary and secondary antibodies along with DAPI for visualization of the nuclei. Fluorescence images were acquired with a Leica TCS SP8 confocal microscope.

Expression of GST or GST-tagged EIF4A3 in E. coli BL21(DE3) cells was induced using 0.1 mM IPTG; cells were allowed to grow for 2 h at 37°C. GST fusion proteins were affinity purified on glutathione Sepharose 4B beads (GE Healthcare) according to the manufacturer's instructions. HEK293T cells were then transfected with Flag-tagged β-catenin or Myc-tagged TCF7L2. These proteins of cell lysates were incubated with GST or GST-tagged EIF4A3 with continuous rotation at 4°C for overnight. Proteins were washed four times to remove non-specific bound proteins from the beads. After incubation, proteins bound to the agarose beads were eluted in SDS sample buffer by boiling and analyzed by standard western blotting.

ChIP assays were conducted using a ChIP Assay Kit (Millipore) according to the manufacturer's protocol. Briefly, HEK293T cells (2×10⁷) were fixed with fresh formaldehyde. Chromatin in cell lysates was sheered to ∼300-900 bp using a VCX 130 Sonicator (Sonics & Materials; 18×10 s on, 17×10 s pulses, 30% amplitude, ChIP samples kept on ice water). Precipitated DNA samples were analyzed by semi-quantitative and quantitative PCR. The qPCR data were expressed as the percentage of input DNA. Samples from three independent experiments were used and each sample was measured in duplicate. The representative figures of semi-quantitative PCR are shown.

Data are presented as means±s.e.m. Two-tailed, unpaired Student's t-test was used for comparisons between two groups or one-way analysis of variance followed by Tukey's post-hoc test was used for comparisons among multiple groups. Differences among groups were analyzed using GraphPad Prism version 5.01 and significance was defined as P<0.05 or smaller P values.

We are grateful to Dr Cunming Duan from the University of Michigan (Ann Arbor, USA) for critical suggestions. We thank Dr Hongyan Li from the Ocean University of China for helping to generate plasmids of TALENs, and Dr Anming Meng from Tsinghua University, Dr Jun Chen from Zhejiang University, Dr Ying Cao from Nanjing University and Dr Ming Shao from Shandong University for providing reagents.