|Home | About | Journals | Submit | Contact Us | Français|
Posttranslational modifications of the RelA subunit of NF-κB, including acetylation and methylation, play a key role in controlling the strength and duration of its nuclear activity. Whether these modifications are functionally linked is largely unknown. Here, we show that the acetylation of lysine 310 of RelA impairs the Set9-mediated methylation of lysines 314 and 315, which is important for the ubiquitination and degradation of chromatin-associated RelA. Abolishing the acetylation of lysine 310 either by the deacetylase SIRT1 or by mutating lysine 310 to arginine enhances methylation. Conversely, enhancing the acetylation of lysine 310 by depleting SIRT1 or by replacing lysine 310 with acetyl-mimetic glutamine inhibits methylation, thereby decreasing ubiquitination, prolonging the stability of chromatin-associated RelA, and enhancing the transcriptional activity of NF-κB. The acetylation of lysine 310 of RelA interferes with its interaction with Set9. Based on structural modeling of the SET domain of Set9 with RelA, we propose that the positive charge of lysine 310 is critical for the binding of RelA to a negatively charged “exosite” within the SET domain of Set9. Together, these findings demonstrate for the first time an interplay between RelA acetylation and methylation and also provide a novel mechanism for the regulation of lysine methylation by acetylation.
In several cases an interplay between different posttranslational modifications has been identified for histone and nonhistone proteins, with one modification either enhancing or inhibiting another modification (11, 37, 42). For example, the phosphorylation of serine 10 of histone H3 interferes with the methylation of lysine 9 and stimulates the acetylation of lysine 14 (4, 5, 23, 24). The methylation of lysine 372 of p53 promotes its acetylation (19, 21). These effects can be direct, in which one modification alters the conformation of the protein, thereby influencing the second modification, or can be indirect, in which the first modification results in the recruitment of an effector, which alters the other modification. The interplays between these modifications together with the distinct combinations of covalent modifications form the basis of the “histone code” and, probably, the “protein code” hypotheses (20, 35).
The inducible transcription factor NF-κB plays an important role in regulating inflammatory responses, apoptosis, cell proliferation and differentiation, and tumorigenesis (12). The prototypical NF-κB complex, a heterodimer of p50 and RelA, is sequestered in the cytoplasm by its assembly with the inhibitor IκBα. Upon stimulation, the IκB kinase (IKK) complex is activated, leading to the phosphorylation and degradation of IκBα, the nuclear translocation of NF-κB, and the activation of its target genes (12). Once in the nucleus, the RelA subunit of NF-κB undergoes a series of stimulus-coupled posttranslational modifications, including phosphorylation, acetylation, and methylation. These modifications impose various effects on nuclear NF-κB, regulating both the strength and duration of NF-κB activity (1, 28, 40).
RelA is acetylated at a number of lysine residues by p300/CBP or PCAF. Acetylation at different lysines regulates distinct functions of NF-κB, including its association with IκBα, DNA binding, and transcriptional activation (1). Among all the acetylation sites, lysine 310 is the most well studied, as its acetylation is required for the full transcriptional potential of NF-κB and is important for modulating NF-κB-dependent inflammatory responses (13, 17, 18, 39) and for maintaining constitutive NF-κB activity in tumors (2, 6, 22). Acetylated lysine 310 enhances the transcriptional activation of NF-κB by creating a docking site for the recruitment of the bromodomain-containing factor Brd4 to activate CDK9 and the RNA polymerase II-mediated transcription of NF-κB target genes (16). The abolishment of the acetylation of lysine 310, either by mutating lysine 310 to arginine or by histone deacetylases (HDACs), significantly inhibits the transactivation of NF-κB and sensitizes cells to tumor necrosis factor alpha (TNF-α)-induced apoptosis (2, 43). Interestingly, cross talk between phosphorylation and acetylation is found for RelA. The phosphorylation of serine 536 in the transactivation domain of RelA enhances the acetylation of lysine 310 of RelA by increasing the recruitment of p300 and the dissociation of the corepressor SMRT1 and increases the overall transcriptional activity of NF-κB (3, 14, 32, 38).
In addition to acetylation, our recent studies demonstrated that RelA is also subject to lysine methylation. RelA is monomethylated by the lysine methyltransferase Set9 (also called Set7 or KMT7) at lysines 314 and 315 in vitro and in vivo in response to stimulation (40). The methylation of RelA at these two residues negatively regulates the functions of NF-κB by triggering the ubiquitination and proteasome-mediated degradation of promoter-associated RelA. The methylation of RelA serves as a “death” signal for the destruction of DNA-bound activated NF-κB (40). However, it remains unclear how methylation is regulated and whether a cross-reaction also exists for RelA methylation and acetylation, which impose opposite effects on the transcriptional activation of NF-κB.
In this study, we have explored the possible interplay between the acetylation and methylation of RelA, especially the involvement of the acetylation of lysine 310 in regulating the methylation of lysines 314 and 315. We demonstrate that the acetylation of lysine 310 of RelA inhibits the methylation of lysines 314 and 315 and the subsequent ubiquitination and degradation of promoter-associated RelA. Furthermore, we provide evidence that acetylated lysine 310 interferes with the assembly of RelA with Set9, likely through destroying direct contact between the positively charged lysine 310 and a negatively charged “exosite” of the SET domain of Set9.
Human U2OS, HEK293T, and mouse embryonic fibroblast (MEF) cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS). RelA-deficient MEFs reconstituted with wild-type (WT) RelA or RelA-K314/315R, RelA-K310R, and RelA-K310Q mutants were generated as previously described (40). T7-RelA, Flag-Set9, and Flag-Set9-H279A plasmids and a bacterial expression vector for His-Set9 were used as previously described (40). An expression vector for SIRT1 small hairpin RNA (shRNA) was kindly provided by K. Pruitt (Louisiana State University Health Sciences Center). RelA point mutations were generated by site-directed mutagenesis (Stratagene). Recombinant RelA and His-Set9 proteins were purified as previously described (40). Anti-RelA (C20 and F6), anti-Set9 (s4E5), antiubiquitin, and anti-Flag antibodies were obtained from Santa Cruz Biotechnology Inc.; Flag agarose beads (M2) and anti-SIRT1 and anti-tubulin antibodies were obtained from Sigma. Anti-Set9 (07-314) and anti-T7 antibodies were obtained from Upstate and Novagen, respectively. Anti-acetylated lysine 310 antibodies were kindly provided by W. C. Greene (3). Polyclonal antibodies against monomethylated lysine 314/315 RelA (anti-Me-K314/315) were generated by New England Peptide with a synthesized peptide corresponding to amino acids (aa) 308 to 320 of RelA (NH2-TFKSIMK[Me]K[Me]SPFSGC-COOH) as the antigen.
Peptides corresponding to amino acids 305 to 324 of RelA (TYETFKSIMKKSPFSGPTDP) or aa 308 to 320 of RelA (TFKSIMKKSPFSG), including unmodified, K314/315-monomethylated, or K310-acetylated and K314/315-monomethylated peptides, were chemically synthesized by Cambridge Research Biochemicals or New England Peptide, respectively. For autoradiography, in vitro-methylated or -acetylated peptides were boiled in 2× SDS sample buffer and separated by 15% Tris-Tricine PAGE as described previously (33). The dried gel was then exposed to X-ray film. For dot blot analysis, peptides were dissolved and diluted in distilled water to different concentrations and spotted onto a nitrocellulose membrane. The air-dried membrane was then subjected to immunoblot analysis.
The in vitro methylation of recombinant RelA and peptides by Set9 was performed as previously described (27, 40). Methylation or acetylation was visualized either by autoradiography or by immunoblotting with anti-Me-K314/315 or -Ac-K310 antibodies, respectively.
Cells were stimulated with TNF-α (20 ng/ml) for the indicated time periods, immediately washed with cold phosphate-buffered saline (PBS), and lysed as described previously (40). Anti-Me-K314/315 or -Ac-K310 antibodies were used to immunoprecipitate methylated or acetylated RelA, respectively. Methylation and acetylation levels were assessed by immunoblotting of immunoprecipitates with anti-RelA monoclonal antibodies.
U2OS cells were left unstimulated or stimulated with TNF-α (20 ng/ml) for 30 or 60 min. A chromatin immunoprecipitation (ChIP) assay using anti-Me-K314/315 RelA antibodies was performed as previously described (3), and the levels of methylated RelA on interleukin-6 (IL-6) and IL-8 promoters were assessed by quantitative real-time PCR. The sequence of ChIP primers will be provided upon request.
For RelA stability assays, 2 × 106 cells were stimulated with TNF-α (20 ng/ml) for 15 min and then chased in DMEM containing cycloheximide (CHX) (10 μg/ml) for various time periods. After treatment, cells were immediately washed with cold 1× PBS buffer, and chromatin-associated proteins were prepared as described previously (40) and immunoblotted for RelA. To assess RelA ubiquitination, 2 × 106 cells were stimulated with TNF-α for 15 min, washed with PBS, and then treated with 20 μM MG132 for 4 h. Chromatin-associated proteins were diluted with chromatin extraction buffer without NaCl into a buffer containing 200 mM NaCl and precleared with protein A beads for 30 min. Finally, the supernatant was immunoprecipitated with anti-RelA agarose beads and immunoblotted for ubiquitin.
A total of 0.5 μg unmodified or K310-acetylated RelA peptides (aa 305 to 324) was incubated with 1 μg recombinant His-tagged Set9 protein in binding buffer (50 mM Tris [pH 9.0], 150 mM NaCl, 0.5 mM dithiothreitol [DTT], 1 mM phenylmethylsulfonyl fluoride) for 30 min at 4°C. Ni-nitrilotriacetic acid (NTA) agarose beads were added for 30 min and washed three times with the binding buffer. The association of peptides with Set9 was assessed by Tris-Tricine gel electrophoresis followed by silver staining.
The crystal structures of the complexes of ERα and Set9 (36) and of p53 and Set9 (7) were used to develop a structural model for the RelA/Set9 complex. The sequence of the RelA peptide was then threaded over the structure of the peptide substrate in the complex. The obtained structure was subjected to energy minimization and structural relaxation in order to remove the overlap between the side chains of the modeled peptide and Set9. The model was then briefly simulated while constraining the side-chain nitrogen atom of lysine 310 to the position of arginine 300 in the ERα peptide. The final structure was then simulated briefly to ensure the conformational stability of the model. The energy minimization and the molecular dynamic simulations were performed using the program NAMD2 (30) and the CHARMM22 set of force field parameters (25). During all the optimization steps and the performed simulations, all the heavy atoms of Set9 as well as those in the backbone of the peptide were kept fixed in order to constrain them to the conformations observed in the crystal structure. As such, only the side chains of the modeled peptide were conformationally manipulated during these steps.
Our previous studies demonstrated that RelA is monomethylated by Set9 at lysines 314 and 315 (40). To better understand the methylation of endogenous RelA and its regulation, we generated a site-specific anti-monomethylated lysine 314/315 antibody by immunizing rabbits with a RelA peptide monomethylated at lysines 314 and 315. First, we used this antibody to determine if it detects in vitro-methylated RelA by Set9. Immunoblotting revealed a reactivity of this antibody with wild-type RelA in the presence of Set9 (Fig. (Fig.1A,1A, lane 2). However, no methylation signal was observed when lysines 314 and 315 were mutated to arginines (RelA-K314/315R) or when the catalytically inactive mutant of Set9 (Set9-H297A) was used (Fig. (Fig.1A).1A). Similarly, when we tested the reactivity of the antibody for in vivo-methylated RelA, we found that RelA was methylated on lysines 314 and 315 in the presence of wild-type Set9 but not Set9-H297A (Fig. (Fig.1B).1B). Furthermore, RelA-K314/315R displayed no reactivity with this antibody (Fig. (Fig.1B).1B). These results confirm that Set9 monomethylates RelA on lysines 314 and 315 both in vitro and in vivo.
With this antibody, we next examined the methylation of endogenous RelA. Immunoprecipitation of methylated RelA with anti-methylated lysine 314/315 RelA antibodies followed by immunoblotting with an anti-RelA antibody revealed that endogenous RelA was unmethylated in unstimulated cells, and TNF-α stimulation led to a substantial methylation of RelA (Fig. (Fig.1C).1C). The stimulus-dependent RelA methylation required endogenous Set9, as the depletion of Set9 by small interfering RNA (siRNA) significantly reduced the TNF-α-induced methylation of RelA (Fig. (Fig.1C).1C). More importantly, when the TNF-α-induced methylation of RelA was examined in RelA-deficient mouse embryonic fibroblasts (MEFs) reconstituted with either WT RelA or RelA-K314/315R, the methylation of RelA was completely abolished in RelA-K314/315R-reconstituted MEFs (Fig. (Fig.1D).1D). These results further confirm that Set9 monomethylates RelA at lysines 314 and 315 in vivo in response to TNF-α.
Since TNF-α stimulates the recruitment of Set9 to the promoters of NF-κB target genes (40), we next assessed the methylation of RelA on the promoters of NF-κB target genes by chromatin immunoprecipitation (ChIP) assays with quantitation by real-time PCR. In the absence of stimulation, there was no detectable methylated RelA on the IL-8 or IL-6 promoter (Fig. (Fig.1E).1E). However, TNF-α stimulated the methylation of RelA on both promoters at 30 and 60 min, although the levels of methylated RelA on the different promoters varied (Fig. (Fig.1E).1E). These results support the conclusion that lysines 314 and 315 of RelA are inducibly methylated following TNF-α stimulation and that the Set9-mediated methylation of RelA plays an important role in regulating nuclear NF-κB function (40).
To explore the possible interplay between the RelA acetylation of lysine 310 and the methylation of lysines 314 and 315, which are only a few amino acids apart (Fig. (Fig.2A),2A), we first examined the effect of the methylation of lysines 314 and 315 on the acetylation of lysine 310. In an in vitro acetylation assay using RelA peptides as substrates, methylated peptides were acetylated by p300 to the same extent as unmethylated peptides (Fig. (Fig.2B),2B), indicating that methylation does not affect the acetylation of lysine 310. However, when unacetylated and acetylated lysine 310 RelA peptides were used as substrates in an in vitro methylation assay, there was substantial Set9-mediated methylation of unacetylated peptides, while the methylation of acetylated peptides was barely detectable (Fig. (Fig.2C),2C), suggesting a direct inhibition of methylation by the acetylation of lysine 310.
Next, we investigated whether the acetylation of lysine 310 inhibits the methylation of RelA in the context of the full-length protein. Consistent with the data for the RelA peptides, the prior acetylation of full-length RelA at lysine 310 by p300 inhibited the methylation of lysines 314 and 315, visualized either by tritium labeling or by immunoblotting with anti-methylated lysine 314/315 RelA antibodies (Fig. (Fig.2D).2D). However, the prior methylation of full-length RelA at lysines 314 and 315 by Set9 did not affect the acetylation of lysine 310 (Fig. (Fig.2E2E).
To exclude the possibility that the reduced methylation of acetylated RelA (Fig. (Fig.2D,2D, lane 3) is due to the inability of the antibody to access methylated lysines 314 and 315 when lysine 310 is concomitantly acetylated, we synthesized various RelA peptides, including an unmodified peptide, one with only lysines 314 and 315 methylated, and one with both lysines 314 and 315 methylated and lysine 310 acetylated, and immunoblotted them with anti-acetylated or anti-methylated antibodies. Anti-methylated lysine 314/315 RelA antibodies were able to recognize methylated lysines 314 and 315 regardless of the acetylation status of lysine 310 (Fig. (Fig.2F,2F, top), indicating that the acetylation of lysine 310 does not block the access of the antibodies to methylated lysines 314 and 315. Anti-acetylated lysine 310 RelA antibodies were also able to recognize acetylated lysine 310 when lysines 314 and 315 were methylated (Fig. (Fig.2F,2F, middle). Together, these data demonstrate a unidirectional regulation; namely, the acetylation of lysine 310 inhibits the methylation of RelA on lysines 314 and 315, but methylation has no effect on the acetylation of lysine 310.
To evaluate the role of acetylation on methylation in vivo, we first examined the effect of p300 on the methylation of RelA. When wild-type p300 or its histone acetyltransferase (HAT) domain deletion mutant (p300-ΔHAT) was transfected into HEK293T cells and TNF-α-induced acetylation and methylation were measured, we found that the level of TNF-α-induced acetylation of lysine 310 was enhanced in the presence of wild-type p300 but was reduced in the presence of p300-ΔHAT (Fig. (Fig.3A).3A). Conversely, the level of TNF-α-induced methylation of lysines 314 and 315 was decreased in the presence of p300 but increased in the presence of p300-ΔHAT (Fig. (Fig.3A).3A). These data indicate that lysine 310 acetylation of RelA might inhibit the methylation of lysines 314 and 315 in vivo.
Since SIRT1 specifically deacetylates lysine 310 of RelA (43), we next determined the effect of the SIRT1-mediated deacetylation of lysine 310 on the methylation of RelA. The TNF-α-induced acetylation of lysine 310 was impaired when SIRT1 was expressed in HEK293T cells (Fig. (Fig.3B),3B), confirming that SIRT1 deacetylates lysine 310 in vivo. Consistent with the notion that the acetylation of lysine 310 inhibits the methylation of lysines 314 and 315, the TNF-α-induced methylation of RelA at lysines 314 and 315 was enhanced (Fig. (Fig.3B).3B). In contrast, when the enzymatically inactive form of SIRT1 (H363Y), which likely functions as a dominant negative to inhibit the activity of endogenous SIRT1, was expressed in the cells, the TNF-α-induced acetylation of lysine 310 was enhanced, and as a result, the level of methylation of lysines 314 and 315 was decreased (Fig. (Fig.3B).3B). To confirm the effect of SIRT1 on RelA methylation, we knocked down the expression of SIRT1 in HEK293T cells with an shRNA against SIRT1 (31). The depletion of SIRT1 by shRNA enhanced the TNF-α-induced acetylation of lysine 310 (Fig. (Fig.3C)3C) and consistently decreased the methylation of lysines 314 and 315 (Fig. (Fig.3C).3C). Collectively, these data indicate that the acetylation of lysine 310 interferes with the methylation of lysines 314 and 315 in vivo.
To explore the potential functional consequence of the interference, we examined the transcriptional activity of NF-κB in SIRT1-knocked-down cells using a κB-luciferase reporter assay. Compared to cells transfected with control shRNA, cells transfected with SIRT1 shRNA displayed a higher level of TNF-α-induced NF-κB activity (Fig. (Fig.3D),3D), suggesting that interfering with the methylation of lysines 314 and 315 by acetylated lysine 310 might enhance the transcriptional potential of NF-κB.
To assess whether the effects from SIRT1 and p300 are derived directly from the acetylation of lysine 310, we generated RelA mutants with lysine 310 replaced with arginine or with the acetyl-mimetic glutamine (designated RelA-K310R and RelA-K310Q, respectively). When the Set9-mediated methylations of these different RelA mutants were examined in vivo, we found that RelA-K310Q displayed a much lower methylation signal, whereas RelA-K310R displayed a higher methylation level (Fig. (Fig.4A),4A), consistent with the findings that the acetylation of lysine 310 of RelA inhibits the methylation of lysines 314 and 315 (Fig. (Fig.22 and and3).3). Interestingly, the methylation levels of RelA mutants correlated with their abilities to interact with Set9. While RelA-K310Q reduced its binding to Set9, RelA-K310R increased its binding to Set9 (Fig. (Fig.4A),4A), indicating that the acetylation status of RelA might regulate its association with Set9.
To investigate the effect of different lysine 310 mutations on the TNF-α-induced methylation of RelA in vivo, we tested the methylation of RelA in RelA-deficient MEFs reconstituted with these different RelA mutants. TNF-α stimulated the methylation of RelA in WT RelA-reconstituted MEFs (Fig. (Fig.4B).4B). However, RelA methylation was significantly impaired in RelA-K310Q-reconstituted MEFs (Fig. (Fig.4B)4B) but enhanced in RelA-K310R-reconstituted MEFs. These data further support the conclusion that the acetylation of lysine 310 inhibits the methylation of lysines 314 and 315 under physiological conditions in response to a stimulus.
Since the transcriptional potential of NF-κB appears to be increased when the acetylation of RelA interferes with its methylation (Fig. (Fig.3D),3D), we next examined TNF-α-induced NF-κB activation in various reconstituted MEFs. Similar to the findings shown in Fig. Fig.3D,3D, the transcriptional activity of each mutant inversely correlated with its methylation level. The TNF-α-stimulated activation of NF-κB was enhanced in RelA-K310Q-reconstituted MEFs (Fig. (Fig.4C),4C), which displayed lower levels of TNF-α-induced RelA methylation (Fig. (Fig.4B).4B). However, in RelA-K310R-reconstituted MEFs, which displayed higher levels of TNF-α-induced RelA methylation (Fig. (Fig.4B),4B), the level of TNF-α-induced activation of NF-κB was decreased (Fig. (Fig.4C).4C). These data, together with the findings shown in Fig. Fig.3D,3D, suggest that the inhibition of the methylation of RelA by acetylation enhances the transcriptional potential of NF-κB.
Our previous studies demonstrated that the methylation of RelA by Set9 triggers the ubiquitination and degradation of chromatin-associated RelA (40). We next determined whether an enhanced acetylation of RelA might decrease the ubiquitination of RelA and therefore enhance its stability. To explore this possibility, we first examined the ubiquitination of RelA in SIRT1-deficient MEFs. TNF-α stimulated the ubiquitination of chromatin-associated RelA in WT MEFs (Fig. (Fig.5A).5A). However, in SIRT1-deficient MEFs, the ubiquitination of RelA was reduced (Fig. (Fig.5A,5A, lane 2). Furthermore, when the TNF-α-induced ubiquitination of RelA in RelA-K310Q- or RelA-K310R-reconstituted MEFs was examined, we observed that the ubiquitination of RelA was remarkably reduced in RelA-K310Q-reconstituted MEFs but was slightly enhanced in RelA-K310R-reconstituted MEFs (Fig. (Fig.5B5B).
Next, we investigated the stability of chromatin-associated RelA in SIRT1-deficient MEFs. SIRT1-deficient MEFs were first pulse stimulated with TNF-α for 15 min, and cycloheximide (CHX) was then added to the cells to prevent the resynthesis of IκBα, since resynthesized IκBα is able to bind NF-κB and remove it from DNA (12). When WT MEFs were pulse stimulated with TNF-α and the level of chromatin-associated RelA was examined, we found that chromatin-associated RelA was gradually degraded (Fig. (Fig.5C).5C). In contrast, chromatin-associated RelA was more stable in SIRT1-deficient MEFs. These data are consistent with the finding that the inhibition of SIRT1 reduces the ubiquitination of chromatin-associated RelA (Fig. (Fig.5A5A).
We also examined the stability of various RelA lysine 310 mutants in reconstituted MEFs. Correlating with their ubiquitination levels, the stability of chromatin-bound RelA-K310Q was enhanced after pulse stimulation with TNF-α in the presence of CHX, and the stability of RelA-K310R was slightly reduced (Fig. (Fig.5D).5D). Together, these data demonstrate that the acetylation of lysine 310 interferes with the methylation of lysines 314 and 315, increasing the stability of RelA by reducing its ubiquitination.
The decreased or increased interaction of RelA-K310Q or RelA-K310R, respectively, with Set9 (Fig. (Fig.4A)4A) prompted us to investigate whether the acetylation status of lysine 310 regulated the assembly of the complex of RelA and Set9. First, we examined the binding of Set9 to the unacetylated or acetylated lysine 310 RelA peptides. For these assays, recombinant His-tagged Set9 was incubated with RelA peptides with or without acetylated lysine 310. When Set9-associated peptides were visualized by silver staining, significant amounts of unmodified RelA peptides were pulled down by Set9 (Fig. (Fig.6A),6A), whereas few acetylated RelA peptides were pulled down by Set9. These findings further support the notion that the acetylation of lysine 310 impairs the interaction of RelA with Set9.
The effect of the acetylation of lysine 310 on the recruitment of Set9 was also evaluated with coimmunoprecipitation experiments with RelA-deficient MEFs reconstituted with WT RelA, RelA-K310Q, or RelA-K310R. After treatment with TNF-α, RelA interacted with Set9 in WT RelA-reconstituted MEFs (Fig. (Fig.6B).6B). However, the TNF-α-induced association between RelA and Set9 was reduced in RelA-K310Q-reconstituted MEFs and increased in RelA-K310R-reconstituted MEFs (Fig. (Fig.6B).6B). Taken together, these results support a model in which the acetylation of lysine 310 interferes with the effective assembly of RelA with Set9.
To understand the detailed molecular mechanism underlying the inhibitory effect of the acetylation of lysine 310 on the assembly of RelA with Set9, we employed computational modeling to examine the interaction between RelA and Set9. Several crystal structures of the SET domain of Set9 in complex with its substrates have been determined thus far (7, 36). Taking advantage of these existing structures (the complex of ERα and Set9 and that of p53 and Set9), we developed a molecular model for Set9 in complex with the RelA peptide (Fig. (Fig.6C).6C). In this model, as expected, the methylation site (lysine 314 or 315) in the peptide is in close contact with the active site of Set9 (Fig. (Fig.6C).6C). In addition to the methylation site, we identified a close contact between the enzyme and the RelA substrate involving lysine 310. The contact at this site is furnished by a salt bridge interaction between the basic side chain of lysine 310 and a clamp-like region, which is rich in acidic (negative) residues, including aspartic acids 256, 259, and 338 and glutamic acid 348 (shown as red patches in Fig. Fig.6C).6C). This negatively charged clamp likely acts as an “exosite” in the enzyme, increasing the binding affinity of the substrate. According to this model, any modification of the positive charge carried by lysine 310 in RelA will have an effect on its binding to Set9 and, thus, on its methylation. This model successfully accounts for the inhibitory effect of the acetylation of lysine 310 on methylation, as the main effect of acetylation is the neutralization of the lysine side chain. Lysine 310 is the only positively charged amino acid available in RelA in the region close to the exosite in the enzyme, so the replacement of lysine 310 with glutamine has a similar neutralization effect. Conversely, not only is an arginine side chain replacing lysine 310 capable of establishing similar salt bridge interactions with the exosite, it likely does so more extensively due to its larger terminal group. Since the clamp-like exosite region in the enzyme is large enough to accommodate either lysine or arginine side chains, steric hindrance is not a concern. Therefore, acetylation of lysine 310 is predicted to interfere with RelA's association with the exosite of Set9, thus resulting in the decreased methylation of RelA.
In this study, we demonstrate a functional interplay between RelA acetylation and RelA methylation. By reducing the binding of RelA to Set9, the acetylation of RelA on lysine 310 inhibits the methylation of lysines 314 and 315 and the subsequent methylation-triggered ubiquitination and degradation of chromatin-associated RelA (Fig. (Fig.7).7). By modeling the structure of the SET domain of Set9 in complex with a RelA peptide, we also provide a mechanistic explanation for this interplay. Our studies reveal new insights into the regulation of NF-κB function through cross talk between acetylation and methylation.
Supporting the idea that RelA undergoes stimulus-coupled lysine methylation (40), we used site-specific anti-methylated lysine 314/315 RelA antibodies and found that endogenous RelA was methylated by Set9 at lysines 314 and 315 in response to TNF-α (Fig. (Fig.1).1). Furthermore, promoter-associated RelA was effectively methylated at lysines 314 and 315 in vivo in a stimulus-coupled manner (Fig. (Fig.1).1). Although the exact kinetics and stoichiometry of methylation are not precisely clear, the finding of a stimulus-coupled methylation of endogenous RelA in various cell lines, including U2OS cells (Fig. (Fig.1C),1C), MEFs (Fig. (Fig.1D),1D), HEK293T cells (Fig. (Fig.3),3), and macrophages (data not shown), strengthens the notion that the methylation of lysines 314 and 315 plays a role in regulating the function of NF-κB (40). Also note that the level of TNF-α-induced methylation of RelA decreased at later time points (Fig. 1C and E). This decreased methylation, similar to other posttranslational modifications including phosphorylation and acetylation, might reflect oscillatory NF-κB activation, in which the amount of nuclear NF-κB decreases due to the resynthesis of IκBα (15). It is also possible that an unidentified demethylase contributes to the decreased signal as well, since methylation, like phosphorylation and acetylation, is also a reversible event (34).
The acetylation of lysine 310 is required for the full transcriptional activation of NF-κB (2). Consistently, the depletion of SIRT1 by shRNA or the replacement of lysine 310 with acetyl-mimetic glutamine enhances the transcriptional activity of NF-κB (Fig. (Fig.3D3D and and4C).4C). Interfering with methylation and enhancing the stability of RelA would certainly account for the acetylation-dependent transcriptional activation of NF-κB. Notably, the acetylation of lysine 310 also creates a docking site for the binding of the bromodomains of Brd4, which further recruits and activates CDK9 for the RNA polymerase II-mediated transcription of NF-κB target genes (16). Therefore, the mechanistic impact of the acetylation of lysine 310 is multifaceted. The ubiquitination and degradation of promoter-associated RelA play an important role in the termination of NF-κB signaling (26), and the methylation of lysines 314 and 315 by Set9 is critical for triggering this process, since the methylation of RelA by Set9 enhances the ubiquitination of RelA (40) and the depletion of Set9 decreases the TNF-α-induced ubiquitination of chromatin-associated RelA (data not shown). The acetylation of lysine 310 inhibits the methylation of lysines 314 and 315 (Fig. (Fig.2,2, ,3,3, and and4);4); therefore, acetylated RelA needs to be deacetylated by an HDAC in order for methylation to occur. The deacetylation of lysine 310 by SIRT1 enhances the methylation of RelA and promotes the ubiquitination and degradation of RelA (Fig. (Fig.33 and and5).5). Conversely, the inactivation of SIRT1 inhibits methylation and stabilizes RelA (Fig. (Fig.3,3, ,5,5, and and7).7). It appears that the deacetylation of RelA lysine 310 by SIRT1 or by other deacetylases functions as a cellular signal to determine the “on” or “off” state of the transcriptional activity of NF-κB. However, the exact signal triggering the activation of HDACs and the deacetylation of lysine 310 remains unclear.
It is well known that, on histones, one specific posttranslational modification can determine the existence of another modification in a cis or trans manner either by creating an altered protein surface or by recruiting an effector protein (10, 42, 44). Such a cross-reaction between different modifications can also be observed for nonhistone proteins (42). Here we show that the acetylation of lysine 310 of RelA clearly inhibits the methylation of lysines 314 and 315 in vitro and in vivo (Fig. (Fig.2,2, ,3,3, and and4).4). Interestingly, peptides with acetylated lysine 310 are poor substrates for methylation (Fig. (Fig.2C),2C), suggesting that acetylation alone, without the recruitment of any effector proteins, is sufficient to inhibit methylation. In fact, acetylation neutralizes the positive charge of the side chain of the lysine residue, which is required for its interaction with Set9, thereby inhibiting the Set9-mediated methylation of RelA (Fig. (Fig.6).6). The interplay between lysine acetylation and methylation has also been observed for p53 (19, 21). Different from the interplay of RelA, p53 acetylation does not regulate its methylation. Instead, methylation enhances p53 acetylation, and the regulation appears to be an indirect effect. The Set9-mediated methylation of p53 (at human K372 or mouse K369) serves as a docking site to recruit Tip60 and/or other acetyltransferases for the subsequent acetylation of p53 (19, 21). The unique regulation for RelA and p53 suggests that acetylation and methylation cross-react in distinct manners to regulate the functions of different proteins.
Although the acetylation of lysine 310 inhibits the methylation of lysines 314 and 315 (Fig. (Fig.2,2, ,3,3, and and4),4), both modifications could be detected at the same time point after TNF-α stimulation (Fig. (Fig.3).3). The detectable acetylated or methylated signals likely represent signals from different RelA molecules, since it is clear from our in vitro experiments that the methylation of lysines 314 and 315 is inhibited when all the RelA molecules are acetylated at lysine 310 (Fig. (Fig.2C).2C). The interplay between RelA acetylation and methylation is unidirectional, since a prior methylation of lysines 314 and 315 has no effect on the acetylation of lysine 310 (Fig. (Fig.2)2) and the depletion of Set9 does not affect the TNF-α-induced acetylation of lysine 310 (data not shown). In addition to the inhibition of RelA methylation by acetylation, the phosphorylation of RelA at serines 276 and 536 was previously shown to enhance the acetylation of lysine 310 by increasing RelA binding to p300 (3, 14). These different modifications appear to form a regulatory cascade, with one modification affecting the next. Note that acetylation and methylation occur within a small region (aa 310 to 315) (Fig. (Fig.2A)2A) where other modifications also occur. For example, serine 311 is phosphorylated by PKCζ, which enhances transcription activation by NF-κB (9). Whether there is also a cross talk between phosphorylation and acetylation or phosphorylation and methylation remains to be determined.
Examination of the crystal structures of Set9 in complex with its substrates and our structural model of Set9 in complex with RelA reveals that an exosite exists for the SET domain of Set9 and that this exosite region is consistently bound to a basic residue from the substrates in all the structure models (Fig. (Fig.6C)6C) (7, 36). In the crystal structures of ERα/Set9 and p53/Set9 (7, 36), the basic residues engaged with the exosite region (arginine 300 in ERα and lysine 370 in p53) are separated from the methylation site by only one residue. RelA lacks a basic residue at this position (position 312); however, it can establish a salt bridge interaction with the exosite region through lysine 310, which can bind to exactly the same site as the basic residues in p53 and ERα (Fig. (Fig.6C).6C). We note that basic side chains (lysines and arginines) are very long and can easily reach the exosite region by slight variations in their orientation despite their slightly different positions in these substrates. As such, any basic residue in the region corresponding to residues 310 to 312 in RelA would be sufficient to interact with the proposed exosite region, resulting in the increased binding affinity and methylation level of the substrate. In contrast, the neutralization of the positive charge of lysine 310 via acetylation or mutation to glutamine is expected to reduce the binding affinity and, thus, the rate of methylation, in accordance with our experimental results (Fig. (Fig.6B).6B). Note that in our model (Fig. (Fig.6C),6C), only the configuration in which lysine 314 is inserted into the methylation active site has been captured, and this model is the only structural model that we have to work with in order to investigate the interaction between the peptide and the enzyme, as we do not have any structural models for the configuration in which the other lysine residue (lysine 315 in this case) is being methylated. Future crystal structures of Set9 in complex with two methylated lysines would provide insight into the mode of interaction involved in the methylation of lysine 315 as well.
Previous studies by Couture et al. suggested that Set9 recognizes a consensus motif of its substrates for methylation: K/R-π-K-X (where π is a small residue, K is the methylation site, and X is any residue) (8). This motif corresponds with the P−2 to P+1 positions of the methylation region in each substrate, and the basic residue in the P−2 position is required for methylation by Set9 (8). Our studies demonstrate that the lysine (residue 310) in the P−4 position in RelA plays a role similar to that of K/R in the P−2 position in the proposed consensus motif (Fig. (Fig.6D),6D), suggesting that the positively charged residue is not necessarily limited to the P−2 position. In fact, K/R in any of the P−2 to P−4 position could meet the requirement of a basic residue for Set9 recognition. This extended consensus motif can accommodate all known Set9 substrates and could help predict and identify new Set9 substrates (Fig. (Fig.6D6D).
RelA undergoes several posttranslational modifications (1, 29). By analogy to the histone code, a “protein code” or “NF-κB code” may also exist. First, different RelA modifications might regulate each other, as demonstrated in the current study. Second, various posttranslational modifications of RelA might create specific marks for the recruitment of different effectors to control the temporal and spatial activation of NF-κB (1, 29). Supporting this, the acetylation of lysine 310 is specifically recognized by the bromodomains of Brd4, enhancing transactivation by NF-κB (16). Similarly, methylated lysines 314 and 315 are likely recognized by some domain-containing E3 ligase for the ubiquitination and degradation of NF-κB (40, 41). Therefore, modifications of RelA by methylation or acetylation at specific sites could, in fact, dictate specific biological responses, reflecting the gain or loss of selective cofactors whose association with RelA is regulated by its state of modifications.
Overall, our data uncover a novel interplay between RelA acetylation and methylation and further support the “NF-κB code” hypothesis. Methylation is contingent upon the prior deacetylation of RelA at lysine 310 (Fig. (Fig.22 and and3).3). The acetylation of lysine 310 interferes with the interaction between RelA and the exosite of the SET domain of Set9 by neutralizing the positively charged side chain. Our studies also provide a better understanding of the regulation of methylation by acetylation. Since many nonhistone proteins are modified by acetylation as well as Set9-mediated methylation, it will be of great interest to explore whether an interplay and mechanism similar to those identified with RelA would apply to the regulation of other proteins.
We thank X. L. Lin and M. McBurney for providing the SIRT1-deficient MEFs, A. Beg for the RelA-deficient MEFs, W. C. Greene for reagents, D. J. Shapiro for critical reading of the manuscript, and members of the Chen laboratory for discussion.
This work is supported in part by ICR provided by the University of Illinois at Urbana-Champaign and a biomedical research grant from the American Lung Association.
Published ahead of print on 16 February 2010.