ZMYM2 controls transposable element transcription through distinct co-regulatory complexes

Reviewer #1 (Public Review):

Owen D et al. investigated the protein partners and molecular functions of ZMYM2, a transcriptional repressor with key roles in cell identity and mutated in several human diseases, in human U2OS cells using mass spectrometry, siRNA knockdown, ChIP-seq and RNA-seq. They tried to identify chromatin bound complexes containing ZMYM2 and identified known and novel protein partners, including ADNP and the newly described partner TRIM28. Focusing mainly on these two proteins, they show that ZMYM2 physically interacts with ADNP or TRIM28, and co-occupies an overlapping set of genomic regions with ADNP and TRIM28. By generating a large set of knockdown and RNA-seq experiments, they show that ZMYM2 co-regulates a large number of genes with ADNP and TRIM28 in U2OS cells. Interestingly, ZMYM2-TRIM28 do not appear to repress genes directly at promoters, but the authors find that ZMYM2/TRIM28 repress LTR elements and suggest that this leads to gene deregulation at distance by affecting the chromatin environment within TADs.

A strength of the study is that, compared to previous studies of ZMYM2 protein partners, it investigates binding partners of ZMYM2 using the RIME method on chromatin. The RIME method makes it possible to identify low-affinity protein-protein interactions and proteins interactions occurring at chromatin, therefore revealing partners most relevant for gene regulation at chromatin. This allowed the identification of novel ZMYM2 partners not identified before, such as TRIM28.

The authors present solid interaction data with appropriate controls and generated an impressive amount of datasets (ChIP-seq for TRIM28 and ADNP, RNA-seq in ZMYM2, ADNP and TRIM28 knockdown cells) that are important to understand the molecular functions of ZMYM2. These datasets were generated with replicates and will be very useful for the scientific community. This study provides important novel insights into the molecular roles of ZMYM2 in human U2OS cells.

The authors could have been more precise in the manuscript title and abstract to emphasize that these findings apply to human cells, as indeed there is no demonstration yet that the findings presented here can be transposed to mouse cells.

The manuscript's main conceptual advance is that the authors propose a novel model of gene regulation whereby transcriptional repressors of transposable elements could regulate genes at distance by modulating the local chromatin environment within TADs. Additional experiments would be needed to strengthen this model. For example the authors could have performed TRIM28 ChIP in ZMYM2-kd cells to test if ZMYM2 favors the recruitment of TRIM28 to its genomic targets, as well as ChIP-seq of repressive chromatin marks (such as H3K9me3) in ZMYM2-kd cells to investigate if the loss of ZMYM2 leads to reduced H3K9me3 in ERVs and over large regions surrounding the ERVs.

Reviewer #2 (Public Review):

In this study the authors investigate functional associations made by transcription factor ZMYM2 with chromatin regulators, and the impact of perturbing these complexes on the transcriptome of the U2OS cell line. They focus on validating two novel chromatin-templated interactions: with TRIM28/KAP1 and with ADNP, concluding that via these distinct chromatin regulators, ZMYM2 contributes to transcriptional control of LTR and SINE retrotransposons, respectively.

Strengths and weakness of the study:

- The co-localization of ZMYM2 with ADNP and TRIM28 is validated through RIME, ChIP-seq and co-IP. (Notably, since both RIME and ChIP-seq rely on crosslinking, and the co-IP with TRIM28 required crosslinking due to being SUMO-dependent, only the ZMYM2-ADNP co-IP experiment demonstrates an interaction in the absence of crosslinking).

- It is good that uniquely-mapped reads are used in the ChIP-seq analysis given the interest in repetitive elements. Likewise, though the RT-qPCR data in Fig5 should be complemented by analysis of the RNA-seq data that the authors already have, it seems that the primers are carefully designed such that a single retrotransposon copy is amplified.

- The top-scoring interactors are highly-abundant nuclear proteins: for example, data from the contaminant repository for affinity purification mass-spec data (https://reprint-apms.org/) show that TRIM28 is identified in 466 / 716 AP-MS experiments with a mean spectral count of 16. While this does not indicate that the ZMYM2-TRIM28 interaction is not 'true', it would have been helpful to further dissect the interaction to strengthen this conclusion. For example, it would be nice to see the co-IP (fig 3A) repeated from the cells expressing the ZMYM2 mutant that is no longer competent to bind SUMO (used in the ChIP-seq data of Fig 2). Alternatively - if the model is that ZMYM2 recruits SUMOylated TRIM28 - with well-characterized TRIM28 mutants that lack SUMOylation.

- The transcriptional response using bulk RNA-seq in ZMYM2-depleted cells is rather gene-centric despite the title of the paper being about TE transcription. In fact the only panels about TE transcription are the RT-qPCR data in Fig 5D,F. I may be missing something (and there aren't many details given about the RNA-seq experiments) but why not look at TE transcription in an unbiased way with the transcriptomic data at hand? I appreciate potential hazards of multi-mapping etc but it would be interesting to see at least some subfamily analysis (e.g. using the TEtranscripts tool). On a similar point, why not show some RNA-seq in the genome browser snapshots of the epigenomics - together with a RepeatMasker annotation track of TEs...

While the results broadly support the authors' conclusions, I have the overall impression that the central claim of TE transcriptional regulation by ZMYM2 could be strengthened a lot with some fairly straightforward additional experiments and analyses.

Reviewer #3 (Public Review):

ZMYM2 is a transcriptional repressor known to bind to the post-translational modification SUMO2/3. It has been implicated in the silencing of genes and transposons in a variety of contexts, but lacking sequence-specific DNA binding, little is known about how it is targeted to specific regions. At least two reports indicate association with TRIM28 targets (Tsusaka 2020 Epigenetics & Chromatin, Graham-Paquin 2022 bioRxiv) but no physical association with TRIM28 targets had been observed. Tsusaka 2020 theorizes an indirect, potentially SUMO-independent, interaction via ATF7IP and SETDB1.

Here, Owen and colleagues show that a subset of ZMYM2-binding sites in U2OS cells are clearly TRIM28 sites, and further find that hundreds of genes are silenced by both ZMYM2 and TRIM28. They next demonstrate that ZMYM2 homes to chromatin, and interacts with TRIM28, in a SUMOylation-dependent manner, suggesting that ZMYM2 is recognizing SUMOylation on TRIM28 itself. ZMYM2 separately homes to SINE elements bound by the ChAHP complex, in an apparently SUMOylation independent manner. Although this is not the first report to show physical interaction between ZMYM2 and ChAHP, it is the first to show that ZMYM2 homes to ChAHP-binding sites and functions as a corepressor at these sites.

The mode by which ZMYM2 and TRIM28 coregulate genic targets remains somewhat unclear. TRIM28/ZMYM2 bind to LTR elements, loss of these proteins results in upregulation of genes distal to (but in the same TAD as) these binding sites.

Overall, the manuscript is well-written, convincing, and fills a significant hole in our understanding of ZMYM2's mechanistic function.