DOT1L complex regulates transcriptional initiation in human erythroleukemic cells

Significance

DOT1L, the only H3K79 methyltransferase in human cells, forms a complex with AF10/AF17 and ENL/AF9, is dysregulated in mixed-lineage leukemia (MLLr), and is believed to regulate transcriptional elongation without direct evidence. Here, functional genomic, proteomic, and biochemical studies aimed at an understanding of the role of DOT1L in erythroleukemic cells show no major role in elongation but, surprisingly, reveal that DOT1L evidently has a role in transcriptional initiation by recruiting initiation factor TFIID and enhancing H2B mono-ubiquitination by restricting deubiquitination by the Spt-Ada-Gcn5-acetyltransferase (SAGA) complex. The results thus implicate the DOT1L complex in transcriptional initiation and raise the intriguing possibility that MLLr leukemic fusion proteins may also promote transcriptional initiation through modified DOT1L complexes.

Abstract

DOT1L, the only H3K79 methyltransferase in human cells and a homolog of the yeast Dot1, normally forms a complex with AF10, AF17, and ENL or AF9, is dysregulated in most cases of mixed-lineage leukemia (MLLr), and has been believed to regulate transcriptional elongation on the basis of its colocalization with RNA polymerase II (Pol II), the sharing of subunits (AF9 and ENL) between the DOT1L and super elongation complexes, and the distribution of H3K79 methylation on both promoters and transcribed regions of active genes. Here we show that DOT1L depletion in erythroleukemic cells reduces its global occupancy without affecting the traveling ratio or the elongation rate (assessed by 4sUDRB-seq) of Pol II, suggesting that DOT1L does not play a major role in elongation in these cells. In contrast, analyses of transcription initiation factor binding reveal that DOT1L and ENL depletions each result in reduced TATA binding protein (TBP) occupancies on thousands of genes. More importantly, DOT1L and ENL depletions concomitantly reduce TBP and Pol II occupancies on a significant fraction of direct (DOT1L-bound) target genes, indicating a role for the DOT1L complex in transcription initiation. Mechanistically, proteomic and biochemical studies suggest that the DOT1L complex may regulate transcriptional initiation by facilitating the recruitment or stabilization of transcription factor IID, likely in a monoubiquitinated H2B (H2Bub1)-enhanced manner. Additional studies show that DOT1L enhances H2Bub1 levels by limiting recruitment of the Spt-Ada-Gcn5-acetyltransferase (SAGA) complex. These results advance our understanding of roles of the DOT1L complex in transcriptional regulation and have important implications for MLLr leukemias.

Cell type-specific transcription is fundamental to the development of multicellular organisms, and mutations of transcription factors (TFs) are found in more than 50% of all cancer cases (1). Among the three eukaryotic nuclear RNA polymerases (Pols I, II, and III), the regulation of Pol II transcription is the major focus of research because of the large number of protein-coding genes with varying lengths, functions, and complexities of underlying regulatory mechanisms. Transcription is divided into multiple stages that include initiation, elongation, and termination. Initiation by Pol II transcription is dependent upon formation of a preinitiation complex (PIC) in association with general transcription factors (GTFs) TFIIA, -B, -D, -E, -F, and -H at the transcription start sites (TSSs) and the creation and stabilization of a transcription bubble (2). TFIID contains TATA binding protein (TBP) and 13 TBP-associated factors (TAFs), and is critical for initial recognition of core promoter elements and subsequent PIC assembly (3). Interestingly, three of the TAFs (TAF9, -10, and -12) are also components of the universal Spt-Ada-Gcn5-acetyltransferase (SAGA) complex, which contains TBP loading, activator binding, acetyltransferase, and deubiquitinase modules (4, 5). Studies in Saccharomyces cerevisiae have clearly established SAGA functions in TBP loading and transcriptional initiation, as well as redundant functions with TFIID for some genes (4–6). Studies in metazoans have established a postinitiation role for SAGA (7), although functions at the level of transcriptional initiation have not been thoroughly investigated. Nevertheless, compared with our understanding of the molecular details of ordered PIC assembly by the GTFs (3), the regulation of PIC assembly by TFs and epigenetic regulators is less well understood.

In metazoans, elongation by Pol II includes promoter clearance, promoter-proximal pause release of Pol II, and productive elongation, with initiation and pause release both recognized as critical checkpoints of transcriptional regulation (8). The binding of NELF and DSIF to elongating Pol II 20 to 80 nt downstream of TSSs stabilizes its promoter-proximal pause, and the release requires kinase activity of P-TEFb, a heterodimer of CDK9 and Cyclin T1, and the PAF1 complex (PAF1C) (9, 10). P-TEFb activity is elaborately regulated through incorporation into both the 7SK snRNP ribonucleoprotein complex, in which its kinase activity is constrained (11), and the multiprotein super elongation complex (SEC), in which it is active. The SEC is composed of AFF1 or AFF4, AF9 or ENL, and ELL1, in addition to P-TEFb (12, 13), with some differences in specificity between the different complexes (12, 13). Among the SEC subunits, AFF1, AFF4, AF9, ENL, and ELL1 are common fusion partners of MLL1 in mixed-lineage leukemia (MLLr), which accounts for over 70% of infant leukemia cases and ∼10% of adult acute myeloid leukemia cases (14). MLL1 is a member of the human SET1/MLL family of methyltransferases that affect chromatin structure and gene expression by methylation of H3K4 at key regulatory regions of the genome (15, 16). In particular, promoter-associated H3K4 trimethylation (H3K4me3) has been shown to stimulate transcriptional initiation by serving as a binding site for the TAF3 subunit of TFIID (17). MLL fusion proteins (MLL-FPs), arising from chromosomal translocations that lead to in-frame fusions of an MLL1 5′ fragment to 1 of the more than 60 fusion partner genes, are sufficient to induce MLLr, but usually require DOT1L for the activation of key target genes through incompletely understood mechanisms (18).

DOT1L, the only known H3K79 methyltransferase in human cells and a homolog of yeast Dot1, normally exists in a complex with AF10, AF17, and AF9 or ENL, two subunits shared with the SEC (19, 20); and these components are also found in a larger complex (DOTcom) containing other components (21). Among the DOT1L complex components, DOT1L is known to antagonize deacetylase SIRT1-mediated epigenetic silencing (22), AF10 stimulates the conversion of H3K79 monomethylation to di- and tri-methylation (23), AF9 and ENL bind acetylated H3 to facilitate recruitment of DOT1L (24, 25), and the function of AF17 is less well understood. Moreover, like AF9 and ENL, AF10 and AF17 are also fusion partners of MLL1 in MLLr (14). The DOT1L complex has been assumed to be involved in transcriptional elongation on the basis of its association with Pol II, its shared subunits with SEC, and the association of H3K79 di- and tri-methylation with the bodies of active genes, although evidence for a direct role in elongation is lacking (20). Mechanistically, H3K79 methylation by Dot1/DOT1L requires binding to monoubiquitinated H2BK120 (H2Bub1) in both yeast and metazoans (26). Of interest in this regard, Dot1 has been reported to promote H2Bub1 in yeast (27), likely by directly acting on Bre1-mediated ubiquitination, whereas the homologous DOT-1.1 suppresses H2Bub1 in Caenorhabditis elegans (28). Possible effects of DOT1L on H2Bub1 in human cells have not been investigated.

To understand the role of the DOT1L complex in transcriptional regulation, and ultimately in MLLr, we performed functional genomic, proteomic, and biochemical studies in human erythroleukemic cells. While our results show only minimal effects of DOT1L on transcriptional elongation in these cells, they do indicate a role of the DOT1L complex in transcriptional initiation, likely by facilitating the recruitment of TFIID in both non-MLLr and MLLr cells, and also show that DOT1L stimulates H2B ubiquitination by limiting chromatin occupancy of the SAGA complex.

Results

DOT1L Promotes the Chromatin Association of Pol II in Human Cells.

We chose two human erythroleukemia cell lines, HEL and K562, to investigate the role of the DOT1L complex in transcriptional regulation in non-MLLr cells in view of the DOT1L requirement for erythropoiesis (29) and the ease of performing functional genomic studies with these cells. Promoter-associated Pol II and gene body-associated Pol II are highly phosphorylated on serine 5 (ser-5) and serine 2 (ser-2) of their C-terminal domain (CTD), respectively, and changes in CTD phosphorylation usually reflect changes in transcriptional status. To assess effects of DOT1L on transcription, we knocked it down in HEL cells by a lentiviral short-hairpin RNA and monitored effects on cellular levels (by immunoblot) and chromatin-associated levels (by chromatin immunoprecipitation [ChIP]-qPCR) of total and phosphorylated forms of Pol II (SI Appendix, Supplementary Methods). DOT1L knockdown (KD) markedly reduced the global level of CTD ser-2 phosphorylated Pol II, suggestive of effects on transcriptional elongation (SI Appendix, Fig. S1A). Consistent with the immunoblot result, the chromatin occupancy of ser-2 phosphorylated Pol II on the c-MYC and CTNNB1 genes decreased in parallel to decreased DOT1L occupancy after DOT1L KD (SI Appendix, Fig. S1 B and C). Interestingly, the chromatin occupancies of total and ser-5 phosphorylated Pol II on these genes also decreased, suggesting that the effects of DOT1L loss on transcription are likely not limited to elongation (SI Appendix, Fig. S1 D and E).

To determine if DOT1L regulates global Pol II occupancy in HEL cells, we performed ChIP-sequencing (ChIP-seq) experiments for total and ser-2 phosphorylated Pol II (SI Appendix, Supplementary Methods). Notably, DOT1L KD reduced their global occupancies, pointing to a general role for DOT1L in promoting the chromatin association of Pol II in human cells (Fig. 1 A–C). To assess the effect of DOT1L KD on the transcriptome of HEL cells, we performed RNA-sequencing (RNA-seq) experiments (SI Appendix, Supplementary Methods). A subset of genes showed significant expression changes with 820 down-regulated and 552 up-regulated after DOT1L KD (SI Appendix, Fig. S2). To confirm that DOT1L plays a general role in promoting chromatin association of Pol II in non-MLLr human leukemic cells, we also performed total Pol II ChIP-seq experiments in control and DOT1L knockout (KO) K562 cells generated by the CRISPR/Cas9 technique (Fig. 1D and SI Appendix, Supplementary Methods). The reduced global occupancy of Pol II after DOT1L KO in these cells provided further support for a general role of DOT1L in promoting Pol II occupancy (Fig. 1 E and F). In an assessment of the effect of DOT1L loss on the transcriptome of K562 cells, an RNA-seq analysis revealed 1,105 down-regulated and 1,551 up-regulated genes in DOT1L KO cells relative to control cells (SI Appendix, Fig. S3 A and B). To identify direct targets of DOT1L, we first performed corresponding ChIP-seq experiments in K562 cells and identified 11,040 peaks, corresponding to 8,619 DOT1L-bound genes, with a false-discovery rate (FDR) < 0.01 (SI Appendix, Fig. S3C). In addition, in agreement with a recent study (30), we found that DOT1L occupies both promoters and enhancers in K562 cells (SI Appendix, Fig. S3C). Subsequently, we compared the DOT1L ChIP-seq data with the RNA-seq data, and identified 894 direct target genes, with 452 being down-regulated and 442 being up-regulated (SI Appendix, Fig. S3 D and E).

Fig. 1.

DOT1L promotes the chromatin association of Pol II in human cells. (A and B) Comparison of the occupancies of total Pol II (A) and Pol II (ser-2p) (B) on an average gene in DOT1L KD versus control HEL cells by ChIP-seq. (C) Normalized read distribution of total and ser-2 phosphorylated Pol II ChIP-seq experiments within the c-MYC locus in DOT1L KD versus control HEL cells. (D) Characterization of a DOT1L KO K562 cell line by Western blot. (E) Comparison of total Pol II occupancies on an average gene in control and DOT1L KO cells. (F) Normalized read distribution of total Pol II ChIP-seq within the c-MYC locus in DOT1L KO versus control K562 cells. (G) A bar graph comparing Pol II occupancy changes on direct target and nontarget genes of DOT1L. Genes were divided into four groups according to their DOT1L binding status and their expression changes upon DOT1 loss as follows: group I: DOT1L-bound genes with expression changes; group II: DOT1L-bound genes without expression changes; group III: DOT1L-nonbound genes with expression changes; group IV: DOT1L-nonbound genes without expression changes. Genes with Pol II occupancies below the cutoff were included in the figure but not analyzed for occupancy changes.

To compare changes of Pol II occupancy on DOT1L direct target and nontarget genes we focused on K562 cells and divided the gene population into four groups: group I, DOT1L-bound genes with mRNA level changes upon DOT1L loss (herein defined as DOT1L direct targets); group II, DOT1L-bound genes without mRNA level changes upon DOT1L loss; group III, DOT1L-nonbound genes with mRNA level changes upon DOT1L loss; and group IV, DOT1L-nonbound genes without mRNA level changes upon DOT1L loss (Fig. 1G). We consider group I genes as DOT1L direct targets and groups II to IV genes as DOT1L nontargets. As summarized in Fig.1G, we found 1) that in both group I and group II, over one-third of the genes show significant reductions of Pol II occupancy on TSSs, with much lower percentages (less than 5%) of genes showing significant increases of Pol II occupancy on TSSs; 2) that in group III, the percentage (less than 15%) of genes showing significant reductions of Pol II occupancy at TSSs is much lower than the percentages for groups I and II and, further, that the percentages of genes showing significant reductions of Pol II occupancy on TSSs and genes showing significant increases of Pol II occupancy on TSSs are comparable; and 3) that in group IV, less than one-quarter of the genes exhibit Pol II occupancy above the cutoff, with the percentage (8%) of genes showing significant reductions of Pol II occupancy on TSSs being even lower than that of group III (Fig. 1G). Together, these results indicate, most notably for DOT1L-bound genes, that Pol II occupancy on a significant fraction of these genes is dependent on DOT1L. An interesting, but as yet unanswered, question is why mRNA expression levels of group I genes are more sensitive than those of group II genes to DOT1L loss. Possible explanations (see also below) include: 1) nonlimiting interactions and transcription functions of factors (e.g., TFIID) regulated by DOT1L, and 2) reciprocal adjustments between mRNA synthesis and degradation rates (transcript buffering) (31).

The general role of DOT1L in the regulation of Pol II occupancy, in particular on DOT1L-bound genes where effects could be more direct, also led us to wonder if there is a positive correlation between the occupancies of DOT1L and Pol II. To this end, we performed more detailed comparative analyses of DOT1L and Pol II ChIP-seq data in K562 cells, and uncovered a positive correlation between them (SI Appendix, Fig. S3F). The Pol II occupancy levels on genes are generally known to positively correlate with the corresponding mRNA levels (32), which together with the positive correlation between the DOT1L and Pol II occupancies observed here, suggest that levels of DOT1L occupancy may positively correlate with mRNA level of genes as well. Indeed, subsequent comparative analyses of the DOT1L ChIP-seq and RNA-seq data confirmed a positive correlation between them (SI Appendix, Fig. S3G).

Having uncovered a general role for DOT1L in promoting the chromatin occupancy of Pol II in non-MLLr cells, we next asked if DOT1L regulates Pol II occupancy in MLLr cells. For this purpose, we treated THP1 and MOLM-13 (MLLr) cells with a potent DOT1L inhibitor (SGC0946) previously shown to effectively inhibit DOT1L activity in these cells (33). As shown by the H3K79me2 ChIP-seq analyses in SI Appendix, Fig. S4 A and B, DOT1L inhibition significantly reduced the global levels of H3K79me2 in THP1 cells, with 19,743 and 14,896 observed peaks in DMSO- and SGC0946-treated cells, respectively. Beyond effecting a dramatic reduction of DOT1L occupancy, the DOT1L inhibitor also reduced both the Pol II occupancy and expression of common key target genes (HOXA9 and MEIS1) of DOT1L and MLL-AF9 (SI Appendix, Supplementary Methods and Fig. S4 C–H). These data thus suggest that DOT1L is a regulator of Pol II chromatin occupancy not only in non-MLLr leukemic (HEL and K562) cells but also in MLLr leukemic cells. However, since we, like others (34), have found that SGC0946 and its analogs are capable of dissociating DOT1L from chromatin, the observed loss of Pol II promoter occupancy and gene expression cannot necessarily be attributed to loss of H3K79 methylation per se and, instead, may reflect the concomitant loss of chromatin binding by DOT1L and interacting factors such as TFIID.

DOT1L May Not Play a Major Role in Transcriptional Elongation in Human Erythroleukemic Cells.

Transcriptional elongation mainly includes pause release and productive elongation. DOT1L has been considered an elongation factor based mainly on correlations but without direct measurements of elongation per se; and indeed, a recent study revealed that DOT1L does not regulate pause release in mouse embryonic stem cells, but may affect far downstream elongation (35). To assess if DOT1L affects pause release in erythroleukemic cells, we calculated the traveling ratio (TR) of Pol II (36) with total Pol II ChIP-seq data from HEL and K562 cells, respectively (Fig. 2A and SI Appendix, Supplementary Methods). DOT1L KD (HEL cells) and KO (K562 cells) had little effect on the TR of Pol II (Fig. 2 B and C), consistent with results of the earlier embryonic stem cell study (35).

Fig. 2.

DOT1L may not play a major role in transcriptional elongation in human erythroleukemic cells. (A) Schematic representation describing the calculation used to determine the TR at each Pol II-bound gene. (B) Comparison of the TR of total Pol II in control and DOT1L KD HEL cells. (C) Comparison of the TR of total Pol II in control and DOT1L KO K562 cells. (D) Comparison of the distribution of engaged Pol II near TSSs in control and DOT1L KD HEL cells by PRO-seq. (E) Normalized read distribution of PRO-seq within the c-MYC locus in DOT1L KD versus control HEL cells. (F) Comparison of the TR of engaged Pol II in control and DOT1L KD HEL cells. (G) A volcano plot of 4sUDRB-seq showing DOT1L-bound genes with significant expression change in DOT1L KO versus control K562 cells. (H) Heatmap of 4sUDRB-seq showing DOT1L-bound genes with significant expression change in DOT1L KO versus control K562 cells. (I) Normalized read distribution of 4sUDRB-seq experiments comparing Pol II elongation rate on UVRAG in DOT1L KO versus control K562 cells. (J) Hidden Markov model (HMM) of elongation rate calculation for UVRAG in control and DOT1L KO K562 cells. (K) Range of Pol II elongation rate on DOT1L-bound genes in control and DOT1L KO K562 cells. (L) Pol II elongation rate change of DOT1L-bound genes in DOT1L KO versus control K562 cells. (M) A volcano plot of DOT1L-bound genes with Pol II elongation rate changes in DOT1L KO versus control K562 cells.

To validate this finding, we performed precision nuclear run-on sequencing (PRO-seq) in control and DOT1L KD HEL cells to analyze the distribution of engaged Pol II and to calculate the TR (SI Appendix, Supplementary Methods). Notably, DOT1L KD reduced the global occupancy of Pol II but had almost no effect on the TR of engaged Pol II (Fig. 2 D–F), further supporting the idea that DOT1L is unlikely to play a major role in the regulation of pause release, at least in the cells analyzed here. To determine if DOT1L affects the rate of productive elongation, we employed a method (4sUDRB-seq) based on the reversible inhibition of transcriptional elongation with DRB and the labeling of newly transcribed RNA with uridine analog 4-thiouridine (4sU) (37) (SI Appendix, Supplementary Methods). While these analyses identified 1,243 down-regulated and 1,734 up-regulated genes in the total population (SI Appendix, Fig. S5), along with 518 down-regulated and 311 up-regulated DOT1L direct target genes (Fig. 2 G and H), they showed only a minimal effect of DOT1L loss on the productive elongation rates of Pol II (Fig. 2 I–M), thus suggesting that DOT1L may not play a major role in productive elongation in these cells. Together, these data suggest that DOT1L may not play a major role in transcriptional elongation in human erythroleukemic cells.

DOT1L Appears to Regulate Transcriptional Initiation in Human Erythroleukemic Cells.

Ordered binding of initiation factors TFIID/TBP, TFIIA, and TFIIB to promoters precedes and facilitates Pol II recruitment in transcriptional initiation in eukaryotic cells. The reduced Pol II occupancy near TSSs in DOT1L-depleted K562 cells (Figs. 1 A and E and 2D), along with the apparently minimal effects of DOT1L on transcriptional elongation in these cells, raised the possibility that DOT1L may regulate transcriptional initiation. To examine this possibility, we performed ChIP-seq experiments for TBP, TFIIA, and TFIIB in control and DOT1L KO K562 cells. Notably, DOT1L loss markedly reduced the global occupancies of TBP and TFIIA and the occupancy of TFIIB on a subset of genes (Fig. 3 A–G and SI Appendix, Supplementary Methods), suggesting that DOT1L may play a relatively general role in the regulation of transcriptional initiation. To determine if there is any correlation between the chromatin occupancy of DOT1L and those of TFIID and TFIIA, we performed comparative analyses of the related ChIP-seq data. We found that similar to what was observed for Pol II (SI Appendix, Fig. S3F), the chromatin occupancies of TFIID and TFIIA each positively correlate with that of DOT1L (SI Appendix, Fig. S6 A and B).

Fig. 3.

DOT1L regulates transcriptional initiation in human cells. (A–C) Genome-wide metagene profiles and heatmaps of ChIP-seq comparing the chromatin occupancies of TBP (A), TFIIA (B), and TFIIB (C) in DOT1L KO versus control K562 cells. (D) Normalized read distribution of ChIP-seq experiments comparing the occupancies of TBP, TFIIA, and TFIIB within the c-MYC locus in DOT1L KO versus control K562 cells. (E–G) Occupancy changes of TBP (E), TFIIA (F), and TFIIB (G) on promoters. (Left) Dot and density plots of occupancy changes on promoters in two replicates. Consistency between the replicates was measured by Pearson correlation coefficient. (Right) MA plots of differential occupancies on promoters upon DOT1L loss calculated from the replicates. (H) A bar graph comparing TBP occupancy changes on direct target and nontarget genes of DOT1L. Genes were divided into four groups as described in the legend to Fig. 1. Genes with TBP occupancies below the cutoff were included in the figure but not analyzed for occupancy changes.

To assess and compare effects of DOT1L loss on TFIID and TFIIA occupancies on DOT1L direct target and nontarget genes, we performed analyses on the aforementioned four groups of genes (Fig.1G). We found that upon DOT1L loss, the percentages of genes showing significant reductions of TFIID occupancies at TSSs are over 30% for both group I and group II genes, whereas the percentages of genes showing significant increases of TFIID occupancy at TSSs of these genes are very low (less than 1%). In contrast, for groups III and IV, the percentages (less than 10% in each case) of genes showing significant reductions of TFIID occupancies on TSSs are much lower than the percentages for groups I and II (Fig. 3H). Together, these results support a role for DOT1L in facilitating TFIID recruitment. Unfortunately, we did not see the same pattern for TFIIA because of variations between the two biological repeats (SI Appendix, Fig. S6C).

To determine if there is any difference between group I and II genes in TFIID and TFIIA occupancies after DOT1L loss, we compared their occupancies for these two groups of genes. We found that whereas DOT1L promotes TFIID and TFIIA occupancies on these DOT1L-bound genes regardless of whether or not their mRNA levels are affected by DOT1L loss, the group I genes that show greater expression changes upon DOT1L loss have overall lower TFIID and TFIIA occupancies (SI Appendix, Fig. S6 D and E). Considering that Pol II occupancy levels on genes generally correlate positively with mRNA levels (32), and that we have shown correlations between TFIID and Pol II occupancies on the DOT1L-bound genes, these results may help rationalize the differential effects of DOT1L loss on group I and group II gene expression, namely that lower levels of DOT1L-facilitated TFIID occupancies and associated TFIID-dependent Pol II recruitment may preferentially sensitize group I gene expression to DOT1L loss.

ENL May Regulate Transcriptional Initiation in Erythroleukemic Cells.

The ENL and AF9 paralogues are shared subunits between the DOT1L and SEC complexes, and are each capable of facilitating recruitment of the two complexes to chromatin through interactions of their YEATS domains with acetylated H3 residues (24, 25). In addition, as subunits of the SEC, ENL and AF9 have been shown to affect the chromatin occupancy of Pol II and to regulate pause release (25). Our discovery of DOT1L as an apparent general regulator of transcriptional initiation in this study and the previously demonstrated roles of both AF9 and ENL in DOT1L recruitment (24, 25) raised the possibility that ENL and AF9 might also regulate transcriptional initiation in the context of the DOT1L complex. To test this idea, we performed ChIP-seq experiments for TBP in control and ENL KO K562 cells generated by CRISPR/Cas9 methodology (Fig. 4A). The loss of ENL markedly reduces the global occupancy of TBP (Fig. 4 B, D, and E), including TBP occupancy on the representative gene MYC that is a DOT1L direct target (Fig. 4C), thereby supporting a general role in TFIID/TBP recruitment and, thus, transcriptional initiation.

Fig. 4.

ENL regulates transcriptional initiation in human erythroleukemic cells. (A) Characterization of an ENL KO K562 cell line by Western blot. (B) Genome-wide metagene profiles and heatmaps of ChIP-seq comparing the chromatin occupancies of TBP in ENL KO versus control K562 cells. (C) Normalized read distribution of ChIP-seq comparing the occupancies of TBP within the c-MYC locus in ENL KO versus control K562 cells. (D) A dot and density plot of TBP occupancy changes on promoters in two replicates. Consistency between the replicates was measured by Pearson correlation coefficient. (E) An MA plot of differential occupancies on promoters upon ENL loss calculated with the replicates. (F) Two heatmaps showing the TBP occupancy changes of DOT1L bound genes (Upper) and DOT1L direct target genes (Lower), respectively, upon ENL and DOT1L loss. (G) Normalized read distribution of 4sUDRB-seq experiments comparing the Pol II elongation rates on UVRAG in ENL KO versus control K562 cells. (H) HMM model of Pol II elongation rate calculation for UVRAG in control and ENL KO K562 cells. (I) Range of Pol II elongation rates on DOT1L-bound genes in control and ENL KO K562 cells. (J) Pol II elongation rate changes of DOT1L-bound genes in ENL KO versus control K562 cells. (K) A volcano plot of DOT1L-bound genes with Pol II elongation rate changes in ENL KO versus control K562 cells.

Since ENL is shared between the DOT1L and SEC complexes (12), we next sought to gain support for ENL functions in initiation through the DOT1L complex by comparing effects of ENL loss versus DOT1L loss on TBP occupancy changes on DOT1L-bound genes (groups I and II) and DOT1L direct target genes (group I). As indicated in Fig. 4F, we found: 1) that 37% of total DOT1L-bound genes and 36% of DOT1L direct target genes exhibit significantly reduced TBP occupancies upon either ENL or DOT1L loss; 2) that 14% of DOT1L-bound genes and 17% of DOT1L direct target genes exhibit significantly reduced TBP occupancies upon DOT1L loss, but not upon ENL loss; 3) that 27% of DOT1L bound genes and 20% of DOT1L direct target genes exhibit significantly reduced TBP occupancies upon ENL loss, but not upon DOT1L loss; 4) that 21% of DOT1L bound genes and 28% of DOT1L direct target genes exhibit no significant changes in TBP occupancies upon either ENL or DOT1L loss; and 5) that no genes exhibit opposite changes in TBP occupancy upon ENL and DOT1L loss. These correlations of overlapping effects of DOT1L and ENL loss on TBP occupancies of significant fractions of DOT1L-bound and, especially, DOT1L direct target genes are consistent with ENL functions in initiation through the DOT1L complex, although the possibility of functions through the SEC cannot be completely ruled out.

To assess whether ENL regulates productive elongation in K562 cells, we performed 4sUDRB-seq analyses in control and ENL KO cells. We found that ENL KO minimally affected the productive elongation rate of Pol II (Fig. 4 G–K), although it is possible that AF9 compensates for the loss of ENL in the context of the SEC. Notably, the absence of a requirement for ENL in productive elongation in K562 cells, coupled with a demonstrable effect of ENL on TBP recruitment on DOT1L direct target genes, is further indicative of a role for the DOT1L complex in transcriptional initiation.

DOT1L Can Recruit TFIID via Physical Interactions.

Mechanistically, DOT1L could potentially regulate transcriptional initiation either through physical interactions with initiation factors or through modulation of their chromosomal binding sites by H3K79 methylation. We first chose to understand mechanisms underlying DOT1L-mediated transcriptional initiation by unbiased analysis of its interacting proteins. To this end, we performed a large-scale coimmunoprecipitation (co-IP) using a DOT1L antibody in conjunction with nuclear extract from K562 cells and then characterized the immunoprecipitated proteins by mass spectrometry (Fig. 5A and SI Appendix, Supplementary Methods). Besides subunits of the DOT1L complex, identified proteins included subunits of TFIID (TAFs), TFIIH, SAGA, and Mediator that were previously unknown to interact with DOT1L (SI Appendix, Fig. S7A). The identification of TAFs raised the possibility that the DOT1L complex might interact with TFIID to facilitate either its recruitment or its stabilization. An association between the DOT1L and TFIID complexes in nuclear extract was confirmed by reciprocal co-IPs (Fig. 5 B and C). To assess direct interactions, a minimal recombinant DOT1L complex (comprised of DOT1L, AF9, and AF10) was purified from High Five cells infected with corresponding baculoviruses (Fig. 5D) and TFIID was purified from HeLa cells expressing FLAG-TBP (Fig. 5E). The results of a co-IP analysis indicate a strong interaction of the DOT1L complex with TFIID as monitored by TBP and TAF80 (Fig. 5F and SI Appendix, Supplementary Methods). These results are consistent with our previous demonstration of AF9 interactions with TFIID-specific TAFs (38).

Fig. 5.

DOT1L interacts with transcriptional initiation factors. (A) Silver staining of proteins immunoprecipitated by rabbit IgG (lane 1) and a DOT1L antibody (lane 2) and separated on an SDS/PAGE gel. (B) Western blot analyses of TFIID subunit TAF7 coimmunoprecipitated with DOT1L. (C) Western blot analyses of DOT1L coimmunoprecipitated with TAF7. (D and E) Silver staining of purified DOT1L complex (D) and TFIID (E) separated on SDS-PAGE gels. (F) Western blot analyses of co-IP assays with TFIID as a bait and DOT1L complex as prey. Lanes 1 and 2 contained 10% of input samples. A longer exposure is also shown for TAF80 to better reveal the TFIID input. (G) Immobilized template assays of the recruitment of TFIID by DOT1L complex on chromatin templates assembled with unmodified H2B versus H2Bub1. Lanes 1 and 2 contained 10% and 5% inputs for DOT1L complex and TFIID, respectively, with a longer TBP exposure also shown to better reveal the input TFIID.

Previous studies have shown: 1) that DOT1L, through various domains, can bind directly to nucleosomes; 2) that an additional direct DOT1L interaction with H2Bub1 enhances H3K79 methylation; and 3) that AF10 interacts not only with DOT1L but also with H3 (through H3K27) (reviewed in ref. 26). The above results led us to predict that the DOT1L complex would interact stably with a recombinant chromatin template and thereby allow us to test the ability of a chromatin-bound DOT1L complex to recruit TFIID. For this purpose, we used an immobilized chromatin template assay reconstituted with either unmodified H2B or monoubiquitinated H2B plus other core histones (SI Appendix, Supplementary Methods). As shown in Fig. 5G, and as anticipated, the DOT1L complex alone bound to chromatin containing unmodified H2B and, at a significantly higher level, to chromatin containing H2Bub1. Notably, while TFIID alone showed only marginal binding to either chromatin template (only evident upon gel overexposure to reveal input), in the presence of the DOT1L complex, it showed significantly enhanced binding to the unmodified H2B chromatin template and, consistent with the higher level of DOT1L complex binding, a corresponding higher level of binding to the H2Bub1 chromatin template. These results were confirmed in the related immobilized template assay in SI Appendix, Fig. S7B. In addition, comparing unmodified chromatin and chromatin premodified with H3K79me2 showed a decrease in DOT1L complex-mediated TFIID recruitment. These results, overall, indicate a DOT1L complex-mediated recruitment of TFIID that is enhanced by H2Bub1 and that is likely independent of DOT1L-mediated H3K79 methylation.

Neither DOT1L nor ENL Loss Affects Global Chromatin Accessibility.

H3K79 methylation is globally associated with active genes and a conformation change of H3 is required for the methylation of K79 by DOT1L (26), which raised the possibility that DOT1L may regulate global chromatin accessibility and therefore transcriptional initiation. To test this idea, we performed ATAC-seq in control, DOT1L KO, and ENL KO K562 cells (SI Appendix, Supplementary Methods). We found that the DOT1L KO minimally increased global chromatin accessibility, whereas the ENL KO had no effect (SI Appendix, Fig. S8 A and B). A closer examination of the peaks revealed that the DOT1L KO slightly decreased the percentage of peaks on promoters, whereas the ENL KO exhibited little effect (SI Appendix, Fig. S8C). However, further analyses revealed no global reduction of promoter accessibility after DOT1L or ENL KO, whereas an expected positive correlation between promoter accessibility and mRNA level was observed (SI Appendix, Fig. S8D). Altogether, these data suggest that although H3K79 methylation is associated with active genes, neither DOT1L nor ENL affects global chromatin accessibility.

To further understand the relationships among DOT1L, TFIID/TBP occupancy, and promoter accessibility, we divided promoters into five groups according to their values of accessibility changes after DOT1L or ENL loss and compared their TBP occupancy changes. As expected, TBP occupancy was found to be positively correlated with promoter accessibility (SI Appendix, Fig. S8E). Notably, we also found that even promoters with increased accessibility exhibited decreased occupancies of TBP after DOT1L loss (SI Appendix, Fig. S8E), strongly supporting a DOT1L contribution to TFIID recruitment regardless of promoter accessibility changes.

DOT1L Promotes H2Bub1 by Limiting Recruitment of the SAGA Complex.

H3K79 methylation by Dot1/DOT1L requires H2Bub1 in both yeast and metazoans, and the SAGA complex is known to deubiquitinate H2B through the DUB module (4, 5). Related, connections between DOT1L homologs and H2B ubiquitylation have been described both in yeast and in C. elegans (27, 28) but not yet in human cells. These observations, along with our identification of the SAGA complex in DOT1L immunoprecipitates (Fig. 5A and SI Appendix, Fig. S7A), led us to further study this matter. The intracellular association between DOT1L and SAGA complexes was confirmed by reciprocal co-IP experiments (Fig. 6 A and B). To determine if the two complexes are able to directly interact, we performed co-IP analysis using purified recombinant human DOT1L complex and purified SAGA from HeLa cells (Fig. 6C and SI Appendix, Supplementary Methods). We found that the SAGA complex was unable to pull down the DOT1L complex, suggesting that it is unlikely to directly interact with DOT1L complex (Fig. 6D). To determine if DOT1L affects the recruitment of SAGA or the H2Bub1 level, we performed ChIP-seq experiments for PCAF, a subunit of the SAGA acetyltransferase module, and H2Bub1 in control and DOT1L KO K562 cells. Notably, the loss of DOT1L increased the chromatin occupancy of PCAF, suggesting that DOT1L may somehow limit the recruitment of SAGA in human cells (Fig. 6 E and G). The loss of DOT1L also decreased the level of H2Bub1, suggesting that, as in yeast (27), DOT1L promotes the stabilization of H2Bub1 in human cells (Fig. 6 F and G). Since the quality of the PCAF ChIP-seq data were not high, and to validate the results, we performed CUT&Tag for PCAF in control and DOT1L KO cells and observed the same results (Fig. 6 H and I and SI Appendix, Supplementary Methods). These cumulative data, together with the recently confirmed H2Bub1 binding capability of DOT1L (26), suggest that the binding of DOT1L to H2Bub1 may prevent its deubiquitination by SAGA in human cells.

Fig. 6.

DOT1L promotes H2Bub1 by limiting recruitment of the SAGA complex. (A) Western blot analyses of SAGA subunits coimmunoprecipitated with DOT1L. (B) Western blot analyses of DOT1L complex subunits coimmunoprecipitated with TRRAP. (C) Silver staining of purified SAGA complex separated on an SDS/PAGE gel. (D) Western blot analyses of co-IP assays with antibodies to GCN5 and PCAF. (E and F) Genome-wide metagene profiles and heatmaps of ChIP-seq comparing the occupancies of PCAF (E) and H2Bub1 (F) in DOT1L KO versus control K562 cells. (G) Normalized read distribution of ChIP-seq comparing the occupancies of PCAF and H2Bub1 within the TMEM18 locus in DOT1L KO versus control K562 cells. (H) Genome-wide metagene profiles and heatmaps of CUT&Tag comparing the occupancies of PCAF in DOT1L KO versus control K562 cells. (I) Normalized read distribution of CUT&Tag comparing the occupancies of PCAF within the TMEM18 locus in DOT1L KO versus control K562 cells. (J) Model for the function of the DOT1L complex in transcription initiation and enhancement of H2Bub1 levels. DOT1L complex recruitment to promoters is facilitated by DOT1L interactions with H2Bub1 and other histone domains (not shown), AF9/ENL YEATS domain interactions with acetylated H3 tail residues, and AF10 interactions with H3K27. The DOT1L complex in turn facilitates TFIID recruitment, and subsequently initiation, through AF9/ENL interactions with specific TAFs. DOT1L interaction with H2Bub1 prevents the recruitment of, and H2Bub1 deubiquitination by, the SAGA complex. Black diamonds represent direct interactions. See text for references and further discussion.

To compare occupancy changes of SAGA on direct target and nontarget genes of DOT1L, we also performed analyses on the aforementioned four groups of genes (Figs. 1G and 3H). We found that in groups I and II, the percentages (over 7%) of genes showing significant increases in SAGA occupancy are higher than those (4.3% and 1.4%, respectively) in groups III and IV (SI Appendix, Fig. S9A), supporting a role for DOT1L in limiting SAGA recruitment on DOT1L-bound genes. To determine if there is any correlation between the chromatin occupancy of DOT1L and that of SAGA, we performed comparative analysis of the related ChIP-seq data. We found that similar to the results for Pol II, TFIID and TFIIA (SI Appendix, Figs. S3F and S6 A and B), the chromatin occupancy of SAGA positively correlates with that of DOT1L (SI Appendix, Fig. S9B), suggesting that the binding of DOT1L to H2Bub1 is just one of the mechanisms of limiting SAGA recruitment in human cells.

Discussion

DOT1L normally exists in a complex that also includes AF10, AF17, and ENL/AF9, has generally been considered to be an elongation factor without direct evidence, and is dysregulated in most cases of MLLr by incompletely understood mechanisms. Results in this study show little or no requirement for DOT1L in either pause release or productive elongation, or for ENL in productive elongation, in erythroleukemic cells. Instead, the results suggest an involvement of DOT1L and DOT1L complex in transcriptional initiation by facilitating the recruitment of TFIID and in the maintenance or elevation of H2Bub1 levels by limiting the recruitment of the SAGA complex. A model for these functions of the DOT1L complex, through various interactions that are discussed further below, is presented in Fig. 6J.

An Apparent Role for the DOT1L Complex in Transcriptional Initiation in Erythroleukemic Cells.

The DOT1L complex has been considered an elongation factor on the basis of its colocalization with Pol II, sharing of the ENL or AF9 subunits with the super elongation complex (containing elongation factors P-TEFb and ELL1), and the localization of H3K79me2 marks on the bodies of active genes. Our analyses of both TRs (by ChIP-seq and PRO-seq) and elongation rates (by 4sUDRB-seq) of Pol II indicate that the DOT1L complex is unlikely to play a major role in pause release and productive elongation in human erythroleukemic cells. This conclusion is in agreement with results of a recent mouse ES cell study that showed a DOT1L contribution to elongation only when SEC elongation functions were compromised (35). In contrast to our failure to observe a DOT1L contribution to elongation, our demonstrations of the loss of intracellular initiation factor (TFIIA and TFIID) binding to promoters upon loss of either DOT1L or ENL, a direct interaction of TFIID with the DOT1L complex, and the recruitment of TFIID to chromatin-bound DOT1L in biochemical assays provide evidence for global contributions of the DOT1L complex to transcriptional initiation, likely through direct physical interactions with TFIID in an H3K79 methylation-independent manner. Our previous demonstration of a direct interaction of AF9/ENL with TFIID through TAF5 and TAF6 (38) suggests a probable molecular basis for the direct TFIID–DOT1L complex interaction, and we view the demonstrated DOT1L complex-dependent recruitment of TFIID to a chromatin template as a TFIID recruitment/stabilization mechanism that could act cooperatively with previously established mechanisms involving interactions of TFIID subunits (TBP, TAFs) with core promoter elements, transcriptional activators, and modified histones (39). Related, mechanisms for the recruitment of the DOT1L complex to promoters (and other gene regions) include: 1) interactions of ENL/AF9 YEATS domains with acetylated H3 (24, 25), 2) interactions of AF10 with unmodified H3K27 (23), and 3) interactions of the DOT1L complex with other transcription (co)factors (21, 40). Interestingly, the high representation of Mediator in our DOT1L immunoprecipitates (Fig. 5A and SI Appendix, Fig. S7A) raises the possibility that it too may facilitate the recruitment of DOT1L to promoters to promote transcriptional initiation. Together, our results strongly suggest that the DOT1L complex is a general regulator of transcriptional initiation, although mechanistic details and potential cell- and gene-specific functions remain to be further elucidated.

The Binding of DOT1L to H2Bub1 Prevents Its Deubiquitination by the SAGA Complex.

H3K79 methylation by Dot1/DOT1L requires an interaction with H2Bub1 in both yeast and metazoans (26). Reciprocally, yeast Dot1 was found to promote H2B ubiquitylation in a manner independent of its enzymatic activity, H2Bub1 deubiquitination by the Ubp8 subunit of SAGA, and recruitment of Bre1/Rad6 by the Paf1 complex (26), and likely acts through an effect on ubiquitination by Bre1 (27). In contrast, C. elegans DOT-1.1 and ZFP-1 (homologs of human DOT1L and AF10) were found to induce Pol II pausing and to concomitantly suppress H2Bub1, indicating alternate, context-dependent functions of Dot1 homologs (28). Our results in human cells show that, as in yeast, DOT1L positively regulates the level of H2Bub1, but apparently by a different mechanism. Thus, our genomic analyses showing an increase in SAGA occupancy and a concomitant decrease in H2Bub1 following loss of DOT1L, together with the H2Bub1 binding capability of DOT1L (26), suggest that the binding of DOT1L to H2Bub1 may prevent its deubiquitination by SAGA. Of note, the ability of DOT1L to restrict H2Bub1 deubiquitination by SAGA indicates yet another role for the DOT1L complex beyond the earlier proposed roles in elongation (20, 35), in suppression of SIRT1 activity through H3K79 methylation (22), in maintaining an open chromatin configuration at a subset of enhancers (30), and in our newly indicated role in initiation. In this regard, interactions of the DOT1L complex with H2Bub1 (through DOT1L), acetylated H3 (through AF9/ENL), and TFIID (as shown here) could potentially act cooperatively both to restrict the function of the SAGA DUB module and to enhance transcriptional initiation and elongation, consistent with earlier reports of roles for H2Bub1 in initiation and elongation (41, 42). A related question concerns the potential impact of the DOT1L restriction on SAGA occupancy to known coactivator functions of SAGA through the HAT domain. However, this may not present a problem in view of the modular nature of SAGA, with separate HAT and DUB domains and distinct recruitment mechanisms for its distinct functions (5, 43), and the possibility of dynamic, position-specific interactions.

MLL FPs Could Potentially Promote Transcriptional Initiation through the DOT1L Complex.

MLL-FPs are known to induce MLLr leukemias by maintaining the expression of several key target genes, most notably HOXA9 and MESI1, that normally are highly expressed in hematopoietic progenitor cells for promotion of proliferation and protection from stress (44, 45). Considering that major fusion partners of MLL1 are subunits of either SEC or DOT1L complexes, that SEC is a critical elongation factor, and that DOT1L has been considered an elongation factor, MLL-FPs similarly are believed to act primarily by stimulating transcriptional elongation. Our discovery of the DOT1L complex, with several MLL fusion partner proteins (AF10, AF17, and ENL or AF9) as subunits, as a general regulator of transcriptional initiation suggests that MLL-FPs, through incorporation into modified (MLL-FP containing) DOT1L complexes, may also be able to stimulate both transcriptional initiation and elongation. Apart from possible recruitment of modified DOT1L complexes by mechanisms described above for the normal cellular DOT1L complexes, their binding to promoter-proximal or transcribed regions through interactions of the MLL1 N terminus with CpG islands (through the CXXC domain) (46) or with the PAF1C complex (47, 48) could also contribute to recruitment. Collectively, the various recruitment mechanisms for modified DOT1L complexes could contribute, through effects on initiation and elongation, to maintenance of the expression of key target genes of MLL-FPs in MLLr leukemias.

Materials and Methods

Cells and Cell Culture.

Human cells HEL, THP1, and MOLM-13 were cultured in RPMI-1640 + 10% FBS + 2% penicillin/streptomycin + 2 mM l-glutamine + 55 μM β-mercaptoethanol. Human cells K562 were cultured in 90% DMEM + 10% FBS + 2% penicillin/streptomycin + 2 mM l-glutamine.

See SI Appendix for additional details and descriptions of RNAi, CRISPR-Cas9 genome editing, ChIP-seq, RNA-seq, PRO-seq, 4sUDRB-seq, ATAC-seq, CUT&Tag, co-IP, immobilized template assay, and mass spectrometry analyses.

Data Availability

Next-generation sequencing data have been deposited in the Gene Expression Omnibus (GEO) database, https://www.ncbi.nlm.nih.gov/geo (accession no. GSE161367). The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (49) partner repository (dataset identifier PXD024874). All other study data are included in the article and SI Appendix.