AP-2α and AP-2β cooperatively function in the craniofacial surface ectoderm to regulate chromatin and gene expression dynamics during facial development

Essential revisions:

1) Validation of the genetic mouse models employed is needed. The activity of the ectodermal Cre used in this study (Crect) was shown by Schock et al., 2017 to be variable. It is not clear how variable Crect activity was controlled for in this study and the extent to which this may impact variability. Include images of the Crect recombination pattern, and AP-2a and AP-2b expression at the critical stages analysed in the mutants, as this would facilitate a better understanding of the phenotypes produced and also show whether AP-2a and AP-2b are co-expressed at all times in the facial ectoderm and if Crect encompasses their full expression domains.

3) Further validate the bioinformatic analysis with in vivo expression analysis of these genes during development. For example, do Tfap2a and Tfap2b expression patterns overlap in the same cells? Are Tfap2a and Tfap2b expressed extensively in the facial prominence epithelium or are they more specifically associated with the most affected regions (e.g. lambdoid junction, mandible fusion site, etc.) in the knockout mice?

We have provided additional information in Figure 1, including a new Figure (Figure 1—figure supplement 1) and incorporated pertinent information in the Materials and methods section to address these two issues.

First, regarding the Crect transgene, we generated this line and have maintained it in our colony for many years. We quickly noticed that, like many transgenes, variable expression can be evident between generations. Therefore, we choose males that have continued to provide a relatively stable expression pattern in the ectoderm over many generations. Further, we test every new batch of Crect males against a Rosa26 reporter line expressing LacZ or GFP to ensure that we only keep males that produce the required expression pattern in numerous embryos. The desired recombination pattern is shown in whole mount in Figure 1—figure supplement 1A, and in sectioned material in Figure 1—figure supplement 1B-D, with reporter expression largely confined to the surface and oral ectoderm, with additional expression noted in otic and optic epithelium. Any animals that produce the mesenchymal pattern reported in Schock et al., (2017) are discarded. Following these procedures, we have observed a consistent set of mutant phenotypes in the EAKO, EBKO and EDKO embryos used in electron microscopy, skeletal staining, and WMISH analysis, as well as in embryos used for DNA and RNA preparation.

We agree with the reviewers that the efficacy and specificity of Cre recombination, as well as the overlap between expression of Crect/Tfap2a/Tfap2b in the ectoderm, are critical issues that require further documentation, and we are now providing additional figures to validate our findings. Data pertinent to the overlap between the expression of Tfap2a, Tfap2b, and the Cre transgene in the ectoderm are shown in both Figure 1 panel B and Figure 1—figure supplement 1, panel E. These panels show feature plots based on the tSNE data from Figure S2 of the paper Li et al., 2019 which used Crect to mark the ectoderm with the R26R LacZ reporter. In Figure 1B, we have outlined the various cell types present in the single cell clustering analysis and shown feature plots for Tfap2a and Tfap2b. The most concentrated expression of the two Tfap2 genes maps to the surface ectoderm and periderm, but expression is also apparent in mesenchyme as previously reported in Van Otterloo et al., 2018. Blending of these plots illustrates the considerable overlap between expression of these two AP-2 genes in the surface ectoderm and periderm at cellular resolution. Similarly, Figure 1—figure supplement 1 shows how Cre is specifically expressed in the ectoderm, and how it significantly overlaps in that tissue layer with Tfap2a and Tfap2b.

With respect to the efficacy of Crect in targeting Tfap2a and Tfap2b in the ectoderm we have now included Figure 1—figure supplement 2. The figure shows the results of RNAseq analysis performed on two separate isolated ectoderm samples from control and EDKO mutant embryos examining expression associated with the Tfap2a and Tfap2b loci. Targeting of the floxed Tfap2a allele removes exons 5 and 6, while exon 6 is lost from the floxed Tfap2b locus. These data show a >95% reduction in transcripts containing these floxed exons.

We have now amended the manuscript to provide further information concerning Crect and the Tfap2a and Tfap2b genes. Specifically, we have rewritten the first paragraph of the Results section to include a sentence referring to overlap between Tfap2a and Tfap2b expression in the surface ectoderm “Further mining of single cell RNA-seq data derived from facial prominences indicated that Tfap2a and Tfap2b expression also displayed significant overlap within cells of the surface ectoderm and periderm” and changed Figure 1 to show these data.

We have also added a new section in the Materials and methods section concerning the mouse strains:

“Although the Crect transgene is used here to target the early embryonic ectoderm, a previous report has indicated that it frequently produces a broader pattern of recombination (Schock et al., 2017). We used several approaches to avoid this broader expression pattern. First, the Crect transgene was always introduced into the experimental embryos via the sire to reduce global recombination sometimes seen with transmission from the female. Second, all sires were tested using reporter lines such as (Gt(ROSA)26Sor^tm1Sor) or mT/mG, Gt(ROSA)26Sor^{tm4(ACTB-tdTomato,-EGFP)Luo/J} to ensure that they consistently produced the desired pattern of recombination before they were used to generate EAKO, EBKO, or EDKO animals (Figure 1—figure supplement 1). We also confirmed the overlap between the expression of Cre, Tfap2a, and Tfap2b in the ectoderm by mining a previously published single cell RNAseq dataset (Li et al., 2019). Lastly, we determined that the Crect transgene was highly efficient at targeting the Tfap2a and Tfap2b loci based upon RNA expression associated with these genes in Cre expressing cells (Figure 1—figure supplement 2).”

2) Can the authors provide an explanation for why the phenotype of the Tfap2b mutant (EBKO) mice is milder than that of the Tfap2a mutant (EAKO) mice or the double knockout (EDKO) mice.

4) The authors need to analyze the compensatory mechanism of Tfap2a and Tfap2b during craniofacial development. Is it possible that Tfap2a can compensate for the loss of Tfap2b, while Tfap2b cannot do the same for Tfap2a? The authors need to provide further analysis, such as testing the expression of Tfap2a and Tfap2b in EBKO and EAKO mice, respectively, and compare the data to the control.

We have considered several scenarios that could account for these observations, but a thorough experimental understanding of this phenomenon is beyond the scope of the current manuscript. One possibility is that there might be subtle differences in the temporal or spatial expression profiles of these factors. An alternative possibility is that there are differences in the mRNA and/or amino acid sequence that might differentially affect stability or localization of the transcripts or the proteins. Third, even though the proteins are quite similar in overall sequence and binding specificity, they may nevertheless interact with suites of unique and/or overlapping co-factors with AP-2α having more critical interactions than AP-2ß. A further possibility is that there may be cross-regulation between Tfap2a and Tfap2b at some level. To examine this last possibility, as requested, we have performed preliminary studies to examine the expression of Tfap2a and Tfap2b in the ectoderm of both EAKO and EBKO mice. However, these studies have not produced a definitive answer concerning compensation caused by loss of one or other transcription factor. Therefore, the mechanism(s) by which AP-2α has the more critical function remains unclear. We believe it will require considerable time and effort to perform studies that will definitively test any of these hypothesis—for example by substituting Tfap2a into the Tfap2b locus and vice versa—and that this is beyond the scope of the current analysis. At the same time, we also note that there are multiple examples throughout the literature where TFAP2 paralogs function redundantly/cooperatively, yet loss of Tfap2a alone provides a more dramatic phenotype than deletion of an alternative paralog. For example, we found that, while AP-2α and AP-2ß function cooperatively in the neural crest, loss of AP-2α alone in the neural crest results in a more severe phenotype than loss of AP-2ß alone with respect to both craniofacial development and melanocyte biology (Van Otterloo et al., 2018 and Seberg et al., 2017). This phenomenon has also been noted in the zebrafish model system (e.g., tfap2a/tfap2c, Li and Cornell, 2007; tfap2a/tfap2e, Van Otterloo et al., 2010).

However, we have now added the following short statement in the discussion to provide possible explanations for why the phenotype of the Tfap2b mutant (EBKO) mice is milder than the Tfap2a mutant (EAKO) mice.

“The reasons behind the more prominent function of AP-2a in these systems remains unclear, especially since the two proteins are over 70% identical and bind to the same DNA recognition sequences. Nevertheless, it is possible that some alterations in the mRNA and/or amino acid sequence could differentially affect stability, localization, or interaction with cofactors. Alternatively, there might be subtle differences in the timing, distribution, or levels of functional AP-2α and AP-2ß protein in these tissues that account for the changes in susceptibility to pathology. Further studies will be needed to determine how the consequence of Tfap2a loss is greater than Tfap2b.”

5) Co-localize the expression of Tfap2a and Tfap2b with Wnt-related molecules to confirm whether Tfap2 genes in the epithelium regulate Wnt signaling molecules in a cell-autonomous manner or through cell-cell interaction?

6) Axin2-lacZ reporter mice showed decreased expression of Axin2 in both facial ectoderm and mesenchyme in EAKO and EDKO mice. The conclusion that loss of AP-2 proteins in the ectoderm leads to reduced Wnt signaling in both the ectoderm and underlying mesenchyme should be better supported. Confirm reduced Wnt signaling activity in the EDKO using immunofluorescent staining for active β-catenin. To explain whether this decrease is directly caused by decreasing expression of Wnt related molecules in facial ectoderm, the authors should include more data on those genes in the mesenchyme.

We have added two new Supplementary figures to address (1) the issue of the overlap between the AP-2 transcription factors and WNT signaling pathway genes in the ectoderm and (2) how loss of AP-2 in the ectoderm affects the expression of WNT pathway genes in the mesenchyme.

The first figure (Figure 4—figure supplement 1) shows feature plots derived from the E11.5 single cell sequencing data (Li et al., 2019) for multiple Wnt ligands and associated genes involved in WNT pathway regulation. In addition, we show the overlap with Tfap2a expression for each of these 18 genes in the whole dataset and the surface ectoderm. (Tfap2b expression is not included due to the considerable overlap between Tfap2a and Tfap2b shown in new Figure 1B). Overall, the data show that the ligands Wnt3, Wnt4, Wnt6, Wnt9b and Wnt10a, as well as other WNT pathway modulators including Kremen2 and to a lesser extent Dkk4 are almost exclusively expressed in the ectoderm and that the majority of cells expressing these ligands also co-express the Tfap2 genes in the surface ectoderm. The data on ectodermal bias in expression for these genes are in agreement with the analyses of Hooper et al., 2020 which used RNAseq analysis to examine expression in the facial ectoderm and mesenchyme compartments. We used the analysis of Hooper et al., as shown in Supplementary File 3 to stratify genes according to ectodermal/mesenchymal ratios of expression in Figure 7E and Figure 7—figure supplement 1.

To summarize, our WMISH, RT-PCR and RNAseq findings demonstrate that there is a significant reduction in the ectodermal expression of Wnt related molecules in the EDKO mutants compared to controls (Figures 4, 7, Figure 4—figure supplement 4 and Figure 7—figure supplement 2). Further, there is significant overlap for many of these genes with Tfap2 expression in the ectoderm and we also we show that several of them are associated with AP-2 dependent ATAC-seq peaks (Figures 3 and 4). Together, these findings would support a hypothesis that regulation of these genes is direct and cell autonomous in nature. However, definitive proof that regulation is exclusively cell autonomous will require considerable further experimentation.

The second figure (Figure 7—figure supplement 3) provides further information regarding how Wnt pathway gene expression is affected in the mesenchyme when Tfap2a and Tfap2b are lost from the ectoderm. In addition to the information concerning the reduction of Axin2 expression which we reported in Figure 4 and Figure 4—figure supplement 4, we now show results from RT-PCR analysis of 5 other Wnt pathway genes that show significant expression in the mesenchyme. Samples of facial prominence mesenchyme were collected from E11.5 control and mutant embryos, RNA extracted, cDNA synthesized, and real-time PCR conducted on these 5 transcripts. We identified an ~ 2 to 5 fold reduction in expression of all 5 transcripts within the mesenchyme of EDKO’s versus controls. These findings are consistent with loss of ectodermal Tfap2 having a non-cell autonomous impact on these Wnt-associated genes in the mesenchyme.

We have amended the text of the results in several places to incorporate the data included in these two new Supplemental figures.

7) It's unclear whether Wnt1 overexpressing mice have an endogenous phenotype, and this should have been presented as a control. It's also important to show that overexpressing Wnt1 in the surface ectoderm is sufficient to elicit a Wnt signaling response in the ectoderm and underlying mesenchyme for these mice possibly via Axin2-lacZ, but other downstream markers would be required. Alternatively, conditionally overexpress B-catenin in the facial ectoderm or treat the faces of E11.5 conditional AP-2a and AP-2b mutants in explant culture with exogenous Wnt3a or Wnt9b.

We have added a new Supplementary figure (Figure 8—figure supplement 1) to address the issue of the endogenous Crect Wnt1 overexpression (OX) phenotype. These images show the phenotype of E12.5 control and Wnt1 overexpression mice, a time point when facial fusion has occurred in both genotypes. Although we have not analyzed the Wnt1 overexpression phenotype in detail, there are certain noticeable differences from controls. First, the angle between the forebrain and snout is more pronounced in the overexpressing animals. Second, the eye is abnormal in overexpressing mice, with an apparent suppression of lens formation. Note that at later time points the Crect Wnt1 OX animals develop severe overall edema and do not survive until term. We have not explored these phenotypes further as it is not the focus of the manuscript.

With respect to the other suggestions, we have also examined the Crect Wnt1 OX phenotype in combination with Axin2-LacZ, but do not see major quantitative or qualitative differences in staining. We also note that there were no appreciative changes in b-catenin protein distribution between mutants and controls. These observations fit with our current working hypothesis for many types of cleft lip and primary palate—that just subtle changes in signaling can prevent normal attachment of the prominences at the lambdoid junction. As a result, subsequent continued growth of the prominences—unrestrained by the absence of fusion—causes major morphological defects. Lastly, in respect to this model, the eye phenotype resembles that observed in the study by Smith et al., 2005 (Dev Biol, vol 285, pp 477-489, “The duality of β-catenin function: a requirement in lens morphogenesis and signaling suppression of lens fate in periocular ectoderm”). In this study, a form of ß-catenin that is unable to be degraded is expressed in the presumptive lens placode and this gain of function in WNT signaling also suppresses lens formation.

We also considered using expression of stabilized ß-catenin under the control of the Crect transgene to rescue the phenotype. In this respect, we have previously conditionally overexpressed the stabilized form of ß-catenin in the surface ectoderm using Crect (Dev Biol (2011), vol 349 pp 261-269). However, facial and vascular defects are more severe in this ß-catenin model, with the mice failing to survive beyond mid embryogenesis and so they are not as useful for the current analysis. We also note that stabilization of ß-catenin in the ectoderm is not equivalent to overexpression of the Wnt1 ligand; the former will have its primary effect on the ectoderm, whereas the latter will act both on the ectoderm in a paracrine fashion, but also on the underlying mesenchyme. Finally, we appreciate the suggestion of using explant culture to study the interaction between the AP-2 transcription factors and the WNT pathway, and we hope to explore this in the future, but we believe that this is outside the scope of the current manuscript, especially since we postulate that changes in the overall morphogenesis of the head are involved in the AP-2 dependent pathology.

In addition to the new Supplemental Figure, we have added the following statement in the Results section:

“First, though, we examined how Crect mediated Wnt1 overexpression in the ectoderm might impact face development to assess its suitability as a rescue model (Figure 8—figure supplement 1). In common with controls, E12.5 Crect Wnt1^ox embryos had completed fusion of the face to form an intact upper lip. However, there were developmental changes in that mutant animals had a more pronounced angle between the forebrain and snout than controls and there were also defects in eye formation, consistent with activation of the Wnt pathway in this process (Smith et al., (2005)). Nevertheless, based on the overall facial phenotype, we reasoned that this approach was feasible to supplement Wnt ligand expression in the facial ectoderm of the EAKO and EDKO mice.”

8) The paper would benefit from demonstrating that exogenous Wnt signaling can restore proliferation and patterning, and from additional analyses of potential changes in expression of other signaling pathways in the ectoderm and neural crest cells, known to be associated with syngathia, maxillo-mandibular patterning and facial clefts.

The analyses we have performed to date point to the WNT pathway as being the most severely affected signaling system that is both heavily altered in our model, associated with craniofacial clefting defects, and able to rescue aspects of the mutant phenotype. Therefore, we have highlighted this pathway in our studies. Nevertheless, we cannot exclude other mechanisms contributing to the phenotype—either alone, or in association with the WNT pathway.

At the level of RNA transcript analysis, we report in Table 1 changes in components of CXCL, EDN, and FGF signaling, but these do not appear to occur at levels that would account for the observed phenotypes (this is mentioned within the Results section). With respect to jaw patterning and syngnathia, we do not find any major differences at E10.5 in the expression levels of genes previously associated with causing similar defects in these tissues such as Foxc1, Six1, Bmp4, Fgf8 or Dlx family members. As noted, there was some slight change in Ednra expression, and a major change in the level of Gbx2 (which has previously been reported to be a target of DLX5/6), but at this point nothing that would account for the observed jaw phenotypes. The analysis of the roles of AP-2 in jaw patterning and syngnathia are the subject of ongoing studies in our laboratories and our current data indicate that this likely utilizes a mechanism parallel to the DLX, EDN, SIX, and FOXC1 pathways.

9) Consider an in vitro cell culture model to test whether Wnt3 and Wnt9b have any role in perturbing the mesenchymal gene expression program. Furthermore, there is a possibility that Wnt1 may not function exactly the same as Wnt3 and Wnt9b. Could the authors confirm that this ectopic Wnt1 can restore the reduced Wnt activity associated with the reduced Wnt3 and Wnt9b in EAKO and EDKO mice?

We appreciate the reviewers’ suggestion to consider an in vitro model in examining WNT3 and WNT9B in more detail and agree that these specific WNTs are good candidates given their previously reported involvement in craniofacial development (Carroll et al., 2005; Juriloff et al., 2006; Niemann et al., 2004; Chiquet BT et al., 2008; Yao T et al., 2011; Fontoura C et al., 2015). However, we note that the expression of a wide range of Wnt ligands are impacted in our model, suggesting the likelihood of a combinatorial influence of multiple WNTs on the phenotype observed. Further, given WNT3/WNT9B’s activation of the canonical WNT pathway (Carroll et al., 2005) and WNT1’s primary ability to activate this pathway, including previous evidence of WNT1’s ability to substitute for WNT9B signaling (Carroll et al., 2005), the utilization of over-expressed WNT1 we feel is well justified. We further note that the Wnt1 OX model has also been used by Ferretti et al., (Dev Cell, vol 621, pp 627-641 (2011) “A Conserved Pbx-Wnt-p63-Irf6 Regulatory Module Controls Face Morphogenesis by Promoting Epithelial Apoptosis”) as a Wnt rescue experiment for a Pbx-based CL/P model.