Tet2 and Tet3 cooperate with B-lineage transcription factors to regulate DNA modification and chromatin accessibility

[…] All three reviewers thought the manuscript was interesting and represented a significant advance in defining roles of Tet2/3 in B cell development. All three reviewers also thought that additional mechanistic studies were needed to more fully support some of the major conclusions. The manuscript was viewed to present two parallel themes; one focused on roles of Tet2/3 on κ locus methylation, germline transmission and DNA rearrangement and the other on broader roles of Tet2/3 in controlling the chromatin landscape. With respect to the first theme, the studies of the κ locus do not in themselves explain the developmental block at the pro- to pre-B cell transition. With respect to broader roles of Tet2/3, the authors suggest roles of PU.1 and E2A in recruiting these enzymes to specific genomic locations, but do not examine this possibility directly. In addition, while the manuscript addresses global roles of demethylating enzymes, there is there is no global analysis of differentially methylated regions in the genome.

Essential revisions:

1) Additional experiments are needed to establish that PU.1 and/or E2A function to recruit Tet2/3 to specific genomic targets. These would include; ChIP-seq analysis of PU.1 and E2A in the DKO pro-B cells to demonstrate that these factors can still access their sites in the absence of Tet2 or Tet3.

We have performed ChIP-seq analysis for PU.1 in WT and Tet2/3 DKO pro-B cells with two biological replicates each (new Figure 6—figure supplement 2). Consistent with our model that PU.1 as a pioneer transcription factor to recruit Tet proteins, the global PU.1 binding patterns are virtually indistinguishable between WT and DKO. Of >46,600 peaks identified from both genotypes, only 18 and 4 peaks were selectively occupied (with adj. p ≤ 0.01) in WT and DKO, respectively. Upon examination, none of these 22 peaks were near the Igκ locus or other loci known to be involved in Igκ recombination (new Supplementary file 2). Thus, we conclude that the loss of TET proteins has a very minor effect, if any, on PU.1 binding. Unfortunately, we encountered technical difficulties in performing E2A ChIP and the results were inconclusive.

ChIP or ChIP-seq analysis of Tet2 and/or Tet3 in PU.1 and/or E2A knockdown cells should be performed to demonstrate that these factors are required for Tet2/3 recruitment. If Abs directed against Tet2 or Tet3 are an issue, an epitope tagged catalytic domain of Tet2 which is exploited in the complementation analysis can be used (see point 2).

We appreciate the excellent points brought up by the reviewers. We performed ChIP-qPCR for endogenous Tet2 and for ectopically expressed Tet2CD and Tet2CD HxD mutant. All three proteins associate with 3’Eκ and dEκ (new Figure 7D). We then analyzed the binding of endogenous Tet2 at these enhancers in PU.1- and E2A-depleted cells and found that Tet2 binding to dEκ was substantially decreased when either transcription factor was depleted with shRNAs (new Figures 6D and E), consistent with increased DNA methylation at this enhancer upon PU.1 or E2A knockdown (Figures 6A and B).

Co-immunoprecipitation experiments should be performed to evaluate whether or not PU.1 and/or E2A form protein complexes with Tet2/3 in solution.

We have performed co-immunoprecipitation of endogenous PU.1 or E2A with Tet2 in Abl-transformed pre-B cells and have showed that Tet2 directly associates with both PU.1 and E2A (new Figure 6C). To exclude potential indirect interactions due to the presence of contaminating nucleic acids, the experiments were performed in the presence of ethidium bromide and benzonase, a DNA intercalator and a nuclease that cleaves both DNA and RNA, respectively.

2) ChIP or ChIP-seq experiments are needed to establish that the epitope tagged Tet2 catalytic domain is specifically targeted to the κ enhancers. The ability of the Tet2 catalytic domain to drive κ rearrangement should be tested using the assay employed to analyze the consequences of loss of Tet2,3 on κ rearrangement.

As mentioned above, we found that ectopically expressed Tet2 catalytic domain (Tet2CD) and its catalytically inactive mutant, Tet2CD HxD, were well expressed in Tet2/3 DKO pro-B cells (Western blotting, Figure 8) and associated with the 3’Eκ and dEκ enhancers (ChIP-qPCR, new Figure 7D). However only Tet2CD, but not Tet2CD HxD, was able to induce Igκ rearrangement and restore chromatin accessibility at the Igκ locus (Figure 7G), demonstrating that Tet2 catalytic activity is required and that the decrease in chromatin accessibility and Igκ rearrangement in Tet2/3 DKO pro-B cells is a reversible phenomenon.

3) There is no global analysis of differentially methylated regions in the genome, though the paper deals with de-methylating enzymes. Instead, ATAC seq is used as a means of assessing the ability of transcription factors to establish regions of open chromatin. However, increases or decreases in ATAC seq peaks could result from effects of the Tet2/3 DKO that are not due to local changes in DNA methylation (e.g., changes in expression of transcription factors). Thus, a global analysis of DNA methylation in the Tet2/3 DKO would substantially strengthen the relationship between DNA methylation, ATAC sensitivity and PU.1/E2A and contribute to an improved understanding of the developmental block.

As recommended by the reviewers, we performed whole-genome bisulfite sequencing (WGBS) for WT and Tet2/3 DKO pro-B cells (two replicates each) and the data are shown in the new Figure 5. In general, as opposed to global changes, loss of Tet2/3 resulted in specific localized changes of DNA methylation, with 872 and 258 regions with increased and decreased methylation, respectively (new Figure 5).

In interpreting these data, it should be noted that bisulfite sequencing does not distinguish between 5mC and 5hmC. Given that the regions that lose accessibility in Tet2/3 DKO compared to WT are enriched in 5hmC in WT cells (Figure 4B), and that 5hmC levels are dramatically decreased in DKO cells (new Figure 3—figure supplement 2), the increase in 5mC+5hmC observed at the 872 regions in DKO cells compared to WT underestimate the real magnitude of the increase in methylation (5mC). Nonetheless, our data show clearly that the regions with increased methylation in Tet2/3 DKO pro-B cells are concomitantly less accessible (new Figure 5A), suggesting that the decreased chromatin accessibility in DKO cells correlates both with loss of 5hmC and with increased DNA methylation as expected. To further demonstrate the direct causal relationship between TET proteins and accessibility, expression of Tet2CD, but not Tet2CD HxD, was able to restore accessibility to Igκ enhancers and globally at other regions (Figures 7D, E). These data strongly suggest TET protein activity (hydroxymethylation followed by demethylation) facilitates chromatin accessibility, although it has yet to be established whether loss of 5hmC or increased 5mC plays a more significant role in decreased chromatin accessibility in Tet2/3 DKO pro-B cells.

Regarding the relationship between transcription factors, TET proteins, and chromatin accessibility, we performed ChIP-seq analysis for PU.1 as mentioned above. The overall binding patterns of PU.1 are virtually indistinguishable between WT and DKO (new Figure 6—figure supplement 2). Given that PU.1 directly binds TET2 (new Figure 6C), and PU.1 depletion results in lower TET2 binding to dEκ (new Figure 6E), we propose that PU.1 first binds to less-accessible nucleosomal DNA and recruits TET2/3, which then remodels the dEκ enhancer for further recruitment of additional transcription factors and/or chromatin regulators. For the convenience of the reader, this hypothesis is now depicted schematically in the new Figure 8D.