Arabidopsis SDG proteins mediate Polycomb removal and transcription-coupled H3K36 methylation for gene activation

Reviewer #1 (Public Review):

Summary:

In this manuscript, the authors reveal a new role for SDG7 in the regulation of H3K36me2 and me3. SDG7 appear to be functionally redundant to SDG8 as the double mutant presents lower levels of H3K36me and stronger phenotypes than either single mutant, however, their mechanisms of action might differ as the proteins displayed different localization on their target genes, with SDG7 localizing preferentially to TSS and TES while SDG8 covers the gene body. SDG7 binds preferentially to PREs, which recruit PRC2 for H3K27me3 deposition. The authors therefore present an interesting model where SDG7 evicts PRC2 from silenced genes, leading to a loss of H3K27me3. This would allow the transcriptional activation of the genes and the deposition of H3K36me3.

Strengths:

Overall, the manuscript is well-written and organized, although some paragraphs need clarifications. The figures are clear and well designed and the proposed model is compelling. While the manuscript is already interesting as it is, I think addressing the following questions would elevate it even more and refine the proposed model:

Weaknesses/potential aspects to address:

(1) It is still unclear whether SDG7 directly catalyzes H3K36me or if it promotes its deposition simply by eviction of PRC2. The AlphaFold and structure analyses show a significant similarity between the catalytic domains which would support the first possibility, but some more experiments would be required to prove this more definitively.

(2) Does SDG7 directly recognize the PRE (as suggested by the model in Figure 5F) or is it recruited by some transcription factors? Is SDG7 known to interact with any of the PRC2 recruiters?

(3) Line154/Figure 2A: The metagene plot for H3K36me3 shows a lower level on the gene body but a higher peak in sdg7sdg8 double mutants compared to the Wild-type, which is a bit surprising, especially considering that the immunostaining in reference 19 showed a near complete loss of H3K36me3 signal in the same double mutant. Can this higher peak be an artifact from the normalization strategy, or due to the existence of different subpopulations of genes?

Indeed, on the genome tracks presented by the authors, the hypomethylated genes show a loss of signal on the entire gene body, and not a higher peak near the TSS. It might be interesting to generate metagene plots for H3K36me3 hypo and hyper-methylated genes, to see if the higher peak at the TSS is solely due to the hyper-methylated genes.

(4) Figure 2C: More than 40% of differentially methylated genes are actually hypermethylated, but the authors do not discuss this at all. What are those genes, are they targeted by SDG7 or 8? Could they be responsible for the higher peak at the TSS observed in the double mutant? (see previous comment).

(5) Figure 2C and D: The method section states that the ChIP-seq was performed on 5-day-old seedlings, while the legend of this figure mentions root and shoot samples but this does not appear in the figure itself. There is also mention of shoot and root samples in Supplementary Tables 1 and 2. The authors should clarify which tissue was used for the data presented in Figure 2 and correct the legends or the methods accordingly.

(6) Line 270/Fig 4K and L: The text mentions looking at the 838 genes "downregulated in clf sdg7 sdg8 relative to sdg7 sdg8" and in the overlap, the authors identified FLC. However, in Figure 5D, FLC is upregulated in clf sdg7-sdg8 compared to sdg7-sdg8, not downregulated as mentioned in line 270. The Venn diagram in Figure 4L mentions "sdg vs clf sdg up", which would fit the pattern seen in Figure 5D, but the number of genes (838) matches the number of downregulated genes in the sdg7sdg8 vs clf s dg7sdg8 volcano plot.

I would actually expect the phenotype rescue to be caused by genes that are up in Wt vs clf, down in Wt vs sdg7-sdg8, and back up in sdg7-sdg8 vs clf-sdg7-sdg8, not "up/down/down" as mentioned in the text: genes would be downregulated in sdg7-sdg8 because of a loss of H3K36me and therefore hypermethylation of H3K27, but in the absence of CLF, this hypermethylation is reversed and the genes are upregulated in the triple mutant compared to the sdg7-sdg8 mutant. This is also what the authors see and describe in their cluster analysis in Figure 4M and line 280, mentioning an upregulation in clf-sdg7-sdg8 vs sdg7-sdg8. Could the authors please clarify these discrepancies between the different subplots and within the text itself? Was there maybe some error plotting the volcano plot and/or Venn diagram?

In general, as this part is quite complicated, maybe it would benefit from a clearer explanation from the authors as to why they look at those particular overlaps, so that the reader can more easily follow their train of thought.

(7) Figure 4N/Line 286: How were these 828 genes identified? Is it stemming from a clf-sdg7-sdg8 vs sdg7-sdg8 comparison? The legend says "genes shown by white color in Fig. 4M", do the authors mean the two clusters previously described?

(8) Line 300: "suggesting that SDG8 primarily mediates target gene expression in conjunction with PAF1C". This statement is based on overlapping genes that are downregulated in sdg7-sdg8 double mutant and paf1c mutants but concludes only on the role of SDG8. I feel that to state that SDG8 regulates expression in conjunction with PAF1C, the authors should rather examine the genes downregulated in the sdg8 mutant, especially considering the reduced overlap between genes downregulated in sdg8 and sdg7-sdg8 (according to Figure 2C, only 30% of the genes downregulated in sdg8 are also downregulated in the double mutant), or this statement should be corrected to also include SDG7.

Maybe it would be easier to read the figure if the authors created a master list of genes downregulated in at least one of the paf1c mutants they examined (as they anyway do not examine in detail the contribution of each individual paf1c mutant), and overlap it with the genes downregulated in sdg7, sdg8 or sdg7-sdg8.

(9) Line 326: "We also discovered that SDG7 and SDG8 overcome PRC2-mediated silencing, leading to a switch from H3K27 methylation to H3K36 methylation during growth and development." While part of this statement is supported by the ChIP data presented in Figure 4E, I think a ChIP for H3K36me2 and/or me3 is necessary to prove the existence of a K27me to K36me switch.

(10) Line 347: The authors state that SDG8 is located at the TSS and 3' end of genes, but on line 187 they state that it occupies the gene body (which is supported by the plot in Figure 3A).

(11) Line 351: The authors suggest a role of RNApolII in the deposition of K36me, but their data are not sufficient to support this hypothesis. The transcriptome data show that both SDGs and PAF1C regulate a similar set of genes, but they do not show data demonstrating that RNApolII is necessary for the deposition of K36me. It might be interesting to examine H3K36me levels in a paf1c mutant to further consolidate their hypothesis.

Reviewer #2 (Public Review):

Summary:

In this manuscript, the authors combined imaging approaches with molecular and genetic experiments to

(i) for the first time establish a chromatin regulatory role for SDG7;
(ii) examine its role in H3K36 methylation, along with its homolog SDG8;
(iii) examine its potential role in mediating antagonism to PRC2, thereby mediating transcriptional activation.

Strengths:

The manuscript explores interesting and relevant mechanistic hypotheses about chromatin-mediated gene regulation by combining a range of experimental tools and a genome-wide perspective. The writing is very clear. The study makes good connections to existing data and generates datasets that complement existing datasets, providing a valuable resource to the community.

Weaknesses:

Some of the claims appear to need further supporting evidence to establish their robustness.

Review of the main conclusions and supporting evidence:

(1) SDG7 contributes to H3K36 methylation at several loci along with SDG8:
This conclusion is supported by the genome-wide differences (measured by ChIP-seq) in H3K36me3 and me2 levels observed between WT, the sdg7 and sdg8 single mutants, and the double mutant. The reduction in H3K36 methylation levels observed at a large number of genes in the sdg7 mutant, and the further reduction in H3K36 methylation levels in the double mutant compared to sdg8 (observed at multiple genes) indicates that SDG7 promotes this transcription-associated modification. The significantly larger number of H3K36 hypomethylated genes in the double mutant (3380) compared to either of the single mutants (523 and 605) indicates lower overall H3K36me3 levels in the double mutant.
While direct evidence of SDG7 methyltransferase activity on H3K36 is lacking, the study reports structural comparisons using AlphaFold predictions that suggest the possibility of such a role.

(2) SDG7 influences gene expression together with SDG8:
This conclusion is supported by a range of phenotypic differences observed between sdg8 and the double mutant sdg7 sdg8. The reported genome-wide differential gene expression between the WT and the double mutant as well as expression differences in specific genes detected by imaging are also consistent with SDG7 and SDG8 together influencing gene expression. However, a role specifically for SDG7 could be further supported by a direct differential expression analysis between the single mutant sdg8 and the double mutant sdg7 sdg8. A role for SDG7 in gene expression is also consistent with its observed effect on H3K36 methylation, which is generally associated with productive transcription.

(3) SDG8 is exclusively nuclear, but SDG7 localises to the cytosol and the nucleus in meristematic cells:
This conclusion is supported by imaging of fluorescent-tagged SDG8 and SDG7 in the root tip, which shows nuclear localisation of SDG8-GFP, consistent with previous reports, and a combination of nuclear and cytosolic localisation for SDG7-VENUS. However, the study presents only one replicate - more replicates are needed to establish the robustness of this conclusion.

(4) Distinct binding patterns of SDG7 and SDG8 on chromatin:
This conclusion is supported by the analysis of genome-wide binding patterns of these proteins measured by ChIP-seq (using fluorescent tagged versions). This data indicates that while SDG7 tends to localise to TSS and TES regions of genes, SDG8 tends to be more uniformly spread across gene loci. These patterns suggest that SDG7 and SDG8 may influence H3K36 methylation through distinct mechanisms, but do not indicate what these mechanisms may be.

(5) SDG7/SDG8 antagonism with PRC2:
a. SDG7 overlaps with PRC2 and its recruiters on chromatin and can bind to PREs
This conclusion is supported by statistically significant overlaps genome-wide between cis-elements at SDG7 binding peaks and those previously reported to be Polycomb associated. This is further supported by the observation of SDG7 binding peaks close to PRC2 subunit binding peaks at known PRC2 targets.
b. SDG7/SDG8 antagonism with PRC2
This conclusion is supported by the observation that the sdg7 sdg8 double mutant phenotypes are partially rescued by disrupting CLF, one of the PRC2 methyltransferases. However, this does not necessarily suggest a direct antagonism with PRC2, but could be part of a more general antagonism between productive transcription (in which H3K36 methylation plays an active role), and PRC2-mediated silencing.
c. SDG7 can evict PRC2 from PREs to overcome H3K27me3-mediated silencing
This conclusion - that SDG7 directly antagonises PRC2 interaction with cis-elements that mediate its targeting - is partly supported by the observation that inducing overexpression of SDG7 (through dexamethasone induction) can cause changes in CLF occupancy and H3K27me3 levels at certain designated PREs. To establish the robustness of these conclusions, it will be necessary to examine designated regions to be used as a negative control and include a non-transgenic control for the SDG-7-HA ChIP. A suitable negative control may be a non-PRE region known to have a CLF peak and high H3K27me3 levels, where the levels would not be expected to change in this experiment.
It remains to be established whether this apparent antagonism of PRC2 by SDG7 is only part of a more general antagonism of PRC2 by productive transcriptional activity, or whether SDG7 can evict PRC2 from PREs independent of transcripts and therefore is a precursor to switching to an active transcriptional state.
Another aspect that would be interesting to examine in the light of recent findings is the single-cell-level behaviour at these genes targeted by SDG7-to examine whether SDG7 induction is causing individual copies of these genes to stochastically switch to an active transcriptional state so that the observed changes in H3K27me3 result from a fraction of copies completely losing silencing rather than all copies losing some H3K27me3.

(6) H3K36me3 levels are co-regulated by SDG8 and Paf1c, SDG8 associates with Pol II to deliver transcription-coupled H3K36 methylation
This conclusion is supported by analysis of a subset of loci previously reported to be regulated by Paf1c, which also exhibit changes in the sdg7 sdg8 double mutant as well as a clf mutant. This analysis is based on a qualitative examination of ChIP-seq signal at a small number of loci, and therefore provides indirect support for this conclusion. It is, therefore, difficult to draw mechanistic insights from this analysis that go beyond correlation.