PREreview of Dynamic DNA methylation turnover in gene bodies is associated with enhanced gene expression plasticity in plants

In this study the authors analyse DNA methylation patterns in plants, suggesting that a subset of genes display non-canonical intermediate levels of methylation in their gene bodies.

This follows on from the author’s previous report of Arabidopsis somatic mutants for all 4 DNA demethylases (drdd mutants) (https://doi.org/10.1093/plcell/koab319) where they characterised these plants and their overall phenotype. It also showed that a subset of genes gain methylation in their gene bodies in these mutants. This new study follows on from this by investigating these genes in greater detail.

The authors define dynamic gene body methylation genes (dynamic GbM) as those that gain methylation in the drdd mutants. They are approximately 10-times less frequent than the canonical genes (termed stable genes, stable GbM) which are often used as a control group in the study.

The study suggests that the dynamic gene group:

- Show intermediate DNA methylation levels across many plant tissues and developmental stages.

- Are not targeted by the canonical de novo methylation pathways in Arabidopsis but are maintained by the maintenance methyltransfersase met1.

- Have intermediate DNA methylation levels in orthologues in other plants.

- Are associated with different chromatin marks to the stable gene group.

- Have more variable expression across different tissues and conditions.

Overall, the study is interesting and suggests that there are different types of gene body methylation in plants. However, given the authors suggestion that their observations are consistent with dynamic variation in DNA methylation and expression at these loci, we felt the manuscript would benefit from some single molecule and single cell analyses. We also felt that it would benefit from further detail in the methods as it was not always clear how the measurements plotted on many of the figure panels were derived.

Specific comments:

- In figure 1 the authors show that dynamic GbM genes have CpGs which are methylated in a proportion of cells. This observation is consistent with either two populations of cells/alleles in the sample analysed or all cells/alleles having heterogenous methylation levels at the single-molecule level. Single read analysis could help clarify which of these is the case. Similar analyses could be applied to many of the other sections of the manuscript.

- We were unclear what was measured on the x-axes in Figure 1D, 2B and C, 3B and C. ‘CG methylation heterogeneity’ and could not find it defined in the manuscript.

- The authors show that methylation is lost at dynamic genes in heterozygous met1 mutants and low-level recovery is observed at some genes in homozygous wildtype progeny. They argue that, combined with the lack of difference in mutants for de novo pathways, that met1 could be the de novo methyltransferase acting at dynamic genes. However, here no comparison is made to stable genes to show if dynamic genes behave differently in these experiments. As the authors note, these data are not conclusive of a role for met1 in de novo methylation. In our opinion the observations made in this section are interesting, but the manuscript would benefit from further experiments to clarify the role of met1 or discussion of alternative possibilities.

- We were unclear how S2 quantifies de novo methylation events? We could not find a detailed definition of the quantity plotted in the manuscript.

- We were unclear why for the analysis of orthologues the dynamic GbM genes were compared to random genes and not stable GbM genes.

- For the histone marks, it would be interesting to know if these loci show a heterogeneous chromatin structure in terms of their marks as well as DNA methylation.

- For the analysis of expression variance the authors select the coefficient of variation. In RNA-seq the variance shows a relationship with the mean. This may confound the analysis and conclusions that can be drawn from it. However, many simpler methods of controlling for this such as the coefficient of variation fail to do so adequately. Most analyses prefer more sophisticated methods such as the variance stabilisation normalisation employed in DE-Seq2.

- It would have been interesting to see a single cell analysis of expression (either genome wide or for selected loci) to understand further the nature of the postulated variance in expression in greater detail.

Minor comments:

- The major difference between dynamic and stable gene body methylation here is in levels of methylation. However, it is worth noting that high levels of DNA methylation can also be associated with high levels of methylation turnover.

- The authors usually use green for stable genes and blue for dynamic genes. However, they occasionally reverse this colour scheme (eg Figure 1).

- Figures would benefit from an indication of the number of observations included (eg how many CpGs/genes in each analysis).

- P-values are indicated with asterisks rather than values.

This pre-review was written by Duncan Sproul on behalf of the members of the Sproul group following a journal club discussion during labmeeting. Note that as a mammalian epigenetics group we are not experts in the plant epigenetics literature and may have missed relevant previous studies in plants.

Competing interests

The author declares that they have no competing interests.