The bromodomain containing protein BRD-9 orchestrates RAD51–RAD54 complex formation and regulates homologous recombination-mediated repair

Homologous recombination (HR) is important for error-free DNA double strand break repair and maintenance of genomic stability. However, upregulated HR is also used by cancer cells to promote therapeutic resistance. Therefore, inducing HR deficiency (HRD) is a viable strategy to sensitize HR proficient cancers to DNA targeted therapies in order to overcome therapeutic resistance. A bromodomain containing protein, BRD9, was previously reported to regulate chromatin remodeling and transcription. Here, we discover that following DNA damage, the bromodomain of BRD9 binds acetylated K515 on RAD54 and facilitates RAD54’s interaction with RAD51, which is essential for HR. BRD9 is overexpressed in ovarian cancer and depleting BRD9 sensitizes cancer cells to olaparib and cisplatin. In addition, inhibitor of BRD9, I-BRD9, acts synergistically with olaparib in HR-proficient cancer cells. Overall, our results elucidate a role for BRD9 in HR and identify BRD9 as a potential therapeutic target to promote synthetic lethality and overcome chemoresistance.

In the current study entitled "Bromodomain Containing Protein BRD-9 Orchestrate the RAD51-RAD54 Complex and Coordinately Regulate Homologous Recombination" Zhou and colleagues examine the potential role of BRD9 in DNA repair. Studying the underlying mechanisms associated with DNA repair remains pertinent, given their fundamental importance in many aspects of biology. Moreover, these mechanisms frequently become perturbed in diseases including multiple cancers; likely contributing to disease development and also presenting therapeutic vulnerabilities. In this study the authors suggest that BRD9 plays and important role in DNA repair through interactions with the key DNA repair proteins RAD51 and RAD54. Based on these reports, the authors further suggest that targeting BRD9 function may provide an opportunity to achieve improved therapeutic responses in tumours treated with PARP inhibitors. Although the findings of this study are interesting, there are currently a number of shortcomings with the experimentation that should be addressed in order to strengthen the findings.
Major Points: 1. The authors begin by examining mutational profiles in published TCGA cancer genomics data. They suggest that BRD9 together with some other bromodomain containing proteins (ASH1L, BRWD3, KMT2A, ZMYND8 and SMARCA4) are mutated with greater frequency in Signature 3 ovarian cancers. It is important to note here that the multiple distinct cancer mutational signatures that have been described have different mutational frequencies/preferences. As such it remains important to delineate whether the observed mutational rate for these genes is simply an indirect consequence of their underlying DNA sequence. How or whether the underlying DNA sequence of these genes differ from other bromodomain containing genes, and other genes in general, may explain the observed differences. Moreover, it could strengthen (or rule out) the suggestion that mutation of these genes may be causally linked to oncogenesis in these tumours. 2. The experimentation used to suggest that BRD9 is specifically important (compared to most other bromodomain containing proteins) for homologous recombination (HR) are weak. For example, none of the shRNA sequences used in Figure 1B are validated with regard to their knockdown efficiency. It is impossible to conclude that BRD9 is more important that many other bromodomain containing proteins in these assays without providing evidence that all shRNAs used in these assays work effectively. Moreover, in relation to these assays no positive control shRNAs (ie. those knocking down expression of known HR regulators) are used to support the robustness of the data presented in Fig. 1a. 3. The authors suggest that BRD9 bromodomain inhibition using a published small-molecule inhibitor reduces HR levels in cells. It is important to note here that this, and other BRD9 targeting small-molecules have been reported to induce broad transcriptional changes. As such it's hard to conclude that the effects observed here are a direct consequence of bromodomain inhibition and not related to altered global transcriptional dynamics. Moreover, the doses used (10uM and 20uM) are extremely high and therefore potentially leading to additional off-target effects. 4. The immunofluorescence (IF) images and experimentation presented in Figures 1/S1, 2/S2 have no positive controls (ie. knockdown of known regulators of HR) and in several instances only use a single BRD9 shRNAs. These experiments need to be strengthened significantly to support the authors suggestions. 5. The resolution of the images presented in Fig. S2A is very poor making it nearly impossible to see the underlying numbers. Moreover, the cell cycle profiles themselves do not match the authors assertion that there is no significant effect on cell cycle profile following BRD9 knockdown. In fact, it is quite clear to see that there are proportionally fewer cells in G2/M (and possibly S) following BRD9 depletion. This is clearly an important point given that these DNA double-strand break repair mechanisms preferentially occur at certain points throughout the cell cycle. Shifting cell cycle dynamics as appears to be the case here could alter these mechanisms and their regulation indirectly. 6. In Fig. S2B the authors present Western blot data that they claim is representative of the subcellular "chromatin" fraction. However, it cannot be concluded based on the presented data that this is in fact the case. The authors have not run cytoplasmic and/or nucleoplasmic fractions on these blots to demonstrate that they have in fact adequately fractionated protein samples. This should be done and control Western blots for proteins present in non-overlapping fractions should be included to support the validity of the claims. 7. In Fig. 4 the authors over-express GCN5 and PCAF in cells and demonstrate that acetylation levels of RAD54 increases in this setting. However, no conceptual rationale was presented as to why these particular enzymes were chosen. Are these enzymes actually the primary (and biologically relevant) mediators of RAD54 acetylation? Its hard to conclude based on the presented data (and lack of rationale) whether this is the case.
Minor Points: 1. The grammar throughout the manuscript is poor and should be improved to make the text easier to read/follow. 2. The authors focus their computational analysis of cancer mutational signatures in Figure 1 exclusively on ovarian cancer. Many other cancers, in particular BRCA1/2 mutated breast cancers, have overlapping mutational signatures (ie. Signature 3). It would be interesting to note whether or not the observations made here in ovarian cancer carry over to other Signature 3 tumours; or whether they are in fact specific to ovarian cancer (also see Major Point 1).
In this manuscript, the authors report that BRD9 participates in HR repair by facilitating the removal of RAD51 from the sites of DNA damage. The authors found that RAD51 mediated the recruitment of BRD9 to DNA lesions. The bromodomain of BRD9 recognized acetylated K515 of RAD54, and mediated the recruitment of RAD54 to the sites of DNA damage, which replaced RAD51 for the completion of HR repair. Moreover, BRD9 is overexpressed in ovarian cancer, and inhibition of BRD9 sensitized ovarian cancer cells to PARP inhibitor treatment. Overall, the authors reveal a novel molecular mechanism in DNA damage repair field, which may impact for the clinical treatment for ovarian cancer patients. The specific points listed below should be addressed before publication.: 1. The figure legend should be revised by including relatively detailed information, so that readers may easily understand the results of each assay.
Response: Thank you very much for your suggestion. We have added additional details to the figure legends to improve readability and facilitate comprehension of the results.
2. The authors need to provide the evidence or discuss the specificity of the acetylated K515 that is recognized by the bromodomain of BRD9, since other members in the BRD family may not bind RAD54.
Response: Thank you for this suggestion. We have addressed this comment by screening other bromodomain containing proteins such as BAZ1B, BAZ2A, CECR2, TRIM66, SMARCA4, BRD7 and BRD9 for their ability to interact with RAD54 K515 peptide. As displayed in Fig.S4B, we found that none of these other bromodomain containing proteins could interact with RAD54 K515, suggesting that the K515 RAD54 acetylation site may be specific for binding to the bromodomain of BRD9.
3. The author only examined GCN5 and PCAF. However, other acetyltransferases are known to play key roles in DNA damage repair, such as TIP60, p300 and MOF. The authors need to examine the possibilities that other enzymes may be also involved in the acetylation of RAD54.
Response: Thank you very much for your suggestion. As requested, we screened whether other key acetyltransferase MOF, TIP-60 and P300 can acetylate RAD54. As displayed in Figure.S3E onlyGCN5 and PCAF, and not MOF, TIP60 or P300 acetylate RAD54.

The retention of RAD51 should be examined in cells only expressing the K515A mutant of RAD54 and the bromodomain deletion mutant of BRD9.
Response: We appreciate the suggestion. As requested, we have further examined the retention of RAD51 in cells expressing K515R mutant of RAD54 or bromodomain deletion mutant of BRD9. As shown in Fig.S6A-B, RAD51 retention can be rescued by RAD54-WT but not RAD54-K515R, and BRD9-WT but not BRD9 bromodomain deletion.

Reviewer #2 (Remarks to the Author):
The authors of this manuscript made the interesting observation that mutations of several BRD proteins are associated with the mutation signature 3 in cancer, indicating that these BRD proteins may be involved in HR. Their subsequent experiments confirmed that BRD9 is important for HR. Interestingly, ablation of BRD9 did not alter RAD51 foci but reduced RAD54 foci. Moreover, BRD9 binds RAD51 and is required for the damage induced binding between RAD51 and RAD54. Through a set of careful biochemistry experiments, the authors found that the bromo domain of BRD9 is important for binding RAD54, and the K515 of RAD54 is acetylated after damage and required for BRD4 binding. Deletion or inhibition of the bromo domain of BRD9, or mutation of the K515 of RAD54, increased the sensitivity of cells to PARPi and cisplatin. Loss or inhibition of BRD9 also increased the chemo sensitivity of tumors in mouse xenografts. These results strongly suggest that BRD9 is a regulator of HR that affects chemo sensitivity of tumors. The results of this study (especially the biochemistry data) are compelling, and the conclusion is important. I have a few suggestions to the authors to strengthen the main conclusions. A satisfactorily revised manuscript should be suitable for publication in Nature Communications.
1. In Fig. 2A, merged images of RAD51 and RAD54 foci should be shown. The colocalization of RAD51 and RAD54 looks quite partial from the images.
Response: Thank you for this suggestion. As requested, the merged images of RAD51 and RAD54 foci are now shown. As displayed in figure 2A, RAD51 and RAD54 co-localized following IR. For the WT OVCAR8 cells, about 60% RAD51 can co-localize with RAD54 foci, which is consistent with a previous publication (BLM helicase stimulates the ATPase and chromatin remodeling activities of RAD54,Vivek Sivastava et al.2009.J Cell Sci.). For the BRD9 KD cells, only 5% of RAD51 colocalized with RAD54 foci, suggesting BRD9 plays an important role in RAD51-RAD54 colocalization. Fig. 2G and S2E should be quantified.

The IF data in
Response: As requested, we have quantified the IF foci data in Fig. 2Gand S2E. 3. In Fig. S2C, BRD9i seems to inhibit RAD51 foci formation. Why?

Response:
We quantified the RAD51 foci in Fig S2C. BRD9i did not inhibit and instead increased RAD51 foci formation. We selected high resolution images to represent the data. Fig. S3A, is the acetylation of endogenous RAD54 at K515 detectable by mass spec after IR? Response: As suggested, we immunoprecipitated endogenous RAD54 and performed mass spectrometry analysis. However, the K515 acetylation of endogenous RAD54 were not detected by mass spectrometry. It may be due to technical limitations including the quality and quantity of RAD54 IP sample. Exogenous RAD54 K515 acetylation can be readily detected by mass spectrometry (Fig S3). Fig. 4, can the authors show that BRD9 binds to an acetylated peptide encompassing the K515 of RAD54? The K515R mutant may be defective for BRD9 binding because of a change in the binding surface but not the lack of acetylation. Direct evidence for K515-Ac mediated interaction is important.

In
Response: Thank you for this comment. We synthesized two peptides which encompass the K515 or K515-Ac site of RAD54 and performed pull down assay. As shown in Figure S4A, BRD9 binds with the K515-Ac peptide but not K515 peptide, suggesting that K515-Ac of RAD54 is important for BRD9 binding.
6. In Fig. 4, does knockdown of GCN5, PCAF, DAC6, or HDAC11 affect RAD54 acetylation? The specificity of these enzymes to RAD54 would be strengthened by depletion experiments.

Response:
We utilized shRNA to knockdown GCN5, PCAF, HDAC6 and HDAC11, then transfected RAD54 to detect its acetylation. Decreased GCN5 and PCAF level reduced RAD54 acetylation. On the other hand, knocking down HDAC6 and HDAC11 increased RAD54 acetylation signal. This data is now shown in FigS8. Fig. 5C, the differences between BRD9 levels in cancer cell lines seem to quite small. For example, the BRD9 levels in OVCAR8 and 10 are only 2-3 folds higher than OVCAR7. If BRD9 is only reduced 2-3 folds in OVCAR8/10 or only increased 2-3 folds in OVCAR7, does it significantly increase or decrease PARPi/cisplatin sensitivities?

In
Response: Thank you for this question. For further clarification, we quantified the WB signal in Fig.5A and B. Data showed that BRD9 expression in OVCAR8/10 are 6-7 folds higher than OVCAR7, while BRD9 KD OVCAR8 express the protein at 19-20 folds lower level than WT OVCAR8 cells. We also tested endogenous and exogenous BRD9 in BRD9 overexpressed OVCAR7 cells. Our data showed that exogenous BRD9 expression in OVCAR7 cells is dramatically higher than endogenous to confer resistance to PARPi and Cisplatin (Fig5 I and J). Response: Thank you for this comment. We have assessed the expression of BRD9 in several cell lines and found that the BRD9 expression level is lower in normal cell lines such as MEF and MCF10A compared to cancer cell lines, potentially providing a therapeutic window (Fig.S7C). Indeed, BRD9i did not increase PARPi sensitivity in normal cells (Fig.S4H). Similar findings were also observed in HOSE cells (Fig.S7D).These data suggest that inhibition of BRD9 does not further lead to PARPi sensitivity in normal cells.

Reviewer #3 (Remarks to the Author):
In the current study entitled "Bromodomain Containing Protein BRD-9 Orchestrate the RAD51-RAD54 Complex and Coordinately Regulate Homologous Recombination" Zhou and colleagues examine the potential role of BRD9 in DNA repair. Studying the underlying mechanisms associated with DNA repair remains pertinent, given their fundamental importance in many aspects of biology. Moreover, these mechanisms frequently become perturbed in diseases including multiple cancers; likely contributing to disease development and also presenting therapeutic vulnerabilities. In this study the authors suggest that BRD9 plays and important role in DNA repair through interactions with the key DNA repair proteins RAD51 and RAD54. Based on these reports, the authors further suggest that targeting BRD9 function may provide an opportunity to achieve improved therapeutic responses in tumours treated with PARP inhibitors. Although the findings of this study are interesting, there are currently a number of shortcomings with the experimentation that should be addressed in order to strengthen the findings.
Major Points: 1. The authors begin by examining mutational profiles in published TCGA cancer genomics data. They suggest that BRD9 together with some other bromodomain containing proteins (ASH1L, BRWD3, KMT2A, ZMYND8 and SMARCA4) are mutated with greater frequency in Signature 3 ovarian cancers. It is important to note here that the multiple distinct cancer mutational signatures that have been described have different mutational frequencies/preferences. As such it remains important to delineate whether the observed mutational rate for these genes is simply an indirect consequence of their underlying DNA sequence. How or whether the underlying DNA sequence of these genes differ from other bromodomain containing genes, and other genes in general, may explain the observed differences. Moreover, it could strengthen (or rule out) the suggestion that mutation of these genes may be causally linked to oncogenesis in these tumours.

Response:
Thank you for this comment. In our initial screen we identified mutations in 6 bromodomain containing genes (BRD9, ASH1L, BRWD3, KMT2A, ZMYND8 and SMARCA4) in ovarian cancer to be associated with signature 3 which is known as a homologous recombination deficiency gene signature. We cannot totally rule out the possibility that mutation in these genes could be an indirect consequence of their underlying DNA sequence (i.e. passenger mutations in the context of repair deficiency of a different etiology). However, we still prefer our explanation because of the following reasons.
1. If the Signature 3 has the mutational preferences on the DNA sequence of these genes (BRD9, ASH1L, BRWD3, KMT2A, ZYMND8 and SMARCA4), it might be also found in other Signature 3 cancer types, i.e. BRCA. However, we did not find that. (Refer to the response to your minor points 2); 2. As we know, during DNA replication, HR repair is the only known error-free repair way for cells to guarantee high-fidelity of genome transmission. If there are sequence preferences for HR repair on these bromodomain-containing proteins, it means these genes should be mutated with greater frequency in Signature 3 ovarian cancers. We also didn't found that in ovarian cancer patient database(430 patient in total). And it seems to be not reasonable for the lack of error-free repair way for some specific genes with specific potential sequence in cells. 3. Moreover, BRD9 mutations were associated with functional HR deficiency by the well established DR GFP assay (Fig.1B-C). ZYMND8 was also reportedly involved in DSB repair in recent studies. Therefore, it is highly likely that loss of BRD9 was the source of the HR deficiency and signature 3 in these tumors..

2.
The experimentation used to suggest that BRD9 is specifically important (compared to most other bromodomain containing proteins) for homologous recombination (HR) are weak. For example, none of the shRNA sequences used in Figure 1B are validated with regard to their knockdown efficiency. It is impossible to conclude that BRD9 is more important that many other bromodomain containing proteins in these assays without providing evidence that all shRNAs used in these assays work effectively. Moreover, in relation to these assays no positive control shRNAs (ie. those knocking down expression of known HR regulators) are used to support the robustness of the data presented in Fig. 1a.
Response: Thank you for your comment. We have confirmed the shRNA knockdown efficiency of a majority of the bromodomain containing proteins using RT-PCR(See the following table). We also employed RAD51shRNA and 53BP1 shRNA as positive control, respectively, in  Response: To address the reviewer's question, we depleted BRD9 in cells using shRNA and further treated the cells with BRD9 inhibitor (20µM), and examined the RAD51/RAD54 foci formation and HR. As shown in Fig. S9A-B, BRD9 inhibitor did not further affect RAD51/RAD54 foci formation and HR in BRD9 cells, suggesting there is no off-target effect of BRD9i in HR regulation. We analyzed the RNA-seq data from previous publication and found that both BRD9knockdown and BRD9i treatment did not affect the expression of DDR gene (Sensitivity and engineered resistance of myeloid leukemia cells to BRD 9 inhibition,Anja Hohmann et al.2016.Nature Chem bio.). We also examined RAD51/RAD54 transcription level in these cells. As shown in Fig. S9C-D Response: Thank you for this comment. We are now displaying positive controls (BRCA1 and 53BP1 knockdown) which we believe has strengthened the DR-GFP and IF data in Figure 1B-J,2A and Figure S1A-C. We also employed two shRNAs for the data presented in Figure1  Figure2 and Figure S2. Fig. S2A  Response: Thank you for this comment. We have repeated the cell cycle analysis 3 independent times. The cell cycle profile is now displayed with improved resolution. Moreover, we have quantified the percentage of cells in G1, S, and G2/M phases and provided this data in a color coded bar graph. As displayed in supplementary figure 2A, we do not see evidence of a significant change in the cell cycle profile following BRD9 knockdown.

The resolution of the images presented in
6. In Fig. S2B the authors present Western blot data that they claim is representative of the subcellular "chromatin" fraction. However, it cannot be concluded based on the presented data that this is in fact the case. The authors have not run cytoplasmic and/or nucleoplasmic fractions on these blots to demonstrate that they have in fact adequately fractionated protein samples. This should be done and control Western blots for proteins present in non-overlapping fractions should be included to support the validity of the claims.
Response: Thanks for this suggestion. As suggested we have blotted both the cytoplasmic and nuclear fractions with GAPDH and Histone H3 antibodies, respectively. As displayed in Fig.S2A there is no significant overlap between cytoplasmic and chromatin binding fractions. Fig. 4 the authors over-express GCN5 and PCAF in cells and demonstrate that acetylation levels of RAD54 increases in this setting. However, no conceptual rationale was presented as to why these particular enzymes were chosen. Are these enzymes actually the primary (and biologically relevant) mediators of RAD54 acetylation? Its hard to conclude based on the presented data (and lack of rationale) whether this is the case.

In
Response: Thank you very much for your suggestion. We also screened for other key acetyltransferases like MOF, TIP-60 and P300 to assess if they can acetylate the RAD54 (new Figure.S3E). Data showed that GCN5 and PCAF, but not MO, TIP60 or P300 can mediate RAD54 acetylation.
Minor Points: 1. The grammar throughout the manuscript is poor and should be improved to make the text easier to read/follow. Response: We have edited the manuscript for grammar. Figure 1 exclusively on ovarian cancer. Many other cancers, in particular BRCA1/2 mutated breast cancers, have overlapping mutational signatures (ie. Signature 3). It would be interesting to note whether or not the observations made here in ovarian cancer carry over to other Signature 3 tumours; or whether they are in fact specific to ovarian cancer (also see Major Point 1).

The authors focus their computational analysis of cancer mutational signatures in
Response: Thank you for this suggestion. We analyzed 42 BRD proteins in Breast cancer. The data shows that the mutations in the 6 bromodomain containing genes (BRD9, ASH1L, BRWD3, KMT2A, ZMYND8 and SMARCA4) associated with signature 3 in ovarian cancer were not associated with signature 3 in breast cancer, another tumor type enriched for signature 3.