Lysine demethylase 4A is a centrosome‐associated protein required for centrosome integrity and genomic stability

  1. Pratim Chowdhury
  2. Xiaoli Wang
  3. Richard I. Han
  4. Manga Motrapu
  5. Ashley G. Boice
  6. Yuya Nakatani
  7. Sofia Vargas‐Hernandez
  8. Julia F. Love
  9. Claude Chew
  10. Sandra L. Grimm
  11. Dereck Mezquita
  12. Frank M. Mason
  13. Elisabeth D. Martinez
  14. Cristian Coarfa
  15. Cheryl L. Walker
  16. Anna‐Karin Gustavsson
  17. Ruhee Dere

Reviewed by

Read the full article See related articles

Discuss this preprint

Start a discussion What are Sciety discussions?

Listed in

Log in to save this article

Abstract

Centrosomes play a fundamental role in nucleating and organizing microtubules in the cell and are vital for faithful chromosome segregation and maintenance of genomic stability. Loss of structural or functional integrity of centrosomes causes genomic instability and is a driver of oncogenesis. Here we identify lysine demethylase 4A (KDM4A), an epigenetic ‘eraser’ of chromatin methyl marks, as a centrosome‐localized protein, visualized at the nanometer‐scale resolution. We additionally uncovered that KDM4A demethylase enzymatic activity is required to maintain centrosome homeostasis and integrity; a previously unknown functionality unlinked to altered expression of genes regulating centrosome number. We find that KDM4A interacts with and localizes to the centrosome in all stages of mitosis, where it maintains centrosome numbers and centriole engagement during mitosis. Loss of KDM4A results in supernumerary centrosomes and accrual of chromosome segregation errors including chromatin bridges and micronuclei, markers of genomic instability. In summary, these data highlight a previously unknown role for an epigenetic ‘eraser’ regulating centrosome integrity, mitotic fidelity, and genomic stability at the centrosome.

Article activity feed

  1. Version published to 10.1111/febs.70240
  2. Review Commons

    Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

    Learn more at Review Commons

    Reply to the reviewers

    Reviewer 1

    Comment 1: A gallery of different cell cycle stages should be included to define KDM4A centrosomal localization at G1, S and G2 phases and whether it is localized to duplicating centrosomes.

    Response: We thank the reviewer for this excellent suggestion. We have now included Fig S1H demonstrating the persistence/retention of KDM4A at the centrosome through the cell cycle. The text in the Results section has been updated to reflect this addition.

    Comment 2: The immunoprecipitations in Fig. 1 and Supp. Fig. 1 must include appropriate controls. There is no positive control in Fig. 1E and the negative controls for the tagged pulldowns are not appropriate in that there is no other HA-tagged protein in cells. Antibody controls and the reciprocal immunoprecipitations should also be included in the same figure (with controls).

    Response: To address the first point, we have included Histone H3 as a positive control for the KDM4A antibody in Figures 1E and 1F. As for the second point raised by the reviewer, the empty vector is an HA-tagged empty vector and so the antibody controls are already included in the Figure as the ‘empty vector’. We have now included detailed information in the Figure legend to clarify the same. In addition, as suggested by the reviewer we have moved the reverse IPs to the main Figure 1 (Figures 1G and 1I).

    Comment 3: Fig. 1H: The use of overexpressed GFP-centrin for immunoprecipitations is questionable; centrin overexpression can cause centrosome amplification, so the level of centrin relative to the endogenous level should be demonstrated.

    Response: This is with regards to the renumbered Fig. 1J. We have generated hTERT RPE-1 GFP-Centrin expressing stable cell lines that were used for our studies. This is a commonly used cell line in the field and although transient over-expression of GFP-centrin does cause centrosome defects, stable cells are less likely to have elevated centrosome defects. Importantly, the concern of overamplification of centrosomes in these cells is less of a concern given that we have only used these cells to validate the localization of KDM4A to centrosomes using centrin as a centrosome marker. Nonetheless, to ensure that we do not have an aberrant increase in centrosome defects in these cells we have included IF images of our cells (green channel in low-mag and high-mag images below) and are happy to report that we did not observe significantly elevated incidence of centrosome amplification in these stable cell lines.

    Comment 4: The precise localization of KDM4A should be determined more clearly with respect to known centrosomal structures/ regions. One would speculate a PCM localization from the data presented here, but the use of centrobin as a marker does not allow the mother centriole's location to be determined with great clarity. It is unclear why the authors chose centrobin as a marker; further explanation of this might be helpful to the reader. Centrobin is usually cited as a daughter centriole marker (PMID: 16275750, but see 29440264). Supp. Fig. 3J appears to shows 2centrioles labelled with centrobin but the paper does not specify whether centrobin is chosen as a daughter marker or otherwise.

    Response: We thank the reviewer for this astute observation. Our initial rationale for choosing centrobin was simply to use a centrosome marker that worked robustly and reliably with minimal background staining, essential for the single-molecule super-resolution imaging. The question we wanted to address was generating a geographic region in the cell showing nano-scale localization of KDM4A. The 2D images shown in Fig. 2 can be understandably static and hard to visualize the 3D distribution of KDM4A which is not exclusive to centrioles (centrobin although more daughter centriole, does show weaker signal at the mother centriole as well). We have now extensively re-worked Figure 2, including the inclusion of a video in Supplemental Information. We have now included new nano-scale imaging of KDM4A with g-tubulin (a more traditional centrosome marker), which shows a similar distribution of KDM4A across the centrosome and have also included distribution measurements along the x- and y-axis showing both KDM4A and centrobin/g-tubulin. We have modified the text to refer to centrobin as a centrosome marker (centrobin as the reviewer rightly noted can localize to both centrioles although predominantly at the daughter centriole).

    Comment 5: Related to this localization issue, Fig. 2D is unclear to this reviewer. What is this normalized to- a marker or just a set of coordinates? This is an unusual means of representing a localization that does not help the reader understand the (sub-) centrosomal location of KDM4A. The analysis in Supp. Fig. 4 is of somewhat limited value and might be omitted.

    __*Response: *__We apologize for the confusion with the Figure and have simplified the graphs to indicate the single-molecule distribution plotted along the x- and y-axis showing both KDM4A and the centrosome markers i.e. centrobin and g-tubulin.

    Comment 6: Fig. 3 shows the amplification of gamma-tubulin signals, but there is no control for cell cycle stage. The Kdm4a knockout cells appear to be twice the size of the controls, suggesting a G2 phase arrest, which can potentiate centrosome overduplication, or cytokinetic failure in a previous cell cycle (this may also be the case in Figs. 6C and 7B). Therefore, these cells should be phenotyped more robustly with respect to their proliferative characteristics and cell cycle phase distribution. Cell cycle phenotypes should also be checked in the rescue experiments.

    Response: We thank the reviewer for the comments above. The cells shown in Fig. 3 are interphase cells evaluated for centrosome numbers in Kdm4a-deficient cells, independent of mitosis. We apologize for the lack of clarity and the confusion generated by our erroneous statement at the beginning of the paragraph “we next investigated a functional role for KDM4A at mitotic centrosomes”. In fact, we started by first evaluating interphase cells to interrogate consequences of losing Kdm4a, followed by evaluations of the mitotic phenotypes once we observed increased centrosome numbers. This error has now been corrected in the Results.

    As for the reviewer’s comment on phenotyping the cells further, we have now performed these evaluations and have included them in Figure 3 (as new panels Figures 3D, 3E, 3F). Our MTT proliferation assays showed the Kdm4a-null cells proliferated slower than control non-targeted MEFs, although this did not result in any significant issues with cell cycle progression with both cell lines progressing without any arrests and importantly without accumulating increased DNA content/aneuploidy. The rescue cell lines were also phenotyped (new Figures 7C, 7D and 7E) and similarly did not show any altered cell cycle progression.

    Comment 7: Related to the previous point, in the DAPI staining in Figure 5A, 'pseudo-bipolar' cells #1 and #3 (from the top) seem to have greatly increased levels of DNA, suggesting failed cytokinesis as a mechanism of centrosome abnormality. This is a very different process to a centrosome overduplication within a single cell cycle; given that these are knockouts, it is not clear what conclusions should be drawn from the current analysis.

    Response: The reviewer makes an excellent point, about the increased centrosome numbers arising from failure to complete cytokinesis. We have performed further phenotyping of the Kdm4a-null cells, included as new Figures 3D, 3E and 3F. Although the Kdm4a-deficient cells grew slower than their Kdm4a-proficient counterparts, there were no significant issues with cell cycle progression and importantly no evidence of increased aneuploidy. We have also now performed further analysis using centrin as a centriole marker to quantify centrosome numbers (new Figures 4C, 4D and 4E) and have found that there is a significant increase in disjointed centrioles (Figure 4E) suggesting that in addition to any potential amplification there also appears to be an increased loss of cohesion in cells deficient for KDM4A. We have also further confirmed presence of single/disjointed centrioles using TEM analysis (new Figure 4F)

    Comment 8: The JIB-04 result may suggest that KDM4A inhibition causes fragmentation of spindle poles, given that it is a relatively short treatment that would probably not be long enough for centrosome overduplication. Whether this arises during M phase, distinct from the over duplication phenotype seen where there are >4 centrioles, should be posed as a separate question- these may be distinct outcomes from KMD4A inhibition at different cell cycle times.

    Response: We completely agree with the reviewer that the JIB-04 treatment is relatively short and does in fact suggest that this is independent of any over duplication phenotype observed in the Kdm4a-CRISPR knockouts. We thank the reviewer for the suggestion of posing two separate questions to address this point and have made the changes in the manuscript (see Results). In addition, our new data discussed in Comment 7 above, corroborates this hypothesis.

    Comment 9: It is unclear why the authors call the cell shown in Fig. 4B 'pseudo-bipolar'- there are clearly four poles here (as in the multipolar example shown in Fig 5A). This makes the data in Fig. 5 difficult to interpret. The authors should review their classification.

    Response: We thank the reviewer for catching this error. We apologize for the misrepresentation of the representative image and have now included the correct image that shows pseudo-bipolar spindles (new Figure 5D) replacing the multipolar spindle. In addition, we have reviewed our data and the quantitation remains unchanged.

    Comment 10: Expression of the vector control in the Kdm4a nulls in Fig. 7A appears to show a decline in the H3K36me3 levels, confusing the outcome of this experiment. Quantitation should be provided for these blots.

    Response: We have now included the requested quantitation (new Figure 7B) for Figure 7A.

    Comment 11: A rescue experiment should be included for the siRNA knockdown of KDM4A.

    Response: A rescue experiment with the siRNA experiments is challenging as we use a pooled siRNA (4 siRNAs) targeting KDM4A. Rescue with a KDM4A construct would result in the knockdown of the exogenously expressed KDM4A as well. The rescue experiments have been therefore performed with the CRISPR knockout cell lines.

    Comment 12: Size markers should be shown in all immunoblots.

    Response: We have now included size markers as requested by the reviewer for all Figures showing immunoblots (Figures 1, 5, 7 and Supplementary Figures 1, 5).

    Comment 13: p.6, 11 'the resulting payment' and 'caustic chromosome environment' are strange usages and should be rephrased.

    Response: The text has been rephrased.

    Comment 14: Are all panels shown at the same magnification in Fig. 1B? (The telophase DAPI appears different to the anaphase)

    Response: We have confirmed that the magnification is the consistent across the entire panel of images in Figure 1.

    Comment 15: Blow-up panels should be shown so that the centrosomes can be visualised more clearly (Fig. 1 and Supp. Fig. 1).

    Response: We have now included blow-up panels for all centrosome images in Figure 1 and Supplemental Figure 1.

    Comment 16: The MT labelling in Fig. 1D is not of good quality; this imaging should be improved.

    Response: We believe that microtubule densities are impacted by modulating KDM4A in cells likely arising from alternate mechanisms that we are currently investigating. However, to the reviewer’s point we have placed the transient overexpression images in Supplementary information (Supplemental Figure 1I) and have replaced with new Figure 1D, using our stable clones expressing RFP-vector or RFP-KDM4A.

    Reviewer 2

    Comment 1: Coimmunoprecipitation and GFP-trap analyses demonstrated interactions between KDM4A and centrobin, CP110, and centrin-2 (Fig. 1). While the authors suggest a functional a functional association with the centrosome, it is noteworthy that no known centriole protein has been identified to interact simultaneously with centrobin, CP110, and centrin-2, located in distinct sub-centriolar regions. Additionally, 3D super-resolution microscopy indicates that KDM4A is not restrained to a particular region of the centrosome, surely not at the centriole (Fig. 4D). These results hint that centrobin, CP110 and centrin-2 may be potential substrates of KDM4A. Therefore, it is worth to conduct immunostaining and coimmunoprecipitation analyses with the JIB-04-treated cells.

    Response: The reviewer makes an excellent point. The co-immunoprecipitation studies were not conducted to show a direct interaction between the centrosome proteins and KDM4A, but more as a proof-of principle that KDM4A is interacting with centrosome proteins (we do not know if this is direct or indirect, although the data would likely suggest an indirect mechanism). Given that we had used centrobin, centrin and CP110 in our immunofluorescence analysis we also used them for our co-IP studies to provide further evidence of a centrosome localization for KDM4A. It is intriguing that any one of these proteins could in fact be substrates for KDM4A, although an in-depth study would be required to prove this since the super-resolution localization would suggest that KDM4A is not at the centrioles per se and is in fact more of a pericentriolar protein. We have clarified this point in the Discussion. Although the experiments suggested with the JIB treatment would be intriguing, identifying a bone fide centrosome substrate for KDM4A’s demethylase activity is not trivial and would require identification of methylation on a substrate followed by then determining if KDM4A can demethylate the target. Methylation on non-chromatin substrates such as centrosome proteins is not currently well characterized.

    Comment 2: The generation supernumerary centrioles in Kdm4a KO MEFs is intriguing yet warrants careful description (Fig. 3). First, supernumerary centrioles should be coimmunostained with multiple centriole markers, such as centrin-2, CP110 and centrobin antibodies at synchronized populations such as G1, S and M phases. Second, the number of centrioles per cells may be counted and statistically analyzed.

    Response: We thank the reviewer for making this suggestion. We have now included new Figures 4C, 4D, 4E and 4F where we show immunofluorescence with Centrin 2 in Kdm4a-deficient cells. Having found an increased incidence of unpaired centrioles in cells deficient for Kdm4a we have further performed TEM to show the presence of these unpaired/disjoint centrioles.

    Comment 3: The high proportion of pseudo-bipolar cells in the NT group requires attention (Fig. 5).

    Response: We thank the reviewer for this astute observation. To obtain enough mitotic cells for analysis we synchronized the MEFs, which appeared to increase the baseline of pseudo-bipolar spindles reflected in the quants. Despite this increase the differential between the controls and Kdm4a-null cells is significant, as indicated, and we have now made this evident in the text for clarity.

    Comment 4: The KO-rescue cells should be valuable tools to confirm specific roles of KDM4A at the centrosome (Fig. 7). The authors may generate stable cell lines in which wild type and H188A mutant KDM4A are expressed in the KO cells, and use them for centrosome localization of the ectopic proteins, spindle formation and supernumerary centriole generation.

    Response: The reviewer makes an excellent point and in fact we generated the stables (Figure 7) with this idea in mind. Unfortunately (but not completely surprising as this is frequently observed in comparable settings) we observed decreased mitotic abnormalities and genomic instability in the Kdm4a-null cells over time in culture. This is likely arising from a compensatory mechanism/redundancy that perhaps kicks in to enable survival of these cells. The process of generating the stables was therefore tricky with us only being able to reliably analyze genomic stability as a downstream readout of mitotic abnormalities that might have occurred in these cells (early passages analyzed for genomic stability).

    Reviewer 3

    Comment 1: Figure 1D: the RFP vector alone localizes to the centrosome. How was the signal across the cells? Can the authors provide a fluorescence intensity measurement comparing the negative control RFP and RFP-KDM4A to demonstrate the localization at centrosomes of the enzyme? While I found the endogenous staining convincing, the fusion protein is less.

    Response: The MEFs were transiently transfected with the RFP-vector/KDM4A for the images shown. In our experience it is not uncommon for the RFP/mCherry/GFP tags to be prominent at the spindle and often tagged vector controls are omitted from many prominent publications. However, in our case there is a significant increase in RFP-KDM4A signal observed at the spindle poles and we have now included the quantification of signal from the two poles in Supplemental Figure 1J where the signal is 3 times higher in the RFP-KDM4A expressing cells compared to vector. We have also included new Figure 1D demonstrating the RFP-KDM4A localization to spindle poles in our stable cell lines where the signal for the control RFP-vector is negligible. The transient transfection data has been moved to Supplemental Figure 1 (1I).

    Comment 2: Figure 1E-F: How specific do the authors think the interactions with CP110 and centrobin are? Do they IP the entire centrosome proteome or do they think that they reveal some specific interactions within the centrosome? Can the authors comment on this? What is the significance of these interactions? Do the authors think that KDM4A is a centriolar component? Or a PCM component? This is only briefly mentioned in the discussion, it should be extended. Did they try to IP PCM components as well?

    Response: The reviewer brings up an excellent point. The purpose of the immunoprecipitation was to demonstrate the ability of KDM4A to pull down centrosome associated proteins and vice versa. We are unable to comment on the interactions being direct or indirect, although we suspect that most of the interactions are likely indirect, given that KDM4A is not specifically localized to the centrioles. As per the reviewer’s suggestion, we have now expanded the Discussion to speculate on the potential significance of these interactions and how they might enable identification of novel KDM4A interactors and potential substrates.

    Comment 3: Fig.S3: the signal of KDM4A seems broader than that of centrobin, with an average diameter of 749 nm. What is the diameter of centrobin for comparison using this method? The interpretation of the authors concerning this localization is not clear to me: "The quantification data of the diameter of the KDM4A distribution, independently in the different axes (x, y, z), revealed a relatively uniform/circular distribution (Fig. 2D) suggesting that KDM4A was not restrained to a particular region of the centrosome". Is KDM4A at centriole or at centrosomes? PCM or centriole component? From the interpretation stated above, it seems that KDM4A is everywhere from the proximal to the distal axis of the centriole, is it correct? But isn't more PCM?

    Response: We would like to apologize for the lack of clarity with respect to the centrobin measurements compared to those of KDM4A. We have attempted to clarify the distributions measurements by showing the distributions for both the centrobin and KDM4A signals. In addition, we have anow included new data with g-tubulin to show co-localization of KDM4A signal with g-tubulin and to also demonstrate that the signal for KDM4A is not centriole specific but is essentially more uniformly distributed throughout the centrosome. We have also included a video (Video 1) as Supplemental data to clarify this point.

    Comment 4: Fig.4B: The authors established that there is an increase in centrosome number upon short inactivation of KDM4A by JIB-04, which affects its enzymatic activity and not the scaffolding function. In addition, the loss of KDM4A phenocopies the effect of the drug: this means that the enzymatic activity is required to control the centrosome number. This is also re-enforced by the rescue with WT enzyme and not the enzymatically dead mutant of KDM4A (looking at micronuclei formation-Fig.7). Could the author speculate on this? The fast action of the inhibitor would exclude a block in S phase as stated in the discussion. The authors mention centrosome fragmentation but there is no evidence that this is happening here. The authors mentioned several possible mechanisms in the discussion without really exploring them. The authors also mention here that the chronic loss of KDM4A could arise through a distinct mechanism than that of the inhibitor, this statement was surprising. Could the authors check if they have a cell cycle delay or block in their KO cells? While it seems that the authors would like to address these points in the future, I think that the mechanistic aspect is lacking in this study or at least some hints of it.

    Response: We agree with all the points brought up by the reviewer. We have elaborated the discussion as recommended, however the challenge with a demethylase is identifying a potential methyltransferase that can lay a methyl mark on a potential substrate followed by then establishing KDM4A as an eraser for the same substrate. To address, the comment about a cell cycle delay as also brought up by Reviewer 1 (Comment 6), we have performed additional phenotyping of the cells and these data are now included in Figure 3 (as new panels Figures 3D, 3E, 3F) and new Figures 7C, 7D and 7E (for the rescue cell lines) which did not show any altered cell cycle progression.

    Comment 5: In general, the figures are organized in an unconventional manner with the panels from one figure distributed on several pages. Could the authors group the panels of each figure in one page to ease the understanding and the reading?

    Response: Although we do understand how having multiple panels on several pages makes its difficult to read, the immunofluorescence images would be extremely difficult to observe clearly. Also, this comment will be resolved once the manuscript is accepted for publication as we will re-format per journal guidelines.

    Comment 6: Figure S1F-G: the authors provide a large field of view showing a dozen of nuclei. While I acknowledge that this is to show the overall staining, itis difficult to really see the foci of KDM4A or g-Tubulin or centrin. The quality of the images looks really pixelated; this might be due to the PDF compression, but I cannot see any red signal on the panels. Could the authors enhance the B/C of the images so that one can see the signal corresponding to the centrosomes? Is it also possible to have a zoom on the centrosome itself with split channels to illustrate the co-localization? As it is, it is not clearly shown. In the panel G, there are many foci of KDM4A in the nucleus and 2 associated with a centrin staining, which correspond to the centrosomes. However, the signals do not seem to fully colocalize. What do the authors think about this?

    Response: We have provided larger zoomed in view of the cells in Figure S1 as requested.

    Comment 7: Figure 1A: same comment as above concerning the quality of the image. I am also concerned by the g-Tubulin staining as it looks not on focus and I do not see any foci that would correspond to the centrosome position, while the merge image clearly shows yellow signal, proof of co-localisation. Could the author correct this? In the inset, can the authors zoom on the centrosomes and display the split channels so that one can appreciate the co-localization of the 2 signals?The quality and display of Fig.1B is much better. Could we have the same rendering for the interphase cells of 1A?

    Response: The picture in Figure 1A is a raw image. This image has not undergone the same post-image deconvolution applied to the other images in the manuscript. The deconvolved images reduce the KDM4A signal in the nucleus and only demonstrate the highly intense signal at the centrosomes especially in mitotic cells. If we show the deconvolved image here it would lead to the erroneous perception that there is no KDM4A signal in the nucleus and the rest of the cell. To clarify this point we have modified the figure legends to state that this is a raw image. In addition, we have also provided blow-ups of the centrosomes specifically.

    Comment 8: Fig.3D: the nucleus of the cell is really affected with many blobs or micronuclei. Is this cell dying? The authors count the number of g-tubulin foci in interphase (Fig. 3C). Could they do it in mitosis and use centrin? In mitosis, there should be 4

    Response: The cell in question is not dying and is micronucleated. The question of genomic instability is addressed later in the manuscript and hence the point was not made in this figure. We thank the reviewer for suggesting use of centrin. We have now included these data as new Figures 4C, 4D and 4E.

  3. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #3

    Evidence, reproducibility and clarity

    Summary:

    Centrosomes are microtubules-based structures surrounded by a pericentriolar material, serving as Microtubule Organizing Center (MTOC) and are thus important during cell division. Ensuring proper segregation of the genetic material is crucial and defects occurring during this step can lead to drastic consequences like aneuploidy and chromosome instability. It is well established that centrosome defects (number, function, structure) can give rise to defective mitosis. In the present study, Chowdhury et al. demonstrate that the lysine demethylase 4A (KDM4A), known as a chromatin methyl marks eraser, localizes to centrosomes both in interphase and mitosis and is important for centrosome homeostasis. Intriguingly, the authors propose that the novel role of KDM4A in regulating centrosome integrity is unrelated to its function in regulating gene expression but linked to its enzymatic activity without providing a mechanistic advance.

    Major Comments

    This manuscript reports convincingly the localization of KDM4A at centrosomes both in interphase and in mitosis as well as the phenotype linked to the loss of KDM4A. While these are interesting observations, there are some important aspects for improvements that are listed below:

    • Figure 1D: the RFP vector alone localizes to the centrosome. How was the signal across the cells? Can the authors provide a fluorescence intensity measurement comparing the negative control RFP and RFP-KDM4A to demonstrate the localization at centrosomes of the enzyme? While I found the endogenous staining convincing, the fusion protein is less.
    • Figure 1E-F: How specific do the authors think the interactions with CP110 and centrobin are? Do they IP the entire centrosome proteome or do they think that they reveal some specific interactions within the centrosome? Can the authors comment on this? What is the significance of these interactions? Do the authors think that KDM4A is a centriolar component? Or a PCM component? This is only briefly mentioned in the discussion, it should be extended. Did they try to IP PCM components as well?
    • Fig.S3: the signal of KDM4A seems broader than that of centrobin, with an average diameter of 749 nm. What is the diameter of centrobin for comparison using this method? The interpretation of the authors concerning this localization is not clear to me: "The quantification data of the diameter of the KDM4A distribution, independently in the different axes (x, y, z), revealed a relatively uniform/circular distribution (Fig. 2D) suggesting that KDM4A was not restrained to a particular region of the centrosome". Is KDM4A at centriole or at centrosomes? PCM or centriole component? From the interpretation stated above, it seems that KDM4A is everywhere from the proximal to the distal axis of the centriole, is it correct? But isn't more PCM?
    • Fig.4B: The authors established that there is an increase in centrosome number upon short inactivation of KDM4A by JIB-04, which affects its enzymatic activity and not the scaffolding function. In addition, the loss of KDM4A phenocopies the effect of the drug: this means that the enzymatic activity is required to control the centrosome number. This is also re-enforced by the rescue with WT enzyme and not the enzymatically dead mutant of KDM4A (looking at micronuclei formation-Fig.7). Could the author speculate on this? The fast action of the inhibitor would exclude a block in S phase as stated in the discussion. The authors mention centrosome fragmentation but there is no evidence that this is happening here. The authors mentioned several possible mechanisms in the discussion without really exploring them. The authors also mention here that the chronic loss of KDM4A could arise through a distinct mechanism than that of the inhibitor, this statement was surprising. Could the authors check if they have a cell cycle delay or block in their KO cells? While it seems that the authors would like to address these points in the future, I think that the mechanistic aspect is lacking in this study or at least some hints of it.

    Minor comments

    In general, the figures are organized in an unconventional manner with the panels from one figure distributed on several pages. Could the authors group the panels of each figure in one page to ease the understanding and the reading?

    • Figure S1F-G: the authors provide a large field of view showing a dozen of nuclei. While I acknowledge that this is to show the overall staining, it is difficult to really see the foci of KDM4A or g-Tubulin or centrin. The quality of the images looks really pixelated; this might be due to the PDF compression, but I cannot see any red signal on the panels. Could the authors enhance the B/C of the images so that one can see the signal corresponding to the centrosomes? Is it also possible to have a zoom on the centrosome itself with split channels to illustrate the co-localization? As it is, it is not clearly shown. In the panel G, there are many foci of KDM4A in the nucleus and 2 associated with a centrin staining, which correspond to the centrosomes. However, the signals do not seem to fully colocalize. What do the authors think about this?
    • Figure 1A: same comment as above concerning the quality of the image. I am also concerned by the g-Tubulin staining as it looks not on focus and I do not see any foci that would correspond to the centrosome position, while the merge image clearly shows yellow signal, proof of co-localisation. Could the author correct this? In the inset, can the authors zoom on the centrosomes and display the split channels so that one can appreciate the co-localization of the 2 signals? The quality and display of Fig.1B is much better. Could we have the same rendering for the interphase cells of 1A?
    • Fig.3D: the nucleus of the cell is really affected with many blobs or micronuclei. Is this cell dying? The authors count the number of g-tubulin foci in interphase (Fig. 3C). Could they do it in mitosis and use centrin? In mitosis, there should be 4 centrin dots (2/spindle pole), it would nicely complement the phenotype of increased number of centrioles. How do the authors interpret these supernumerary centrioles? Is it due to overduplication? Centriole fragmentation? De novo centriole formation? Or failure of cytokinesis?

    Significance

    Controlling centrosome number is key to ensure faithful chromosome segregation. Increased number of centrosomes through diverse mechanisms can lead to abnormal mitosis despite the well-known centrosome clustering mechanism that permits the formation of a bipolar spindle. In this manuscript, the authors describe the presence of a chromatin eraser KDM4A at centrosomes across the cell cycle without specific localization within the centriole. While the role of KDM4A on chromatin has been described, the authors uncovered a novel function for KDM4A through its enzymatic activity in regulating centrosome numbers that would be independent on its impact on gene expression. The findings described in this manuscript are interesting despite lacking a mechanistical understanding ("Further studies will be necessary to understand the mechanistic underpinnings and molecular targets of KDM4A enzymatic activity at the centrosome"). Conceptually, it is interesting for scientists from centrosome and mitosis fields to consider uncanonical proteins, as exemplified by the enzyme KDM4A, in regulating centrosome function.

  4. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    In this study, the authors explored the involvement of lysine demethylase 4A (KDM4A), a chromatin regulatory factor, in centrosome dynamics. They observed that KDM4A localizes to the centrosomes and identified physical interactions between KDM4A and key centriole proteins, such as centrobin, CP110, and centrin-2 through coimmunoprecipitation. Loss of KDM4A function resulted in various mitotic abnormalities including spindle defects, supernumerary centriole formation, chromatin bridges, and micronuclei formation. Additionally, treatment with JIB-04, a KDM4A inhibitor, exacerbated spindle defects, implicating enzymatic activity of KDM4A in spindle pole function during mitosis. From these findings, the authors inferred a novel role for KDM4A in maintaining centrosome integrity, ensuring mitotic fidelity, and preserving genomic stability.

    A comprehensive characterization of centrosome phenotypes in KDM4A-null cells would be invaluable. While the quality of microscopic images immunoblots is commendable, the preliminary nature of some results prompts further inquiry. The follows are major questions to be answered.

    1. Coimmunoprecipitation and GFP-trap analyses demonstrated interactions between KDM4A and centrobin, CP110, and centrin-2 (Fig. 1). While the authors suggest a functional a functional association with the centrosome, it is noteworthy that no known centriole protein has been identified to interact simultaneously with centrobin, CP110, and centrin-2, located in distinct sub-centriolar regions. Additionally, 3D super-resolution microscopy indicates that KDM4A is not restrained to a particular region of the centrosome, surely not at the centriole (Fig. 4D). These results hint that centrobin, CP110 and centrin-2 may be potential substrates of KDM4A. Therefore, it is worth to conduct immunostaining and coimmunoprecipitation analyses with the JIB-04-treated cells.
    2. The generation supernumerary centrioles in Kdm4a KO MEFs is intriguing, yet warrants careful description (Fig. 3). First, supernumerary centrioles should be coimmunostained with multiple centriole markers, such as centrin-2, CP110 and centrobin antibodies at synchronized populations such as G1, S and M phases. Second, the number of centrioles per cells may be counted and statistically analyzed.
    3. The high proportion of pseudo-bipolar cells in the NT group requires attention (Fig. 5).
    4. The KO-rescue cells should be valuable tools to confirm specific roles of KDM4A at the centrosome (Fig. 7). The authors may generate stable cell lines in which wild type and H188A mutant KDM4A are expressed in the KO cells, and use them for centrosome localization of the ectopic proteins, spindle formation and supernumerary centriole generation.

    Significance

    This manuscript presents a novel function of KDM4A at the centrosome. Their immunostaining results clearly showed centrosome localization at the centrosome. It is surprising that a chromatin regulator plays a role at the centrosome. It is likely that KDM4A is essential for maintenance of centriole integrity, although the specific mechanisms remain unexplored. Nevertheless, the descriptive nature of the study leaves questions regarding precise contribution of KDM4A to centrosome function unanswered.

  5. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    Chowdhury, Dere and colleagues here explore potential centrosomal roles for the lysine demethylase, KDM4A, better known for its functions in chromatin regulation. They show that it localises to centrosomes during mitosis and that its loss causes centrosome numerical aberrations and spindle multipolarity, along with micronuclei.

    The major issue is the limited insight provided into the potential mechanism(s) by which KDM4A loss impacts centrosome/ spindle pole integrity. There are also several technical issues that should be improved upon.

    Major points

    1. A gallery of different cell cycle stages should be included to define KDM4A centrosomal localisation at G1, S and G2 phases and whether it is localised to duplicating centrosomes.
    2. The immunoprecipitations in Fig. 1 and Supp. Fig. 1 must include appropriate controls. There is no positive control in Fig. 1E and the negative controls for the tagged pulldowns are not appropriate in that there is no other HA-tagged protein in cells. Antibody controls and the reciprocal immunoprecipitations should also be included in the same figure (with controls).
    3. Fig. 1H: The use of overexpressed GFP-centrin for immunoprecipitations is questionable; centrin overexpression can cause centrosome amplification, so the level of centrin relative to the endogenous level should be demonstrated.
    4. The precise localisation of KDM4A should be determined more clearly with respect to known centrosomal structures/ regions. One would speculate a PCM localisation from the data presented here, but the use of centrobin as a marker does not allow the mother centriole's location to be determined with great clarity. It is unclear why the authors chose centrobin as a marker; further explanation of this might be helpful to the reader. Centrobin is usually cited as a daughter centriole marker (PMID: 16275750, but see 29440264). Supp. Fig. 3J appears to shows 2 centrioles labelled with centrobin but the paper does not specify whether centrobin is chosen as a daughter marker or otherwise.
    5. Related to this localisation issue, Fig. 2D is unclear to this reviewer. What is this normalized to- a marker or just a set of coordinates? This is an unusual means of representing a localization that does not help the reader understand the (sub-)centrosomal location of KDM4A. The analysis in Supp. Fig. 4 is of somewhat limited value and might be omitted.
    6. Fig. 3 shows the amplification of gamma-tubulin signals, but there is no control for cell cycle stage. The Kdm4a knockout cells appear to be twice the size of the controls, suggesting a G2 phase arrest, which can potentiate centrosome overduplication, or cytokinetic failure in a previous cell cycle (this may also be the case in Figs. 6C and 7B). Therefore, these cells should be phenotyped more robustly with respect to their proliferative characteristics and cell cycle phase distribution. Cell cycle phenotypes should also be checked in the rescue experiments.
    7. Related to the previous point, in the DAPI staining in Figure 5A, 'pseudo-bipolar' cells #1 and #3 (from the top) seem to have greatly-increased levels of DNA, suggesting failed cytokinesis as a mechanism of centrosome abnormality. This is a very different process to a centrosome overduplication within a single cell cycle; given that these are knockouts, it is not clear what conclusions should be drawn from the current analysis.
    8. The JIB-04 result may suggest that KDM4A inhibition causes fragmentation of spindle poles, given that it is a relatively short treatment that would probably not be long enough for centrosome overduplication. Whether this arises during M phase, distinct from the overduplication phenotype seen where there are >4 centrioles, should be posed as a separate question- these may be distinct outcomes from KMD4A inhibition at different cell cycle times.
    9. It is unclear why the authors call the cell shown in Fig. 4B 'pseudo-bipolar'- there are clearly four poles here (as in the multipolar example shown in Fig 5A). This makes the data in Fig. 5 difficult to interpret. The authors should review their classification.
    10. Expression of the vector control in the Kdm4a nulls in Fig. 7A appears to show a decline in the H3K36me3 levels, confusing the outcome of this experiment. Quantitation should be provided for these blots.
    11. A rescue experiment should be included for the siRNA knockdown of KDM4A.
    12. Size markers should be shown in all immunoblots.

    Minor points

    1. p.6, 11 'the resulting payment' and 'caustic chromosome environment' are strange usages and should be rephrased.
    2. Are all panels shown at the same magnification in Fig. 1B? (The telophase DAPI appears different to the anaphase)
    3. Blow-up panels should be shown so that the centrosomes can be visualised more clearly (Fig. 1 and Supp. Fig. 1).
    4. The MT labelling in Fig. 1D is not of good quality; this imaging should be improved.

    Significance

    • General assessment: Strengths:

    Potential novel localisation and functions for KMD4A.

    Weaknesses:

    Limited mechanistic detail.

    Technical concerns.

    Key biological points not addressed.

    • Advance: While these observations are of potential interest, there are many questions that are not resolved clearly, resulting in a mainly descriptive advance.
    • Audience: Basic research, primarily. The study is likely to be of interest to the centrosome/ cell cycle field, with potential ramifications in genome stability and cancer biology.
  6. Version published to 10.1101/2024.02.20.581246 on bioRxiv