Author response:
The following is the authors’ response to the original reviews.
Reviewer #1 (Public Review)
The weaknesses are in the clarity and resolution of the data that forms the basis of the model. In addition to whole embryo morphology that is used as evidence for convergent extension (CE) defects, two forms of data are presented, co-expression and IP, as well as a strong reliance on IF of exogenously expressed proteins. Thus, it is critical that both forms of evidence be very strong and clear, and this is where there are deficiencies; 1) For vast majority of experiments general morphology and LWR was used as evidence of effects on convergent extension movements rather than Keller explants or actual cell movements in the embryo. 2) The study would benefit from high or super resolution microscopy, since in many cases the differences in protein localization are not very pronounced. 3) The IP and Western analysis data often show subtle differences, and not apparent in some cases. 4) It is not clear how many biological repeats were performed or how and whether statistical analyses were performed.
(1) To more objectively assess the convergent extension phenotypes, we developed a Fiji macro to automatically quantify the LWR in various injected Xenopus embryos, as detailed in the Methods section. We acknowledge that a limitation in the current manuscript is how to link our mechanistic model at the molecular level with the actual cellular behavior during convergent extension, and we plan to perform cell biological studies in the future to elucidate the link;
(2) We have repeated some of the imaging experiments in DMZ explants using a Zeiss LSM 900 confocal equipped with Airyscan2 detector that can increase the resolution to ~100 nm. The new data are in Suppl. Fig. 4, 9, 11, 16;
(3) We have repeated all IP and western blots at least three times and provided quantification and statistical analyses;
(4) We have added the information on biological repeats and statistical analyses in all figures and figure legends.
Reviewer #2 (Public Review):
The protein localization experiments in animal cap assays are for the most part convincing, but with the caveat that the authors assume that the proteins are acting within the same cell. As Fzd and Vangl2 are thought to localize to opposite cell ends in many contexts, can the authors be sure that the effects they observe are not due to trans interactions?
In our previous publication, we provided evidence that Vangl is necessary and sufficient to recruit Dvl to the plasma membrane within the same cell (Figure 3 in 10.1093/hmg/ddx095). In a more recent publication ( 10.1038/s41467-025-57658-0 ), we further elucidated a mechanism through which Dvl oligomerization switches its binding from Vangl to Fz, and determined that Dvl binding to Vangl and Fz are differentially mediated by its PDZ and DEP domain, respectively. In the current manuscript, we also performed co-IP experiment under various conditions to demonstrate binding between Dvl and Vangl. We feel that these evidences together provide a strong argument for our model where Vangl2 acts within the same cell to sequester Dvl from Fz.
In regards to the Dvl patches induced by Wnt11 (Fig. 3 and Suppl. Fig. 9), we performed separate injection of EGFP- and mSc-tagged Dvl into adjacent blastomeres, and demonstrated that the Wnt11-induced patches arise from symmetrical accumulation of Dvl at contact of two neighboring cells (Suppl. Fig. 9a-c’). This scenario is different from epithelial PCP where Fz/Dvl and Vangl/Pk are asymmetrically accumulated at the contact between two adjacent cells.
The authors propose a model whereby Vangl2 acts as an adaptor between Dvl and Ror, to first prevent ectopic activation of signaling, and then to relay Dvl to Fzd upon Wnt stimulation. This is based on the observation that Ror2 can be co-IPed with Vangl2 but not Dvl; and secondly that the distribution of Ror2 in membrane patches after Wnt11 stimulation is broader than that of Fzd7/Dvl, while Vangl2 localizes to the edges of these patches. The data for both these points is not wholly convincing. The co-IP of Ror2 and Vangl2 is very weak, and the input of Dvl into the same experiment is very low, so any direct interaction could have been missed. Secondly, the broader distribution of Ror2 in membrane patches is very subtle, and further analysis would be needed to firm up this conclusion.
(1) We repeated the co-IP experiment with Myc-tagged Vangl or Dvl. Using the same anti-Myc antibody and experimental condition (including the expression level of Vangl, Dvl and Ror2), we still found that Ror2 could be pulled down by Vangl but not Dvl (Suppl. Fig. 15b). Whereas this data confirms our previous conclusion, we acknowledge that a negative data does not fully exclude the possibility for direct biding between Ror and Dvl.
(2) We re-analyzed the signal intensity of Dvl and Ror in Wnt11-induced patches. By quantifying the intensity ratio between Ror and Dvl along the patches, we found an increase over two folds at the border of the patches (Fig. 7j, bottom panel). We interpret this data to suggest that Ror is accumulated to a higher level than Dvl at the patch borders.
A final caveat to these experiments is that in the animal cap assays, loss of function and gain of function both cause convergence and extension defects, so any genetic interactions need to be treated with caution i.e. two injected factors enhancing a phenotype does not imply they act in the same direction in a pathway, in particular as there are both cis/trans and positive/negative feedbacks between the PCP proteins.
We agree with the reviewer that a difficulty in studying PCP/ non-canonical signaling is that both loss and gain of function of any its components can cause convergence and extension defects. Genetic interactions, especially synergistic interactions, should be interpreted with caution. But we do want to point out that, in a number of case, we were also able to demonstrate epistasis. For instance, we found that Dvl2 over-expression induced CE defects can be rescued by Pk over-expression (Fig. 1e and f), whereas Vangl/ Pk co-injection induced severe CE defects can be reciprocally rescued by Dvl2 over-expression (Fig. 1g). Likewise, we showed that Fz2/ Dvl2 co-injection induced CE defects can be rescued by wild-type Vangl2 but not Vangl2 RH mutant (Suppl. Fig. 6b), and Ror2 can rescue Vangl2 overexpression induced CE defect (Suppl. Fig. 14). Collectively, these functional interaction data consistently demonstrate an antagonism between Dvl/ Fz/ Ror2 and Vangl2/ Pk, which is correlated with our imaging and biochemical studies.
As you can see from the reviews, the referees generally agree that your paper is a potentially valuable contribution to the field. Your observations are important because of the novel model based on the inhibitory feedback regulation between planar cell polarity (PCP) protein complexes. However, the reviewers also stated that the model is only partly supported by data because of insufficient clarity and missing controls in several experiments supporting the proposed model. The paper would be significantly improved if your conclusions are backed up by additional experimentation. Specifically, the referees wanted to see the reproducibility of the results shown in Figures 3, 4, 8, S3, S7, S12.
We hope that you are able to revise the paper along the lines suggested by the referees to increase the impact of your study on the current understanding of PCP signaling mechanisms.
We thank the reviewers for careful reading of our manuscript and for their constructive critiques and suggestions. We have repeated the animal cap studies in original Figures 3, 4, 8 and S3 with DMZ explants, and the new data are in Supplementary Fig. 9, 11, 16 and 4, respectively. We also repeated the biochemical studies in original Figure S 7and 12, and the new data are in Supplementary Fig. 8 and 15.
Reviewer #1 (Recommendations For The Authors):
Major points:(1) The author conducted an analysis of the subcellular localization of PCP core proteins, including Vangl2, Pk, Fz, and Dvl, within animal cap explants (ectodermal explants). To validate the model proposing that 'non-canonical Wnt induces Dvl to transition from Vangl to Fz, while PK inhibits this transition, and they function synergistically with Vangl to suppress Dvl during Convergent Extension (CE),' it is crucial to assess the subcellular localization of PCP core proteins in dorsal marginal zone (DMZ) cells, which are known to undergo CE. Notably, the overexpression of Wnt11 alone, as employed by the author, does not induce animal cap elongation. Therefore, the use of animal cap explants may not be sufficient to substantiate the model during Convergent Extension (CE). Indeed, previous knowledge indicates that Vangl2 and Pk localize to the anterior region in DMZ explants. However, the results presented in this manuscript appear to differ from this established understanding. Consequently, to provide more robust support for the proposed model, it is advisable to replicate the key experiments (Figures 3, 4, 8, and Figure S3) using DMZ explants.
We repeated the experiments in Figure 3, 4, 8 and Figure S3 with DMZ explant and the new data are in new Supplementary Fig. 9, 11, 16 and 4, respectively.In regards to “previous knowledge indicates that Vangl2 and Pk localize to the anterior region in DMZ explants”, we are aware Vangl/ Pk localization to the anterior cell cortex in neural epithelium from the studies by the Sokol and Wallingford labs, but are not aware of similar reports in DMZ explants. When we examined the localization of small amount of injected EGFP-mPk2 (0.1 ng mRNA) in DMZ explants, we saw a somewhat uniform distribution on the plasma membrane (Suppl. Fig. 4). In addition, in a related recent publication, we examined endogenous XVangl2 protein localization in activin induced animal cap explants that do undergo CE. What we observed was that whereas low level injected Dvl2 and Fz form clusters on the plasma member, endogenous XVangl2 remains uniformly distributed on the plasma membrane (Suppl. Fig. 3S-Z in 10.1038/s41467-025-57658-0 ). These observations may suggest potential differences of PCP protein localization during neural vs. mesodermal convergence and extension.
(2) The author suggests that 'Vangl2 and Pk together synergistically disrupt Fz7-Dvl2 patches.' As shown in Figure 4 (panels J' to I'), it is evident that the co-expression of Pk and Vangl2 increases Fz7 endocytosis. Nevertheless, a significant amount of Fz7 still co-localizes with Dvl2. To strengthen the author's hypothesis, additional clear assay is required such as Fluorescence resonance energy transfer (FRET) assay.
We appreciate this valuable advice. Since none of the tagged Fz/ Dvl/ Vangl proteins we had were suitable for FRET, we made proteins tagged with mClover and mRuby2, which were reported as optimized FRET pairs. But in our hands mRuby2 seems to require very long time (~2 days) to mature and become detectable at room temperature, and is not suitable for our Xenopus experiments. We are in the process of establishing a luciferase based NanoBiT system to detect Fz-Dvl and Dvl-Vangl interactions in live cells and cell lysates, and will use it in future studies to investigate their interaction dynamics.
For the current manuscript, we reason that a substantial reduction of Fz7-Dvl2 clusters with Vangl2/ Pk co-injection would still support our idea that Vangl2 and Pk act synergistically to sequester Dvl from Fz to prevent their clustering in response to non-canonical Wnt ligands.
(3) The IP data is less clear and evident. A couple of examples are: a) Fig 2g where the authors report that the Vangl2 R177H variant reduced Vangl2 interaction with Pk and recruitment of Pk to the plasma membrane, but it appears that the variant interacts slightly better than WT Vangl2 with Pk. In Fig. S7a, the authors state that Pk overexpression can indeed significantly reduce Wnt11-induced dissociation of EGFP-Vangl2 and Flag-Dvl2 in the DMZ. However, there is a minimal impact when compared to the Wnt11 absent control. Based on the results presented in Fig S12a the authors indicate that Wnt11 reduces the association between Vangl2 and Dvl2, which can be discerned, but loss of Ror2 does not change this in any obvious way - but the authors indicate it does. In S12b, the authors have suggested that Ror and Dvl do not form a direct binding interaction. However, the interpretation of Figure S12b is not entirely convincing due to several issues. Notably, the expression levels of each protein appear inconsistent, the bands are not sufficiently clear, and there is the detection of three different tag proteins on a single blot. To strengthen the validity of these findings, it is advisable to repeat this experiment with improved quality.
We repeated all the co-IP and western blot analyses pointed out by the reviewer, and performed quantification and statistical analyses.
Fig 2g had a mistake in the labeling and is replaced with new Figure 2g;
Fig. S7a is replaced by new data in Supplementary Figure 8a and b;
Fig. S12a and 12b are replaced by new data in Supplementary Figure 15a, a’ and b, respectively. In 15a and a’, we noticed a consistent decrease of Dvl2-Vangl2 co-IP in Xror2 morphant. The reason for this is not yet clear and will need further study in the future.
Minor points: (1) In all the whole embryo injection assays examining morphology, no Western analysis is performed to show roughly equivalent and appropriate levels of the various proteins are being expressed. Differences will affect the data.
Although we did not do western analyses to examine the protein levels in various functional interaction assays, we did examine how co-expression of Vangl2, mPk2 or Dvl2 may impact each other’s protein levels in Supplementary Fig. 2, which did not reveal any significant change when co-injected in different combination.
(2) The author's prior publication (Bimodal regulation of Dishevelled function by Vangl2 during morphogenesis, Hum Mol Genet. 2017) presented clear evidence of Vangl2 overexpression inducing Dvl2 membrane localization. However, Figure S4 in the current manuscript did not provide clear evidence of membrane localization. To strengthen the hypothesis that Vangl2-RH mutant also induces Dvl2 membrane localization, further comprehensive imaging analysis is needed.
We re-analyzed the imaging data and replaced old Figure S4 with a new Supplementary Fig. 5.
(3) In Supplementary Figure 9, the authors propose that the overexpression of Vangl2/Pk induces Fz7 endocytosis, as indicated by its co-localization with FM4-64. However, it raises a question: how does the Fz7-GFP protein internalize into the cells without endocytosis, as seen in Figures S9a-c'? To enhance readers' understanding, a discussion addressing this point should be included.
We think that this might be a technical issue. As detailed in the Method section, we only incubated the embryos transiently with FM4-64 for 30 minutes, and the embryos were subsequently washed and dissected in 0.1X MMR without the dye. Therefore, only the Fz7-GFP protein endocytosed during the 30 minute-incubation would be labeled by FM-64, whereas that endocytosed before or after the incubation would not. Alternatively, the very few Fz7-GFP puncta occasionally observed in the absence of Vangl2/Pk overexpression could be vesicles trafficking to the plasma membrane.
(4) Statistical analyses are absent for several results, including those in Figure 2f, Figure S4d, and Figure S7b.
We repeated these experiments and included statistical analyses. The new data are in Figure 2f, Supplementary Fig. 5d and Supplementary Fig. 8b.
(5) This manuscript lacks any results regarding Ck1. Therefore, it is advisable to consider removing the discussion or mention of CK1.
We agree, and tune down the discussion on CK1 and removed CK1 from our model in Fig. 9.
Reviewer #2 (Recommendations For The Authors):
(1) In all the convergence and extension assays, the authors should report n numbers (i.e. number of animals), what statistical test is used, and what the error bars show. Ideally dot-plots would be used instead of bar charts as they give a better insight into the data distribution. It might be useful to give a section on the statistical analyses used in the M&M, including e.g. any power calculations carried out, as now required by many journals.
We have follow the advice to use dot-plots for all the quantification analyses in the manuscript. We include in the figure legends the statistical test used and what the error bars show. The number of embryos analyzed were included in each panel in the figures. We also provided more details in the Methods section on how the LWR quantification was carried out.
(2) I think Figure 2g is wrongly labelled? FLAG bands are in all three lanes in the western blot, but not labelled as such in the schematic.
We corrected the schematic labeling in Figure 2g, and thank the reviewer for catching this mistake.
(3) In Figure S7, the authors show that co-IP of Dvl and Vangl2 is reduced by Wnt11 and the effects of Wnt are blocked by Pk. Does Pk have any effect in the absence of Wnt?
We examined the effect of Pk over-expression on Dvl2-Vangl2 co-IP as advised, and did not see a significant impact in the absence of Wnt11 co-injection. The data is included in the new Supplementary Figure 8a. We interpret the data to suggest that “at least under the condition of our co-IP experiment, Pk may not directly impact the steady-state binding between Vangl and Dvl”.
(4) In Figure 3, the authors show (as published previously) that Wnt11 induces patches of Dvl at the plasma membrane. It would be useful to see Dvl in the absence of Wnt and Vangl2/Dvl in the absence of Wnt.
Dvl is widely known as a cytoplasmic protein and its localization has been published by many labs over the past 20-30 years. In our recent publication (10.1038/s41467-025-57658-0 ), we also re-examined Dvl localization when injected at various dosages. So we did not feel it was necessary to show its localization in the absence of Wnt11 again, but included a reference to our prior publication. In regards to Vangl/Dvl distribution in the absence of Wnt11, the readers can see Suppl. Fig. 5b as an example, in addition to our previous publications referenced in the manuscript.
(5) In the review figures, the difference in Fz7-GFP patch formation in d' and e' (vs e.g. a') is not very clear. Could the images be improved or (better) quantified in some way?
We assume that “review figures” refer to Figure 3 or 4? If so, we felt that Fz7-GFP patch formation was clear in Fig. 3d’, e’ or Fig. 4d’, e’. Nevertheless, we repeated these experiments in DMZ explants as advised by Reviewer 1, and additional examples of Fz7-EGFP patch formation can be seen in the new Suppl. Fig. 9d-f’ and Suppl. Fig. 11d-f’.
(6) In Figure 6d, I'm concerned that the loss of flag-Dvl2 might occur via dephosphorylation in the IP reaction. Also the M&M don't include methodological details about buffers and whether phosphatase inhibitors were used. A compelling control would be anti-FLAG pulldown showing retention of phosphorylation. Also Figure 6f shows a reduced ratio of fast-to-slow migrating bands of Dvl with Vangl2/Pk - unless I have misunderstood, is this ratio the wrong way round?
We added co-IP buffer and protease inhibitor information in Methods.
We agree that the concern about dephosphorylation during IP reaction is valid, and that direct pull down of Dvl to show the phosphorylated form is a compelling control. We therefore note that in Suppl. Fig. 8a and 15b, direct pull down of Flag-Dvl or Myc-Dvl (with anti-Flag or anti-Myc) did show the slower migrating, phosphorylated form. Additional examples in which Vangl only co-IP the faster migrating unphosphorylated Dvl include Suppl. Fig. 15a, and in a related paper we published recently (Fig. 3R and R’ in 10.1038/s41467-025-57658-0 ).
Finally, we did wrongly label Figure 6f in the last submission, and the ratio should have been “slow/fast”. We have made the correction, and appreaicte the reviewer for the meticulousness in perusing our manuscript.
(7) In Figure 7, what does Ror2 look like in the absence of Wnt11?
We included new Figure 7a-c to show that without Wnt11 co-injection, Ror2 is uniformly distributed on the plasma membrane.
(8) Also in Figure 7, Ror2 patches are said to be slightly wider than Dvl2 patches "reminiscent of Vangl2" - I wouldn't describe them as being similar. Vangl2 shows a distinct dip in the center of the Dvl patches, Ror2 does not show a dip, and is only (at best) in a slightly wider patch, and I would want to see further examples to be convinced that the localization domain is reproducibly wider. The merge of many samples in 7d may actually be making the distribution harder to see and if the Xror2 and Dvl2 intensities were normalized I'm not sure how different the curves would appear. (i.e. the Xror2 curve looks like a flattened version of the Dvl2 curve).
We have added an additional panel in the new Figure 7j to compare the intensity ratio of Ror/ Dvl2 along the patches, and this analysis reveals an over two folds increase of the ratio at the border region. This quantification may make a more convincing argument that at the patch border region, Dvl is diminished whereas Ror2 accumulate with Vangl2.
(9) In Figure S12a, the authors suggest Wnt11 induced dissociation of Dvl from Vangl2 (by co-IP), and this is reduced after Ror2 MO. This would be more convincing with replicates and quantitation.
We have repeated this experiment with Vangl2 pull down and added quantification. The data is in the new Suppl. Fig. 15a.
(10) In Figure S12b, the authors suggest Ror2 can co-IP Vangl2 but not Dvl. This is not very convincing, as the Dvl input band is very weak, and the Vangl2 co-IP band is very weak.
We repeated the co-IP experiment with Myc-tagged Vangl or Dvl. Using the same anti-Myc antibody and experimental condition (including the expression level of Vangl, Dvl and Ror2), we still found that Ror2 could be pulled down by Vangl but not Dvl (Suppl. Fig. 15b).
(11) "Prickle" spelled "Prickel" in the abstract (and abbreviated to "PK" not "Pk" at one place in the abstract and several places in text)
We have corrected these typos.
(12) Quite a lot of interesting observations are in supplemental figures. Normally it might be expected that extra data supporting a conclusion would be in supplemental, but here some of the supplemental data feels like it is more than simply additional evidence. For instance supplemental Figures 2 and 3 feel more than just supplemental (and Supplemental Figure 3 if merged with Figure 2 would make it easier for the reader). Moreover, for example, the description of the results in Figure 2 is punctuated by references to supplemental Figures 4 and 5 that contain key data to support the conclusions, which means the reader has to flick backwards and forwards from place to place in the manuscript to follow the argument. It is of course up to the authors, but in some cases putting supplemental data back into the main figures (for which there is no size or number limit) would increase clarity.
These are excellent points; in the resubmitted manuscript we have a total of 24 data figures, and we used 8 as main figures since we felt that they provide the most relevant and conclusive evidence to our model. We will consult the copy editors at eLife on how to arrange the rest as main vs. supporting figures when requesting publication as version of record.
