Prmt5, which forms distinct homo-oligomers, is a member of the protein-arginine methyltransferase family.

We found that JBP1, known as a human homolog (Skb1Hs) of Skb1 of fission yeast, interacts with NS3 of the hepatitis C virus in a yeast two-hybrid screen. Amino acid sequence analysis revealed that Skb1Hs/JBP1 contains conserved motifs of S-adenosyl-l-methionine-dependent protein-arginine methyltransferases (PRMTs). Here, we demonstrate that Skb1Hs/JBP1, named PRMT5, is a distinct member of the PRMT family. Recombinant PRMT5 protein purified from human cells methylated myelin basic protein, histone, and the amino terminus of fibrillarin fused to glutathione S-transferase. Myelin basic protein methylated by PRMT5 contained monomethylated and dimethylated arginine residues. Recombinant glutathione S-transferase-PRMT5 protein expressed in Escherichia coli also contained the catalytic activity. Sedimentation analysis of purified PRMT5 on a sucrose density gradient indicated that PRMT5 formed distinct homo-oligomeric complexes, including a dimer and tetramer, that comigrated with the enzyme activity. The PRMT5 homo-oligomers were dissociated into a monomer in the presence of a reducing agent, whereas a monomer, dimer, and multimer were detected in the absence or at low concentrations of a reducing agent. The results indicate that both covalent linkage by a disulfide bond and noncovalent association are involved in the formation of PRMT5 homo-oligomers. Western blot analysis of sedimentation fractions suggests that endogenous PRMT5 is present as a homo-oligomer in a 293T cell extract. PRMT5 appears to have lower specific enzyme activity than PRMT1. Although PRMT1 is known to be mainly located in the nucleus, human PRMT5 is predominantly localized in the cytoplasm.

Protein arginine methylation is an irreversible, post-translational covalent modification. Protein-arginine methyltransferases (PRMTs) 1 transfer the methyl group from S-adenosyl-L-methionine to the guanidino nitrogen atoms of an arginine residue (1). PRMTs are classified into two major types, I and II, based on substrate and reaction product specificity. Both type I and II PRMTs are common in the formation of monomethylarginine, but the two differ in that type I PRMT catalyzes asymmetric dimethylarginine, whereas type II PRMT produces symmetric dimethylarginine. Type I PRMTs methylate arginines in the Arg-Gly-Gly-rich region, known as the RGG motif, present in many RNA-binding proteins (2)(3)(4), or in the Arg-Xaa-Arg motif in poly(A)-binding protein II (5). Myelin basic protein (MBP) and the spliceosomal D1 and D3 proteins are the only known in vivo substrates for type II PRMT (6 -8). However, the classification of type I and II protein-arginine methyltransferases is tenuous because they are based only on substrate and reaction product specificity (1).
Although five different kinds of genes for protein-arginine methyltransferases PRMT1, HRMT1L1 (human arginine methyltransferase-1 L1)/PRMT2, PRMT3, CARM1 (coactivator-associated arginine methyltransferase-1)/PRMT4, and Skb1Hs (Shk1 kinase-binding protein-1 Homo sapiens)/JBP1 (named PRMT5) have been cloned in mammalian cells, proteinarginine methyltransferase activities of the gene products are demonstrated in only PRMT1, PRMT3, and CARM1/PRMT4 (9 -11). The gene for rat PRMT1 is the first mammalian gene cloned 30 years after the discovery of protein arginine methylation (12,13). PRMT1 interacting with the mammalian immediate-early protein (TIS21/PC3), known as BTG2 (14,15), is a predominant protein-arginine methyltransferase in mammalian cells and tissues (16,17). Subsequently, human PRMT1, which is almost identical to rat PRMT1, was found to be associated with the intracytoplasmic domain of the interferon-␣/␤ receptor (18). The gene for PRMT2 was found by screening expressed sequence tag data bases (19,20), but its protein methyltransferase activity has not been detected. PRMT3 containing a zinc finger domain in its amino terminus (21) was found by conducting a yeast two-hybrid screen using rat PRMT1 as a bait (10). CARM1/PRMT4 regulates transcription as an interacting molecule with GRIP1 (glucocorticoid receptorinteracting protein-1), a p160 family of transcriptional coactivators (22). Pollack et al. (23) found that Skb1Hs (24,25), named JBP1 (Janus kinase-binding protein-1), interacts with Janus kinases and contains the protein methyltransferase activity. However, it was not clearly determined whether the protein methylation activity of JBP1 is arginine-specific. Any biological consequences of an interaction between Janus kinases and JBP1 are unknown.
Independently, we found that NS3 (nonstructural protein-3) of the hepatitis C virus (HCV) interacts with an Skb1Hs protein, a human homolog of Skb1 (Shk1 kinase-binding protein-1) of fission yeast, in a yeast two-hybrid screen. HCV causes acute and chronic liver diseases such as liver cirrhosis and hepato-cellular carcinoma (26,27). The NS3 protein of HCV not only contains serine protease (28 -31) and RNA helicase (32,33) activities, both of which appear to be essential for the virus replication, but also has been implicated in cellular transformation (34,35). For example, NIH3T3 mouse fibroblasts transfected with the N-terminal domain of NS3 become transformed and are tumorigenic in nude mice (34). In fission yeast, Skb1 interacts with the Shk1 kinase, which is a yeast homolog of human p21 Cdc42/Rac1 -activated kinase (24). Skb1 has been suggested to regulate mitosis negatively and can be functionally replaced with its human homolog (Skb1Hs) in fission yeast (25). In this context, an interaction between Skb1Hs and NS3 may play a role in liver diseases caused by HCV. To investigate a function of Skb1Hs interacting with the viral protein, its amino acid sequence was compared with those of genes registered in the GenBank TM /EBI Data Bank. The comparison revealed that the C-terminal domain of Skb1Hs contains an extensive homology to a family of proteins with arginine-specific protein methyltransferase activity.
In this study, we focused on the biochemical properties of PRMT5 and found that the arginine residue present in MBP is methylated by PRMT5 and that homo-oligomerization is important for the catalytic activity. A homomeric complex of PRMT5 was detected in vivo by co-immunoprecipitation analysis. The homomeric complex of PRMT5 could be separated into a dimer and multimer by sucrose gradient sedimentation. Purified PRMT1 also forms homo-oligomers. PRMT5 forms distinct homo-oligomeric complexes different from those formed by PRMT1, but both covalent and noncovalent associations are involved in the homo-oligomerization of PRMT5 and PRMT1. The homo-oligomeric complexes of PRMT5 and PRMT1, both of which methylate MBP in vitro, may account for the controversial polypeptide compositions of protein-arginine methyltransferases previously purified from cells and tissues (3,36,37).

EXPERIMENTAL PROCEDURES
Cell Cultures-293T, COS-1, and Chang liver monolayer cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin (Life Technologies, Inc.) in an atmosphere of 5% CO 2 and air humidified at 37°C.
Cloning of Human PRMT5-A yeast two-hybrid screen was conducted using a HeLa cell cDNA library and a bait plasmid that encodes the LexA DNA-binding domain fused to the N-terminal domain (amino acids 1027-1297) of HCV NS3. The isolation of positive clones and subsequent analysis were carried out as previously reported (38). One of the positive clones, named clone 4-7, encodes a partial region of human PRMT5 cDNA. To obtain full-length 4-7, primers and a HepG2 cell cDNA library were used in polymerase chain reactions. A partial clone missing ϳ200 base pairs from the N terminus of clone 4-7 was obtained and sequenced. The missing N-terminal portion was cloned by reverse transcription-polymerase chain reaction of total mRNAs from HepG2 cells. The full-length PRMT5 cDNA was inserted into the HincII site of pBluescript KS (Stratagene) and sequenced, yielding pBS-4-7F.
Electrophoretic Analysis of Proteins-Proteins or methylated proteins were boiled in SDS-PAGE sample buffer at 100°C for 5 min (except where indicated) and separated on slab gels prepared from 29% (w/v) acrylamide and 1% (w/v) N,N-methylenebisacrylamide (1.5 mm ϫ 5.5 cm-resolving gel) using the buffer system described by Laemmli (39) at a constant current of 50 mA for ϳ2 h. Following electrophoresis, gels were fixed with a fixing solution (45% (v/v) methanol and 10% (v/v) acetic acid) and then soaked in Amplify (Amersham Pharmacia Biotech) according to the manufacturer's instructions. Gels were dried in a vacuum, and radioactivity was visualized by exposing the gels to an x-ray film at Ϫ80°C for 3-14 days or by a Fuji BAS1000 PhosphorImager. 14 C incorporation was quantitated using a PhosphorImager. Gels were stained with Coomassie Brilliant Blue R-250 for 20 -30 min and destained with the fixing solution to visualize the protein bands. Proteins on the gels were transferred to polyvinylidene difluoride membranes (PerkinElmer Life Sciences). Membranes were blocked with Tris-buffered saline (50 mM Tris-HCl (pH 7.4) and 150 mM NaCl) containing 5% skim milk and then incubated with mouse anti-FLAG or anti-GST antibodies at a concentration of 5 g/ml. Membranes were washed three times with Tris-buffered saline and incubated with goat anti-mouse IgG conjugated to horseradish peroxidase (Sigma) at 1:5000 dilution. After being washed three times, the reactive proteins were detected by enhanced chemiluminescence (ECL, Amersham Pharmacia Biotech).
Analysis of Methylated Arginines-The enzyme assay was done in a 500-l reaction volume containing 80 g of MBP, 10 g of purified FLAG-PRMT5, and 1 Ci of [ 14 C]AdoMet. MBP was then precipitated with 500 l of 25% (w/v) trichloroacetic acid. The precipitate was washed with acetone twice and then dissolved in distilled water. The solution was dried in a 6 ϫ 50-mm glass vial and hydrolyzed with 6 N HCl at 110°C for 24 h in a Waters Pico-Tag work station (Pico-Tag ® System, Waters, Milford, MA). Released amino acids were labeled with phenylisothiocyanate and separated on a Pico-Tag column according to the recommended instructions of the manufacturer. Monomethylated and symmetrically dimethylated arginines purchased from Sigma were also labeled in the same manner and used as standards. The sample and the standards were injected onto an HPLC column (Waters Pico-Tag, 3.9 ϫ 300 mm) equilibrated with 140 mM sodium acetate buffer containing 0.05% (v/v) triethylamine and 6% (v/v) acetonitrile at 46°C, respectively. Amino acids were eluted from the column with the acetonitrile gradient recommended by the manufacturer. Phenylisothiocya-nate-and 14 C-labeled amino acids were simultaneously detected on a serially connected UV detector (269 nm) and a Model 150TR on-line flow scintillation analyzer (Packard Instrument Co.), respectively.
In Vivo Binding Assay-To detect in vivo interactions between PRMT5 proteins, 293T cells were transfected with various sets of PRMT5 expression plasmids. After transfections, the cells were broken in lysis buffer. Aliquots of cell lysates were incubated with glutathione or anti-FLAG antibody beads for 2 h at 4°C. The beads were then washed five times with lysis buffer. The bead-bound proteins dissolved in 1ϫ SDS-PAGE sample buffer with a final concentration of 100 mM DTT were then boiled. Protein samples were separated by SDS-PAGE and analyzed by Western blotting with anti-GST or anti-FLAG antibodies.
Purification of FLAG-PRMT5 and FLAG-PRMT1 Proteins-293T cells (5 ϫ 10 8 ) transfected with pCMV2-PRMT5 or pCMV2-PRMT1 DNA were broken in lysis buffer as described above. The cleared lysates were applied to an anti-FLAG affinity column (1 ϫ 10 cm) equilibrated in lysis buffer, and the column was then washed twice with lysis buffer. PRMT proteins were eluted with 100 mM glycine HCl buffer (pH 3.5). The protein elutes were collected in 1 M Tris base buffer (pH 8.0) and then dialyzed against PRMT buffer. The dialysates were concentrated using a Nanospin Plus centrifugal filter (Gelman Instrument Co.).
Preparation of GST Fusion Proteins and Mouse Polyclonal Antibody against PRMT5-GST fusion proteins were expressed and purified as described by the manufacturer (Amersham Pharmacia Biotech). In brief, cells harboring GST fusion expression plasmids were induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 2-3 h at 30°C. Cells were washed with phosphate-buffered saline buffer, resuspended in lysis buffer, and then sonicated. Soluble protein extracts were loaded onto glutathione-agarose columns (1 ϫ 10 cm). The columns were washed four times with lysis buffer. Bead-bound proteins were eluted with PRMT buffer containing 10 mM reduced glutathione. After dialysis, purified proteins were stored at Ϫ70°C.
A GST-PRMT5-C (amino acids 315-637) fusion protein was expressed as a soluble form in E. coli. The fusion protein was purified as described above. The purified fusion protein, which appeared to be homogeneous on an SDS-polyacrylamide gel, was used to immunize BALB/c mice for antibody production.
Sedimentation Analysis of PRMT5 and PRMT1 Proteins-The purified FLAG-PRMT5 or FLAG-PRMT1 protein (20 -40 g) was overlaid on a 35-ml gradient of 5-45% sucrose in sedimentation buffer (50 mM Tris-HCl (pH 7.5), 1 mM EDTA, 1 mM EGTA, and 1 mM phenylmethylsulfonyl fluoride). After centrifugation at 10°C for 24 h at 25,000 rpm in a Beckman SW 28 rotor, the fractions were collected from the bottom. To analyze a homo-oligomer of endogenous PRMT5 in human cells by sedimentation, 293T cells (1 ϫ 10 8 ) were washed once with phosphatebuffered saline and resuspended in 3 ml of PRMT buffer. The resuspension was sonicated for 30 s on ice. Cell debris was removed by centrifugation. The soluble protein extract was sedimented on a 35-ml 5-45% sucrose density gradient under the same conditions as described above.
Chemical Cross-linking of PRMT5 Proteins by Glutaraldehyde Treatment in Vitro-Glutaraldehyde (40) was used to form chemical bridges in the homo-oligomers of FLAG-PRMT5. The purified FLAG-PRMT5 or FLAG-PRMT5-C protein (100 ng) was incubated with glutaraldehyde (0.00006 -0.0006%) in 20 l of PRMT buffer for 5 min at room temperature. The samples were boiled in SDS-PAGE sample buffer containing 100 mM DTT for 5 min and then separated by 6% SDS-PAGE. PRMT5 proteins were detected by Western blotting with an anti-FLAG monoclonal antibody.
Subcellular Localization of Human PRMT5-COS-1 or Chang's liver cells (10 5 cells/ml) were cultured on a polylysine-coated slide glass before transfection. As a control, pEGFP-PRMT5 or pEGFP-C1 (5 g) was transfected. After 24 h, the cells were observed with a confocal microscope.

PRMT5 Contains the Conserved Sequence Motifs Present in a Family of S-Adenosyl-L-methionine-dependent Protein-arginine
Methyltransferases-Searches of the GenBank TM /EBI Data Bank to identify the biochemical function of PRMT5 revealed that full-size PRMT5 does not have significant homology to any characterized proteins. We separated N-and C-terminal regions of PRMT5 and searched to find whether the N-or Cterminal region has amino acid sequence homology to characterized proteins. The search revealed that whereas the N-terminal region does not have any homology to the known genes, the C-terminal portion of PRMT5 contains amino acid sequences homologous to domains I-III present in AdoMetutilizing methyltransferases including PRMT1 (9, 41-43) (Fig.  1). The amino acid sequence of the post-I domain of PRMT5 (Lys-Tyr-Ala-Val-Glu) matches well with that of human PRMT2 (Val-Tyr-Ala-Val-Glu) (19,20). Furthermore, intervals between the conserved domains in PRMT5, PRMT1, and PRMT3 are also well preserved. These observations led us to examine the possibility of whether PRMT5 contains intrinsic protein-arginine methyltransferase activity.
PRMT5 Contains Intrinsic Protein-arginine Methyltransferase Activity-The GST-PRMT5 plasmid, with GST at the N terminus of the gene for PRMT5, and a control plasmid (GST) were transfected into 293T cells, respectively. The cell lysates were pulled-down with glutathione-agarose beads. The bead suspensions were incubated with [ 14 C]AdoMet and MBP. The reaction products were visualized by SDS-PAGE and fluorography ( Fig. 2A). MBP was radioactively labeled only by the beads incubated with the cell lysate transfected with the GST-PRMT5 plasmid, but not by the beads from the control plasmid. The GST-PRMT5 protein bound to the beads was detected by Western blotting with anti-GST antibody ( Fig. 2A). To further demonstrate the protein methyltransferase activity of PRMT5, the FLAG-PRMT5 protein was partially purified from the 293T cell lysate transfected with the FLAG-PRMT5 plasmid via anti-FIG. 1. Human PRMT5 contains conserved domains of AdoMetdependent protein-arginine methyltransferases. The amino acid sequence of human PRMT5 (hPRMT5) was deduced from human PRMT5 cDNA sequenced in our laboratory. To clone a murine homolog of human PRMT5, primers 5Ј-GTCCTCCACGTAATGCCTATGACC-3Ј, 5Ј-CTCTAGCGTCACCACGGCATT-3Ј, 5Ј-GTCCTCCACCTAATGCCT-ATGACC-3Ј, and 5Ј-CAAGGCTCTGGACACTTGGC-3Ј were used in the polymerase chain reactions to amplify a partial fragment of PRMT5 from a mouse T-cell hybridoma cDNA library. The partial murine PRMT5 (mPRMT5) clones were sequenced. The amino acid sequence of murine PRMT5 missing perhaps six amino acids at its N terminus was deduced from the partial murine PRMT5 cDNA clones and GenBank TM /EBI accession numbers AA120484, AA066095, AA838892, AA266241, AA921641, AA560154, AA270442, AA755892, AA637429, AA139488, AA869821, AA967893, and AA638905. Murine PRMT5 is almost identical to human PRMT5 at the amino acid level. The amino acid sequences of human PRMT1 and PRMT3 were deduced from GenBank TM /EBI accession numbers AA307385 and Q99873. Conserved domains I, post-I, II, and III of protein-arginine methyltransferases are indicated by lines. Identical amino acid residues are indicated in boldface.
FLAG antibody-conjugated agarose beads. Two closely migrating protein bands of ϳ72 kDa in size in the enzyme preparation were detected by Western blotting with anti-PRMT5-C antibody and Coomassie Blue staining (Fig. 2B). The fast migrating band of 72 kDa was not detected by Western blotting with anti-FLAG antibody (data not shown), suggesting that this protein is endogenous PRMT5 and that a homomeric complex between FLAG-PRMT5 and endogenous PRMT5 is formed. MBP was methylated by purified FLAG-PRMT5 in a dose-dependent manner (Fig. 2B, 14 C-Methylation). Next, we determined whether PRMT5 methylates the arginine residue in MBP. The acid lysate of MPB methylated by FLAG-PRMT5 was subjected to separation by HPLC as described under "Experimental Procedures." 14 C-Labeled monomethylarginine and dimethylarginine comigrated with the standards monomethylarginine and symmetric dimethylarginine (Fig. 2C). To further demonstrate that the catalytic activity of PRMT5 is intrinsic and not derived from a contaminant from human cells during purification, we examined the enzyme activity of GST-PRMT5 purified from E. coli. It was reported that E. coli does not contain any protein-arginine methyltransferase (3). Furthermore, the conserved GXGXG motif (domain I) in the AdoMetdependent protein methyltransferases is known to be an AdoMet-binding site and is critical for the catalytic activity (41)(42)(43). Therefore, we constructed a domain I mutant in which the amino acids in the GAGRG motif of PRMT5 are substituted with RARLE. The GST-PRMT5 protein expressed in E. coli methylated MBP, whereas the domain I mutant (GST-PRMT5-M) did not (Fig. 3A). Taken together, the results indicate that PRMT5 contains intrinsic protein-arginine methyltransferase activity.
The C-terminal Region of PRMT5 Containing All the Methyltransferase Motifs Is Not Sufficient for the Catalytic Activity and Interacts with the N-terminal Domain of PRMT5-The domains for the catalytic activity are localized in the C terminus of PRMT5 from amino acids 359 to 637 (Fig. 1). To determine whether only the C-terminal domain of PRMT5 containing canonical methyltransferase motifs is sufficient for the enzyme activity, the purified FLAG-PRMT5-C-(308 -637) and GST-PRMT5-C-(315-637) proteins bound to glutathione-agarose beads were used in the enzyme reactions using MBP and histone as substrates, respectively. Neither the FLAG-PRMT5-C-(308 -637) nor GST-PRMT5-C-(315-637) protein showed the enzyme activity, whereas the full-size FLAG-PRMT5 and GST-PRMT5 proteins did (Fig. 3, B and C). The GST-PRMT5-C-(315-637) protein expressed in E. coli also did not show the enzyme activity (Fig. 3A). The results suggest that the Cterminal domain of PRMT5 by itself is not sufficient for the catalytic activity.
It was suggested that Skb1, a yeast homolog of human PRMT5, forms a homomeric complex in fission yeast and that the N terminus of Skb1 may be important for the homomeric complex formation (24). Furthermore, purified FLAG-PRMT5 appeared to form a homomeric complex (Fig. 2B). Therefore, we examined whether human PRMT5 forms a homomeric complex in vivo by co-immunoprecipitation. FLAG-PRMT5 and glutathione-PRMT5 plasmids were cotransfected into 293T cells. The cell lysates were precipitated with anti-FLAG antibody-or GST-conjugated agarose beads. The immunoprecipitates or the GST-bead-bound proteins were separated by SDS-PAGE and analyzed by immunoblotting using anti-GST or anti-FLAG antibodies (Fig. 4). GST-PRMT5 was found to be associated with FLAG-PRMT5 and vice versa. To determine whether an interaction between N-and C-terminal domains occurs, we cotransfected the FLAG-PRMT5-C-(308 -637) and GST-PRMT5-N-(1-309) plasmids into 293T cells. FLAG-PRMT5-C-(308 -637) was coprecipitated with GST-PRMT5-N-(1-309) and vice versa (Fig. 4). The results indicate that PRMT5 forms a homomeric complex in vivo and suggest that an interaction between the Nand C-terminal regions of PRMT5 is involved in the homomeric complex formation.
PRMT5 Forms Homomeric Complexes, Including a Dimer and Tetramer-The homomeric complex of PRMT5 was further characterized by sedimentation analysis on a sucrose gradient. As PRMT1 was shown to form an oligomeric complex in Rat1 cells (9, 10), we also characterized the oligomeric complex of purified PRMT1. The purified FLAG-PRMT5 and FLAG-PRMT1 proteins were sedimented onto 35-ml 5-45% sucrose gradients. The fractions were collected from the bottom of the gradients. Equal volumes of every other gradient fraction boiled in sample buffer containing a reducing agent were separated by SDS-PAGE and subjected to Western blot analyses with anti-FLAG antibody (Fig. 5A). The enzyme assay was carried out with equal volumes of the gradient fractions and MBP as a substrate. The radioactivity incorporated into MBP was quantified by a PhosphorImager (Fig. 5B). The FLAG-PRMT5 protein peaked mainly at two positions. The first position at which it peaked is fraction 41, which sedimented between standard molecular masses of 68 and 158 kDa. Most likely, this fraction represents a homodimer, although we cannot rule out that it may partially contain a monomer. The second position at which it peaked is fraction 25, which corre-sponds to a molecular mass of slightly Ͼ240 kDa. This fraction may represent multimers, including a tetramer. The enzyme activity comigrated with the dimer and multimer fractions of FLAG-PRMT5. Most of FLAG-PRMT1 sedimented between standard molecular masses of 158 and 240 kDa, suggesting the presence of a multimer, including a tetramer (Fig. 5A). A minor portion of FLAG-PRMT1 sedimented at a standard molecular mass of 68 kDa, suggesting the presence of a dimer. The MBPmethylating activity of FLAG-PRMT1 was mainly detected in the multimer fractions (Fig. 5B).
The presence of a dimer and multimers of purified FLAG-PRMT5 was further demonstrated by a chemical cross-linking experiment. Purified FLAG-PRMT5 or FLAG-PRMT5-C-(308 -637) was treated with glutaraldehyde, which forms intra-or intercovalent chemical bridges between proteins depending on its concentration (40). The reaction products boiled in sample buffer containing a reducing agent were separated by SDS-PAGE and analyzed by Western blotting with anti-FLAG antibody (Fig. 6A). Three protein bands, namely a monomer of 72 kDa, a homodimer migrating between standard molecular mass markers of 200 and 97.4 kDa, and a multimer Ͼ200 kDa in size, were detected in the FLAG-PRMT5 sample. FLAG-PRMT5-C-(308 -637) did not form oligomeric complexes. Taken together, the results indicate that PRMT5 forms homo-oligomers, including a dimer and tetramer.
Both Covalent Linkage via Disulfide Bond and Noncovalent Association Are Involved in the Formation of Homo-oligomers of PRMT5 and PRMT1-We examined whether the homo-oligomers of FLAG-PRMT5 and FLAG-PRMT1 are linked covalently by disulfide bond or by noncovalent association. Purified FLAG-PRMT5 boiled in SDS-PAGE sample buffer with or without a reducing agent was separated by SDS-PAGE. The PRMT5 protein was detected by Western blotting with anti-FLAG antibody (Fig. 6B, Pooled). Only a monomer was detected in the presence of 100 mM DTT, whereas a monomer and smeared bands containing presumably a dimer and multimer of FLAG-PRMT5 were detected in the absence of DTT, indicating that a portion of PRMT5 homo-oligomers are covalently linked by a disulfide bond. The presence of a monomer in the absence of a reducing agent suggests either that a monomer is present in purified FLAG-PRMT5 at a small quantity or that the oligomeric complexes are formed partially by noncovalent interaction of a monomer. To further determine the presence of covalent and noncovalent associations in the oligomeric complexes, we used the dimer and multimer fractions of FLAG-PRMT5, which were separated by sedimentation (fractions 25 and 41) (Fig. 5). The dimer and multimer fractions boiled in SDS-PAGE sample buffer with or without a reducing agent or in low concentrations of a reducing agent were separated by SDS-PAGE. Protein bands were detected by Western blotting with anti-FLAG antibody (Fig. 6B). In the presence of 10 and 100 mM DTT, the dimer fraction was dissociated completely into a monomer, whereas in the absence or at low concentrations of DTT, both the monomer and dimer were detected, indicating that a portion of the dimer is covalently associated by disulfide bond (fraction 41). The multimer fraction was also completely dissociated into a monomer in the presence of 10 and 100 mM DTT, whereas the monomer, dimer, and multimer were detected in the absence or at low concentrations of DTT (fraction 25). The multimer fraction apparently did not contain FLAG-PRMT5 present as a monomer, at least in the sedimentation analysis. The presence of a monomer and dimer in the absence or at low concentrations of DTT indicates that a portion of the multimer is formed by noncovalent interaction of a monomer or a disulfide-linked dimer and is SDS-labile. Furthermore, the presence of a multimer in the absence of DTT indicates that a portion of the multimer is formed by covalent association by disulfide bonds and is SDS-resistant. Thus, the results indicate that a PRMT5 multimer can be formed either by covalent association by disulfide bond or by noncovalent association of a monomer or dimer linked by a disulfide bond. A multimer can also be heterogeneous depending on the association properties. A monomer and dimer in the multimer fraction would not have been detected in the absence of a reducing agent if a noncovalent association were not involved in the homo-oligomerization.
Multimer fraction 33 of FLAG-PRMT1 separated by sedimentation (Fig. 5) was dissociated into a monomer in the presence of 10 and 100 mM DTT, indicating that covalent linkage by a disulfide bond is involved in the homo-oligomerization (Fig.  6B, Fraction #33). A monomer and dimer, as well as a smeared band Ͼ97.4 kDa in size, were detected in the absence or at low concentrations of the reducing agent, indicating that homooligomeric complexes of PRMT1 are also formed by noncovalent association of a monomer or a covalently linked dimer.
PRMT5 Is Present as a Homo-oligomer in Vivo-To determine whether endogenous PRMT5 is also present in oligomeric forms and as a predominant enzyme methylating MBP in cells, we sedimented the protein extract of 293T cells on a sucrose gradient. Equal volumes of the gradient fractions were separated by SDS-PAGE. An endogenous PRMT5 protein was then detected by Western blotting with anti-PRMT5-C antibody (Fig. 7A). The enzyme activities in the sedimentation fractions were measured using MBP as a substrate. 14 C-Labeled proteins were visualized by SDS-PAGE and fluorography (Fig.  7C). The radioactivity incorporated into MBP was quantified by a PhosphorImager (Fig. 7B). The endogenous PRMT5 protein peaked at fraction 17, which appears to correspond to the molecular mass of the tetramer of PRMT5 (4 ϫ 72 ϭ 288 kDa). Monomeric and dimeric forms of endogenous PRMT5 were not detected. The MBP-methylating enzyme activity was detected in a broad range of fractions containing the PRMT5 protein, and it appeared to peak at the fractions in which PRMT5 was not detected, suggesting that PRMT5 is not predominant (Fig.  7B). However, we cannot completely exclude that PRMT5 is a predominant MBP-methylating enzyme because endogenous proteins methylated by cellular enzymes (Fig. 7C, arrowheads) interfered with specific detection of 14 C-labeled MBP.
PRMT5 Contains Lower Specific Enzyme Activity Compared with PRMT1 and Differs in Cellular Localization from PRMT1-To compare the specific enzyme activity of PRMT5 with that of PRMT1, we used a GST-GAR (glycine/argininerich) protein expressed in E. coli as a substrate in the enzyme assay. GST-GAR is a recombinant protein containing the first 148 amino acids of the human fibrillarin protein fused in frame to GST. The amino-terminal region of fibrillarin contains 14 arginine residues, the majority of which are present in RGG consensus methylation sites (44 -46). The enzyme assay was carried out at different molar concentrations of purified FLAG-PRMT1 and FLAG-PRMT5 proteins. The reaction products were visualized by SDS-PAGE and fluorography, and then they were quantified by a PhosphorImager (Fig. 8). PRMT5 contained ϳ2.5-fold weaker specific enzyme activity compared with PRMT1.
PRMT1 is localized in the nucleus of Rat1 cells, whereas PRMT3 is predominantly cytoplasmic (10). We examined where human PRMT5 is localized in the cell. We fused the gene for GFP in frame to the N terminus of PRMT5. The GFP-PRMT5 plasmid was transfected into COS-1 and Chang liver cells, and then we examined green fluorescence under fluorescence microscopy (Fig. 9). Green fluorescence was detected mainly at in the cytoplasm of COS-1 and Chang liver cells, suggesting that PRMT5 is predominantly localized in the cytoplasm. DISCUSSION We demonstrated here that PRMT5 (Skb1Hs/JBP1) contains an intrinsic protein-arginine methyltransferase activity. First, the conclusion that the catalytic activity of PRMT5 is intrinsic is supported by the facts that primarily the conserved motifs of the known AdoMet-dependent protein-arginine methyltransferases are present in PRMT5 and that spacings between the motifs are not greatly different from those of other PRMTs (Fig.  1). Second, monomethyl-and dimethylarginines were detected in the acid lysate of MBP methylated by PRMT5 (Fig. 2C). Third, GST-PRMT5 expressed in E. coli, but not GST-PRMT5 domain I mutant, contained the enzyme activity (Fig. 3A). Finally, the PRMT activities comigrated with the oligomeric complexes of purified FLAG-PRMT5 sedimented on a sucrose gradient (Fig. 5).
The presence of a homomeric complex of PRMT5 in vivo was detected by co-immunoprecipitation (Fig. 4). We provide biochemical evidence that the homomeric complex of FLAG-PRMT5 is composed of a dimer and multimer by sedimentation analysis, chemical cross-linking, and SDS-PAGE in the presence or absence of a reducing agent (Figs. 5 and 6). The oligomerization was not due to a FLAG epitope tagged at the N terminus of PRMT5 because FLAG-PRMT5-C-(308 -637) did not form a homo-oligomer (Fig. 6A). Although the presence of the endogenous PRMT5 multimer in vivo was detected in the protein extract of 293T cells by sedimentation, an endogenous PRMT5 dimer was not detected (Fig. 7). It is likely that cells expressing FLAG-PRMT5 overproduce dimers simply as an intermediate for the formation of multimers. Alternatively, a FLAG epitope tagged at the N terminus of PRMT5 may partially interfere with efficient formation of homo-oligomers larger than dimers. The multimer might be composed of a tetramer, a hexamer, an octamer, and so on. However, sedimentation analyses suggested that endogenous PRMT5 or FLAG-PRMT5 is present most likely as a tetramer (Figs. 5 and 7), although we cannot rule out the presence of a multimer larger than a tetramer in size in vivo, which is likely unstable during purification or sedimentation analysis. PRMT1 also appears to be present as multimers, including a tetramer, because it migrates between standard molecular masses of 158 and 240 kDa (Fig. 5). In fact, Lin et al. (9) reported that endogenous PRMT1 migrated as a high molecular mass complex of ϳ180 kDa on a Superdex 200 gel filtration column, although it eluted in a broad peak ranging between 200 and 440 kDa on a Sephacryl S300HR gel filtration column (10).
The specific enzyme activity of GST-PRMT5 from E. coli was several hundredfold lower than that of GST-PRMT5 from mammalian cells (Fig. 3A), presumably because GST-PRMT5 expressed in mammalian cells is likely to form a homomeric complex with endogenous PRMT5. GST-PRMT5 from E. coli appeared not to form distinct homomeric complexes (data not shown). The C-terminal portion of PRMT5, which contains three important motifs for the PRMT activity, showed by itself neither the enzyme activity nor oligomerization (Figs. 3 and  6A). An interaction between the N-and C-terminal domains of the PRMT5 protein, which was detected by co-immunoprecipitation, appears to be involved in oligomerization of PRMT5 (Fig. 4). The multimer fractions of FLAG-PRMT5 contained higher specific enzyme activity compared with the dimer fractions, given that the dimer fractions contained much more PRMT5 protein than the multimer fractions as determined by Western blot analysis (Fig. 5). These results suggest that homooligomerization of PRMT5 is important for the catalytic activity.
The gene for type II PRMT, which produces symmetric dimethylarginine, has not been cloned. PRMT5 methylated MBP, histone, and GST-GAR as a designed methyl acceptor (Figs. 2,  3, and 8). Histones and GST-GAR have been shown to be efficient methyl acceptors for PRMT1, PRMT3, and yeast Rmt1 (arginine methyltransferase-1) (9, 10, 47), which represent type I enzymes. Previous studies on the mammalian PRMTs showed that the enzyme responsible for methylating MBP is distinctly different from the histone-methylating enzyme (48). The MBPspecific methyltransferase preferentially methylates MBP and histone to a much lesser extent. The histone-specific enzyme methylates only histone. We could not determine conclusively whether dimethylarginine detected in the acid lysate of MBP methylated by PRMT5 is symmetric or asymmetric and, consequently, whether PRMT5 is a type I or II enzyme. The HPLC column used for the detection of methylated arginines could not readily resolve symmetric and asymmetric dimethylarginines, although dimethylarginine in the acid lysate of MBP methylated by PRMT5 comigrated with a standard symmetric dimethylarginine (Fig. 2C). As PRMT5 methylates MBP and histone, PRMT5 may be an enzyme known as an MBP-specific methyltransferase.
Immunoaffinity-purified FLAG-PRMT1 methylated MBP (Fig. 5). However, it was reported that rat PRMT1 and Rmt1, a yeast homolog of human PRMT1, did not methylate MBP in vitro (9,47). Purified FLAG-PRMT1 was not contaminated with PRMT5 since the FLAG-PRMT1 enzyme preparation did not contain PRMT5 as determined by Western blot analysis and FLAG-PRMT1 did not interact with PRMT5 as determined by co-immunoprecipitation analysis (data not shown). The discrepancy from the previous results may result from MBP derived from different sources. We used MBP derived from guinea pig brain (Figs. 2, 3, and 5), whereas bovine brain MBP was used as a methyl acceptor in the cases of rat GST-PRMT1 and yeast GST-Rmt1 proteins expressed in E. coli (9,47). However, MBP derived from bovine brain was methylated by PRMT5 (JBP1) (23). These findings make it highly unlikely that MBP from different sources determines whether or not it functions as a methyl acceptor for protein-arginine methyltransferases. On the contrary, in vitro substrate specificity may depend on an enzyme preparation. FLAG-PRMT1 purified from mammalian cells may have different substrate specificity compared with recombinant GST-PRMT1 expressed in E. coli. This may account for the MBP-methylating activity of FLAG-PRMT1.
Presumably, the substrate and reaction product specificity of protein-arginine methyltransferases may be partially dependent on the heterogeneity of homo-oligomeric complexes. This is presumed since purified PRMT1 and PRMT5 preparations are composed of dimers and multimers, which can be separated by sucrose gradient sedimentation (Fig. 5). Furthermore, PRMT1 and PRMT5 multimers are likely to be composed of at least three different kinds of multimers formed by noncovalent association of a monomer or a disulfide-linked dimer or by covalent association via disulfide linkage (Fig. 6B). It is possible that noncovalently associated, large homo-oligomeric complexes of native PRMTs present in vivo might be dissociated into a more stable homo-oligomeric complex during purification by chromatographic methods from cells or tissues. This dissociation would result in different substrate and reaction product specificity. For example, although the spliceosomal D1 protein in HeLa cells is found to be symmetrically dimethylated in vivo, the D1 protein overexpressed in baculovirus-infected insect cells is asymmetrically arginine-methylated. A heptapeptide (GRGRGRG) also present in the D1 protein is asymmetrically methylated by the type II MBP-specific methyltransferase in vitro (8). Therefore, the findings suggest that the tenuous classification of type I and II PRMTs based on substrate and reaction product specificity in vitro should be reconsidered.
A heterogeneity of the homo-oligomeric complexes of PRMTs resulting from the association characteristics (Figs. 5 and 6) and the presence of various protein-arginine methyltransferases methylating MBP and histones in cells (1,22,23,47) might have also resulted in multiple protein bands in the highly purified PRMT enzyme preparation. In fact, although a number of purifications of protein-arginine methyltransferase have been reported (3,36,37), the polypeptide composition of protein-arginine methyltransferase has been controversial. For example, histone-arginine methyltransferase from calf brain contains two polypeptides of 110 and 75 kDa (36), whereas Rawal et al. (37) showed that the enzyme from rat liver contains only a single 110-kDa polypeptide. The purest proteinarginine methyltransferase fraction from HeLa cells contains about eight protein bands upon SDS-PAGE and silver staining, and the two most prominent bands have molecular masses of 45 and 100 kDa (3). Presumably, the two bands may represent a monomer and disulfide-linked dimer of PRMT1. Our result that the disulfide-linked dimer of PRMT1 is not completely dissociated into a monomer even in the presence of 1 mM DTT (Fig. 6B, Fraction #33) supports this possibility.
PRMT5 has been shown to bind the pICln (I ϭ current, Cl ϭ chloride, and n ϭ nucleotide-sensitive) protein, which is correlated with the appearance of a nucleotide-sensitive chloride current (49). The pICln protein was shown previously to exist in several discrete complexes with other cytosolic proteins (50). PRMT5 binds to Janus kinases, which are localized in the cytoplasm and are involved in an interferon-signaling pathway (23,(51)(52)(53). We found PRMT5 as a molecule interacting with the NS3 protein of HCV, which is also localized in the cytoplasm (54). The sedimentation profile of purified PRMT1 and PRMT5 indicates that they are distinctly different enzymes (Fig. 5). Thus, the results suggest that although PRMT1 functions mainly in the nucleus (10), PRMT5 plays important roles in the cytoplasm.
The HCV NS3 protein is a multifunctional protein containing serine protease and RNA helicase activities (28 -33). A specific interaction between the C-terminal domain of PRMT5 and the N-terminal domain of NS3 was detected in vitro and in vivo. 2 However, PRMT5 appeared not to affect the enzyme activities of HCV NS3. A reason, then, why PRMT5 interacts with NS3 is that HCV NS3 is post-translationally argininemethylated by a protein-arginine methyltransferase, and it may possibly modulate the enzyme activity in HCV-infected cells. In fact, we found that HCV NS3 as an RNA-binding protein contains several Arg-Gly motifs and that arginine residues in the NS3 helicase domain are methylated by cellular protein-arginine methyltransferase. 2 Therefore, it will be of great interest to determine the functions of PRMT5 interacting with the HCV NS3 protein in mammalian cells and the biological consequences of protein arginine methylation. The findings that PRMT5 and PRMT1 homo-oligomers formed by covalent and noncovalent interactions can be heterogeneous may contribute to understanding the polypeptide compositions of protein-arginine methyltransferases and identifying their specific substrates and reaction products.