Crystal Structure of the COMPASS H3K4 Methyltransferase Catalytic Module

Overall structure of the yeast COMPASS catalytic module

To reveal the complete architecture of the yeast COMPASS catalytic module, we co-expressed and purified the K. lactis Set1 catalytic domain together with the full-length or near full-length Swd3, Swd1, Bre2, and Sdc1 proteins from insect cells (Figure 1A). The yeast COMPASS catalytic module displayed strong distributive H3K4 methyltransferase activity toward a reconstituted nucleosome substrate (Patel et al., 2009) (Figure 1B). In the presence of SAM and an H3 peptide bearing a K4M mutation, the intact complex was crystallized in the C2 space group with one molecular assembly in the asymmetric unit (Table S1). The catalytic module measures a total of ~145 Å in its longest dimension, and ~70 Å in its shortest. In agreement with previously published low-resolution structural results (Shinsky and Cosgrove, 2015; Takahashi et al., 2011), the catalytic module adopts a Y-shaped overall structure, with the Bre2-Sdc1 sub-complex serving as the “base” and the two β-propeller proteins Swd1 and Swd3 forming the two “arms” (Figure 1C). The SET domain of Set1 is situated in the center of the complex, making direct contacts with all other subunits except Sdc1. With both the SAM cofactor and the H3 mutant peptide bound, the Set1 catalytic domain is trapped in a pre-reaction state with a compact conformation.

The COMPASS catalytic module is organized by the WD40 repeat protein, Swd1, which sports a long and winding C-terminal tail wrapping around three of the four neighboring subunits, Set1, Swd3, and Bre2 (Figure 1C and1D). Swd1 alone buries a total of ~3430 Ų surface area, consistent with its critical role in nucleating COMPASS in vivo. Swd3 is anchored adjacent to Swd1, with the “top” surface of its β-propeller domain facing away from Swd1 (Figure 1C). A loop projecting from the Swd3 β-propeller contacts Set1 and an adjacent Swd1 tail loop through a small, albeit strategic, interface near the active site of the enzyme. Opposite to the Swd1-Swd3 “arms”, Bre2 and Sdc1 packs against Set1 as a globular three polypeptide sub-complex with an asymmetric Sdc1 homo-dimer cuddling a long Bre2 C-terminal α-helix. The five subunits, Set1, Swd1, Swd3, Bre1 and Sdc1, therefore, assemble the COMPASS catalytic module subunits with a stoichiometry of 1:1:1:1:2.

Bre2 and ASH2L contain a non-canonical SPRY domain

Previous studies of an aggressively truncated human ASH2L central fragment revealed a canonical “SPRY-only” β-sandwich domain characterized by two tightly packed anti-parallel β-sheets (Chen et al., 2012; Zhang et al., 2015a). An RBBP5 C-terminal sequence-binding site has been subsequently mapped to a shallow surface pocket on one side of the β-sandwich fold (Cao et al., 2010; Zhang et al., 2015a) (Figure 2A). Although these properties of ASH2L are also found in Bre2 (Figure 2B), the structure of the intact yeast ortholog in the context of the entire COMPASS catalytic module unveils a far more structurally and functionally complex architecture that is conserved in the animal orthologs, but missing from previous studies.

The annotated “SPRY-only” domain of Bre2 is flanked by a ~90 amino acid N-terminal “pre-SPRY” region and a 20 amino acid C-terminal tail (Figure 1A). The central SPRY fold also features two long insertions loops, Ins-1 and Ins-2. Together, these sequences adopt a highly intertwined topology, which embraces and expands the SPRY-only β-sandwich into a much larger globular fold (Figure 2B). At the center of this SPRY domain extension is a twisted auxiliary β-sheet, which is constructed by two pairs of anti-parallel β-strands, one from the pre-SPRY region and the other from the Ins-1 loop (Figure 2B). Closer sequence analysis indicates that these key structural elements of the auxiliary β-sheet are conserved in the sequences of all yeast and animal Bre2/ASH2L orthologs (Figure S1). Intriguingly, next to the SPRY-only β-sandwich, the auxiliary β-sheet provides an extended platform, whose surface is decorated by a cluster of highly, if not strictly, conserved residues, including P46, N48, K49, F52, Y54, D248, I251, Y253, K254, and E260 (Figure 2C and S1). Their mixed hydrophobic and polar side chains and their close proximity to the H3 peptide binding pocket of the Set1 catalytic domain hint at a functional role in either binding the nucleosomal substrate or a COMPASS regulatory factor. The potential functional relevance of the auxiliary β-sheet is further underscored by the α-helix formed by the Bre2 C-terminal tail, which buttresses the extended platform from the bottom (Figure 2D).

Sdcl forms an asymmetric dimer to stabilize Bre2

Sdc1 is the smallest subunit of COMPASS and interacts exclusively with Bre2 at the distal end of the catalytic module. Previous yeast genetics and biochemical analysis have shown that loss of Sdc1 downregulates H3K4 methylation by Set1 (Dehé et al., 2006; Kim et al., 2013; Krogan et al., 2002; South et al., 2010). The complete spatial separation between Sdc1 and Set1 strongly suggests that Sdc1 positively regulates the COMPASS catalytic module by stabilizing the non-canonical Bre2 SPRY domain.

Sdc1 and its animal ortholog, DPY-30, share a highly conserved central sequence (Figure S2). In DPY-30, this region forms three short a-helices and binds to the isolated C-terminal α-helix of ASH2L as a homo-dimer (Tremblay et al., 2014; Wang et al., 2009b). In complex with the full-length Bre2 protein, Sdc1 clasps the Bre2 helical tail in the same binding mode, but also makes additional interfaces with the expanded Bre2 non-canonical SPRY domain (Figure 2D and 3A). Superposition analysis of the two Sdc1 protomers (designated Sdc1-A/B) reveals substantial conformational variation within their N-terminal a-helices (Figure 3B). This structural asymmetry is attributable to the distinct interfaces made by the two copies of Sdc1 with Bre2. For example, in one protomer (Sdc1-A), the tip of its N-terminal α-helix uses the strictly conserved residue Arg88 to lock the C-terminal α-helix of Bre2 onto its non-canonical SPRY domain with a salt bridge and a hydrogen bond (Figure 2D, and 3C). By contrast, the N-terminal region of the other Sdc1 copy curves underneath the Bre2 auxiliary β-sheet with its R88 residue pointing away from Bre2 (Figure 2D, 3C and 3D). By clamping the helical tail of Bre2 onto its non-canonical SPRY domain, the asymmetric Sdc1 dimer appears to reinforce the entire Bre2 protein fold. To validate such a structural role, we show that a 17 amino acid Sdc1 N-terminal region (Asp74 to Arg89), which plays a minor role in tethering the isolated ASH2L C-terminal α-helix to DPY-30 (Tremblay et al., 2014), is critical for the production of soluble Bre2 and the formation of a stable Bre2-Sdc1 complex when co-expressed in insect cells (Figure 3E).

Swd1 organizes the catalytic module and creates an inter-subunit pocket next to active site

As the largest subunit of the COMPASS catalytic module, Swd1 and its orthologs contain a WD40 repeat β-propeller domain, which is preceded by a short N-terminal extension and followed by a long C-terminal tail that varies in length between different species (Figure 1A and S3A). The Swd1 C-terminal tail plays a central role in nucleating the entire complex and can be divided into two segments: a highly conserved WD40 repeat proximal (WDRP) region organizing Set1, Swd3 and Bre2, and a more variable distal portion responsible for further securing Swd3 binding (Figure 4A). These two regions flank a short disordered loop invisible in the crystal.

The Swd1 WDRP segment emerges from the end of the β-propeller fold and spans roughly 25 amino acids (Figure 4A). This short sequence represents the most conserved region of the entire protein with 21 out of 25 residues either strictly identical or highly similar between yeast and animal orthologs (Figure S3A). Remarkably, the vast majority of the WDRP residues are directly involved in binding Set1, while about one third of these residues also participate in binding either Swd3 or Bre2 at the two ends. The N-terminal half of the WDRP sequence folds into a stable loop, which uses more than five invariant hydrophobic residues, including Trp356 and Phe363, to anchor at a hydrophobic surface covering the junction of the SET-N/C and SET-I subdomains, which sandwich substrate and cofactor in between (Figure 4B and 4C). The tip of this Swd1 WDRP loop joins the two Set1 subdomains and together encloses a well-defined pocket in close vicinity of the SAM-binding site (Figure 4C). Importantly, this Swd1-Set1 intersubunit pocket houses a Swd3 loop protruding from its WD40 repeat domain and is characterized by disease mutations as discussed in the following section. Beyond this Set1-Swd1-Swd3 interface, the Swd1 WDRP segment continues to meander its way around the SET domain and ends at the other tri-subunit interface, which has been previously revealed for the mammalian counterparts of Swd1, Set1, and Bre2 (Li et al., 2016) (Figure S3A and S4A). As a whole, the Swd1 WDRP sequence snakes around one side of the Set1 catalytic domain making multiple inter-subunit interactions. Despite its short length, the condensed interfaces made by the Swd1 WDRP region provides a structural explanation for its critical role in maintaining the integrity of Set1 and the entire COMPASS (Dehé et al., 2006; Halbach et al., 2009; Kim et al., 2013; Mersman et al., 2012).

Besides the WDRP region, Swd1 uses two additional “tentacle”-like structural elements to recruit Swd3 to the COMPASS catalytic module. The N-terminal extension of Swd1 folds into a short helix and latches onto the “bottom” surface of the Swd3 β-propeller domain next to the interface formed between Swd3 and the WDRP sequence (Figure 4A). The C-terminal distal tail of Swd1, meanwhile, utilizes a stretch of residues between Asp395 and Thr403 to dock to a hydrophobic cleft in between two blades of the Swd3 β-propeller, as previously revealed for the mammalian orthologs, WDR5 and RBBP5 (Figure S4B) (Avdic et al., 2011; Odho et al., 2010). These Swd1-Swd3 interfaces are further augmented by the extreme end of the Swd1 C-terminal distal tail, which funnels into a canal sandwiched between the WD40 domains of the two proteins and ends with a short α-helix that nestles in another inter-blade hydrophobic cleft on Swd3 (Figure 4D). In a co-purification experiment, removal of this last α-helix of Swd1, but not the N- terminal extension, resulted in a complete abrogation of Swd3 binding, indicating that the distal interface formed by this last helix is critical for the recruitment of Swd3 to the catalytic module (Figure 4E). Overall, by interacting with Swd3 through a multi-angled interface, Swd1 not only recruits Swd3, but also precisely positions the Swd3 loop protruding out of its β-propeller domain to the Swd1-Set1 inter-subunit pocket adjacent to the active site.

Swd3 regulates Set1 enzymatic activity via a SMART motif targeting the Kabuki pocket

The hallmark interactions of the Set1-centric COMPASS catalytic module lie at the tri-subunit interface, where the prominent Swd3 projecting loop inserts a tryptophan residue, Trp197, into the inter-subunit pocket formed between Set1 and Swd1 (Figure 4C and 5A). This Swd3 loop is found in all yeast orthologs (Figure S5) and is stabilized by a network of hydrogen bonds formed among backbone groups of the loop as well as hydrophobic side chain clustering (Figure 5A). Next to the Swd3 loop, the Swd1 WDRP loop provides continuous support by complementing the edge of the Swd3 propeller domain. Sitting snugly at the center of the Swd1-Set1 pocket, the indole ring of Swd3-Trp197 is sandwiched between Phe363 of Swd1 and Ile909 of Set1 (Figure 5B). It is further surrounded by a mixture of hydrophobic and polar residues highly conserved among all Swd1 and Set1 proteins from yeast to humans. Remarkably, the side chain of this Swd3 tryptophan residue is only about 5 Å away from the carboxyl end of the SAM cofactor. It also buries two Set1 residues, Tyr913 and Arg933, which are found in the majority of SET domain-containing enzymes with essential roles in catalysis (Zhang et al., 2002) (Figure 5B).

The structural configuration of this prominent tri-subunit interface strongly suggests that Swd3-Trp197 might play an important role in modulating the enzymatic activity of Set1 through the Set1-Swd1 pocket. In support of this notion, alanine mutation of Trp197 drastically compromised the catalytic activity of Set1 by slowing down the production of all methylation marks on nucleosomal substrates (Figure 5C). The functional importance of its targeting pocket is also manifested by mutations of two MLL4 /SETD1B residues found in both Kabuki syndrome and autism patients, which correspond to two yeast Set1 residues in direct contact with Swd3- Trp197: Arg933 and its adjacent Thr926 (Hiraide et al., 2018; Shinsky et al., 2014). To highlight their functional roles, we hereafter name the Swd3 loop sequence, DWIAE, as the Set1 Methyltransferase Activity RegulaTor (SMART) motif, and its targeting site as the “Kabuki” pocket.

The WD40 repeat domain of WDR5, the animal ortholog of Swd3, has been previously shown to possess a central H3 peptide-binding site, which is also utilized for engaging the WIN motif preceding the SET domains of all human SET1/MLL family members (Couture et al., 2006; Dharmarajan et al., 2012; Patel et al., 2008; Ruthenburg et al., 2006; Schuetz et al., 2006). In our structure, the equivalent site on Swd3 is more than 35 Å away from the active site of Set1 (Figure S4A), arguing against a cis-acting substrate adapter function of WDR5 if it adopts the same binding mode in the catalytic module. Interestingly, yeast Swd3 lacks a critical H3 peptide- binding phenylalanine residue (Ruthenburg et al., 2006) and does not display any detectable affinity toward the H3 peptide (Figure S5, data not shown), nor does yeast Set1 contains a WIN motif in front of its catalytic domain. Thus, the central peptide-binding site of the WDR5 β- propeller domain most likely has been co-evolved with the WIN motif in the expanded SET1/MLL family members in animals. Curiously, concurrent with the evolutionary gain of these functional elements, the animal WDR5 orthologs have lost the SMART motif found in Swd3 (Figure S5), which targets the Kabuki pocket and regulates the enzymatic activity of yeast Set1. Because the Swd1 and Set1 residues constructing the Kabuki pocket are highly conserved from yeast to humans, the inter-subunit pocket is expected to exist in all yeast and animal COMPASS and COMPASS-like complexes and potentially serves as a common regulatory site. Minimally, the catalytic module of the animal SETD1A and SETD1B COMPASS complexes is anticipated to contain this regulatory pocket, which could be targeted by a structural element similar or analogous to the yeast Swd3 SMART motif. In fact, we were able to identify a putative SMART motif, DWLND, immediately upstream of the WIN motif in the vertebrate SETD1A and SETD1B protein themselves (Figure 5D). To our surprise, when we compared the recombinant WIN-SET domain of SETD1B, which was used for reconstituting the enzymatic activity of its catalytic core, with a longer construct that includes this upstream putative SMART motif, the activity of the WRAD-bound human enzyme was substantially inhibited, suggesting an auto- inhibitory mechanism in contrast to the stimulating effect of the Swd3 SMART loop (Figure 5C). Although the exact mode of action of this putative SETD1B SMART motif awaits future investigation, this result highlights the susceptibility of the mammalian COMPASS catalytic module to activity modulation.

Structural determinant of substrate specificity among SET1/MLL family members

The substrate specificity of a SET domain-containing enzyme is defined by its intrinsic ability to mono- or multi-methylate the lysine substrate. Although this property is usually dictated by a “Phe/Tyr” switch (Couture et al., 2008), the SET1/MLL enzymes share an identical Tyr residue at the relevant position, which fails to explain the differences in their enzymatic product profiles (Figure S3B). With the first structure of a robust prototypical tri-methyltransferase in the SET1/MLL family, we set to identify the structural determinant of substrate specificity among different family members (Li et al., 2016; Southall et al., 2009; Zhang et al., 2015b).

Previous studies have suggested that the catalytic domain of SET1/MLL enzymes can adopt open and closed conformations, which involve a hinge motion between the SET-N/C and SET-I subdomains (Li et al., 2016; Southall et al., 2009) (Figure 6A). Superposition analysis indicates that the Set1 catalytic domain is trapped in the closed conformation, which has also been reported for the mono-methyltransferase, MLL3 (Figure 6A). Structural comparison of the two enzymes revealed a strictly conserved substrate-binding pocket with few significant conformational variations. A closer examination with sequence alignment analysis, nevertheless, unveiled a variable “door” loop, which demarcates the Kabuki pocket and the cofactor binding site right above the substrate peptide (Figure 6B). Among yeast Set1 and the six human SET1/MLL family members, this 4–5 amino acids loop features a common central glycine residue flanked by variable amino acids on each side, [GI/NR]G[V/I/C/SS] (Figure 6C). Strikingly, sequence variations in this “door” loop seem to correlate with the substrate specificity of the enzymes (Shinsky et al., 2015) (Figure S6).

To investigate the possible role of this structural element in substrate specificity, we first probed the importance of the second half of the “door” loop, which directly contacts the SAM cofactor in yeast Set1 and forms a channel for the target lysine to tunnel through. Substitution of the tandem serine residues with single residues found in MLL1–4 gradually reduced the tri- methyltransferase activity with bulkier hydrophobic side chains (Figure 6D, lane 1–4). Similarly, replacing the single valine residue in MLL3 with either cysteine or tandem serine had discernable effects on the product pattern, either slightly enhancing multi-methylation, or compromising mono-methylation (Figure 6D, lane 5–7). This portion of the loop, therefore, can affect, but not dramatically alter, substrate specificity of the enzymes. We next examined the role of the two N-terminal amino acids of the door loop, which belong to the Kabuki pocket. Mutation of Gly-Ile in Set1 to Asn-Arg rendered the yeast catalytic module completely insoluble. This result was not unexpected, because the hydrophobic isoleucine residue, Ile909, is a key component of the Set1 Kabuki pocket and directly interacts with Swd3-Trp197 of the SMART loop (Figure 5B). Strikingly, in our more challenging “up-engineering” attempt, the mono-methyltransferase MLL3 can be effectively converted into a robust tri-methyltransferase by simply replacing its Asn-Arg residues in the door loop with Gly-Ile found in MLL1/2 and SETD1A/B (Figure 6C and 6D, lane 5 vs. lane 8 & 9).

Based on our superposition analysis, we propose that these changes most likely eliminated a “doorstop” mechanism that keeps MLL3 from multi-methylating its substrate. When the complete RBBP5/Swd1 WDRP loop is modeled on MLL3 SET domain in its closed conformation, the arginine residue in the MLL3 door loop, Arg4822, is predicted to form a perfect salt bridge with the strictly conserved RBBP5/Swd1 aspartate residue, RBBP5-Asp335/Swd1-Asp362 (Figure 6E). Although an open conformation of MLL3 has recently been determined in complex with a partial RBBP5/Swd1 WDRP loop (Li et al., 2016), the two doorstop residues would sterically collide when the aspartate residue is modeled in with the rest of the WDRP loop (Figure 6F). The opening of the MLL3 SET-I sub-domain, therefore, is likely to be more limited in the presence of the complete RBBP5/Swd1 WDRP loop. In contrast, the isoleucine residue (Ile3880) in the equivalent position would not induce clashing in the widely open form of MLL1 (Figure 6F). Together, these analyses suggest that the signature Asn-Arg sequence most likely restricts the activity of MLL3/4 to mono-methylation by acting as a doorstop to effectively limit the large breathing motion of the SET-I sub-domain, which is necessary for up-taking and accommodating mono- and particularly di-methylated substrate lysine (Figure S6D). In congruence with this model, the RBBP5 Asp335Ala mutant, which is expected to partially release the doorstop mechanism, allowed the wild type MLL3 to catalyze higher order methylation (Figure 6G). While a minimized ASH2L-RBBP5 WDRP fragment has been previously shown to confer catalytic activity to the SET1/MLL catalytic domains by reducing their inherent flexibility, our results suggest that further restricted subdomain movement is involved in controlling substrate specificity.

Molecular biology and protein purification

All COMPASS subunits were cloned from K. lactis genomic DNA and subcloned into pFastBac vectors for protein expression in insect cells. Recombinant viruses were produced per manufacturer instructions (Life Technologies) and were amplified three times in Sf9 monolayer cells to produce high titer P4 viruses for use in infection of HighFive monolayer cells to produce protein.

Cells were harvested after 2–3 days post-infection and lysed by sonication in a buffer containing 40 mM HEPES pH 7.5, 350 mM NaCl, 5 mM βME, 10% glycerol. Lysates were clarified by ultracentrifugation. The cleared lysate was purified using a glutathione affinity column (GE Healthcare) pre-equilibrated in the lysis buffer. Protein was released from the column by overnight on-column cleavage of the GST-tag by TEV protease at 4^oC. Eluted material was further purified on a 5 mL HiTrap Q-HP column (GE Healthcare), followed by a Superdex 200 Increase gel filtration column (GE Healthcare) in a final buffer containing 10 mM HEPES pH 7.5, 100 mM NaCl, 1 mM DTT. Proteins were concentrated to 10–20 mg/mL and flash frozen in liquid nitrogen for future use.

Mutants were generated using overlap extension PCR, and purified similarly as wild type Set1 complexes. SET1/MLL family members and WRAD subunits were all PCR amplified from a human cDNA library, and subcloned into pFastBac vectors for co-expression in insect cells. Wild-type and mutant complexes of SET1/MLL were purified similarly to the yeast Set1 complex, concentrated to 2–5 mg/mL and flash frozen in liquid nitrogen.

Nucleosome reconstitution

Histones were purified according to the one-pot refolding method described previously (Lee et al., 2015). The 147-bp 601 DNA was purified from a plasmid encoding 20 repeats of the sequence, following a previously established protocol (Dyer et al., 2004). The plasmid was a kind gift from R. McGinty (McGinty et al., 2016). Small scale reconstitutions were performed to find a 1: 1 match of histones:DNA. Large scale reconstitutions were subsequently performed by dialyzing the sample with a buffer containing high salt (2 M NaCl) to low salt (0 M NaCl) over a 36 hour gradient. Nucleosomes were concentrated to ~1 mg/mL and stored at 4^oC.

COMPASS catalytic core crystallization

Set1 complex samples were mixed with a synthetic H3K4M peptide (Genscript) and SAM (NEB) to final concentrations of 0.4 and 0.5 mM respectively, and incubated on ice for one hour. Trypsin (Sigma-Aldrich) was subsequently added to the sample at a w/w trypsin:protein ratio of 1:300. The samples were further incubated on ice for another 30 minutes prior to crystallization. Protein was mixed with precipitant (0.1 M Bicine pH 9.0, 7–10% PEG20000, 100 mM Potassium Sodium Tartrate) at a ratio of 2:1 in the hanging drop format at 4^oC. Crystals appeared in 2–3 days, and would complete growth in approximately a week. Crystals were then slowly transferred into mother liquor supplemented by 30% ethylene glycol by increasing the ethylene glycol concentrations in 10% intervals, with a minimum of 30 minutes per transfer step. Crystals were then frozen in liquid nitrogen for synchrotron data collection.

X-ray data collection and structure determination

X-ray diffraction data was collected at the Advanced Light Source at Berkeley on beam lines 8.2.1 and 8.2.2. Due to radiation damage and the relatively low symmetry of the crystal (C2), several datasets with isomorphous crystals were collected and merged. Diffraction data was indexed, integrated, and scaled using the HKL2000 package (Otwinowski and Minor, 1997). Resolution cutoffs were determined using completeness (>80%) and I/σ>1 as primary criteria.

Initial phases were determined using a sequential molecular replacement search, first with the structure of WDR5 (PDB: 2H14) as a polyAla model, followed by a search using the structure of MLL3-ASH2LSPRY (PDB: 5F6K) also in polyAla chains. The RBBP5 peptide in this structure was removed for this search. With the successful identification of these three subunits, another search was performed using WDR5 to locate the WD40 repeat domain of Swd1 in the complex. Density modification was subsequently performed with RESOLVE (Terwilliger, 2000) to obtain an interpretable electron density map for model building using Coot (Emsley and Cowtan, 2004). After cycles of model building and refinement using Phenix (Adams et al., 2010), a final model was obtained with Rwork/Rfree of 22.8/27.1 (see Figure S4C for representative densities). All structure figures were drawn using PyMOL (Delano) and all structural modeling were based on structural superposition.

Protein co-purification pulldown

Three 150 mm plates of HighFive monolayer cells were co-infected with viruses encoding the proteins of interest. Cells were harvested after 2–3 days of infection and lysed in lysis buffer. Clarified lysates were loaded on to a 50 μL glutathione affinity column equilibrated in lysis buffer, and then washed with 20 column volumes of buffer. Material was eluted from the column by boiling the beads in SDS-PAGE loading buffer. After electrophoresis, the gel was stained by Coomassie for analysis. For inputs, lysates were loaded on a gel, and then transferred to PVDF membranes. The membranes were stained with Ponceau to assess even loading of lysates on the membranes, and then blocked by 5% milk in TBS-T for 30 minutes at room temperature with gentle rocking. The blots were probed by anti-His and anti-GST antibodies in 5% milk in TBS-T overnight at 4^oC. In the following morning, blots were washed with TBS-T, and appropriate secondary antibody coupled with HRP was added to the blots and incubated for one hour at room temperature. Blots were washed with TBS-T to remove excess secondary. Finally, ECL reagent was added, and blots were exposed on an imager for analysis.

Methyltransferase assays

1 μM of enzyme and 0.5 μM nucleosomes were incubated in a buffer containing 20 mM HEPES pH 7.5, 100 mM NaCl, 1 mM DTT, 0.2 mM SAM for 30 minutes at 30^oC. Reactions were quenched with SDS-PAGE loading buffer, resolved on a 15% gel, and transferred to PVDF membranes. Membranes were blocked in 5% milk for 30 minutes at room temperature, and then probed with antibodies against H3K4me1 (CST D1A9), H3K4me2 (Abcam ab7766), H3K4me3 (Abcam ab8580) and H3 (Abcam ab1791) overnight. Blots were washed with TBS-T and then incubated with secondary antibody for 1 hour at room temperature. Blots were subsequently washed with TBS-T to remove excess secondary. Finally, ECL reagent was then added and blots were exposed on an imager for analysis.