Paracingulin recruits CAMSAP3 to tight junctions and regulates microtubule and polarized epithelial cell organization

ABSTRACT Paracingulin (CGNL1) is recruited to tight junctions (TJs) by ZO-1 and to adherens junctions (AJs) by PLEKHA7. PLEKHA7 has been reported to bind to the microtubule minus-end-binding protein CAMSAP3, to tether microtubules to the AJs. Here, we show that knockout (KO) of CGNL1, but not of PLEKHA7, results in the loss of junctional CAMSAP3 and its redistribution into a cytoplasmic pool both in cultured epithelial cells in vitro and mouse intestinal epithelium in vivo. In agreement, GST pulldown analyses show that CGNL1, but not PLEKHA7, interacts strongly with CAMSAP3, and the interaction is mediated by their respective coiled-coil regions. Ultrastructure expansion microscopy shows that CAMSAP3-capped microtubules are tethered to junctions by the ZO-1-associated pool of CGNL1. The KO of CGNL1 results in disorganized cytoplasmic microtubules and irregular nuclei alignment in mouse intestinal epithelial cells, altered cyst morphogenesis in cultured kidney epithelial cells, and disrupted planar apical microtubules in mammary epithelial cells. Together, these results uncover new functions of CGNL1 in recruiting CAMSAP3 to junctions and regulating microtubule cytoskeleton organization and epithelial cell architecture.

Overall, it is a nice manuscript that focuses on understanding the role of CGNL1 and CAMSAP3 in orchestrating microtubule organization in polarized epithelial cells. Some data is somewhat conformational since some of the findings (CGNL1 interaction with CAMSAP3) have been reported before. Still, it is important study since it systematically analyses the differential roles of CGNL1 and PLEKHA7 in recruiting CAMSAP3 and organizing microtubules. Most data is very strong and use of couple different cell lines, as well as in vivo model, makes it quite convincing. The manuscript does have few issues (see listed below), mostly associated with quantifications and statistical analyses. All of them should be easy to address. 1) Figure 1-2. Overall, very nice and very convincing data. However, it is unclear to me why CAMSAP3 is completely cytosolic in ZO1-KO cells? In those cells CGLN1 is still recruited to AJs via binding to PLEKHA7. Thus, I would expect CAMSAP3 (if it is recruited via binding to CGLN1) to go to AJ.
2) Figure 3A-C. In these images CAMSAP3 and ZO1 has very limited co-localization (in both wt and CGNL1-KO cells). Quantification should be provided about this co-localization.
3) Figure 4A-B. Defects are not super convincing. Higher magnification images would be useful (they can clearly do it using in vivo tissues as shown in Figure 5). Data also needs quantification. Figure 6F-K. Quantification is needed before one can interpret the data. 5) Figure 7F. GFP-CAMSAP3-5A mutants do not seem to localize with MTs anymore. That rises the concern that this mutation affected overall CAMSAP3 structure, making it hard to interpret. At the very least, authors need to show that this CAMSAP3 mutant can bind some other known CAMSAP3 interactors.

4)
6) It is important that in the quantifications (where individual cells are plotted as separate data points, for example see Figure 2C or F) all data is color coded so reviewer can determine how different is the data from different experimental repeats. That is especially important since it appears (based on Statistical analysis section in Methods) that some experiments were only repeated twice.
Comments for the author see above

Reviewer 2
Advance summary and potential significance to field This study addresses the mechanisms of microtubule minus-end organization at the cell-cell junctions in epithelial cells. Previous well-cited work has proposed a mechanism whereby a scaffolding protein PLEKHA7 recruited microtubule minus ends decorated by CAMSAP3 to adherens junctions to organise apicobasal microtubule arrays in a cell culture model of intestinal cells, Caco2. Later work showed that the main mechanism actually responsible for the generation of apico-basal microtubule polarity in intestinal epithelium depends on CAMSAP3 binding to the apical cortex through spectraplakin MACF1/ACF7. In the current study, the authors re-examine microtubule minus-end organization at cell-cell junctions in several epithelial cell models and in knockout mice. The authors show that a protein bound to tight junctions, Paracingulin (CGNL1), is responsible for junctional recruitment of CAMSAP3 and microtubule organization. The great strength of this study is the use of multiple cell and mouse knockout models, as well as biochemical assays characterizing CAMSAP3-CGNL1 interaction in detail. Most of the experiments are very well controlled, the overall quality of the images is good and appropriate quantitative analysis is included. Based on these strengths, the study represents an excellent addition to the field, and is certainly suitable for publication in J Cell Sci.

1.
The major functional readout used throughout the paper is the localization of CAMSAP3, and it would be important to establish the specificity of the antibodies used, by using either a knockout or depletion of CAMSAP3. There are several types of localization detected -at the junctions, in the cytoplasm and in the nucleus (e.g., Fig. 2B), and it would be important to determine how specific they are. Even more importantly, it would be good to know whether CAMSAP3 is continuously distributed along CGNL1-positive junctions or whether it is specifically accumulated at the minus-ends of microtubules colocalizing with the junctions. The images in Figure 3 imply the latter (but their quality is a bit limited, as there are numerous CAMSAP3 dots which are not colocalizing either with the junctions or microtubules particularly in WT cells), whereas all the other images in the paper (including rescue experiments in Figure 7) support the former idea. Two not-too-complicated ways of addressing this question further are through microtubule disassembly by nocodazole, to check if CAMSAP3 would still localize to junctions when microtubules are absent, and by more detailed imaging of GFP-CAMSAP3 in the rescue experiments.

2.
It is interesting that in cells lacking CGNL1, microtubules at the junctions become more abundant. This point is hardly discussed although it represents an unexpected outcome. It would be important that the authors provide a clearer view of the overall organization of microtubules in WT and CGNL1-knockout cells. A scheme illustrating such organization at the end of the paper would seem appropriate. Do the authors think that binding of CAMSAP3-decorated microtubule minus ends to the junctions promotes their organization perpendicular to the junctions? Some additional discussion of this point would be welcome.

3.
A great strength of this study is the generation of two new mouse knockout models. Understandably, the authors would like to save detailed phenotypic characterization of these knockout mice for another study. Still, some short description of the viability and major phenotypes of the mice related to the analyzed tissues would seem appropriate.

Reviewer 3
Advance summary and potential significance to field The manuscript by Flinois et al., explores the role of junctional proteins in epithelial microtubule organisation and cell morphology in-vitro and in-vivo. They define the molecular interactions between CGNL1 and the microtubule minus end protein CAMSAP3 at tight junctions, independent of the AJ protein PLEKHA7. The strength of the study is the classic combination of in-vivo and in-vitro models, with quantitative imaging approaches, and super resolution microscopy complemented by biochemical analyses. It describes a previously unappreciated relationship between the TJ protein CGNL1 and the regulation of the microtubule cytoskeleton organisation in epithelial cells. However, it could be strengthened by including stronger critique of the effects on cell shape/architecture in the absence of CGNL1.

Comments for the author
I have very few experimental recommendations as the data was clear and quantitated sufficiently to support the statements. However, in places, the author?s statements are not supported by the data and I would recommend rewording.
Microtubule organisation in response to CGNL1-KO In figure 3 where authors are imaging the apical microtubule network in Eph4 cells. They state "peri-junctional tubulin labelling was more intense in CGNL1-KO cells suggesting a collapse of MTs into peri-junctional bundles". The data as it stands doesn't support this statement. I'd recommend softening it to "suggesting an alteration in the apical organisation of the microtubule network".
Also "controls the organization of MTs" soften to "contributes" Loss of cellular and tissue architecture The authors state that there is a loss of tissue architecture, but don't ever clarify what that is. Yes, proteins change localisation-but what does this mean for the cell or tissue shape or function?
For example: In the CCGNL1 KO cells, if the apical microtubule organisation is lost ("collapsed") and the microtubules are "spatially separated" from TJ ( figure 3A), what was the consequence on the cellular architecture? Was there a loss of cell height? Change in cell size? nuclear orientation/shape?
In Figure 4-yes there is a loss of nuclear apex distance-but what does that mean at the cellular/tissue scale? Toya et al., PNAS figure 3 use Caco2 cysts and show a change in architecture (lumen number/size, acini size) in CAMSAP3 siRNA acini. Do the cell lines featured in this manuscript grow as polarised 3D cysts? If so-could a 3D acini model be used to investigate changes in architecture? I might have misunderstood 3H. ZO1 intensity should be quantitated at the junction for both WT and CGNL1-KO. A simple line scan of existing images. Particularly as previous reports show ZO3 increases in CGNL1-depleted cells https://doi.org/10.1091/mbc.e08-06-0558 The manuscript clarity. The manuscript was dense in sections, and difficult to understand as several parallel concepts were in one sentence. There was *way* too many acronyms, I would recommend at least spelling out microtubule.
Colour-blind friendly images I would recommend changing red to magenta Figure 3A-B red arrows in magnified inset -make these yellow or another colour-not red/the same as the IF pseudo colour. It was difficult to see.

Typographical errors
Line 67 zonula adhaerens, should read zonula adheren Latin-either italics or hyphen de novo should read de-novo or de novo "in mice" should read "in tissue"

First revision
Author response to reviewers' comments

Response to reviewers
Reviewer 1 Advance Summary and Potential Significance to Field: Overall, it is a nice manuscript that focuses on understanding the role of CGNL1 and CAMSAP3 in orchestrating microtubule organization in polarized epithelial cells. Some data is somewhat conformational, since some of the findings (CGNL1 interaction with CAMSAP3) have been reported before. Still, it is important study since it systematically analyses the differential roles of CGNL1 and PLEKHA7 in recruiting CAMSAP3 and organizing microtubules. Most data is very strong and use of couple different cell lines, as well as in vivo model, makes it quite convincing. The manuscript does have few issues (see listed below), mostly associated with quantifications and statistical analyses. All of them should be easy to address. 1) Figure 1-2. Overall, very nice and very convincing data. However, it is unclear to me why CAMSAP3 is completely cytosolic in ZO1-KO cells? In those cells CGLN1 is still recruited to AJs via binding to PLEKHA7. Thus, I would expect CAMSAP3 (if it is recruited via binding to CGLN1) to go to AJ.

Response.
In ZO-1-KO Eph4 cells there is a faint residual junctional staining for CAMSAP3, that could be attributed to the PLEKHA7-associated CGNL1. However, quantifications indicate that the contribution of such AJ-associated pool of CGNL1 to CAMSAP3 recruitment is minor. This could be due to multivalent interactions or post-translational modifications of CGNL1 at AJ. The text of results and discussion was revised to illustrate this point more clearly. On the other hand, in mCCD cells the KO of PLEKHA7 has no effect on junctional CGNL1/CAMSAP3 labeling, suggesting that all CGNL1 is associated with TJ, and only the TJ (ZO-1)-associated pool of CGNL1 recruits CAMSAP3.
2) Figure 3A-C. In these images CAMSAP3 and ZO1 has very limited co-localization (in both wt and CGNL1-KO cells). Quantification should be provided about this co-localization.

Response. CAMSAP3 and ZO-1 actually do not co-localize, not only because they do not interact together, but especially because ZO-1 is localized very close to the TJ membrane (approximately 25 nm), whereas the rod of CGNL1 (where CAMSAP3 binds) is located at a much higher distance from the TJ membrane (approximately 200 nm). ZO-1 is a large molecule, the antibody to ZO-1 binds to the alpha domain (close to the membrane), CGNL1 binds to the extreme C-terminus of ZO-1 (ZU5 domain) and CGNL1 has an asymmetric shape. Thus, ZO-1 is being usedhere exclusively as a spatial marker of the peri-junctional region when comparing the localization of CAMSAP3/microtubules in WT and CGNL1-KO cells. Because ZO-1 and CAMSAP3 do not co-localize precisely, no quantification can be made. The only quantification we can do is the number of CAMSAP3-decorated MT in the peri-junctional region, e.g. within approximately 200 nm (=1 micron in expanded images) of the ZO-1 signal.
3) Figure 4A-B. Defects are not super convincing. Higher magnification images would be useful (they can clearly do it using in vivo tissues as shown in Figure 5). Data also needs quantification. Fig. 4A Figure 7F. GFP-CAMSAP3-5A mutants do not seem to localize with MTs anymore. That rises the concern that this mutation affected overall CAMSAP3 structure, making it hard to interpret. At the very least, authors need to show that this CAMSAP3 mutant can bind some other known CAMSAP3 interactors.

5)
Response. Toya et al. (PNAS 2016) already showed that this mutant still interacts with MT, thus its structure is not affected by the mutation in a way to impact on its interaction with MTs. The text of the Discussion was revised to make this clear.
6) It is important that in the quantifications (where individual cells are plotted as separate data points, for example see Figure 2C or F) all data is color coded so reviewer can determine how different is the data from different experimental repeats. That is especially important since it appears (based on Statistical analysis section in Methods) that some experiments were only repeated twice.

Response. We thank this Reviewer for this suggestion. A color code was used for different experimental repeats when showing individual data points.
Reviewer 1 Comments for the Author: see above

Reviewer 2
Advance Summary and Potential Significance to Field: This study addresses the mechanisms of microtubule minus-end organization at the cell-cell junctions in epithelial cells. Previous well-cited work has proposed a mechanism whereby a scaffolding protein PLEKHA7 recruited microtubule minus ends decorated by CAMSAP3 to adherens junctions to organise apicobasal microtubule arrays in a cell culture model of intestinal cells, Caco2. Later work showed that the main mechanism actually responsible for the generation of apico-basal microtubule polarity in intestinal epithelium depends on CAMSAP3 binding to the apical cortex through spectraplakin MACF1/ACF7. In the current study, the authors re-examine microtubule minus-end organization at cell-cell junctions in several epithelial cell models and in knockout mice. The authors show that a protein bound to tight junctions, Paracingulin (CGNL1), is responsible for junctional recruitment of CAMSAP3 and microtubule organization. The great strength of this study is the use of multiple cell and mouse knockout models, as well as biochemical assays characterizing CAMSAP3-CGNL1 interaction in detail. Most of the experiments are very well controlled, the overall quality of the images is good and appropriate quantitative analysis is included. Based on these strengths, the study represents an excellent addition to the field, and is certainly suitable for publication in J Cell Sci.
Reviewer 2 Comments for the Author: There are several points that would require clarifications and additional controls, to ensure that the conclusions are fully supported by data.
1. The major functional readout used throughout the paper is the localization of CAMSAP3, and it would be important to establish the specificity of the antibodies used, by using either a knockout or depletion of CAMSAP3. There are several types of localization detected -at the junctions, in the cytoplasm and in the nucleus (e.g., Fig. 2B), and it would be important to determine how specific they are.  Fig. 2E).

Response
Even more importantly, it would be good to know whether CAMSAP3 is continuously distributed along CGNL1-positive junctions or whether it is specifically accumulated at the minus-ends of microtubules colocalizing with the junctions. The images in Figure 3 imply the latter (but their quality is a bit limited, as there are numerous CAMSAP3 dots which are not colocalizing either with the junctions or microtubules, particularly in WT cells), whereas all the other images in the paper (including rescue experiments in Figure 7) support the former idea. Two not-too-complicated ways of addressing this question further are through microtubule disassembly by nocodazole, to check if CAMSAP3 would still localize to junctions when microtubules are absent, and by more detailed imaging of GFP-CAMSAP3 in the rescue experiments. Fig-S7) show that CAMSAP3 still localizes at junctions, whereas the cytoplasmic/apical pool becomes undetectable upon MT depolymerization. This allows us to conclude that CAMSAP3 can bind to CGNL1 independently of MT cytoskeleton integrity, and that the 5A mutation affects the binding to CGNL1 and not to MT. The text was revised accordingly.

Response. We thank this Reviewer for this insightful comment. We carried out the analysis of CAMSAP3 localization upon treatment of cells with nocodazole, and the results (new
2. It is interesting that in cells lacking CGNL1, microtubules at the junctions become more abundant. This point is hardly discussed, although it represents an unexpected outcome. It would be important that the authors provide a clearer view of the overall organization of microtubules in WT and CGNL1-knockout cells. A scheme illustrating such organization at the end of the paper would seem appropriate. Do the authors think that binding of CAMSAP3-decorated microtubule minus ends to the junctions promotes their organization perpendicular to the junctions? Some additional discussion of this point would be welcome. Fig. 8K), with a summary at the beginning of the Discussion.

3.
A great strength of this study is the generation of two new mouse knockout models. Understandably, the authors would like to save detailed phenotypic characterization of these knockout mice for another study. Still, some short description of the viability and major phenotypes of the mice related to the analyzed tissues would seem appropriate.
Response. Additional information about the KO mice was included in the methods section. As the Reviewer correctly understands, mouse tissues are used here to validate the phenotypes obtained in cultured cells, and additional phenotypic characterization of the phenotypes of these mice belongs to future studies.

Reviewer 3
Advance Summary and Potential Significance to Field: The manuscript by Flinois et al., explores the role of junctional proteins in epithelial microtubule organisation and cell morphology in-vitro and in-vivo. They define the molecular interactions between CGNL1 and the microtubule minus end protein CAMSAP3 at tight junctions, independent of the AJ protein PLEKHA7. The strength of the study is the classic combination of in-vivo and in-vitro models, with quantitative imaging approaches, and super resolution microscopy, complemented by biochemical analyses. It describes a previously unappreciated relationship between the TJ protein CGNL1 and the regulation of the microtubule cytoskeleton organisation in epithelial cells. However, it could be strengthened by including stronger critique of the effects on cell shape/architecture in the absence of CGNL1.
Reviewer 3 Comments for the Author: I have very few experimental recommendations as the data was clear and quantitated sufficiently to support the statements. However, in places, the author's statements are not supported by the data and I would recommend rewording.
Microtubule organisation in response to CGNL1-KO In figure 3 where authors are imaging the apical microtubule network in Eph4 cells. They state "peri-junctional tubulin labelling was more intense in CGNL1-KO cells suggesting a collapse of MTs into peri-junctional bundles". The data as it stands doesn't support this statement. I'd recommend softening it to "suggesting an alteration in the apical organisation of the microtubule network".
Also "controls the organization of MTs" soften to "contributes" Loss of cellular and tissue architecture The authors state that there is a loss of tissue architecture, but don't ever clarify what that is. Yes, proteins change localisation-but what does this mean for the cell or tissue shape or function?

Response. We observe a consistent change in the apico-basal positioning of the nuclei in intestinal epithelial cells of CGNL1-KO mice. In this sense, it is an alteration of cell architecture. The text
was revised to make this clear. We have not examined the overall shape or function of any tissue yet, this will belong to a future phenotypic characterization of the mice.
For example: In the CCGNL1 KO cells, if the apical microtubule organisation is lost ("collapsed") and the microtubules are "spatially separated" from TJ ( figure 3A), what was the consequence on the cellular architecture? Was there a loss of cell height? Change in cell size? nuclear orientation/shape?
Response. It is difficult to measure cell height in cysts, since not all cysts have the same size. Effects on cellular architecture were also hard to assess on cultured cells, as the polarization is much less pronounced than in columnar epithelial tissues. Cell size differences were not detected, either, and the nucleus occupies the entirety of the cell, making it impossible to find any difference between WT and KO cells. Intestinal epithelial (columnar) cells are ideally suited for these measures, this is why we generated and analyzed KO mice.
In Figure 4-yes there is a loss of nuclear apex distance-but what does that mean at the cellular/tissue scale?
Response. At the cellular scale, this means that the overall cell architecture is perturbed by the disorganization of cytoplasmic microtubules, as described in Toya et al (PNAS 2016) for CAMSAP3 mutant mice. The implication of a similar cellular phenotype (for CAMSAP3 mutant mice) at the tissue level were described in Mitsuhata et al (Sci. Rep. 2021), for example cysts in kidney proximal convoluted tubules. As stated above, a detailed phenotypic analysis of CGNL1-KO epithelial tissues/organs has not been carried out yet and will be the object of future studies. We did not notice altered histology of the kidney of CGNL1-KO micey, but we will expand our analyses in future work. The present study is focused on cells and subcellular structures, not on tissues and organs.
Toya et al., PNAS figure 3 use Caco2 cysts and show a change in architecture (lumen number/size, acini size) in CAMSAP3 siRNA acini. Do the cell lines featured in this manuscript grow as polarised 3D cysts? If so-could a 3D acini model be used to investigate changes in architecture? Figure 1E-I and Fig. 8D-J). The KO of CGNL1 has an impact on cyst morphogenesis, and this phenotype requires the Rod-1 region of CGNL1 for a rescue.

Response. We thank this Reviewer for this constructive comment. We carried out analysis of WT and CGNL1-KO cysts grown in Matrigel (new panels in
I might have misunderstood 3H. ZO1 intensity should be quantitated at the junction for both WT and CGNL1-KO. A simple line scan of existing images. Particularly as previous reports show ZO3 increases in CGNL1-depleted cells https://doi.org/10.1091/mbc.e08-06-0558

Response. In this Figure, ZO-1 labeling is exclusively used to identify the outermost cell area. The intensity of ZO-1 and ZO-3 labeling is not affected by KO of CGNL1 in different epithelial cell lines, as shown in our recent paper Vasileva et al. (JBC 2022). Instead, the KO of CGN does reduce both. The publication this Reviewer refers to studies the effect of acute depletion of CGNL1 by shRNA (not KO) and we observed a small increase in ZO-3 mRNA and protein levels, but no change in the intensity of immunofluorescent ZO-3 labeling at junctions.
The manuscript clarity. The manuscript was dense in sections, and difficult to understand as several parallel concepts were in one sentence. There was *way* too many acronyms, I would recommend at least spelling out microtubule.
Response. We used microtubule throughout, instead of MT, and reduced acronyms as much as possible.
Colour-blind friendly images I would recommend changing red to magenta Response. Magenta color is very similar to DAPI blue, and we feel the clarity of the images would be affected by changing the color. Figure 3A-B red arrows in magnified inset-make these yellow or another colour-not red/the same as the IF pseudo colour. It was difficult to see.

Response.
Green and red arrows were modified to have an outer white line in order to make them more visible.

Typographical errors
Line 67 zonula adhaerens, should read zonula adherens Latin-either italics or hyphen de novo should read de-novo or de novo "in mice" should read "in tissue" Response. We have now reached a decision on the above manuscript.
To see the reviewers' reports and a copy of this decision letter, please go to: https://submitjcs.biologists.organd click on the 'Manuscripts with Decisions' queue in the Author Area.
(Corresponding author only has access to reviews.) As you will see, all reviewers are extremely favourable regarding the publication of your paper. However, Reviewer 2 maintains a critical point of validation of the anti-CAMSAP3 antibodies, which was raised in their original review. I hope that you will be able to carry this out because I would very much like to be able to accept your paper, depending on this validation.
Please ensure that you clearly highlight all changes made in the revised manuscript. Please avoid using 'Tracked changes' in Word files as these are lost in PDF conversion.
I should be grateful if you would also provide a point-by-point response detailing how you have dealt with the points raised by the reviewers in the 'Response to Reviewers' box. Please attend to all of the reviewers' comments. If you do not agree with any of their criticisms or suggestions please explain clearly why this is so.

Reviewer 1
Advance summary and potential significance to field Authors have addressed all my concerns and incorporated all suggestions. As the result, in my opinion, this manuscript is now ready for publication.

Comments for the author
Authors have addressed all my concerns and incorporated all suggestions. As the result, in my opinion, this manuscript is now ready for publication.

Reviewer 2
Advance summary and potential significance to field As indicated in my previous review, overall, I find this a strong paper potentially suitable for publication in the Journal of Cell Science.

Comments for the author
Unfortunately, the authors chose to ignore my request to characterize the specificity of the anti-CAMSAP3 antibodies, on which the majority of the conclusions of this paper rests. This is in my view inappropriate because the characterization of these antibodies is described in the Meng et al Cell 2008 paper -the very paper the conclusions of which the authors of the current manuscript bring into question. The specificity of this anti-CAMSAP3 antibody is questionable because nuclear staining is observed (e.g. Fig. 2B), whereas GFP-CAMSAP3 is never found in the cell nucleus. This is an essential point that needs proper clarification before the paper is published. Performing stainings with this antibody in CAMSAP3-depleted or CAMSAP3-knockout epithelial cells that the authors study would address this comment.

Reviewer 3
Advance summary and potential significance to field The manuscript by Flinois et al., explores the role of junctional proteins in epithelial microtubule organisation and cell morphology in-vitro and in-vivo. They define the molecular interactions between CGNL1 and the microtubule minus end protein CAMSAP3 at tight junctions, independent of the AJ protein PLEKHA7. The study brings new knowledge regarding the relationship between the microtubule cytoskeleton, cell adhesion and cell morphology/architecture.

Comments for the author
I am happy with how the authors have addressed my concerns.

Second revision
Author response to reviewers' comments

Response to Reviewer2
Reviewer 2 Comments for the Author: Unfortunately, the authors chose to ignore my request to characterize the specificity of the anti-CAMSAP3 antibodies, on which the majority of the conclusions of this paper rests. This is in my view inappropriate because the characterization of these antibodies is described in the Meng et al Cell 2008 paper -the very paper the conclusions of which the authors of the current manuscript bring into question. The specificity of this anti-CAMSAP3 antibody is questionable because nuclear staining is observed (e.g. Fig. 2B), whereas GFP-CAMSAP3 is never found in the cell nucleus. This is an essential point that needs proper clarification before the paper is published. Performing stainings with this antibody in CAMSAP3-depleted or CAMSAP3-knockout epithelial cells that the authors study would address this comment.

Response.
We did not ignore the request to characterize the specificity of the anti-CAMSAP3 antibodies. We responded that the antibodies had already been validated. Indeed, the Takeichi lab polyclonal antibodies described in Meng et al 2008Meng et al (doi:10.1016Meng et al / j.cell.2008

NOTE:
We have removed unpublished data that had been provided for the referees..
-The Takeichi lab polyclonal antibodies were used in essentially all the experiments reported in our paper, on which the majority of the conclusions of our paper rest, as the Reviewer correctly states. The exception is Fig. S5C, where we used a home-made guinea-pig polyclonal antibody to label kidney sections in order to do double labeling with anti-polyE rabbit antibodies (NB the labeling was identical using the rabbit polyclonal against CAMSAP3). The specificity of the guinea pig polyclonal antibody is shown below, compared to the antibodies from the Takeichi  NOTE: We have removed unpublished data that had been provided for the referees in confidence.
Regarding the nuclear staining mentioned by the Reviewer (Fig. 2B), in our experience it can happen that (especially) polyclonal antibodies occasionally give non-specific staining of the nuclei and/or the cytoplasm, depending on fixation procedure, length of incubation, washing steps (length of time, number of washes, composition of buffer), blocking step (length of incubation, composition of blocking buffer), high expression of transgenes which are recognized by crossreacting secondary antibodies, direct non-specific cross-reactivity of secondary antibodies with nuclear antigens, etc etc. Indeed, in Fig. 2B a similar weak background nuclear staining is visible in the channel of the polyclonal anti-PLEKHA7 antibody. In addition, we see a weak nuclear labeling in the CAMSAP3 channel in Fig. S1B (bottom panel), where there is a high expression of CFP-HA (which we believe is the source of the nuclear non-specific staining in the CAMSAP3 channel). In contrast, in Fig. 1, Fig. 2A, Fig. 4, Fig. 7, Fig. 8, Fig. S1A, Fig. S1B (top panel), Fig. S7 there is no nuclear non-specific background staining with anti-CAMSAP3 polyclonal antibodies. All together we have about 30 immunofluorescence microscopy panels (including XY, Z images of cells, and tissue sections) with no nuclear labeling by the rabbit polyclonal anti-CAMSAP3, versus 3 panels with a weak nuclear labeling. We feel it is reasonable to conclude that the occasional nuclear labeling is due to non-specific cross-reaction that can rarely happen but does not automatically invalidate the specificity of the reagent. And, most importantly, the presence of the nuclear labeling does not impact the conclusion about the presence/absence of the junctional localization of CAMSAP3.