Promiscuous Modification of the Nuclear Poly(A)-binding Protein by Multiple Protein-arginine Methyltransferases Does Not Affect the Aggregation Behavior*

The mammalian nuclear poly(A)-binding protein, PABPN1, carries 13 asymmetrically dimethylated arginine residues in its C-terminal domain. By fractionation of cell extracts, we found that protein-arginine methyltransferases (PRMTs)-1, -3, and -6 are responsible for the modification of PABPN1. Recombinant PRMT1, -3, and -6 also methylated PABPN1. Our data suggest that these enzymes act on their own, and additional polypeptides are not involved in recognizing PABPN1 as a substrate. PRMT1 is the predominant methyltransferase acting on PABPN1. Nevertheless, PABPN1 was almost fully methylated in a Prmt1-/- cell line; thus, PRMT3 and -6 suffice for methylation. In contrast to PABPN1, the heterogeneous nuclear ribonucleoprotein (hnRNP) K is selectively methylated only by PRMT1. Efficient methylation of synthetic peptides derived from PABPN1 or hnRNP K suggested that PRMT1, -3, and -6 recognize their substrates by interacting with local amino acid sequences and not with additional domains of the substrates. However, the use of fusion proteins suggested that the inability of PRMT3 and -6 to modify hnRNP K is because of structural masking of the methyl-accepting amino acid sequences by neighboring domains. Mutations leading to intracellular aggregation of PABPN1 cause the disease oculopharyngeal muscular dystrophy. The C-terminal domain containing the methylated arginine residues is known to promote PAPBN1 self-association, and arginine methylation has been reported to inhibit self-association of an orthologous protein. Thus, arginine methylation might be relevant for oculopharyngeal muscular dystrophy. However, in two different types of assays we have been unable to detect any effect of arginine methylation on the aggregation of bovine PABPN1.

Dimethylation of arginine side chains is a relatively common post-translational modification of proteins (1)(2)(3)(4)(5). Most proteins modified in this manner are involved in DNA or RNA function; many of them bind nucleic acids directly, and the modification is frequently found in RNA binding domains. Still, modulation of protein-protein rather than protein-nucleic acids interactions seems to be the predominant function of arginine dimethylation (2,3). There are two types of arginine dimethylation as follows. Symmetric dimethylation (-N G ,NЈ G -dimethylarginine) has been found in some of the so-called Sm proteins in the core of spliceosomal small nuclear ribonucleoprotein particles (6) and in several other proteins and is catalyzed by the type II proteinarginine methyltransferases (PRMT) 2 -5, -7, and -9 in mammalian cells (4). Asymmetric dimethylation (-N G ,N G -dimethylarginine) is the more frequent modification and is catalyzed by the type I protein-arginine methyltransferases PRMT1, -3, -4, -6, and -8 (4). PRMT2 is considered a type I enzyme because of sequence similarities, but catalytic activity has not been demonstrated. The human genome contains two more potential PRMTs, which have not been characterized (4).
The substrate preferences of different type I enzymes are poorly understood, and in many cases it is not clear whether a particular protein is methylated by a specific PRMT or by multiple enzymes. For some proteins, in vivo methylation by one specific methyltransferase has been confirmed by genetic or quasi-genetic means. For example, a knock-out of Prmt1 abolishes the methylation of the proteins Sam68 (7), hnRNP K (8), and histone H4 (9). The role of PRMT1 in H4 methylation has also been confirmed by RNA interference experiments (10). A hypomorphic allele of Prmt3 leads to a loss of methylation of ribosomal protein S2 in mice (11), and similar results have been obtained with a prmt3 deletion in Schizosaccharomyces pombe (12). A knock-out of Prmt4 abolishes the modification of the cytoplasmic poly(A)-binding protein and the transcriptional cofactor p300 (13), and knockdown of PRMT6 interferes with histone H3 methylation at Arg-2 (14,15). These data suggest specific enzyme-substrate relationships, and a published survey also indicates that PRMT4 and -6 modify unique substrates (3). In contrast, most identified PRMT3 substrates can also be modified by PRMT1 according to the same survey; this argues for redundant PRMT function.
The features of substrate proteins directing their modification by one or more PRMTs also remain to be defined. RGG domains with the consensus sequence (F/G)GGRGG(G/F) (1,16) were initially identified as prototypical arginine methylation sites. However, many identified substrates do not correspond to this consensus, and a more general definition of a methylation substrate is called a glycine-and arginine-rich sequence. A GST-glycine-and arginine-rich fusion protein (17) and synthetic peptides with similar sequences indeed serve as generic methyl acceptor sites, being methylated, in vitro, by PRMT1, -3, -6 and by the type II enzyme PRMT7 (3) but not by PRMT4 (18,19). This supports the idea that at least some of the PRMTs may be largely redundant with respect to their substrates.
One unbiased approach to the identification of the PRMT(s) relevant for the modification of a specific substrate is purification from a cell extract. For example, this approach has led to the identification of PRMT1 as being responsible for the methylation of Arg-3 in histone H4 (9). In contrast, experiments based solely on the use of recombinant PRMTs are potentially flawed. Many experiments have used GST fusion proteins, and this type of fusion affects the activity of some PRMTs but not others, skewing their relative efficiencies in the modification of a substrate (8). More importantly, for most PRMTs it is currently unknown whether they act by themselves or are merely the catalytic subunits of larger complexes that may influence the substrate specificity. Locus-specific arginine methylation of histones strongly suggests roles for additional proteins in targeting methylation (2,20). In fact, the methylation of histone H3 Arg-17 by PRMT4 (ϭCARM1) depends on the recruitment of the enzyme by a transcriptional coactivator (18). PRMT4 has also been identified as a component of a hetero-oligomeric complex, which can methylate histone H3 within nucleosomes, whereas PRMT4 by itself prefers the free histone as a substrate (21). The type II enzyme PRMT5 forms a complex with two other proteins, WD45 and pICln. The latter is responsible for substrate recognition by binding to the Sm domain (22)(23)(24).
The nuclear poly(A)-binding protein (PABPN1) binds the growing poly(A) tails during 3Ј-processing of mRNA precursors in the cell nucleus (25)(26)(27)(28). It stimulates the activity of poly(A) polymerase by increasing the processivity of the enzyme and contributes to limiting the poly(A) length to about 250 nucleotides (27,29,30). The protein contains a C-terminal domain of ϳ49 amino acids, in which all 13 arginine residues are quantitatively asymmetrically dimethylated (31,32). The methylated sites in mammalian PABPN1 are clusters of RXR sequences, which reside in a sequence environment reminiscent of RGG domains, being rich in phenylalanine and tyrosine and low in sequence complexity. The C-terminal domain contributes both to poly(A) binding and to the stimulation of polyadenylation, but neither function is noticeably influenced by arginine methylation (29,33). PABPN1 was found to be methylated by recombinant PRMT1 and -3 (32). However, this analysis was incomplete; other PRMTs were unknown at the time and thus not tested, and only recombinant GST-PRMT fusion proteins were used. Thus, the question of which enzymes are responsible for the modification of PABPN1 deserved to be re-evaluated.
Mutations in the human gene encoding PABPN1 cause a genetic disease, oculopharyngeal muscular dystrophy (OPMD). These mutations lead to the extension of the polyalanine tract at the N terminus of the protein from 10 residues in the wildtype up to 17 residues (34). The mutant protein forms regular fibrillar aggregates in muscle cell nuclei of afflicted individuals (35). Wild-type PABPN1 has a tendency to form ill-defined amorphous aggregates in vitro (31,36,37). Formation of these PABPN1 aggregates relies to a large extent on the C-terminal domain containing the dimethylated arginine residues (33,38). It has been proposed that formation of amorphous aggregates may promote formation of fibrils (38). Several observations have indicated that arginine methylation might affect protein aggregation. In immunoprecipitation experiments with cells lacking PRMT1, the background was strongly increased compared with wild-type cells, suggesting a generally increased tendency of proteins to aggregate (8). The Saccharomyces cerevisiae protein Npl3, which normally carries asymmetric dimethylarginine, has an increased tendency to self-associate in a methylation-deficient yeast strain (39). The PABPN1 orthologue in S. pombe (termed Pab2) was recently identified and shown to be arginine-methylated in RG dipeptides in its C-terminal domain. Again, genetic inactivation of the methyltransferase RMT1 increased the tendency of Pab2 to self-associate, and it was proposed that this observation might be relevant for PABPN1 fibril formation in OPMD (40).
In this study, we identify the type I arginine methyltransferases PRMT1, -3, and -6 as the enzymes modifying mammalian PABPN1. Among them, PRMT1 is the dominant enzyme. Our data suggest that PRMT1, -3, and -6 recognize their substrates on their own, independently of associated polypeptides, and that their overlapping but distinct substrate specificities are determined both by local amino acid sequences and by accessibility of methyl-accepting side chains in the context of a folded protein. Comparison of methylated and unmethylated PABPN1 in two different assays revealed no methylation-dependent difference in self-association, whereas the same assays showed that a His tag does favor aggregation.

EXPERIMENTAL PROCEDURES
Plasmids-Bovine His-PABPN1 and His-PABPN1⌬C49 were expressed from the plasmids pUK-synPABPN1 and pUK-synPABPN1⌬C49, respectively (33). For expression of nontagged PABPN1, the NdeI/BamHI fragment of pUK-syn-PABPN1 containing the PABPN1 open reading frame was cloned into the corresponding sites of pET11a (Novagen). PABPN1-KS and PABPN1-KL are fusion proteins in which the C-terminal domain of PABPN1 was replaced by fragments of hnRNP K containing the methyl-accepting arginines. The fragment encoding amino acids 258 -306 in the expression construct for His-tagged bovine PABPN1 was removed by means of an introduced KpnI site and the BamHI site. This fragment was replaced either by a DNA sequence encoding amino acids 241-333 of hnRNP K, amplified with the primers CGGGGTACCT-TTGATGACCGTCGC and CGCGGATCCTCAGTAACGG-TCTCCGAGC (PABPN1-KL fusion), or by a DNA sequence encoding amino acids 251-309, amplified with the primers CGGGGTACCGGATTTCCATGCGG and CGCGGATCCT-CAAGGAAGATTCCGAGC (PABPN1-KS fusion). The plasmid for recombinant expression of the llama single chain antibody fragment anti-PABPN1 (3F5) was a gift of Peter Verheesen, University of Utrecht (41). The plasmid pET28b-PRMT1 (42) used to express rat His-PRMT1 with a deletion of the first 10 amino acids was a gift of Xiaodong Cheng, Emory University. Human His-PRMT3 was expressed from the plasmid pET28a-PRMT3 (8). The associated human proteins PRMT5, His-pICln, and WD45 were encoded by the plasmid pET28a-pICln-PRMT5-WD45, obtained from Lucy Handoko and Utz Fischer, University of Würzburg, Germany. For expression of His-PRMT6, the human PRMT6 coding sequence was cut out from a GST fusion construct (8) with NdeI and BamHI and subcloned into pET19b (Novagen).
Recombinant Proteins-Expression and purification of His-PABPN1 and His-PABPN1⌬C49 were performed as described in Ref. 33 except that sometimes a ResourceS column (GE Healthcare) was used as the second step. Both PABPN1-K fusions were purified by the same procedure with a Mono Q column as the second step. For the purification of nontagged PABPN1 from Escherichia coli, the lysate was applied to a Blue-Sepharose column. Elution and further purification were as described (45). Recombinant hnRNP K was generously provided by Bodo Moritz (University of Halle). Purified recombinant SmD3 and SmB/BЈ were a kind gift of Utz Fischer. His-PRMT1, -3, and -6 were expressed and purified as described in Ref. 8. For use in kinetic measurements, His-PRMT1 and -3 were further purified on a Mono Q column (GE Healthcare) and His-PRMT6 on a Mono Q and a phenyl-Superose column. The nontagged PRMTs were purified by slight variations of the following procedure. The bacterial lysate was loaded on a DEAE-Sepharose column (GE Healthcare; 10 mg of protein/ml column volume) equilibrated in 50 mM Tris-HCl, pH 7.4, 50 mM KCl, 1 mM EDTA, 0.5 mM dithiothreitol, and 10% glycerol. The PRMTs were eluted with a KCl gradient from 50 to 500 mM KCl. Fractions active in the methylation assay (see below) were pooled and loaded on a hydroxyapatite column (Bio-Gel HTP, Bio-Rad; up to 5 mg of protein/ml column volume). For PRMT1 and -3, the equilibration solution contained 10% glycerol, 0.5 mM dithiothreitol, and either 1 mM MgCl 2 (PRMT1) or 1 mM NaCl (PRMT3). Protein was eluted in five steps from 0.01 to 0.3 M potassium phosphate, pH 8. For PRMT6, DEAE peak fractions were dialyzed against 25 mM Hepes, 1 M KCl, 10% glycerol, 1 mM dithiothreitol, and 0.01% Nonidet P-40, pH 7.5, and loaded onto a hydroxyapatite column. Following a washing step with 25 mM Hepes, 1 mM potassium phosphate, 10% glycerol, 1 mM dithiothreitol, and 0.01% Nonidet P-40, pH 7.5, the protein was eluted with a linear potassium phosphate gradient. Peak fractions from all hydroxyapatite columns were concentrated with a Centriprep device (Millipore), and 200 l were loaded on a Superdex 200HR 10/30 column (GE Healthcare) run in 20 mM Tris-HCl, pH 7.4, 100 mM KCl, and 10 mM dithiothreitol (for PRMT1 and -3) or 25 mM Hepes, pH 8.0, 100 mM KCl, 10% glycerol, 1 mM dithiothreitol, for PRMT6. For gel filtrations, fractions containing PRMT1 or -3, 10% glycerol was added for storage.
Concentrations of PRMT1 and -6 and all methylation substrates were determined by photometric measurements based on calculated extinction coefficients. Concentrations of PRMT3 were determined by SDS-PAGE, subsequent Coomassie staining, and comparison with known amounts of bovine serum albumin.
Synthetic Peptides-Peptides were synthesized with a free N terminus and an amidated C terminus by N-(9-fluorenyl)methoxycarbonyl/tert-butyl strategy and solid-phase peptide synthesis with the help of a peptide robot system (Syro; MultiSynTech, Witten, Germany) as described (46). After cleavage from the resin, the peptides were purified by preparative HPLC and analyzed by reversed-phase analytical HPLC. Identity of the peptides was confirmed by MALDI-TOF mass spectrometry (Ultraflex III; Bruker Daltonic).
Anti-PRMT1 and anti-PRMT3 were affinity-purified. The respective His-tagged antigens were immobilized on CnBr-activated Sepharose beads (GE Healthcare). Residual activated groups were blocked with 10 mM Tris-HCl, pH 8.0. The beads were washed with 10 volumes of 1ϫ PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 2 mM KH 2 PO 4 , pH 7.4), 1 volume of 100 mM glycine, pH 2.7, and three times with 10 volumes of 1ϫ PBS. Antiserum was diluted with 1 volume of 10 mM Tris-HCl, pH 7.4, 2 mM EDTA, 4 mM EGTA, 10 g/ml leupeptin, 5 g/ml pepstatin, 5 M PMSF, added to the beads, and incubated for 2 h at room temperature. The beads were washed three times with 10 volumes of 1ϫ PBS, and antibody was eluted with 100 mM glycine, pH 2.7, and immediately neutralized by the addition of 100 mM Tris-HCl, pH 8.0 (final concentration).
Purification of PRMT1 from Cytoplasmic HeLa Extract-All steps were performed in the cold. Extract or pooled fractions were dialyzed against the column equilibration buffer and centrifuged at 10 000 ϫ g if not indicated otherwise. All fractions were assayed for their methylation activity as described below. The distributions of PRMT1, -2, -3, -4, and -6 were analyzed by Western blot assays.
DEAE-Sepharose chromatography of HeLa cell extract was the column described in Ref. 8. Methylation activity directed against PABPN1 was found in two overlapping peaks. Fractions corresponding to the first peak, containing PRMT1 and -3, were pooled and loaded on a Macroprep-S column (7 mg of protein per ml column volume) equilibrated in buffer 1 (50 mM Hepes, pH 7.4, 50 mM KCl, 1 mM EDTA, 0.5 mM dithiothreitol, 10% glycerol). The column was washed with 1 volume of buffer 1. Both PRMTs were recovered in the flow-through and wash fractions. They were pooled and adjusted to the conductivity of buffer 2 (20 mM Tris-HCl, pH 7.4, 500 mM ammonium sulfate, 1 mM EDTA, 0.5 mM dithiothreitol, and 10% glycerol) by addition of ammonium sulfate. Precipitates were removed by centrifugation, and the supernatant was applied to a phenyl-Sepharose column (8 mg of protein per ml column volume) equilibrated in buffer 2. Proteins were eluted with a gradient from 500 to 0 mM ammonium sulfate. The methylation activity eluted from 150 to 0 mM ammonium sulfate and could be divided into two activity peaks as follows: the first contained mainly PRMT1 and the second PRMT1 and PRMT3. Fractions containing mainly PRMT1 were concentrated by centrifugation through an Amicon filter and loaded on a Superdex 200 Hiload 16/60 gel filtration column (GE Healthcare) equilibrated in 25 mM Tris-HCl, pH 7.4, 100 mM KCl, 1 mM EDTA, and 1 mM EGTA (17).
ES Prmt1 Ϫ/Ϫ cells were harvested and lysed in 50 mM Tris-HCl, pH 7.4, 300 mM NaCl, 1% Triton 100, 1 mM dithiothreitol, 10 g/ml leupeptin, 5 g/ml pepstatin, 5 M PMSF. Lysate was centrifuged for 30 min at 20,800 ϫ g, and the supernatant was dialyzed and loaded on a DEAE-Sepharose Fast Flow column (GE Healthcare) (10 mg of protein per ml column volume) equilibrated in buffer A containing 50 mM KCl. The column was washed with 1 volume of the same buffer, and proteins were eluted by a gradient from 50 to 500 mM KCl in buffer A. The methylation activity found in the fractions of the wash step and subsequent first fractions coeluted with PRMT3 and PRMT4. These fractions were pooled (pool IIa), and 25% was loaded on a 1-ml Mono S column, equilibrated in buffer B containing 50 mM KCl. Proteins were eluted by a gradient from 50 to 500 mM KCl in buffer B. The main methylation activity of the DEAE-Sepharose eluted from 150 to 250 mM KCl (pool IIb) and contained PRMT5 and -6. One representative column fraction of pool IIb was concentrated by centrifugation through an Amicon filter and loaded on a Superdex 200 HR 10/30 equilibrated in buffer A containing 50 mM KCl. This column separated the methylation activity into two pools as follows: pool IIIba contained PRMT5, and pool IIIbb contained PRMT6 and a negligible amount of PRMT5. Both pools were separately loaded on a 1-ml Mono Q column equilibrated in buffer C containing 50 mM KCl. The column was washed with 2 volumes of buffer C, and proteins were eluted by a gradient from 50 to 500 mM KCl in buffer C.
Purification of PABPN from Tissue-Purification from calf thymus was according to the published procedure (45). For purification from Prmt1 Ϫ/Ϫ ES cells, the same protocol was used. Fractions of the DEAE-Sepharose fast flow column (see above) containing PABPN1 were identified by Western blotting and used for further purification to homogeneity.
In Vitro Methylation Assay-Methyltransferase activity was assayed in a reaction volume of 25 l containing 50 mM Hepes, pH 8.0, 40 mM potassium acetate, 0.2 mg/ml bovine serum albumin, 0.01% Nonidet P-40, 10% glycerol, 1 mM EDTA, 0.5 mM dithiothreitol, 40 M S-adenosyl-L-[methyl-14 C]methionine (GE Healthcare), and 500 ng of PABPN1 (ϳ0.6 M). The reaction was incubated at 30°C for the time indicated. The incorporation of methyl groups was measured by SDS-PAGE, phosphorimaging, and quantification by the ImageQuant software (GE Healthcare). A standard curve with known quantities of labeled S-adenosylmethionine was generated for each experiment and used to calculate molar quantities of methyl group incorporation. One unit was defined as 1 pmol of methyl groups incorporated in 3 h.
For steady-state kinetics, reactions were performed in a total volume of 20 l under standard conditions. S-Adenosylmethionine was titrated up to 85 M with PABPN1 held constant at 0.9 M. PABPN1 and hnRNP K were titrated up to 1.8 or 1.7 M, respectively, with S-adenosylmethionine held constant at 42 M. Peptides were titrated up to a concentration of 10ϫ K m or higher except the K1 peptide in the PRMT1 reaction and the K2 peptide in the PRMT3 reaction, which were titrated up to 4ϫ K m . A linear dependence of the reaction velocity on enzyme concentration was experimentally verified, and reaction times were chosen to reflect initial rates. Typically, each data point was done in duplicate or triplicate, and each titration was carried out at least twice. K m and V max values were calculated either directly from hyperbolic fits or from linear fits to the Lineweaver-Burk equation. k cat was calculated from V max . In view of the highly probable ordered sequential mechanism (48), all kinetic constants given should be considered apparent constants.
Mass Spectrometric Analysis-After desalting with ZipTip C4 (Millipore), protein masses of PABPN1 purified from PRMT1 Ϫ/Ϫ ES cells were determined with a Q-TOF 2 mass spectrometer (Waters) equipped with a modified nano-ESI source to hold a pico-tip (New Objective, Cambridge, MA). The samples were injected with a syringe pump (Harvard Apparatus, MA) with a flow rate of 300 nl/min. The Max-Ent TM algorithm was used for deconvoluting the data for single charge state. Mass accuracy was Ϯ2 Da.
For the analysis of in vitro methylation, 500 ng of His-PABPN1 was methylated by either 1 g of His-PRMT1 (3 h of incubation), 2 g of His-PRMT3 (1 h of incubation with 1 g followed by 1 h with an additional 1 g), 1 g of His-PRMT6 (4 h of incubation), or 50 units of partially purified cellular PRMT1. 500 ng of Sm B/BЈ was methylated with 2 g of PRMT5/His-pICLn (4 h of incubation). Proteins were separated by SDS-PAGE. PABPN1 and Sm B/BЈ protein bands were cut out of the gels, washed three times with water, twice with 50 mM ammonium bicarbonate, and finally with 50 mM ammonium bicarbonate in 50% acetonitrile. The gel pieces were dried under a stream of nitrogen, reswollen in 20 l of 50 mM ammonium bicarbonate, pH 8.0, and digested with trypsin (Promega) overnight at 37°C. The peptide mass fingerprint spectra were recorded on an Ultraflex-II TOF/TOF mass spectrometer (Bruker) equipped with MALDI source, nitrogen laser, LIFT cell for fragment ion post-acceleration, and gridless ion reflector. The software Flex Control 2.4, Flex Analysis 2.4, and Biotools 3.0 were used to operate the instrument and analyze the data. For external calibration, a peptide calibration mixture (Bruker) was used. The methylated peptides were fragmented using the LIFT method (49). For MALDI sample preparation, 0.5 l of a saturated solution of ␣-cyano-4-hydroxycinnamic acid in acetone was deposited onto the anchor target. 1 l of the digest was added to the matrix layer. After evaporation of the solvent the spot was recrystallized with 1 l of ethanol, acetone, 0.1% trifluoroacetic acid (6:3:1).
Immobilization of PABPN1 and Pulldown Experiments-NHS-activated Sepharose beads (GE Healthcare) were washed three times with 1 mM HCl. 185 g of either His-PABPN1, recombinant nontagged PABPN1, PABPN1 purified from calf thymus, or bovine serum albumin was coupled separately to 185-l packed volume of NHS-activated beads overnight in the cold in 50 mM Hepes, pH 8.3, 150 mM KCl, 10% glycerol. Less than 10% of the input protein was detected in the supernatant after coupling. Residual activated groups were blocked with blocking buffer (500 mM ethanolamine, pH 8.3, 500 mM NaCl, 10% glycerol) overnight in the cold. Beads were washed three times with blocking buffer, three times with 100 mM sodium acetate, pH 4.0, 300 mM NaCl, 10% glycerol, again three times with blocking buffer, and five times with binding buffer (50 mM Hepes, pH 7.6, 150 mM KCl, 10% glycerol, 1 mM EDTA, 0.5 mM dithiothreitol). 2.3 mg of cytoplasmic or nuclear extract from HeLa cells (51) was incubated with 20 l of packed volume of beads. After incubation for 4 h in the cold, the beads were washed four times with binding buffer. Bound proteins were eluted with 40 l of SDS sample buffer (250 mM Tris-HCl, pH 6.8, 4% SDS, 20% glycerine, 1.43 M 2-mercaptoethanol, 0.02% bromphenol blue). 5 l of the eluted fractions were loaded onto a 10% SDS-polyacrylamide gel, and Western blot analysis was performed with a rabbit polyclonal antibody directed against PABPN1.

PRMT1
Methylates PABPN1-HeLa cell extracts were initially employed in an unbiased approach to identify, by protein fractionation, protein-arginine methyltransferases acting on PABPN1. We have previously described the fractionation of HeLa cell PRMTs by DEAE chromatography (8). In vitro methylation assays of these DEAE column fractions using recombinant bovine PABPN1 as a substrate showed two peaks of methylation activity, which corresponded roughly to the two peaks of PRMT1 protein, detected by Western blotting. The two PRMT1 peaks overlapped with the peaks of PRMT3 and -6, whereas PRTM2 and -4 were present in the flow-through and in the earliest fractions, which were devoid of methylation activity (supplemental Fig. 1). When the first methylation activity peak, containing both PRMT1 and -3 but not PRMT6, was further purified by cation exchange chromatography followed by phenyl-Sepharose chromatography, active fractions were obtained that contained PRMT1 as the only detectable type I arginine methyltransferase (Fig. 1). The type II enzyme PRMT5 was undetectable as well (data not shown). Methylation activity also coeluted with PRMT1 in subsequent gel filtration (data not shown). Methylation was specific because the PABPN1⌬C49 variant (33), lacking the methyl accepting arginines, was not methylated (data not shown).
Methylation of PABPN1 in Cells Lacking PRMT1-To confirm a role of PRMT1 in the methylation of PABPN1, we made use of a mouse ES cell line with a homozygous disruption of the Prmt1 gene (Prmt1 Ϫ/Ϫ cells) (52). Extracts from these cells had an 8-fold reduced methylation activity for PABPN1, but activity was still clearly detectable ( Fig. 2A). These data confirm a major role of PRMT1 but show the involvement of additional PRMTs. Methylation of PABPN1 by other PRMTs was confirmed by an experiment in which proteins were labeled in vivo by [ 3 H]methylmethionine under conditions of translation inhibition. PABPN1 was then precipitated by the single chain antibody 3F5 (41) directed against an epitope that is not subject to methylation. Incorporation of radioactivity was readily detectable in the Prmt1 knock-out cells (Fig. 2B). As noted previously in a similar experiment (8), the background of immunoprecipitation was increased with extracts from the Prmt1 knock-out cells.
If PRMT1 is the main enzyme responsible for the methylation of PABPN1, one would expect that the latter should not be fully methylated in the Prmt1 Ϫ/Ϫ ES cell line. However, we were unable to detect significant amounts of unmethylated PABPN1 with the help of an antibody specifically directed against an unmethylated PABPN1 peptide (data not shown). Also, when recombinant PRMT1 was added to extract of Prmt1 Ϫ/Ϫ ES cells, endogenous PABPN1 was not methylated to any significant extent, suggesting that the protein was already methylated (data not shown). For a definitive analysis, PABPN1 was purified from these cells by column chromatography (see "Experimental Procedures"). Western blots showed that PABPN1 behaved as a single entity during purification (data not shown); thus, selective loss of a significant fraction of the protein differing in its methylation status can be excluded. Analysis of the purified material by mass spectrometry showed a major peak with a mass of 32573 Da (supplemental Fig. 2). The difference to the calculated mass (32,206 Da, assuming removal of the start methionine and acetylation of the N terminus, as found for the bovine protein (32)) was 366 Da, indicating the presence of 26 methyl groups. A second smaller peak with a mass of 32544 Da corresponded to protein containing 24 methyl groups. Because PABPN1 purified from calf thymus contains 26 methyl groups (32), the data indicate that the majority of PABPN1 in Prmt1 Ϫ/Ϫ cells is fully methylated. Thus, other PRMTs can substitute for PRMT1 very efficiently.
PRMT3 and -6 Also Methylate PABPN1 but PRMT2 and -4 Do Not-Extracts of the Prmt1 knock-out cells were used to facilitate the search for additional PRMTs acting on PABPN1. Three separable activities were found, as summarized in Fig. 3. Upon DEAE chromatography, weak methylation activity was found in the wash and the first column fractions (fraction IIa), whereas a major methylation peak was eluted in later fractions of the salt gradient (fraction IIb). Fraction IIa contained  3 H]methionine, cycloheximide, and chloramphenicol (see "Experimental Procedures"). After cell lysis, PABPN1 was immunoprecipitated with the single chain antibody 3F5 (see "Experimental Procedures"). Extract (Input), the immunoprecipitate (IP), or control precipitation without antibody (buffer) were loaded on a 12% SDS-polyacrylamide gel, and proteins were visualized by fluorography. In the same experiment, hnRNP K was labeled in the PRMT1 ϩ/ϩ but not in the Prmt1 Ϫ/Ϫ cells (Fig. 5C in Ref. 8). This provides a control that there was no incorporation of radioactivity via translation and that PRMT1 activity was indeed absent. PRMT2, -3, and -4. Because the activity profile was distinct from the PRMT2 elution profile, and no catalytic activity has been reported for this enzyme (see Introduction), PRMT2 was not further considered. Subsequent Mono S chromatography (fraction IIIa) resulted in a strong increase in methylation activity, suggesting removal of an inhibitor. The activity was distributed over one major and one minor peak, each corresponding to a PRMT3 peak. PRMT4 was separated from PRMT3 and the methylation activity (Fig. 4). The same column fractions did not methylate PABPN1⌬C49 (data not shown). These results identify PRMT3 as a second methyltransferase acting on PABPN1.
Fraction IIb of the DEAE column contained PRMT6; other type I PRMTs were not detectable in these fractions but PRMT5 was (data not shown). Further analysis of a representative DEAE column fraction by gel filtration separated two methylating activities, which were further purified by Mono Q chromatography. One of these activities (fraction IVbb; Fig. 3) closely corresponded to the elution profile of PRMT6 (Fig. 5). Residual PRMT5 was also present, but its elution profile was clearly distinct from that of the methylation activity. PABPN1⌬C49 was not methylated by the column fractions (data not shown). Thus, PRMT6 is a third enzyme modifying PABPN1. A separate activity (fraction IVba; Fig. 3) contained no detectable type I PRMTs but, surprisingly, corresponded closely to the chromatographic behavior of PRMT5 (data not shown). This enzyme is not expected to act on PABPN1, because no symmetric dimethylation was detectable in the protein (32). Accordingly, the recombinant PRMT5⅐pICln heterodimer did not accept PAPBN1 as a substrate but was active in control reactions with its authentic substrates, SmD3 and SmB/BЈ (23) (data not shown). Attempts to determine whether PRMT5 in the column fractions was responsible for methyla-   tion of PABPN1 did not lead to conclusive results. Because these fractions made a minor contribution to the total activity, the question was not pursued.
In summary, we conclude that the type I methyltransferases PRMT1, -3, and -6 and one minor unidentified enzyme methylate PABPN1 but PRMT2 and -4 do not. PRMT8 is expressed specifically in the brain (53) and was therefore not considered. Methylation of PABPN1 by this enzyme in those cells in which it is expressed obviously can not be excluded. Type II enzymes PRMT5, -7, and -9 are not expected to act on PABPN1, in which exclusively asymmetric dimethylarginine has been detected (32).
Methylation of PABPN1 by Recombinant PRMTs-The assignment of PABPN1 methylation activity to PRMT1, -3, and -6 was confirmed by assays with recombinant enzymes. Because an N-terminal GST tag (in comparison to a His tag) strongly inhibits the activity of PRMT1 (8), the possibility was considered that any N-terminal tag might interfere with methyltransferase activity. Therefore, PRMT1, -3, and -6 were all purified both as nontagged and as His-tagged proteins (supplemental Fig. 3). All three enzymes were able to methylate PABPN1. The specific activities of the His-tagged and the nontagged versions of PRMT1, -3, and -6 were similar (Table 1 and data not shown). Therefore, all subsequent experiments were done with the His-tagged enzymes. In contrast to PRMT1, -3, and -6, GST-PRMT4 did not methylate PABPN1, although it was active with the cytoplasmic poly(A)-binding protein as a substrate (54) (data not shown).
The specific activities of the recombinant proteins were also compared with those of the "authentic" enzymes partially purified from cell extracts. For this purpose, the column fractions described above, in which individual PRMTs had been separated from all others, were used. The abundance of each PRMT was estimated by Western blotting in comparison with a standard curve generated with the corresponding recombinant protein, and the activity was measured under standard conditions. The results (Table 1) show that the specific activities of the recombinant proteins deviate at most by a factor of 2 from those of the authentic enzymes. High yields indicated that no factors promoting PABPN1 methylation were lost during the purification. Thus we conclude that the recognition of PABPN1 as a substrate is mediated by the catalytic polypeptides themselves and does not depend on any additional subunits or factors present in cell extracts.

PABPN1 Is Methylated in an Asymmetric Fashion in Vitro-
The type of arginine dimethylation generated in vitro was determined after incubation of PABPN1 with recombinant PRMT1, -3, and -6. The isolated PABPN1 was digested with trypsin, and a peptide mass fingerprint spectrum was recorded. All peptide ions containing dimethylated arginine were further fragmented in the mass spectrometer. A characteristic neutral loss of m/z 45 (dimethylamine) in the MS/MS spectra was a proof of asymmetric dimethylation in all cases. (supplemental Fig. 4; supplemental Table 1). As a control, after reaction of PRMT5 with SmD and SmB/BЈ, symmetric dimethylarginine was detectable by a typical neutral loss of m/z 70 (dimethylcarbodiimide) (55) (data not shown).
Of the partially purified cellular PRMTs, only PRMT1 was sufficiently active for this type of analysis; again, only asymmetric dimethylarginine was found (supplemental Table 1 and data not shown). These data are consistent with the exclusive presence of asymmetric dimethylarginine in PABPN1 purified from tissue (32).
PRMT1 Is the Predominant Methyltransferase for PABPN1-The abundance of PRMT1, -3, and -6 was estimated by Western blot analysis of HeLa cell extract in comparison with known quantities of recombinant polypeptides (Table 1). In agreement with earlier data (56), PRMT1 was found to be the most abundant methyltransferase. The relative contributions of the three enzymes to the methylation of PABPN1 were estimated from their abundance and their specific activities under standard conditions ( Table 1). The results show that close to 90% of the total methylation activity directed against PABPN1 is because of PRMT1 because this enzyme is the most abundant and has the highest activity for this substrate. This calculated contribution of PRMT1 is in good agreement with the assays of wildtype and Prmt1 Ϫ/Ϫ ES cell extract. PRMT3 and -6 contribute 5 and 7% because of a combination of lower activities and lower abundance.
Steady-state parameters for the methylation of PABPN1 by PRMT1, -3, and -6 were determined. PRMT3 activity is limited by a relatively high K m value, whereas PRMT6 has a low V max value. As a result, the catalytic efficiency of PRMT1, expressed as k cat /K m , is more than 10-fold higher compared with PRMT3 and -6 ( Table 2). In similar experiments, K m values for the other substrate, S-adenosylmethionine, were 1.5, 14.2, and 1.9 M for PRMT1, -3, and -6, respectively (data not shown).
Substrate Specificity of PRMTs-We have previously reported that the protein hnRNP K is methylated exclusively by PRMT1 (8). In agreement with these data, steady-state kinetic parameters show that methylation of hnRNPK by PRMT3 and -6 is 120-and 8-fold less efficient than methylation of PABPN1,

TABLE 1 Specific activities of PRMTs under standard reaction conditions and their contributions to PABPN1 methylation
His-PRMT1 was from rat, and all other recombinant PRMTs were human (see "Experimental Procedures"). Cellular PRMT1 was also human, and cellular PRMT3 and -6 were from mouse. The concentrations of non-tagged PRMTs were estimated from the intensity of the PRMT band in SDS gels in comparison with a bovine serum albumin standard. The specific activities of the cellular enzymes were based on the analysis of partially purified fractions. Their PRMT content was estimated by Western blots as described in the text. The accuracy of the specific activity determinations of the recombinant proteins is limited by their tendency to lose activity during storage; therefore, no standard deviations are given. For the same reason, the small differences observed between tagged and non-tagged proteins are not considered significant. Numbers in the last column were calculated from the two preceding columns. For assay conditions and unit definition, see "Experimental Procedures."

Enzyme
Specific activity (no tag)

Arginine Methylation of the Nuclear Poly(A)-binding Protein
which is relatively inefficient to begin with, whereas methylation of the two proteins by PRMT1 proceeds with similar efficiencies. As a consequence, PRMT3 and -6 are 3000-and 200fold less efficient than PRMT1 in the methylation of hnRNP K ( Table 2). In a first attempt to investigate which features determine the substrate properties of the two proteins, synthetic peptides were used. Peptide sequences are shown in the legend to Table 3 and aligned with protein sequences in supplemental Fig. 6. The sequence of the RGG peptide corresponds to the RGG consensus and has been used as a generic methylation substrate before (57). The RXR-1 peptide sequence was modified from the C-terminal domain of PABPN1 to keep its length and arginine content directly comparable with the RGG peptide. Because this peptide turned out to be a less efficient substrate than PABPN1, two longer peptides from this protein, RXR-2 and -3, were also used. Because methylated arginines in hnRNP K are spread over a larger region (8), this protein was represented by two peptides, K1 and K2. Results of methylation assays are summarized in Table 3.
For PRMT1, catalytic efficiencies obtained with the best peptides (RXR-2 and K2) were similar to those seen with the corresponding protein substrates. A slightly lower efficiency with the RXR peptides is similar to what we have observed with partial C-terminal deletions of PABPN1 with reduced numbers of arginine residues. 3 In contrast, the "standard" RGG peptide was not used efficiently by PRMT1. For PRMT6, the synthetic peptides also seemed to be good representatives of the proteins from which they were derived. The RXR-2 peptide was nearly as good a substrate as PABPN1, and the two K peptides were poor substrates. The RGG peptide was also a poor substrate. For PRMT3, the RXR peptides were reasonably good substitutes for PABPN1. However, the two peptides representing hnRNP K, in particular K1, were much better substrates than the entire protein, suggesting some inhibitory effect of the environment in which these sequences are found in hnRNP K. The best substrate for PRMT3 was the RGG peptide. For every PRMT, there was at least one peptide that was methylated with an efficiency comparable with that of the best protein substrate. Thus, the enzymes recognize their substrates via local amino acid sequences and not by contacting additional domains.
For a further investigation of substrate recognition by PRMT3, two fusion proteins were generated in which the methyl-accepting C-terminal domain of PABPN1 was replaced with amino acid sequences representing the domain of hnRNP K containing the methyl-accepting arginine residues (positions 256, 258, 268, 296, and 299) (8). Both domains are likely to be unstructured. The fusion protein PABPN1-KS contained hnRNP K amino acids 251-309 at the C terminus, and fusion protein PABPN1-KL contained hnRNP K amino acids 241-333 in the same position (supplemental Fig. 6). As expected, both fusion proteins were excellent substrates for PRMT1 (Table 2). PRMT3 methylated both fusion proteins with reasonable efficiency. More specifically, both fusion proteins as well as both K peptides were methylated with a high k cat , whereas poor methylation of full-length hnRNP K was limited by a very low k cat . Unexpectedly, both fusion proteins were also very good substrates for PRMT6, being methylated ϳ20-fold more efficiently than hnRNP K and 2-3-fold more efficiently than PABPN1 (Table 2). These data suggest that the inability of hnRNP K to serve as a substrate for PRMT3 and -6 must be due to the structure of the protein somehow impairing access of the methyltransferases.
Arginine Methylation Does Not Affect the Propensity of PABPN1 to Aggregate-Two types of experiments were carried out to determine whether arginine methylation has any influence on the tendency of PABPN1 to self-associate. First, three

hnRNP K, and fusion proteins by PRMT1, -3, and -6
Experiments were carried out with His-tagged recombinant enzymes (see "Experimental Procedures"). The fusion proteins PABPN1-KS and -KL are described under "Experimental Procedures" and supplemental Fig. 6. Standard deviations of V max were generally below 10%, and those of K m were below 20%.  variants of PABPN1 were covalently immobilized on Sepharose beads as follows: PABPN1 purified from calf thymus, which is fully methylated (32); His-tagged PABPN1 purified from E. coli, and nontagged PABPN1 also purified from E. coli; the latter two proteins are both unmethylated. Bovine serum albumin was immobilized as a control. The beads were then used for a pulldown experiment. They were incubated with HeLa cell extract and washed; proteins remaining bound to the beads were eluted with SDS-containing sample buffer, separated on an SDS-polyacrylamide gel, and analyzed by Western blotting with PABPN1-specific antibodies. All three species of immobilized PABPN1 reproducibly pulled down PABPN1 from the extract, whereas the albumin control did not. Results were the same with or without micrococcal nuclease treatment of the extract; thus, the self-association of PABPN1 was not RNA-mediated (data not shown). Whereas the His-tagged immobilized PABPN1 always pulled down about 3-fold more PABPN1, there was no difference between the nontagged recombinant PABPN1 and the protein purified from calf thymus (Fig. 6A). We conclude that the His tag favors PABPN1 self-association but the absence of methylation does not.

Enzyme
In a second type of experiment, the same three PABPN1 variants were allowed to associate with poly(A) of a length of ϳ280 nucleotides, and the complexes were analyzed by native gel electrophoresis. With both calf thymus PABPN1 and nontagged recombinant PABPN1, clean ladders of bands were obtained, reflecting increasing occupancy of the polynucleotide with protein (Fig. 6B). These data confirm that arginine methylation has no effect on the RNA binding affinity of PABPN1 (33). In contrast, the His-tagged protein showed no clear banding pattern but a strongly retarded, poorly defined smear instead, suggesting nonspecific aggregation on the polynucleotide (Fig. 6B). Thus, this experiment also suggests that the His tag promotes self-association of PABPN1, whereas the lack of methylation does not. This effect of the His tag in gel shift analyses was not noticed in earlier experiments presumably because of the use of shorter RNA ligands and different buffer conditions.
With a view to a potential role of arginine methylation of PABPN1 in the fibril formation occurring in OPMD patients, we also examined whether the polyalanine expansion affects the ability of the protein to serve as a PRMT substrate. Titrations of the mutant protein in activity assays with PRMT1, -3, and -6 showed that its ability to be methylated was at most 2-fold reduced in comparison with wild-type (supplemental Fig. 5 and data not shown). This small difference is unlikely to be biologically important, given that PABPN1 methylation is almost complete even in the Prmt1 Ϫ/Ϫ cell line.

DISCUSSION
In this work we have identified PRMT1, -3, and -6 as the enzymes responsible for arginine methylation of PABPN1. A previous study (32) was incomplete, as only PRMT1 and -3, the two enzymes then known, were tested. At the time, we suggested that PRMT3 might be the more important enzyme for the modification of PABPN1. This erroneous conclusion was because of the use of GST fusions of the PRMTs. The N-terminal GST part strongly inhibits the enzymatic activity of PRMT1 but not of PRMT3 (8). We now find that an N-terminal His tag does not affect the activities of PRMT1, -3, and -6. The enzymes partially purified from cell extracts had very similar specific activities to the recombinant enzymes, and no stimulatory factors were lost during the purification. Thus, methylation of PABPN1 is carried out by the catalytic polypeptides on their own, without the help of additional subunits. This is in contrast to the methylation of histones by PRMT1, which depends on targeting by transcriptional coactivators, and also to the methylation of Sm proteins by PRMT5, which relies on substrate recognition by the associated subunit pICln (see Introduction).
The ability of PABPN1 to be methylated by PRMT1, -3, and -6 is different from other PRMT1 substrates, like Sam68 or hnRNP K, which are methylated exclusively by this enzyme (7,8). Thus, the substrate specificities of PRMT1, -3, and -6 are FIGURE 6. Lack of methylation does not favor PABPN1 aggregation, but a His tag does. A, His-tagged PABPN1, nontagged PABPN1, completely methylated PABPN1 purified from calf thymus, and bovine serum albumin were covalently linked to NHS-activated beads, and cytoplasmic (CXT) and nuclear (NXT) HeLa cell extracts were incubated with the beads. Beads were washed, and bound protein was eluted and analyzed by SDS-PAGE and Western blotting with antibodies against PABPN1 (see "Experimental Procedures" for details). B, radioactively labeled A 280 was incubated with increasing amounts of different PABPN1 preparations as indicated (c t., protein purified from calf thymus; no tag, nontagged recombinant PABPN1; His, His 6 -tagged recombinant PABPN1). PABPN1⅐poly(A) complexes were resolved by native gel electrophoresis (see "Experimental Procedures"). Note that the ladders of bands of intermediate mobility present at limiting concentrations of PABPN1 (1st two lanes in each titration) represent partial occupancy of the polynucleotide. In contrast, the slightly different pattern of bands of intermediate mobility remaining at high concentrations of PABPN1 (last two lanes in the titrations) represents fully occupied but shorter polynucleotides contaminating the preparation (36). distinct but overlapping. This is also observed in the ability of the enzymes to use synthetic peptides as substrates. Some peptides are relatively good substrates for all three PRMTs (e.g. RXR-2), whereas others are methylated almost exclusively by one enzyme. For example, K2 is methylated by PRMT1 240-fold more efficiently in comparison to PRMT3 and 1700-fold more efficiently in comparison to PRMT6. In contrast, the RGG peptide is by far the best substrate for PRMT3 but a very poor substrate for PRMT1 and -6 ( Table 3).
The preferences of the individual PRMTs for certain peptides are in good agreement with their preferences for the proteins from which the peptides are derived, and the kinetic constants characterizing the methylation of entire proteins are similar to those determined with peptide substrates. Thus, PRMT1, -3, and -6 recognize local amino acid sequences as their substrates. This is in contrast to PRMT5, in which recognition of the Sm fold outside the methylated domain is important for methylation (22)(23)(24). However, the substrate specificities of the PRMTs are not solely determined by their preferences for certain peptide sequences. Even though hnRNP K is a very poor substrate for PRMT3 and -6, sequences taken from this protein can be efficiently methylated by the two enzymes when the substrates are offered either as free peptides or as C-terminal domains appended to a carrier protein (in this case, the RNP domain and preceding N-terminal domain of PABPN1). It is obviously possible that specific interactions within the full-length protein may limit the accessibility of the methyl-accepting amino acids. However, the mere presence of folded K homology domains on either side of the methylated regions in hnRNP K might be sufficient to prevent access of the methyl-accepting peptide sequences to the active sites of the PRMTs, which are facing the central cavity of a ring-shaped homodimer (42, 58 -61). In contrast, unstructured sequences located at the end of a protein, as in PABPN1, in the PABPN1-K fusions, and in many naturally occurring methylated proteins, can be assumed to thread easily into the central cavity. Interestingly, the inhibitory effect of the neighboring sequences on the methylation of arginine residues in hnRNP K is specific for PRMT3 and -6, whereas PRMT1 is not affected. It is tempting to speculate that PRMT1 might accommodate internal peptide sequences more easily due to more frequent opening of the dimeric structure.
We are aware of only one previous study in which the methylation of a protein and a derived peptide has been compared sytematically. Full-length histone H4 and a 21-amino acid peptide from its N terminus, carrying the methyl-accepting Arg-3, were methylated by PRMT1 with very similar k cat and K m values (62). Unfortunately, the data of Osborne et al. (62) cannot easily be compared with ours in a quantitative manner. The same RGG peptide was used in both studies. Although very similar k cat values were found, Osborne et al. (62) reported a 10-fold higher K m . One reason for this discrepancy could be that they used peptides with a free carboxyl end, whereas ours were C-terminally amidated. An influence of different reaction conditions cannot be excluded either.
The "RGG box" sequence was previously considered a good substrate for arginine methylation (16). Surprisingly, the RGG peptide, which represents this sequence, is a rather poor sub-strate for PRMT1, the major methyltransferase; several peptides deviating very strongly from the RGG box sequence were up to 40-fold better. Osborne et al. (62) also found that the RGG peptide was a relatively poor substrate compared with the histone H4 peptide. The identification of the RGG box as a consensus PRMT substrate was based mostly on the comparison of sequences surrounding methylated arginines identified in proteins. This may be misleading as the derived consensus may reflect the function of the methylated protein domain, RNA binding in the case of the RGG box, rather than the substrate specificity of the methyltransferase.
Osborne et al. (62) concluded that a patch of basic amino acid residues at the C terminus of their peptide substrate, separated by 12 amino acids from the methyl-accepting position, was important for recognition by PRMT1. This cannot be seen with the peptides tested here. In the best substrate, K2, we have not determined which of the four arginine residues is methylated, but the two most distant arginine residues are separated by only 8 amino acids, and there is no "patch" of several basic residues. In the RXR-1 peptide, still a reasonably good substrate, the middle arginine is the dominant methylation site, 3 and with the exception of the two arginines in the immediate vicinity, there is no basic residue on either side.
The ability of PRMT1, -3, and -6 to substitute for each other in the modification of a subset of their substrates may contribute to the viability of Prmt1 Ϫ/Ϫ cells (52) and Prmt3 knock-out mice (11). Even though PRMT1 appears to contribute about 90% of the total capacity of a cell to methylate PABPN1, the removal of the enzyme leads only to a very small loss in the overall methylation of this particular substrate. Thus, an investigation of the phenotypic effects of a lack of PABPN1 methylation will require RNA interference-mediated knockdown of PRMT3 and -6 in the Prmt1 Ϫ/Ϫ background. The small effect of a Prmt1 deletion on the methylation level of PABPN1 is surprising in view of the low activity of the PRMTs with k cat values of 2/min or less. This low activity was not improved by the use of synthetic peptide substrates or by any of a number of additives that we tested, including poly(A), the ligand for PABPN1, or an ATP-regenerating system in a crude extract. Thus, the PRMTs seem to be intrinsically slow. We also considered the possibility of synergistic effects of multiple PRMTs in the modification of PABPN1. Because PRMT6 prefers monomethylated arginine as a substrate (15,48), the enzyme might efficiently use products generated by PRMT1 or -3. However, in all pairwise combinations of the three PRMTs, activities were roughly additive (data not shown).
As explained in the Introduction, several pieces of evidence suggested that arginine dimethylation might reduce the inclination of proteins to associate in a presumably nonspecific manner. Surprisingly, our data suggest that self-association of PABPN1 is not strongly influenced by its modification. The pulldown assays with immobilized PABPN1 clearly demonstrate the ability of the protein to form oligomers. Moreover, an increase in self-association dependent on the N-terminal His tag was easily detectable. In contrast, no influence of methylation was observed. In the formation of RNA-protein complexes, assayed by native gel electrophoresis, aggregate formation dependent on the His tag was again evident, whereas no such effect was seen for the lack of arginine methylation. In analytical ultracentrifugation, Histagged, nonmethylated PABPN1 sediments as a monomer at a concentration of 2 M (29). Unfortunately, our attempts to investigate the concentration dependence of aggregation more systematically were not successful because the protein tended to aggregate in an erratic manner during the purification when present at high concentration, as reported previously (37). Thus, the effect of His tag and arginine methylation could not be examined more thoroughly.