Arginine methylation of scaffold attachment factor A by heterogeneous nuclear ribonucleoprotein particle-associated PRMT1.

Components of the heterogeneous nuclear ribonucleoprotein (hnRNP) complex and other nucleic acid-binding proteins are subject to methylation on specific arginine residues by the catalytic activity of arginine methyltransferases. The methylation has been implicated in transcriptional regulation and RNA and protein trafficking and signal transduction, but the mechanism by which these functions are achieved has remained undetermined. We show here that the predominant arginine methyltransferase in human cells, protein arginine methyltransferase 1 (PRMT1), is associated with hnRNP complexes, dependent on the methylation status of the cell, and that it methylates its preferred substrates in situ. Binding of PRMT1 occurs through physical interaction with scaffold attachment factor A (SAF-A), also known as hnRNP-U, which is quantitatively methylated by PRMT1 in all investigated cell lines as determined by a novel, highly specific, methylation-sensitive antibody.

Methylation of arginine residues in proteins is a post-translational modification that increases the structural diversity of proteins and might modulate their function in the living cell. The major targets of arginine methylation have been identified as nuclear proteins that interact with nucleic acids, such as components of the hnRNP 1 particle involved in pre-mRNA processing and transport, or nucleolar components such as fibrillarin and nucleolin, which are involved in pre-rRNA processing and ribosome biogenesis (1)(2)(3)(4). More than 85% of the methylation of these substrates is catalyzed by PRMT1, the predominant protein arginine methyltransferase in human cells, which is located both in the nucleus and in the cytoplasm (5)(6)(7)(8). PRMT1 is a type I methyltransferase that transfers a methyl group from S-adenosylmethionine to guanidino nitrogens of arginine residues to form monomethylarginine and asymmetric dimethylarginine (for review, see Ref. 9). Other type I methyltransferases, such as PRMT3, -4, -6, and -7 have also been identified in higher eukaryotes, but not in yeast, and are characterized by differences in substrate specificity and subcellular localization (8,10,11,13). Recently, disruption of the Prmt1 gene in mice has shown that PRMT1 plays an essential and nonredundant role, as Prmt1Ϫ/Ϫ embryos die shortly after implantation, and ES cells derived from these embryos are defective in their ability to differentiate in vitro (5,14). However, these ES cells are viable and able to assemble apparently normal hnRNP complexes, suggesting that arginine methylation is not required for vital processes under cell culture conditions. This is consistent with findings in yeast, where disruption of the only known arginine methyltransferase, RMT1, does also not affect cell viability (15,16).
Studies on known substrates reveal that PRMT1 preferentially methylates arginine residues in an RG or RGG context, a common feature of RNA-binding proteins, suggesting that methylation might modulate protein-RNA or protein-protein interactions (1). This might then affect a variety of functions such as signal transduction, protein trafficking, or transcriptional regulation. In fact, methylation of the transcription factor Stat1 suppresses interactions with its inhibitory protein PIAS1 and thereby enhances Stat1-driven transcription (17). In addition, PRMT1 and also PRMT4/CARM1 act as transcriptional co-activators, e.g. of nuclear hormone receptors, most probably because they methylate histones H3 and H4 and thereby facilitate histone acetylation and chromatin remodeling (10, 18 -21). Importantly, PRMT1 has been found to interact physically with several proteins, such as the interferon ␣,␤-receptor (22,23), the interleukin enhancer-binding factor 3 (ILF3 (24)) or the mitogen-induced proteins TIS21 and BTG1 (25). It is not clear in all cases whether PRMT-interacting proteins are also substrates that are methylated by PRMT1 or whether binding to PRMT1 itself has as regulatory role independent of the enzymatic activity.
Even though several different functional consequences of arginine methylation have been described, the precise biological or biochemical role of arginine methylation has not been determined. One major problem concerns the finding that arginine methylation appears to be both constitutive and irreversible. The constitutively high activity of PRMT1, as well as the apparent lack of an arginine demethylase to remove the methyl group again, is difficult to reconcile with a dynamic role in signaling, as classical signals depend on transience. Indeed, most known substrates appear to be extensively methylated, with no detectable regulation (14). This does not rule out that some proteins may exist in a hypomethylated state, e.g. because they are bound to other proteins that inhibit methylation due to steric hindrance, and only become methylated in response to different physiological stimuli. Up to now, however, no such proteins have been reported. It will therefore be important to investigate more substrate proteins with regard to the level of their methylation in vivo, and potential differences of the methylation status in different cell types and under different conditions.
In the present article, we have investigated the methylation of scaffold attachment factor A (SAF-A). SAF-A is a multifunctional nucleic acid-binding protein that interacts with both DNA and RNA in vivo (26,27). Because of the specific interaction with putative architectural DNA elements in the genome, the so-called scaffold attachment regions, the protein has been implicated in higher order chromatin structure and in the functional architecture of the cell nucleus (28 -30) and DNA replication (31). Interaction of SAF-A with RNA polymerase II, transcription factors such as the glucocorticoid hormone receptor, and the histone acetylase CBP/p300 has indicated a role in the regulation of gene expression (32)(33)(34)(35)(36), and a recent study suggested a role in X-chromosome specific transcriptional inactivation (37). In addition, a subpopulation of SAF-A is present in hnRNP particles, suggesting that the protein might also be involved in RNA maturation and transport. On the basis of these experiments, the protein is also known as hnRNP-U (38). We show here that SAF-A physically interacts with PRMT1 and is methylated by its enzymatic activity. Using a novel methylation-sensitive antibody, we demonstrate that SAF-A is present in fully methylated state in cell lines from a multitude of different origins. These findings are compatible with a role of arginine methylation in protein structure, or "labeling," rather than a dynamic role in signaling processes.

EXPERIMENTAL PROCEDURES
Cell Culture, Transfection, and Treatment with Periodate-oxidized Adenosine-Human embryonic kidney cells (HEK293) were cultivated on plastic dishes in Dulbecco's modified Eagle's medium with 10% fetal calf serum in a humidified atmosphere containing 5% CO 2 and were split 1:5 every second day. All cell lines used for the experiment shown in Fig. 6 were kept under conditions recommended by the American Type Culture Collection.
HEK293 cells were passaged the day before transfection, and transfection was performed with polyethylenimine as described previously (39). Transfection efficiency as determined by fluorescent cell sorting was routinely between 25 and 30%. Stable cell lines expressing SAF-A: Myc were created by selection with G418 for 5 weeks.
For inhibition of methylation and preparation of hypomethylated cell extract, the medium of HEK293 cells was supplemented with 15 M periodate-oxidized adenosine (adenosine-2Ј-3Ј-dialdehyde, Sigma A7154), and cells were cultured for an additional 48 h before harvesting.
Preparation of Total Cell Extract-Cells were harvested by scraping off the culture dishes with a rubber policeman after two washes with PBS and collected by centrifugation (200 ϫ g, 5min). For the preparation of hypomethylated extract, cells were resuspended in water and disrupted by sonication. Extract for immunoprecipitation was prepared by resuspending cells in lysis buffer (1.5ϫ PBS, 1% Triton X-100, and Complete EDTA-free protease inhibitor mixture (Roche Applied Science)), incubating them on ice for 5 min, and clearing by centrifugation for 15 min in an Eppendorf microcentrifuge at full speed.
Immunoprecipitation, Western Blotting, and Immunofluorescence-For immunoprecipitation, cell extract from 1-5 ϫ 10 6 cells was supplemented with 5 g of antibodies against SAF-A (26) or the Myc tag (clone 9B11, Cell Signaling) and incubated for 2 h at 4°C, and immune complexes were collected by incubation with protein G-Sepharose for an additional hour. Immune complexes were washed thoroughly eight times before bound proteins were eluted with SDS-PAGE sample buffer.
A methylation-sensitive antibody against SAF-A was developed by injecting rabbits with the synthetic peptide CRGNYNQNFRGRGN-NRG, corresponding to amino acid residues 778 -793 in wild type SAF-A, plus a single cysteine residue added to the amino terminus for purification and immobilization on thiopropyl-Sepharose (29). Sera from different animals were tested, and a methylation-sensitive antibody, designated A232, was affinity-purified over immobilized peptide.
Immunofluorescence was performed on periodate-oxidized adenosine (Adox)-treated and untreated HEK293 cells, using either the antibody K371 against total SAF-A or the methylation-sensitive antibody A232 in concentrations of 1:300 for 1 h at 37°C. Alexa568-coupled secondary antibody (Molecular Probes) was used for fluorescence detection.
Glycerol Gradient Centrifugation-Native total cell extract was prepared as described for immunoprecipitation and layered over a 10 -30% glycerol gradient (in lysis buffer) in a Beckman SW41 tube. The gradient was spun at 4°C for 20 h at 25,000 rpm, optimized for 40 S hnRNP particles sedimenting in the middle of the gradient, and fractionated in 0.5-ml aliquots from the top. Aliquots of each fraction were resolved by SDS-PAGE, and proteins were either stained with Coomassie Blue or detected by Western blotting.
Purification of Recombinant PRMT1 and PRMT2 and Methylation Reactions-Vectors for bacterial expression of GST-fused PRMT1 and PRMT2 were kindly provided by Dr. Harvey Herschman (UCLA, Los Angeles) and Dr. Michael Henry (University of Medicine and Dentistry of New Jersey, Stratford, NJ). Recombinant PRMTs and GST were purified from transformed Escherichia coli (strain BL21RIL, Stratagene) by affinity chromatography over glutathione-Sepharose as described previously (40) using the soluble protein fraction of enzyme from bacteria permeabilized with B-Per (Pierce). This fraction contained ϳ10% of the expressed PRMT proteins, whereas the vast majority present in inclusion bodies was discarded; in contrast, GST was quantitatively recovered from the soluble fraction. All purified proteins showed the expected size in SDS-PAGE (see Fig. 2 for purified PRMT1-GST) and were quantified by the Bio-Rad protein assay. In addition, GST and GST fusion proteins were quantified by the CDNB (1-chloro-2,4-dinitrobenzene) colorimetric assay for the enzymatic activity of the GST part of the fusion protein (41) to ensure the use of identical amounts of properly folded and active protein in comparative assays (e.g. in Figs. 2 and 5).
For in vitro methylation, extract from hypomethylated cells was pretreated at 70°C for 10 min to inactivate endogenous methyltransferases and then centrifuged for 10 min at full speed. The supernatant was adjusted to 1ϫ PBS (final concentrations: 1.5 mM KH 2 PO 4 , 12.7 mM K 2 HPO 4 , 138 mM NaCl, 2.7 mM KCl, pH 7.5) by adding concentrated PBS stock solution and centrifuged again. Per reaction, 20 l of extract was combined with 500 ng of recombinant PRMT and 5 Ci of S-adenosyl-L-[methyl-3 H]methionine (Amersham Biosciences TRK865, specific activity 2.96 TBq/mmol), and reactions were incubated for 2 h at 37°C. The reaction mixture was then resolved by SDS-PAGE, and radioactively labeled proteins were visualized by fluorography with enHance reagent (PerkinElmer Life Sciences).

RESULTS
Motivated by earlier studies reporting that the multifunctional nuclear protein SAF-A is subject to arginine methylation (1,42), we began to study this modification of SAF-A in more detail, with the ultimate goal of understanding its functional role in vivo. With at least six active protein arginine methyltransferases known to be present in human cells (13), we first sought to identify the methyltransferase responsible for the methylation of SAF-A. In an initial experiment, we performed immunoprecipitation of SAF-A and used commercially available antibodies against various PRMTs to identify potential interactions (Fig. 1). We found that only PRMT1, but not the other tested PRMTs 3, 5, and 6, could be co-precipitated with SAF-A, suggesting a physical linkage between SAF-A and PRMT1. However, as SAF-A is a component of hnRNP complexes and PRMT1 can also be co-precipitated with antibodies against hnRNP-C1/C2, the interaction between SAF-A and PRMT1 could be indirect via other hnRNP proteins (Fig. 1C). To investigate this possibility, we performed a far Western analysis, in which a protein sample is resolved by SDS-PAGE, blotted to a membrane, renatured, and incubated with recombinant protein. After extensive washing, immunodetection of the recombinant protein revealed signals for each protein that has bound the "prey." We immunoprecipitated endogenous SAF-A, together with the attached hnRNP proteins, and incubated the blot with recombinant PRMT1 (Fig. 2). Immunode-tection with either anti-GST or anti-PRMT1 antibodies resulted in clearly detectable signals at the position of SAF-A, providing evidence for a direct interaction of SAF-A and PRMT1. No signal was obtained with GST protein alone, even when 10-fold more protein was used ( Fig. 2A), or in mock immunoprecipitations with protein G-Sepharose alone (Fig.  2B). Thus, SAF-A and PRMT1 directly interact with each other in the hnRNP complex. With the possible exception of very weak additional signals in the far Western analysis, no evidence for binding of PRMT1 to other hnRNP proteins was found in these experiments.
The direct interaction of PRMT1 with SAF-A suggests that PRMT1 is also the enzyme that is responsible for methylation of SAF-A, which is indeed the case. As shown in Fig. 3A, immunoprecipitated SAF-A can be methylated in vitro by recombinant PRMT1 (Fig. 3A). In addition to SAF-A, coprecipitated hnRNP proteins are also methylated by PRMT1 in this reaction, demonstrating that these proteins are accessible to the enzyme in the native hnRNP complex. Moreover, endogenous PRMT1 co-precipitated with SAF-A is catalytically active both in trans and in cis. Methylation in trans is observed after the addition of hypomethylated extract from cells treated with Adox, a well characterized inhibitor of cellular methylation reactions (Fig. 3B, left panel). More interestingly, co-precipitated PRMT1 is also able to methylate SAF-A and other substrate proteins in cis, in the hnRNP particle, after the addition of radioactive S-adenosylmethionine only (Fig. 3B, right panel). Consistent with a methylation by PRMT1, immunoprecipitated SAF-A was recognized by an antibody against asymmetric, but not symmetric, dimethylarginine (data not shown).
Even though hnRNP proteins are well known as the preferred substrates of PRMT1, we were surprised to see a stable protein interaction between hnRNP particles and the enzyme. To validate the result from our immunoprecipitations, we performed glycerol gradient centrifugation to prepare native 40 S hnRNP particles and tested for the presence of PRMT1 in these complexes (Fig. 4). Although we were able to detect a subpopulation of PRMT1 that co-sedimented with the hnRNP particle, confirming the immunoprecipitation experiments, its amount was quite small. This changed dramatically when the experiment was repeated with extracts from cells that had been FIG. 2. PRMT1 and SAF-A interact directly in vitro. A, endogenous SAF-A was immunoprecipitated from HEK293 cells, and the precipitated protein was split into two identical aliquots that were resolved by SDS-PAGE and blotted to a polyvinylidene difluoride membrane. The blot was cut in two strips and, after renaturation in PBS, was incubated with purified recombinant PRMT1-GST (50 ng/ml in PBS) or 10-fold more GST protein, respectively. Bound protein was detected by a horseradish peroxidase-conjugated anti-GST antibody and chemiluminescence, revealing a signal only in the PRMT1-treated lane (lower panel). A Western blot with SAF-A antibodies confirm the presence of SAF-A in both lanes (upper panel). Note some weak additional signals in the PRMT1 lane. Asterisks, IgG heavy and light chains, the latter of which fortuitously binds GST for unknown reasons. B, a similar experiment was performed, but immunoprecipitation was controlled by mock precipitation with protein G-Sepharose alone (right lane), and detection of bound PRMT1 was performed by a PRMT1-specific antibody. IP, immunoprecipitate.

FIG. 1. PRMT1 is associated with SAF-A in hnRNP complexes.
A, PRMT1 but not PRMT3, -5, or -6 is co-immunoprecipitated with Myc-tagged SAF-A. Extract from HEK293 cells transfected with a SAF-A:Myc expression plasmid or the empty pCMV-Tag1 vector (control) was immunoprecipitated with 3 g of anti-Myc antibody. Precipitated proteins were separated by SDS-PAGE, blotted, and probed with mixtures of anti-Myc antibody and the indicated anti-PRMT antibody (left panel). The control lane was probed with PRMT1 antibody. Arrow, SAF-A:Myc; asterisk, IgG heavy chain. The presence of PRMTs in the cell lysates was verified by Western blot with the same anti-PRMT antibodies (right panel). B, PRMT1 is associated with endogenous SAF-A. Total cell extract (Input) was immunoprecipitated with antibodies against endogenous SAF-A or empty protein G-Sepharose as a control, and the precipitate was probed with antibodies against SAF-A and PRMT1. C, SAF-A and PRMT1 are co-immunoprecipitated with hnRNP-C1/2. Total cell extract (Input) was immunoprecipitated (IP) with antibodies against endogenous hnRNP-C1/2 protein, and precipitated proteins were detected by Western blot analysis using anti-hnRNP-C, anti-SAF-A, and anti-PRMT1 antibodies, respectively. pretreated with Adox to inhibit methylation. In these cells, a high amount of PRMT1 was found in association with the 40 S hnRNP particle and multimers thereof, whereas the amount of soluble PRMT1 was significantly reduced. We conclude that PRMT1 preferentially associates with the hypomethylated hnRNP complex by protein-protein interaction that is independent of the actual methylation reaction.
Methylation by PRMT1 primarily occurs in amino acid sequences RG and RGG, of which 15 of 19 in total SAF-A are located in the carboxyl-terminal sixth of the protein. The most prominent clustering of such sites is present in the short region from amino acids 778 to 793 in SAF-A, where four RG sites are found within 16 residues (Fig. 5A, wt sequence). This peptide, when expressed in E. coli as a fusion with a synthetic protein A binding site (Fig. 5A, lower panel) or with maltose binding protein (not shown), effectively serves as an in vitro substrate for methylation by recombinant PRMT1 but not PRMT2, which served as a control. Methylation occurs only in this peptide but not the fusion partner, as a similar construct with the relevant arginine residues replaced by lysines is not methylated. This finding confirms that methylation of SAF-A is catalyzed by PRMT1, suggesting that the region between amino acids 778 and 793 of SAF-A might also be a target for PRMT1 in vivo. If this was the case, it should be possible to discriminate between methylated and unmethylated SAF-A by probing this region. We therefore created a polyclonal antibody against an (unmethylated) synthetic 16-mer peptide of the wild type sequence RGNYNQNFRGRGNNRG. Fig. 5B shows that the resulting antibody detects SAF-A in Adox-treated HEK293 cells but did Extracts from cells treated with oxidized adenosine (ϩAdox) or untreated cells (ϪAdox) were investigated in parallel. Note that PRMT1 is enriched in high molecular weight complexes in Adox-treated cells. B, 40 S hnRNP particles were purified by glycerol gradient centrifugation and probed for SAF-A and PRMT1, demonstrating a preferred association of PRMT1 with the hypomethylated hnRNP complex.

FIG. 5. A novel antibody against a cluster of potential methylation sites allows discriminating between methylated and unmethylated SAF-A.
A, a peptide corresponding to amino acids 778 and 793 of SAF-A (wild type (wt) sequence, containing four potential methylation sites) and a mutated variant (mut) were bacterially expressed as fusion proteins with a synthetic protein A-binding site and used for in vitro methylation assays with S-adenosyl-L-[methyl-3 H]methionine and recombinant PRMT1 and PRMT2. Enzymatic activity of the two PRMTs was monitored in parallel assays with hypomethylated cell lysate as substrate. Note that PRMT2 does not exhibit detectable methyltransferase activity on any tested substrate as described earlier (40) and does also not methylate the SAF-A substrate. B, rabbits were immunized with a synthetic peptide of the wild type sequence and developed antibodies that detected unmethylated but not methylated SAF-A. Identical amounts of extracts from Adox-treated cells and untreated control cells were separated by SDS-PAGE, blotted, and probed with antibodies against total SAF-A (K371) or the synthetic peptide (A232). Note that A232 does not produce any signal on extract from untreated cells. An identical result was obtained for immunoprecipitated Myc-tagged SAF-A probed with either anti-Myc or the peptide antibody (right panel). C, unmethylated SAF-A was detected by immunofluorescence with A232 antibody in Adox treated cells (ϩAdox) but not untreated controls (ϪAdox). Note the identical subcellular localization of unmethylated (A232, ϩAdox) and methylated (K371, ϪAdox) SAF-A. not reveal any signal in extracts from untreated cells; in contrast, an antibody against total SAF-A gave identical signals for both extracts. Thus, the peptide antibody is methylationsensitive and allows the detection of unmethylated SAF-A only. This is confirmed in immunofluorescence experiments, where the peptide antibody reacts only on Adox-treated cells but not on untreated controls (Fig. 5C). In addition, the immunofluorescence experiment also demonstrates that the subcellular localization of SAF-A is independent of its methylation status, as the nuclear staining pattern is identical for both antibodies. It is also important to note that the total amount of SAF-A is not changed after treatment of cells with the methylation inhibitor Adox. This suggests that methylation does not affect the stability of SAF-A and that there is also no up-or downregulation of SAF-A expression.
Characterization of the antibody against unmethylated SAF-A revealed no signal in cultured HEK293 cells even after prolonged exposure. Thus, the amount of unmethylated SAF-A in these cells is undetectably small. We wondered whether SAF-A, which is a housekeeping protein present in all nucleated cells, is differently methylated in cells from different origins. This could provide insight into the regulation of arginine methylation in different cells and could also facilitate determining its functional role in vivo. Using the new methylationsensitive antibody as a tool, we therefore investigated 23 cultured human and one simian cell line for the methylation status of SAF-A (Fig. 6). Selected cell lines contained cells growing in suspension (e.g. Jurkat T-lymphocytes, CA46 Blymphocytes, and HL-60 promyeloblasts) as well as adherent cells (e.g. HeLa cervix adenocarcinoma cells, TE671 cerebellar medulloblastoma cells, and Saos-2 osteosarcoma cells). Investigated lines were chosen to be as diverse as possible and were very heterogeneous with respect to origin (solid tumors of e.g. liver, colon, and epidermis, different forms of leukemia, and normal tissue), morphology (e.g. epithelial, fibroblastic, lymphoblastic, myeloblastic), and culture conditions. While SAF-A was detectable in all of these lines by an antibody against total SAF-A, we did not find a single cell line with detectable amounts of unmethylated SAF-A. An identical result was also obtained with normal human peripheral blood lymphocytes of the authors (data not shown). We conclude that the methylation of SAF-A is a constitutive, ubiquitous, and quantitative modification.

DISCUSSION
In this article, we demonstrate that the predominant arginine methyltransferase of human cells, PRMT1, is associated physically with hnRNP particles and is able to methylate its target proteins in situ. The interaction of PRMT1 with hnRNP complexes occurs via SAF-A, a well known constituent of hnRNP complexes and the nuclear scaffold, and is strongly enhanced when methylation is blocked by periodate-oxidized adenosine. Physical interaction of PRMT1 with its major substrates, the components of hnRNP particles, might provide a means to ensure a high level of methylation of the preferred target proteins and might also contribute to substrate specificity.
Methylation of arginine residues in proteins is a widespread posttranslational modification, but its functional consequences are much less understood than those of e.g. phosphorylation or acetylation (for review, see ref. 9). In contrast to these other modifications, arginine methylation is currently assumed to be an irreversible process, as no protein demethylase has yet been identified in mammalian cells. In addition, the majority of target proteins is extensively methylated by constitutively active methyltransferases, with no apparent regulation process that would restrict methylation to specific cellular conditions such as the cell cycle (14). In this article, we have used a novel methyl-sensitive peptide antibody to determine how much of a particular PRMT1 target protein, SAF-A, is methylated in vivo.
In very good agreement with data from other target proteins (3)(4)(5), we have found that SAF-A is quantitatively methylated in all investigated cells lines. This leaves very little room for speculation that arginine methylation might be a regulatory modification, where at least some degree of dynamics must be inferred. In addition, our data suggest that unmethylated SAF-A protein can exist only in a very small time window directly after translation. The quantitative and apparently unregulated nature of arginine methylation would be compatible with a more static function, for example in protein folding, stability, or localization. In fact, the effects of methylation on the subcellular localization have been reported for a variety of proteins including RNA helicase A, Sam68, and hnRNP-A (7,43,44). In the context of our studies on SAF-A, however, we did not see any significant difference in the localization of methylated versus nonmethylated protein (Fig. 5). In addition, SAF-A seems to be present in identical amounts in both untreated and Adox-treated cells, where methylation is blocked, suggesting that the absence of methylation in SAF-A does not affect its stability. There is also no up-regulation of the SAF-A protein level, which could compensate for the lack of methylation if the activity of methylated and unmethylated protein differed. However, we found that PRMT1 becomes enriched on hnRNP complexes under conditions in which PRMT1 is present but cannot perform its enzymatic reaction because of the lack of S-adenosylmethionine. Thus, the binding of PRMT1 to hnRNP complexes is independent of the actual catalytic activity, and the observed interaction is more complex than a simple transient enzyme-substrate contact. When such complexes were purified by immunoprecipitation and S-adenosylmethionine was added, PRMT1 readily methylated SAF-A and other attached substrate proteins in place. This is clearly at variance with earlier experiments that failed to detect methyltransferase activity in hnRNP complexes (1). However, this apparent discrepancy is easily resolved by the fully methylated state of hnRNP proteins unless methylation is inhibited before complex purification. Our finding of a direct interaction between PRMT1 and SAF-A might also explain why PRMT1 has consistently been detected in high molecular weight complexes (1, 12, 25).

FIG. 6. SAF-A is fully methylated in a wide variety of cell lines.
Total cell extracts from 23 human and one simian (COS7) cell line were resolved by SDS-PAGE, blotted, and probed with antibodies against total SAF-A (K371) or the novel methyl-sensitive peptide antibody, A232. Note that SAF-A is ubiquitous, but none of the investigated cell lines contains an A232-detectable amount of unmethylated SAF-A.
As in the case of other target proteins, the reason for the methylation of SAF-A is currently unclear. There is a significant discrepancy between the high constitutive methylation activity in all investigated cell lines types (see Fig. 6) and the recent finding that this methylation is dispensable for basic cellular functions, as ES cells from Prmt1 knock-out mouse embryos are viable (5). Nevertheless, methylation by PRMT1 is essential, as knock-out embryos die shortly after implantation, and the ES cells cannot differentiate in vitro (5). This discrepancy can possibly be resolved by the idea that arginine methylation acts as a protein "tag." In contrast to a classical "signal," a tag could be irreversibly attached to a protein to label it for its entire lifetime. A tag could convey information about the protein, which is dispensable for most cellular processes but become important when cellular conditions change, e.g. during differentiation processes. This is comparable to a price tag on a product, which is not necessary for its function per se but is important only when the product is passed from one person to the next. In the case of the methylation tag, proteins might become labeled "functional," "properly folded," or "correctly localized" after they have reached their final destination for the first time. This is compatible with our finding that PRMT1 is a component of hnRNP complexes, the other components of which are the main target proteins of the enzyme. Thus, a newly synthesized hnRNP protein could be methylated as soon as it successfully enters the complex. If that assumption was correct, an almost complete methylation of target proteins would be expected in "undisturbed" cells, which is fully in line with our results and that of other publications (3)(4)(5). We can currently only speculate about the processes for which the methylation tag might become important, and how the readout of the enclosed information is accomplished. We suppose that the affected process(es) will be as ubiquitous as the methylation itself and can potentially affect any cell. On the other hand, DNA replication and transcription and all other vital processes occur in the absence of methylation by PRMT1 (which cannot be compensated by other methyltransferases) in knock-out cells (5,14). Protein folding and the assembly of functional complexes also do not seem to be affected, as demonstrated by the ability of knock-out cells to form normal hnRNP complexes from unmethylated components. These findings suggest that methylation might be more important for changing the status of a cell rather than for maintaining it, for example by keeping cells flexible to react to extracellular stimuli. Presumably, methylated proteins are preferentially reused to save energy or resources for the synthesis of newly expressed gene products during differentiation and developmental processes. This would be compatible with the finding that prmt1Ϫ/Ϫ ES cells lack the ability to differentiate in vitro and cannot give rise to embryos after implantation. Our future experiments will focus on testing this hypothesis.