The chromatin remodeler DEK promotes proliferation of mammary epithelium and is associated with H3K27me3 epigenetic modifications
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (Review Commons)
Abstract
The DEK chromatin remodeling protein was previously shown to confer oncogenic phenotypes to human and mouse mammary epithelial cells using in vitro and knockout mouse models. However, its functional role in normal mammary gland epithelium remained unexplored. We developed two novel mouse models to study the role of Dek in normal mammary gland biology in vivo . Mammary gland-specific Dek over-expression in mice resulted in hyperproliferation of cells that visually resembled alveolar cells, and a transcriptional profile that indicated increased expression of cell cycle, mammary stem/progenitor, and lactation-associated genes. Conversely, Dek knockout mice exhibited an alveologenesis or lactation defect, resulting in dramatically reduced pup survival. Analysis of previously published single-cell RNA-sequencing of mouse mammary glands revealed that Dek is most highly expressed in mammary stem cells and alveolar progenitor cells, and to a lesser extent in basal epithelial cells, supporting the observed phenotypes. Mechanistically, we discovered that Dek is a modifier of Ezh2 methyltransferase activity, upregulating the levels of histone H3 trimethylation on lysine 27 (H3K27me3) to control gene transcription. Combined, this work indicates that Dek promotes proliferation of mammary epithelial cells via cell cycle deregulation. Furthermore, we report a novel function for Dek in alveologenesis and histone H3 K27 trimethylation.
Article activity feed
-
Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.
Learn more at Review Commons
Reply to the reviewers
Response to Reviewers
We thank the reviewers for their fair and thorough review. With regards to the reviewers’ comments, both largely focused on something that is a misunderstanding. For unclear reasons, both reviewers thought that most of the data shown was in pregnant or previously pregnant mice and both requested a significant amount of preliminary data regarding virgin mice (R1 comment #3, #4, R2 comment #1). This may be due to a (now-corrected) typo in the results section despite the methods section being correct, or the very few instances of pregnant mice being used for analyses that led to confusion. As mentioned below, the entire manuscript …
Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.
Learn more at Review Commons
Reply to the reviewers
Response to Reviewers
We thank the reviewers for their fair and thorough review. With regards to the reviewers’ comments, both largely focused on something that is a misunderstanding. For unclear reasons, both reviewers thought that most of the data shown was in pregnant or previously pregnant mice and both requested a significant amount of preliminary data regarding virgin mice (R1 comment #3, #4, R2 comment #1). This may be due to a (now-corrected) typo in the results section despite the methods section being correct, or the very few instances of pregnant mice being used for analyses that led to confusion. As mentioned below, the entire manuscript evaluates virgin mice, with a few specific exceptions, so the preliminary revisions have emphasized the parity status of the mice used in every experiment. We regret this misunderstanding happened and we are concerned this may have led to reviews that were biased towards a negative viewpoint. We hope the completed preliminary revisions (indicated in red text in the manuscript) and the planned revisions will, combined, satisfy the reviewer’s concerns and clarify points of confusion, while leading to a greatly improved manuscript.
Reviewer 1:
Major points
Major Comment 1: “Several of the conclusions are made based on a limited number of replicates (often n=3) which is not a robust sample size to make a rigorous conclusion.”
We have consulted with a biostatistician (Adam Lane, now included in acknowledgements) and plan to add at least 3 more mice per group to bring the total sample size to 6-7. Given our results are already statistically significant with an n=3-4, we do not anticipate any changes in the overall results of our data. We have already collected at least 3 more age-matched and parity-matched mice per group for the molecular analyses and are working on performing the immunohistochemical stains, western blots, etc.
Major Comment 2: The main text for Figure 1C mentions repression of luciferase expression by doxycycline chow, however the figure does not show any discernable repression in the Dek-OE conditions.
We believe the reviewer may have mis-interpreted the figure. The mouse on the far left (“control”) with no luciferase signal is the dox chow-repressed condition. We have revised the figure label to specify that “Control” is the “+dox condition” and throughout the manuscript have specified “+dox controls” instead of just “controls.”
Major Comments 3: To evaluate the impact of prolonged Dek overexpression on mammary epithelium in Figure 1G and 1H, the authors used multiparous females. One confounding factor with this experimental set up is the impact of previous pregnancies on the development of the mammary epithelium and in lowering tumorigenesis. Therefore, the impact of Dek on tumorigenesis cannot be determined in multiparous animals alone. To get a full picture, nulliparous animals should also be examined.
__We have revised the text on page 6 to explain that we have monitored tumor growth in both aged virgins and in multiparous mice (our female breeders) and neither group develops tumors. __
Major Comment 4: “To elucidate the molecular underpinning of Dek-OE phenotypes, the authors performed bulk RNA sequencing in Figure 2. Similar to point 2 however, only multiparous animals were used. As it has been previously shown that pregnancy significantly impacts the transcriptome of mammary glands, the effects of Dek overexpression can't be generalized to mammary glands as a whole. To make it generalizable, nulliparous Dek-OE animals must also be characterized.”
As mentioned in the introduction to the review, the reviewer has misunderstood the experimental design, perhaps through a single typo in the Results section when the Methods were correct, or through poor writing on our behalf. Regardless, the RNA-Seq, whole mounts, and all subsequent molecular validations were conducted on virgin mice. The only exceptions are in Figure 4, where we do explore the expression of endogenous Dek during pregnancy and the impact of pregnancy in the transgenic model. We have revised the typographical error, confirmed the parity status of all mice in the study to date, and have specifically added the parity status to each experiment in Results section and/or Figure Legend.
Major Comment 5: To validate findings from their transcriptomics work, the authors used IHC and western blots of candidate proteins that were found to be down regulated. In Figure 3A and 3C, the decrease in p21 protein levels through western blot seem much more modest than what the decrease seen in 3A would suggest.
We thank the reviewer for pointing this out. With increased sample sizes, as requested, we hope this will resolve. We plan to increase sample size and quantify the p21 western blot to potentially resolve the concern. In addition, we would like to note that the p21 IHC is specific for mammary epithelium signal while the western blot is whole mammary gland lysate that includes quiescent stromal cells, which may explain the slight discrepancy between the two methods.
Major Comment 6: In Figure 3G-3I, the authors test the CDK4/6 inhibitor palbociclib to establish a direct link between the phenotypes seem in Dek-OE and cell cycle progression in organoid culture. Have the authors verified these findings with treatment of Dek-OE mice with palbociclib? In addition, have the authors checked to see if palbociclib corrected any of the transcriptional features associated with the Dek-OE model found in their transcriptomics data? In addition, the authors claim that the effect is specific to Dek-OE organoids as the effects of palbociclib on growth are not seen in control organoids. However, the data on unperturbed growth of control cells are not seen. To determine the specificity of the effects of palbociclib on Dek-OE derived organoids, the authors must show a time course tracking the growth of organoids with and without palbociclib. Rather than conclude the effects of palbociclib being specific to Dek-OE organoids, the authors most likely wanted to conclude that the increased growth of Dek-OE organoids compared to control organoids is dependent on the increase in cell cycle factors. (The validity of this is also weird though because even if division and growth were triggered through other transcriptional changes they found, like increased metabolism, growth in that scenario would be stopped by palbo as well)
- Because the hyperplasia phenotype accumulates over the lifetime of the animal, the amount of treatment time required to abrogate the hyperplasia phenotype could be from days to weeks to months. For this reason, we believe it is outside the scope of this revision to test the effects of palbociclib in vivo.
- We plan to re-do this experiment with palbociclib treatment to test organoid growth over time as suggested and, time permitting, perform immunofluorescence for some of the transcription targets such as cyclins, CDKs, Ki67, and p27/p21
- We have revised the text on page 8 to say “____We observed that the increased growth of Dek over-expressing organoids was dependent on the Dek-induced increase in CDK4/6, since palbociclib treatment resulted in smaller Dek over-expressing organoids that were comparable to organoids from +dox controls.” We also agree that CDK inhibitor treatment may impact multiple downstream signaling pathways. However, the authors do not see this as a negative because cell proliferation, induced by cyclin/CDK complexes, requires metabolic regulation to support physical growth of the cell. The two processes are intricately integrated and have a bidirectional relationship. Thus, it is possible that DEK induces both processes, or it may only promote one process (i.e.: cell cycle) and the other one (i.e.: metabolism) is induced as a secondary result of cell cycle demands. This is one reason why we indicate that metabolic dysregulation should be further studied in the Discussion section. Indeed, a colleague in the DEK field (Susanne Wells) is already working on the relationship between DEK over-expression and metabolic dysfunction, thus this particular aspect of the request is outside the scope of this manuscript.
Major Comment 7: In the main text of Figure 4, the authors conclude that markers for luminal hormone sensing cells were unchanged in Dek-OE mammary glands, however the data to show this is not shown. This is problematic because the authors are directly drawing the conclusion that Dek-OE specifically upregulates luminal alveolar markers using this data.
We have revised the manuscript to include a new supplementary figure (now Fig S4) to include a western blot for HER2 and ERa and a summary of RNA expression data from the bulk RNA-Seq experiment. We will also perform additional western blots to increase the sample size to demonstrate this negative data as part of our planned
Major comment 8: In figure 7, the authors look at a conditional knockout of Dek and conclude that pup death in the knockout was due to insufficient milk production by dams. While the authors establish that H3K27me3 and Ezh2 expression are abrogated, morphological analysis of the ducts is missing and would present convincing data. For instance, in the Dek conditional knockout, are luminal alveolar cells unable to differentiate fully, or are there far fewer? Decreased levels of histone modifications does not tell you much about whether repressive chromatin has changed its landscape in Dek KO mice, which is actually what influences transcription.
__We plan to add histological and whole mount imaging of Dek knockout mammary glands in the revision. We have preliminary data that supports this from 2 mice and will be collecting more samples for the revision. However, as noted in Fig 7C-D, heterozygous females also have small litter sizes and this will pose a breeding challenge for generating knockout females for this experiment in a timely manner. __
Minor Points:
All figures need some sort of reformatting. Several of the conclusions are made based on a limited number of replicates (often n=3) which is not a robust sample size to make a rigorous conclusion. Many figures have text that is stretched. Histology and whole mount images are missing scale bar. IHC quantifications are obscure - what is an optical density? how many animals were analyzed and how many fields of vision were captured? Figure 2F is absolutely impossible to understand. Neither figures nor legends disclose the number of animals or samples analyzed. The statistical test utilized across all figures is not appropriated. Fig5B GSEA plots are missing statistical significance, and without this information one cannot properly access the relevance of the findings. Fig5C - how were co-expressed genes defined? is this just random genes that are expressed in cells that have higher levels of DEK? The term co-expressed suggests a specific type of analysis that would investigate linkage of expression between genes, which i dont think is the case here.
__As the reviewer already mentioned in major comment #1, there was a concern with sample size, which we addressed above in the planned revisions. We believe this concern about sample size was the rationale for the minor comment about “The statistical test utilized across all figures is not appropriate.” We have consulted with a biostatistician, Adam Lane PhD, who has confirmed that our statistical approaches were correct but were limited by our sample size. Thus, we do not agree with the reviewer’s view of statistical analyses. We have revised the text to include sample size information in figure legends and statistical significance information for GSEA plots in Fig 5, With regards to figure text being stretched, it does not look like that in our version of the document and reviewer 2 did not comment on this, so we would like the reviewer to identify a specific instance of this. We plan to capture images with size bars for IHC while we are performing the additional sample size collection. The reviewer asked about the number of fields of view for IHC quantification and we would like to note that our methods section already had that information in the first submission, “at least 3 fields of view from at least 3 different mice per group.” Our methods section also already had information regarding the identification of co-expressed genes in scRNA-Seq data and quantifying IHC with Image J. However, we have revised the text to add some clarifying sentences that we hope helps the reviewer better understand our methods. Finally, we are not sure what is “absolutely impossible to understand” about Fig 2F, which is a network visualization of functional enrichment analyses for differentially expressed genes in our RNA-Seq data. Is the text too small, or does the reviewer not understand the network? We would appreciate it if the reviewer could clarify this concern in their next review. __
Minor Point 1: Throughout, it would be better to indicate the genotype of the "Control" animals on each figure so as the rigor the experiment can be evaluated fully.
It appears that the reviewer was not aware that all controls were the same genotype and were the bitransgenic mice on dox chow. We have revised the manuscript to better clarify that “controls” = “+dox chow” bitransgenics and have added text on page 5 to directly state this. We have also revised Fig 1C to specify that the mouse with no luciferase signal is the “+dox” control.
Minor Point 2: Standard nomenclature for gene names and protein names should be corrected throughout the text.
__We have revised the text to confirm gene and protein names are correct. We have followed convention in using italics for gene names, non-italics for protein names, all capital letters for human genes/proteins (i.e.: DEK) and only first letter capitalization for non-human gene names (i.e.: Dek). __
Minor Point 3: Similar to the point above, the use of Dek-OE to either refer to the mouse model or function as an acronym for "Dek overexpression" is inconsistent throughout the text.
We thank the reviewer for pointing out this inconsistency and we have revised the text so that the “-OE” notation is only used when discussing the mice and have changed to writing out “over-expression” for function.
Minor Point 4: In the main text for Figure 4I-J, the authors state that DEK was previously published as an Erα target gene, however there is no citation to support this.
We have revised the text to include this citation, which is:
#16. Privette Vinnedge, L.M., et al., The DEK Oncogene Is a Target of Steroid Hormone Receptor Signaling in Breast Cancer. PLoS One, 2012. 7(10): p. e46985.
Minor Point 5: It is unclear what the conclusion drawn from the experiments shown in Figure 4G-H and Figure 4I-J mean with respect to the goal of Figure 4, which was to show that Dek-OE mice have an expanded luminal alveolar compartment.
We have revised the text to better explain that we were investigating the impact of ovarian hormones and pregnancy on endogenous Dek expression in wild-type mice, since this information has not been previously reported and adds context to our study.
Minor Point 6: Optical density was used to quantify IHC experiments, which was performed using color deconvolution in ImageJ. Something that is unclear is whether the authors are measuring density in the entire field of view, or if the authors are measuring optical density per cell. This has implications whether there are more cell expressing the protein of interest, or if the existing cells are expressing a higher level of the protein of interest.
We have revised the text to include more information in the methods. The Methods now states: “____Image J color deconvolution was utilized to measure the staining intensity only within mammary epithelial cells from at least 3 fields of view from at least 3 different mice per group. Specifically, cross-sections of similarly sized ducts were outlined such that only the collective epithelial cells within that cross section were measured, removing background signal from the stromal cells. Only single cross-sections of ducts were analyzed to minimize the impact of epithelial hyperplasia in experimental mice compared to controls fed dox chow.”
Minor Point 7: In the main text for Figure 6D, the system being used to overexpress DEK protein is not described. It is not the same genetic system as is used in the Dek-OE mice, as doxycycline is inducing Dek expression.
We have revised the figure 6 legend to specify “____DEK over-expression was accomplished with a dox-inducible pTRIPZ vector while DEK knockdown was accomplished with a pLKO.1 shRNA vector” and we kindly point the reviewer to the Methods section (“human cell lines” subsection) as written in the first submission which included detailed information for the subcloning of DEK cDNA into the pTRIPZ vector.
Reviewer 2
____All comments____
Comment 1: This study would be improved by sharing important data including virgin mammary gland development in the DEK-OE and DEK-KO models (ductal growth and branching) and the expression of markers including ESR1, PGR, and ERBB2 (data not shown, page 8). Although there may be no differences, this is important data to share regarding the goal of this study. For example, in the DEK-OE model, data are only evaluated in the aged/multiparous stage and in the DEK-KO model, data are only evaluated during lactation. Furthermore, the DEK-KO model resembles germline DEK loss (under control of the CMV promoter), and there is limited validation of a MEC-intrinsic function.
We have revised the manuscript to include data on Esr1/ERa and Erbb2/Her2 by western blot in new Fig S4 as well as the bulk RNA-Seq mRNA levels (by FPKM) for select basal and hormone sensing cell populations. The concern regarding parity was also mentioned by Reviewer 1 (major comments 3&4 above). Briefly we have clarified that ____*nearly all *data in the manuscript is from nulliparous (virgin) females and have revised the text throughout to more clearly state this fact. We have also revised the text to address the limitation of the CMV promoter. The Discussion section now states “____However, it is noted that one weakness of this CMV-Cre knockout model, is that there is a constitutive loss of Dek, which limits the interpretation for mammary epithelial cell-specific Dek functions.” * *
Comment 2: Another major concern with this manuscript is the use of immunohistochemistry (IHC) and bulk mammary gland lysate western blots. IHC is non-quantitative, and the images are low resolution. For example, using IHC DEK expression is observed in all MECs (control and DEK-OE mice, Figure 1F), however, in the scRNAseq data DEK expression is confined to basal cells and a subset of stem/progenitor cells (Figure 5A). Furthermore, the hyperplasia in the DEK-OE model will bias bulk analysis (such as western blot and RNAseq) towards increased expression of MEC markers.
- __We have revised the text to point out that IHC images for Dek in control tissues show some cells have higher expression than others, which is what would be predicted by scRNA-Seq. The text now states on page 16 “____The scRNA-Seq data suggests that Dek is more highly expressed in specific subpopulations of cells, and the variable intensity of immunohistochemical staining for Dek in epithelial cells within control mouse tissue supports this (see Fig 3I, 4I, 4K, and 7H).” Furthermore, on page 10 in the Result section we have revised the text to state “The mammary gland undergoes substantial hormone-induced remodeling across the murine lifespan. We show that Dek is highest during pregnancy and minimally expressed during lactation and involution (Fig 4K), and that Dek protein expression is not uniform across all epithelial cells in wild-type glands (Fig 3I, 4I-K). This suggests that certain epithelial subpopulations express more Dek than others.” __
- __We acknowledge that IHC and western blots are only semi-quantitative, which is why we attempt to perform both as orthogonal approaches or find additional ways to support our findings throughout the manuscript (i.e.: co-expression at the RNA level from other sources, small molecule inhibitor treatment, etc). We also note that these methods are used to validate the quantitative method of RNA-Seq, and (often) validation of differentially expressed genes can be limited by antibody availability and the applications those antibodies are suitable for. __
- We also have revised the text to acknowledge that we knew the bulk RNA-Seq would be biased towards the hyperplastic cells. We wanted to take advantage of that bias to identify a gene signature that could be used to determine which cell type was leading to the hyperplasia phenotype. We used the differentially expressed genes to identify biomarkers for specific cell populations. On pages 6-7 the text now reads “____We performed bulk RNA sequencing on whole mammary tissue from two +dox control and two Dek-OE adult virgin females at 15 months of age to discover molecular targets regulated by Dek over-expression and to reveal a gene signature that could help identify the expanded cell population(s) in hyperplastic glands.” And “DEGs were plotted as a heatmap and ontologies for biomarkers of cell populations were defined to help identify the expanded cell population driving Dek-induced hyperplasia.”
Comment 3: A third major concern is the mechanistic link between DEK and H3K27me3. Most of the data are correlative and rely on bulk analysis or IHC. For example, in the DEK-OE organoid model, is there an increase in H3K27me3. Additionally, in the DEK-OE organoids, can loss of EZH2 block the increased cell proliferation?
__We plan to revise the manuscript to include an experiment in which we treat primary mammary epithelial cell organoids from Dek-OE mice with EZH2 inhibitor, GSK-126, +/- doxycycline for a mechanistic or functional link between DEK and H3K27me3 levels. We will then determine organoid size and attempt molecular characterization with IF. This will support the biochemical studies in Fig 6 showing DEK interacts with the PRC2 complex. __
-
Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.
Learn more at Review Commons
Referee #2
Evidence, reproducibility and clarity
In the manuscript "The chromatin remodeler DEK promotes proliferation of mammary epithelium and is associated with H3K27me3 epigenetic modifications", Johnstone et al. investigate the proto-oncogene, DEK, in control of normal mammary gland development. The authors utilize transgenic mouse models, including conditional DEK-overexpression (OE) and DEK-knockout (KO) model, highlighting the role for DEK in control of mammary epithelial cell (MEC) proliferation and differentiation during pregnancy and lactation. Furthermore, the authors demonstrate DEK expression correlates with EZH2 and H3K27me3, which have previously been reported to …
Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.
Learn more at Review Commons
Referee #2
Evidence, reproducibility and clarity
In the manuscript "The chromatin remodeler DEK promotes proliferation of mammary epithelium and is associated with H3K27me3 epigenetic modifications", Johnstone et al. investigate the proto-oncogene, DEK, in control of normal mammary gland development. The authors utilize transgenic mouse models, including conditional DEK-overexpression (OE) and DEK-knockout (KO) model, highlighting the role for DEK in control of mammary epithelial cell (MEC) proliferation and differentiation during pregnancy and lactation. Furthermore, the authors demonstrate DEK expression correlates with EZH2 and H3K27me3, which have previously been reported to control mammary gland lactation [Pal et al., Cell Reports, 2013].
Overall, this manuscript is interesting and well prepared. This group have previously established a role for DEK in breast cancer, however, the function of DEK in normal mammary gland development is unknown. Towards this goal, two novel mouse models were developed, conditional DEK-OE and DEK-KO. This manuscript would be substantially improved by formal investigation of virgin mammary gland development in these models, high-resolution analysis of the MEC subpopulations at different stages, and strengthening the mechanistic link between DEK and EZH2. The following are detailed major concerns.
- This study would be improved by sharing important data including virgin mammary gland development in the DEK-OE and DEK-KO models (ductal growth and branching) and the expression of markers including ESR1, PGR, and ERBB2 (data not shown, page 8). Although there may be no differences, this is important data to share regarding the goal of this study. For example, in the DEK-OE model, data are only evaluated in the aged/multiparous stage and in the DEK-KO model, data are only evaluated during lactation. Furthermore, the DEK-KO model resembles germline DEK loss (under control of the CMV promoter), and there is limited validation of a MEC-intrinsic function.
- Another major concern with this manuscript is the use of immunohistochemistry (IHC) and bulk mammary gland lysate western blots. IHC is non-quantitative, and the images are low resolution. For example, using IHC DEK expression is observed in all MECs (control and DEK-OE mice, Figure 1F), however, in the scRNAseq data DEK expression is confined to basal cells and a subset of stem/progenitor cells (Figure 5A). Furthermore, the hyperplasia in the DEK-OE model will bias bulk analysis (such as western blot and RNAseq) towards increased expression of MEC markers.
- A third major concern is the mechanistic link between DEK and H3K27me3. Most of the data are correlative and rely on bulk analysis or IHC. For example, in the DEK-OE organoid model, is there an increase in H3K27me3. Additionally, in the DEK-OE organoids, can loss of EZH2 block the increased cell proliferation?
Significance
Using genetic approaches in mice, this paper explores the role of the chromatin remodeler and oncogene DEK in the development of the normal mammary gland. This work will be of interest to researchers in the mammary development and breast cancer fields.
-
Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.
Learn more at Review Commons
Referee #1
Evidence, reproducibility and clarity
Johnstone & Leck et al. report their findings on the DEK chromatin remodeler and its newly discovered role in the development of the mammary gland through the use of a mammary epithelium-specific Dek overexpression model (Dek-OE). Using immunohistochemistry (IHC) and whole mounts of mammary glands, they show that the Dek-OE model is characterized by epithelial hyperplasia in multiparous, 15-month-old females. Through performing and analyzing bulk RNA sequencing of whole mammary tissue, they find that overexpression of Dek is correlated with cell cycle entry and progression, and the expression of luminal alveolar and mammary …
Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.
Learn more at Review Commons
Referee #1
Evidence, reproducibility and clarity
Johnstone & Leck et al. report their findings on the DEK chromatin remodeler and its newly discovered role in the development of the mammary gland through the use of a mammary epithelium-specific Dek overexpression model (Dek-OE). Using immunohistochemistry (IHC) and whole mounts of mammary glands, they show that the Dek-OE model is characterized by epithelial hyperplasia in multiparous, 15-month-old females. Through performing and analyzing bulk RNA sequencing of whole mammary tissue, they find that overexpression of Dek is correlated with cell cycle entry and progression, and the expression of luminal alveolar and mammary progenitor genes. The deregulation of cell cycle inhibitors was confirmed through IHC and western blot. To further support the connection between Dek and the cell cycle, it was also shown that palbociclib treatment of mammary epithelial organoids derived from Dek-OE mice was able to rescue the hyperplastic phenotype. To validate their transcriptomic findings of increased expression of luminal progenitor genes, IHC and western blots for alveolar markers and milk proteins were performed. By performing ovariectomy and looking at DEK expression throughout the development of the mammary gland, it was also found that Dek expression was promoted by ovarian hormones. Analysis of single-cell data from a previously published single cell gene atlas of the mammary gland, the authors found that Dek expression heavily overlapped with mammary stem cells and luminal progenitor populations, and was heavily correlated with expression of PRC2 components. Using western blots and a GFP-trap assay, it was found that Dek overexpression leads to increased H3K27me3, and PRC2 components directly interact with DEK. Using a conditional knockout of Dek, the authors found that Dek loss leads to decreased expression of PRC2 components in mammary epithelial cells by IHC and a failure for dams to lactate efficiently. While the authors findings are novel, there are major points that need to be strengthened and elaborated for clarity.
Major points:
- Several of the conclusions are made based on a limited number of replicates (often n=3) which is not a robust sample size to make a rigorous conclusion.
- The main text for Figure 1C mentions repression of luciferase expression by doxycycline chow, however the figure does not show any discernable repression in the Dek-OE conditions.
- To evaluate the impact of prolonged Dek overexpression on mammary epithelium in Figure 1G and 1H, the authors used multiparous females. One confounding factor with this experimental set up is the impact of previous pregnancies on the development of the mammary epithelium and in lowering tumorigenesis. Therefore, the impact of Dek on tumorigenesis cannot be determined in multiparous animals alone. To get a full picture, nulliparous animals should also be examined.
- To elucidate the molecular underpinning of Dek-OE phenotypes, the authors performed bulk RNA sequencing in Figure 2. Similar to point 2 however, only multiparous animals were used. As it has been previously shown that pregnancy significantly impacts the transcriptome of mammary glands, the effects of Dek overexpression can't be generalized to mammary glands as a whole. To make it generalizable, nulliparous Dek-OE animals must also be characterized.
- To validate findings from their transcriptomics work, the authors used IHC and western blots of candidate proteins that were found to be down regulated. In Figure 3A and 3C, the decrease in p21 protein levels through western blot seem much more modest than what the decrease seen in 3A would suggest.
- In Figure 3G-3I, the authors test the CDK4/6 inhibitor palbociclib to establish a direct link between the phenotypes seem in Dek-OE and cell cycle progression in organoid culture. Have the authors verified these findings with treatment of Dek-OE mice with palbociclib? In addition, have the authors checked to see if palbociclib corrected any of the transcriptional features associated with the Dek-OE model found in their transcriptomics data? In addition, the authors claim that the effect is specific to Dek-OE organoids as the effects of palbociclib on growth are not seen in control organoids. However, the data on unperturbed growth of control cells are not seen. To determine the specificity of the effects of palbociclib on Dek-OE derived organoids, the authors must show a time course tracking the growth of organoids with and without palbociclib. Rather than conclude the effects of palbociclib being specific to Dek-OE organoids, the authors most likely wanted to conclude that the increased growth of Dek-OE organoids compared to control organoids is dependent on the increase in cell cycle factors. (The validity of this is also weird though because even if division and growth were triggered through other transcriptional changes they found, like increased metabolism, growth in that scenario would be stopped by palbo as well)
- In the main text of Figure 4, the authors conclude that markers for luminal hormone sensing cells were unchanged in Dek-OE mammary glands, however the data to show this is not shown. This is problematic because the authors are directly drawing the conclusion that Dek-OE specifically upregulates luminal alveolar markers using this data.
- In figure 7, the authors look at a conditional knockout of Dek and conclude that pup death in the knockout was due to insufficient milk production by dams. While the authors establish that H3K27me3 and Ezh2 expression are abrogated, morphological analysis of the ducts is missing and would present convincing data. For instance, in the Dek conditional knockout, are luminal alveolar cells unable to differentiate fully, or are there far fewer? Decreased levels of histone modifications does not tell you much about whether repressive chromatin has changed its landscape in Dek KO mice, which is actually what influences transcription.
Minor points:
All figures need some sort of reformatting. Several of the conclusions are made based on a limited number of replicates (often n=3) which is not a robust sample size to make a rigorous conclusion. Many figures have text that is stretched. Histology and whole mount images are missing scale bar. IHC quantifications are obscure - what is an optical density? how many animals were analyzed and how many fields of vision were captured? Figure 2F is absolutely impossible to understand. Neither figures nor legends disclose the number of animals or samples analyzed. The statistical test utilized across all figures is not appropriated. Fig5B GSEA plots are missing statistical significance, and without this information one cannot properly access the relevance of the findings. Fig5C - how were co-expressed genes defined? is this just random genes that are expressed in cells that have higher levels of DEK? The term co-expressed suggests a specific type of analysis that would investigate linkage of expression between genes, which i dont think is the case here.
- Throughout, it would be better to indicate the genotype of the "Control" animals on each figure so as the rigor the experiment can be evaluated fully.
- Standard nomenclature for gene names and protein names should be corrected throughout the text.
- Similar to the point above, the use of Dek-OE to either refer to the mouse model or function as an acronym for "Dek overexpression" is inconsistent throughout the text.
- In the main text for Figure 4I-J, the authors state that DEK was previously published as an Erα target gene, however there is no citation to support this.
- It is unclear what the conclusion drawn from the experiments shown in Figure 4G-H and Figure 4I-J mean with respect to the goal of Figure 4, which was to show that Dek-OE mice have an expanded luminal alveolar compartment.
- Optical density was used to quantify IHC experiments, which was performed using color deconvolution in ImageJ. Something that is unclear is whether the authors are measuring density in the entire field of view, or if the authors are measuring optical density per cell. This has implications whether there are more cell expressing the protein of interest, or if the existing cells are expressing a higher level of the protein of interest.
- In the main text for Figure 6D, the system being used to overexpress DEK protein is not described. It is not the same genetic system as is used in the Dek-OE mice, as doxycycline is inducing Dek expression.
Significance
The role of Dek in tumorigenesis and in maintaining stem-like qualities in breast cancer cell lines have been previously reported. However, Dek has never been studied in the context of the normal mammary gland. The authors work revealing the role of Dek in normal development of the mammary gland is significant as understanding it has the potential of revealing additional roles Dek may have as an oncogene in breast cancers.
-