Bromodomain-Containing Protein BRD4 Is Hyperphosphorylated in Mitosis

Round 1

Reviewer 1 Report

The manuscript “Bromodomain-containing protein BRD4 is hyperphosphorylated in mitosis” by Ranran Wang, June F Yang, Flora Ho, Erle S Robertson, and Jianxin You reports that bromodomain-containing protein 4 (BRD4), a critical therapeutic target for wide variety of cancers, is hyperphosphorylated during mitosis. The authors identified CDK1 as a potential kinase responsible for phosphorylation of at least four sites on BRD4, one in N-terminal region and a cluster of three sites in C-terminal region. By combining auxin-inducible degron, the authors generated stable cell lines in which endogenous BRD4 is replaced by Dox-induced WT BRD4 or the non-phosphorylatable 4A mutant. The 4A mutant expressing cell line exhibited more sensitivity to BRD4 targeted cancer drug JQ1. As many of the cancers have acquired JQ1 resistance, the authors derived from the obtained data the hypothesis predicting possible synergistic effects of co-administering CDK1 inhibitors and JQ1. In the last part of the paper this synergy was demonstrated by treating a cell line with different combinations of JQ1 and CDK1 inhibitor RO3306.

The synergistic effect of these two inhibitors is interesting and may have therapeutic value. Therefore, in general, I am enthusiastic about the work. However, there are several serious issues and missing experiments in parts described under the figures 1-3 that must be addressed.

1) The authors should explain the intermediate effect, in Figure 2A, on BRD4 mitotic hyperphosphorylation shift in case of RO-3306 (inhibitor of CDK1) compared to a very strong effect in case of BMS-265246 (targeting both CDK1 and 2). Secondly, the observed effect of BMS-265246 would seemingly suggest also the involvement of Cdk2, but why the treatment with K03861 (targeting CDK2) would promote the BRD4 mitotic hyperphosphorylation? In fact, K03861 uplifted the shift more than any other inhibitors, even to a higher level than the nocodazol. How could this be interpreted? Does suppression of CDK2 somehow promote the expression of Cyclin B (clearly, lower cyclin expression in both RO-3306 and BMS-265246 lanes was observed, but insufficiently discussed)? But then again, how can one explain the shift patterns in case of RO-3306 and BMS-265246?

2) Recombinant GST-CDK1 and GST-Cyclin B1 were expressed and purified from E. Coli (Figure 2C). This is a very questionable way to produce active CDK complexes. The authors do not even use CAK for activating phosphorylation. This a very serious issue with the whole study. Controls with commercially available Cdk1 and Cdk2 complexes should be added. The authors should provide clear quantitative comparison of specific activities of their GST-tagged complexes and the ones commercially obtained or previously well characterized by other labs (produced in eukaryotic expression system, e.g. in Sf9 cells). As discussed below, the CDK1 activity of the preparation that the authors use, seems to be very low and puts a huge question mark onto the validity of the whole study. Is it really CDK1, or could it be an indirect downstream effect? The study certainly does not follow the gold standard experimental procedures for kinase target validation. Also, some explanation should be given why only the selected 4 sites are phosphorylated. There should also be an in vitro proof that.

3) Why RO-3306 has no effect whatsoever on the shift in Figure 2B compared to NS lane, while BMS-265246 has a strong effect?

4) Related to point 2 above. In Figure 3C no phosphorylation shift is observed when the CBB and autorad gels are compared. This suggests that the kinase preparation is really very weak. The authors should run this assay in PhosTag gels. Besides, single alanine mutant forms of the four sites, and other candidate sites, should be included in the assay. As it is now, the assay has no relevance at all. Contrarily, it would suggest that the in vitro kinase assay does not produce hyperphosphorylated forms. This would leave open the question of possible other, yet unknown, factors (kinases) that may be involved in vivo.

5) The authors hypothesized that CDK1-activated BRD4 mitotic hyperphosphorylation allows BRD4 to bind more strongly to its specific bookmarked genes during mitosis to stimulate their expression, promote downstream cellular proliferation, and consequently contribute to BETi resistance. To test this, the authors examined if WT BRD4 Flp-In cells are more resistant to JQ1 inhibition compared to the cells expressing the non-phosphorylatable 4A mutant. The IC50 was reduced from 0.79 uM for WT cells to 0.27 uM for the 4A mutant cells. The authors conclude that this result suggests that blocking CDK1-mediated BRD4 mitotic hyperphosphorylation may reduce its affinity for acetylated chromatin at key mitotic bookmarked genes, making it more easily dissociated by BET inhibitors from these genes and thus dampening cell viability. However, IC50 of relative viability (Figure 4E) is the most indirect way to test this very biochemical hypothesis, and provided experiment is certainly not enough to prove or disprove the hypothesis. Given the necessity to redo several experiments with proper CDK1 complexes (points 2 and 4), ones that are able to produce phosphorylation shifts in vitro, it would also be necessary to do the quantitative binding assay of hyperphosphorylated BRD4 and acetylated chromatin, and to evaluate the dose-dependent effects of JQ1 on this binding event in vitro.

 

Minor points

1) In Figure 4C, how the size markers 250 and 300 wree determined. As PhosTag gels show highly differential migration of proteins with similar MW but different phosphorylation status (that has hardly any significant contribution to the MW of such large proteins), the MW markers are meaningless in this context

2) It is not really clear what the lane M in Figure 2B means. The horizontal line covers the lanes non-synchroneous (NS). It is confusing for the reader, how a lane can be simultaneously NS and M

3) A bit unclear wording: “Since the BRD4 4A mutations could inhibit BRD4 mitotic hyperphosphorylation in HEK293T cells, we also performed an in vitro kinase assay to determine whether these mutations can abolish CDK1-mediated BRD4 phosphorylation.” It leaves an impression that 4A mutated form of BRD4 is a pseudosubstrate inhibitor for CDK.

4) Typo: “Purified recombinant GST, GST-CDK1, and/or GST-CCND1 proteins were also included in the reaction as indicated in the figure legends”. Perhaps CCNB1, not CCND1?

Author Response

The manuscript “Bromodomain-containing protein BRD4 is hyperphosphorylated in mitosis” by Ranran Wang, June F Yang, Flora Ho, Erle S Robertson, and Jianxin You reports that bromodomain-containing protein 4 (BRD4), a critical therapeutic target for wide variety of cancers, is hyperphosphorylated during mitosis. The authors identified CDK1 as a potential kinase responsible for phosphorylation of at least four sites on BRD4, one in N-terminal region and a cluster of three sites in C-terminal region. By combining auxin-inducible degron, the authors generated stable cell lines in which endogenous BRD4 is replaced by Dox-induced WT BRD4 or the non-phosphorylatable 4A mutant. The 4A mutant expressing cell line exhibited more sensitivity to BRD4 targeted cancer drug JQ1. As many of the cancers have acquired JQ1 resistance, the authors derived from the obtained data the hypothesis predicting possible synergistic effects of co-administering CDK1 inhibitors and JQ1. In the last part of the paper this synergy was demonstrated by treating a cell line with different combinations of JQ1 and CDK1 inhibitor RO3306. The synergistic effect of these two inhibitors is interesting and may have therapeutic value. Therefore, in general, I am enthusiastic about the work. However, there are several serious issues and missing experiments in parts described under the figures 1-3 that must be addressed.

 

Response: We thank this reviewer for the encouraging and constructive comments, which have allowed us to further strengthen our manuscript. As outlined below, we have added new data sets and revised the manuscript to address these issues raised by the reviewer and to support our conclusion.

 

1) The authors should explain the intermediate effect, in Figure 2A, on BRD4 mitotic hyperphosphorylation shift in case of RO-3306 (inhibitor of CDK1) compared to a very strong effect in case of BMS-265246 (targeting both CDK1 and 2). Secondly, the observed effect of BMS-265246 would seemingly suggest also the involvement of Cdk2, but why the treatment with K03861 (targeting CDK2) would promote the BRD4 mitotic hyperphosphorylation?

 

Response: We thank the reviewer for pointing this out. BMS-265246 can inhibit CDK1/Cyclin B, CDK2/Cyclin E, and CDK4/Cyclin D with IC50 of 6 nM, 9 nM, and 0.23 μM, respectively (Misra et al. 2003, Bioorg. Med. Chem. Lett. 13, 2405). On the other hand, RO-3306 selectively inhibits CDK1/cyclin B1 activity with a Ki value of 35 nM (Vassilev et al. 2006, PNAS. 103, 10660). Because RO-3306 is an ATP-competitive kinase inhibitor, the Ki value represents about one-half of its IC50 value. Therefore, the results from Misra et al. 2003 and Vassilev et al. 2006 together suggest that BMS-265246 inhibits CDK1/Cyclin B1 with an IC50 that is nearly 10 times lower than RO-3306. This much stronger effectiveness of BMS-265246 in inhibiting CDK1/Cyclin B1 may contribute to its stronger effect in blocking BRD4 mitotic hyperphosphorylation as shown in Fig.  2A.

 

We agree with the reviewer that the observed effect of BMS-265246 could suggest that other targets of BMS-265246, such as CDK2 and CDK4, could be involved in BRD4 mitotic hyperphosphorylation. We have tested K03861 (selectively inhibits CDK2 with Kd of 50 nM) and Palbociclib (selectively inhibits CDK4/6 with IC50 of 11 nM/16 nM) in many experimental repeats. Our studies consistently showed that K03861 and Palbociclib could not inhibit BRD4 mitotic hyperphosphorylation (See Fig. 2A and the new Fig. S4). Our finding therefore suggests that CDK2 and CDK4 are unlikely to contribute to BRD4 mitotic hyperphosphorylation. BMS-265246 belongs to the Type I family of kinase inhibitors, which are ATP-competitors that mimic the purine ring of the adenine moiety in ATP. As summarized by Bhullar et al. 2018, Molecular Cancer 17, 48, the targeted ATP pocket is highly conserved throughout the kinome. Therefore, Type I inhibitors tend to show low kinase selectivity with high potential for off-target effects. As described above, the higher potency of BMS-265246 in competing the kinase ATP-binding sites as compared to RO-3306 is likely to reduce its specificity, thereby targeting other unknown kinase(s), which could contribute to BRD4 phosphorylation. In addition, as shown in the new Fig. S5, we have treated HCC2429 cells with BMS-265246, RO-3306, K03861, or the combination of RO-3306 and K03863. The results show that dual treatment of RO-3306 (CDK1 inhibitor) and K03861 (CDK2 inhibitor) could not achieve the same effect in blocking BRD4 phosphorylation as the treatment with BMS-265246, which inhibits both CDK1 and CDK2. This finding further supports that BMS-265246 may target other kinases, which could phosphorylate BRD4. These new data presented in the new Fig. S4 and S5 and the comments above are discussed in the “Discussion” section (Line 409-430).

 

In fact, K03861 uplifted the shift more than any other inhibitors, even to a higher level than the nocodazol. How could this be interpreted?

 

Response: Among our many experimental repeats, this is the only time we observed this K03861 effect. For example, K03861 does not show this strong uplifting effect in the new Fig.  S4. Therefore, we don’t want to overinterpret this one time observation.

 

Does suppression of CDK2 somehow promote the expression of Cyclin B (clearly, lower cyclin expression in both RO-3306 and BMS-265246 lanes was observed, but insufficiently discussed)? But then again, how can one explain the shift patterns in case of RO-3306 and BMS-265246?

 

Response: This comment is a bit confusing because there is no CDK2 suppression in Fig.  2. So we are guessing that this reviewer is referring to CDK1 inhibition in Fig. 2. Also, suppression of CDK1 does not “promote the expression of Cyclin B”. Instead, Cyclin B1 level was slightly reduced in cells treated with BMS-265246 and RO-3306. This is consistent with a previous landmark study showing that adding CDK1 inhibitors to cells in mitosis “induces cytokinesis and other normal aspects of mitotic exit, including cyclin B degradation” (Potapova et al., 2006 Nature 440, 954-958). To avoid this cell cycle effect, we only treated the cells with the indicated kinase inhibitors for 1h so that RO-3306 and BMS-265246 treatment only caused very minor reduction of Cyclin B1, which does not account for the dramatic inhibition of BRD4 mitotic hyperphosphorylation. However, the data from Fig. 2A indicates that this Cyclin B1 reduction could contribute to some level of loss in BRD4 mitotic hyperphosphorylation. Therefore, to rule out this cell cycle effect, we performed the in vitro kinase assay to confirm that CDK1 could directly phosphorylate BRD4 (Fig. 2C-2E). This discussion has been added to the manuscript (Line 147-154).

 

2) Recombinant GST-CDK1 and GST-Cyclin B1 were expressed and purified from E. Coli (Figure 2C). This is a very questionable way to produce active CDK complexes. The authors do not even use CAK for activating phosphorylation. This a very serious issue with the whole study. Controls with commercially available Cdk1 and Cdk2 complexes should be added. The authors should provide clear quantitative comparison of specific activities of their GST-tagged complexes and the ones commercially obtained or previously well characterized by other labs (produced in eukaryotic expression system, e.g. in Sf9 cells). As discussed below, the CDK1 activity of the preparation that the authors use, seems to be very low and puts a huge question mark onto the validity of the whole study. Is it really CDK1, or could it be an indirect downstream effect? The study certainly does not follow the gold standard experimental procedures for kinase target validation.

 

Response: Yes, we agree with this reviewer. In fact, we have already performed the BRD4 in vitro phosphorylation assay using recombinant human CDK1/CyclinB1 complex expressed in baculovirus-infected Sf9 cells (Cat. No. 14-450, Millipore). We detected strong and dose-dependent phosphorylation of BRD4 using this CDK1/CyclinB1 kinase complex produced by the insect cells. This result is now presented in the new Figure S2. We were excited about this finding. At the same time, we were concerned that other eukaryotic kinases carried over from the insect cells might be present in this kinase sample and contributed to the BRD4 phosphorylation kinase activity. Therefore, we switched to the CDK1/CyclinB1 complex purified from E.Coli. As shown in Figure 2E and discussed in the manuscript, incubation with GST-CDK1, GST-Cyclin B1, or GST alone did not lead to BRD4 phosphorylation in vitro. Only when GST-CDK1 was combined with GST-Cyclin B1, which promotes formation of the active CDK1 kinase complex, was BRD4 phosphorylation clearly detected. On the other hand, no phosphorylation of the negative control TII protein was detected under any of the conditions tested. Cyclin B1 was also clearly phosphorylated by CDK1 in both the BRD4-TII and TII reaction, providing an internal positive control for the CDK1 kinase activity in these reactions. We admit that the bacterial expressed CDK1/CyclinB1 kinase complex shows relatively weak activity toward BRD4, but with all of the positive and negative controls included in the experiment, it provides clear evidence that the CDK1/Cyclin B1 complex but not other contaminating eukaryotic kinase(s) could directly phosphorylate BRD4 in vitro. We thank the reviewer for the great suggestion of using CAK for activating phosphorylation. We will definitely include this idea in our future studies. We have included new Figure S2 and related discussion in the manuscript (Line 167-180).

 

Also, some explanation should be given why only the selected 4 sites are phosphorylated. There should also be an in vitro proof that.

 

Response: As described in Figure 3 and the Result Section 2.3 (Line 198-259), in order to identify the BRD4 amino acid residues phosphorylated by CDK1 during mitosis, we first used the ScanSite online kinase-specific phosphorylation site analysis server and found 21 predicted CDK1 phosphorylation sites on the BRD4 protein. We then performed alanine (A) substitution mutagenesis on these predicted CDK1 recognition sites one by one to determine their impact on BRD4 hyperphosphorylation. The single-A substitution mutants of BRD4 were individually expressed in HEK293T cells and the mitotic cell lysates were analyzed in Phos-tag gels followed by western blotting. From these studies, we discovered that the BRD4 mutants T249A, S1045A, S1117A and S1126A each showed reduced BRD4 mitotic phosphorylation, as indicated by the faster migrating BRD4 bands compared to wild type BRD4 protein (Figure 3). Based on the single-A mutagenesis results, we generated the BRD4 4A mutant containing the T249A, S1045A, S1117A and S1126A mutations. When expressed in HEK293T cells, the BRD4 4A mutant extracted from mitotic cells migrates to the same position as the WT BRD4 protein isolated from the unsynchronized cells (Figure 3B). This result suggests that the 4A mutations eliminate almost all of BRD4’s mitotic phosphorylation. We also performed an in vitro kinase assay to determine whether these mutations can abolish CDK1-mediated BRD4 phosphorylation. The autoradiography of the in vitro kinase assay samples show that the BRD4 4A mutant dramatically reduced levels of CDK1/Cyclin B1-mediated BRD4 phosphorylation compared to wild type BRD4 (Figure 3C). To confirm this in vitro result in cells, we expressed BRD4 and GFP-Cyclin B1 in HEK293T cells. While overexpression of Cyclin B1 was sufficient to induce hyperphosphorylation of the wild type BRD4 molecule, BRD4 4A quadruple mutagenesis efficiently abolishes Cyclin B1/CDK1-mediated BRD4 hyperphosphorylation (Figure 3D). Together, these studies identified four BRD4 residues, T249, S1045, S1117, and S1126, as the major sites for CDK1-mediated BRD4 mitotic hyperphosphorylation. Since BRD4 4A quadruple mutagenesis abolishes most but not all of BRD4’s hyperphosphorylation in mitotic cells as well as in the in vitro kinase reaction (Figure 3), we also suggested that other BRD4 residues could also be phosphorylated by CDK1. This discussion has been added to the end of the Result Section 2.3 (Line 256-259). In addition, we have mentioned several times in the manuscript that “we also noticed that BRD4 4A quadruple mutagenesis abolishes most but not all of BRD4’s hyperphosphorylation in mitotic cells as well as in the in vitro kinase reaction (Figure 3), suggesting that other BRD4 residues could also be phosphorylated by CDK1 in mitosis to contribute to BRD4’s mitotic function” (Line 446-449) and “As discussed above, it is also possible that BRD4 residues other than those represented in the 4A mutant could still be hyperphosphorylated by CDK1, since BRD4 4A quadruple mutagenesis did not completely abolish the ability of BRD4 to be phosphorylated by CDK1 (Figure 3)” (Line 484-487).

 

3) Why RO-3306 has no effect whatsoever on the shift in Figure 2B compared to NS lane, while BMS-265246 has a strong effect?

 

Response: Figure 2B is showing the inhibitor treatment results for not-synchronized (NS) cells including both interphase and mitotic cells. RO-3306 has no effect likely because its target CDK1 only phosphorylates BRD4 in mitotic cells, which represent a small portion of the NS cells. The strong effect of BMS-265246 may reflect its strong ability to inhibit BRD4 hyperphosphorylation in mitotic cells in a way that it became obvious in the NS population. Also as discussed above, BMS-265246 may have other kinase targets that may contribute to BRD4 phosphorylation in interphase.

 

4) Related to point 2 above. In Figure 3C no phosphorylation shift is observed when the CBB and autorad gels are compared. This suggests that the kinase preparation is really very weak. The authors should run this assay in PhosTag gels. Besides, single alanine mutant forms of the four sites, and other candidate sites, should be included in the assay. As it is now, the assay has no relevance at all. Contrarily, it would suggest that the in vitro kinase assay does not produce hyperphosphorylated forms. This would leave open the question of possible other, yet unknown, factors (kinases) that may be involved in vivo.

 

Response: This is an important point. As discussed in our response to point 2 above, we agree that the CDK1/CyclinB1 kinase complex produced in bacterial has weak activity toward BRD4, but it allows us to conclude that the CDK1/Cyclin B1 complex but not other contaminating eukaryotic kinase(s) could directly phosphorylate BRD4 in vitro. Because of this weak activity, it is reasonable that the enzymes used in the assay were not able to completely hyperphosphorylate the large amount (g level) of BRD4 used in the reaction. This could be the reason why we are not seeing the phosphorylation shift with the standard SDS-PAGE gels shown in Figure 3C. Since we don’t anticipate that all of the recombinant BRD4 used in the reaction to be hyperphosphorylated, we chose not to use the Phos-Tag gel, but instead, use the incorporation of radioactive ATP as the way to detect BRD4 phosphorylation. In addition, we have included a set of new data in the new Fig. S2 to show that, when human CDK1/CyclinB1 complex produced by infected Sf9 cells with much higher activity were used in the same assay, we could detect in the autoradiography gel the significantly shifted radioactive band (marked by the bracket) representing hyperphosphorylated BRD4. The single-A substitution mutants of BRD4 were individually expressed in HEK293T cells and analyzed in Phos-tag gels in Fig.  3A. For the in vitro kinase assay, we felt that it is more meaningful to look at the 4A mutant instead of the single-A substitution mutants. The Fig. 3C autoradiography of the in vitro kinase assay samples show that, even with such weak kinase and so much purified BRD4 substrate, the BRD4 4A mutant still shows dramatically reduced levels of CDK1/Cyclin B1-mediated BRD4 phosphorylation compared to wild type BRD4 (Figure 3C). We also performed in vivo experiment to show that compared to wild type BRD4 molecule, BRD4 4A quadruple mutagenesis efficiently abolishes Cyclin B1/CDK1-mediated BRD4 hyperphosphorylation (Figure 3D). Together, these studies identified BRD4 T249, S1045, S1117, and S1126, as the major sites for CDK1-mediated BRD4 mitotic hyperphosphorylation. As we discussed above in our response to Point 1 and 2, there are other sites in BRD4 could be phosphorylated by CDK1 or other kinases. But for this manuscript, we decided to focus on the 4A mutant and CDK1 phosphorylation. The data presented in our manuscript support our conclusion that BRD4 is hyperphosphorylated in mitosis.

 

5) The authors hypothesized that CDK1-activated BRD4 mitotic hyperphosphorylation allows BRD4 to bind more strongly to its specific bookmarked genes during mitosis to stimulate their expression, promote downstream cellular proliferation, and consequently contribute to BETi resistance. To test this, the authors examined if WT BRD4 Flp-In cells are more resistant to JQ1 inhibition compared to the cells expressing the non-phosphorylatable 4A mutant. The IC50 was reduced from 0.79 uM for WT cells to 0.27 uM for the 4A mutant cells. The authors conclude that this result suggests that blocking CDK1-mediated BRD4 mitotic hyperphosphorylation may reduce its affinity for acetylated chromatin at key mitotic bookmarked genes, making it more easily dissociated by BET inhibitors from these genes and thus dampening cell viability. However, IC50 of relative viability (Figure 4E) is the most indirect way to test this very biochemical hypothesis, and provided experiment is certainly not enough to prove or disprove the hypothesis. Given the necessity to redo several experiments with proper CDK1 complexes (points 2 and 4), ones that are able to produce phosphorylation shifts in vitro, it would also be necessary to do the quantitative binding assay of hyperphosphorylated BRD4 and acetylated chromatin, and to evaluate the dose-dependent effects of JQ1 on this binding event in vitro.

 

Response: We agree with the reviewer that the Fig. 4E experiment is not an ideal and direct method to test our hypothesis and that quantitative binding is a better way to determine the affinity between hyperphosphorylated BRD4 and acetylated chromatin. As we described in the Result Section 2.4, in order to determine the effects of mitotic hyperphosphorylation on BRD4’s affinity for chromatin, we have performed a BRD4 co-immunoprecipitation experiment and found that the BRD4 WT protein and 4A mutant could pull down similar amount of acetylated histone H4 (Figure 4D). We have also used increasing salt concentrations to extract WT and 4A BRD4 from nuclear lysates. Through this method, we also did not detect any significant difference in the bulk chromatin binding affinities of WT or 4A BRD4. This data is now presented in the new Figure S3. Since BRD4 is a mitotic bookmarker that preserves the epigenetic memory of key M/G1 growth-associated genes through mitosis to ensure their rapid postmitotic transcriptional re-activation, we reasoned that CDK1-activated BRD4 mitotic hyperphosphorylation may allow BRD4 to bind more strongly to these specific bookmarked genes during mitosis to stimulate their expression, promote downstream cellular proliferation, and consequently contribute to BETi resistance. This is why we started to examine if WT BRD4 Flp-In cells are more resistant to JQ1 inhibition compared to the cells expressing the non-phosphorylatable 4A mutant. While the IC50 method is not directly measuring the BRD4 affinity for chromatin, it provides some quantitative measurement to suggest that blocking CDK1-mediated BRD4 mitotic hyperphosphorylation may reduce its affinity for acetylated chromatin at key mitotic bookmarked genes (but not all genes), making it more easily dissociated by BET inhibitors from these genes and thus dampening cell viability.

 

Minor points

 

1) In Figure 4C, how the size markers 250 and 300 wree determined. As PhosTag gels show highly differential migration of proteins with similar MW but different phosphorylation status (that has hardly any significant contribution to the MW of such large proteins), the MW markers are meaningless in this context

 

Response: We very much appreciate this comment from the reviewer. For all Phos-tag gels presented in this study, standard High Range Protein Ladder (Cat. No. 26625, Thermo Fisher Scientific) are resolved in each gel along with protein samples tested. However, as indicated by the manufacturer, the molecular weight markers are frequently distorted during Phos-tag gel electrophoresis, and therefore they can only be used as rough estimate of the molecular weights. We have clarified this in Section 4.3.

 

2) It is not really clear what the lane M in Figure 2B means. The horizontal line covers the lanes non-synchroneous (NS). It is confusing for the reader, how a lane can be simultaneously NS and M

 

Response: Thanks for pointing this out. For these gels, we have loaded control samples from untreated mitotic and not-synchronized cells to show the relative position of BRD4 hyperphosphorylation and low phosphorylation bands on this Phos-tag gel. M lane in Fig.  2B represents the sample from untreated mitotic cells. We have clarified this in the Fig.  2 legend and also fixed the line issue.   

 

3) A bit unclear wording: “Since the BRD4 4A mutations could inhibit BRD4 mitotic hyperphosphorylation in HEK293T cells, we also performed an in vitro kinase assay to determine whether these mutations can abolish CDK1-mediated BRD4 phosphorylation.” It leaves an impression that 4A mutated form of BRD4 is a pseudosubstrate inhibitor for CDK.

 

Response: We have changed the sentence into: “Since the BRD4 4A mutations could block BRD4 mitotic hyperphosphorylation in HEK293T cells, we also performed an in vitro kinase assay to determine whether these mutations can prevent CDK1-mediated BRD4 phosphorylation.” (Line 247-249)

 

 

4) Typo: “Purified recombinant GST, GST-CDK1, and/or GST-CCND1 proteins were also included in the reaction as indicated in the figure legends”. Perhaps CCNB1, not CCND1?

 

Response: Yes, it should be CCNB1. We have corrected it (Line 553). Thank you!

 

Author Response File: Author Response.docx

Reviewer 2 Report

This study demonstrates hyper phosphorylated BRD4 by CDK1 in mitosis contributes to oncogenesis by increasing interaction with acetylated histones and then chromatin with much higher affinity, that subsequently stimulate abnormal induction of target oncogenes. The second importance could be their finding of  a synergistic antitumor effect of combinatory inhibition of CDK1 and BRD4 binding to chromatin, by using JQ1 and RO-3306.

 

This is an excellent paper and I have no serious criticisms regarding methodology, results, methods and interpretation of results.

A few minor comments are listed below:

1. The authors demonstrated recombinant BRD4 was directly phosphorylated by recombinant CDK1 by in vitrophosphorylation experiment. According to the Method section, the reaction was performed at 30^oC for 1 hour. I wonder if the condition was suitable for CDK1, because it seems to be too long for kinase assays. The reviewer suggests authors to cite related reference(s) and describe whether the assay conditions were optimized.
2. Phosphorylation sites in BRD4 were not determined by proteomics analysis, although the authors addressed to the conclusion by using alanine-substitution mutants. In general, phosphorylation sites could be “determined” by combining both methods. Therefore, the reviewer suggests the title of Figure 3 “Identification…” could be reconsidered carefully.
3. It is unclear why the authors select DLD-1 cells to introduce knock-out/knock-in BRD4 mutant. Is it because the cell showed hyper-phosphorylated BRD4? The cell line is derived from colorectal carcinoma and expected to have multiple genomic mutations.
4. Line 309-310: the authors mentioned “more sensitive to JQ1 and more resistant to RO-3306”, which seems not supported by the data.

Author Response

This study demonstrates hyper phosphorylated BRD4 by CDK1 in mitosis contributes to oncogenesis by increasing interaction with acetylated histones and then chromatin with much higher affinity, that subsequently stimulate abnormal induction of target oncogenes. The second importance could be their finding of  a synergistic antitumor effect of combinatory inhibition of CDK1 and BRD4 binding to chromatin, by using JQ1 and RO-3306.

 

This is an excellent paper and I have no serious criticisms regarding methodology, results, methods and interpretation of results.

A few minor comments are listed below:

 

1. The authors demonstrated recombinant BRD4 was directly phosphorylated by recombinant CDK1 by in vitrophosphorylation experiment. According to the Method section, the reaction was performed at 30^oC for 1 hour. I wonder if the condition was suitable for CDK1, because it seems to be too long for kinase assays. The reviewer suggests authors to cite related reference(s) and describe whether the assay conditions were optimized.

 

Response: We greatly appreciate all the supportive comments from this reviewer. Our kinase assay protocol was adapted from Devaiah et al., 2012 PNAS 109, 6927. We have optimized the ratio of hot and cold ATP in the reaction but keep the other assay conditions such as the reaction time. We have included the citation and clarification in Section 4. 4 of the manuscript.

 

2. Phosphorylation sites in BRD4 were not determined by proteomics analysis, although the authors addressed to the conclusion by using alanine-substitution mutants. In general, phosphorylation sites could be “determined” by combining both methods. Therefore, the reviewer suggests the title of Figure 3 “Identification…” could be reconsidered carefully.

 

Response: This point was well taken. We have revised the title of Fig. 3 to “Determination of BRD4 mitotic phosphorylation sites by alanine-substitution mutagenesis.”.

 

3. It is unclear why the authors select DLD-1 cells to introduce knock-out/knock-in BRD4 mutant. Is it because the cell showed hyper-phosphorylated BRD4? The cell line is derived from colorectal carcinoma and expected to have multiple genomic mutations.

 

Response: DLD-1 was one of the cell lines that show hyperphosphorylation in mitosis (Fig. 1). It was selected to generate Flp-In BRD4 WT/4A because we have previously obtained a DLD-1 cell line with the Flp-In site of the Flp-In^TM T-REx^TM reconstitution system integrated into the genome.

 

4. Line 309-310: the authors mentioned “more sensitive to JQ1 and more resistant to RO-3306”, which seems not supported by the data.

 

Response: We have deleted this statement in the manuscript (Line 349). 

 

Author Response File: Author Response.docx

Reviewer 3 Report

This paper aims to determine the role of hyperphosphorylation of BRD4 in regulation of its activity. BRD4 is a epigenetic reader that associates with chroatin at higher affinity when hyperphosphorylated and the authors use a small molecule screen to identify underlying mechanisms that control its phosphorylation. The discovery that inhibition of microtubule formation and enrichment of phosphorylation in M-phase leads to the identification of CDK1 activity being primarily (although not entirely) responsible for BRD4 phosphorylation.

The authors show through well thought out and performed experiemtns that BRD4 is phosphorylated at multiple positions by CDK1. The use of AID and CRISPR to engineer a BRD4 degrader cell line supports this work. The complementation of the AID-BRD4 with tagged-WT and 4A non-phosphorylatable BRD4 constructs enable the dissection of this pathway. The further characterisation of the role of cyclin B-CDK1 in the phosphorylation  of BRD4 shows that CDK1 can indeed mediate site specific phosphorylation.

Finally, the authors test the hypothesis that CDK1 inhibition and BET inhibition may sensitise BET inhibitor resistant cell lines in combination. The proliferation assays broadly support this hypothesis.

I only have minor comments:

(i) In figure 4E the log of drug concentration is shown. This would be more transparent with the concentration in micromolar concentrations as show in Fig 5.

(ii) In Figure 5, the symbol for micro is shown as 'u' and not mu. Please replace with the correct symbol.

I would be supportive for publication of this work with completion of these very minor points. It is a well written and logical paper that identifies that BRD4 is regulated in mitosis by CDK1 mediated phosporylation.

Author Response

This paper aims to determine the role of hyperphosphorylation of BRD4 in regulation of its activity. BRD4 is a epigenetic reader that associates with chroatin at higher affinity when hyperphosphorylated and the authors use a small molecule screen to identify underlying mechanisms that control its phosphorylation. The discovery that inhibition of microtubule formation and enrichment of phosphorylation in M-phase leads to the identification of CDK1 activity being primarily (although not entirely) responsible for BRD4 phosphorylation.

The authors show through well thought out and performed experiemtns that BRD4 is phosphorylated at multiple positions by CDK1. The use of AID and CRISPR to engineer a BRD4 degrader cell line supports this work. The complementation of the AID-BRD4 with tagged-WT and 4A non-phosphorylatable BRD4 constructs enable the dissection of this pathway. The further characterisation of the role of cyclin B-CDK1 in the phosphorylation of BRD4 shows that CDK1 can indeed mediate site specific phosphorylation.

Finally, the authors test the hypothesis that CDK1 inhibition and BET inhibition may sensitise BET inhibitor resistant cell lines in combination. The proliferation assays broadly support this hypothesis.

I only have minor comments:

(i) In figure 4E the log of drug concentration is shown. This would be more transparent with the concentration in micromolar concentrations as show in Fig 5.

(ii) In Figure 5, the symbol for micro is shown as 'u' and not mu. Please replace with the correct symbol.

 

I would be supportive for publication of this work with completion of these very minor points. It is a well written and logical paper that identifies that BRD4 is regulated in mitosis by CDK1 mediated phosporylation.

 

Response: We appreciate the reviewer’s supportive and insightful comments on our manuscript. We have revised the relevant figures in the manuscript as suggested the reviewer. For point i), we have added the drug dose in micromolar concentrations to the new Fig. 4E. For point ii) the correct symbol is used in the new Fig. 5.

 

Author Response File: Author Response.docx

Reviewer 4 Report

In the presented study authors aimed to address a possible link between BRD4 phosphorylation and resistance to BET inhibitors. They found that BRD4 is massively phosphorylated during mitosis and using alanine mutants, they identified several sites that are phosphorylated by CDK1 both in vitro and in vivo. They further constructed an inducible system in which they can reconstitute endogenous BRD4 either with wild type or non-phophorylatable version of BRD4. Using this system they found that WT and 4A cells recruit comparable amounts of transcriptional activators and also comparably bind chromatin. Surprisingly, they find that 4A mutant cells are more sensitive to BET inhibitor. From this they conclude that hyperphosphorylation of BRD4 may be one of the mechanism of BET inhibitor resistance. Although this may be an interesting concept, authors do not provide direct evidence for this. In Fig 4E, the 4A cells seem to generally proliferate slower than WT cells so it is not surprising that the curve is shifted also for 4A cells in the presence of JQ1 inhibitor. Also they should rather use phosphorylation mimicking 4D mutant to prove its resistance to the inhibitor. Observation that inhibition of CDK1 and BET inhibitor could act synergistically is potentially interesting, however, results in Figure 5A,B argue against the possibility that this is the effect of BRD4 phosphorylation. Authors state that CDK1 could be abnormally activated causing hyperphosphorylation of BRD4 in interphase but again, they do not provide any experimental evidence for this. As CDK1 activity is tightly regulated by inhibitory phosphorylation in normal and cancer cells, this scenario seems highly unlikely. Rather it is possible that some cancers may have higher mitotic index resulting in the presence of phosphorylated BRD4. How these multiple mitotic phosphorylations of BRD4 affect its function still remains to be determined. In summary, the manuscript shows that BRD4 is phosphorylated in mitosis but fails to provide mechanistic insight how this affects its function and to what extent it regulates the sensitivity of cancer cells to BET inhibitors.

 

Other points

1. It is unclear from the methods how authors expressed and purified BRD4 in bacteria. Clearly, it is not full length BRD4 as it migrates substantially lower on SDS-PAGE than BRD4 from mammalian cells. Other phosphorylations by CDK1 likely exist on full length BRD4. 3C shows that cdk1 phosphorylates more than the four sites mutated in 4A mutant
2. Throughout the manuscript, authors do not control for the level of mitotic cells in individual experiments. Cyclin B is expressed already in G2 and cannot be used to distinguish between cells in G2 and mitosis. Authors should use another established marker of mitosis (such as pS10-H3).
3. It is unclear whether data in 4E and Figure 5 represent technical or biological replicates.

Author Response

In the presented study authors aimed to address a possible link between BRD4 phosphorylation and resistance to BET inhibitors. They found that BRD4 is massively phosphorylated during mitosis and using alanine mutants, they identified several sites that are phosphorylated by CDK1 both in vitro and in vivo. They further constructed an inducible system in which they can reconstitute endogenous BRD4 either with wild type or non-phophorylatable version of BRD4. Using this system they found that WT and 4A cells recruit comparable amounts of transcriptional activators and also comparably bind chromatin (Fig.4D). Surprisingly, they find that 4A mutant cells are more sensitive to BET inhibitor. From this they conclude that hyperphosphorylation of BRD4 may be one of the mechanism of BET inhibitor resistance. Although this may be an interesting concept, authors do not provide direct evidence for this. In Fig 4E, the 4A cells seem to generally proliferate slower than WT cells so it is not surprising that the curve is shifted also for 4A cells in the presence of JQ1 inhibitor.

 

Response: We truly appreciate this reviewer’s constructive comments. Regarding Fig. 4E, we don’t think cell viability is an issue because, as described in the figure legend, the data presented in Fig. 4E is “Relative cell viability”, which was determined for WT and 4A cell line by normalizing the (+)-JQ1 values to the DMSO control values from the same cell line. This approach allows us to rule out the influence from cell viability. In fact, the first data points for WT and 4A cell line show that, at this low JQ1 concentration, the cells proliferate at almost identical rates. Also, during all of our experiments working with the WT and 4A cell lines, we have never observed any proliferation difference between the two cell lines.

 

Also they should rather use phosphorylation mimicking 4D mutant to prove its resistance to the inhibitor.

 

Response: This is a good idea. We will consider this experiment in our future studies to further explore the detailed mechanism of BRD4 hyperphosphorylation. One of the reasons for using 4A mutant in these first set of experiments is to get a loss-of-BRD4- phosphorylation phenotype and then determine if this loss of function phenotype synergizes with CDK1 inhibitor to achieve the “double hits” synergistic effect in BRD4 function. The other reason we did not try the 4D mutant is because BRD4 phosphorylation could induce intramolecular contact switch to alter its interaction with acetylated chromatin and several other key cellular proteins such as p53 (Wu et al., 2013, Molecular Cell 49, 843–857). We were concerned that introducing 4D mutations in BRD4 could alter this complex BRD4 inter- and intra-molecular interaction that defines its functional activity and stability, making the results difficult to interpret.

 

Observation that inhibition of CDK1 and BET inhibitor could act synergistically is potentially interesting, however, results in Figure 5A,B argue against the possibility that this is the effect of BRD4 phosphorylation.

 

Response: This is a good point. We have in depth discussion on this issue in the manuscript (Line 482-487). We specifically pointed out two possible explanations for our data: (1) “The observation that the BRD4 WT and 4A mutant cells were similarly susceptible to RO-3306 and BETi combination treatment also suggests that CDK1 has other functions in BETi-resistant oncogenesis in addition to its role in phosphorylating BRD4.” (2) “It is also possible that BRD4 residues other than those represented in the 4A mutant could still be hyperphosphorylated by CDK1, since BRD4 4A quadruple mutagenesis did not completely abolish the ability of BRD4 to be phosphorylated by CDK1 (Figure 3).” These discussions show that we are open to these possibilities. We also agree with the reviewer that the synergistic effect of the CDK1 inhibitor and BETi observed in our studies is interesting and may provide the basis for exploring the combinatorial inhibition of CDK1 and BRD4 to improve treatments for a diverse array of BRD4-associated cancers.

 

Authors state that CDK1 could be abnormally activated causing hyperphosphorylation of BRD4 in interphase but again, they do not provide any experimental evidence for this. As CDK1 activity is tightly regulated by inhibitory phosphorylation in normal and cancer cells, this scenario seems highly unlikely. Rather it is possible that some cancers may have higher mitotic index resulting in the presence of phosphorylated BRD4. How these multiple mitotic phosphorylations of BRD4 affect its function still remains to be determined. In summary, the manuscript shows that BRD4 is phosphorylated in mitosis but fails to provide mechanistic insight how this affects its function and to what extent it regulates the sensitivity of cancer cells to BET inhibitors.

 

Response: We agree with the reviewer that CDK1 activity is tightly regulated by inhibitory phosphorylation in normal cells. There are many papers such as those we cited in the manuscript support that CDK1 and/or Cyclin B1 are frequently upregulated or overexpressed in different types of cancers (Bednarek et al. 2016, Tumour Biol. 37, 11115-11126; Hoffmann et al. 2011, Anticancer Res. 31, 3151-3157; Salaun et al. 2008, Advances in experimental medicine and biology 617, 41-56; Malumbres et al. 2009, Nat. Rev. Cancer 9, 153-166; Zheng et al. 2019, Am J Transl Res 11, 7233-7254; Li et al. 2020, J Int Med Res. 48, 300060519897508; Zhuang et al. 2018, Biomed Res Int. 2018, 7897346; Zhang et al. 2017, J Ovarian Res. 10, 60). In some cases, CDK1 is activated because CDC25A, which dephosphorylates the inhibitory phospho-Thr14/Tyr15 groups on CDK1, is overexpressed in cancers (Ray et al. 2008, Cancer Res. 68, 1251-1253; Shen et al. 2012, Anticancer Agents Med Chem. 12, 631‐639). As described in the manuscript, by analyzing the TCGA transcriptomic dataset, we discovered that CDK1 is highly expressed in the vast majority of TNBC tumors. In Figure 5C, we showed that, compared to the noncancerous HDFs, higher levels of cyclin B1 and/or CDK1 were detected in a number of the cancer cell lines. The TNBC cell line MDA-MB-231 was chosen for Fig 5D experiments because it showed the highest expression of both cyclin B1 and/or CDK1.

 

In this manuscript, we main goal is to provide sufficient data to show that BRD4 is hyperphosphorylated in mitosis. In addition, we found that overexpression of CDK1 or its functional partner, Cyclin B1, stimulates BRD4 hyperphosphorylation outside of mitosis in asynchronous cells (Figure 3D). We have also found that BRD4 is hyperphosphorylated throughout the cell cycle in many different types of cancers (Wang et al., 2017, PNAS 114, E5352-e5361). In addition, both CDK1 upregulation and BRD4 hyperphosphorylation have been observed in BETi-resistant cancer cells (Shu et al., 2016, Nature, 529, 413-417). Based on these evidences, we hypothesized that dysregulated CDK1 activation in cancer cells may trigger aberrant BRD4 hyperphosphorylation that persists outside of mitosis, supporting aberrant induction of its mitotic-bookmarked oncogenes, thereby driving tumor cellular proliferation and BETi resistance. To test this hypothesis, we provided additional data to establish that blocking BRD4 hyperphosphorylation (by 4A mutation or CDK1 inhibitor) could increase BETi sensitivity in cancer cells. How CDK1 could be abnormally activated in cancer cells to cause BRD4 hyperphosphorylation is beyond the scope of this manuscript. But we agree with the reviewer that this is an important topic that remains to be studied in the future.

 

Other points

1. It is unclear from the methods how authors expressed and purified BRD4 in bacteria. Clearly, it is not full length BRD4 as it migrates substantially lower on SDS-PAGE than BRD4 from mammalian cells. Other phosphorylations by CDK1 likely exist on full length BRD4.

 

Response: We have included more details on this method and molecular weight markers in Section 4.1 and Section 4.3, respectively. The construct pET23a(+)-BRD4-TII used in our study encodes the full length BRD4 with TII tag (including one TEV cleavage site and two IgG binding domains) on the C-terminus. BRD4-TII protein was purified from bacteria lysates with IgG sepharose (GE healthcare). TII tag adds about 16 kDa to the fusion protein. So the predicted molecular weight of BRD4-TII should be 216 kDa and the protein should migrate between the 180~250 kDa markers on the Coomassie Brilliant Blue stained SDS-PAGE gel (Fig. 2D and 3C).  Because BRD4-TII expressed in E.coli is not modified by phosphorylation or other post-translational modification as in mammalian cell system. It is reasonable that it appears smaller compared to the sample isolated from mammalian cells. Also, for Phos-tag gels presented in this study, standard High Range Protein Ladder (Cat. No. 26625, Thermo Fisher Scientific) were resolved in each gel along with protein samples tested. However, as indicated by the manufacturer, the molecular weight markers are frequently distorted during Phos-tag gel electrophoresis, and therefore they can only be used as rough estimate of the molecular weights and do not show the accurate molecular weigh of BRD4 expressed in mammalian cells.

 

3C shows that cdk1 phosphorylates more than the four sites mutated in 4A mutant

 

Response: Yes, we have discussed this extensively in the manuscript. This possibility was first mentioned at the end of the Result Section 2.3 (Line 256-259). We also added in the Discussion section that “we also noticed that BRD4 4A quadruple mutagenesis abolishes most but not all of BRD4’s hyperphosphorylation in mitotic cells as well as in the in vitro kinase reaction (Figure 3), suggesting that other BRD4 residues could also be phosphorylated by CDK1 in mitosis to contribute to BRD4’s mitotic function” (Line 446-449) and “As discussed above, it is also possible that BRD4 residues other than those represented in the 4A mutant could still be hyperphosphorylated by CDK1, since BRD4 4A quadruple mutagenesis did not completely abolish the ability of BRD4 to be phosphorylated by CDK1 (Figure 3)” (Line 484-487).

 

2. Throughout the manuscript, authors do not control for the level of mitotic cells in individual experiments. Cyclin B is expressed already in G2 and cannot be used to distinguish between cells in G2 and mitosis. Authors should use another established marker of mitosis (such as pS10-H3).

 

Response: Cyclin B1 protein level gradually increases during G2 phase, and it reaches the highest level at prometaphase and metaphase. In most of our experiments, the cells were synchronized at prometaphase with nocodazole treatment. Therefore, Cyclin B1 protein level was used to distinguish the synchronized cells and non-synchronized cells. As shown in Fig. 3A, the Cyclin B1 marker has worked out very nicely for this purpose. Besides our own data, many others have used Cyclin B as loading control for mitotic cells (Frisa et al. 2009, PLoS ONE 4(9), e7064; Van Horn et al. 2010, J. Biol. Chem. 2010 285, 21849-21857; Roy et al. 2018, Mol Cancer Res. 16(11), 1785-1797; Sakurikar et al. 2012, J Biol Chem. 287(46), 39193–39204). We agree with the reviewer that pS10H3 is a more specific mitotic marker than Cyclin B1. We will consider using pS10H3 in our future studies

 

3. It is unclear whether data in 4E and Figure 5 represent technical or biological replicates.

 

Response: The data in Fig.4E and 5 represent biological replicates.

 

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Review of the revised manuscript

I do not really understand the argument in the response to my point 3. The Phos Tag gel should show even a small fraction of hyperphosphorylated protein in autoradiography, this is the whole point. It is not a question if the kinase “completely hyperphosphorylates” it. The point is to show if the kinase is able to produce any hyperphosphorylated fraction of the target at all. Otherwise there is nothing to compare with the in vivo shifts and the whole central idea remains questionable. This experiment is absolutely necessary and easy to do.

“Because of this weak activity, it is reasonable that the enzymes used in the assay were not able to completely hyperphosphorylate the large amount of BRD4 used in the reaction. This could be the reason why we are not seeing the phosphorylation shift with the standard SDS-PAGE gels shown in Figure 3C. Since we don’t anticipate that all of the recombinant BRD4 used in the reaction to be hyperphosphorylated, we chose not to use the Phos-Tag gel, but instead, use the incorporation of radioactive ATP as the way to detect BRD4 phosphorylation.”

Since the answer to my point 5 was just another way to explain the same shortcoming, and it remains unclear how is it possible that “hyperphosphorylation may reduce its affinity for acetylated chromatin at key mitotic bookmarked genes (but not all genes)”, (Do they mean not all acetylated genes?), I suggest that the authors could tone down this whole hypothesis in the text.

After these two points are corrected, I will have no problems with recommending the manuscript for publishing.

Author Response

Reviewer 1

1. I do not really understand the argument in the response to my point 3. The Phos Tag gel should show even a small fraction of hyperphosphorylated protein in autoradiography, this is the whole point. It is not a question if the kinase “completely hyperphosphorylates” it. The point is to show if the kinase is able to produce any hyperphosphorylated fraction of the target at all. Otherwise there is nothing to compare with the in vivo shifts and the whole central idea remains questionable. This experiment is absolutely necessary and easy to do.

 

Response: With all due respect, we don’t agree with the reviewer on this point. Here are our reasons:

1) In the first round of review, this reviewer asked for the kinase assay to be done using the enzymes purified from insect cells. We provided this data in Fig.  S2, which we showed in the autoradiography gel the significantly shifted radioactive band (marked by the bracket) representing hyperphosphorylated BRD4. In addition to that, throughout the manuscript, we provided sufficient evidence to prove our main point that BRD4 is hyperphosphorylated in mitosis. These evidences include Fig 1. BRD4 is hyperphosphorylated during mitosis; Fig 2A-2B, CDK1 inhibition abolishes BRD4 mitotic hyperphosphorylation; Fig. 2C-2E and Fig S2, recombinant CDK1/cyclin B1 from both bacterial and insect cells can phosphorylate BRD4; and Fig 3A-3C and 4C, blocking the putative CDK1 phosphorylation sites in BRD4 abolishes the hyperphosphorylation of BRD4 in mitotic cells and in vitro. We also showed that blocking BRD4 CDK1 phosphorylation sites inhibits its hyperphosphorylation by over-expressed Cyclin B1 in cells in Fig. 3D. Also, Fig.  3B, 3D and 4C were all done using Phos-tag gel to show that BRD4 hyperphosphorylation is blocked by the 4A mutant.  So we don’t understand why we still need to do the Phos-tag gel analysis of BRD4 phosphorylation using CDK1 and Cyclin B1 purified from insect cells.

2) Most of the published classic kinase assays were done before the development of the Phos-tag gel method by using the hot ATP incorporation and detection in standard SDS-PAGE, as we have done in our in vitro kinase assay.  In fact, our kinase assay protocol was adapted from Devaiah et al., 2012 PNAS 109, 6927. No Phos-tag gel and only standard SDS-PAGE gel was used in that paper.

3) All the other 3 reviewers have no problem with our in vitro kinase data shown in Fig.  2E and 3C.

4) As we discussed in our response to the first round of reviews, even if we get the perfect result for Phos-tag gel analysis of BRD4 phosphorylation using CDK1 and Cyclin B1 purified from insect cells, we still could not rule out the possibility that other eukaryotic kinases carried over from the insect cells may be present in this kinase sample and therefore contribute to the BRD4 phosphorylation kinase activity.

 

Because of these reasons, we believe that the Phos-tag gel analysis of insect kinase assay as suggested by this reviewer will not significantly improve the science of our manuscript. Thank you for considering our opinions.

 

2. Since the answer to my point 5 was just another way to explain the same shortcoming, and it remains unclear how is it possible that “hyperphosphorylation may reduce its affinity for acetylated chromatin at key mitotic bookmarked genes (but not all genes)”, (Do they mean not all acetylated genes?), I suggest that the authors could tone down this whole hypothesis in the text.

After these two points are corrected, I will have no problems with recommending the manuscript for publishing.

 

Response: We agree with the reviewer and have deleted this sentence “This result suggests that blocking CDK1-mediated BRD4 mitotic hyperphosphorylation may reduce its affinity for acetylated chromatin at key mitotic bookmarked genes, making it more easily dissociated by BET inhibitors from these genes and thus dampening cell viability.” in Line 306.

Reviewer 4 Report

In the revised manuscript authors addressed some but not all critical points. Mostly, they included some information to the material and methods but the do not include any new data. My major concern thus remains unresolved. One of the main statements of the paper is that BRD4 hyperphosphorylation by CDKs happens outside mitosis. Unfortunately this statement is not supported by experimental data. Authors did not include the experiment with a 4D mutant that could support their hypothesis that BRD4 phosphorylation contributes to resistance to BET inhibitors. Instead they argue that this mutant could induce intra-molecular interactions such as the ones observed in Wu et al 2013. This seems unlikely as CDKs will induce other set of phosphorylations than CK2-mediated phosphosites. Overall, the revised manuscript is almost identical as the original version and I cannot recommend its publication.

Author Response

Reviewer 4

In the revised manuscript authors addressed some but not all critical points. Mostly, they included some information to the material and methods but the do not include any new data. My major concern thus remains unresolved.

 

Response: As we highlighted in our response to the first round of reviews, we have added new data in Fig. S2, S3, S4, and S5 to further support the conclusion of our manuscript.

 

One of the main statements of the paper is that BRD4 hyperphosphorylation by CDKs happens outside mitosis. Unfortunately this statement is not supported by experimental data.

 

Response: We apologize for causing this confusion for the reviewer. In Fig. 3D, we showed that overexpression of CDK1 functional partner, Cyclin B1, stimulates BRD4 hyperphosphorylation outside of mitosis in asynchronous cells. Also CDK1 is frequently overexpressed or activated in cancer cells (see citations 68-71 of the manuscript). In addition, we have previously shown that BRD4 is hyperphosphorylated throughout the cell cycle in many different types of cancers (Wang et al., PNAS 2017, 114, E5352-e5361.).  Based on these findings, we hypothesized that dysregulated CDK1 activation in cancer cells has the potential to trigger aberrant BRD4 hyperphosphorylation that persists outside of mitosis, supporting stronger chromatin binding and aberrant induction of its mitotic-bookmarked oncogenes, thereby driving tumor growth and BETi resistance. This is our hypothesis not our conclusion. We have modified the manuscript in Line 22 and Line 470 to make this clear.

 

Authors did not include the experiment with a 4D mutant that could support their hypothesis that BRD4 phosphorylation contributes to resistance to BET inhibitors. Instead they argue that this mutant could induce intra-molecular interactions such as the ones observed in Wu et al 2013. This seems unlikely as CDKs will induce other set of phosphorylations than CK2-mediated phosphosites. Overall, the revised manuscript is almost identical as the original version and I cannot recommend its publication.

 

Response: Our main reason for using 4A mutant instead of 4D mutant in these first set of experiments is to get a loss-of-BRD4-phosphorylation phenotype and then determine if this loss of function phenotype synergizes with CDK1 inhibitor to achieve the synergistic effect in blocking BRD4 function. As we discussed in our response to the first round of reviews, we agree with the reviewer that it is a good idea to test 4D mutant and we will consider this experiment in our future studies.

 

Round 3

Reviewer 1 Report

The response of the authors is not sufficient. If they refuse to do the Phostag experiment, I do not believe that they can state in the abstract or anywhere else that "In this study, we discovered that BRD4 is hyperphosphorylated by CDK1 during mitosis and identified determined the major CDK1 phosphorylation sites in BRD4."

It is a very easy experiment to do. Especially, the claim that other reviewers did not ask it, is highly unusual. I have never seen anyone having such an argument in review response. It is a very easy experiment, and this sounds like a very desparate answer.

Purity of the kinase from insect cell is not an excuse. There have been hundreds of kinases purified from insect cells and used for research. One can use inhibitors, high-sensitive staining etc to exclude the impurities. If kinase field would take such arguments seriously, we would lag behind in our knowledge of signaling for couple of decades.

So, I am not confident that authors can claim this central conclusion, as other kinases that can also be involved.

 

Reviewer 4 Report

Opinion of the reviewer on the manuscript has not changed. Authors still misinterpret their own data and data from others concerning CDK1 activation outside mitosis. Overexpression of the protein does not mean it is enzymatically active and that this happens outside mitosis. Asynchronic cultures contain substantial amounts of mitotic cells so authors cannot use experiments in asynchronic cells to conclude that modification occurs in interphase. Also in the opinion of the reviewer phenotype of the 4D mutant needs to be shown in this manuscript not in the future study. How does mitotic phosphorylation of BRD4 affect its function is not addressed in the manuscript. Clearly much more data is required to support all conclusions of the authors.