Lipopolysaccharide-induced methylation of HuR, an mRNA-stabilizing protein, by CARM1. Coactivator-associated arginine methyltransferase.

The RNA-binding protein HuR stabilizes labile mRNAs carrying AU-rich instability elements. This mRNA stabilization can be induced by hypoxia, lipopolysaccharide, and UV light. The mechanism by which these stimuli activate HuR is unclear and might be related to post-translational modification of this protein. Here we show that HuR can be methylated on arginine. However, HuR is not a substrate for PRMT1, the most prominent protein-arginine methyltransferase in mammalian cells, which methylates a number of heterogeneous nuclear ribonucleoproteins. Instead, HuR is specifically methylated by coactivator-associated arginine methyltransferase 1 (CARM1), a protein-arginine methyltransferase previously shown to serve as a transcriptional coactivator. By analyzing methylation of specific HuR arginine-to-lysine mutants and by sequencing radioactively methylated HuR peptides, Arg(217) was identified as the major HuR methylation site. Arg(217) is located in the hinge region between the second and third of the three HuR RNA recognition motif domains. Antibodies against a methylated HuR peptide were used to demonstrate in vivo methylation of HuR. HuR methylation increased in cells that overexpressed CARM1. Importantly, lipopolysaccharide stimulation of macrophages, which leads to HuR-mediated stabilization of tumor necrosis factor alpha mRNA in these cells, caused increased methylation of endogenous HuR. Thus, CARM1, which plays a role in transcriptional activation through histone H3 methylation, may also play a role in post-transcriptional gene regulation by methylating HuR.

Heterogeneous nuclear ribonucleoproteins (hnRNPs) are among the most commonly methylated proteins in mammalian cells, but other RNA-binding proteins are also substrates for methylation in vitro and in vivo (25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35). Protein arginine methyltransferases (PRMTs) transfer the methyl group from S-adenosylmethionine (AdoMet) to specific arginines in substrate proteins (26,36). Thus far, cDNA clones encoding six genetically distinct but related mammalian PRMTs have been isolated. PRMT1 is the major arginine methyltransferase in mammals (37) and is essential for early development of mouse embryos (38). Besides methylating hnRNPs, PRMT1 can methylate other nucleic acid-binding proteins, such as poly(A)-binding protein 2, fibrillarin, nucleolin, and histone H4 in vivo (26 -28, 37, 39 -41). Coactivator-associated arginine methyltransferase 1 (CARM1/PRMT4), which was originally discovered by its interaction with glucocorticoid receptor-interacting protein 1 in a yeast two-hybrid screen (42), methylates histone H3 (42)(43)(44) and p300/CBP (45) in vivo. Specific protein substrates have also been identified for PRMT3, JBP1 (PRMT5), and PRMT6, but no substrates have been reported yet for PRMT2 (see Refs. 46 and 47 and references therein). Thus far, arginine methylation of proteins has been implicated in the regulation of gene transcription, signal transduction, and nuclear-cytoplasmic protein transport (36,47). Recent studies have revealed changes in cellular levels of protein arginine methylation after a variety of stimuli. For example, the increased methylation of STAT1 parallels its enhanced transcription activation function in response to interferon ␣/␤ (48). CARM1-directed arginine methylation of histone H3 in the promoters of steroid hormone-responsive genes is induced by steroid hormone treatment of cells (43). CARM1 methylation also appears to inhibit cAMP-mediated gene expression, preventing the association between transcription factor CREB and cofactor CBP/p300 by methylation of the p300 domain required for CREB binding (45). Thus, it appears that, like phosphorylation and acetylation, methylation may play an important role in the transduction of extracellular signals to the transcription machinery.
Here, we present evidence that CARM1-directed methylation may also be involved in gene regulation at the post-transcriptional level. We show that HuR associates with and is specifically methylated in vitro by CARM1. Using antibodies that specifically recognize the methylated form of HuR, we demonstrate that HuR is methylated in vivo. The level of HuR methylation is altered by overexpression of CARM1 or by stimulation of macrophages with LPS, which is known to cause HuR-mediated mRNA stabilization (18). Our data suggest that methylation by CARM1 plays a role in mRNA stabilization by HuR.

EXPERIMENTAL PROCEDURES
Cell Culture-COS-7 and RAW 264.7 cells were maintained in Dulbecco's modified Eagle's medium, and Jurkat cells were maintained in RPMI 1640, each supplemented with 1% penicillin and streptomycin and 10% fetal bovine serum. LPS (Sigma) was added into the medium at 10 g/ml for 1 h to stimulate RAW 264.7 cells.
Plasmids, Peptides, and Antibody Production-Construction of plasmids pSG5-HA-CARM1 and pGEX-4T1-CARM1 has been previously reported (42). To construct plasmid pET3d-HuR for the expression of full-length recombinant human HuR protein, the coding region was amplified from pGST-HuR (3) by high fidelity PCR using oligonucleotides to engineer an NcoI-compatible BspHI site at the ATG and an NotI site immediately following the last codon of the HuR cDNA. The BspHI-NotI HuR cDNA fragment was inserted into an NcoI-NotI-digested derivative of the pET3d vector (Novagen) encoding a C-terminal hexahistidine and c-myc epitope tag (49), yielding pET3d-HuR. The HuR hinge region mutants were made by first creating a BssHII site (GCG/ CGC at alanine 204/arginine 205) and an ApaI site (GGG/CCC at glycine 209/proline 210) with silent mutations, resulting in plasmid H918. Bold indicates which nucleotide was mutated to generate the restriction site. The HuR hinge deletion mutant was then made by digestion of H918 with BssHII and NcoI (at methionine 223), blunting, and religation, resulting in deletion of amino acids Arg 206 -Pro 222 . The individual arginine-to-lysine hinge region mutants were made by digestion of H918 at preexisting sites PvuII (at glutamine 198/leucine 199) or NcoI (at methionine 223) or the engineered BssHII or ApaI sites (see above), and replacing the intervening sequences with double-stranded oligonucleotides encoding the desired mutation. This approach yielded HuR arginine-to-lysine mutants R205K, R206K, R217K, and R219K in the pET3d derivative. The eukaryotic HuR expression plasmids were created from the pET3d constructs by excising the coding region using an EcoRI site just upstream of the ATG and the NotI site just downstream of the last codon and inserting this HuR fragment into a modified vector derived from pSG5-HA (42), so that an N-terminal hemagglutinin tag and a C-terminal hexahistidine tag were fused on either end. The resulting HA-HuR protein carried the following additional amino acids: N-terminal to HuR methionine 1, MGYPYDVPDYAEF; C-terminal to HuR Lys 326 , AAAHHHHHH-stop. GST-PRMT1 (41), GST-PRMT3 (50), GST-GAR (41), and GST-hnRNP K (51) expression vectors have been previously described. GST-PRMT2 was kindly provided by Dr. Steven Clarke (UCLA). Peptide P1 (HYHSPARRFGGPVH-HQAQRFRFSPMGV-NH 2 ) (AnaSpec Inc.) encoding the HuR hinge do-main was generated for in vitro methylation and radiosequencing analysis. To raise an antibody that could specifically recognize methylated HuR (anti-me(R217)HuR antibody), methylated peptide HHQAQ(DMA)FRFSPGC (where DMA represents asymmetric dimethylarginine) was synthesized and conjugated to keyhole limpet hemocyanin before injection into rabbits. Specificity of the antiserum was confirmed by enzyme-linked immunosorbent assays with the DMAcontaining peptide and a corresponding unmethylated peptide.
Immunoprecipitation and Immunoblot-Cell lysates, prepared as described above (one 100-mm dish of cells/immunoprecipitation sample), were cleared with protein A/G beads (Santa Cruz Biotechnology, Inc.) for 1 h at 4°C. 2 g of anti-met(R217)HuR or anti-mono/dimethyl arginine (Ab412; Abcam Ltd.) were added to the cell lysates and incubated overnight at 4°C on a rotator. Protein A/G beads were added and incubated for another 3 h. Beads were washed three times with radioimmune precipitation buffer and subjected to SDS-PAGE. Blots were probed with mouse monoclonal anti-HuR (Santa Cruz Biotechnology) at 1 g/20 ml of blocking buffer (5% (w/v) nonfat milk in TBST: 150 mM NaCl, 10 mM Tris-HCl, pH 8.0, and 0.1% (v/v) Tween 20). Horseradish peroxidase-conjugated secondary antibodies (Santa Cruz Biotechnology) were used at 1 g/10 ml of blocking buffer and ECL reagents (Amersham Biosciences) were used for detection.
In Vitro Protein-Protein Interaction-GST fusion proteins were produced in Escherichia coli BL21 as described previously (52). Hexahistidine fusion proteins were purified with His-Bind Buffer Kit (Novagen) according to the manufacturer's protocol. Pull-down assays were conducted as described previously (53). CARM1 protein was translated in vitro in the presence of [ 35 S]methionine using the TNT-T7-coupled reticulocyte lysate system (Promega). The binding assay was conducted by incubating beads containing 2 g of His-HuR with slow rotation overnight at 4°C with 10 l of the translation reaction in a 500-l total volume of NETN-0.01% (100 mM NaCl, 1 mM EDTA, 20 mM Tris-HCl, pH 7.0, and 0.01% (v/v) Nonidet P-40). Beads were washed four times with NETN-0.01% and analyzed by SDS-PAGE and autofluorography.
Methylation Assay-Methyltransferases were prepared as recombinant GST fusion proteins and eluted from glutathione-agarose beads (Sigma) with 20 mM glutathione. 0.5-3 g of commercially obtained histone H3 (Roche Molecular Biochemicals) or recombinant proteins were incubated with 1 g of methyltransferase in the presence of 6 M S-adenosyl-[methyl- 3 (42). SDS-loading buffer or radioimmune precipitation buffer was used to stop the reaction. Labeled proteins were identified by SDS-PAGE and autofluorography for 12 h. To prepare the in vitro methylated samples for immunoprecipitation, nonradiolabeled AdoMet was used instead at 0.1 mM.
Radiosequencing-Approximately 5 g (1.6 nmol) of [ 3 H]methyl-HuR hinge peptide P1 was purified from a CARM1-catalyzed methylation reaction by reversed phase high performance liquid chromatography on a 2.1 ϫ 30-mm RP-300 (C8) column (Brownlee Laboratories). Elution was carried out at 1.0 ml/min using a linear gradient of 5-40% (v/v) acetonitrile in 0.1% (v/v) trifluoroacetic acid, and the peptide was detected by monitoring absorbance at 214 nm. The purified [ 3 H]methyl peptide was found to contain ϳ0.07 mol of [ 3 H]methyl/mol of peptide. Two samples (650 pmol each) of the purified peptide were sequenced for 22 cycles by automated Edman degradation on a Hewlett-Packard G1105A sequenator. The first sample was sequenced by standard procedures to verify the correct amino acid sequence and to establish the repetitive yield. For the second sample, the phenylthiohydantoin-derivatized amino acids from each cycle were collected for measurement of 3 H by liquid scintillation counting.

HuR Is Specifically Methylated by CARM1 in Vitro-Because
it is unclear how extracellular signals mediate HuR-induced stabilization of labile ARE-containing mRNAs, we sought to establish whether post-translational modification of HuR might play a role. No modifications of HuR (including phosphorylation) have been found as yet. Because arginine methylation is a common modification of RNA-binding proteins, we tested whether HuR could be methylated by one of the PRMTs. Various recombinant PRMTs were incubated with recombinant hexahistidine-tagged HuR (His-HuR) in the presence of [ 3 H]AdoMet, and the products were examined by SDS-PAGE and autofluorography. His-HuR was methylated by CARM1 but not by PRMT1, PRMT2, or PRMT3 (Fig. 1A, lanes 4 -7). Staining of the gel demonstrated that His-HuR and PRMT proteins were present in the appropriate reactions (Fig. 1B,  lanes 4 -7), and methylation of known substrates for PRMT1, PRMT3, and CARM1 (hnRNPK, GAR, and histone H3, respectively) verified the specificity and activity of these PRMTs (Fig.  1A, lane 1-3). (No known substrate for PRMT2 is currently available as a control.) Methylation of degradation products of GST-hnRNP K and His-HuR and an aggregation product of histone H3 was also observed (Fig. 1A, lanes 1, 2, and 7). CARM1 and HuR Associate in Vitro-To further test the substrate specificity of CARM1 for HuR, we investigated whether CARM1 could bind HuR directly in vitro. Either CARM1 or PRMT1 was translated in vitro in the presence of [ 35 S]methionine (Fig. 2, lane 1) and incubated with protein-free Ni 2ϩ beads, Ni 2ϩ beads to which His-HuR had been bound, glutathione-agarose beads with bound GST, or glutathioneagarose beads to which GST-hnRNP K had been bound (Fig. 2,  lanes 2-5). Strong interactions between CARM1 and His-HuR (lane 3) and between PRMT1 and hnRNP K (lane 5) were observed, whereas no association between the other combinations was detected. Thus, the substrate specificity observed in the in vitro methylation assay is supported by the selective association between the two PRMTs and their respective substrates.
The HuR Hinge Region Harbors the Methylation Site(s)-Whereas PRMT1 methylation sites are frequently located in Arg-Gly-Gly (RGG) repeats (26), neither histone H3, the first identified in vivo substrate for CARM1 (42), nor HuR contains RGG repeats. Based upon two CARM1 methylation sites in histone H3, a potential consensus motif, KAXRK, has been proposed (44). Whereas HuR does not contain KAXRK sequences, we noted that two arginines in the hinge region were preceded by alanine at the Ϫ2 position (PARRF and QAQRF) ( Fig. 3A). We therefore tested whether an HuR mutant from which a 16-amino acid tract containing the two possible methylation sites had been deleted (Fig. 3A) could still be methylated. In contrast to intact HuR, which was strongly methylated (Fig. 3B, left panel, lane 1), no in vitro methylation of this deletion mutant by CARM1 was seen (Fig. 3B, left panel, lane 2), although similar quantities of mutant and wild type HuR proteins were used as substrates (Fig. 3B, right panel). This result suggests that the hinge region of HuR contains the methylation site(s) or is important for substrate recognition by CARM1.
To determine which of the four arginines located in hinge region (Fig. 3A) might be methylated, we generated four point mutants in the context of full-length HuR, each one containing a single arginine-to-lysine substitution. Mutation of arginine 217 to lysine (R217K) completely abolished in vitro methylation of HuR by CARM1, whereas each of the other three HuR mutants (R205K, R206K, and R219K) was methylated to a level comparable with that of wild type HuR (Fig. 3C). This suggests an essential role of Arg 217 as the methylation site and/or as a residue that is important for recognition of HuR by CARM1.
We next synthesized a HuR hinge peptide, P1, containing all four arginines (Fig. 3A). CARM1 efficiently methylated the P1 peptide in vitro, whereas PRMT1 did not (Fig. 4A). Thus, the determinants that mark HuR as a CARM1 substrate are retained in the 27-amino acid P1 sequence. In order to establish directly the location of the methylation site(s), a sample of the peptide, [ 3 H]methyl-labeled by CARM1, was subjected to radiosequencing by automated Edman degradation. Approximately 25% of the 3 H was recovered in cycles 7-9, whereas the remaining 75% was recovered in cycles 18 -22 (Fig. 4B). The first peak indicates selective methylation of Arg 206 . 3

CARM1-mediated Methylation of HuR
ration, respectively. Thus, Arg 217 is the major methylation site in the HuR hinge peptide when modified by CARM1 in vitro.
HuR Is Methylated by CARM1 in Vivo-To test whether HuR is methylated in vivo, we developed antibodies directed against HuR methylated at Arg 217 (anti-me(R217)HuR). We also used an antibody that generally recognizes proteins containing mono-or dimethylarginine (anti-M/DMA), which has been used successfully to demonstrate the in vivo methylation of STAT1 (48) and hepatitis C virus NS3 protein (55). Both antibodies were characterized for discrimination between HuR methylated in vitro by CARM1 (using unlabeled AdoMet) and unmethylated HuR (Fig. 5). (The in vitro methylation was confirmed in a parallel reaction with [ 3 H]AdoMet (data not shown).) Following immunoprecipitation by antime(R217)HuR or anti-M/DMA antibodies, the precipitated HuR protein was detected by immunoblot using antiserum raised against unmodified HuR protein. Both the antime(R217)HuR and anti-M/DMA antibodies specifically immunoprecipitated methylated HuR (Fig. 5). Although no HuR was detected in immunoprecipitates obtained with appropriate isotype control antibodies (data not shown), a small amount of HuR was precipitated from the unmethylated reaction with anti-me(R217)HuR antiserum (Fig. 5B, lane 1). This weak signal might be due to low levels of antibodies directed against parts of the peptide flanking the methylation site. Nevertheless, both anti-me(R217)HuR and anti-M/DMA antibodies were highly specific for methylated versus unmethylated HuR. Neither antibody could detect the in vitro methylated HuR on immunoblots (data not shown).
In order to determine whether HuR can be methylated by CARM1 in vivo, we co-transfected COS-7 cells with expression vectors encoding HA-tagged CARM1 and either HA-HuR (Fig. 6A, lane 2) or HA-HuR(R217K) (lane 4). COS-7 cells transfected with HA-HuR alone (lane 1) or HA-HuR(R217K) alone (lane 3) were used as controls. Cell lysates were then subjected to immunoprecipitation with either anti-M/DMA antibody or anti-me(R217)HuR antibody, and immunoprecipitated HA-HuR was detected by immunoblotting using an anti-HA antibody. Whereas the expression level of HA-HuR was similar in all four samples (Fig. 6A), expression of HA-CARM1 substantially increased the amount of HA-HuR immunoprecipitated by both methyl-directed antibodies (Fig.  6, B and C, compare lanes 1 and 2) but had no effect on the amount of HA-HuR(R217K) immunoprecipitated (compare lanes 3 and 4). These results demonstrate that HuR can be methylated in vivo by CARM1 and support the notion that Arg 217 is the major methylation site in vivo as well as in vitro.
HuR Methylation Is Enhanced in RAW 264.7 Cells Treated with LPS-Having demonstrated that HuR methylation occurs in vivo, we next asked whether this post-translational modification could play any role in the regulation of gene expression by HuR. HuR can bind to AREs present in a large variety of mRNAs, including those encoding mediators of inflammation (13,17,18,23). This binding of HuR has been linked to stabilization of mRNAs encoding interleukin-3, nitric-oxide synthase II, and TNF-␣, which are induced by ionophores, 12-Otetradecanoylphorbol 13-acetate, cytokines, or LPS, depending on the cell type (13,17,18,23). Recently, interferon (IFN) signaling, another component of the inflammatory response, was shown to be mediated at least in part through increased methylation of STAT1, an IFN-activated transcription factor (48). A role for arginine methylation in IFN signaling was supported by the observation that the IFN-␣ receptor interacts with PRMT1 and that antisense PRMT1 oligonucleotides mitigated the IFN response (56). Whereas CARM1 appears to play a role in steroid hormone and cAMP-mediated signaling, there has as yet been no evidence for its involvement in the inflammatory response. However, its ability to methylate HuR, combined with the observed role of HuR in the mRNA stabilization following inflammatory stimuli in a variety of cells, suggested that CARM1 methylation might play a role in the inflamma-

CARM1-mediated Methylation of HuR
tory response as well. We therefore tested whether the in vivo methylation status of HuR might change in LPS-stimulated RAW 264.7 cells (a murine macrophage-like cell line). RAW 264.7 cells respond to LPS by strongly inducing TNF-␣ expression, mediated in part through TNF-␣ mRNA stabilization by HuR binding (18).
We first determined by immunoblot that RAW 264.7 cells contain CARM1 protein (data not shown). Then cell lysates from either untreated or LPS-treated RAW 264.7 cells were analyzed by immunoprecipitation with anti-me(R217)HuR antibody (Fig. 7A) followed by immunoblotting with anti-HuR. The total level of HuR protein in the cells was not increased by LPS treatment (Fig. 7A, input samples). Whereas HuR was barely detectable when lysates from unstimulated RAW 264.7 cells were used for immunoprecipitation, a strong HuR band became visible upon LPS stimulation (Fig. 7A, lanes 3-5). These observations suggest that methylation of HuR occurs in response to LPS stimulation. A similar observation was made when Jurkat cells, a human T cell leukemia cell line that responds to a wide variety of agents, including LPS (57), was examined for induction of HuR methylation (Fig. 7B). Since the anti-me(R217)HuR antibodies specifically recognize methylated HuR and CARM1 is the only one of the four PRMTs tested that can utilize HuR as a substrate, our results strongly suggest that CARM1 methylates HuR in response to LPS and indicate that CARM1 might play a role in LPS-mediated mRNA stabilization. DISCUSSION We have identified a new target for arginine methylation by CARM1: the RNA-binding protein HuR. HuR is methylated by   FIG. 4. Mapping of the major HuR methylation site to Arg 217 . A, 5 g of synthetic peptide P1 (Fig. 3A) or recombinant GST-hnRNP K were tested for in vitro methylation by either GST-CARM1 or GST-PRMT1 and analyzed by autofluorography (left panel) and Coomassie staining (right panel). Minor contaminants in P1 or degradation products of GST-hnRNP K were methylated as well. B, radiosequencing of in vitro methylated P1 was carried out as described under "Experimental Procedures." C, a refinement of data from B in which corrections were made for the ragged N terminus, for sequencing lag, and for a measured repetitive yield of 95%. Numbers above the bars refer to amino acid residue numbers in HuR.

CARM1-mediated Methylation of HuR
CARM1 in vitro but not by PRMT1, -2, or -3 (Fig. 1). Analysis of arginine-to-lysine mutants of HuR indicated that Arg 217 is the major methylation target site (Fig. 3). Radiosequencing of a HuR hinge peptide methylated in vitro supported this conclusion but also identified Arg 206 as a possible minor methylation site (Fig. 4). However, since Arg 206 was not methylated in full-length HuR carrying the R217K mutation (Fig. 3), methylation at Arg 206 is either dependent on the presence of Arg 217 (in the methylated or unmethylated form) or does not happen in the context of the full-length protein, perhaps because it is not accessible. Major and minor methylation sites have been observed in other CARM1 substrates (44,45,58), but their significance is presently unclear.
Whereas PRMT1 methylates many RNA-binding proteins at arginine residues in so-called RGG repeats (26,47), the sequence specificity of CARM1 is not clear at present. However, a sequence or structural context that affects specificity must exist, because Arg 205 and Arg 219 in the same peptide are not methylated (Fig. 4), and CARM1 methylated only a few proteins in a large membrane-bound array of polypeptides (58). Recently, CBP/p300 and poly(A)-binding protein 1 (PABP1) were also identified as CARM1 substrates (45,58). Comparison of all known major CARM1 target sites with those of HuR does not yield any obvious consensus, apart from the alanine at position Ϫ2, which is present in four of six previously identified major CARM1 methylation sites (Table I) and the fact that two major methylation sites are found in each protein, separated by 23 amino acids or fewer.
Since antibodies against a HuR peptide methylated at Arg 217 (Fig. 5) immunoprecipitated increased amounts of HuR, but not HuR(R217K) from COS-7 cells overexpressing CARM1 (Fig. 6), we conclude that Arg 217 is the major CARM1 methylation site in HuR, both in vitro and in vivo. To determine whether HuR methylation levels might be modulated in response to certain signals, we studied HuR methylation in RAW 264.7 cells stimulated with LPS. HuR has been shown to mediate stabilization of TNF-␣ mRNA in macrophages following LPS stimulation. The mechanism by which this occurs appears to be competition of HuR with the destabilizing protein tristetraproline for binding to the TNF-␣ mRNA ARE (18,23,59). Using our antimethyl-HuR antibody, we showed that HuR methylation was increased in RAW 264.7 cells following LPS treatment (Fig. 7). This observation suggests that methylation by CARM1 might play a role in post-transcriptional gene regulation following inflammatory stimuli.
What could the effect of HuR methylation be, and how could this lead to mRNA stabilization? Immunostaining experiments had previously suggested that HuR was mainly nuclear, but heterokaryon analyses and cell fractionation experiments have demonstrated that the protein shuttles between the nucleus and cytoplasm (8 -11). Only cytoplasmic HuR would be able to stabilize mRNA. Thus, one possible function of HuR methylation might be to raise its cytoplasmic concentration by increasing HuR export from the nucleus. Changes in intracellular localization caused by methylation have been observed for hnRNP A2 in mammalian cells and the yeast protein Np13p, which are substrates of mammalian PRMT1 and its yeast homologue RMT1, respectively. Intriguingly, methylation can have opposite effects on the intracellular location of these proteins, promoting a nuclear localization for hnRNPA2 (60) but stimulating a cytoplasmic location for Np13p (61). Methylation does not affect the charge of arginine residues but could cause steric changes and prevent hydrogen bond formation. Thus, it is assumed that changes in intracellular location caused by methylation result from altered protein/protein interactions.
Shuttling of HuR is mediated by sequences in the hinge region between RRMs 2 and 3 (amino acids 205-237 (Fig. 3), referred to as the HuR nucleocytoplasmic shuttling sequence, or HNS (8)). Interestingly, Arg 206 and Arg 217 are located within  (44); PABP1 (58), p300 (45), and HuR (3) from man. The methylated arginine is indicated in bold, the alanine at Ϫ2 is highlighted in gray, and residues flanking the arginines, which are identical within one protein, are underlined.  1 and 4) or were treated with LPS at 10 g/ml in cell culture medium for 1 h (lanes 2, 3, and 5). Cell extracts were incubated overnight at 4°C with 1 g of either normal mouse IgG or anti-me(R217)HuR antibodies as indicated, resolved by SDS-PAGE, and probed with anti-HuR. 5% of each cell extract used for immunoprecipitation was loaded for comparison (input lanes; a shorter exposure confirms that signal in both input lanes is very similar). B, Jurkat cells were untreated or treated with LPS at 10 g/ml in cell culture medium for 1 h. Cell extracts were incubated overnight at 4°C with 1 g of anti-me(R217)HuR antibodies, resolved by SDS-PAGE, and probed with anti-HuR. 5% of each cell extract used for immunoprecipitation was loaded for comparison (input lanes).
the HNS. Shuttling of HuR is thought to be influenced by proteins that interact with the HNS (such as pp32 and APRIL), thereby linking HuR to one or more nuclear export pathways (16,62,63). Methylation of the hinge domain could modify the interaction between HuR and the export machinery. In this respect, it is of great interest that the HuR export pathway appears to change following exposure of cells to stress, which increases export of HuR by the CRM1 pathway (62)(63)(64). The molecular mechanism for this switch has not yet been identified and may be related to methylation.
Whereas it is attractive to propose that methylation might increase nuclear export, an increase in cytoplasmic RNA-bound HuR could also be explained by alternative mechanisms. For example, methylation could prevent reimport of cytoplasmic HuR into the nucleus, lead to stabilization of cytoplasmic HuR protein, or alter HuR RNA-binding properties to favor binding to AU-rich sequences, thereby indirectly contributing to cytoplasmic accumulation of HuR. Whether methylation and shuttling are directly linked may be difficult to determine; distinguishing methylation effects from shuttling effects will be challenging, because the Arg 217 methylation site is located in the shuttling sequence.
Whether methylation of proteins can affect their nucleic acid-binding properties remains unclear. Methylation of hnRNP A1 appears to have a moderate effect on its ability to bind to homopolymers (65). The RGG domains methylated in many RNA-binding proteins can participate in RNA binding, and their methylation may therefore influence the RNA/protein interaction (66). In the case of HuR, Arg 217 lies outside the three RNA-binding domains, and a prominent role for the hinge region in RNA binding seems unlikely. Nevertheless, methylation could indirectly influence RNA binding, by affecting the interaction with hinge-binding proteins, which could in turn modulate RNA binding, perhaps by controlling the access of RRM2 and/or RRM3 to RNA (67).
One of the most intriguing questions that remains unanswered is whether methylation is dynamic. To date, no demethylase has been identified, and if such an enzyme does not exist, a signal achieved through increased methylation must be silenced by protein degradation or sequestration of the methylated protein. It has also been suggested that methylation could be an incremental signal, accumulating slowly on proteins that carry multiple methylation sites (36). This is not likely to be the case for HuR, since the response to LPS is rapid (within 1 h) and there appears to be only one or two methylation sites.
In conclusion, CARM1, which was previously identified as a transcriptional coactivator for nuclear receptors (42,43), now appears to extend its role into the post-transcriptional domain of gene regulation. The potential for CARM1 involvement in post-transcriptional regulation of genes in response to a variety of stimuli is considerable, since one of its targets (HuR) not only mediates the stabilization of TNF-␣ mRNA upon stimulation with LPS (18) but also the stabilization of p21 mRNA upon exposure of cells to UV light (19), of vascular endothelial growth factor mRNA by hypoxia (15), and of nitric-oxide synthase II mRNA upon cytokine stimulation (13). How broad the regulatory network modulated by CARM1 might be will depend on how many target proteins it methylates in vivo. Thus far, CARM1 appears to have markedly fewer substrates than PRMT1 (36,47,58), but its recently demonstrated ability to methylate poly(A)-binding protein 1 (58) further supports its potential role in post-transcriptional gene regulation. Our results suggest that methylation by CARM1 might create a link between transcriptional and post-transcriptional events, affecting RNA synthesis as well as RNA transport and/or stability, and that methylation of the shuttling RNA-binding protein HuR may be one way in which nuclear and cytoplasmic regulation is coupled.