All NPB TFs are required for PPE formation
To elucidate which of the early ectodermal TFs are required for the proper establishment of PPE and NC territories, we first investigated how MO-mediated knockdown of these TFs affects Six1, Eya1, Sox3 and FoxD3 expression at the NPB. The efficacy and specificity of all MOs used has been validated in previous studies which included rescue experiments of various genes expressed at the neural plate border (Additional file 1: Table S1). Moreover, with the exception of Hairy2b and Vent2 (for which no orthologous genes are known in rodents), mutants in genes encoding these TFs (FoxI1, Msx1, AP2, Pax3, Zic1) in mouse and/or zebrafish are perturbed, like morphants, in NPB-derived tissues (Additional file 1: Table S1). Similar to previous studies, we initially injected high doses of MOs (10–20 ng) for each TF. This resulted in strong reduction of Six1 and Eya1 expression in the PPE of most embryos after knockdown of each TF (Additional file 1: Table S2). To reduce the probability of unspecific side effects of MOs, we then performed a more extensive analysis of MO knockdown phenotypes after injecting much lower doses (1–2 ng) of these MOs. Even at these low doses, MOs perturbed NPB development at a relatively high frequency; however, the phenotypes tended to be less severe than after injection of higher doses (e.g. resulting in relatively mild rather than strong reduction of NPB marker expression) (Fig. 1, Additional file 2: Figure S1, Additional file 1: Table S3).
Six1 and Eya1 expression in the PPE was downregulated in a high proportion of embryos after knockdown of both the dorsally restricted TFs Zic1, Pax3 and Hairy2b and the ventrally restricted TFs AP2, Vent2 and FoxI1a and in a smaller proportion after knockdown of the ventrally restricted Msx1 (Fig. 1a–g, Additional file 2: Figure S1, Additional file 1: Table S3). This indicates that all of these TFs are required for establishing Six1 and Eya1 expression in the PPE. Sox3 expression in the PPE was also reduced after knockdown of most TFs but rarely or never after Pax3 and Msx1 knockdown (Fig. 1o–u, Additional file 1: Table S3). Whether this reduction of Sox3 is due to a direct requirement of these TFs for placodal Sox3 expression or an indirect consequence of the downregulation of Six1 and Eya1 in the absence of these TFs remains to be determined. In contrast to other TFs, Msx1 and Pax3 loss of function typically resulted in the expansion of Sox3 expression from the PPE into adjacent non-neural ectoderm and this was occasionally also observed after Hairy2b and FoxI1a knockdown. Taken together this suggests a requirement of Zic1, AP2 and Vent2 (and to some extent Hairy2b and FoxI1a) in activating Sox3 expression in the PPE, whereas Msx1 and Pax3 (and to some extent Hairy2b and FoxI1a) are mostly required for repression of Sox3 throughout the non-neural ectoderm.
Knockdown of each TF also led to reduction of FoxD3 expression in the NC in at least some embryos (Fig. 1h–n, Additional file 1: Table S3). However, only Zic1, Pax3, Msx1 and Vent2 loss of function showed reduction of FoxD3 in the majority of embryos. In contrast, after Hairy2b, AP2 and FoxI1a loss of function, FoxD3 was reduced only in some embryos but was increased in others. This suggests that while each TF is required for establishing FoxD3 expression in the NC, Hairy2b, AP2 and FoxI1a play additional roles for restricting FoxD3 expression to the NC domain in distinct and partly counteracting pathways.
Finally, Sox3 expression in the neural plate was broadened (and the expression domains of FoxD3, Six1 and Eya1 were laterally displaced) in most embryos after knockdown of Pax3, Msx1 and Vent2 and in a minority of embryos (and usually only mildly) after knockdown of Zic1, Hairy2b and FoxI1a (Fig. 1o–u, Additional file 1: Table S3) suggesting that these TFs—in particular Pax3, Msx1 and Vent2—contribute to define the lateral limit of Sox3 expression in the neural plate.
The dorsally restricted TFs Zic1 and Pax3 are cell-autonomously required for PPE formation
In our knockdown experiments, MOs were injected at 2–8 cell stages and, thus, potentially could exert their effects by blocking translation of their target mRNAs in all germ layers. Since Msx1 and Vent2 are expressed in both mesoderm and ectoderm during gastrulation and early neural plate stages [11, 36], we can thus not rule out that some of the deficiencies in NPB marker expression after knockdown of Msx1 or Vent2 may reflect mesodermal rather than ectodermal functions of these genes in NPB formation. In contrast, Zic1, Pax3, Hairy2b, AP2 and FoxI1 are predominantly ectodermally expressed during gastrulation and early neural plate stages [27], suggesting that the phenotypes observed reflect a function of these TFs in the embryonic ectoderm.
However, during gastrulation, Zic1 and Pax3 become confined to a dorsal, neural ectodermal territory with a progressively decreasing degree of overlap and increasingly sharper boundary with the expression domains of Six1 and Eya1 in the PPE or with expression of Dlx3, GATA2 or FoxI1a in the non-neural ectoderm [14, 15, 19, 27]. This raises the possibility that Zic1 and Pax3 may be non-cell-autonomously required for PPE formation in the adjacent neural plate (e.g. by promoting the formation of signaling molecules required for PPE formation). To determine whether Zic1 and Pax3 are cell-autonomously required for PPE formation in the presumptive PPE ectoderm (presumably before the end of gastrulation when expression domains still overlap) or are instead required in the adjacent neural plate, we grafted the neural plate from embryos injected with Zic1 MO or Pax3 MO orthotopically into uninjected embryos or vice versa (Additional file 3: Figure S2), thereby juxtaposing Zic1 MO- or Pax3 MO-injected neural plate ectoderm with uninjected ectoderm in the PPE region. Control experiments with grafts from GFP-injected embryos showed that the grafting procedure itself did not affect Six1 expression (Additional file 3: Figure S2 A). Similarly, no reduction of Six1 expression was observed after grafting neural plates from Zic1 MO- or Pax3 MO-injected embryos into uninjected hosts (Additional file 3: Figure S2 B, D). Conversely, grafting neural plates from uninjected embryos into Zic1MO- or Pax3MO-injected embryos was unable to rescue reductions of Six1 expression observed in the host PPE (Additional file 3: Figure S2 C, E). Taken together, this indicates that both Zic1 and Pax3 are required cell-autonomously for PPE formation.
AP2 and Msx1 are sufficient to promote PPE markers in neural ectoderm
We next tested whether overexpression of any of the early ectodermal TFs is sufficient to promote the activation of PPE or NC markers. Since injection of mRNAs encoding these TFs (Additional file 1: Table S4) often affected early development and may lead to gastrulation defects (especially for Hairy2b and Vent2), we also injected hormone-inducible constructs of TFs, which were activated by dexamethasone treatment at the end of gastrulation (Fig. 2a–g, Additional file 4: Figure S3, Additional file 1: Table S5). Overexpression of all dorsally restricted TFs, Zic1, Pax3 and Hairy2b, reduced Six1 and Eya1 expression in the PPE. Pax3, in particular, resulted in very strong and often complete repression of Six1 or Eya1, while Zic1 and Hairy2b had milder effects. Overexpression of the ventrally restricted TFs also led to occasional reductions of Six1 and Eya1 expression (most frequently for AP2 and Vent2). However, overexpression of AP2 and Msx1 also promoted ectopic expression of Six1 and Eya1 not only in the non-neural ectoderm but also in the neural plate similar to what was previously described after Dlx3 overexpression [15]. This suggests that AP2 and Msx1 play a central role in PPE formation possibly by endowing ectoderm with non-neural ectodermal competence as previously shown for AP2 in zebrafish [29]. Sox3 expression in the PPE was reduced after the overexpression of Pax3, Hairy2b, AP2 and Vent2 but unaffected by Zic1, Msx1 or Vent2 overexpression (Fig. 2o–u, Additional file 1: Table S5) indicating that its regulation in the PPE depends on different combinations of TF than Six1 or Eya1.
The effects of overexpression of most TFs on the NC were more complex and variable. While overexpression of each TF led to reduced FoxD3 expression in a subset of embryos, overexpression of each TF except AP2 and Msx1 also led to the expansion of FoxD3 expression (but never to ectopic expression) in another subset of embryos (Fig. 2h–n, Additional file 1: Table S5). Taken together, this suggests that these TFs act in a complex and combinatorial fashion to promote NC expression.
Sox3 expression in the neural plate was reduced in scattered cells after overexpression of Pax3, Msx1 and Vent2 (Fig. 2o–u, Additional file 1: Table S5) in accordance with the proposed role of these TFs in defining the lateral border of neural Sox3 expression.
AP2, Msx1 and Dlx3 promote PPE formation via different pathways
The observation that many dorsally restricted TFs including Zic1 and Pax3 (see above) but also Sox3 (Additional file 1: Table S6) repress Six1 and Eya1 expression in the PPE suggests that the ability of AP2, Msx1 and Dlx3 [15] to ectopically activate Six1 and Eya1 in the neural plate may depend on their ability to repress some or all dorsally restricted TFs (Msx1 and Dlx3 repress Sox3: see above and [15]; AP2 represses Zic1: see [19]). To test this, we determined whether coinjection of Zic1, Pax3 or Sox3 could prevent ectopic neural expression of Six1 after AP2, Dlx3 or Msx1 injection (Fig. 3a, b; Additional file 1: Table S6). The frequency of ectopic neural Six1 expression was indeed significantly reduced after coinjection of AP2 with Zic1 (but not with Sox3 or Pax3) or coinjection of Dlx3 or Msx1 with Sox3 (but not with Zic1 or Pax3 in the case of Dlx3; these were not tested for Msx1) (Fig. 3a; Additional file 1: Table S6). This suggests that AP2 and Dlx3/Msx1 promote PPE formation in neural ectoderm via different pathways, viz. by inhibition of Zic1 and Sox3, respectively. Indeed, Sox3 immunostaining in vibratome sections of embryos in which Six1 was ectopically expressed in the neural plate after overexpression of Dlx3 or Msx1, shows that Sox3 is specifically reduced in the injected part of the neural plate in which Six1 is ectopically expressed (Additional file 5: Figure S4). Whether Zic1 is similarly reduced in the area of AP2 overexpression remains to be determined once a specific antibody recognizing Xenopus Zic1 becomes available.
Zic1 and Pax3 promote PPE formation only in Dlx3-expressing ectoderm
While coinjection of Zic1 or Pax3 with Dlx3 does not significantly alter the frequency of ectopic Six1 expression in the neural ectoderm, it significantly increases the frequency of ectopic Six1 expression in the non-neural ectoderm compared to injection of either Dlx3 or Pax3 alone, which never promote non-neural Six1 expression or to Zic1 alone, which promotes Six1 only in a small subset of embryos (Fig. 3b, Additional file 1: Table S6). Conversely, coinjection of Dlx3 MO with Zic1 completely blocks the ability of Zic1 to promote Six1 expression (Fig. 3b Additional file 1: Table S6). This suggests that Zic1 and Pax3 can promote Six1 only in Dlx3-expressing ectoderm. Coinjection of Zic1 (but not Pax3) and Dlx3 also significantly reduces the frequency of decreased Six1 or Eya1 expression in the PPE compared to overexpression of Zic1 or Dlx3 alone (Fig. 3c, Additional file 1: Table S6) suggesting that the combination of both TFs protects against the repressive effect of each TF alone.
Similarly, coinjection of Dlx3 MO with Zic1 significantly reduces the ability of Zic1 to promote FoxD3 expression (Fig. 3d, Additional file 1: Table S6). However, Dlx3 overexpression represses FoxD3 at high frequency, which is significantly reduced by coinjection of Zic1 (Fig. 3d, Additional file 1: Table S6). Taken together, this indicates that Zic1 also requires Dlx3 for NC formation, and protects FoxD3 from repression by Dlx3.
Our results demonstrate that Zic1 and Pax3 are required for the cell-autonomous activation of Six1 in the PPE but do so only in conjunction with Dlx3. However, Dlx3 and another ventrally restricted TF GATA2 were previously shown to repress Zic1 and Pax3 [15]. Taken together, this suggests that dorsally restricted TFs Zic1 and Pax3 may be required for the initiation of PPE formation in Dlx3-expressing ectoderm but subsequently become excluded from the Dlx3-expressing part of the ectoderm. To determine whether cross-repressive interactions contribute to the sharpening of the boundary between non-neural ectoderm expressing the ventrally restricted TFs Dlx3, GATA2 and FoxI1a and neural ectoderm expressing Zic1 and Pax3, we injected Zic1 and Pax3 and analysed the effect on Dlx3, GATA2 and FoxI1a expression (Fig. 3e, Additional file 1: Table S7). While FoxI1a and GATA2 expression was reduced, Dlx3 was not affected indicating that Zic1 and Pax3 indeed repress some but not all ventrally restricted TFs.
Cross-regulation of NPB TFs by Six1 and Eya1
We finally analysed the expression of NPB TF genes (Zic1, Pax3, AP2, Msx1, FoxI1a, Dlx3 and GATA2) as well as dedicated PPE (Six1, Eya1), NC (FoxD3) and neural plate markers (Sox3) using injection of Six1 and Eya1 MOs (Fig. 4, Additional file 6: Figure S5, Additional file 1: Table S8) and mRNAs (Fig. 5, Additional file 7: Figure S6, Additional file 1: Table S9) to determine whether Six1 and Eya1 cross-regulate these other TFs. Again, the efficacy and specificity of the Six1 and Eya1 MOs used has been validated in previous studies (Additional file 1: Table S1). Since Six1 and Eya1 MOs were injected at 2–8 cell stages, we cannot completely rule out that some of the observed phenotypes reflect early embryonic or non-ectodermal roles of Six1 and Eya1. However, up to neural plate stages expression of both genes is largely confined to the NPB ectoderm as well as to a domain in the paraxial mesoderm, which is much more medial and posterior than the NPB [37, 38] suggesting that the deficits observed in the NPB after Six1 or Eya1 knockdown reflect mostly their ectodermal function.
Knockdown of either Six1 or Eya1 leads to reductions of Eya1, Six1 and Sox3 expression in the PPE; reductions of FoxD3 in the NC; lateral displacement of Six1, Eya1 and FoxD3; and broadening of Sox3 expression in the neural plate (Fig. 4a–c, Additional file 6: Figure S5 A-C, Additional file 1: Table S8). This suggests that Six1 and Eya1 themselves are required for PPE as well as NC formation. It remains possible that gastrulation defects (impaired convergence-extension), which are sometimes observed after knockdown of Six1 or Eya1 contribute to the observed shift of the neural plate border. However, lateral displacement of PPE domains of Eya1 or Six1 after Six1 or Eya1 knockdown, respectively, was also observed in embryos with relatively normal Six1 or Eya1 expression in the paraxial mesoderm (which should also be affected by gastrulation defects), suggesting that Six1 and Eya1 also play a more direct role in setting the lateral border of the neural plate.
To gain insights into how Six1 and Eya1 modulate the establishment of different ectodermal territories at the NPB, we also analysed the effects of Six1 and Eya1 knockdown on earlier ectodermal TFs. Knockdown of either Six1 or Eya1 slightly reduces the level of expression for genes encoding ventrally restricted TFs FoxI1a, Dlx3 and GATA2 and shifts their expression boundaries laterally (Fig. 4, Additional file 6: Figure S5, Additional file 1: Table S8) suggesting that Eya1 and Six1 appear to be required for the maintenance of high-level expression of ventrally restricted TFs in the PPE. Conversely, knockdown of either Six1 or Eya1 results in broader and stronger expression of Zic1, Pax3, AP2 and Msx1 in the neural plate and NC (Fig. 4, Additional file 6: Figure S5, Additional file 1: Table S8). This indicates that Six1 and Eya1 are required for repressing and laterally delimiting Zic1, Pax3, AP2 and Msx1 at the NPB, thereby helping to confine strong expression of these TFs to the NC.
We next analysed the effect of Six1 or Eya1 overexpression at the NPB. Overexpression of Eya1 often broadens Six1 and Sox3 expression in the non-neural ectoderm (although it reduces non-neural Sox3 expression in another subset of embryos) and promotes Six1 even ectopically in the neural plate (Additional file 7: Figure S6, Additional file 1: Table S9). It also results in increased or ectopic FoxD3 expression in NC and neural plate, but causes reduction of Sox3 expression in the neural plate suggesting that Eya1 promotes both PPE and NC but represses dedicated neural plate markers. While Six1 overexpression causes similar but less pronounced reductions of Sox3 in the neural plate than Eya1, it leads to reductions of Eya1 in the PPE and of FoxD3 in the NC, different from Eya1. Taken together, this suggests that Six1 despite being required for PPE and NC formation similar to Eya1 negatively regulates NPB markers in additional, Eya1-independent pathways. The ability of Six1 to interact not only with the coactivator Eya1 but also alternatively with corepressors [30] may at least partly account for these effects although this has to be confirmed in further studies.
Overexpression of Eya1 and Six1 causes a reduction of expression of some genes encoding ventrally restricted TFs such as Dlx3 and GATA2 expression, whereas, overexpression of Eya1 causes an increase in FoxI1a expression and overexpression of Six1 has variable effects on FoxI1a (Fig. 5, Additional file 7: Figure S6, Additional file 1: Table S9). Thus, while our knockdown experiments indicated that Eya1 and Six1 appear to be required for the maintenance of ventrally restricted TFs, high levels of Six1 and Eya1 seem to repress Dlx3 and GATA2.
Somewhat paradoxically, overexpression of Eya1 and Six1 has rather similar effects on NC-enriched TFs Zic1, Pax3, AP2 and Msx1 than Six1 or Eya1 knockdown generally resulting in broadening and stronger expression in the neural plate and NC with the exception that Six1 (but not Eya1) overexpression typically resulted in repression of Pax3, whereas Eya1 (but not Six1) overexpression led to reduced Msx1 expression (Fig. 5, Additional file 7: Figure S6). Thus, while our knockdown experiments demonstrate that Six1 and Eya1 are both required (possibly in a cooperative fashion) for repressing and laterally delimiting Zic1, Pax3, AP2 and Msx1 in the NC, these overexpression experiments indicate that they act as inhibitors of these TFs only in certain contexts, for example only in cooperation with other cofactors or in a dosage dependent way. Moreover, while Six1 and Eya1 may jointly promote Zic1 and AP2, they independently promote Msx1 and Pax3, respectively, presumably in conjunction with other binding partners.