Loss of MGA repression mediated by an atypical polycomb complex promotes tumor progression and invasiveness

  1. Haritha Mathsyaraja
  2. Jonathen Catchpole
  3. Brian Freie
  4. Emily Eastwood
  5. Ekaterina Babaeva
  6. Michael Geuenich
  7. Pei Feng Cheng
  8. Jessica Ayers
  9. Ming Yu
  10. Nan Wu
  11. Sitapriya Moorthi
  12. Kumud R Poudel
  13. Amanda Koehne
  14. William Grady
  15. A McGarry Houghton
  16. Alice H Berger
  17. Yuzuru Shiio
  18. David MacPherson
  19. Robert N Eisenman

  • Curated by eLife

    eLife logo

    Summary:

    The reviewers agreed that the paper provides strong in vivo data for a tumor-suppressive role for Mga in lung carcinogenesis. The authors convincingly show that MGA is important in oncogenesis. We note here that MGA is highly understudied (~200 publications) in and of itself despite its involvement with the MYC network for oncogenesis (~41,000 publications at the current time). Given a protein of 3000 amino acids, the number of potential protein partners and PTMs that might modify its tumor suppressor functions are staggering. However, the reviewers also noted that a previous paper has addressed the same topic and the novelty of the data presented here needs to be better explained and additional experiments are needed to strengthen and expand the new aspects.

    Reviewer #1 opted to reveal their name to the authors in the decision letter after review.

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

MGA, a transcription factor and member of the MYC network, is mutated or deleted in a broad spectrum of malignancies. As a critical test of a tumor suppressive role, we inactivated Mga in two mouse models of non-small cell lung cancer using a CRISPR-based approach. MGA loss significantly accelerated tumor growth in both models and led to de-repression of non-canonical Polycomb ncPRC1.6 targets, including genes involved in metastasis and meiosis. Moreover, MGA deletion in human lung adenocarcinoma lines augmented invasive capabilities. We further show that MGA-MAX, E2F6, and L3MBTL2 co-occupy thousands of promoters and that MGA stabilizes these ncPRC1.6 subunits. Lastly, we report that MGA loss also induces a pro-growth effect in human colon organoids. Our studies establish MGA as a bona fide tumor suppressor in vivo and suggest a tumor suppressive mechanism in adenocarcinomas resulting from widespread transcriptional attenuation of MYC and E2F target genes mediated by MGA-MAX associated with a non-canonical Polycomb complex.

Article activity feed

  1. Version published to 10.7554/elife.64212 on eLife
  2. eLife

    Author Response:

    Reviewer #3:

    However, a lot of the data presented in the manuscript are not novel and were previously published. A recent Molecular Cancer Research paper by Llabata and collaborators published in April 2020 (referred to in the text) has already identified the same MGA interactors by Mass Spectrometry and the same binding sites by ChIP-Seq using human lung adenocarcinoma cell lines. Llabata et al. found that MGA interacts with the non-canonical PCGF6-PRC1 complex (named PRC1.6) that includes L3MBTL2 and that the complex also contains MAX and E2F6 but not MYC. They clearly show that MAG binds to and represses genes that are bound and activated by MYC convincingly showing that MYC and MGA have opposite functions. This unfortunately tempers the enthusiasm of the reviewer.

    This reviewer states that "... a lot of the data presented in the manuscript are not novel and were previously published". The reviewer goes on to write that the Llabata et al. 2020 paper (referring to doi: 10.1158/1541-7786.MCR-19-0657 [https://mcr.aacrjournals.org/content/18/4/574]) "has already identified the same MGA interactors by Mass Spectrometry and the same binding sites by ChIP-Seq using human lung adenocarcinoma cell lines. Llabata et al. found that MGA interacts with the non-canonical PCGF6-PRC1 complex (named PRC1.6)..." ​

    We strongly disagree with the reviewer's statements.

    1. A major focus of our paper is that it provides and validates a mouse model in which we delete MGA and demonstrate its tumor suppressive activity. The experiments in Llabata et al., including the biological assays and the ChIP_Seq, were done by overexpressing MGA in cells which already express endogenous MGA. Therefore, all their data monitor the consequences of overexpression of MGA, a situation without clear biological relevance. In the experiments reported in our paper, we delete MGA. Therefore our molecular data refer to a comparison between MGA null and the same cells expressing endogenous MGA. This is important since MGA is a tumor suppressor and its loss of function is what is crucial biologically, as we show here or the first time in our lung adenocarcinoma model. Furthermore, by deleting MGA we were able to show that its loss corresponds to an increase in a core set of target genes previously associated with PRC1.6. Furthermore, we show that members of this core group are relevant to the proliferation of tumors that lack MGA.

    2. The PRC1.6 complex has been known to be associated with MGA since at least 2012 as indicated in our references cited. Llabata et al confirmed that result. Our paper reports PRC1.6 subunits are associated with MGA through the DUF4801 domain of MGA. This is the first identification of the interface between PRC1.6 and MGA. It is important and relevant because multiple frame shift mutants in MGA have the consequence of deleting this region in a wide range of tumor types.

  3. eLife

    Reviewer #3:

    Mathsyaraja and collaborators analyzed the role of the MAX-Gene associated protein, referred to as MAG, in mouse models and human cell lines and organoids of Non-Small Cell Lung Cancer. MAG is a repressor, a MYC antagonist that opposes its transcriptional activity. It has TBX and bHLH domains. They found that MGA loss by shRNA or CRIPSR accelerated tumor development in vivo in the KP mouse models. Using RNA-Seq, the authors showed that MGA loss leads to the de-repression of the atypical/non-canonical PRC1.6 polycomb complex, E2F and MYC targets as well as increased invasion. ChiP-Seq/cut and run as well as proteomics, revealed that MGA, E2F6 and L3MBTL2 co-occupy thousands of promoters and that MGA interacts with E2F6, and many core members of PRC1.6. Finally, they mapped the DUF domain as required to bind the PRC1.6 complex and bring it to promoters.

    Overall, the experiments are well executed, the paper clearly written and the conclusions justified by the data.

    The new data in the present report are the in vivo data in the mouse models, the role of MGA in repressing invasion, in increasing IFN signaling and the anti-tumor response, and the identification of the DUF domain required for binding to the PRC1.6 complex.

    However, a lot of the data presented in the manuscript are not novel and were previously published. A recent Molecular Cancer Research paper by Llabata and collaborators published in April 2020 (referred to in the text) has already identified the same MGA interactors by Mass Spectrometry and the same binding sites by ChIP-Seq using human lung adenocarcinoma cell lines. Llabata et al. found that MGA interacts with the non-canonical PCGF6-PRC1 complex (named PRC1.6) that includes L3MBTL2 and that the complex also contains MAX and E2F6 but not MYC. They clearly show that MAG binds to and represses genes that are bound and activated by MYC convincingly showing that MYC and MGA have opposite functions. This unfortunately tempers the enthusiasm of the reviewer.

  4. eLife

    Reviewer #2:

    This manuscript by Mathsyaraja et al. studies the oncogenic loss of the Max-gene-associated (MGA) protein due to deletion or mutation in cell-lines, in mice and in human cancers (cell-lines and tumors). The authors knocked out MGA by aerosol-delivered, CRISPR-CAS expressing lentiviruses that simultaneously Cre-activated a Lox-stop Kras oncogene. The loss of MGA accelerated proliferation and oncogenesis, and shortened survival. Oncogenesis was further enhanced by enforced TP53 deletion in these lung tumors. RNA-seq and ChIP-seq of MGA+ or - cell-lines demonstrated the up and downregulation of various gene classes (thousands of genes) according to function and regulation including of PRC1.6 targets, meiosis regulators, TGF-beta signaling pathway components, EMT regulators, anti-tumor immunity, as well as of MYC, E2F, etc. Different cell lines exhibited both overlapping and distinct target sets. MGA knockout cells were more migratory and invasive and displayed actin-protrusions in accord with this behavior. They show that a Domain of Unknown Function in the mid-region of MGA engages PRC1.6 and is required to depress proliferation. The DUF is also required to limit actin-protrusions. Human colon organoids were studied since MGA mutations and deletions are also apparent in colon cancer. Again, shared and distinct targets of MGA action were inferred.

    The authors make a strong case that MGA is an important tumor suppressor that operates through PRC1.6 for some of its actions.

  5. eLife

    Reviewer #1:

    The authors report the analysis of a Mga deletion and provide convincing evidence that Mga functions as a tumor suppressor during lung carcinogenesis. The data shown are clear, the message is important and the discussion is very careful. There is a certain overlap with a recent study by Llabata et al., but there is sufficient novelty in the current study.

    Comments:

    It seems that the investigation of publicly available datasets is essentially identical to the Schaub et al . analysis and not new data. If the authors want to maintain this, they would need to better explain what is new. One important piece of information that seems to be missing is whether the mutations are homozygous or heterozygous. So data on MGA and MYC protein expression in human tumors would greatly strengthen this part.

    Conceptually, one would to know whether tumor development in an MGA-delete situation depends on MYC. One would also like to know whether the polycomb complex that is assembled by MGA is tumor-suppressive. Therefore,the authors should perform a similar analysis as they did for MGA (introduce sgRNAs into the lung models) and score the phenotypes they get. Both experiments could be done in cell lines established from this model and either in vitro (that would allow a mechanistic analysis, e.g. RNA seq) or upon re-transplantation. This would also prevent simply reporting negative results.

    The interpretation of the VENN diagram and the heatmaps in Figure 5A,B is somewhat uncertain. If one plots these for MYC, occupancy often simply parallels occupancy by RNAPII, so essentially being bound by MYC simply says the promoter is open/active. Is this the case for MGA and its complex partners? Or is there a specificity in binding? The authors should do RNAPII ChipSeqs in these cells, preferentially +/- MGA, and then show these alongside (and plot a correlation between MYC, RNAPII and MGA occupancy).

    Along these lines, it is hard to understand how one obtains the extreme p-values shown in figure 5E and 5H, I would challenge this. If the authors want to maintain this, they should not use ENCODe data, but simply determine what genes are active in the cells (e.g. what promoters are bound by RNAPII) and then use those as background list and calculate P-values for overlap between MYC, MAX and E2F6.

    Based on the description, the ChIPSeq analyses are not spike-normalized and I could not find information about the number of repeats. If it is n=1, the authors need to find a way to exclude that the differences are due to experimental variation.

    I think the Llabata reference is missing in the list.

  6. eLife

    Summary:

    The reviewers agreed that the paper provides strong in vivo data for a tumor-suppressive role for Mga in lung carcinogenesis. The authors convincingly show that MGA is important in oncogenesis. We note here that MGA is highly understudied (~200 publications) in and of itself despite its involvement with the MYC network for oncogenesis (~41,000 publications at the current time). Given a protein of 3000 amino acids, the number of potential protein partners and PTMs that might modify its tumor suppressor functions are staggering. However, the reviewers also noted that a previous paper has addressed the same topic and the novelty of the data presented here needs to be better explained and additional experiments are needed to strengthen and expand the new aspects.

    Reviewer #1 opted to reveal their name to the authors in the decision letter after review.

  7. Version published to 10.1101/2020.10.16.334714v1 on bioRxiv