Tumour-associated missense mutations in the dMi-2 ATPase alters nucleosome remodelling properties in a mutation-specific manner

ATP-dependent chromatin remodellers are mutated in more than 20% of human cancers. The consequences of these mutations on enzyme function are poorly understood. Here, we characterise the effects of CHD4 mutations identified in endometrial carcinoma on the remodelling properties of dMi-2, the highly conserved Drosophila homologue of CHD4. Mutations from different patients have surprisingly diverse defects on nucleosome binding, ATPase activity and nucleosome remodelling. Unexpectedly, we identify both mutations that decrease and increase the enzyme activity. Our results define the chromodomains and a novel regulatory region as essential for nucleosome remodelling. Genetic experiments in Drosophila demonstrate that expression of cancer-derived dMi-2 mutants misregulates differentiation of epithelial wing structures and produces phenotypes that correlate with their nucleosome remodelling properties. Our results help to define the defects of CHD4 in cancer at the mechanistic level and provide the basis for the development of molecular approaches aimed at restoring their activity.

In this paper, the authors systematically characterise C HD4 mutations found in endometrial cancers. The interesting finding is that they identify mutations that increase, as well as decrease, the enzyme's activity. The in-vitro characterisation of the mutants is complemented by in-vivo experiments in Drosophila, which show that the expression of cancer-derived C HD4 mutants is sufficient to misregulate differentiation of epithelial structures. I thought this was a carefully designed and well thought out study, which is also clearly presented, and it deserves to be published.
In the Abstract, I think the authors should conclude with "Our results define the defects of C HD4 in cancer…".
In the Introduction, the authors argue that C HD4 mutants do not contribute to disease by lowering C HD4 activity (haploinsufficiency). However, it is not clear to me why the results don't simply suggest that C HD4 levels are critical and that both down-regulating and up-regulating NuRD activity could lead to misregulation of genes and disease? In fact, to me it would seem that all the evidence points towards this being the case, and I think this would also be consistent with earlier work (much of it from the authors) where either over-expression or knock down of C HD4 results in altered chromosome structure. C ould the dominant-negative and gain-of-function effects associated with the expression of C HD4 mutants in Drosophila (that the authors argue for in both the Introduction and Discussion) not involve either down-or up-regulation of NuRD? I would like to see some discussion of how C HD4 mutations could alter NuRD (not just C HD4) activity.
In the results, the authors use a gel shift assay to analyse binding of wild-type and mutant dMi-2 proteins to a mononucleosome using an electrophoretic mobility shift assay ( Figure 2B). It was not clear to me that they are seeing true complex formation here. It looks as if that, with increasing protein concentration, everything aggregates and the proteins and nucleosomes no longer run into the gel. I think these experiments need to be repeated perhaps with a different type of gel (maybe agarose?) which allows them to demonstrate that the complexes run into the gel and show where the different complexes formed by the wild-type and mutant proteins run. Alternatively, if they cannot get good data using EMSA's perhaps it would be sufficient to rely on the other assays?
In summary, however, I think this is a nice paper and, with some revision, suitable for publication.

Reply to reviewers
We thank both reviewers for carefully considering our manuscript, for finding our data convincing and important and for their constructive criticism. We have prepared a revised version of our manuscript that contains additional experimental results and incorporates the reviewers' s uggestions. Please, find a detailed point-by-point response to the reviewers' comments below: We do not claim that the changes in remodeling activities of the mutants are the sole reason for the wing phenotype and human cancer pathologies. Indeed, we have noted that overexpression of wild type dMi-2 in the developing wing is already s ufficient to elicit a gain-of-PCV phenotype ( Figure 5). In the revised version we have included additional genetic experiments (s ee below and new Supplementary Figure 6) which s upport our findings. In addition, we explicitly s tate in the discussion that the cancer pathologies are the result of a combination of mutations and that it is unclear to what extent CHD4 mutations contribute: "Cancer-associated dMi-2 mutants disrupt differentiation of wing structures Endometrial cancer arises when epithelial cells of the uterus become transformed and overprolif erate. In serous endometrial carcinoma, the most aggressive from of this cancer, CHD4 mutations are most f requent and occur alongside mutations in tumour suppressor or oncogenes including TP53, PIC3CA , PPP2R1A or FBXW7 6 . The individual contribution of CHD4 mutations to malignant transformation is, theref ore, difficult to assess." The authors have already generated flies carrying these mutants, it should be relatively easy to provide more insights on the in vivo molecular mechanisms underlying the wing phenotypes. I therefore recommend publication with a revision to strengthen the in vivo data in the manuscript. This is a very good s uggestion. We have tes ted expression of a set of eight genes encoding TGF-beta/BMP components in imaginal wing discs of transgenic flies by RT-qPCR. We have determined expression levels in wing imaginal discs overexpressing dMi-2 WT, a dMi-2 mutant with increased nucleosome remodeling activity in vitro (dMi-2 H1198Y) and a dMi-2 mutant with s trongly reduced remodeling activity (dMi-2 R1164Q). Out of eight genes encoding TGFbeta/BMP signaling components tested, s ix dis played s tatistically s ignificant increases in expression ranging from 1.3-fold to 2-fold relative to the control. These effects are mild but it is important to note that the engrailed driver us ed expresses the transgene exclusively in the posterior half of the wing imaginal disc, whereas RNA for RT-qPCR analysis was prepared from the entire disc. As a consequence, RNA levels measured reflect an average from transgene-expressing and non-expressing cells, thereby reducing the magnitude of gene expression changes. Intriguingly, upregulation of gene expression was detected irrespective of whether the dMi-2 transgene encoded a wild type enzyme, an enzyme with increased or an enzyme with decreased remodeling capacity. This correlates with the observation that expression of all dMi-2 trans genes resulted in gain-of-PCV phenotypes ( Figure 5). It is conceivable that upregulation of TGFbeta/BMP s ignaling genes and the gain-of-PCV phenotype are the result of dMi-2 overexpression via a mechanism that is independent of dMi-2 remodeling activity. Expres s ion of remodeling-defective dMi-2 generates additional loss-of-PCV phenotypes that were not observed upon expression of remodeling-competent dMi-2 ( Figure 5). Given that overexpression of both remodeling-defective and remodeling-competent dMi-2 produces similar changes to the genes we have tes ted it is unlikely that these gene expression changes are causally related to the loss-of-PCV phenotype.
These new results are described on p.11 of the revised manuscript: "PCV dif f erentiation is very sensitive to changes in BMP/TGFbeta signaling. In order to assess if expression of genes encoding components these signaling pathw ays were derepressed by ectopic expression of dMi-2 w e measured RNA levels in developing w ing imaginal discs. We prepared RNA from w ing discs from control larvae that carry a UAS-dMi-2 WT transgene but do not express GAL4 (UAS-dMi-2 wt) and larvae overexpressing dMi-2 WT (en-GAL4>>UAS-dMi-2 wt), the hyperactive remodeler dMi-2 H1196Y (en-GAL4>>UAS-dMi-2 H1196Y) or the remodeling-def ective dMi-2 R1162Q mutant (en-GAL4>>UAS-dMi-2 R1162Q). We then perfomed RT-qPCR analysis (Supplementary Figure 7). Measuring dMi-2 mRNA levels conf irmed overexpression of the three dMi-2 transgenes. We then tested expression of the BMP/TGFbeta genes b rk, Mad, Med, bi, cv, gbb, dpp and tkv. Tw o of these genes (brk and Med) did not show a signif icant upregulation of RNA levels in w ing discs expressing dMi-2 transgenes. The six remaining genes w ere upregulated between 1.25-and 2-fold with tkv showing the strongest effect. While these expression changes are mild it is w orth noting that while RNA was prepared f rom w hole w ing imaginal discs the engrailed promotor drives expression of dMi-2 transgenes in the posterior half of the wing disc only, suggesting that the observed expression changes are an underestimation. Interestingly, genes encoding BMP/TGFbeta components were similarly upregulated by expression of WT, hyperactive and remodeling-defective dMi-2 proteins. This correlates with the observation that all three types of dMi-2 proteins generate gain of PCV phenotypes and indicates that this phenotype may be caused by expression of ectopic dMi-2 irrespective of its remodeling properties.
Taken together, our results demonstrate that expression of cancer-derived dMi-2 point mutants in the background of endogenous, wild-type dMi-2 protein upregulates genes encoding BMP/TGFbeta components and is suf ficient for derailing the dif f erentiation of epithelial structures." We agree with the reviewer that it would be very interesting if the changes in gene expression observed in the imaginal wing disc would reflect the abnormal gene expression observed in human cancers. Our results do not support s uch far reaching conclusions. However, it is important to s tress that in endometrial cancer mutations in CHD4 (17% frequency) are generally accompanied by mutations in other genes such as TP53 (71%), PIK3CA (31%), PPP2R1A (25%) and others which also contribute to gene expression changes. In our experimental sys tem we investigate the effects of mutations in dMi-2 only. While this informs us about the general effects of dMi-2 mutations on epithelial differentiation and gene epression our experimental system does not recapitulate endometrial cancer generation and progression.
It will also be advisable if the authors can map the in vivo nucleosome patterns on these affected genes in the mutants to understand the role of the remodeling activities.
This would be a very good experiment in principle. However, we feel that this is outside of the scope of our manuscript. The experiment would require us to dissect wing discs and perform (e.g.) an MNas e-ChIP experiment. Indeed, we have used s uch an assay in the past to study the effects of dMi-2 depletion on chromatin accessiblity of ecdysoneresponsive genes in S2 cells (Kreher et al., Nature Communications 2017). However, S2 cells are a homogeneous cell population whereas wing discs are not. In fact, as explained above only the pos terior half of the wing disc is expressing the transgene resulting in any effects on nucleosome positioning (and these effects are not dramatic even in S2 cells where dMi-2 has been almost completely depleted; Kreher et al. 2017) being diluted by signals from unaffected cells.
Provided that expressing the wild type Mi-2 alone already gives a wing phenotype due to overexpression, it may be better to express the wild-type/mutants in a heterozygous Mi-2 null-mutant b ackground. It can imitate the heterozygous CHD4 mutant situations in human cancers and it hopefully also reduce the overall Mi-2 level to close to physiological endogenous level.
We agree with the reviewer that expression levels of the transgenes and the ratio between mutant and endogenous WT dMi-2 could influence the wing phenotypes observed. A heterozygous s ituation would indeed mimick the s ituation in cancer cells most closely. However, the generation of heterozygous fly lines would be cumbersome and time consuming. We have chosen a s impler approach that addresses the s ame issue: The activity of GAL4 in flies is highly temperature dependent. We have conducted the fly experiments at the normal temperature of 29 0 C which affords optimal GAL4 activity (Duffy, Genesis, 2002). By contrast, at 18 0 C GAL4 activity is minimal (Duffy, Genesis, 2002). A wide range of transgene expression levels can be realised by simply varying the temperature at which the flies are kept. In contrast to transgene expression, expression of endogenous genes is largely independent of temperature. Accordingly, lowering the temperature does not only reduce transgene expression but also lowers the ratio between ectopic and endogenous dMi-2 protein, thereby bringing it clos er to a heterozygous scenario. We have repeated the UAS/GAL4 crosses at 18 0 C (new Supplementary Figure 6). As expected, this reduced the general penetrance of the phenotypes. By contrast, it did not alter the qualitative and quantitative differences between remodeling-competent and remodeling deficient dMi-2 mutants: Firs t, expression of all WT and mutant dMi-2 trans genes generated gain-of-PCV phenotypes irrespective of whether the encoded enzyme was remodeling-competent od remodeling-defective. Second, dMi-2 H1196Y, which displays increased remodeling activity in vitro, still displayed the highest penetrance of gain-of-PCV phenotypes. In fact, the penetrance of dMi-2 H1196Y-induced gain-of-PCV phenotypes at 18C (54%) was s till greater than the penetrance of dMi-2 WT-induced phenotypes at 29C (48%). Third, loss-of-PCV phenotypes, although detected in a much s maller proportion of flies, were still restricted to the expression remodeling-defective mutants. Thus, PCV phenotypes generated by expression of ectopic dMi-2 proteins are robust over a wide range of transgene expression levels and ratios of endogenous vs . ectopic dMi-2 expression levels. We thank the reviewer for this excellent suggestion. We have carried out both experiments (mononucleosome binding and ATPas e assays). These results are included as the new Supplementary Figure 4 in the revised manuscript. We have tes ted binding to three types of nucleosomes: 0-77 nucleosomes (with a 77bp overhang on one side of the nucleosome), 0-44 nucleosomes and 0-22 nucleosomes. Both dMi-2 WT and dMi-2 H1196Y bind most efficiently to 0-77 nucleosomes. Binding becomes less efficient when the DNA overhang is shortened. Interestingly, binding of dMi-2 H1196Y is more s trongly affected by the shortening of the DNA overhang than binding of dMi-2 WT: In agreement with our results with the 0-80 nucleosome s hown in Figure 3C s imilar amounts of dMi-2 WT and dMi-2 H1196Y are required to s hift the 0-77 nucleosome (lower panel). By contrast, compared to dMi-2 WT, twice the amount of dMi-2 H1196Y is required to s hift the 0-22 nucleosome (upper panel). This s upports the reviewer's hypothesis that dMi-2 H1196Y s enses DNA length differently than dMi-2 WT and could contribute to the apparent loss of directionality in this mutant. We also measured the ATPas e activities of dMi-2 WT and dMi-2 H1196Y in presence of the three types of mononucleosomes (new Supplementary Figure 4). This revealed no differences in ATPas e activities. Finally, we also attempted to perform nucleosome s liding assays with the 0-22 nucleosome that displays the s trongest difference in dMi-2 binding affinities. However, due to the s hort length of the overhang into which the histone octamer can be moved we failed to adequately resolve remodeled and non-remodeled nucleosomes with our gel sys tem (data not shown).

Authors reported an interesting observation that H1196Y mutation renders
We have added this data as a new Supplementary Figure 4 and have described the results on p.9 of the revised manuscript: "In addition, f or some remodelers the direction of nucleosome sliding is determined by the length of the tw o DNA segments extending f rom the nucleosome core particle 34 . In order to investigate if dMi-2 H1196Y senses the length of these DNA extensions differently than dMi-2 WT w e tested binding to mononucleosomes w ith 22bp (0-22 nucleosome), 44bp (0-44 nucleosome) and 77bp (0-77bp nucleosome) extensions (Supplementary Figure 4A). Shortening the DNA segment extending f rom the core decreased binding af finity for both dMi-2 WT and f or dMi-2 H1196Y . How ever, dMi-2 H1196Y was more strongly affected by this (Supplementary Figure 4A , compare upper and low er panels). These dif f erences in binding af f inity did not translate into dif ferences in A TPase activity. Both dMi-2 WT and dMi-2 H1196Y A TPases were equally stimulated by all three types of mononucleosomes (Supplementary Figure 4B). Taken together, the results of our remodeling and nucleosome binding assays are consistent w ith the hypothesis that dMi-2 H1196Y senses DNA extending f rom the nucleosome core in a dif ferent manner. Thus, dMi-2 H1196Y might respond to the DNA sequence of nucleosomal DNA and/or the length of DNA extending f rom the nucleosome differently resulting in altered nucleosome positioning."

Minor suggestions: 1. Simply increasing wild type Mi-2 protein amount is enough to cause a phenotype in the wing. This suggests that total protein level is important to interpret the result. The authors may provide western b lot showing protein level of exogenous mutated Mi-2 and endogenous Mi-2 in these transgenic flies. The mutants are expressed from the same genomic loci, b ut the protein stability of different mutants may vary.
We agree with the reviewer that wing phenotypes might be influenced by expression levels . As explained above, we have carried out the UAS/GAL4 cros ses at 18°C and 26°C which reflect low and high expression levels, respectively. While the penetrance of the phenotypes were indeed influenced by expression levels, the phenotype itself (gain-of-PCV versus loss-of-PCV) was not. Thus, we are confident that the phenotypes reflect b ona fide properties of the dMi-2 enzymes tes ted.

The title and abstract may cause misunderstanding. Almost the whole manuscript is b ased on the Drosophila protein Mi-2, b ut the abstract led me one think that the in vitro experiments were done with human proteins and human mutations. Drosophila was only mentioned in the genetic experiments. It would be good the make that clearer.
We did not intend to confuse the reader. We have now changed the title to read "Tumour-associated missense mutations in the dMi-2 A TPase alters nucleosome remodelling properties i n a mutation-specif ic manner" We have also changed the abstract: "A TP-dependent chromatin remodellers are mutated in more than 20% of human cancers. The consequences of these mutations on their enzymatic activities and their in vivo f unction are poorly understood. Here, we systematically characterise the effects of CHD4 mutations identified in e ndometrial carcinoma on the remodeling properties of dMi-2, the highly conserved Drosophila homolog of CHD4. Mutations from different patients have surprisingly diverse defects on nucleosome binding, A TPase activity and nucleosome remodelling. Unexpectedly, we identif y mutations that decrease as w ell as mutations that increase enzyme activity. Our results define both the chromodomains and a novel regulatory region adjacent to the ATPase domain as essential f or nucleosome remodelling. Genetic experiments in Drosophila demonstrate that expression of cancer-derived dMi-2 mutants is sufficient to misregulate dif f erentiation of epithelial w ing structures and produces phenotypes that correlate with their nucleosome remodelling properties. Our results help to define the def ects of CHD4 in cancer at the mechanistic level and provide the basis f or the development of molecular approaches aimed at restoring their activity."

It is not clear to me if how L1215P disrupt nucleosome binding. Does the b race II region b ind directly to nucleosomal DNA and L1215P disrupt the interaction with the nucleosome, or b race II bind to the ATPase core to inhibit its interaction with the nucleosome, and L1215P enhance the interaction with ATPase. Does the b race II region alone binds to nucleosome? Mayb e the authors can elaborate.
We currently do not have an answer to these questions. In related remodelers, s uch as Snf2, recent structural data s uggests that the brace-I and brace-II regions are disordered and adopt their alpha-helical structures when the enzme binds to the nucleosome (Liu et al, Science, 2017). In this case, the two helices protrude from the core1 domain and make direct contact with the core2 domain. From this s tructural work, there is no indication that brace-II directly contacts the nucleosomal DNA. The L1215P introduces a proline which is likely to disrupt the formation of the corresponding alpha-helix, potentially affecting the core1-core2 interaction in dMi-2/CHD4. Thus, we would hypothesise that this mutation indirectly reduces nucleosome binding by causing s tructural changes to the core1-core2 conformation. However, this hypothesis will have to be validated by analysing the structure of dMi-2/CHD4 its elf, an undertaking that is outside of the s cope of this manuscript.

It is unclear how the mutation in the PHD domain may affect the ATPase activity.
In order to understand the function of this mutation, the authors may try to use nucleosomes that are methylated at H3K4 residue to perform their b inding assay and ATPase assay. Only a minor fraction of endogenous histone octamer extracted from the cell contains H3K4me3. If it has b een reported before in the literature, please elaborate.
Indeed, analys es of the role of the CHD4 PHD fingers in nucleosome binding and ATPas e function have been published by several labs (including the Laue, Mancini, Kutateladze and Mackay labs). The PHD fingers bind to the H3 tails of nuceosomes both individually and cooperatively in a modification s ensitive manner where K4 methylation generally decreases and K9 methylation and K9 acetylation enhances binding. While we agree with the reviewer that it would be informative to tes t the activity of the PHD finger mutant on recombinant nucleosomes carrying defined methylation patterns we feel that this is not the focus of our study. However, we have changed the discussion of this mutant to better reflect the current s tate of knowledge by elaborating on the work of the above mentioned labs.
The relevant section in the discussion now reads as follows.
"Mutation of the zinc binding C464 resulted in a mild reduction in dMi-2 A TPase and nucleosome remodelling activity. These results are consistent with the weak effect observed upon expression of this mutant in the developing w ing ( Figure 5) and with our previous finding that deletion of the PHD f ingers of dMi-2 has only mild consequences for ATPase and nucleosome sliding activities 22 . A ddition of the PHD f ingers to a CHD4 construct containing chromodomains and A TPase domain moderately increases A TPase and remodeling activities 34,39 . In some cases these effects are inf luenced by the nucleosome methylation status 34 . Indeed, the PHD f ingers of CHD4 bind nucleosomes both individually and cooperatively in a H3 methylation-sensitive manner 35,40 . The nucleosomes we have used in this study w ere assembled w ith histone octamers purified f rom Drosophila embryos. These contain a mixture of unmodif ied and methylated nucleosomes. It is, theref ore, conceivable that the PHD f inger mutation has a stronger effect in the context of the human CHD4 protein or in the presence of nucleosomes carrying def ined H3 methylation patterns. We can also not exclude that this mutation affects intra-or intermolecul ar interactions that are not detected by our in vitro assays." In this section we cite the following papers:

Figure 1 and 2 could be easily comb ined.
This is in principle true. However, we prefer to devote one figure to each class of mutants and keep Figure 1 s eparate as an "overview". The total number of figures in our manuscript is not so large that we deem s uch a consolidation neccessary.

Reviewer #2 (Remarks to the Author):
Referee's Report for the manuscript entitled "Tumour-associated missense mutations in CHD4 ATPases alter their nucleosome remodelling properties in a patient-specific manner" by Kristina Kovač et. al. In this paper, the authors systematically characterise CHD4 mutations found in endometrial cancers. The interesting finding is that they identify mutations that increase, as well as decrease, the enzyme's activity. The in-vitro characterisation of the mutants is complemented by in-vivo experiments in Drosophila, which show that the expression of cancer-derived CHD4 mutants is sufficient to misregulate differentiation of epithelial structures. I thought this was a carefully designed and well thought out study, which is also clearly presented, and it deserves to be published.
In the Ab stract, I think the authors should conclude with "Our results define the defects of CHD4 in cancer…".
We have followed the reviewer's suggestion and have changed the abstract accordingly.
In the Introduction, the authors argue that CHD4 mutants do not contribute to disease by lowering CHD4 activity (haploinsufficiency). However, it is not clear to me why the results don't simply suggest that CHD4 levels are critical and that b oth down-regulating and upregulating NuRD activity could lead to misregulation of genes and disease? In fact, to me it would seem that all the evidence points towards this b eing the case, and I think this would also be consistent with earlier work (much of it from the authors) where either overexpression or k nock down of CHD4 results in altered chromosome structure.
The reviewer makes a very good point. Indeed, we cannot rule out that the CHD4 mutations contribute to disease simply be altering CHD4 activity levels. Our reasoning is based on the absence of any major deletions in the CHD4 gene in endometrial cancer (s ee LeGallo et al., 2012) that would likewise lower CHD4 activity levels. Therefore, we favour the hypothesis that the point mutated dMi-2 enzymes exert dominant negative or gain of function effects. Nevertheless, we agree that both scenarios merit equal cons ideration. We have changed the relevant s ection in the introduction to: "How ever, it is not know n whether and how disease-associated CHD4 mutations affect its enzymatic activities at the molecular level and how this impacts the epigenetic landscape, gene expression and genome stability. A nalysis of the CHD4 mutation spectrum reveals two remarkable f eatures: First, the majority of CHD4 alterations are missense mutations (89% in endometrial carcinoma) 4,6,7 . Deletions, f rameshift and nonsense mutations that would result in a complete loss or a truncated CHD4 protein are rare. Second, patients are heterozygous for CHD4 missense mutations and retain one w ild type copy of CHD4 4,6,7 . It is possible that mutations in one of the two dMi-2 alleles lowers the total CHD4 activity in affected cells sufficiently to result in the misregulation of genes. In addition, CHD4 point mutants m ight contribute to disease not or not only by lowering overall CHD4 activity: Rather, the absence of de letions, frameshift and nonsense mutations might suggest that defective CHD4 enzymes exert a dominant negative effect or that CHD4 mutations result in a gain of function." We have also changed a s ection in the discussion to give equal weight to both s cenarios: "It is clear that the majority of CHD4 mutations identif ied in endometrial cancers negatively impact A TPase and/or remodelling activity. It is conceivable that a reduction of overall CHD4 activity in endometrial cancer cells contributes to cancerogenesis. However, CHD4 gene deletions or f rameshift mutationsw hich would likew ise low er overall CHD4 activity -are very rare in endometrial cancer 6 . By contrast, missense mutations predominate (89%) arguing that it might also be the mutated CHD4 protein itself that contributes to cancerogenesis via gain-of-function or dominant negative mechanisms." Could the dominant-negative and gain-of-function effects associated with the expression of CHD4 mutants in Drosophila (that the authors argue for in b oth the Introduction and Discussion) not involve either down-or up-regulation of NuRD? I would like to see some discussion of how CHD4 mutations could alter NuRD (not just CHD4) activity.
We thank the reviewer for his s uggestion to include a discussion on the possible consequences of CHD4 mutations on NuRD activity. We have now included the following paragraph in the discussion: "The ef f ects of CHD4 mutations might also manif est themselves by an altered activity of the NuRD complex. NuRD combines CHD4 nucleosome remodeling w ith HDAC1/2 histone deacetylase activities.
Early w ork has suggested that remodeling is a pre-requisite f or efficient nucleosome deacetylation 2 . Thus, altered CHD4 activity is likely to have an impact not only on nucleosome positioning but also on histone acetylation." In the results, the authors use a gel shift assay to analyse b inding of wild-type and mutant dMi-2 proteins to a mononucleosome using an electrophoretic mobility shift assay ( Figure  2B). It was not clear to me that they are seeing true complex formation here. It looks as if that, with increasing protein concentration, everything aggregates and the proteins and nucleosomes no longer run into the gel. I think these experiments need to be repeated perhaps with a different type of gel (maybe agarose?) which allows them to demonstrate that the complexes run into the gel and show where the different complexes formed by the wild-type and mutant proteins run. Alternatively, if they cannot get good data using EMSA's perhaps it would be sufficient to rely on the other assays?
As explained in our response to reviewer #1 above, we have now performed additional bandshifts with a s eries of nucleosomes with s horter DNA overhangs (new Supplementary Figure 4). In particular, the 0-22 and 0-44 nucleosomes are s mall enough to allow the dMi-2/nucleosome complex to enter the gel (upper and middle panel). We conclude that dMi-2 does form a proper complex with the nucleosome and does not aggregate under our bandshift conditions. However, when the DNA overhang gets too long (as in the 0-80 nucleosomes used in Figures 2, 3 and 4) this complex is too large to enter the gel.
In summary, however, I think this is a nice paper and, with some revision, suitable for pub lication.