Arginine Methylation of the Histone H3 Tail Impedes Effector Binding*

Histone tail post-translational modification results in changes in cellular processes, either by generating or blocking docking sites for histone code readers or by altering the higher order chromatin structure. H3K4me3 is known to mark the promoter regions of active transcription. Proteins bind H3K4 in a methyl-dependent manner and aid in the recruitment of histone-remodeling enzymes and transcriptional cofactors. The H3K4me3 binders harbor methyl-specific chromatin binding domains, including plant homeodomain, Chromo, and tudor domains. Structural analysis of the plant homeodomains present in effector proteins, as well as the WD40 repeats of WDR5, reveals critical contacts between residues in these domains and H3R2. The intimate contact between H3R2 and these domain types leads to the hypothesis that methylation of this arginine residue antagonizes the binding of effector proteins to the N-terminal tail of H3. Here we show that H3 tail binding effector proteins are indeed sensitive to H3R2 methylation and that PRMT6, not CARM1/PRMT4, is the primary methyltransferase acting on this site. We have tested the expression of a select group of H3K4 effector-regulated genes in PRMT6 knockdown cells and found that their levels are altered. Thus, PRMT6 methylates H3R2 and is a negative regulator of N-terminal H3 tail binding.

The tight packing of DNA into chromatin creates a need for mechanisms to relax chromatin and expose DNA for transcription, replication, and DNA repair (1). One of the mechanisms used by the cell to access DNA is the post-translational modification of histone tails. Specifically, methylation of histone tails generates a docking site for effector proteins, which aid in the recruitment of other enzymes necessary for the function at hand. In general, methylation of histone residues lysines 4 and 36 on H3 are correlated with active gene regions, whereas methylation of lysines 9 and 27 on H3 is correlated with repressed gene regions, although exceptions exist (2). The domain types that bind histone tails include the Chromodomain, tudor domains, MBT domains, WD40 repeats, and PHD 5 fingers (3)(4)(5).
Recently, two groups reported that select PHD fingers have the propensity to bind trimethyl lysine 4 on H3 (6,7). The structures further showed important aromatic residues in the PHD that cage the methylated lysine but also revealed critical contacts made between the arginine at the second position of the H3 tail and the PHD (8,9). During this same time, the WD40 domain of WDR5 was reported to complex with the H3 tail (10). The structure of the WD40 repeats of WDR5 revealed arginine 2 of H3, and not lysine 4, buried within the donut hole of the large domain (11,12). Specifically, four amino acids in WDR5 critically interact with arginine 2 (11). In addition, the tudor domains of JMJD2A also bind in an H3K4me3-dependent manner, and again, the H3R2 residue forms critical interactions with an Asp residue of one of the tudor domains (13). The analysis of the structures of these three different domain types bound to the N-terminal tail of histone H3 led us to ask whether methylation of this arginine, reportedly by CARM1 (14), could block binding of these domain types to the histone. If true, methylation of arginine 2 would be a critical epigenetic mark that will antagonize lysine 4 methylation by blocking effector protein binding. Currently, little is known about H3R2 methylation, including the in vivo enzyme acting on this site, whether it can exist in combination with H3K4 methylation, and how broad an effect H3R2 methylation has on H3 effector binding.
To answer these questions, we screened a collection of PRMTs to identify the H3R2-methylating enzyme. Here we show that PRMT6, an enzyme with restricted nuclear localization (15), and not CARM1, methylates H3R2 in vitro, and H3R2 methylation is still present in CARM1-null embryos. We also screened a panel of histone binding domains for sensitivity to H3R2 methylation. We found that a large number of H3 binding domains are sensitive to H3R2 methylation and add additional domains to the list of methyl-specific chromatin binders. In vitro methylation experiments show that R2me2 and K4me3 can exist on the same histone molecule. In addition, finally, the transcriptional activity of select genes, known to be affected by the transcriptional regulatory complexes containing the H3R2me2a-sensitive effectors, is altered when PRMT6 levels are reduced or increased.

EXPERIMENTAL PROCEDURES
In Vitro Methylation Reactions-The GST-PRMT1, GST-CARM1, and GST-PRMT6 were expressed and purified as described previously (18). In vitro methylation reactions were * This work was supported by National Institutes of Health Grants DK62248 and DA023449. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1  Sciences), and exposed to film for 1-3 days at Ϫ80°C. Acid Extraction of Histones-HEK293 and U2OS cells were grown to 80% confluency. Cells were suspended in reticulocyte standard buffer (10 mM Tris-HCl, pH 7.4; 10 mM NaCl; 3 mM MgCl 2 ) and then centrifuged. The pellet was resuspended in reticulocyte standard buffer plus 0.5% Nonidet P-40, placed on ice for 10 min, and then centrifuged again (2500 ϫ g). Nuclei were resuspended in 5 mM MgCl 2 , an equal volume of 0.8 M HCl was added, and histones were extracted for 20 min on ice. Histones (in supernatant) were precipitated with 50% (w/v) trichloroacetic acid and centrifuged at 8,000 ϫ g. The pellet was washed twice with cold acetone and then resuspended in deionized water and 2 l of 1.0 M Tris-HCl, pH 8.8.
Histological Analysis and Antibodies-E18.5 embryos with their abdomens perforated were fixed in formalin and embedded in paraffin wax. Embryos were sectioned at 3 m and subjected to immunohistochemical localization of ␣CARM1 (Upstate Biotechnology), ␣H3R2me2a (Abcam), and ␣H3R17me2a (Upstate Biotechnology). Staining was performed using the En-Vision system (DAKO), and the counterstain was hematoxylin. The H3R2me1 antibody is from Abcam, and the PRMT6 antibody is from Bethyl Laboratories, Inc.
Peptide Pulldowns-Biotinylated histone tail peptides (15 g) were immobilized on 8 l of streptavidin beads (Pierce) in 500 l of pulldown buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM dithiothreitol, and 0.1% Nonidet P-40 (v/v), 1 M ZnSO 4 ) for 2 h at room temperature. Immobilized peptide bead complexes were washed three times with pulldown buffer. 1 g of GST fusion protein and 400 l of pulldown buffer were added to beads and rocked overnight at 4°C. The beads were then washed five times with pulldown buffer, boiled in protein loading buffer, fractionated by 10% SDS-polyacrylamide gel electrophoresis, and subjected to Western blot analysis using an anti-GST antibody.
Stable shRNA Line Generation-Plasmids designed to express shRNAs targeting nucleotides 995-1014 of human PRMT6 (GenBank TM accession number AY043278) were constructed by using pSUPER RNA interference system from Oligoengine (Seattle, WA) according to the manufacturer's protocol. The insert target sequences are 5Ј-gatccccGCAA-GACACGGACGTTTCAttcaagagaTGAAACGTCCGTGT-CTTGCtttttggaaa-3Ј (forward) and agcttttccaaaaaGCAAG-ACACGGACGTTTCAtctcttgaaTGAAACGTCCGTGTCT-TGCggg-5Ј (reverse). U2OS cells were then transfected with pSUPERretro vector encoding the PRMT6-shRNA using Lipofectamine 2000 (Invitrogen). A polyclonal population was selected in 2 g/ml puromycin, and ring cloning was performed. Expression of PRMT6 in each clone was analyzed by Western blots in triplicate.
RNA Isolation and Quantitative real-time-PCR-Cells at 80% confluency were trypsinized and collected. The pellet was washed in phosphate-buffered saline, and 10% was used to make a whole cell extract for Western analysis on PRMT6 protein levels. The remaining cells were spun down and then RNAisolated following the Qiagen RNeasy mini prep manufacturer's protocol. An on-column DNA digestion was performed in each RNA sample preparation. cDNA was prepared from total RNA using the Applied Biosystems high capacity cDNA archive kit following the manufacturer's protocol. Real-time prevalidated gene primer sets were purchased from the Applied Biosystems "Assays-on-Demand" and analyzed with the Applied Biosystems 7900HT real-time PCR instrument using the TaqMan universal master mix. 18 S RNA was used as the internal control.
ChIP Analysis-U2OS cells at 80% confluency were transfected with 20 g of 3ϫFLAG-ING2 vector (6). 24 h after transfection, ChIP analysis was performed following the Upstate Biotechnology ChIP assay kit protocol (catalog number . For the immunoprecipitation, 2 g of the ␣FLAG antibody (Sigma, catalog number F-3165) was used for each condition and incubated with the cross-linked complexes overnight at 4°C. PCR of the cyclin D1 promoter on the input and isolated DNA was performed using Advantage 2 polymerase (Clontech 639201). The cyclin D1 promoter primer sequences were as follows: forward, 5Ј-GATTTTCTTTCAAACAACGTGGTTAC-3Ј, and reverse, 5Ј-TCTTGGTGACCATTTGGAGACA-3Ј.

RESULTS AND DISCUSSION
PRMT6 Is the H3R2 Arginine Methyltransferase-CARM1 is reported to methylate H3R2, based on peptide mapping of in vitro methylated H3 (14). We decided to further study this using recombinant PRMTs and CARM1 knock-out mice. Based on the previous reports, we hypothesized that CARM1-null embryos would lose immune reactivity with an H3R2me2aspecific antibody. Surprisingly, immunostaining of wild-type and CARM1-null E18.5 embryos reveals a loss of H3R17 methylation, but not of H2R2 methylation, upon CARM1 loss (Fig.  1A). We then performed in vitro methylation experiments on calf thymus core histones using a set of recombinant PRMTs. Both CARM1 and PRMT6 robustly methylated H3 (Fig. 1B). PRMT1 methylates histone H4 and H2A. Next, calf thymus H3 was again methylated in vitro with this same set of PRMTs. The methylated H3 was then used for fluorography and Western analysis, using antibodies specific for histones methylated at different sites. The specificity of the antibodies was tested and only recognizes H3 when modified at the indicated site (data not shown). Surprisingly, although CARM1 dramatically increased the amount of R17 methylation, it did not increase the levels of arginine 2 methylation. PRMT6, and to a lesser degree PRMT1, catalyzed the methylation of H3R2 in vitro (Fig.  1C). Therefore we conclude that PRMT6 is the primary enzyme responsible for H3R2 methylation.
Knockdown of PRMT6 Decreases H3R2me2a, whereas Overexpression Increases H3R2me2a-To further validate the data obtained from the in vitro methylation experiments, we overexpressed PRMT6 in HEK293 cells (and HeLa cells, data not shown), which clearly led to an increase in H3R2me2a on bulk ACCELERATED PUBLICATION: PRMT6 Methylates H3R2 FEBRUARY 8, 2008 • VOLUME 283 • NUMBER 6 histones ( Fig. 2A). If overexpression of PRMT6 results in increased H3R2me2a levels, one would expect that by knocking down endogenous PRMT6 levels, we would see a decrease in H3R2me2a methylation levels on bulk histones. Indeed, this is what was observed when we generated a stable PRMT6 knockdown cell line in U2OS cells (Fig. 2B). These data further support the finding that PRMT6 is the H3R2 methyltransferase.
H3 Binding Domains Are Sensitive to H3R2 Methylation-The published structures of the ING2 PHD finger, the WDR5 WD40 domain, and the JMJD2A tudor domains all show important contacts between domain residues and arginine 2 of H3. Thus, we were interested in looking at the effects of H3R2 methylation on the binding of these domain types. To address this, we synthesized biotinylated peptide and tested the ability of recombinant PHD, tudor, and WD40 domains to bind unmodified (H3K4me0), H3K4me, and the H3R2me2aK4me3 "dual" modified peptide using a peptide pulldown approach. The selected domains tested included the tudors of JMJD2A, the WD40 domains of WDR5, and several PHD domains that have been reported to bind H3K4me3 (ING2 and BPTF) and others tested here for the first time (PHF2, DATF1, and RAG2). Clearly, the majority of domains tested in this assay are sensitive to H3R2 methylation, as seen by reduced binding to the dual peptide as compared with the H3K4me3 peptide (Fig. 3A). Importantly, not all domains (RAG2-PHD) are sensitive, implying that certain H3-binding proteins are not regulated by this mechanism (19,20). WDR5 binds the unmodified H3 N-terminal tail peptide well, regardless of H3K4 methylation. However, again, we see that H3R2 methylation reduces this binding.
PRMT6 Can Methylate H3 Peptides Regardless of Lysine 4 Methylation Levels-Up to this point, it is unknown whether the combination of H3K4me3 and H3R2me2 exists on the same histone molecule in vivo. To address this question, we performed an in vitro methylation experiment using histone peptides. Peptides corresponding to amino acids 1-18 of H3 containing varying degrees of methylation were subjected to in vitro methylation by PRMT6. Fluorography of the peptides revealed that PRMT6 has the ability to methylate all the H3 peptides regardless of the lysine 4 methylation status (Fig. 3B), although methylation of the H3K4me3 substrate is less robust than the other lysine-methylated peptides tested. Thus, PRMT6 can lay down the R2me2 mark, whereas H3K4 is methylated in vitro. The endogenous baseline levels of the dual mark and the stability of the dual mark are currently unknown. However, ChIP-on-chip analysis of these two marks indicate that they do not co-segregate and are in fact mutually exclusive (21). The ability of PRMT6 to in vitro methylate a peptide that is already methylated at H3K4 suggests that this dual modification can exist in vivo and raises the possibility that it is not very stable. Perhaps, after the dual mark is generated, efficient lysine demethylation causes the rapid loss of the H3K4me3 mark. We have not tested the ability of SET domain-containing proteins  PRMT6 methylates H3R2 in vitro. A, immunohistochemical analysis of E18.5 CARM1 wild-type (ϩ/ϩ) and knock-out (Ϫ/Ϫ) embryo brains. Paraffin-embedded sections were stained with ␣CARM1, ␣H3R17me2, and ␣H3R2me2 antibodies. S ϭ skin, N ϭ neopallial cortex, I ϭ intermediate zone, and V ϭ ventricular zone. B, calf thymus core histones were methylated in vitro with recombinant PRMT1, CARM1, and PRMT6. A radioactive methyl donor is used to label reaction products, which were then subjected to a fluorograph to visualize methylation events. C, calf thymus H3 was methylated in vitro with PRMT1, CARM1, and PRMT6 and subjected to a fluorograph to visualize methylation events. The same histone samples were also used for Western blot analysis using antibodies specific for H3R2me1 and me2a and for H3R17me2a.

FIGURE 2. PRMT6 regulates H3R2 methylation levels in cells.
A, PRMT6 was transiently overexpressed in HEK293 cells by transfection of a pCAGGS-PRMT6 vector (ϩPRMT6). 24 h after transfection, the cells were harvested. Analysis of total cell lysate revealed PRMT6 levels. Core histones were isolated and analyzed for H3R2me2a levels. B, a stable PRMT6 knockdown U2OS cell line was generated using shRNA. Core histones were isolated from the mock stable line (Mock) and the PRMT6 knockdown line and analyzed by Western blot for H3R2me2a levels. Western analysis of total cell lysates confirms efficiency of the PRMT6 knockdown.
to methylate an H3R2me2a peptide. A large number of enzymes have been reported to methylate the H3K4 site (22).
PRMT6 Activity Affects HOXA5 and Cyclin D1 Gene Expression-As H3R2 methylation reduces binding of WDR5 and ING2 chromatin binding domains to H3 N-terminal tail peptides, we hypothesized that the recruitment of these complexes to gene promoters may be hindered when PRMT6 is overexpressed, or conversely, overactive when PRMT6 is knocked down. We collected RNA from mock stably transfected U2OS cells, U2OS cells that were transiently transfected with a PRMT6 expression vector (ϩPRMT6), and shPRMT6 U2OS cells for real-time PCR analysis on HOXA5 and cyclin D1 genes. The expression of these genes is regulated by the WDR5-MLL and ING2-HDAC1-mSin3a protein complexes (6,23). Specifically, the WDR5-MLL complex positively regulates HOXA5 transcription, whereas the ING2-HDAC1 complex negatively regulates cyclin D1 transcription after DNA damage. In agreement with our hypothesis, overexpression of PRMT6 decreased HOXA5 expression, whereas PRMT6 knockdown increased expression more than 2-fold (Fig. 3C). In addition, repression of cyclin D1 after doxorubicin treatment, by the ING2-HDAC1 complex, was very robust in the PRMT6 knockdown cells and weakened in the PRMT6-overexpressing cells (Fig. 3D). The robust repression of cyclin D1 after doxorubicin treatment in the shPRMT6 cells is likely due to the increased binding of the ING2 repressive complex; H3R2 methylation antagonizes ING2 binding, and PRMT6 knockdown reduces bulk H3R2 methylation levels. Indeed, ChIP analysis of FLAG-ING2 demonstrates the increased physical association of ING2 at the cyclin D1 locus when PRMT6 is knocked down (Fig. 3, E  and F).
Here we identified a number of H3-binding proteins sensitive to arginine 2 dimethylation (Fig. 3A). We also showed that the H3R2 can be methylated by PRMT6 when lysine 4 is previously methylated. One can imagine, perhaps during cellular differentiation, PRMT6 being recruited to active genes to start a shutdown procedure. H3R2 methylation by PRMT6 would knock off K4me3-binding proteins and their associated transcriptional coactivators. Following this event, H3K4 demethylases may be recruited to remove this activating mark to permanently silence the region. Supporting this hypothesis, Guccione et al. (21) show a counter-correlation between R2me2 and K4me3 on genomic regions bound by the transcription factor Myc. Their result implies that the dual mark is transient and/or present at very low amounts. H3K4 demethylase recruitment is one possible mechanism resulting in a transient dual R2K4 modified state on H3.
This study highlights the importance of a combinatorial histone code and will open the door to many future studies on the effects adjacent histone modifications have on each other and effector protein binding. Already described is the interplay Biotinylated histone tail peptides were immobilized onto streptavidin-conjugated Sepharose beads. Different GST fusion protein (1 g) were added to the beads and allowed to bind overnight. Three washes removed unspecific bound protein. The bound fraction was boiled off beads into sample buffer and subjected to SDS-PAGE and Western analysis with an ␣GST antibody. IN, input. B, biotinylated peptides (15 g) were methylated with GST-PRMT6 in the presence of radioactive methyl donor and then subjected to a fluorograph to visualize methylation events. The same blot was probed with streptavidin-horseradish peroxidase (streptavidin-HRP) to confirm roughly equal loading of the indicated peptides. C, U2OS cells were transfected with pCAGGS-PRMT6 vector (ϩPRMT6). 36 h after transfection, the cells were harvested, and RNA was isolated. RNA was also isolated from mock stable transfected U2OS (Mock) and the shPRMT6 knockdown line. Quantitative real-time PCR analysis was used to measure gene expression of HOXA5. D, U2OS cells were transfected with pCAGGS-PRMT6. 24 h later, 0.2 g/ml doxorubicin (Dox) was added to the PRMT6 transfected wild type, the Mock stable line, and the PRMT6 knockdown line. 24 h later, total RNA was isolated from cell lysates and analyzed with real-time PCR for the number of cyclin D1 transcripts. The results are expressed as the mean Ϯ S.D. of two independent experiments done in triplicate. E, ChIP analysis of FLAG-ING2 at the cyclin D1 locus. F, controls for the ChIP experiment demonstrating efficient knockdown of PRMT6. between H3K9 methylation and H3S10 phosphorylation. Namely, HP1 binding to H3K9me3 is antagonized by H3S10 phosphorylation during mitosis (24). We performed a limited screen for genes affected by PRMT6 activity and found that HOXA5 and cyclin D1 expression/repression is responsive to PRMT6 levels, presumably through methylation of H3R2me2.