TrxG and PcG Proteins but Not Methylated Histones Remain Associated with DNA through Replication, Cell (2012),
- Result 1 (Fig 1): ‘K27’ ‘K4’ don’t associate with PCNA labeled nuclei. comment: At gastrulation K27 and K4 levels are much much lower than they will get at later stages of development. The strong K27 and K4 signals in these images are entirely from mitotic nuclei, which pack the methyl marks close enough to strengthen the signal much above background. PCNA (like most TFs) is excluded from replicating nuclei due to the lack of a nuclear membrane to keep it localized, so it is not surprising they don’t co-associate. Consequently this result I don’t find convincing of anything. It should be done in late stage embryos with strong interphase K27 and K4 signals.
- Result 2, (Fig 1): using EdU to distinguish early from late S phase. detect modified histone early (when EdU is disperse and strong) but not late. Maybe. Few images look suggestive. Quantification over number of cells would be a nice addition rather than just showing 4 cells each.
- Result 3 (Fig 2): PLA with PCNA and PTM-histones vs PcGs. Well there’s loads more PcG proteins at this point than modified histones. Do this with polII or some highly expressed factor which is not expected to survive the replication fork and show it also fails to create signal. To show a weak K27 signal goes to undetectable is not convincing that it is lost, especially given how much the H3 signal drops as well (from bright beaming nuclei to a scattering of dots in some nuclei.
- Result 4 (fig 4): Pull down of PCNA after the RF find histones and PcGs but not modified histones (PCNA-rechip). These results are much more convincing. In a 2-15 hr collection most of the chromatin is from much later embryos than the imaging and should have plenty of K27me3 chromatin. Unlike the previous results, these results also seem to include appropriate positive and negative controls. PcGs and unmodified histones are present following RFs but not modified histones.
- Result 5 (fig 5) BrdU pull down with pulse chase: Trx and Pc bound before and after replication. Why is this result not also shown with K4 and K27? (e.g. in fig 5C?)
- Fig 6: Again, it would be nice if the results form post-cell-cycle 16 embryos were shown instead of gastrulating embryos, to alleviate the levels problem. (And I wasn’t aware one could still write ‘data not shown’ in the age of 50 page long online supplements. If the data is part of the case I think it should be shown).
- Problem: the PLA on H3 shows a dramatic reduction in signal compared to H3 alone. the H3K27 during gastrulation is dramatically less than the H3 signal, and dramatically less than embryos post-graduation. (E) shows they can detect H3K27me3, but 2 hrs later is a substantial time gap since the start of gastrulation which is sufficient time of K27me3 to accumulate. It would be nice to see the 5 min pulse EDU for K27 performed at the developmental stage 2 hrs afterward (same stage as in panel E).
Discussion / Conclusions:
- It’s not clear to me that any Polycomb dependent epigenetic memory is being transmitted through cell divisions at gastrulation — certainly for the hox and en loci all the original transcription factors which set up this pattern are still robustly expressed, so it may be argued that the Pc machinery is not necessary. So all the imaging work focusing on gastrulation phases is much less convincing than it might have been in later stages when the marks are strong and where removal of Pc function gives phenotypes.
- However, perhaps the rapid replications (under an hour) and short G2 phases may be the reason why there is little K4 and K27 in early Drosophila development (where these marks seem to play a very prominent role from the earliest phases of vertebrate development). (Rather than little K27 because the Pc system isn’t really active, just Pc and company are there maternally).
- these results suggest the Pc doesn’t need any K27me3 to be bound at the appropriate target sites (in agreement I think with your in vitro work?).
- Why are K27/K4 in the parental strand ahead of the replication fork not detected by these assays (e.g. the PLA or CAA experiments?)
- Rather than ‘histone modifications don’t function as epigenetic marks’, how about PcGs relay the memory signal through replication and modified histones through mitosis? The modified histones are easiest to detect in early embryos in mitotic nuclei (dramatically brighter), substantially more so than the PcGs (some of which do remain associated but maybe not at all the correct places based on Nicole’s genomic chip-seq?)