PREreview of Dominant negative effects on H3K27 methylation by Weaver syndrome-associated EZH2 variants

In this manuscript, Deevy and colleagues investigate the molecular impacts of EZH2 variants found in individuals with Weaver syndrome.

EZH2 is a catalytic subunit of the Polycomb Repressive Complex 2 (PRC2) responsible for depositing histone 3 lysine 27 (H3K27) mono(me1), di (me2) and trimethylation (me3). Heterozygous mutations in EZH2 are associated with Weaver syndrome and are inherited in an autosomal dominant manner. To model Weaver syndrome, the authors express human EZH2 with and without Weaver-associated variants in mouse embryonic stem cells heterozygous for Ezh2, where one allele is catalytically inactivated. Weaver syndrome extremely rare, and generating models from each patient is intractable and would introduce confounding genetic variability. Here, by creating a parental Ezh2 heterozygous line and overexpressing human EZH2 variants enables the team to model changes in H3K27 methylation and gene expression in a well-controlled system.

The authors employ a combination of Western Blotting, quantitative ChIP-seq, and STORM to demonstrate Weaver-associated EHZ2 variants generally decrease global levels of H3K27me2/3 as well as chromatin compaction. Further, RNA-seq analysis uncovered a shared set of 18 genes that are ectopically upregulated in 4/5 Weaver-syndrome variants tested, in association with a loss of H3K27me3, PRC2 and PRC1 at their promoter sequences. Importantly, the authors note that 6/18 genes are implicated in human growth phenotypes.

Another EZH2 variant, A733T, is associated with human growth restriction. Of note, most of the 18 upregulated genes in Weaver models are downregulated in cells reconstituted with A733T, coincident with a gain of H3K27me3, PRC2 and PRC1 levels at their promoters. Such reciprocal effects strongly suggests that modulating EZH2 activity has important consequences on human growth, perhaps through the regulation of this shared set of growth-related genes.

The authors also describe significant restriction of intergenic H3K27me2 levels in Weaver-associated mutants, and reciprocal H3K27me3 expansion in A733T and lymphoma-related Y641F mutants. However, the transcriptional consequence of changes in intergenic H3K27 methylation is less obvious and was not assessed.

The number and range of variants associated with Weaver syndrome assayed, as well as those associated with lymphoma and growth restriction, greatly helped contextualize the effect of all these variants.

Altogether, this is a compelling set of experiments that characterize reciprocal H3K27me2/3 and gene expression dynamics that likely contribute to human overgrowth and undergrowth phenotypes. The manuscript is well written, and the figures are clear and support their observations and conclusions.

Major comments

The authors mention Weaver syndrome-associated variants are classically considered hypomorphs because EZH2 has lower histone methylation activity in vitro, leading to the conclusion that haploinsufficiency underlies Weaver syndrome. Indeed, OMIM currently lists Weaver syndrome inheritance as autosomal dominant. However, the authors’ literature search did not uncover cases of “full” heterozygotes, where one allele is genetically ablated, suggesting Weaver syndrome variants are dominant negative. The molecular assays carried out lend support to this claim and is consistent with an autosomal dominant form of inheritance.

It remains unclear to me what impact the heterozygous deletion of the Ezh2 SET domain has on cells. Since there are no reported cases of “full” heterozygous Ezh2 Weaver syndrome (where one allele is fully inactivated, like in a SET domain mutant), it is odd to me that the parental Ezh2 het line shares many molecular signatures with the Weaver-associated EZH2 expressing variants:

·       WB quantification and spike-in normalized ChIP shows global H3K27me3 levels are lower in hets compared to WT, at a similar level (albeit higher) compared to the 10 Weaver syndrome variants (with the exception of A677T).

·       STORM imaging reveals hets have decompacted chromatin compared to WT, like in the Weaver variant cells.

·       At least 9 genes are significantly upregulated in Ezh2 het cells; are any shared with those upregulated in Weaver syndrome variants?

·       It would be interesting to include the quantitative H3K27me3 ChIP data generated from het cells in the meta-analysis over intergenic DNA (Fig. 4b)

·       Essentially, what I’m asking is: is what respect are Ezh2 het cells unlike Weaver syndrome variant expressing cells? Is it possible that hets impart a modest dominant negative effect, and expressing a Weaver variant in these cells pushes these (potentially sensitized) cells further toward PRC2 dysfunction in a dominant negative manner?

It is unclear whether the Weaver variants overlap conserved residues between mouse and human. An amino acid alignment of (at least) human and mouse EZH2, highlighting the position of Weaver variants, would be nice.

Can the authors expand on their speculation in the Discussion regarding why these Weaver mutants have a dominant negative effect? Specifically, why is Weaver-variant EZH2 able to methylate H3K27 (Extended data 1e), but when expressed in a het context reduce global H3K27me3 and restrict intergenic H3K27me3?

From the Introduction: “Furthermore, we previously proposed that the potential disruption of intergenic H3K27me2/3 could contribute to the aetiology of Weaver syndrome5, and we now speculate it may also be central to the aetiology of EZH2-associated growth restriction.”

Can the authors speculate on the relevance of disrupting intergenic H3K27me2/3 levels? Expression analysis focused solely on PcG target gene promoters. Is non-genic transcription disrupted in Weaver variant EZH2 expression cells?

From the Introduction: “This broad deposition profile of H3K27me2 prompted us to propose it could function as a repressive “blanket” to counter H3K27ac accumulation and/or to restrict the inappropriate activation of enhancers of alternative lineages.”

Does H3K27me2 restriction in Weaver variant EZH2 expressing mESCs result in H3K27ac gain at alternative lineage enhancers? Or is it believed to be a cell-type specific effect?

Minor comments

In the Discussion the authors state “Here, we characterize the Weaver syndrome- associated EZH2-A677T variant and find that it acts similarly to ‘gain-of-function’ EZH2 variants causative of B- cell lymphoma [ref 30]”. Repeating (as it was mentioned in the Introduction) the fact that the mutant in ref 30 (A677G) is at the same residue would stress the overlap between Weaver syndrome mutations and cancer-associated mutations at EZH2. 

Can the authors speculate why the Y641F mutant results in the increase of both global H3K27me3 and H3K27ac? Different scales?

WB loading controls used: ACTIN and GAPDH. In a perfect world pan-H3 would have been used.

Figure 1a is extremely helpful in understanding the variants presented. It would be nice to add all the mutants studied, including gain-of-function lymphoma-related ones like Y641F. Or do as in 3b and 5a and show the schema again. I believe only Y641F is missing.

Figure 3d: Seems unnecessary to present WT vs A677T data as overlapping tracks. It’s clear from the tracks above that A677T increases H3K27me3 at the region shown.

I don’t think I interpret Figure 2a correctly. Is it showing K27me3 at intergenic regions? I thought K27me2 should be enriched at these regions.

The introduction mentions p.A738T is a missense mutation associated with growth restriction. But the missense mutation studied was labeled A733T. Differences in human vs mouse amino acid position counting?

Sup 1 d is labelled c

Y677G typo – should be A677G

Competing interests

The author declares that they have no competing interests.