H3K4me1 Facilitates Promoter-Enhancer Interactions and Gene Activation during Embryonic Stem Cell Differentiation

SUMMARY

Histone H3 lysine 4 mono-methylation (H3K4me1) marks poised or active enhancers. KMT2C (MLL3) and KMT2D (MLL4) catalyze H3K4me1, but their histone methyltransferase activities are largely dispensable for transcription during early embryogenesis in mammals. To better understand the role of H3K4me1 in enhancer function, we analyze dynamic enhancer-promoter (E-P) interactions and gene expression during neural differentiation of the mouse embryonic stem cells. We found that KMT2C/D catalytic activities were only required for H3K4me1 and E-P contacts at a subset of candidate enhancers, induced upon neural differentiation. By contrast, a majority of enhancers retained H3K4me1 in KMT2C/D catalytic mutant cells. Surprisingly, H3K4me1 signals at these KMT2C/D-independent sites were reduced after acute depletion of KMT2B, resulting in aggravated transcriptional defects. Our observations therefore implicate KMT2B in the catalysis of H3K4me1 at enhancers, and provide additional support for an active role of H3K4me1 in enhancer-promoter interactions and transcription in mammalian cells.

Graphical Abstract

eTOC Blurb

Despite the association of H3K4me1 with distal enhancers, inactivation of KMT2C/D, the major mono-methyltransferases for H3K4me1, results in mild transcriptional defects in cells. Kubo et al. show that KMT2B contributes to KMT2C/D-independent H3K4me1 at enhancers and transcription. Their results suggest that H3K4me1 indeed facilitates enhancer-promoter contacts and enhancer-driven transcription during differentiation.

INTRODUCTION

Transcriptional enhancers regulate the spatiotemporal gene expression in metazoans, through binding of sequence-specific transcription factors (TFs), which recruit chromatin remodeling complexes such as SWI/SNF proteins and chromatin modifiers^1–4. Mono-methylation of histone H3 lysine 4 (H3K4me1) is a canonical mark of both poised and active transcriptional enhancers and has been broadly utilized to identify and annotate enhancers in the genome^5–12. H3K4me1 at enhancers is catalyzed by the histone methyltransferases (HMTs) including KMT2C and KMT2D (KMT2C/D, also known as MLL3 and MLL4)^13–16. Enhancer activation is accompanied by the presence of histone modification mark H3K27ac, which is catalyzed by histone acetyl transferases such as CBP/p300, recruitment of which is known to be facilitated by KMT2C/D^17,18. KMT2C and KMT2D are integral components of a large multi-protein complexes known as the COMPASS (complex proteins associated with Set1) complex that also include WDR5, RBBP5, ASH2L and DPY30^15,19,20. The COMPASS complex and in particular KMT2C/D play crucial roles in mammalian development and tumorigenesis. Furthermore, Kmt2c/d as well as other components of the COMPASS complex are among the most frequently mutated genes in human cancers and developmental disorders^21–25. Better characterization of KMT2C and KMT2D’s function in transcriptional activation in mammalian cells is therefore of critical importance for developmental biology and cancer research.

Both Kmt2c and Kmt2d encode large proteins encompassing a diverse array of functional domains including the N-terminal catalytic SET domain responsible for methyltransferase activity, and several protein-protein interaction domains such as the PHD (plant homeodomain), PHD-like, TPR (tetratricopeptide repeat), and RRM (RNA recognition motif) domains^15,21,26. In early embryogenesis, KMT2D and KMT2C play partially redundant roles in gene regulation, thus Kmt2d knockout in mice display embryonic lethality^14,27, and development of heart, adipose, muscle, and immune cells is severely impeded after Kmt2c/d depletion^14,28,29. Transcriptional defects after Kmt2c and Kmt2d deletions are not entirely attributed to their HMT activities. A recent study demonstrated that catalytic inactivation of KMT2C/D causes loss of H3K4me1 at enhancers along with partial reduction of H3K27ac in mouse embryonic stem cells (ESCs), but with surprisingly minor effects on gene expression unlike that in Kmt2c/d knockout cells^30,31. Additionally, catalytic inactivation of Trr, the Drosophila homolog of Kmt2c/d, does not impede Drosophila development³² and KMT2C/D catalytic activities are largely dispensable for mouse early embryonic development in a lineage-selective manner³³. These observations certainly support HMT-activity independent functions of KMT2C/D, but also raise question about the functional role of H3K4me1 in enhancer-dependent gene activation.

We previously showed that loss of KMT2C/D proteins resulted in broad chromatin reorganization, altered transcriptional programs, and reduced Cohesin complex occupancy at a large number of enhancers, in the mouse ESCs and Neural Progenitor Cells (NPC)³⁴. We further demonstrated that catalytic activities of KMT2C/D are required for enhancer-promoter contacts and optimal gene expression of the Sox2 gene³⁴. However, these initial observations still left open the question whether KMT2C/D-dependent H3K4me1 facilitates enhancer-promoter (E-P) contacts beyond the Sox2 gene, and whether H3K4me1-dependent E-P contacts contribute to transcriptional activation during mESC differentiation. To gain a better understanding of H3K4me1’s role at enhancers, it is essential to precisely determine the extent to which KMT2C/D catalytic activities are involved in the dynamics chromatin contacts between enhancers and promoters, and expression of their target genes^35,36. Here we used mouse ESCs with catalytically deficient KMT2C/D (hereafter referred to as dCD), which display global defect of H3K4me1 deposition comparable to those in KMT2C/D knockout cells^30,37,38, to delineate the role of KMT2C/D activities in histone H3K4 methylation at enhancers, E-P contacts, and transcriptional induction during ESC differentiation (Figure 1A; Table S1). As a model for cellular differentiation, we focused on retinoic acid (RA)-induced neural differentiation toward neural precursor cells (NPCs) (Methods)^39,40. Our results showed that the HMT activities of KMT2C and KMT2D mediate H3K4me1 and E-P contacts mainly at newly formed candidate enhancers during neural differentiation, but H3K4me1 levels at the majority of candidate enhancers remain unperturbed in the mutant mESC cells. We further showed that KMT2B contributes to H3K4me1 at KMT2C/D-independent enhancers. Acute depletion of KMT2B in mESC resulted in severe decrease of H3K4me1 levels at thousands of additional candidate enhancers in both the WT and dCD cells, and aggravated the transcriptional defects in dCD cells. Our results therefore suggest that KMT2B contributes to H3K4me1 at enhancers, and H3K4me1 actively promotes E-P interactions and enhancer-dependent gene expression.

Figure 1 |. KMT2C/D (MLL3/4) catalytic activity is required for H3K4me1 at candidate enhancers induced during differentiation of ESC to NPC.

(A) Mouse embryonic stem cells (mESCs) expressing wild-type (WT) or catalytically deficient (dCD) mouse KMT2C and KMT2D proteins are induced to differentiate towards neural progenitor cells (NPC). Cells at Day 0, Day 2.5 and Day 5 of differentiation were subject to various genomic assays..

(B)(C) Heatmaps showing histone modifications and chromatin accessibility at candidate enhancers in WT and dCD ESC and NPC cells. In (B), candidate enhancers with increased H3K4me1 in NPCs (FDR < 0.05, log₂FC > 0.5) (N=3,373) were shown. In (C) candidate enhancers showing persistent H3K4me1 levels were shown (N=30,404) (C). The candidate enhancers were further classified into enhancers KMT2C/D-dependent (N=3,028 and 4,150) and KMT2C/D-independent (N=345 and 26,254). Data of two biological replicates are shown.

(D) The top 5 TF binding motifs enriched at KMT2C/D-dependent (left) and -independent candidate enhancers (right) are shown. (see Figure S1I for the same analysis of persistent candidate enhancers). See also Figures S1 and S2.

RESULTS

KMT2C/D Catalytic Activity is required for H3K4me1 at a subset of Candidate Enhancers

We first conducted ChIP-seq experiments to analyze how catalytic inactivation of KMT2C/D methyltransferase altered histone modification genome-wide in dCD cells during ESC differentiation to NPC. Consistent with previous reports^13–15,30, we observed a reduction of H3K4me1 at 19,454 and 25,271 distal elements in ESCs and NPCs, respectively (FDR < 0.05, log₂ FC > 0.5), similar to what we observed previously in KMT2C/D knockout cells³⁴ (Figures S1A–S1F). H3K27ac ChIP-seq signals at the same regions were partially reduced and the degree of changes was positively correlated with the change in H3K4me1 (Figure S1G). Surprisingly, significantly elevated H3K4me1 signals remained at a large number of gene promoters (11,001 and 16,175 loci in ESCs and NPCs, respectively), as well as many distal elements (Figures S1A, S1B, and S1D), possibly due to activities of other histone H3 lysine 4 methyltransferases in the cells^15,19,20. We defined the candidate enhancers as distal regions (distance to transcription start site ≥ 10 kb) with both H3K4me1 and H3K27ac ChIP-seq signals. In total, 35,744 and 33,777 candidate distal enhancers were identified in ESCs and NPCs, respectively. During NPC differentiation, 3,373 candidate enhancers gained H3K4me1 signals (FDR < 0.05, log₂ FC > 0.5) along with increased chromatin accessibility as profiled previously by ATAC-seq^41,42. Interestingly, 90% (N = 3,028) of them failed to acquire H3K4me1 in KMT2C/D dCD cells, suggesting that KMT2C/D catalytic activity plays a key role in the de novo H3K4me1 at these candidate enhancers induced during NPC differentiation (Figure 1B; Table S2). Similarly, acquisition of H3K27ac at these newly formed candidate enhancers in NPC was also impaired in the dCD cells, whereas H3K27me3 signals gained slightly at these regions. To our surprise, H3K4me1 was retained at over 26,000 candidate enhancers in KMT2C/D dCD NPCs. Most of these KMT2C/D-independent candidate enhancers were already associated with H3K4me1 in ESC in general (Figure 1C). Many of them (14,778 loci) were also annotated as poised enhancers in ESCs and gained H3K27ac signals during NPC differentiation, suggesting a transition from poised to active state of these candidate enhancers³⁰. Motif enrichment analysis showed that the newly formed candidate NPC enhancers were likely driven by GATA family of transcription factors, evidenced by enrichment of motifs of GATA family (Figure 1D). Likewise, the KMT2C/D-dependent persistent enhancers were likely driven by NANOG and SOX2, while KMT2C/D-independent persistent candidate enhancers are likely driven by TEAD1, 4 and SOX3 (Figure S2)^43–47. These findings suggest that KMT2C/D catalytic activity plays a crucial role in setting up H3K4me1 at a subset of candidate enhancers, especially those that were induced during NPC differentiation, but is dispensable for H3K4me1 at the majority of candidate enhancers. This result suggests the existence of KMT2C/D-independent mechanisms responsible for H3K4me1 histone modification at enhancers.

KMT2C/D Catalytic Activities are Required for New E-P Contacts in NPC

We next investigated the impact of the loss of KMT2C/D catalytic activity on E-P contacts in ESC and NPC. We performed PLAC-seq (also known as HiChIP)^48,49 using antibodies against the promoter histone mark H3K4me3 to map chromatin contacts anchored at active or poised promoters, which enabled us to identify E-P contacts more effectively than conventional Hi-C analysis. We obtained between 280 and 430 million paired-end reads for each replicate of PLAC-seq experiment (Table S1). We carried out differential chromatin contacts analysis focusing on ~12,000 gene promoters with similar levels of H3K4me3 ChIP-seq signals using a negative binomial model for each distance-stratified 10-kb interval and observed genome-wide changes of E-P contacts at candidate distal enhancers including validated cis-regulatory elements determined by CREST-seq^50–53 (Figure S3, Methods). We observed significant decreases of E-P contacts upon loss of KMT2C/D catalytic activity predominantly in the NPCs (2,252 reduced contacts by FDR < 0.05, 7478 reduced contacts by p value < 0.01), moderate reduction of E-P contacts in the ESCs (43 reduced contacts by FDR < 0.05, 1,778 reduced contacts by p value < 0.01) (Figures S4A and S4B; Table S3). Importantly, the majority of E-P contacts induced during NPC differentiation in WT cells (2,545 induced contacts, FDR < 0.05) failed to form in KMT2C/D dCD cells (529 induced contacts, FDR < 0.05), suggesting that the KMT2C/D-dependent H3K4me1 is required for the formation of E-P contacts induced during NPC differentiation (Figures 2A, 2B, and S4C–S4F). As a control, induction of E-P contacts at the 345 KMT2C/D-independent newly formed NPC enhancers (N=111 E-P contacts) were less affected by the loss of KMT2C/D catalytic activity than at the 3,028 KMT2C/D-dependent enhancers (Figures 2C and S4G). This trend was also observed in E-P contacts anchored at promoters of genes that were activated during NPC differentiation (1303 NPC-induced genes, see also Figure 3C) (Figures 2D and 2E). These results support a general role of KMT2C/D-dependent H3K4me1 at enhancers in the establishment of chromatin contacts, as we previously reported for the Sox2 gene³⁴.

Figure 2 |. KMT2C/D catalytic activities are required for the formation of enhancer-promoter (E–P) contacts during NPC differentiation.

(A) Scatter plots showing differential chromatin contacts anchored on promoters (y-axis) between ESCs and NPCs in WT and KMT2C/D dCD (right) cells. Significantly induced and reduced chromatin contacts are shown as red and blue dots, respectively (FDR < 0.05).

(B) Genome browser snapshots of Zbtb16 gene, which depends on KMT2C/D for induction upon NPC differentiation. The arcs denote chromatin contacts between Zbtb16 promoters and nearby candidate enhancers. The colors of arcs indicate significance of changes (blue to red, −log₁₀(p value)) with + or − marking direction of change.

(C) Heatmaps showing the changes of E-P contacts at the KMT2C/D-dependent de novo enhancers (N=882) and the KMT2C/D-independent de novo enhancers (N=111) upon cell differentiation from ESCs towards NPCs in WT (left column) and KMT2C/D dCD cells (right column). The pseudo color indicates significance of change (−log₁₀(p value)) with + or − marking direction of change. Boxplots show the fold changes of E-P contacts in WT and KMT2C/D dCD cells. All boxplots hereafter are defined as following: Central bar, median; lower and upper box limits, 25th and 75th percentiles, respectively; whiskers, minimum and maximum value within the range of (1st quartile −1.5*(3rd quartile - 1st quartile)) to (3rd quartile + 1.5*(3rd quartile - 1st quartile)). *** p value < 0.001, two-tailed t-test.

(D) Scatter plots showing changes of chromatin contacts anchored on promoters and enhancers of NPC-induced genes (N=1303) in WT cells and KMT2C/D dCD cells.

(E) Heatmaps showing the changes of E-P contacts centered at the 1303 NPC-induced genes in WT (left column) and KMT2C/D dCD cells (right column). See also Figures S3–S6, and STAR Methods.

Figure 3 |. Transcriptional defects upon loss of KMT2C/D catalytic activities in NPC cells.

(A) Microscopic images of mouse ESCs differentiation towards NPCs (at day 2.5, and day 5) in WT and KMT2C/D dCD cells. Alkaline phosphatase staining was performed at each time point to monitor their loss of pluripotency during cell differentiation.

(B) Principal component analysis of gene expression profiles of WT and KMT2C/D dCD cells at specified time points of cell differentiation. Two replicates of each cell line and treatment condition are shown.

(C) Scatter plots showing transcription levels of NPC-differentiation induced genes (FDR < 0.05, FC > 2, RPKM in NPCs > 1.0) in NPCs (x-axis) of WT and KMT2C/D dCD cells (day 5) (y-axis). Blue and light-blue dots mark down-regulated genes in KMT2C/D dCD NPCs (FC > 2 and 2 > FC > 1.5, respectively, FDR < 0.05). Red and orange dots mark up-regulated gene in KMT2C/D dCD NPCs (FC > 2 and 2> FC > 1.5, respectively, FDR < 0.05).

(D) Top three enriched gene ontology (GO) terms in genes that failed to be induced in KMT2C/D cells. p values (Fisher’s exact test) are also indicated.

(E) Volcano plots showing transcriptional changes during NPC differentiation in WT ESCs at genes that have significant chromatin contacts (MAPS, FDR < 0.01) with the KMT2C/D-dependent candidate enhancers (N=3028). Significantly up-regulated and down-regulated genes are marked as red and blue dots, respectively (FDR < 0.05, FC > 2).

(F) Scatter plots showing transcription levels of genes in NPCs of WT and KMT2C/D dCD cells. Only genes that are induced upon NPC differentiation and making significant chromatin contacts with KMT2C/D-dependent candidate enhancers are shown. Down-regulated genes in KMT2C/D dCD NPCs (2> FC > 1.5, FDR < 0.05) are marked as light-blue dots.

(G) Schematic representation of the method used to calculate the ratio of chromatin contact counts mapped at KMT2C/D-dependent or independent candidate enhancers. See STAR Methods for details of the calculation.

(H) Boxplots showing transcriptional changes between WT and KMT2C/D dCD cells during NPC differentiation. NPC-differentiation induced genes were classified into four groups based on the ratios of chromatin contact counts on the KMT2C/D-independent enhancers (ratio of C.C.). * p value < 0.05, *** p value < 0.001, one-tailed t-test.

(I) Boxplots showing the number of KMT2C/D-independent candidate enhancers making significant contacts (MAPS, FDR < 0.01) with each group of NPC-differentiation induced genes classified based on differential gene expression as in Figure 3C. ns p value > 0.05, *** p value < 0.001, one-tailed t-test.

(J) Schematic representation depicting the requirement of KMT2C/D catalytic activities at different groups of candidate enhancers. See also Figure S7–S9.

Meanwhile, we also observed loss of a large number of promoter-promoter (P-P) contacts in KMT2C/D dCD cells (1,233 contacts in NPCs, FDR < 0.05) (Figure 2A). Unlike the E-P contacts, the changes in P-P contacts upon loss of KMT2C/D catalytic activity showed no correlation with the alterations in H3K4me1 signals (Figures S5A and S5B). Interestingly, they were correlated with the changes of E-P contacts anchored at the same promoters (Figures S5C and S5D). The loss of these E-P and P-P contacts is likely not a result of changes in Pol II occupancy at the promoters (Figures S5E), not higher-order chromatin organization such as topologically associated domains (Figures S6). Taken together, loss of KMT2C/D catalytic activity resulted in diminished H3K4me1 at thousands of candidate enhancers induced during ESC differentiation to NPCs, and disruption of E-P contacts formed at these candidate enhancers.

Loss of KMT2C/D Catalytic Activities Delays NPC Differentiation and Activation of Hundreds of Genes

We next investigated the impact of the loss of KMT2C/D catalytic activity on gene activation during neural differentiation. The KMT2C/D dCD cells exhibited a delay in formation of neuronal projections and remained in ESC-like round shape colonies after 5 days of the neural induction (Figure 3A). Consistent with this observation, the overall gene expression profiles at each time point also suggested a delay of transcriptional transition in KMT2C/D dCD cells (Figure 3B). While the up-regulation of most NPC marker genes such as Pax6, Sox3, Map2, and the down-regulation of pluripotent markers such as Pou5f1, Sox2 were not interrupted in KMT2C/D dCD cells, induction of other NPC marker genes such as Tuj1, NeuN, and Olig2 was significantly delayed (Figure S7). As a matter of fact, of the 1,303 genes activated during the NPC differentiation (FDR < 0.05, FC > 2, RPKM in NPCs > 1), 17.6% or 228 genes (FC > 2, FDR < 0.05) failed to be fully induced in KMT2C/D dCD cells (Figures 3C and S8; Table S4). Genes related to organism development and cell differentiation (e.g. Sox11, Hoxb9, Lhx1) were highly enriched in these genes, suggesting that the loss of KMT2C/D catalytic activity impairs NPC differentiation (Figure 3D).

We focused on the induced genes that also displayed dynamic chromatin interactions between promoters and KMT2C/D-dependent NPC enhancers. In general, genes interacting with these candidate enhancers tended to be up-regulated upon cell differentiation in WT cells (Figure 3E). However, over 60% of them continued to be induced in KMT2C/D dCD cells during NPC differentiation despite the reduction of H3K4me1 at distal elements (Figures 3F and S8E). Why do some genes depend on KMT2C/D catalytic activity while others don’t? We hypothesized that genes could be activated by both KMT2C/D-dependent and -independent enhancers during cell differentiation^50,54, and the relative fraction of KMT2C/D-dependent and -independent enhancers that contact promoters may determine their dependence on the KMT2C/D catalytic activity. To test this hypothesis, we analyzed the chromatin contact counts on two groups of candidate enhancers classified based on the dependence of H3K4me1 on KMT2C/D catalytic activity as defined above (Figures 1B, 1C, and 3G). For each gene, we calculated the ratio of total contact counts on the KMT2C/D-independent enhancers to that on all candidate enhancers. Supporting our hypothesis, genes with relatively higher levels of input from KMT2C/D-independent enhancers tended to be unaffected by the loss of KMT2C/D catalytic activity than genes making more contacts with KMT2C/D-dependent candidate enhancers (Figures 3H and S9A). The KMT2C/D-independent genes (692 genes, defined in Figure 3C) were more likely to interact with KMT2C/D-independent candidate enhancers than the KMT2C/D-dependent genes (228 down-regulated and 98 up-regulated genes, defined in Figure 3C) (Figures 3I and S9B). Additionally, these robustly regulated genes were generally located at enhancer dense regions (7 enhancers or more around TSS ± 200 kb) and had relatively short chromatin contacts with distal elements (< 40 kb, PLAC-seq peak signal p-value < 0.01) (Figures S9C–S9F). These findings suggest that the chromatin contacts with multiple KMT2C/D-independent enhancers could sustain gene activation in the majority of genes during neural differentiation in the absence of KMT2C/D catalytic activity (Figure 3J).

KMT2B Contributes to H3K4me1 at KMT2C/D-independent Enhancers

Lastly, we explored KMT2C/D-independent mechanisms of H3K4me1 at distal enhancers by focusing on under-appreciated function of other methyltransferases. Although KMT2B is a major methyltransferase known to catalyze H3K4me3 at promoters⁵⁵, it was also broadly enriched at KMT2C/D-independent H3K4me1 distal peak regions (distance to TSS > 10 kb) (Figures S2B and S10A). Therefore, we hypothesized that KMT2B might also contribute to H3K4me1 depositions at enhancers. We then utilized an inducible degron system (dTAG) to acutely deplete KMT2B protein in WT and KMT2C/D dCD ESC to examine the impact of KMT2B loss on transcription and H3K4 methylation during NPC differentiation⁵⁶ (Figures 4A and S10B). Consistent with a previous report⁵⁷, severe reduction of H3K4me3 level at promoters had little effect on transcriptional induction at most genes during NPC differentiation (Figures 4B and 4C). Nevertheless, the impact of KMT2B loss was more severe than that of KMT2C/D catalytic activity loss on the gene induction, and the majority of KMT2B-dependent down-regulated genes (N=368) were not KMT2C/D-dependent genes (N=126) (Figures 4C and 4D). Furthermore, the KMT2C/D-independent distal candidate enhancers showed a greater decrease in H3K4me1 compared to KMT2C/D-dependent regions upon KMT2B loss in both ESCs and NPCs (Figures 4E, S10C, and S10D). Moreover, genes making more contacts with the 2,765 KMT2B-dependent H3K4me1 regions at candidate enhancers tended to be affected upon KMT2B loss (Figures 4F, S10E–S10G). These findings suggest that H3K4me1 at KMT2C/D-independent enhancers were partially mediated by KMT2B and the KMT2B-mediated H3K4me1 at these sites also affected gene activation during NPC differentiation.

Figure 4 |. KMT2B contributes to H3K4me1 at KMT2C/D-independent enhancers and transcription.

(A) Schematic representation of the dTAG system utilized to acutely deplete KMT2B in wild type and KMT2C/D dCD ESCs.

(B) Average H3K4me3 signals on TSSs (N=14090) in WT ESCs and NPCs. The ChIP-seq signals of WT (light brown), KMT2B-depleted (red), KMT2C/D dCD (light blue), and KMT2C/D dCD + KMT2B-depleted (purple) ESCs and NPCs are shown.

(C) Scatter plots showing transcription levels of NPC-differentiation induced genes (FDR < 0.05, FC > 2, RPKM in NPCs > 1.0) in WT NPCs (x-axis) and each indicated mutated NPCs (left: KMT2B-depleted NPCs, middle: KMT2C/D-dCD NPCs, right: KMT2C/D-dCD + KMT2B-depleted NPCs) (y-axis). Blue and light-blue dots mark down-regulated in mutated NPCs (FC > 2 and 2> FC > 1.5, respectively, FDR < 0.05). Red and orange dots mark up-regulated in mutated NPCs (FC > 2 and 2>FC > 1.5, respectively, FDR < 0.05). Three replicates of each sample were analyzed to determine the differentially expressed genes.

(D) Venn-diagram showing the overlap of down-regulated genes in KMT2C/D dCD (KMT2B(+)) NPCs (N=126) and KMT2B-depleted (KMT2C/D^WT) NPCs (N=368).

(E) Heatmaps showing H3K4me1 signals centered at KMNT2C/D-independent and -dependent candidate enhancers. KMT2C/D-independent enhancers: N=26,254 and 26,599 in ESCs and NPCs, respectively. KMT2C/D-dependent enhancers: N=4,150 and 7,178 in ESCs and NPCs, respectively. Average enrichments of the H3K4me1 signals are also shown on the right with same color scheme as in panel B.

(F) (left) Heatmaps showing H3K4me1 signals at candidate distal enhancers exhibiting significant reduction in KMT2B-depleted NPCs (N=2,765). (center and right) Boxplots showing transcriptional changes between WT and KMT2B-depleted (KMT2C/D^WT) NPCs at NPC-differentiation induced genes, classified based on the ratio of C.C. on the 2,765 KMT2B-dependent enhancers. * p value < 0.05, *** p value < 0.001, one-tailed t-test.

See also Figure S10.

DISCUSSION

In summary, consistent with our previous reports^34,58, we observed a severe disruption of newly formed E-P contacts at the newly induced candidate enhancers in KMT2C/D dCD NPCs, accompanied by a delay in NPC differentiation and activation of hundreds of lineage specific genes. Surprisingly and contrary to a long-held view that KMT2C/D are the major HMTs of H3K4me1 in mammalian cells, H3K4me1 at over 75% of candidate distal enhancers in NPCs was independent of KMT2C/D catalytic activity. We instead found that KMT2B provides redundant functions for H3K4me1 deposition at these KMT2C/D-independent enhancers. Future studies are required to assess whether this finding can be generalized to other cell lineages, and to reveal the mechanisms of redundant H3K4me1 deposition in combination with more different types of histone methyltransferases^20,59. Additionally, further exploration of TFs such as GATA family members as shown in this study would help to unveil the cell type specific role of KMT2C/D⁴⁶.

A recent study also focused on the role of KMT2C/D catalytic activity in mouse embryonic development and ESC differentiation, and reported their altered transcription in a few hundred NPC-induced genes³³. The population and the number of NPC genes affected by loss of KMT2C/D catalytic activity were not the same between the two studies. This might be due to the different culturing conditions and durations for NPC differentiation. Nevertheless, the two histone modifications at the newly formed enhancers were reduced upon loss of KMT2C/D catalytic activity in both studies. Our study further provides an explanation of these KMT2C/D-dependent and -independent gene activations by focusing on the E-P contact dynamics. Previously it was noted that most E-P contacts are formed independent of CTCF/Cohesin complex^50,60. On the other hand, we showed that Cohesin enrichment was indeed altered in correlation with the H3K4me1 signal changes in dCD cells (Figure S6C)³⁴.

Limitations of the study

A major limitation of our study is that we did not address the biochemical mechanisms by which H3K4me1 facilitates E-P contacts. In addition, it is important to address differences in the biological roles of KMT2C/D- and KMT2B-mediated H3K4me1. Further analysis of different lineages and in vivo models will provide new insights into the selection of these enhancers. It should be noted that our analysis of E-P contacts does not include the information of chromatin contacts within 10-kb genomic distance due to a limitation of resolution in the current approach. Our analysis is also not based on acute depletion of KMT2C/D catalytic activity, and some genes and non-histone substrates might be affected by secondary effects during the long-term culturing.

STAR★Methods

RESOURCE AVAILABILITY

Lead contact

Further information and requests for reagents may be directed to and will be fulfilled by the corresponding author Bing Ren (biren@health.ucsd.edu).

Materials availability

All resources used for this study are available upon request to the lead contact.

Data and code availability

• All datasets generated in this study have been deposited to Gene Expression Omnibus (GEO), with accession number GSE160892. Accession numbers of the downloaded data are listed in the Key Resources Table. Original western blot images are deposited in Mendeley data (http://doi.org/10.17632/nfj942xfpj.1).

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit polyclonal anti-H3K4me1	Abcam	ab8895
Rabbit monoclonal anti-H3K4me3	EMD Millipore	04-745
Rabbit polyclonal anti-H3K27ac	Active Motif	39133
Mouse monoclonal anti-H3K27me3	Active Motif	61017
Critical Commercial Assays
TruSeq Stranded mRNA Library Prep Kit	Illumina	RS-122-2101
ThruPLEX DNA-seq 12s kit	Rubicon Genomics	R400428
Accel-NGS 2S Plus DNA Library Kit	Swift Biosciences	21024
Deposited Data
Raw and processed sequencing data	This paper	GSE160892
Wild-type mouse ESCs ATAC-seq	Xu et al., 201741	GSE84646
Wild-type mouse NPCs ATAC-seq	Duren et al., 201742	GSE98479
KMT2C/D and Pol II ChIP-seq in mouse ESCs	Dorighi et al., 201730	GSE98063
GATA6 and 4 ChIP-seq in mouse embryonic fibroblasts	Shu et al., 201561	GSE57849
NANOG ChIP-seq in mouse ESCs	Miller et al., 201662	ERP014597
SOX2 ChIP-seq in mouse ESCs	Festuccia et al., 201963	GSE122589
TEAD4 ChIP-seq in mESC derived hemogenic endothelium	Obier et al., 201664	GSE79320
TEAD4 ChIP-seq in mouse trophoblast stem cell	Home et al., 201265	GSE37350
TEAD1 ChIP-seq in mouse nerve tumor	Wu et al., 201866	GSE99040
SOX3 ChIP-seq in mouse NPCs	Bergsland et al., 201267	GSE33059
KMT2B ChIP-seq in mouse ESCs	Sze et al., 202068	GSE152595
RAD21 ChIP-seq in mouse WT and KMT2C/D dCD ESCs	Yan et al., 201834	GSE74055
Uncropped western blot images	This paper	Mendeley dataset: http://doi.org/10.17632/nfj942xfpj.1
Experimental Models: Cell Lines
WT R1 mESC	ATCC	Cat#SCRC-1036
KMT2C/D (MLL3/4) dCD mESC	Dorighi et al., 2017	N/A
Software and Algorithms
R	R Core Team, 2020	http://www.R-project.org/
Perl	Perl.org	https://www.perl.org
edgeR	Robinson et al., 2012	https://bioconductor.org/packages/release/bioc/html/edgeR.html
MACS	Zhang et al., 2008	http://liulab.dfci.harvard.edu/MACS/00README.html
deepTools	Ramirez et al., 2016	https://github.com/fidelram/deepTools
DESeq2	Love et al., 2014	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
MAPS	Juric et al., 2019	https://github.com/ijuric/MAPS
Homer	Heinz et al., 2010	http://homer.ucsd.edu/homer/motif/
ENCODE ATAC-seq pipeline	ENCODE DCC 2017	https://github.com/ENCODE-DCC/atac-seq-pipeline
STAR	Dobin et al., 2013	https://github.com/alexdobin/STAR
bwa	Li et al., 2013	https://github.com/lh3/bwa
WashU Epigenome Browser	Zhou et al., 2013	http://epigenomegateway.wustl.edu/
igv	Robinson et al., 2011	http://software.broadinstitute.org/software/igv/
bedtools	Quinlan, 2014	http://bedtools.readthedocs.io/en/latest/
pgltools	Greenwald et al. 2017	https://github.com/billgreenwald/pgltools

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Cell lines

Mouse R1 ES cell line was used for KMT2C/D (MLL3/4) catalytically deficient ES cell line, which was reported in the previous study³⁰. ESCs were cultured in KnockOut Serum Replacement containing mouse ES cell media: DMEM 85%, 15% KnockOut Serum Replacement (Gibco), penicillin/streptomycin (Gibco), 1× non-essential amino acids (Gibco), 1× GlutaMax (Gibco), 1000 U/ml LIF (Millipore), 0.4 mM β-mercaptoethanol. The cells were grown on 0.2% gelatin-coated plates with irradiated mouse embryonic fibroblasts (MEFs) (GlobalStem). Cells were maintained by passaging using Accutase (Innovative Cell Technologies) at 37°C and 5% CO2. Medium was changed daily when cells were not passaged. Cells were checked for mycoplasma infection and tested negative.

dTAG knock-in vector construction

The DNA fragments containing 5` homology arm, 3` homology arm, and FKBP12^F36V-3XFLAG was designed to include overlapping sequences to aid in Gibson assembly and synthesized at IDT as gBlock gene fragments (Table S5). The cassette was assembled into the pAW62.YY1.FKBP.knock-in.mCherry backbone (Addgene Cat#104370) using the Gibson Assembly Master Mix (NEB) according to the manufacturer’s instructions.

For the generation of the endogenously tagged lines, 2×10⁶ mESCs were electroporated (Lonza) with 5 ul of the ribonucleoprotein (RNP) complex containing fluorescently labeled Alt-R^™ CRISPR-Cas9 tracrRNA (IDT), 10 ug of MLL2-dTAG-knockin plasmid. Fluorescence positive cells were sorted 48 hours after electroporation. Cells were expanded for 5 days and then split into three 96-well plates by limited dilution (0.5 cell/per well). The cells were grown for approximately one week in 2i medium and genotyped by PCR (Table S5). Clones with a homozygous knock-in tag were further expanded and used for experiments.

METHOD DETAILS

Neural progenitor cell differentiation

NPC differentiation was conducted utilizing retinoic acid (RA) induction^39,40. ESCs were grown on MEFs and passaged on 0.2% gelatin-coated plates without MEFs one day before starting differentiation treatment. On day 0, LIF was deprived from the culture medium. From day 1, 5 uM RA (Sigma, R2625) was added with the LIF-deprived medium. Cells were harvested on day 2.5 and day 5. Alkaline phosphatase staining was performed at each time point using the AP Staining kit II (Stemgent, 00–0055). For the dTAG treatment, 2uM of dTAG-13 (Tocris Bioscience) was added from day 1 of NPC differentiation and cells were harvested on day 4.

ChIP-seq library preparation

ChIP-seq experiments for each histone mark were performed as described in ENCODE experiment protocols (“Ren Lab ENCODE Chromatin Immunoprecipitation Protocol” in https://www.encodeproject.org/documents/) with minor modifications. Cells were crosslinked with 1% formaldehyde for 10 minutes. We used 1.0 million cells for each ChIP sample. Shearing of chromatin was performed using truChIP Chromatin Shearing Reagent Kit (Covaris) according to the manufacturer’s instructions. Covaris M220 was used for sonication with following parameters: 10 minutes duration at 10.0% duty factor, 75.0 peak power, 200 cycles per burst at 5–9°C temperature range. For immunoprecipitation, we used 11 μL anti-rabbit or anti-mouse IgG Dynabeads (Life Technologies) and wash them with cold BSA/PBS (0.5 mg / mL bovine serum albumin in 1x phosphate buffered saline) for 3 times. After washing, 3 μg antibody with 147 μL cold BSA/PBS were added to the beads and incubated over 2 hours at 4°C. After incubation, beads were washed with150 μL cold BSA/PBS for 3 times and mixed with 100 μL Binding Buffer (1% Triton X-100, 0.1% Sodium Deoxycholate, 1x complete protease inhibitor (Roche)) plus 100 μL 0.2 μg/μl chromatin followed by overnight incubation on a rotating platform at 4°C. Beads were washed 5 times with 50 mM Hepes pH 8.0, 1% NP-40, 1 mM EDTA, 0.70% Sodium Deoxycholate, 0.5 M LiCl, 1x complete protease inhibitor (Roche) and washed once with 150 μL cold 1x TE followed by incubation at 65°C for 20 minutes in 150 μL ChIP elution buffer (10 mM Tris-HCl pH 8.0, 1 mM EDTA, 1% SDS). The beads were removed and the samples were incubated at 65°C overnight to reverse crosslinks. The input samples were also processed in parallel with the ChIP samples. Samples were incubated with RNase A (final conc. = 0.2 mg/mL) at 37°C for 1 hour, and Proteinase K (final conc. = 0.4 mg/mL) at 55°C for 1 hour. The samples were extracted with phenol: chloroform: isoamyl alcohol (25:24:1) and precipitated with ethanol. We used 3–5 ng of starting IP materials for preparing Illumina sequencing libraries. The End-It DNA End-Repair Kit (Epicentre) was used to repair DNA fragments to blunt ends. A-tailing 3’ end was performed using Klenow Fragment (3’→5’ exo-) (New England Biolabs), and then TruSeq Adapters were ligated by Quick T4 DNA Ligase (New England Biolabs). Size selection using AMpure Beads (Beckman Coulter) was performed to get 300–500bp DNA and PCR amplification (8–10 cycles) was performed. Libraries were sequenced on HiSeq4000 single end for 50 bp. Two biological replicates were prepared for each sample.

RNA-seq library preparation

Total RNA was extracted using the AllPrep Mini kit (QIAGEN) according to the manufacturer’s instructions and 1 μg of total RNA was used to prepare each RNA-seq library. The libraries were prepared using TruSeq Stranded mRNA Library Prep Kit (Illumina). Libraries were sequenced on HiSeq4000 and NextSeq500 using 50 bp paired-end. Two or three biological replicates were prepared for each sample.

PLAC-seq library preparation

PLAC-seq experiments were performed as previously described⁴⁸. Cells were crosslinked with 1% formaldehyde (w/v, methanol-free, ThermoFisher) for 15 minutes. The crosslinked pellets (2.5–3 million cells per sample) were incubated with 300ul of lysis buffer (10 mM Tris-HCl pH 8.0, 10 mM NaCl, 0.2% Igepal CA630, 33 μL, 1x complete protease inhibitor (Roche)) on ice for 15 min, washed with 500 μL cold lysis buffer, and then incubated in 50uL of 0.5% SDS for 10 min at 62°C. After heating, 160 μL of 1.56% Triton X-100 was added and incubated for 15min at 37°C. To digest chromatin 100U MboI and 25uL of 10X NEBuffer2 were added followed by 2 hours incubation at 37°C with agitation at 900rpm. MboI was inactivated by heating at 62°C. Digestion efficiency was confirmed by performing agarose gel electrophoresis of the samples. The digested ends were labeled with biotin by adding 37.5uL of 0.4mM biotin-14-dATP (Life Tech), 1.5 μL of 10mM dCTP, 10mM dTTP, 10mM dGTP, and 8uL of 5U/ul Klenow (New England Biolabs) and incubating at 37°C for 1 hour with shaking at 900 rpm. Then the samples were mixed with 1x T4 DNA ligase buffer (New England Biolabs), 0.83% Trition X-100, 0.1 mg/mL BSA, 2000U T4 DNA Ligase (New England Biolabs, M0202), and incubated at room temperature for 2 hours with shaking with slow rotation. The ligated cell pellets were resuspended in 125 ul of RIPA buffer with protease inhibitor and incubated on ice for 10 minutes. The cell lysates were sonicated using Covaris M220. After spinning, we saved 20 ul supernatant as input, and for the rest part, 100 ul of antibody-coupled beads were added to the supernatant sample, and then rotated in cold room at least 12 hours. For immunoprecipitation, 300 ul of M280 sheep anti-rabbit IgG beads (ThermoFisher) was washed with cold BSA/PBS (0.5 mg / mL bovine serum albumin in 1x phosphate buffered saline) for 4 times. After washing, 30 ug anti-H3K4me3 (Millipore, 04–745) with 1 mL cold BSA/PBS were added to the beads and incubated on a rotating platform at 4°C for over 3 hours. After incubation, beads were washed with cold BSA/PBS and resuspended in 600 ul RIPA buffer. The beads were washed with RIPA buffer (3 times), RIPA buffer + 0.16M NaCl (2 times), LiCl buffer (1 time), and TE buffer (2 times) at 4°C for 3 minutes at 1000 rpm. For reverse crosslinking, 163 ul extraction buffer (135 ul 1xTE, 15 ul 10% SDS, 12 ul 5M NaCl, 1 ul RnaseA (10mg/ml)) was added and incubated at 37°C for 1 hour at 1000 rpm, and 20 ug of proteinase K was added and incubated at 65°C for 2 hours at 1000rpm. After crosslinking, DNA was purified using Zymo DNA clean & concentrator and eluted with 50 ul of 10mM Tris (pH 8.0). For biotin enrichment, 25 ul of T1 Streptavidin Beads (Invitrogen) per sample were washed with 400 ul Tween wash buffer (5 mM Tris-HCl pH 8.0, 0.5 mM EDTA, 1 M NaCl, 0.05% Tween-20), and resuspended in 50 ul of 2x Binding buffer (10 mM Tris-HCl pH 7.5, 1 mM EDTA, 2 M NaCl). The purified 50 ul DNA sample was added to the 50 ul resuspended beads and incubated at room temperature for 15 minutes with rotation. The beads were washed with 500 ul of Tween wash buffer twice and washed with 100 ul Low EDTA TE (supplied by Swift Biosciences kit). Then beads were resuspended in 40 ul Low EDTA TE. Next, we used Swift Biosciences kit (Cat. No. 21024) for library construction with modified protocol. The Repair I Reaction Mix was added to 40 ul sample beads and incubated at 37°C for 10 minutes at 800 rpm. The beads were washed with 500 ul Tween wash buffer twice and washed with 100 ul Low EDTA TE once. The Repair II Reaction Mix was added to the beads followed by incubation at 20°C for 20 minutes at 800 rpm. The beads were washed in the same way with 500 ul Tween wash buffer and 100 ul Low EDTA TE. Then, 25 ul of the Ligation I Reaction Mix and Reagent Y2 was added to the beads followed by incubation at 25°C for 15 minutes at 800 rpm. The beads were washed with Tween wash buffer and Low EDTA TE. Then 50 ul of the Ligation II Reaction Mix was added to the beads followed by incubation at 40°C for 10 minutes at 800 rpm. The beads were washed and resuspended in 21 ul 10mM Tris-HCl (pH 8.0). The amplification and purification were performed according to the Swift library kit protocols. Libraries were sequenced on Illumina HiSeq 4000. Two biological replicates were prepared for each sample.

QUANTIFICATION AND STATISTICAL ANALYSIS

ChIP-seq data analysis

Each fastq file was mapped to mouse genome (mm10) with BWA⁶⁹ -aln with “-q 5 -l 32 -k 2” options. PCR duplicates were removed using Picard MarkDuplicates (https://github.com/broadinstitute/picard) and the bigWig files were created using deepTools⁷⁰ with following parameters: bamCompare --binSize 10 --normalizeUsing RPKM --operation subtract (or log2). The deepTools was also used for generating heatmaps. Peaks were called with input control using MACS2⁷¹ with broad peak calling. The candidate active enhancer regions were characterized by the presence of both H3K4me1 and H3K27ac peaks, but not H3K4me3 peaks. DEseq2⁷² was used for differential peak analysis. We defined “De novo enhancers in NPC” as the enhancer regions whose H3K4me1 signal was significantly increased from ESCs to NPCs (FDR < 0.05, log₂ FC > 0.5). The same differential peak analysis was performed between WT and KMT2C/D dCD cells to determine the KMT2C/D-dependent and -independent enhancers.

Motif analysis

Enrichment analysis of known DNA binding motifs was performed using HOMER tool⁷³. Default parameters with a fragment size of 1000 bp and “-mask” parameter were used. In the differential motif analysis, regions from the compared KMT2C/D-dependent or -independent enhancers were used as background by adding “-bg” parameter.

RNA-seq data analysis

RNA-seq reads (paired-end) were mapped to the mm10 genome using STAR⁷⁴. The mapped reads were counted using HTSeq⁷⁵ and the output files from two replicates (Figure 3) or three replicates (Figure 4) were subsequently analyzed by edgeR⁷⁶ to detect the differentially expressed genes (FDR < 0.05, FC > 2 or FC > 1.5). RPKM was calculated using an in-house pipeline.

ATAC-seq data analysis

ATAC-seq reads (paired-end) were mapped to the mm10 genome and processed using ENCODE ATAC-seq pipeline (https://github.com/ENCODE-DCC/atac-seq-pipeline). The deepTools⁷⁰ was used to generate bigwig files and heatmaps as described above.

PLAC-seq data analysis

PLAC-seq reads (paired-end) were aligned against the mm10 genome using BWA -mem⁶⁹. PCR duplicate reads were removed using Picard MarkDuplicates. Filtered reads were binned at 10 kb size to generate the contact matrix. Individual bins that were overlapped with H3K4me3 peaks on transcription start sites (TSSs) were used for downstream differential contact analysis. MAPS⁷⁷ was used for peak calling with default settings in 10 kb resolution. For the differential contact analysis⁵⁰, the raw contact counts in 10 kb resolution bins that have the same genomic distance were used as inputs. To minimize the bias from genomic distance, we stratified the inputs into every 10-kb genomic distance from 10 kb to 150 kb, and the rest of the input bins with longer distances were stratified to have a uniform size of input bins that were equal to that of 140–150 kb distance bins. Since these inputs showed negative binomial distribution, edgeR⁷⁶ was used to get the initial set of differential interactions. Only bins that had more than 20 contact counts in each sample of two replicates were used for the downstream analysis. The significances of these differential interactions are either due to the difference in their H3K4me3 ChIP coverage or 3D contacts coverage. Therefore, the chromatin contacts from promoter regions with differential ChIP-seq peaks between the samples (p value < 0.01) were removed and only the chromatin contacts with the same level of H3K4me3 ChIP-seq peaks were processed. We used all bins for inputs that included non-significant interactions that were not identified by MAPS peak caller because the majority of short-range interactions were not identified as significant peaks due to their high background and the changes in these short-range interactions might be also important for gene regulation. We identified a large number of differentially changed short-range interactions even though many of them were not identified by peak caller, and we observed that such differentially changed interactions were positively correlated with the changes of H3K4me1/H3K27ac levels on their anchor sites during neural differentiation (Figure S3C), suggesting these interaction changes might reflect the biological changes. We used significance level with change direction (− log₁₀(p value))*(direction(+ or −)) instead of fold change to show the changes of chromatin contacts because fold change tends to be marginal when it is short-range interaction although their changes are actually significant and biologically meaningful. In Figure 2C and 2E, we also showed fold change. To visualize the chromatin contacts, we used WashU Epigenome Browser⁷⁸.

In Figure 3G, the ratio of chromatin contacts on KMT2C/D-independent enhancers was calculated by simply summing raw contact counts between promoters and all KMT2C/D-independent enhancers and dividing by total contact counts between promoters and all candidate enhancers in each gene in WT NPCs. The contact counts within 10-kb distance (1 bin) were not added. Genes that had less than 50 of the total contacts counts were removed for downstream analysis.

Odds ratio calculation for KMT2C/D-dependent E-P contacts enrichment

For Figure S9C, all genes were classified based on the distance to the nearest interacting enhancer and the number of enhancers around TSS (< 200 kb) (categorized into 3×3 bins). The distance to the nearest interacting enhancer is represented by the shortest genomic distance of significant PLAC-seq peaks on enhancers and promoters (p value < 0.01). Then, we generated 2×2 tables based on whether they are stably-regulated genes or not (FDR < 0.05) and whether they were categorized into the bin or not. Odds ratios and p values on each 2×2 tables were calculated. In Figure S9D and S9E, the same analysis as panel (C) was performed in differentially down-regulated genes and differentially up-regulated genes.

Supplementary Material

2

Table S1. List of next-generation sequencing sample information, related to Figures 1–4.

3

Table S2. List of KMT2C/D catalytic activity-dependent and -independent candidate enhancers and KMT2B-dependent and -independent candidate enhancers, related to Figures 1 and 4.

4

Table S3. List of differentially changed enhancer-promoter contacts, related to Figure 2.

5

Table S4. Gene expression changes during neural differentiation in WT and KMT2C/D dCD cells, related to Figure 3.

6

Table S5. DNA fragments and primers for dTAG knock-in to deplete KMT2B, related to Figure 4.

Highlights.

• H3K4me1 promotes enhancer-promoter contacts in neural progenitor cell differentiation.

• H3K4me1 can occur at candidate enhancers due to KMT2C/D-independent mechanisms.

• KMT2B contributes to H3K4me1 at KMT2C/D-independent candidate enhancers.

• KMT2B loss aggravates the transcriptional defects in KMT2C/D catalytic deficient cells.