PREreview of Tunable DNMT1 degradation reveals cooperation of DNMT1 and DNMT3B in regulating DNA methylation dynamics and genome organization

In this study the authors analyse the impact of removal of the maintenance DNA methyltransferase DNMT1 in human cells using a degron system. This follows on from previous work on the role of DNMT1 in maintaining mammalian DNA methylation patterns and the effect that its removal has on cellular phenotypes. While degron tagged DNMT1 has been reported in the literature, this study did not examine phenotypes in detail (Kikuchi et al 2022, https://doi.org/10.1038/s41467-022-34779-4). The study focuses on characterisation of the system, before examining how degradation of DNMT1 affects DNA methylation patterns, cellular phenotypes and genome organisation.

Based on their study, the authors conclude that:

- Degradation of tagged DNMT1 results in demethylation of the genome to a greater extent than 5-aza-2’-deoxycytidine

- Removal of DNMT3B results in greater hypomethylation upon DNMT1 degradation

- DNMT1 degradation has minor effects on cellular fitness compared to previous reports using alternative approaches

- Changes in heterochromatin organisation are observed upon DNMT1 degradation

Overall, this is a very interesting study making the first use of a degron approach to understand DNMT1 function. However, the study suffers from a lack of discussion about how the main cell line model used might be responsible for differences in cellular phenotypes compared to previous work. In addition, some of the analyses could be conducted in a more statistically robust fashion and to greater depth.

Specific comments:

- The authors engineer the degron tag into both DLD1 and RPE1 cells. They highlight that they use two cell lines in the abstract but the vast majority of studies are undertaken on DLD1 cells with very little work done in RPE1 cells. The study would benefit from greater replication and comparison between the two cell lines.

- The blot showing increased DNMT3B upon degradation of DNMT1 in the DLD1 cells do not appear to be quantified or replicated (Fig 1D).

- There are no replicates for the Infinium array experiment. It is also unclear how generalisable these results are to cell lines/types.

- The losses of DNA methylation observed upon depletion of DNMT1 are reported to be biased towards certain chromatin contexts. Given that DNMT1 is thought to act on the whole genome, we wondered why losses of methylation might show this bias?

- DNMT3B loss is reported to be associated with decreased methylation at heterochromatin. This is puzzling given that DNMT3B is thought to be predominantly recruited to H3K36me3 in gene bodies. Can the authors discuss why this might be the case?

- No quantification or statistics are presented when describing the results shown in heatmaps shown in figure 1F-I. Note that the beta value threshold applied to DMPs displayed in the heatmaps differs from that in other analyses due to the high numbers of DMPs. These results might be better displayed in an alternative fashion to show all DMPs.

- The chromatin states used to characterise the differentially methylated probes come from HEPG2 liver carcinoma cells. It is unclear these data were used as ChIP-seq data are available for colorectal cancer cell lines in ENCODE. It would have also been nice to see verification of major conclusions using ChIP-seq data from DLD1 cells.

- The authors observe some regions that gain methylation in the DNMT3B KO cells. They suggest that this could result from uncontrolled DNMT1 activity but do not provide evidence that this is the case and their appear to be other potential reasons. For example, the DNMT3B KO cells have been through multiple rounds of clonal selection. Could selection for clones that survive due to a higher level of methylation at some sites explain the gains of methylation observed?

- One of the major claims of the manuscript is that DNMT1 depletion by degron is less toxic to cells than deletion of the gene or pharmacological inhibition. However, the authors state that the DLD-1 cell line used is p53 deficient. Previous studies have shown that DNMT1 KO causes p53-dependent apoptosis (Jackson-Grusby et al 2001 https://doi.org/10.1038/83730). In addition, upregulation of p53 was noted upon inducible deletion of DNMT1 in HCT116 cells in the Chen et al (https://doi.org/10.1038/ng1982) that the authors directly compare to (‘In sharp contrast to previous reports using a inducible DNMT1 KO in HCT116 colorectal cancer cells…’). The study would be improved by work to disentangle whether the degron approach p53 deficiency underpin the differences compared to previous work (for example through greater exploration of the differences observed in RPE1 cells).

- The authors include a comparison with GSK3685032 but do not show how the dose used affects DNA methylation levels. A recent study also suggests that a related compound, GSK3484862, results in DNMT1 degradation (Chen et al 2023 https://doi.org/10.1093/narcan/zcad022). It would be interesting to know whether the doses of GSK3685032 used by the authors also cause DNMT1 degradation and whether this is to a level similar that observed with the degron.

- The authors suggest that high-numbers of mitotic errors are unlikely to explain the reduced proliferation potential observed in their experiments because they observed low levels of aneuploidy using single cell sequencing. However, this is an indirect assay and the study would be strengthened by a direct examination of mitosis. In addition, we were unclear how the level of aneuploidy observed in single cell sequencing relates to the rate of mitotic error and note that the percentage of aneuploidy observed increased from 12.2 to 20% in the treated cells. This result does not appear to be analysed by statistical testing.

- The authors observe the inactive genomic compartment associated with H3K9me3 occupies a very small proportion of the genome. How does this compare to the staining in Figure 4A? Also was the loss of H3K9me3 reported upon DNMT1 degradation/ DNMT3B deletion seen in this staining. Can robust conclusions be drawn from the ChIP-seq as it does not appear to have been conducted using a spike-in control?

- How do the changes in genome organisation observed by HiC correspond to changes in DNA methylation. Could there be off target effects of the degron that might affect other proteins required for nuclear organisation?

Minor comments:

- Figure 2D. The distribution of Beta values derived from Infinium arrays is generally bimodal. As such representing the data as a boxplot in Fig 2D obscures the true distribution and an alternative type of plot would be preferable for this figure.

- It was unclear whether hypomethylated probes and hypermethylated probes in the Infinium array analysis were considered together when analysing enrichments in genomic annotations and chromatin annotations (Fig 2E and SH to K).

- On Page 9 the authors suggest that in comparison to DAC ‘The total demethylation achieved using this system is approximately 4 times higher…’. This statement appears to be based on the number of significant probes. However, given that Infinium array data only measure a small percentage of CpG sites in the genome making it difficult to conclude anything about the total level of demethylation. The previous analyses in Figure S2A and D also contradict this statement. This sentence would be clearer if it was rephrased.

- Figure 2E it is unclear if only selected significances are shown or if all were tested and others were not significant. Several of the bars look very similar to each other but only some are starred indicating they are significant.

- Figure S2J,K. We were unsure how this analysis was conducted. The graph appears to imply that 100% of probes on the EPIC array are located within repeat elements.

- The authors cite Lehnertz et al 2003 in support of the statement ‘it has been proposed that DNAme can influence H3K9me3 deposition’. However, that study states that deletion of DNMTs did not alter H3K9me3: ‘In contrast, H3-K9 trimethylation at pericentric heterochromatin is not impaired in Dnmt1 single- or Dnmt3a/Dnmt3b double-deficient ES cells.’.

- It is unclear what the dots represent in figure 4B and C? Are these means from replicates? It would be useful to see the overall distribution of the data from these analyses.

Written by Duncan Sproul on behalf of the members of the Sproul group following a journal club discussion during labmeeting.

Competing interests

The author declares that they have no competing interests.