USP9X regulates centrosome duplication and promotes breast carcinogenesis

Defective centrosome duplication is implicated in microcephaly and primordial dwarfism as well as various ciliopathies and cancers. Yet, how the centrosome biogenesis is regulated remains poorly understood. Here we report that the X-linked deubiquitinase USP9X is physically associated with centriolar satellite protein CEP131, thereby stabilizing CEP131 through its deubiquitinase activity. We demonstrate that USP9X is an integral component of centrosome and is required for centrosome biogenesis. Loss-of-function of USP9X impairs centrosome duplication and gain-of-function of USP9X promotes centrosome amplification and chromosome instability. Significantly, USP9X is overexpressed in breast carcinomas, and its level of expression is correlated with that of CEP131 and higher histologic grades of breast cancer. Indeed, USP9X, through regulation of CEP131 abundance, promotes breast carcinogenesis. Our experiments identify USP9X as an important regulator of centrosome biogenesis and uncover a critical role for USP9X/CEP131 in breast carcinogenesis, supporting the pursuit of USP9X/CEP131 as potential targets for breast cancer intervention.

In the current submission Li et al. investigate the function of deubiquitination enzyme USP9X in chromosome instability. This manuscript shows that USP9X binds directly to CEP131 and rescuing the levels of CEP31 in cancer cells. Upregulated CEP31 and USP9X in turn promoted chromosome instability in cancer cells. Next, the authors demonstrate that USP9X is overexpressed in breast carcinomas with worse survivals of these patients. Furthermore, overexpression of USP9X in cancer cells transplanted into mice caused breast carcinogenesis in vivo. The experimental evidence presented in the study is interesting and novel. They clearly show that USP9X regulates CEP31 stability in breast cancer however, the reviewer is not convinced with the authors' claim/evidence that presented regarding the function USP9X in non-cancerous cells. In addition, it is not clear how chromosome instability presented in the first three figures can be coupled to increase survival and proliferation of cancer cells presented in the last figure. Including data using USP9X cells isolated from knockout animals could strength the conclusion of this manuscript.
Major questions: -What is the function and role of USP9X in non-cancerous cells? If the answer is that the expression of USP9X is low in normal cells, this is not clear by the data shown in Fig. 3F.
-USP9X transgenic animal has been generated previously. Including experiments using USP9X primary cells isolated from knockout animals can strength the conclusion of the authors. This important since all of the experiments present in the manuscript are performed by overexpressing/siRNA treating human cancer cell lines.
- Fig. 1E is not necessary and can be presented in Suppl. figures since the same information is presented in presented in 1D.
-The authors need to confirm direct interaction between USP9X and CEP131 using purified proteins. Fig. 1G shows the binding between "truncated mutant" of USP9X and CEP131.
-Downregulation of USP9X leads to reduced levels of CEP131. What is not discussed is "why downregulation of CEP31 which diminish centrosomal localization of USP9X also reduces the levels of USP9X" shown in Figure 2D right panel?
-The percentage of knock down cells of CEP131 needs to be included in Fig. 2.
-It is not clear to the reviewer why ubiquitin mutant was used to establish the type of ubiquitination. The authors should use different ubiquitin mutants including lysine 6,11,29,33,48 and 63 to establish if poly-ubiquitination of CEP131 is via lysine 48 and/or 29.
-Explanation for in vitro Deub. and Ub. assay were not clear. Did the authors use pure in vitro deub. assay using bacterial purified protein? -Global analysis of human ubiquitin-modified proteome was used to identify site of the ubiquitination. It is not clarified how this analysis was performed, while the other problem with such in silico method is the reliability. In particular, the identified sites and mutation of these sites are not preventing totally the ubiquitination status of CEP131 (Fig. 4F). Another approach could be to perform Masspect to cnfirm or identify new sites.
-The data presented in Fig. 5 is convincing however, it is essential to include live cell imaging showing chromosomal instability in overexpressed and SiRNA treated regulated cells.
-A table is necessary to summarize the percentage of "Different" chromosomal instability phenotypes that occurs in overexpressed and down regulated cells.
-Last figure shows growth retardation cancer cells both in vitro and in vivo in ShRNA treated cells. However, in contrast to this information, previous data presented in figure Fig. 2 and Fig.5 shows no differences in the cell cycle or cell survival in siRNA treated cells. Can the authors' explain the discrepancy between these two figures? -There are previous reports showing the involvement of other deubiquitination enzymes in chromosomal stability. Including these references might strengthen the conclusion of the present manuscript.
-Part of the discussion need to focus on or speculating how CEP131 deubiquitination is involved in genome instability of cells and what is the downstream signaling that can be affected.

Review Song et al Nature Communications
In this manuscript, the authors identify CEP131 as a major interacting protein with the deubiquitinase USP9X. They provide experiments to demonstrate that CEP131 protein abundance is positively regulated by the deubiquitinase activity of USP9X. Overexpression of USP9X or CEP131 leads to centrosome amplification and genomic instability. The authors then postulate that increased levels of USP9X and CEP131 in some breast cancer cells is correlated with their metastatic potential.
Although the authors provide reasonable data for a functional link between USP9X and CEP131, the data lack some controls, the specificity of the observed phenotypes is unclear and they do not provide a compelling link between USP9X/CEP131 and breast cancer. These concerns must be adequately met for publication in Nature Communications.
Major issues: No rescue experiments are performed to show the specificity of the RNAi triggers used. Given the propensity of off-target effects using siRNA, rescue experiments are essential. Rescue of phenotypes for critical experiments should be provided (for example, Figures 3A and D, 5C).
The specificity of the phenotypes observed upon USP9X depletion and over expression needs to be confirmed further: 1) Centriolar satellite proteins exhibit a high degree of interdependence in their localizations. For example, the loss of PCM1 or pericentrin affects the localization of CEP131 to satellites; however the loss of CEP131 does not seem to reciprocally affect these proteins (Staples CJ, JCS, 2012). It would be beneficial to show controls with additional satellite proteins to provide convincing evidence that the link between USP9X/CEP131 is direct and not a consequence of a general loss of satellite proteins. For example, is PCM1 localization disrupted in USP9X-depleted cells? does PCM1 abundance decrease after USP9X depletion?
2) The authors consistently find that expression/silencing of CEP131 only partially rescues the phenotype of depleting/over-expressing USP9X (see Figures 5 B,C and D,7C and E). While it seems that there is a functional link between USP9X and CEP131, perhaps this indicates that USP9X functions actually functions upstream of CEP131.
Lastly, the authors try to demonstrate that elevated levels of USP9X and CEP131 are clinically relevant in a breast cancer model. I have a number of concerns about this section of the manuscript: In panel 7A, what is the difference between the two box plots labeled 'Ma Breast 4'? Have these been mislabeled? 1) Oncomine indicates that CEP131 expression is not significantly increased in the Ma Breast 4, Finak breast and Turashvili breast samples. This argues against the authors model where increased USP9X results in increased CEP131, which itself is responsible for the observed phenotypes.
2) Relating to the correlation between the abundance of USP9X and CEP131 in cancer tissues, can the authors provide a negative control to compare. For example, does the abundance of PCM1, another satellite protein, also correlate with USP9X. Perhaps these cells have more centriolar satellites in general and therefore many of the components will correlate. I believe it is important to show that the correlation between USP9X and CEP131 is specific in these cell types.
3) Can the authors provide the parameters used for the Kaplan-Meier analyses and justification for these settings?
A) The web-based tool allows four options for survival analysis, DMFS (distant metastasis free survival), RFS (relapse free survival), OS (overall survival) and PPS (post progression survival). Using the USP9X identifier '201099_at', the survival of high and low expressors is significant for the DMFS and RFS survival, but not OS or PPS. Can the authors explain why they have only presented the DMFS and RFS survival? Is the OS and PPS not applicable for USP9X? If so, please indicate why. B) The webtool indicates that the 'JetSet' identifier, which in the case of USP9X is 229573_at is the preferred identifier. Using this id, the difference between high and low expression is either not significant, or significant in the opposite direction (high expression correlates with better survival). Can the authors please justify why they did not use the preferred identifier? C) Using CEP131 (id 214742_at), high expression is correlated with worse outcome for DMFS, but not RFS, OS or PPS. Since the model is that increased USP9X stabilizes CEP131, one would think that increased CEP131 expression would also result in poorer outcomes. This link only holds with the DMFS survival.
Minor Points: The extent of CEP131 destabilization in response to USP9X depletion seems to vary between some panels. Is it possible that different siRNA durations can account for this? Please see: Figure 3A vs. 3D Figure 7B Page 7: "elution pattern of CEP131 was largely overlapped with that of PCM1 and USP9X" While the elution pattern of CEP131 largely overlapped with USP9X, I do not agree that it largely overlaps with PCM1. Additionally, it seems that there is a large peak of protein, as defined by the A280 trace that might elute around fraction 17. If so, it is less compelling to see multiple proteins in this fraction. This panel should be labeled more clearly to correlate the western blot to the elution profile.

Page 9:
In the domain mapping studies, the authors show that the N-terminus of USP9X is necessary for the interaction with CEP131 but do not show it is sufficient (as they do for the CEP131 domain mapping). Does the N-terminal domain alone interact with CEP131? "USP9X truncation mutants revealed a high affinity between CEP131 and USP9X/M" Statements of affinity cannot be made from these experiments. Please change text to reflect that this is a non-quantitative experiment.
"demonstrated CEP131 N-terminal region directly interacts with USP9X/M" Is there a reason the authors used USP9X purified from mammalian cells? USP9X could interact another co-purifying protein that could bridge the interaction between USP9X and CEP131. Without performing experiments with purely bacterially (or at least non-mammalian) expressed protein the authors cannot claim a direct interaction.
Page 10: The data in Figure 2B does not match the description in the text. The authors suggest that the colocalization between USP9X and Centrin is evident in S/G2, but less so in G1. However, their images are labeled G1/S and G2 so it is not possible to evaluate this claim. Can the authors provide images of G1 cells? Cyclin A staining should discriminate between G1 and S/G2. Although PCM1 and CEP131 proteins are known to reside in centriolar satellites, very little noncentrosomal staining is apparent for these proteins. USP9X does not co-localize with the detectable satellite staining of PCM1. Can the authors comment on this observation? Page 11: "USP9X and CEP131 in these cells oscillated at a similar pace during" Can the authors explain this statement? I do not observe a strong 'oscillation'. The data suggest CEP131 abundance increases around 2 hours, while USP9X increases at 4 to 6 hours. Perhaps a quantification of multiple westerns would strengthen this point. Also, at 12 hours, the phospho-H3 signal suggests cells have entered mitosis. At this point USP9X and CEP131 abundance has not decreased, yet these proteins are not detectable in mitosis in the immunofluorescence images ( Figure 1A). Does the abundance of these proteins decrease by western plot at later time points?
Additionally, time of interaction between USP9X and CEP131 in Figure 2F does not seem to correlate with the initial increase in CEP131 abundance in Figure 2E.

Page 12:
Please provide biological replicates for the quantification in Figure 3D.
Does the loss of CEP131 from the centrosome specific in USP9X siRNA conditions? i.e. are other components also displaced (other than centrin)? Possibly other satellite components such as CEP290 or PCM1? The loss of CEP131 from centrosomes could represent a general loss of satellite components. Figure 3F. Please provide quantification for this. The CEP131 abundance in MDA-MB-231, ECC1 and Ishkawa cells look reasonably similar while having noticeably different levels of USP9X. Perhaps a correlation plot would be more convincing.

Page 14:
What is FLAG-tagged in Figure4D? Is it FLAG-USP9X? Please indicate. The level of overexpression of the wt and C1566S forms of USP9X is not comparable. Is it possible to use a higher doxycycline concentration for the C1566S protein?
The labels on Figure 4F are confusing. The bottom panel is labeled CEP131/K504, yet CEP131/K254R is indicated as being present in one of the lanes.
Please include biological replicates for the quantification of Figure 4G.

Page 16:
Do the FACS profiles correspond to the cells analyzed in the immunofluorescence panels in Figure  5A and C? If so, the profiles look like mostly G1 cells. Can the authors provide A) a FACS profile of cycling cells to use for a comparison B) another method, such as cyclin A staining or western blot to demonstrate the cells are in the expected cell cycle stage.
For Figures 5B and C it would be nice to see western blots to judge protein abundance. For example, USP9X over expression should stabilize CEP131, but CEP131 siRNA should deplete the protein -it is necessary to judge what the final steady state level of the protein is in this case to interpret the functional data.
Page 18: Can Figure 6C be rearranged so all normal and all tumor lines are together? Page 19: In Figure 7C, shRNA against CEP131 results in a weaker colony formation phenotype that using shRNA against USP9X. However, the abundance of CEP131 is lower using shRNA against CEP131 than in cells using USP9X shRNA. This is not entirely consistent with the model where USP9X regulates CEP131 stability.
Perhaps some of the data that is replicated in different cell lines can be presented in Supplementary information. page7: excessive FLAG peptides: excess FLAG peptide(s) centrioloar satellite protein CEP131 was also identified: were also identified 1. The mechanism of USP9X promoted-centrosome amplification has been investigated and the results point to a role of CEP131-regulated CDK2 localization in USP9X-promoted centrosome biogenesis ( Figure S6E, Figure S6F and Figure S6G).
2. Co-immunoprecipitation assays were performed to analyze association of USP9X with the other interactors identified in our USP9X-pull down experiments. The results validated ITCH, IPO5, PRMT5, and PPM1B as USP9X interactors ( Figure S1A), but the protein abundance of IPO5, PRMT5, and PPM1B are not subjected to USP9X regulation ( Figure S10A). Figure 3B, Figure 4A and Figure 4B have been optimized and repeated and the data have been replaced, and the experiments in previous Figure 3D have been repeated and results from biological triplicate experiments have been shown in Figure 3D, Figure S3C, and Figure S3D. 4. We have tested the effects of other USP9X substrates on USP9X regulated centrosome amplification and breast carcinogenesis. The results provided in Figure  S10 suggest that USP9X-promoted CEP131 stabilization functions in centrosome biogenesis and breast cancer cell survival in an ITCH and MCL1 independent manner.

The experiments in
5. The function and role of USP9X in HMECs (human mammary gland epithelial cells) have been investigated and the results provided in Figure S3B and Figure S6C indicated that USP9X plays a conserved role in non-cancerous cells and cancer cells.
6. We utilized CRISPR/Cas9 system to knockout USP9X in HMECs and demonstrated that in USP9X deficient cells the expression level of CEP131 is downregulated and the downregulation of CEP131 in these cells could be rescued by forced expression of wild type USP9X, but not USP9X mutant (USP9X/C1566S) ( Figure S5A), while the mRNA expression level of CEP131 was not altered ( Figure S5A). Figure 1E has been moved to SUPPLEMENTAL FIGURES and shown as Figure S1B; According to the suggestion of Reviewer #3, Figure 3A, Figure 3D, Figure 4C and Figure 5B have been moved to SUPPLEMENTAL FIGURES and shown as Figure S3A, Figure S3D, Figure S5B and Figure S6B, respectively.

Original
8. The y axis in the right panel of Figure 2D and Figure 3F has been replaced with new label "Relative USP9X intensity in centrosome" and "Relative CEP131 intensity in centrosome". 2 9. The percentage of CEP131 or USP9X knockdown cells has been provided in Figure  2D and Figure 3F, respectively. 10. In vivo deubiquitination assays have been performed with ubiquitin mutant with all lysine residues replaced by arginine except K29 (K29-only) or K48 (K48-only) or K63 (K63-only) to differentiate ubiquitin species opposed by USP9X on poly-ubiqitinated CEP131. The results provided in Figure S5D indicated K48-linked ubiquitin species are the major forms targeted by USP9X.
11. Mass spectrometry analysis of ubiquitin conjugation sites on CEP131 is provided in Figure S5E.
12. Live cell imaging analysis of chromosome instability has been provided in Figure S7 and Supplementary Movie 1.
13. The effect of USP9X-promoted CEP131 stabilization on genome stability and cell death has been investigated and the results have been provided in Figure S9.
14. Previous reports on the involvement of other deubiquitination enzymes in chromosomal stability have been included/cited in the revision (paragraph 1, page 30).
15. How CEP131 deubiquitination is involved in centrosome amplification and genome instability and what is the downstream signaling that can be affected have been discussed (paragraph 2, page 28). 16. Rescue experiments have been performed in USP9X deficient cells and the results have been provided in Figure 3E, Figure S6D and Figure S10C.
17. The effects of USP9X depletion on the localization and abundance of PCM1, CEP290 and Pericentin have been examined and the results have been shown in Figure S4.
18. The second dataset of "Ma Breast 4" ( Figure 6A) has been removed. 19. The expression level of PCM1 and CEP290 in breast cancer samples and tumor adjacent tissues has been examined and the results have been provided in Figure S8A and Figure S8B.
20. Figure 6E of Kaplan-Meier analysis has been removed in the revision. Figure 6E have been removed. 3 22. Description of FPLC results ( Figure 1D) in the text has been re-worded. 23. Description of pull-down between USP9X truncation mutants and CEP131 (Figure1E) has been re-worded.

Kaplan-Meier survival analyses in original
24. USP9X deletion mutants purified from Sf9 cells have been used to re-perform the pull-down experiments and the data has been provided in Figure 1F and Figure 1G. 25. Immunostaining with antibodies against USP9X and Centrin with cyclin A-CFP stably expressing U2OS cells has been provided in Figure S2A. 26. Biological triplicate experiments corresponding to original Figure 3D have been provided in Figure 3D, Figure S3C and Figure S3D, and Biological triplicate experiments corresponding to original Figure 4G have been provided in Figure 4G and Figure S5F. The quantitation and statistical analysis of these results have been provided in Figure 3D, Figure S3D and Figure 4G.
27. The results with Western blotting analysis judging the abundance of CEP131 and USP9X post mitosis, have been shown in Figure S2D.
28. Biological triplicate experiments corresponding to original Figure 2E together with quantification and correlation of USP9X and CEP131 has been provided in Figure 2E and Figure S2C.
29. Myc-tagged and FLAG-tagged proteins have been relabeled in Figure 4C and Figure  4D.
30. Higher concentration of doxycycline was used to induce C1566S expression and the deubiquitination assay has been re-performed as shown in Figure S5C. 31. Figure 4F has been relabeled as indicated. Figure 5A have been provided in Figure S6A. 33. FACS profiles corresponding to control or HU treated cells have been provided in Figure 5C. 34. The results with Western bloting analysis judging protein abundance in immunofluorescence stainings have been provided in Figure S6A, Figure 5B, Figure   4 5C, Figure 5D, Figure S4A, Figure S4C, Figure S6B, Figure S6C, Figure S6D, Figure  S6E, Figure S6F, Figure S6G, Figure S10D and Figure S10E as indicated. 35. The misspelled or excess words pointed by Reviewer #3 have been corrected in the revision. 36. Two additional references relevant to USP9X physiological or pathological function are cited in the manuscript (paragraph 1, page 6). 38. SUPPLEMENTAL FIGURES and SUPPLEMENTAL FIGURE LEGENDS are provided in the revision. 5 Response to Reviewer #1"s comments-Reviewer #1: The first difficulty of the paper is that it attempts to give a role to USP9X in regulating CEP131 function when we do not understand what CEP131 -or indeed centriolar satellites -actually do. The authors set out to study USP9X partners and identify CEP131. Unfortunately they do not really indicate in full the other proteins that they identify as interactors and there must be many. Part of the difficulty appears to be that they present data for gel purified interacting proteins rather then presenting unbiased analysis of all interactors identified in the pull down. This is important particularly in interpreting the subsequent experiments in which USP9X function is removed. I am certain that they are correct and that CEP131 is one of USP9X's partnersthey present good evidence for this in Fig. 1 -but what are ALL its other partners? Nevertheless, an association with centrosomes ( Fig. 2) makes sense if USP9X is a CEP131 partner.

FACS profiles and Western blotting results corresponding to sychronized cells used in
Authors: The cellular function of CEP131 and centriolar satellites remains to be delineated, as the reviewer rightfully points out. A recent study reported that CEP131 is involved in centrosome duplication via regulating centrosomal localization of CDK2 1 , a cyclin-dependent kinase with an established role in centrosome biogenesis 2, 3 . This is consistent with our finding that USP9X-promoted CEP131 stabilization is required for centrosome biogenesis. In responding to the reviewer"s criticism, we examined the impact of USP9X loss-of-function on CDK2 localization in centrosome and found that, similar to CEP131 depletion, USP9X deficiency was associated with an impaired centrosomal localization of CDK2 ( Figure S6E), while the protein level of CDK2 remained unchanged in both USP9X and CEP131 knockdown cells ( Figure S6E). Importantly, USP9X depletion-induced phenotype of CDK2 localization could be rescued by CEP131 overexpression ( Figure S6F). Moreover, USP9X promoted-centrosome amplification was abrogated upon CDK2 depletion ( Figure S6G). These results point to a role of CEP131-regulated CDK2 localization in USP9X-promoted centrosome biogenesis.
Our affinity purification and mass spectrometry analysis showed that USP9X is associated with multiple proteins ( Figure 1A). As stated in the manuscript, USP9X has been reported to target several cytosolic proteins, and its association and potential targeting of CEP131 was particular interesting as the cellular compartmentalization and biological function of CEP131 are distinct from that of the reported substrates of USP9X. Thus, we focused our study on the regulation of CEP131. In responding to the reviewer"s 6 criticism, co-immunoprecipitation assays were performed to analyze the other interactors of USP9X identified in our pull down experiments. The results validated ITCH, IPO5, PRMT5, and PPM1B as USP9X interactors ( Figure S1A).

Reviewer #1:
The effects shown in subsequent figures and used to reach conclusions about USP9X function are in many cases modest. For example, the loss of CEP131 following USP9X RNAi in MCF7 cells is not convincing (Fig. 3B). Moreover, the change in CEP131 half-life following USP9X RNAi particularly in U2OS cells is certainly not dramatic. Similarly, the increase in CEP131 following expression of elevated USP9X is not dramatic (Fig. 4A), as claimed, and neither are the diminution of CEP131 levels following inhibition of the DUB with WP1130 (Fig. 4D).
Authors: In responding to the reviewer"s criticism, the experiments in Figure 3B, Figure  4A and Figure 4B (inhibition of the DUB with WP1130) have been optimized, including improving the knockdown or overexpression efficiency and extending the time of doxycycline or inhibitor treatment to cells, and repeated, and the data have been replaced. The experiments in previous Figure 3D (CHX-chase assay) have been repeated and results from biological triplicate experiments have been shown in Figure 3D, Figure S3C, and Figure S3D. In conclusion, much more needs to be done to understand the roles of CEP131 in centrosome biology and the full range of substrates of UPS9X before it will be possibly to correctly interpret this study.
Authors: In responding to the reviewer"s comments, we have performed a series of experiments with other substrates of USP9X: 1) In agreement with previous reports 4, 5, 6, 7, 8 , we found that USP9X knockdown did result in decreased expression of ITCH ( Figure   S10A), an E3 ligase involved in carcinogenesis 4, 6 , and of MCL1, an anti-apoptotic regulator implicated in cancer 9, 10, 11, 12 , However, USP9X, but not CEP131, could be co-immunoprecipitated by ITCH or MCL1 ( Figure S10B); 2) Colony formation assays demonstrated that while overexpression of ITCH or MCL1 could rescue the growth inhibitory phenotype resulted from USP9X depletion to certain extent as CEP131 did, simultaneous expression of CEP131 and ITCH or MCL1 showed an additive effect ( Figure S10C); 3) while loss of USP9X-associated defects of centrosome amplification could be rescued by CEP131 overexpression ( Figure S10D), overexpression of ITCH or MCL1 could not rescue the phenotype ( Figure S10D); 4) knockdown of either ITCH or MCL1 had no effect on centrosomal localization of USP9X and CEP131 (Figure S10E); 7 and 5) Western blotting analysis of cellular lysates of USP9X-deficient MCF-7 cells revealed that the expression of IPO5, PRMT5 and PPM1B, which are also identified as interactors of USP9X in our experiments ( Figure S1A), was essentially unchanged ( Figure S10A), suggesting that the protein abundance of IPO5, PRMT5 and PPM1B is not subjected to USP9X regulation, although the effect of these proteins on USP9X functionality needs to be further investigated. The data have been added to the revision as indicated. Response to Reviewer #2"s comments-

Major questions:
Reviewer #2 -What is the function and role of USP9X in non-cancerous cells? If the answer is that the expression of USP9X is low in normal cells, this is not clear by the data shown in Fig. 3F.
Authors: USP9X was co-localized with Centrin at centrosome in non-cancerous cells HMECs (human mammary gland epithelial cells) ( Figure S2B), and the protein, but not mRNA expression level of CEP131 was down-regulated upon USP9X depletion in HMECs ( Figure S3B). Consistent with the role of USP9X in cancerous cells, we demonstrated that USP9X promotes centrosome biogenesis in a CEP131-dependent manner in HMECs ( Figure S6C).
Our data indicate that the expression of USP9X is low in non-cancerous cells compared to cancerous cells of the same tissue origin (original Figure 3F, now shown as Figure 3G). Consistently, measurement of the expression of USP9X and CEP131 in human tissues by Western blotting or immunohistological analysis showed the protein levels of both USP9X and CEP131 were substantially elevated in breast carcinoma samples ( Figure 6C and Figure 6D). . Reviewer #2 -USP9X transgenic animal has been generated previously. Including experiments using USP9X primary cells isolated from knockout animals can strength the conclusion of the authors. This important since all of the experiments present in the manuscript are performed by overexpressing/siRNA treating human cancer cell lines.
Authors: We appreciate the reviewer for this comment. However, we have tried to obtain USP9X transgenic animals from the researchers who generated the animals 1, 2 without success. To address the issue, we utilized CRISPR/Cas9 system to knockout USP9X in HMECs and demonstrated that, in USP9X deficient normal mammary cells, the expression level of CEP131 decreased and the downregulation of CEP131 in these cells could be rescued by forced expression of wild type USP9X, but not USP9X mutant (USP9X/C1566S) ( Figure S5A), while the mRNA expression level of CEP131 was not altered ( Figure S5A). These results are consistent with our observations in the manuscript.
Reviewer #2 - Fig. 1E is not necessary and can be presented in Suppl. figures since the same information is presented in presented in 1D.
Authors: Figure 1E has been moved to SUPPLEMENTAL FIGURES and shown as Figure S1B.
Reviewer #2 -The authors need to confirm direct interaction between USP9X and CEP131 using purified proteins. Fig. 1G shows the binding between "truncated mutant" of USP9X and CEP131.
Authors: We agree with the reviewer on this point. However, possibly due to its high molecular mass (~300 kDa), USP9X full length protein is difficult to purify in either bacteria or insect Sf9 cells, despite our intensive efforts. To address the reviewer"s concern, USP9X deletion mutants purified from insect Sf9 cells have been used to repeat the pull-down experiments and the data has been provided in Figure 1F and Figure 1G.
Reviewer #2 -Downregulation of USP9X leads to reduced levels of CEP131. What is not discussed is "why downregulation of CEP31 which diminish centrosomal localization of USP9X also reduces the levels of USP9X" shown in Figure 2D right panel?
Authors: The y axis labeled with "Relative USP9X intensity" represents the relative USP9X intensity in centrosome, not in the whole cell. We apologize for the confusion and have changed the labeling of y axis to "Relative USP9X intensity in centrosome" ( Figure 2D right panel). As demonstrated in the right lower panel of Figure 7B and the right upper panel of Figure 7D, CEP131 downregulation had no effect on the expression of USP9X.
Reviewer #2 -The percentage of knock down cells of CEP131 needs to be included in Fig.  2.
Authors: The percentage of knockdown cells of CEP131 has been included in Figure 2D.
Reviewer #2 -It is not clear to the reviewer why ubiquitin mutant was used to establish the type of ubiquitination. The authors should use different ubiquitin mutants including lysine 6,11,29,33,48 and 63 to establish if poly-ubiquitination of CEP131 is via lysine 48 and/or 29.
Authors: The ubiquitin mutant with all lysine residues replaced by arginine (Ub/mt) was used as a negative control in poly-ubiquitin chain detection. To comply with the reviewer, we have added additional ubiquitin mutants with all lysine residues replaced by arginine except K29 (K29-only) or K48 (K48-only) or K63 (K63-only) to differentiate ubiquitin 11 species opposed by USP9X on poly-ubiqitinated CEP131. The results in Figure S5D indicate that K48-linked ubiquitin species are the major forms of CEP131 targeted by USP9X.
Reviewer #2 -Explanation for in vitro Deub. and Ub. assay were not clear. Did the authors use pure in vitro deub. assay using bacterial purified protein?
Authors: As explained above, full length USP9X protein is difficult to purify. Thus, we used FLAG-USP9X/wt and FLAG-USP9X/C223S purified from mammalian cells in high salt and detergent buffer for in vitro Deub assays. Moreover, since the bona fide E3 ligase for CEP131 is currently unidentified and Ub-conjugated CEP131 could not be generated in vitro, we retrieved HA-Ub-conjugated FLAG-CEP131 from HeLa cells with tandem purification using anti-FLAG and anti-HA affinity gel in high salt and detergent buffer. The information has been clarified in In Vitro Deubiquitination Assay (Supplemental Methods).
Reviewer #2 -Global analysis of human ubiquitin-modified proteome was used to identify site of the ubiquitination. It is not clarified how this analysis was performed, while the other problem with such in silico method is the reliability. In particular, the identified sites and mutation of these sites are not preventing totally the ubiquitination status of CEP131 (Fig. 4F). Another approach could be to perform Masspect to cnfirm or identify new sites.
Authors: Ubiquitination sites were identified by SILAC-based quantitative analysis of human ubiquitin-modified proteome 3 . Detailed information on 19,000 diGly-modified lysine residues within 5000 proteins including CEP131 is shown in Table S2 of the reference 3 .
Although K254R did not totally abolish CEP131 ubiquitination, this mutation rendered CEP131 resistant to USP9X deubiqutination ( Figure 4F) and the half-life of CEP131/K254R was not affected upon USP9X depletion ( Figure 4G and Figure S5F), indicating that K254 residue of CEP131 is the major poly-ubiquitin targeting site opposed by USP9X. It is possible that other ubiquitin conjugating sites exist on CEP131, which are catalyzed or opposed by other E3 ligases or deubiquitinases, respectively.
In responding to the reviewer"s suggestion, we have employed mass spectrometry to analyze ubiquitin conjugation sites on CEP131 and found two CEP131 peptides carrying ubiquitin-modified sites at lysine residues 254 and 714 ( Figure S5E).
Reviewer #2 -The data presented in Fig. 5 is convincing however, it is essential to include live cell imaging showing chromosomal instability in overexpressed and SiRNA 12 treated regulated cells.
Authors: To comply with the reviewer"s requests, we have provided live cell imaging in Figure S7 and Supplementary Movie 1.
Reviewer #2 -A table is necessary to summarize the percentage of "Different" chromosomal instability phenotypes that occurs in overexpressed and down regulated cells.
Authors: Such a table has been provided in Figure 5D.
Reviewer Authors: Cell cycle profiles but not cell survivals were examined in Figure 2E, original Figure 5A and original Figure 5C. Our experiments indicate that USP9X/CEP131-promoted breast cancer cell growth ( Figure 7) is not through cell cycle control, as either overexpression or knockdown of USP9X/CEP131 had minimal effect on cell cycle progression ( Figure S6A and Figure 5C). Since centrosome dysregulation-associated mitotic defects could result in genome instability and cell apoptosis 4, 5, 6 , we examined whether USP9X-promoted CEP131 stabilization plays a role in genome instability and cell apoptosis. Indeed, we demonstrated either USP9X or CEP131 depletion resulted in markedly accumulation of H2AX ( Figure S9A) and severe apoptosis of MCF-7 cells ( Figure S9B). Moreover, USP9X depletion-associated phenotypes could be alleviated by CEP131 overexpression ( Figure S9B). These results indicate that USP9X/CEP131-promoted breast cancer cell survival is through controlling cell apoptosis.  Figures 3A and D, 5C).
Authors: In responding to the reviewer"s concerns, we have performed the following experiments: 1) Control U2OS cells or U2OS cells stably expressing USP9X was transfected with siRNA targeting 3"UTR of USP9X mRNA. Western blotting analysis revealed that USP9X overexpression was able to restore the expression of CEP131 ( Figure 3E) and prolong the half-life of CEP131 ( Figure 3E) in USP9X deficient cells; and 2) Control U2OS cells or U2OS cells stably expressing USP9X was transfected with siRNA targeting 3"UTR of USP9X mRNA, and immunostaining or colony formation analysis revealed that the centrosome amplification defect or growth inhibitory effect induced by USP9X depletion was overcome by USP9X overexpression, respectively ( Figure S6D and Figure S10C). Authors: To comply with the reviewer"s requests, we have examined the localization and abundance of PCM1 after USP9X depletion. The results demonstrated that the centrosomal localization of PCM1 was mildly interrupted ( Figure S4A) and the protein, but not mRNA, level of PCM1 decreased ( Figure S4B) upon USP9X depletion, indicating that PCM1 is a potential substrate of USP9X. However, we noted that, in USP9X-deficient cells, CEP131 could be effectively recruited to centrosome ( Figure   15 S4C), suggesting that mild disruption of PCM1 centrosomal localization associated with USP9X depletion has limited effect on CEP131 recruitment. Nevertheless, we agree that how PCM1 contributes to USP9X-regulated centrosome biogenesis remains to be investigated in the future.
To further exclude the possibility that CEP131 or PCM1 dis-localization from centrosome is indirectly affected by other centrosome components, we examined the localization and expression level of Pericentrin and CEP290, both of which were critical satellite components and reported to be essential for the centrosomal restriction of PCM1 and CEP131 1 . The results in Figure S4A and Figure S4B indicate that the localization and abundance of these proteins were unaffected upon USP9X knockdown.
These observations, together with our findings that USP9X physiologically interacts with CEP131 and promotes CEP131 deubiquitination in vitro and in vivo, support a notion that the link between USP9X and CEP131 is direct and USP9X depletion-associated loss of CEP131 from centrosome is not a consequence of loss of general satellite proteins. Authors: In the second dataset of "Ma Breast 4", the second group labeled as number 2 (original Figure 6A) represents ductal breast carcinoma in situ of stroma, not epithelia cells. We apologize for the inappropriate representation and have removed the data. Authors: To comply with the reviewer"s requests, we have examined the expression level of PCM1 in breast cancer samples and tumor adjacent tissues. As demonstrated in Figure  S8A and Figure S8B, the protein abundance of PCM1 was elevated in breast cancer and correlated with that of USP9X. This observation is consistent with the finding that PCM1 is a candidate substrate of USP9X. However, the expression of CEP290, another essential satellite protein, was not elevated in breast cancer, nor correlated with that of USP9X ( Figure S8A and Figure S8B). However, the analysis in all sub-categories is not convincing due to sample size limitation. Thus, we did not provide these results. Similarly, patient number enlisted into PPS analysis is less than 200. Thus, we did not present these data.

Reviewer #3: 3) Can the authors provide the parameters used for the Kaplan-Meier analyses and justification for these settings? A) The web-based tool allows four options for survival analysis, DMFS (distant metastasis free survival), RFS (relapse free survival), OS (overall survival) and PPS (post progression survival). Using the USP9X identifier '201099_at', the survival of high and low expressors is significant for the DMFS and RFS survival, but not OS or PPS. Can the authors explain why they have only presented the DMFS and RFS survival? Is the OS and PPS not applicable for USP9X? If so, please indicate why. B) The webtool indicates that the 'JetSet' identifier, which in the case of USP9X is 229573_at is the preferred identifier. Using this id, the difference between high and low expression is either not significant, or significant in the opposite direction (high expression correlates with better survival
B) Although the 'JetSet' identifier of USP9X is the preferred Affymetrix probe, the cohort with USP9X identifier '201099_at' that we presented in the manuscript contains more than two times of patients (3554) compared to that in the cohort with USP9X "JetSet" identifier (1660). Thus, we did not use the preferred identifier.
C) In this study, we demonstrated that USP9X regulates the expression of CEP131 at post-translational level through the deubiquitinase activity of USP9X, while the survival information retrieved from K-M plotter is based on mRNA level. Whether and how the abundance of CEP131 protein is correlated with the outcome for RFS, OS or PPS remain to be investigated.
We appreciate the reviewer for these points. To avoid confusions, these data have been removed in the revision.

Minor Points:
Reviewer #3: The extent of CEP131 destabilization in response to USP9X depletion seems to vary between some panels. Is it possible that different siRNA durations can account for this? Please see: Figure 3A vs. 3D Figure 7B Authors: We admit this and agree with the reviewer"s point. The transfection duration for Figure 3A and Figure S3A were 96 hours, while that in Figure 3D, Figure S3C and Figure  S3D were about 110 hours. Chemically synthesized siRNAs were used in Figure 3A, Figure 3D, Figure S3A and Figure S3D in MCF-7 or U2OS cells, while lentiviruses carrying shRNA stably integrated into MCF-7 cells were used in Figure 7B. We hope the reviewer agree that the overall observation that USP9X depletion was associated with CEP131 destabilization can still hold.
Reviewer #3: Page 7: "elution pattern of CEP131 was largely overlapped with that of PCM1 and USP9X" While the elution pattern of CEP131 largely overlapped with USP9X,

I do not agree that it largely overlaps with PCM1. Additionally, it seems that there is a large peak of protein, as defined by the A280 trace that might elute around fraction 17. If so, it is less compelling to see multiple proteins in this fraction. This panel should be labeled more clearly to correlate the western blot to the elution profile.
Authors: We agree with the reviewer"s point that the elution pattern of CEP131 largely overlapped with that of USP9X, but not with that of PCM1, and thus we have modified the statement (the last paragraph, Page 8) in the revision. In addition, we have carefully checked the fraction labels of FPLC elution profiles and Western blotting. The counting for fraction number in silver staining or Western blotting began when samples were collected, which is different from that in A280 trace. The largest peak of protein fraction is the excess 3 × FLAG peptides (original Figure 1E, now shown as Figure S1B).

Reviewer #3: Page 9: In the domain mapping studies, the authors show that the N-terminus of USP9X is necessary for the interaction with CEP131 but do not show it is sufficient (as they do for the CEP131 domain mapping). Does the N-terminal domain alone interact with CEP131?
Authors: Our data indicate that the middle region, not the N-terminus, of USP9X is necessary for the interaction with CEP131. As demonstrated in the lower panel of Figure  1E and in Figure 1F, the middle region of USP9X (USP9X/M) is sufficient for USP9X interaction with CEP131.

Reviewer #3: "USP9X truncation mutants revealed a high affinity between CEP131 and USP9X/M" Statements of affinity cannot be made from these experiments. Please change text to reflect that this is a non-quantitative experiment.
Authors: The text has been modified in the revision.

Reviewer #3: "demonstrated CEP131 N-terminal region directly interacts with USP9X/M" Is there a reason the authors used USP9X purified from mammalian cells? USP9X could interact another co-purifying protein that could bridge the interaction between USP9X and CEP131. Without performing experiments with purely bacterially (or at least non-mammalian) expressed protein the authors cannot claim a direct interaction.
Authors: To address the reviewer"s concern, USP9X deletion mutants purified from insect Sf9 cells have been used to repeat the pull-down experiments and the data has been provided in Figure 1F and Figure 1G.
Reviewer #3: Page 10: The data in Figure 2B does not match the description in the text. 19

The authors suggest that the co-localization between USP9X and Centrin is evident in S/G2, but less so in G1. However, their images are labeled G1/S and G2 so it is not possible to evaluate this claim. Can the authors provide images of G1 cells? Cyclin A staining should discriminate between G1 and S/G2.
Authors: The images of cells in G 1 and S phases have been provided in Figure S2A.
Reviewer #3: Although PCM1 and CEP131 proteins are known to reside in centriolar satellites, very little non-centrosomal staining is apparent for these proteins. USP9X does not co-localize with the detectable satellite staining of PCM1. Can the authors comment on this observation?
After contrast adjustment of the images in Figure 2C, the overlapping signal for PCM1 and USP9X is almost as strong as that of CEP131 and USP9X. To further address the reviewer"s concern, we performed immunostainings with antibodies against USP9X and PCM1, CEP290 or Pericentrin and demonstrated that USP9X is co-localized with satellite staining of these satellite components ( Figure S4A).  Figure 1A). Does the abundance of these proteins decrease by western plot at later time points?
Authors: According to the reviewer"s suggestion, a quantification of multiple westerns has been provided in Figure 2E and Figure S2C and the results indicate that USP9X and CEP131 in these cells change at a similar pace except 2 hours after synchronization, a time point at which the abundance of CEP131, but USP9X, dramatically increased. These results imply that CEP131 could be controlled by other factors at the initial S phase of the cell cycle. Accordingly, We have modified the relevant text.
Similar to CEP131 1 , USP9X is redistributed and exported from centrosome ( Figure 2B and Figure S2B), but the level of these proteins did not decrease, as demonstrated in Figure 2E and Figure S2D. To clearly display the centrosomal signal of USP9X, we 20 performed immunofluorescent assay using antibodies against USP9X in high dilution (1:500). Cytoplasmic staining of USP9X in mitotic cells was weak but could still be detected ( Figure 2B and Figure S2B). Western blotting analysis indicated that the abundance of CEP131 and USP9X deceased at later time points, when cells left mitosis and proceeded into a new cycle ( Figure S2D).
Reviewer #3: Additionally, time of interaction between USP9X and CEP131 in Figure 2F does not seem to correlate with the initial increase in CEP131 abundance in Figure 2E.
Authors: As stated earlier, CEP131 abundance could be regulated by other factors especially in initial S phase (two hours post synchronization) of the cell cycle.
Reviewer #3: Page 12: Please provide biological replicates for the quantification in Figure 3D.
Authors: The information has been provided in Figure 3D, Figure S3C, and Figure S3D. Authors: As demonstrated in Figure S4A, the centrosomal localization of PCM1 was indeed mildly disrupted, while the centrosomal localization of CEP290 and Pericentrin was essentially unchanged in USP9X deficient cells.
Reviewer #3: Figure 3F. Please provide quantification for this. The CEP131 abundance in MDA-MB-231, ECC1 and Ishkawa cells look reasonably similar while having noticeably different levels of USP9X. Perhaps a correlation plot would be more convincing.
Authors: The quantification and correlation on these proteins have been provided in Figure 3G. Authors: FLAG-tagged USP9X is indicated in Figure 4D. Based on the reviewer"s suggestion, we have used higher concentration of doxycycline to induce C1566S 21 expression and the corresponding deubiquitination assay has been re-perfomed as shown in Figure S5C.
Reviewer #3: Page 15: The labels on Figure 4F are confusing. The bottom panel is labeled CEP131/K504, yet CEP131/K254R is indicated as being present in one of the lanes.
Authors: Figure 4F has been carefully relabeled.
Reviewer #3: Please include biological replicates for the quantification of Figure 4G.
Authors: The information has been provided in Figure 4G and Figure S5F Reviewer #3: Page 16: Do the FACS profiles correspond to the cells analyzed in the immunofluorescence panels in Figure 5A and C? If so, the profiles look like mostly G1

cells. Can the authors provide A) a FACS profile of cycling cells to use for a comparison B) another method, such as cyclin A staining or western blot to demonstrate the cells are in the expected cell cycle stage.
Authors: FACS profiles in original Figure 5A corresponded to unsychronized cells, not cells analyzed in the immunofluorescence panels. In the revision, new FACS profiles and Western blotting analysis of the cells used in the immunofluorescence panels have been shown ( Figure S6A), and FACS profiles of control and HU treated unsynchronized cells corresponding to the immunofluorescence panels have been provided in Figure 5C. Authors: Western blots have been provided in Figure 5B, Figure 5C, and other Figures as suggested.
Reviewer #3: Page 18: Can Figure 6C be rearranged so all normal and all tumor lines are together?
Authors: Since the samples we used were not paired from the same patients, to avoid confusing, we grouped the samples as shown in Figure 6C and run them on separate gels to display the different abundances of these proteins in breast cancer and normal breast tissues. 22 Reviewer #3: Page 19: In Figure 7C, shRNA against CEP131 results in a weaker colony formation phenotype that using shRNA against USP9X. However, the abundance of CEP131 is lower using shRNA against CEP131 than in cells using USP9X shRNA. This is not entirely consistent with the model where USP9X regulates CEP131 stability.
Authors: In our opinion, it is hard to compare the effect of USP9X depletion versus that of CEP131 depletion in Figure 7A and 7B, because: 1) the initial cell numbers used in live cell counting or colony formation assays were not the same; and 2) the efficiency of USP9X knockdown versus CEP131 knockdown would, inevitably, differ. Similarly, the discrepancy of the abundance of CEP131 in CEP131 knockdown cells versus USP9X knockdown cells could possibly come from the disparity of shRNA transfection/knockdown efficiency; 3) USP9X also targets substrates other than CEP131.

Reviewer #3: Perhaps some of the data that is replicated in different cell lines can be presented in Supplementary information.
Authors: To comply with the reviewer"s suggestion, data in Figure 3A, Figure 3D, Figure  4C and Figure 5B have been moved to Supplementary information and shown as Figure  S3A, Figure S3D, Figure S5B and Figure S6B, respectively.
Reviewer #3: page7: excessive FLAG peptides: excess FLAG peptide(s) centriolar satellite protein CEP131 was also identified: were also identified Review on the article "The X-linked Deubiquitinase USP9X Regulates Centrosome Duplication and Promotes Breast Carcinogenesis" In this paper, USP9x is found to be a centrosome component that deubiquitinates its partner, CEP131, shown elsewhere to regulate centrosomal localisation of CDK2. USP9X and CEP131 are also overexpressed in breast cancer and it is suggested that USP9X promotes carcinogenesis through stabilizing CEP131. The interaction between USP9X and CEP131 is very well supported by affinity purification and mass spectrometry, as well as by co-immunoprecipitation in cell lines. Importantly, the co-IP experiments on 4 different cell lines give consistent results and the interaction sites of USP9X and CEP131 are properly validated. The effect of USP9X overexpression on centrosome amplification and chromosome stability and the co-dependency of USP9X and CEP131 on these process are well demonstrated as is the mediation of this effect through CDK2 as previously shown.
In conclusion, this this is a valuable article but still requires some revision before publication. Major points: 1) page 10: centrosomal and cytoplasmic localisation of USP9X The immunofluorescence data presented suggests that Usp9X localisation is restricted to the centrosomes to the same extent as ƴ-tubulin, centrin, PCM1 or CEP131. This creates an illusion that this protein is localised exclusively at the centrosomes. However . The authors should comment on this issue. However, further work is still required to show that the mainly cytoplasmic Usp9X relocates to the centrosome in G1/S. This requires localisation of USP9X by immunofluorescence in G1. It would be very surprising and unlikely, however, that all cytoplasmic Usp9X disappears in S/G2-M and relocates to the centrosomes. Since there is no indication, either in Results section or Methods, which anti-Usp9X antibody was used, we do not know how the authors' findings relate to the localisation in the above mentioned papers where a diffuse cytoplasmic localisation was reported.
2) page 14, line 17: rescue of CEP131 level by catalytically inactive USP9X The authors state that "downregulation of CEP131 in these [Usp9X knockout] cells could be reverted by forced expression of wild-type USP9X, but not USP9X mutant (USP7/C1566S)". Here there is a minor problem -"USP7/C1566S" should be corrected to "USP9X/C1566S" (this seems to be a typo graphical error as USP7 is a totally different DUB According to these guidelines, when two samples differ less than 5-fold in RNA levels it is essential to use multiple reference genes. This is critical for the experiment at Figure 6B, where USP9X expression level was determined in normal and malignant tissue and where the average difference seems to be less than 5-fold. The use of beta-actin as a reference gene should be avoided, because it is well described that beta-actin expression is deregulated in many cancer types, including breast cancer and MCF-7 cells (reviewed in Guo et al, 2013). 5) page 24: centrosome amplification in tumors It would be helpful if the authors could show the centrosome amplification in the tumors grown in athymic mice and the reduction of centrosome amplification upon Usp9X knock-down in these tumors. This could support their firm statement made at page 24, line 13: "USP9X promotes breast carcinogenesis, and that USP9X does so, through stabilizing the centriolar satellite protein CEP131". There are other USP9X substrates involved in carcinogenesis (reviewed in Murtaza et al., 2015), which were not investigated by the authors, and their involvement cannot be excluded. The overexpression of CEP131 just partially restores the growth of USP9x knocked-down tumors ( Figure 7E), indicating that other pathways might be involved as well.
Minor remarks: -the year of publication is missing at reference 57 in the reference list -"in vitro ubiquitination assay" is mentioned at page 16, though it should be "in vitro deubiquitination assay" as it can be deduced from the figure legend and methods. This should be corrected because it is misleading.
-pg. 17, 1st line: "able to remove the ubiquitination" should be corrected to "able to remove ubiquitins".
Reviewer #2 (Remarks to the Author): The revised version of the manuscript: "The X-linked Deubiquitinase USP9X Regulates Centrosome Duplication and Promotes Breast Carcinogenesis" has been completely answering all of my questions and concern.
Reviewer #3 (Remarks to the Author): The new experiments included in the revised version of the manuscript seem reasonable and well controlled. However, I still remain only partly convinced of their model and for this reason the lasting impact of this work remains unclear. I note the following issues: 1) The interpretation of the localization data aren't correct. PCM1 and CEP131 co-localize with centrin and show almost no satellite distribution. Although the is not a major issue, it is disturbing that they don't see proper satellites (in U2OS cells).
2) The evidence for USP9X regulating CEP131 is convincing, but unfortunately they also see effects on the stability of PCM1. The authors also observe a good correlation between PCM1 and USP9X in breast cancer cells (better than CEP131!). However, they basically ignore this new data. I think this complicates their interpretation and this needs to be fully integrated in their model. Authors: We thank for the reviewer's comments. To clearly display the centrosomal signal of USP9X, we performed immunofluorescent assays using antibodies against USP9X in high dilution (1:500). We believe that could be the reason why USP9X showed weak staining in the cytoplasm. To address the reviewer's concern, we have repeated the immunofluorescent assays with more concentrated antibodies against USP9X (1:200). Immunofluorescence analysis indicated that USP9X displayed extensive cytoplasmic, weak nuclear and evident centrosomal localization, the observation of which is consistent with the cellular distribution pattern of GFP tagged USP9X. The results have been provided in Figure 2A, Figure 2D, Figure 3F, Figure 4H and Figure S2A to replace the previous ones. Moreover, we demonstrated that the centrosomal signal of USP9X was severely abolished upon USP9X depletion ( Figure 3F), indicating that the centrosomal staining of USP9X was not resulted from antibodies' non-specific binding.
In addition, more concentrated USP9X antibodies (1:200) were used to re-perform the 2 immunofluorescent assays with cells in different cell cycles. The new data in Figure 2B and Figure S2B indicated that in G 1 , S and G 2 phases of the cell cycle USP9X displayed positive staining in both cytoplasm and centrosome, while the centrosomal localization of USP9X was largely diminished in metaphase.
The anti-USP9X antibody information (manufacture and catalogue No.) has been provided in Supplemental Methods. Authors: According to the reviewer' suggestion, we have changed the statement to "overexpression of USP9X mutant could only moderately revert the CEP131 downregulation". In addition, "USP7/C1566S" has been corrected to "USP9X/C1566S".
The rescue effect observed with the catalytically inactive USP9X may come from: 1) residual activity of the mutant in vivo, non-catalytic roles of USP9X or a combination of both; 2) excessive amounts of USP9X/C1566S might bind CEP131, blocking it from accessing its unidentified E3 ligase. ubiquitin linkages are mainly involved in non-proteasomal pathways as a scaffolding modification in signal transduction 1 . Given that the physical association of USP9X with CEP131 was detected primarily in S and G 2 phases of the cell cycle ( Figure 2F), we did not investigate K11 ubiquitin linkages that are preferentially produced during mitosis and early G 1 3, 6 . Thereby, we chose to investigate K29 and K48 polyubiquitin chains with K63 as a likely control.
Consistent with our observations that USP9X deubiquitinates K48 poly-ubiquitylated CEP131, USP9X has been reported to efficiently remove degradative K48-linked polyubiquitin chains on MCL1 7 and XIAP 8 . To clearly reveal the exact nature of CEP131 ubiquitination and USP9X specificity, we performed deubiquitination assays with ubiquitin chain type specific antibodies and revealed that K48 ubiquitin chains on CEP131, but not K63 linkages, was opposed by USP9X ( Figure S5E). This observation is in favor of our previous findings with ubiquitin mutants K48-only and K63-only.
Considering that the preference of USP9X on different types of ubiquitin linkages has been reported, we assume that USP9X opposes specific ubiquitin linkages in a substrate or context dependent manner.
We agree with the reviewer's point that other types of polyubiquitins might be involved in CEP131 ubiquitination. It will be interesting to investigate whether CEP131 is subjected to K6, K11, K27 and K33 polyubiquitin conjugates，whether these types of ubiquitin linkages could be removed by USP9X, and how these modifications impact on the molecular behaviors and functionalities of CEP131.
All of the information has been incorporated into the Results or Discussion sections (paragraph 2, page 30) accordingly in the revision. Authors: In responding to the reviewer's comments, we have re-performed quantitative real-time PCR to determine USP9X expression level in normal and malignant tissues with GAPDH and PUM1 (a normalizer used in breast cancer) 9, 10 as reference genes. The results still hold and the new data has been provided in Figure 6B to replace the previous one with -actin as a normalizer.  Figure 7E), indicating that other pathways might be involved as well.
Authors: To comply with the reviewer's requests, we have examined the centrosome numbers in cultured xenograft tumors with immunofluorescent assays. Since puromycin resistant gene is carried by the integrated shRNA lentivirus, tumor cells were enriched by puromycin selection during culture post isolating by trypsinization. The results in Figure  7F indicated that mild centrosome amplification could be observed in control tumors, while the percentage of cells with centrosome amplification was indeed reduced in USP9X knocked-down tumors. Furthermore, we found that CEP131 gain of function overrode the effects induced by USP9X depletion ( Figure 7F). These observations support the argument that USP9X promotes breast carcinogenesis through stabilizing the centriolar satellite protein CEP131.
We agree with the reviewer's point that other pathways might be involved in USP9X-promoted tumorigenesis and their involvement needs to be considered. To clarify this issue, we have already performed a series of experiments with other substrates of 5 USP9X: 1) In agreement with previous reports 7, 11,12,13 , we found that USP9X knockdown did result in decreased expression of ITCH ( Figure S10A), an E3 ligase involved in carcinogenesis 12,14 , and of MCL1, an anti-apoptotic regulator implicated in cancer 15,16,17,18 . However, USP9X, but not CEP131, could be co-immunoprecipitated by ITCH or MCL1 ( Figure S10B); 2) Colony formation assays demonstrated that while overexpression of ITCH or MCL1 could rescue the growth inhibitory phenotype resulted from USP9X depletion to certain extent as CEP131 did, simultaneous expression of CEP131 and ITCH or MCL1 showed an additive effect ( Figure S10C); 3) while loss of USP9X-associated defects of centrosome amplification could be rescued by CEP131 overexpression ( Figure 5C and Figure S10D), overexpression of ITCH or MCL1 could not rescue the phenotype ( Figure S10D); and 4) knockdown of either ITCH or MCL1 had no effect on centrosomal localization of USP9X and CEP131 ( Figure S10E). These results suggest that CEP131 functions cooperatively with but independently of other USP9X substrates in USP9X-promoted breast cancer cell survival, and also provided an explanation for why overexpression of CEP131 only partially restores the growth of USP9X knocked-down tumors (paragraph 2, page 23).
Minor remarks: Reviewer #1: the year of publication is missing at reference 57 in the reference list.
Authors: The missing information has been added in the revision and the reference is shown as No.22 in the revision.
Reviewer #1: "in vitro ubiquitination assay" is mentioned at page 16, though it should be "in vitro deubiquitination assay" as it can be deduced from the figure legend and methods. This should be corrected because it is misleading.
Authors: It has been corrected.
Authors: It has been corrected.
Authors: It has been corrected. 6 Response to Reviewer #3's comments-Reviewer #3: 1) The interpretation of the localization data aren't correct. PCM1 and CEP131 co-localize with centrin and show almost no satellite distribution. Although the is not a major issue, it is disturbing that they don't see proper satellites (in U2OS cells).
Authors: We appreciate the reviewer for this comment. In our initial investigation, we performed immunofluorescent assay using antibodies against PCM1 and CEP131 in high dilution (1:500 and 1:1000, respectively). We speculate that this could be the reason why satellite information of CEP131 and PCM1 was missed. To avoid confusion, we have repeated the immunofluorescent assays with more concentrated antibodies against PCM1 (1:200) and CEP131 (1:200). Immunofluorescence analysis indicated that PCM1 and CEP131 displayed evident satellite stainings, and the new data has been provided to replace the previous results in Figure 2C, Figure 2D, Figure 3F, Figure S4C and Figure  S10E.
Reviewer #3: 2) The evidence for USP9X regulating CEP131 is convincing, but unfortunately they also see effects on the stability of PCM1. The authors also observe a good correlation between PCM1 and USP9X in breast cancer cells (better than CEP131!). However, they basically ignore this new data. I think this complicates their interpretation and this needs to be fully integrated in their model.
Authors: Indeed, we demonstrated that PCM1 is a potential substrate of USP9X ( Figure  S4A and S4B). Specifically, the centrosomal localization of PCM1 was mildly interrupted ( Figure S4A), and the protein, but not mRNA, expression level of PCM1 decreased ( Figure S4B) in USP9X deficient cells. Although we observed a better correlation between PCM1 and USP9X than that between CEP131 and USP9X in breast cancer samples, the effect of USP9X depletion on the expression of PCM1 was not as dramatic as that of CEP131 ( Figure S4B), suggesting that PCM1 is a potential, but might not be a major substrate of USP9X in centrosome. These observations likely explain why CEP131 overexpression could not fully compensate centrosomal biogenesis defects induced by USP9X depletion.
Since the abundance and localization of PCM1 seem to be regulated by USP9X, we are curious that whether the regulation of CEP131 by USP9X is a secondary effect of 7 USP9X-promoted PCM1 stabilization. To test this, we performed the following experiments and found: 1) In USP9X deficient cells, FLAG-tagged CEP131 could be still effectively recruited to centrosome ( Figure S4C), suggesting that mildly disrupted centrosomal localization of PCM1 associated with USP9X depletion has limited effect on CEP131 recruitment; 2) Severe loss of PCM1 indeed impaired CEP131 centrosomal localization as reported 19 , while mild loss of PCM1 (approximately 40 to 50 percentage left) failed to do so ( Figure S4D); 3) The expression level of CEP131 was essentially not altered upon PCM1 knockdown ( Figure S4D). Combining these observations and the findings that USP9X directly interacts with CEP131 and opposes its poly-ubiquitin linkages in vitro, we get the conclusion that the effect of USP9X on CEP131 stabilization is attributed to the interplay between these two molecules but not resulting from other potential USP9X centrosomal interactors such as PCM1, and USP9X-regulated PCM1 stabilization on centrosome activity, if it does so, seems to be independent of USP9X-promoted CEP131 stabilization. We believe that understanding how USP9X promotes PCM1 stabilization and whether/how PCM1 contributes to USP9X regulated centrosome biogenesis will be helpful in interpreting the function of USP9X in centrosome.
All of the information has been incorporated into the Results or Discussion sections (paragraph 1, page 29) accordingly in the revision.

Reviewer #3:
3) USP9X is already known to affect cancer cells (oddly as both as an oncogene and tumour suppressor...). Therefore, the link between USP9X and CEP131 in the breast cancer model has to be air-tight since the novelty of USP9X is not really there. Additionally, the rescue experiments in this section of the manuscript are not very convincing which makes me question the biological validity of their interpretation.
Authors: As the reviewer mentioned, the conflicting conclusions surrounding USP9X function as either a tumor suppressor or an oncogene indicate that USP9X plays a complex role in carcinogenesis. Specifically, several reports have portrayed USP9X as an oncogene in prostate cancer 20 , lymphomas 7 and colorectal carcinoma 21 , while two studies suggest that interfering with USP9X expression promotes a more rapid onset of PDAC (pancreatic ductal adenocarcinomas) in genetic mouse model 12,22 . However, a recent study pointed that USP9X is an oncogene in the context of established PDAC with xenograft model 7 , which is consistent with findings in other neoplasms. In this regard, USP9X may parallel the behavior of TGF-β observed in some cancers, where TGF-β behaves as a tumor suppressor during early stages, but as an oncogene in later stages 23,24 . Whether this paradigm is applicable for USP9X in tumorigenesis needs to be further characterized. 8 To further establish the link between USP9X and CEP131 in breast carcinogenesis, we isolated breast cancer cells from frozen xenograft tumors, cultured the cells under puromycin selection (puromycin resistant gene together with shRNA cassette carried by the shRNA expressing lentivirus has been integrated into the genome of tumor cells) and examined the centrosome numbers with immunofluorescent assays. The results in Figure  7F indicated that mild centrosome amplification could be observed in control tumors, while the percentage of cells with centrosome amplification was indeed reduced in USP9X knocked-down tumors. Furthermore, we found that CEP131 gain of function overrode the effects induced by USP9X depletion. These observations further strengthen the argument that USP9X-promoted CEP131 stabilization plays a role of importance in breast carcinogenesis.
The novelty of our work is based on the weight of following observations: 1) although USP9X has been implicated in several pathological states including various malignancies and centrosome-associated CEP family proteins are considered to be potent tumor suppressors or oncogenes, mechanistic insights into the role of USP9X and CEP proteins in cancer development and progression, in particular, breast carcinogenesis, remain to be investigated; and 2) defective centrosome duplication is implicated in microcephaly and primordial dwarfism as well as various ciliopathies and cancers, yet how the centrosome biogenesis is regulated remains poorly understood. Through a body of work, we first discovered that USP9X is an integral component of the centrosome and serves to specifically stabilize the centriolar satellite protein CEP131. Second, we revealed that CEP131-regulated centrosomal localization of CDK2, a cyclin-dependent kinase with an established role in centrosome duplication, is required for USP9X-promoted centrosome biogenesis. Third, we demonstrated that USP9X, in doing so, impacts on chromosome stability and eventually promotes breast carcinogenesis.  34,35,36 . We envision the same paradigm could be true for USP9X, in which USP9X targets multiple substrates involved in distinct or even converged signaling pathways and eventually impacts on tumorigenesis. This likely explains why expression of CEP131 only partially rescued the phenotype of depleting USP9X.
To test the above hypothesis, we have performed a series of experiments with other substrates of USP9X: 1) In agreement with previous reports 7, 11, 12, 13 , we found that USP9X knockdown did result in decreased expression of ITCH ( Figure S10A), an E3 ligase involved in carcinogenesis 12,14 , and of MCL1, an anti-apoptotic regulator