EHMT1 Protein Binds to Nuclear Factor-κB p50 and Represses Gene Expression*

Background: Methylation of histone H3 lysine 9 (H3K9) is commonly correlated with gene repression. Results: p50 recruits EHMT1 to repress gene expression. Conclusion: EHMT1 is a negative regulator of NF-κB and type I interferon induction pathways. Significance: p50-dependent changes in the chromatin structure importantly contribute to the repression of the expression genes in resting cells. Transcriptional homeostasis relies on the balance between positive and negative regulation of gene transcription. Methylation of histone H3 lysine 9 (H3K9) is commonly correlated with gene repression. Here, we report that a euchromatic H3K9 methyltransferase, EHMT1, functions as a negative regulator in both the NF-κB- and type I interferon-mediated gene induction pathways. EHMT1 catalyzes H3K9 methylation at promoters of NF-κB target genes. Moreover, EHMT1 interacts with p50, and, surprisingly, p50 appears to repress the expression of type I interferon genes and genes activated by type I interferons by recruiting EHMT1 to catalyze H3K9 methylation at their promoter regions. Silencing the expression of EHMT1 by RNA interference enhances expression of a subset NF-κB-regulated genes, augments interferon production, and augments antiviral immunity.

Transcriptional homeostasis relies on the balance between positive and negative regulation of gene transcription. Methylation of histone H3 lysine 9 (H3K9) is commonly correlated with gene repression. Here, we report that a euchromatic H3K9 methyltransferase, EHMT1, functions as a negative regulator in both the NF-B-and type I interferon-mediated gene induction pathways. EHMT1 catalyzes H3K9 methylation at promoters of NF-B target genes. Moreover, EHMT1 interacts with p50, and, surprisingly, p50 appears to repress the expression of type I interferon genes and genes activated by type I interferons by recruiting EHMT1 to catalyze H3K9 methylation at their promoter regions. Silencing the expression of EHMT1 by RNA interference enhances expression of a subset NF-B-regulated genes, augments interferon production, and augments antiviral immunity.
Covalent modifications of chromatin have emerged as key determinants of the transcriptional competence of the genome. Indeed, a strong correlation exists between the expression status of a gene and specific histone modification (1). Although histone methylation was discovered 4 decades ago (2), its physiological role was unclear until recently. Site-specific methylations of specific histone residues correlate with either activation or repression of transcription. For example, H3K4 2 methylation is a mark of active genes, whereas H3K9 methylation has been correlated with gene repression (3)(4)(5).
There are seven known H3K9 histone methyltransferases (H3K9-HMTs) in the human genome: SUV39H1, SUV39H2, ESET, RIZ, EHMT1 (GLP), EHMT2 (G9a), and SETDB2. The most studied H3K9-HMTs are SUV39H1 and SUV39H2, which promote the formation of pericentric heterochromatin (6,7). Methylation of H3K9 by SUV39H1 creates a docking site for subsequent binding of the heterochromatin protein HP1, providing compelling evidence linking H3K9 methylation to heterochromatin formation and transcriptional repression (8 -10). In contrast, we know that EHMT1 and EHMT2 are the major euchromatic H3K9-HMTs because cells lacking either EHMT1 or EHMT2 show decreased H3K9me2 in euchromatin, whereas the level of H3K9me2 remains unchanged in heterochromatin (11,12). EHMT1 is ubiquitously expressed. Knocking out the EHMT1 gene in mice leads to embryonic lethality (12). The global levels of H3K9me1 and H3K9me2, but not H3K9me3, are lower in EHMT1-deficient ES cells than in wild-type ES cells (12). However, it has been shown that EHMT1 is capable of catalyzing formation of H3K9me1, H3K9me2, and H3K9me3 in an in vitro methyltransferase assay (13).
NF-B is a transcription factor that plays a pivotal role in regulating multiple biological functions including inflammation, immunity, cell proliferation, and apoptosis. In resting cells, NF-〉 is sequestered in the cytoplasm by interaction with IB␣. Upon activation, IB␣ is rapidly degraded by the ubiquitin-proteasome pathway. Free NF-B is then translocated into the nucleus where it turns on expression of a large array of target genes (14,15). Although inducible nuclear translocation is critical for NF-B activation, many post-translational modifications have been shown to regulate the nuclear function of NF-B, including phosphorylation, ubiquitination, nitrosylation, acetylation, and methylation (16,17).
We previously demonstrated that NF-B is positively regulated by methylation of the p65 subunit (17). In this report, we show that NF-B action as well as that of type I interferon are negatively regulated by EHMT1. We show that the expression of 44% of TNF-induced genes is enhanced in cells lacking EHMT1. Using IL8 as a model, we further show that EHMT1 is required for establishing H3K9 methylation at the IL8 promoter. Furthermore, EHMT1 interacts with the p50 subunit of NF-B. Surprisingly, p50 recruits EHMT1 to the promoters of genes that respond to type I interferon and represses the expression of these genes. Silencing the expression of either p50 or EHMT1 augments interferon production and strengthens the interferon-mediated inhibition of virus replication.
Flow Cytometry-A549-PB1 and 293T-PB1 cells were infected with influenza virus for 12 h. Cells were washed and fixed with paraformaldehyde (1% final). Fixed cells were assayed using a BD FACScan flow cytometer (BD Biosciences) and further analyzed with FlowJo software.
RT-PCR-Cells treated with TNF␣ for 2 h were lysed with TRIzol LS reagent (Invitrogen) to isolate total RNAs. cDNAs were synthesized with the SuperScript III Reverse Transcriptase kit (Invitrogen). Quantitative PCR was performed with the 2ϫ SYBR Green PCR Master Mix (Applied Biosystems) and run on the Applied Biosystems ABI 7300 Real-time PCR System. All data were normalized to L32. The sequences of the primers are listed in supplemental Table S2.
ChIP Assay-ChIP assay was carried out using the Fast-ChIP protocol with a slight modification (19). Briefly, cells were cross-linked with 1% formaldehyde at room temperature for 10 min. Excess formaldehyde was neutralized by adding 1 ml of 1.25 M glycine at room temperature for 5 min. Cells were washed three times with PBS and lysed in the hypotonic buffer. The nuclei were isolated by centrifugation at 500 ϫ g 4°C for 5 min and resuspended in 0.5 ml of SDS lysis buffer (50 mM Tris, pH 8.0, 10 mM EDTA, and 1% SDS). DNAs were sheared by sonication with a Branson 450 sonicator (output 2, 90% duty cycle, 10 ϫ10-s pulse). Ten micrograms of sheared DNAs was diluted with a 9ϫ volume of dilution buffer (16.7 mM Tris, pH 8.0, 167 mM NaCl, 1.2 mM EDTA, 1% Triton X-100, 100 g/ml salmon sperm DNA, and 2% BSA). Antibodies and Dynabeads protein A (Invitrogen) were added and incubated at 4°C for 12 h. The beads were then washed two times with a low salt washing buffer (50 mM Tris, pH 7.5, 200 mM NaCl, 1% Triton X-100, 2 mM EDTA, 0.5% Nonidet P-40) followed by two times with a high salt washing buffer (50 mM HEPES, pH 7.9, 500 mM NaCl, 1% Triton X-100, 0.1% SDS, 0.1% 2,5-dimethoxy-4chloramphetamine) and one time with TE buffer. The beads were then resuspended with 100 l of 10% Chelex 100 beads, and cross-linking was reversed by boiling the beads for 10 min. The beads were then treated with 1 l of proteinase K and incubated at 56°C for 30 min. The proteinase K was inactivated by boiling for 10 min. Supernatants containing DNA were collected, and the beads were washed with 120 l of dH 2 O. Supernatants were collected and analyzed by quantitative PCR. The sequences of the primers are listed in supplemental Table S2.
Microarray-For microarray hybridization, 2 g of RNA was labeled, fragmented, and hybridized to an Affymetrix GeneChip Human Genome U133 Plus 2.0 Array. Data were then analyzed with the Rosetta Resolver gene expression data analysis system (Rosetta Inpharmatics). Expression differences Ն2-fold with a p value Ͻ0.01 were considered significant. The Gene Expression Omnibus accession number is GSE39744.
Statistical Analysis-Data were analyzed by Microsoft Excel and are presented as the mean Ϯ S.D. Data are representative of three or more independent experiments. Student's t test was used to determine significant differences between groups.

RESULTS
EHMT1 Is an H3K9-HMT That Negatively Regulates the NF-B Pathway-To test whether H3K9 methylation regulates NF-B-dependent gene expression, we silenced the expression of every known H3K9-HMT in cells using RNA interference (RNAi). Silencing the expression of EHMT1 but not other H3K9-HMTs greatly enhanced TNF␣-induced expression of several NF-B-dependent genes including A20, ICAM1, IL8, FIGURE 1. EHMT1 negatively regulates the NF-B pathway. A, HeLa cells transfected with control or H3K9-HMTs-specific siRNAs were stimulated with TNF␣ for 2 h. The expression of A20, IB␣, IL8, ICAM1, and IP-10 was measured by RT-PCR (*, p Ͻ 0.05; **, p Ͻ 0.01; ns, not significant). B, HeLa cells transfected with control or EHMT1 siRNAs were treated with or without TNF␣ for the indicated time periods. The expression of A20, IL8, and IP10 was measured by RT-PCR. C, BMMs were treated with or without BIX01294 for 24 h then treated with or without TNF␣ for the indicated time periods. The expression of A20 and CCL5 was measured by RT-PCR. D, pie chart represents comparison of 124 TNF␣-induced genes between TNF␣-stimulated HeLa cells transfected with control or EHMT1 siRNAs. A-C, error bars represent the variation range of duplicate experiments. IP10, and IB␣ (Fig. 1A) (20). The efficiency of the RNAi was measured by RT-PCR, and 77-95% of knockdown efficiency was achieved (supplemental Fig. S1A). Although knocking down the expression of EHMT1 enhanced the expression of this suite of genes, it did not affect their expression kinetics (Fig.  1B). For example, mRNA for the immediate early gene, A20, was rapidly induced (peaked at 30 min) and then degraded, whereas that of the late gene, IP10, was not detected until 2 h of TNF stimulation and was further increased to 6 h, regardless of the presence or absence of EHMT1. Similar results were obtained when we treated mouse BMMs with BIX01294, a specific inhibitor of EHMT1 and EHMT2, a closely related H3K9-HMT (13,21). The inhibition of EHMT1 in BMM led to enhanced expression of A20 and CCL5 in response to TNF stimulation (Fig. 1C). These results indicate that EHMT1 is the major H3K9-HMT that negatively regulates the NF-B pathway.
To minimize the possibility that the effect we observed with EHMT1 RNAi was caused by an off-target effect, we tested two different siRNAs specific for EHMT1 (a and b). We found that TNF␣-induced IL8 expression was enhanced in cells transfected with either EHTM1a or EHMT1b siRNAs, and 70 -75% knockdown efficiency was achieved with each (supplemental Fig. S1B).
We next determined whether EHMT1 regulated NF-B activity when it was induced by IL-1␤. As with TNF␣ stimulation, silencing the expression of EHMT1 greatly enhanced the expression of A20, IP10, and IL8 induced by IL-1␤ stimulation (supplemental Fig. S1C), suggesting that EHMT1 negatively regulates NF-B activity induced by various forms of stimulation. We further showed that depletion of EHMT1 blocked neither degradation of IB␣ nor nuclear translocation of p65 (supplemental Fig. S2A). In addition, TNF-induced activation of JNK and p38 was also normal in EHMT1 siRNA-treated cells (supplemental Fig. S2B) suggesting that the hyperresponsiveness of EHMT1-deficient cells is not due to increased receptorproximal signaling.
To examine the range of genes whose transcription is affected by EHMT1, we evaluated the effects of EHMT1 loss of function on TNF␣-induced gene expression by global gene expression profiling. In response to TNF␣ stimulation, 124 genes were up-regulated (Ͼ2-fold) in control siRNA-treated HeLa cells. Treatment with EHMT1-specific siRNA further enhanced the levels of 55 genes (44%) (Ͼ1.5-fold) whereas 53 genes (43%) were not affected to 1.5-fold, and the levels of 16 genes (13%) were reduced ( Fig. 1C and supplemental Table S1). It is evident that EHMT1 blockade augments transcription of many genes in response to TNF␣ induction.
EHMT1 Catalyzes H3K9 Methylation at the Promoters of NF-B-regulated Genes-EHMT1 is a H3K9-HMT. In EHMT1-deficient ES cells, the global H3K9me2 but not H3K9me3 is reduced (12). Consistent with this observation, silencing the expression of EHMT1 by RNAi led to a global reduction of H3K9me2 but not of H3K9me3 ( Fig. 2A). To determine whether EHMT1 catalyzes the H3K9 methylation at the IL8 promoter, we first examined when EHMT1 is present at the IL8 promoter by a ChIP assay. Before induction, EHMT1 was enriched at promoters of IL8 and ␤-globin but not at the pro-moter of GAPDH (Fig. 2B). EHMT1 was also evident in the coding region of the IL8 gene. Knocking down the expression of EHMT1 by RNAi resulted in a 50% reduction of the ChIP signal. These results provide evidence that at the IL8 gene, EHMT1 is present at the inactive locus before induction and reduced but not eliminated by TNF induction.
We then analyzed the levels of H3K9me2 and H3K9me3 in the control or EHMT1 siRNA-transfected cells treated with or without TNF␣. In EHMT1-deficient cells, the basal levels of H3K9me2 and H3K9me3 were reduced by 40% at the IL8 promoter and exon, but not at the ␤-globin promoter (Fig. 2C). This reduction was not caused by nucleosome loss because the data were normalized to the total level of histone H3. Furthermore, TNF␣ induced more pol II recruitment to the IL8 gene in EHMT1-deficient cells than in control siRNA-treated cells. This is consistent with the observation that more IL8 was produced in EHMT1-deficient cells in response to TNF␣ stimulation (Fig. 1A). These results suggest that EHMT1 is the major H3K9-HMT that catalyzes H3K9me2 and H3K9me3 at the IL8 gene. Although EHMT1 is present in the ␤-globin promoter, it is dispensable for methylating H3K9 at the ␤-globin promoter.
p50 contains two functional domains, RHD-N and RHD-C. To determine which domain of p50 is required for interacting with EHMT1, we generated p50 fragments containing either the RHD-N or RHD-C and performed co-immunoprecipitation assay as described above. We found that only the RHD-N of p50 interacted with EHMT1 (Fig. 3B, left panel). Because RHD-N of p50 contains the DNA binding domain, we further tested whether DNA binding is required for interacting with EHMT1. We generated a DNA binding-deficient mutant (p50Y60DG64E (22)) and showed that DNA binding of p50 was dispensable for the binding to EHMT1 (Fig. 3B, right panel).
In contrast, EHMT1 contains an ankyrin repeat domain, a Pre-SET, and a SET domain. To map the p50-interacting domain, serial deletion mutants of EHMT1 were generated. A fragment containing the ankyrin repeats (701-1108 amino acids) was able to interact with p50 in the co-immunoprecipitation assay (Fig. 3C), indicating that the ankyrin repeats are responsible for the interaction with p50. These findings provide further evidence that EHMT1 interacts with p50.
To test whether endogenous p50 and EHMT1 form complexes in cells, we carried out co-immunoprecipitation with nuclear extracts prepared from cells treated with TNF␣ for the indicated time periods. Endogenous EHMT1 was detected in p50 immunoprecipitates, regardless of TNF␣ treatment, but not in IgG or p65 immunoprecipitates (Fig. 3D). Interestingly, p50 was detected in TNF␣-treated p65 immunoprecipitates; however, no EHMT1 was detected, suggesting that EHMT1 interacts with p50 homodimer constitutively but not with p65-p50 heterodimer.
p50 Recruits EHMT1 to the Promoters of Type I Interferonresponding Genes-The p50 homodimer, which lacks a transactivation domain, has been implicated in transcriptional repression (23)(24)(25). Our data show that EHMT1 negatively regulates NF-B target genes and interacts with p50. This suggests that the p50 homodimer may function as a transcriptional repressor by recruiting EHMT1 to the promoter of target genes to promote H3K9 methylation. To test this hypothesis, we first looked for genes that were up-regulated in both p50 and EHMT1 knockdown cells. Surprisingly, we found that expression of many type I interferon-induced genes were up-regulated in EHMT1 knockdown cells (Fig. 4A). Some of these genes, including IFN␤ itself, CCL5, and IFIT2 were also up-regulated in p50 knockdown cells (Fig. 4B). We further showed that the rise in steady mRNA levels of ISGs is not due to an increase in mRNA stability. We treated HeLa cells with BIX01294 for 24 h followed by addition of actinomycin D. We then measured the stability of mRNAs by RT-PCR and found that the mRNA stabilities of IFIT2, IFIT3, and CCL5 mRNAs were similar whether cells were BIX01294-treated and not (supplemental Fig. S3A). Importantly, TNF␣ stimulation does not induce production of IFN␤ under normal conditions. In the absence of p50 or EHMT1, TNF␣ stimulation led to a 3-fold (p50-deficient) and 20-fold (EHMT1-deficient) increase of IFN␤ mRNA. To examine this phenomenon, we determined whether p50 was present at the promoters of IFN␤ and CCL5 by ChIP. Using wild-type and p50-deficient MEFs, we were able to detect a 3-fold increase compared with the knockout of the p50-mediated signal at the promoter of IFN␤ and CCL5 but not at the promoter of ␤-globin, which only showed a background ChIP signal (Fig. 4C, top panel). In contrast, there was no enrichment of p65 at these promoters (supplemental Fig. S3B). Moreover, in the absence of p50, less EHMT1 was bound to the promoters of CCL5 (40% decrease) and IFN␤ (50% decrease, Fig. 4C, second panel). The level of EHMT1 at the promoter of ␤-globin was comparable between wild-type and p50-deficient MEFs.
Consistent with the decreased level of EHMT1, the levels of H3K9me2 at the promoters of CCL5 and IFN␤ were also decreased in the p50-deficient MEFs (Fig. 4C, and data not shown), whereas the levels of histone H3 remained unchanged (Fig. 4C, bottom panel). Similar observations were made when the expression of p50 was knocked down by RNAi in HeLa cells (supplemental Fig. S3C). The levels of EHTM1 and H3K9me2 were decreased at the promoters of IFN␤ and CCL5 in p50specific siRNA-treated HeLa cells. From these data, it appears that p50 represses the expression of IFN␤ and CCL5 in response to TNF␣ by recruiting EHMT1, which in turn methylates H3K9. . EHMT1 interacts with p50. A, HEK293 cells were transfected with EHMT1 along with increasing amounts of FLAG-p65 or FLAG-p50. EHMT1 (middle panels) and FLAG (right panels) immunoprecipitates (IP) as well as inputs (left panels) were immunoblotted (IB) with EHMT1 and FLAG antibodies. B, HEK293 cells were transfected with EHMT1 along with various fragments of p50 as indicated (left panel) or a DNA binding-defective mutant p50 (right panel). FLAG or IgG immunoprecipitates and inputs were immunoblotted with EHMT1 and p50 antibodies. Asterisks indicate different fragments of p50. HC, heavy chains; LC, light chains. C, various deletion mutants of EHMT1 along with FLAG-p50 were co-transfected into HEK293 cells. Inputs and FLAG immunoprecipitates were analyzed with HA and FLAG antibodies. Asterisks indicate different fragments of p50. D, endogenous p50 and p65 were immunoprecipitated from nuclear extracts prepared from HeLa cells treated with or without TNF␣ for the indicated time periods. Immunoprecipitates were probed with various antibodies as indicated. IgG immunoprecipitates were used as a negative control. FIGURE 4. P50 recruits EHMT1 to the promoters of type I interferon-responding genes. A, heat map represents gene expression patterns of type I interferon-responding genes among HeLa cells transfected with control siRNA treated with TNF␣, or with EHMT1-specific siRNA treated with or without TNF␣ (*, p Ͻ 0.05; **, p Ͻ 0.01). B, HeLa cells transfected with control, EHMT1-, or p50-specific siRNAs were treated with or without TNF␣ for 2 h. Expression of IFN␤, CCL5, and IFIT2 was determined by RT-PCR (error bars represent the variation range of duplicate experiments). C, wild-type and p50 Ϫ/Ϫ MEF cells were analyzed by ChIP (p50, EHMT1 and H3K9me2/H3; mean Ϯ S.D., data are representative of three or more independent experiments; *, p Ͻ 0.05).
The ability of p50 to act as a transcriptional repressor through the recruitment of HDAC1 has been reported to be a mode of negative control of expression of some NF-B target genes in resting cells (26). To test whether there is cross-talk between HDAC1-mediated and EHMT1-mediated gene repression, we inhibited the activity of HDAC1 and EHMT1 with TSA and RNAi, respectively. Pretreating the cells with TSA led to higher expression of A20 and IL8 but not TNF␣, IL6, and IFN␤ in response to TNF stimulation (supplemental Fig.  S3D). Interestingly, silencing the expression of EHMT1 with RNAi further enhanced the expression of IL8 and A20 in the presence of TSA. Furthermore, the recruitment of HDAC1 to the promoters of A20, IL8, TNF␣, CCL5, ␤-globin, and IFN␤ remained unchanged regardless of the presence or absence of EHMT1 (supplemental Fig. S3E). These results indicate that HDAC1-mediated and EHMT1-mediated gene repression are independent of one another.
EHMT1 and p50 Are Negative Regulators of the Type I Interferon Pathway-Viral infection of host cells triggers innate and adaptive immune responses that are essential for the survival of the host. IFN␤ is particularly potent as an antiviral agent. Because EHMT1 negatively regulates the expression of IFN␤, we examined the possibility that EHMT1 could affect immune defense against influenza virus. A nonreplicative WSN strain influenza virus carrying mCherry in the place of the PB1 gene was used to infect HEK293 cells stably expressing PB1 protein (HEK293-PB1) (27). In HEK293-PB1 cells treated with a control siRNA, influenza virus infection induced the expression of IFN␤, CCL5, and IFIT2 as measured by RT-PCR (Fig. 5A). The expression of these genes was greatly enhanced in cells treated with siRNAs against p50 or EHMT1. A similar result was obtained when using p50-deficient MEF cells. Infecting the wild-type MEF cells with Sendai virus led to induction of IFN␤, CCL5, and ISG20 (Fig. 5B). The expression of these genes was further enhanced in p50-deficient MEF cells. These data confirm a negative role of EHMT1 and p50 in the type I interferon pathway.
Knocking down the expression of p50 or EHMT1 in HEK293-PB1 decreased the viral titer by 93 and 64%, respectively, compared with control siRNA-treated cells (Fig. 5C). To confirm the negative role of EHMT1 on influenza viral replication, we examined the anti-viral effect of BIX01294, a specific inhibitor of EHMT1 and EHMT2, a closely related H3K9-HMT (13,21). We pretreated the A549-PB1 cells with BIX01294 for 48 h and infected the cells with influenza virus at a multiplicity of infection (m.o.i.) of 1. After 24 h, infected cells were quantified by sorting the virus-infected, mCherry-positive cells using FACS. Pretreating cells with BIX01294 strongly inhibited the viral infection (Fig. 5D). A similar result was observed when using vesicular stomatitis virus. Infection of HeLa cells by vesicular stomatitis virus led to nearly complete killing of the cells within 24 h at a m.o.i. of 0.05 (supplemental Fig. S4A). Pretreating cells with BIX01294 prevented cell killing by vesicular stomatitis virus even at a m.o.i. of 1. In addition, we found that the level of H3K9me2 but not H3K9me3 was reduced in BIX01294pretreated cells, confirming the inhibition of EHMT1 by BIX01294 (supplemental Fig. S4B). Furthermore, we found that the basal level of type I interferon response genes, including CCL5, IFN␤, IFIT2, IFIT3, and ISG20, was up-regulated in BIX01294-treated cells (Fig. 5E), implying that inhibition of EHMT1 activity leads to up-regulation of antiviral genes which in turn protects cells from viral infection. These findings further support the negative role of p50 and EHMT1 in the type I interferon pathway.

DISCUSSION
We have found that the histone methylase EHMT1 is able to negatively regulate a large fraction, but not all, of NF-B-regulated genes. Silencing of EHMT1 causes superactivation of the genes its regulates. It does this through interaction with the p50 subunit of NF-B. Surprisingly, it is carried to the promoters of certain IFN-regulated genes by p50, where it exerts its negative effect on transcription. IFN␤ itself is one of the regulated genes so that silencing of EHMT1 or p50 augments induces production of excess interferon and remarkably strengthens the inhibition of viral infection.
Our findings support a model in which p50-dependent changes in the chromatin structure of the type I interferon response genes importantly contribute to the repression of the expression of these genes in resting cells. This is consistent with the notion that p50 homodimers are involved in preinduction repression of IFN␤ gene transcription as well as subset of interferon-responding genes (28 -31). The function of p50 as a transcriptional repressor involving the recruitment of HDAC1 has been reported to repress the expression of some NF-B target genes in resting cells (26). Here we demonstrated an alternative mechanism involving the recruitment of EHMT1 by p50 to the promoters of IFN␤ and CCL5. EHMT1 in turn promotes H3K9 methylation and results in the formation of chromatin structure that represses transcription (Fig. 6A). How is p50 recruited to the promoters of type I interferon response genes? STAT1 is one of the key transcription factors in the interferon pathway that regulates expression of type I interferon response genes through binding to the IREs (32). It is possible that STAT1 may play a role in recruiting p50 to the promoters of the type I interferon response genes. However, we failed to detect any interaction between p50 and STAT1 (data not shown), and a recent study has shown that p50 binds to G-rich IREs (30). Thus we think that p50 represses type I interferon response genes by directly binding to G-rich IREs. P50 has been shown to bind to some of the NF-B target genes in unstimulated cells (33). P50 may also recruit EHMT1 to repress the expression of a subset of TNF-induced NF-B target genes (Fig. 6A). In addition, EHMT1 can also inhibit the expression of NF-B target genes independent of p50. For example, we were unable to detect any binding of p50 at the promoters of IL8 and IP10 in resting cells (data not shown). The involvement of p50 in the repression of IL8 and IP10 is very unlikely. Thus it seems that EHMT1 is recruited in some other way to the promoters certain NF-B target genes (Fig. 6B). It could involve an undetermined factor X. One strong candidate for X is the p65 subunit of NF-B as it has been recently shown that methylated p65 (p65K310me1) was able to repress the expression of subset NF-B target genes in resting cells by recruiting EHMT1 (34).
Negative regulation of transcription is an important strategy in establishing and maintaining cell-and signal-specific gene expression patterns. Our study indicates that EHTM1 is critical not only for silencing the expression of NF-B-regulated genes and type I interferon-responding genes, but also for establish-ing the signal specificity of gene expression. It has been elegantly shown that different immunostimuli induce different gene expression patterns (35). For example, PAM induces an "inflammatory-like" program, whereas poly(I:C) induces an "anti-viral" program, and LPS induces both the inflammatorylike and the anti-viral programs (35). Under normal conditions, the expression of IFN␤ is induced specifically by a few NF-B agonists, such as LPS and poly(I:C) but not TNF␣ and IL1␤. In the absence of EHMT1, the expression of IFN␤ is induced by treating the cells with TNF␣ or IL1␤, suggesting that EHMT1 is required for establishing the signal-specific induction of IFN␤ ( Fig. 4B and data not shown). Another consequence of using EHTM1 to maintain and establish signal-specific gene expression is that the capacity of gene induction is limited. For example, in response to TNF stimulation, the expression of IL8 (3-fold) and IP10 (2-fold) is higher in cells without the negative regulation of EHMT1 (Fig. 1A).
Our results show that EHMT1 is a negative regulator of type I interferon pathway (Figs. 4 and 5) and thus might be a potential new target for developing new antivirus drugs. Indeed, pharmacological inhibition of EHMT1 with BIX01294 protects cells from virus infection (Fig. 5D). However, EHMT1 has been implicated in playing a role in control of cognition and adaptive behavior in adult mice (36). Conditional EHMT1 and EHMT2 gene inactivation in postnatal mouse neurons led to defects in learning, motivation, and environmental adaptation (36), suggesting that long term inhibition of EHMT1 might affect proper brain functions. Furthermore, EHMT2 has been shown to be involved in cocaine-induced plasticity, and treating mice with BIX01294 increased the preference for cocaine (37). Thus, the potential of using BIX01294 as an antiviral drug might be compromised by adverse side effects related to the biological roles of EHMT1 in other systems such as the CNS. Developing an inhibitor that inhibits the binding between EHMT1 and p50 rather than the enzymatic activity of EHMT1 might be better a strategy but would be very difficult. Further experimentation using animal models will be needed to examine the clinical application of EHMT1 inhibitors.
In preparation of our manuscript, Fang et al. identified EHMT2 (G9a) as a negative regulator of the type I interferon response (38). It has been shown that EHMT1 and EHMT2 form a homodimer and heterodimer in vivo, and deficiency of either EHMT1 or EHMT2 in ES cells results in global reduction of H3K9me2 (39). Therefore, it seems that the EHMT1/ EHMT2 heterodimer is the functional H3K9 methyltransferase in vivo. Our results indicate that EHMT2 in dispensable for the NF-B pathway (Fig. 1A), whereas both EHMT1 and EHMT2 are required for repressing type I interferon pathway (Figs. 4 and 5 and Ref. 38).