Methylation of arginine residues is a widespread post-translational modification of proteins catalyzed by a small family of protein arginine methyltransferases (PRMTs). Functionally, the modification appears to regulate protein functions and interactions that affect gene regulation, signalling and subcellular localization of proteins and nucleic acids. All members have been, to different degrees, characterized individually and their implication in cellular processes has been inferred from characterizing substrates and interactions. Here, we report the first comprehensive comparison of all eight canonical members of the human PRMT family with respect to subcellular localization and dynamics in living cells. We show that the individual family members differ significantly in their properties, as well as in their substrate specificities, suggesting that they fulfil distinctive, non-redundant functions in vivo. In addition, certain PRMTs display different subcellular localization in different cell types, implicating cell- and tissue-specific mechanisms for regulating PRMT functions.

Introduction

Post-translational modification of proteins is an important means to expand the structural and functional diversity of the proteome. Depending on their function, post-translational modifications can either be transient or stable. Although transient modifications, such as phosphorylation or acetylation, form the basis of cellular signaling and act as reversible activity switches, stable modifications such as myristylation, glycosylation, or proteolytic processing, usually affect the physicochemical properties of proteins, often without direct impact on protein activity. The methylation of arginine residues appears to be a rather stable, but nonetheless dynamic modification (Bedford, 2007; Bedford and Richard, 2005; Fackelmayer, 2005b; Wysocka et al., 2006). It has been implicated in a large variety of important cellular functions, such as signaling (Berthet et al., 2002; Blanchet et al., 2006; Mowen et al., 2004; Xu et al., 2004), DNA repair (for a review, see Lake and Bedford, 2007), maturation and nucleocytoplasmic transport of RNA (Boisvert et al., 2002; Cheng et al., 2007; Shen et al., 1998; Yanagida et al., 2004), protein protection (Fackelmayer, 2005b), ribosomal assembly (Bachand and Silver, 2004; Swiercz et al., 2007; Swiercz et al., 2005) and the regulation of gene expression (e.g. An et al., 2004; Balint et al., 2005; Iberg et al., 2008; Xu et al., 2004). However, the molecular mechanisms connecting the modification to any functional consequences are only beginning to emerge. With regard to the regulation of gene expression, arginine methylation appears to act as an epigenetic mark on chromatin, similar to lysine methylation, which contributes to the histone code and affects expression of the underlying genomic information. In this case, methylated arginines on histones (reviewed by Ruthenburg et al., 2007), and possibly non-histone proteins involved in nuclear architecture (Fackelmayer, 2005a; Herrmann et al., 2004), are recognized by protein adaptors termed `effectors', which contain domains that specifically bind to methylated arginine (or lysine) residues. Well-characterized examples for these domains or `modules' are the chromodomain, tudor domain or the PHD finger (reviewed by Ruthenburg et al., 2007). These modules recognize different sets of methylated histones and recruit functional multi-protein complexes to chromatin to exert the appropriate effect, such as the activation or repression of a particular gene. Less is known about the mechanisms by which arginine methylation affects functions unrelated to gene expression, but it is clear that they also involve discrimination of methylated versus non-methylated arginine residues by other proteins, and thereby regulate their function.

To understand the impact of arginine methylation on cellular events, efforts to isolate, clone and characterize the enzymes responsible for this modification have been ongoing in many laboratories, and led to the discovery of a small, distinct family of protein arginine methyltransferases (PRMTs). In human cells, the PRMT family consists of eight canonical members [for an overview, see Bedford (Bedford, 2007)]. They catalyze the addition of one or two methyl groups to the guanidino nitrogen atoms of arginine, resulting in either ω-NG-monomethylarginine, ω-NG,NG-asymmetric or ω-NG,N′G-symmetric dimethylarginine. PRMTs have been classified into two groups based on the end product, both of which first catalyze the formation of monomethylarginine as an intermediate. In a second step, the type I enzymes PRMT1 (Lin et al., 1996), PRMT3 (Tang et al., 1998), CARM1/PRMT4 (Chen et al., 1999), PRMT6 (Frankel et al., 2002) and PRMT8 (Lee et al., 2005a; Sayegh et al., 2007) lead to the formation of asymmetric dimethylarginine, whereas the type II enzyme PRMT5 (Branscombe et al., 2001; Pollack et al., 1999) and possibly PRMT7 (Lee et al., 2005b; Miranda et al., 2004) lead to symmetric dimethylarginine. One more member of the PRMT family, PRMT2, was identified by sequence homology to the known enzymes (Katsanis et al., 1997), but has not been classified into a group, because no enzymatic activity has yet been demonstrated or substrates identified; nevertheless, PRMT2 acts as a coactivator in hormone-dependent gene expression (Meyer et al., 2007; Qi et al., 2002). Recently, an additional arginine methyltransferase activity has been described for FBXO11 (Cook et al., 2006). Unlike the canonical PRMTs, FBOX11 is not a member of the seven-β-strand methyltransferase family. In addition, no methyltransferase activity was reported for the human or worm FBOX11 proteins (Fielenbach et al., 2007). Consequently, further work is needed to determine whether FBXO11 is a PRMT, and we have not studied it the context of this article. In plants, PRMT10 and PRMT11 have been described, which are true members of the PRMT family, but have no known human homologs (Niu et al., 2007; Scebba et al., 2007).

Available data suggest that most PRMTs are housekeeping enzymes, which are ubiquitously expressed in most – if not all – cell types and tissues of the human body. The only known exception is PRMT8, which appears to be restricted to neurons in the brain (Lee et al., 2005a; Sayegh et al., 2007; Taneda et al., 2007). Importantly, PRMTs differ in their substrate specificities, and are therefore probably involved in different physiological processes (for a review, see Bedford, 2007). This has clearly been shown for the two best-characterized family members, PRMT1 and CARM1/PRMT4, where the genetic knockout of either enzyme results in an accumulation of hypomethylated substrates in vivo, demonstrating that these two PRMTs cannot substitute for each other (Kim et al., 2004; Pawlak et al., 2002). The activity of both enzymes is essential for distinct differentiation processes, but is apparently dispensable for basic cellular metabolism and survival (Pawlak et al., 2000; Yadav et al., 2003). In this context, it is interesting to note that the two recently identified PRMTs in plants, PRMT10 and PRMT11, are also required for differentiation processes such as flowering, but are dispensable for basic cellular life (Niu et al., 2007; Scebba et al., 2007).

Fig. 1.

Expression and activity of human PRMT1-PRMT8 fused to EGFP. (A) Schematic representation of the eight canonical members of the human PRMT family. The highly conserved PRMT core region (gray), signature motifs I, post I, II, and III (black) and the conserved THW loop (red), present in all PRMTs, are indicated. Note that PRMT7 has a duplication of these motifs and PRMT2 and PRMT3 have an N-terminal SH3 domain (orange) or a zinc-finger domain (green), respectively. The nuclear localization signal (NLS, blue) targets PRMT6 to the nucleus, whereas N-terminal myristoylation (∼) tethers PRMT8 to the plasma membrane. The size of individual PRMTs is indicated at each C-terminus. (B) Human embryonic kidney (HEK293) cells were stably transfected with expression vectors encoding PRMT1-GFP to PRMT8-GFP. Total cell extracts were separated by SDS-PAGE, blotted to a polyvinylidene difluoride membrane, and probed with antibodies specific for GFP. Equal amounts of total cell extract were used in each lane, verified by using hnRNP-C as a loading control. (C) Methylation activity assays. Individual PRMTs were immunoprecipitated from the total cell extracts described in B, by using anti-GFP antibodies. Activity assays were carried out with hypomethylated extracts (upper panel) or recombinant core histones (lower panel) as described in the Materials and Methods.

Fig. 1.

Expression and activity of human PRMT1-PRMT8 fused to EGFP. (A) Schematic representation of the eight canonical members of the human PRMT family. The highly conserved PRMT core region (gray), signature motifs I, post I, II, and III (black) and the conserved THW loop (red), present in all PRMTs, are indicated. Note that PRMT7 has a duplication of these motifs and PRMT2 and PRMT3 have an N-terminal SH3 domain (orange) or a zinc-finger domain (green), respectively. The nuclear localization signal (NLS, blue) targets PRMT6 to the nucleus, whereas N-terminal myristoylation (∼) tethers PRMT8 to the plasma membrane. The size of individual PRMTs is indicated at each C-terminus. (B) Human embryonic kidney (HEK293) cells were stably transfected with expression vectors encoding PRMT1-GFP to PRMT8-GFP. Total cell extracts were separated by SDS-PAGE, blotted to a polyvinylidene difluoride membrane, and probed with antibodies specific for GFP. Equal amounts of total cell extract were used in each lane, verified by using hnRNP-C as a loading control. (C) Methylation activity assays. Individual PRMTs were immunoprecipitated from the total cell extracts described in B, by using anti-GFP antibodies. Activity assays were carried out with hypomethylated extracts (upper panel) or recombinant core histones (lower panel) as described in the Materials and Methods.

Given the important role of arginine methylation, surprisingly little is known about the cell biology of the PRMT enzymes. In particular, data is scarce with respect to subcellular localization, dynamics and complex formation of PRMTs, which might be central determinants of their regulation and function in vivo. We have recently demonstrated that a variant of PRMT1 is primarily localized in the cytoplasm, spatially separated from the main pool of its substrates, hnRNP proteins and core histones in the nucleus (Herrmann et al., 2005). Importantly, these experiments also demonstrated that PRMT1 is a highly dynamic enzyme with variable subcellular localization and mobility, dependent on the methylation status of its substrate proteins. In ongoing efforts to determine the physiological role of the individual members of the PRMT family, we report here on the first comprehensive comparison of subcellular localization and dynamics of all eight canonical PRMT family members in human cells. Our results provide evidence that the individual members of the family differ significantly in their in vivo properties, compatible with distinct physiological functions.

Results

The methylation of arginine residues in proteins has been intensively studied over the past few years. Many substrates, as well as the family of enzymes that catalyze the methylation reaction, have been identified, revealing that arginine methylation is an important regulator of protein function in a variety of fundamental cellular processes. However, little is known about the properties of protein arginine methyltransferases in live cells. In addition, up to now, studies usually focussed on the properties and functional roles of individual members of the PRMT family. We therefore decided to compare, in a comprehensive way, key properties of all eight canonical PRMT family members under identical conditions. As a first step towards in vivo measurements, we created human embryonic kidney (HEK293) cell lines stably expressing GFP fusion proteins for all eight members of the PRMT family (Fig. 1A). After four to six weeks of G418 selection, populations containing expressing cells were obtained by fluorescence-activated cell sorting (FACS). To ensure that the GFP fusion proteins behave like their endogenous counterparts and can be used with confidence to investigate their properties, expressed proteins where checked by a variety of biochemical means (see Herrmann et al., 2005). All eight PRMT-GFP fusion proteins were well expressed, and migrated at their expected molecular mass on SDS gels with no apparent degradation (Fig. 1B). The expression level of the individual stable cell lines differed noticeably, with PRMT3, PRMT5, PRMT7 and PRMT8 being weaker than PRMT1, PRMT2, PRMT4 and PRMT6. As all eight proteins were expressed equally well in transient transfections, we assume that the different levels of expression in stable lines indicate that elevated levels of PRMT3, PRMT5, PRMT7 and PRMT8 lead to a growth disadvantage, and were thus overgrown by weaker expressing cells. To measure enzymatic activity, we immunoprecipitated all constructs by using an anti-GFP antibody, and used them together with hypomethylated total cell extract (Fig. 1C, upper panel) or recombinant histones H3 and H4 (Fig. 1C, lower panel) as substrates in radioactive methylation assays. As expected, strong activity was observed for PRMT1, followed by PRMT4, PRMT6 and PRMT8. Weak activity, using total cell extract as a substrate, was seen for PRMT5 and PRMT2, whereas PRMT3 and PRMT7 did not show detectable activity in these assays. This result was not surprising, because both enzymes are known to have very low specific activity (Miranda et al., 2004; Tang et al., 1998).

Because the formation of homo-oligomeric complexes is a known prerequisite for enzymatic activity of several PRMTs (Higashimoto et al., 2007; Lim et al., 2005; Zhang and Cheng, 2003; Zhang et al., 2000), we determined the migration of our PRMT constructs on glycerol gradients, for comparison with the migration of the endogenous proteins (Fig. 2). Extracts from cells stably expressing PRMT1 to PRMT7 were sedimented in glycerol gradients; an extract from non-transfected cells was sedimented in parallel. PRMT8 was omitted from this experiment, because it binds to the plasma membrane (Lee et al., 2005a) and cannot be extracted in a membrane-free, soluble form. Fractions from the gradients were collected and probed with anti-GFP antibodies for the PRMT constructs, or with commercial antibodies recognizing endogenous PRMT1 to PRMT7; these experiments were performed for all PRMTs separately (not shown), and, with identical results, with a mixture of extracts from the seven cell lines expressing the proteins, to allow for a direct comparison and to rule out potential gradient-to-gradient variability (shown in Fig. 2, upper panel). We found that all GFP fusion proteins, with the exception of PRMT5 (see below), sediment in identical or very similar way to the endogenous proteins. PRMT3, PRMT6 and PRMT7 sediment in the gradient as expected for a monomeric protein or a small oligomer, whereas PRMT1 and PRMT4, which are known to form homo-oligomers and associate with a variety of other proteins, sediment as high molecular mass complexes. These results indicate that all GFP fusion proteins, except for PRMT5, are properly folded and oligomerize in their correct functional context. Thus, they are suitable tools to investigate the localization and dynamics of PRMTs in vivo. However, PRMT5-GFP did not behave in a similar manner to the endogenous counterpart; it sedimented as if it was part of a very large heterogenous complex, suggesting that it is prone to aggregation (see supplementary material Fig. S2). In vivo mobility measurements for PRMT5 are shown for completeness only (see below), but do not represent the behavior of the endogenous enzyme.

Fig. 2.

Determination of complex sizes of PRMT1-PRMT7 by glycerol gradient centrifugation. A mixture of total cell extracts from cells stably expressing PRMT1-GFP to PRMT7-GFP fusion proteins was loaded on a 10-30% glycerol gradient and proteins were separated by centrifugation for 20 hours at 30,000 r.p.m. The gradient was fractionated from the top, and aliquots of individual fractions were resolved by SDS-PAGE. Proteins were detected by western blotting using anti-GFP antibodies for detection (top panel). PRMT-GFP fusion proteins are denoted by their respective numbers (1-7). The sedimentation of endogenous PRMTs was investigated similarly, using extracts from nontransfected cells and commercially available antibodies for PRMT1 to PRMT7. BSA (66 kDa, 4.2S), β-amylase (200 kDa, 8.9S) and apoferritin (443 kDa, 17.6S) were used as sedimentation markers in an identical gradient run in parallel. Note that the expressed GFP fusion proteins of all PRMTs except PRMT5 sediment similarly to their endogenous counterparts.

Fig. 2.

Determination of complex sizes of PRMT1-PRMT7 by glycerol gradient centrifugation. A mixture of total cell extracts from cells stably expressing PRMT1-GFP to PRMT7-GFP fusion proteins was loaded on a 10-30% glycerol gradient and proteins were separated by centrifugation for 20 hours at 30,000 r.p.m. The gradient was fractionated from the top, and aliquots of individual fractions were resolved by SDS-PAGE. Proteins were detected by western blotting using anti-GFP antibodies for detection (top panel). PRMT-GFP fusion proteins are denoted by their respective numbers (1-7). The sedimentation of endogenous PRMTs was investigated similarly, using extracts from nontransfected cells and commercially available antibodies for PRMT1 to PRMT7. BSA (66 kDa, 4.2S), β-amylase (200 kDa, 8.9S) and apoferritin (443 kDa, 17.6S) were used as sedimentation markers in an identical gradient run in parallel. Note that the expressed GFP fusion proteins of all PRMTs except PRMT5 sediment similarly to their endogenous counterparts.

Localization of PRMT-GFP fusion proteins in live cells

The subcellular localization of members of the PRMT family has been described previously on an individual basis, but often results even for the same PRMT are quite variable (Cote et al., 2003; Frankel et al., 2002; Herrmann et al., 2005; Tang et al., 1998). In the case of PRMT1, most inconsistencies have recently become comprehensible, when Goulet and colleagues reported that PRMT1 isoforms produced by alternative splicing localize differently (Goulet et al., 2007); for our experiments, splicing variant 2 was used, designated PRMT1v2, which has an N-terminal nuclear export signal. Our cell lines stably expressing GFP fusion proteins were an ideal tool to compare the localization of all family members under identical conditions in live cells (Fig. 3). We found that PRMT1v2, PRMT2, PRMT3, PRMT4 and PRMT7 were found both in the nucleus and in the cytoplasm, ranging from a predominant nuclear (PRMT2) to a predominantly cytoplasmic (PRMT1v2, PRMT3, PRMT4 and PRMT7) localization. PRMT6 is exclusively localized in the nucleus, whereas PRMT5 is only cytoplasmic and PRMT8 is tethered to the plasma membrane. PRMT2 was found predominantly in the nucleus, and was enriched in subnuclear structures resembling nuclear speckles (supplementary material Fig. S3A). In most cases, these localizations were in line with earlier reports (Frankel et al., 2002; Lee et al., 2005a; Meyer et al., 2007; Rho et al., 2001; Tang et al., 1998), but especially in the case of PRMT1 and CARM1/PRMT4 there was some discrepancy that we wanted to investigate further. We hypothesized that the localization could be cell-type specific, and created stable cell lines for PRMT1v2 and CARM1/PRMT4 from often-used parental cell lines U2OS, HeLa and MCF7. After identical selection and cell sorting, the cell lines were investigated for localization of the fusion proteins (Fig. 4). Indeed, the localization of both PRMT1v2 and CARM1/PRMT4 differed fundamentally in different cell types. Interestingly, they also did it in an identical way: in U2OS osteosarcoma cells, PRMT1v2 and PRMT4 were predominantly cytoplasmic, as in the original HEK293 embryonic kidney cells, but some cells displayed enrichment in the nucleus. Further investigations suggested that these differences are due to a cell-cycle-dependent relocation, especially in the case of PRMT4 (supplementary material Fig. S4). By contrast, in stably transfected HeLa cervix carcinoma and MCF7 breast cancer cell lines, both PRMT1v2 and CARM1/PRMT4 were predominantly nuclear. We conclude that the localization of the PRMTs is cell-type specific, and can also dynamically change in the cell cycle or under certain conditions (see Herrmann et al., 2005).

Fig. 3.

Subcellular distribution of PRMTs in HEK293 cells. Confocal slices through typical cells that stably express PRMT1-GFP to PRMT8-GFP fusion proteins. For reference, the same cells were analyzed by differential interference contrast (DIC). Scale bar: 25 μm.

Fig. 3.

Subcellular distribution of PRMTs in HEK293 cells. Confocal slices through typical cells that stably express PRMT1-GFP to PRMT8-GFP fusion proteins. For reference, the same cells were analyzed by differential interference contrast (DIC). Scale bar: 25 μm.

Dynamics of PRMT family members in vivo

To gain qualitative insight and quantitative data on the intracellular dynamics of the PRMT family members, we performed photobleaching experiments (FRAP, fluorescence recovery after photobleaching) as described earlier for PRMT1 (Herrmann et al., 2005). For each member of the PRMT family we measured 15 to 20 individual cells, using a rectangular bleaching area encompassing both the cytoplasm and the nucleus; typical examples of the bleaching and the recovery over time are given in Fig. 5. Pixel intensities in the bleached region (and, for reference and normalization, in an unbleached region of the same cell) were determined, and graphically represented (Fig. 6) to show the kinetics of recovery. Furthermore, data from the recovery were fitted to a modified exponential association function (supplementary material Fig. S1) to determine the half-time of fluorescence recovery (t/2) and mobile fraction for all PRMTs. The recovery half-time allowed calculation of the diffusion coefficient for every PRMT (Table 1A), separately for the nucleus and the cytoplasm, or, in the case of PRMT8, on the plasma membrane.

Table 1.

Quantitative diffusion analysis of PRMT photobleaching experiment

Protein n Cytoplasm t/2 ± s.e.m. (seconds) Nucleus t/2 ± s.e.m. (seconds) Cytoplasm D ± s.e.m. (μm2/second) Nucleus D ± s.e.m. (μm2/second)
A. Analysis in untreated HEK293 cells       
PRMT1   20   2.75±0.21   4.61±0.42   1.83±0.12   1.20±0.12  
PRMT2   17   0.97±0.08   3.51±0.33   5.28±0.44   1.56±0.16  
PRMT3   15   0.90±0.08   0.95±0.11   5.53±0.44   5.55±0.52  
PRMT4   15   1.61±0.13   2.63±0.29   3.12±0.21   1.98±0.17  
PRMT5   12   3.59±0.36   n.d.   1.44±0.15   –  
PRMT6   16   n.d.   3.94±0.13   –   1.18±0.04  
PRMT7   20   0.58±0.07   1.04±0.08   9.21±0.64   4.99±0.38  
PRMT8   16   24.78±2.09*  n.d.   0.21±0.03*  n.d.  
EGFP   8   0.23±0.07   0.56±0.08   22.18±3.14   9.59±1.20  
βGal-NLS   19   n.d.   2.68±0.14   –   1.80±0.09  
B. Analysis in HEK293 cells treated with 15 μM oxidized adenosine for 48 hours       
PRMT1   24   2.61±0.16   4.01±0.29   1.93±0.13   1.26±0.07  
PRMT2   21   1.67±0.10   5.35±0.35   2.98±0.19   0.92±0.05  
PRMT3   13   1.28±0.11   2.10±0.22   3.86±0.27   2.72±0.36  
PRMT4   18   1.61±0.12   4.69±0.42   3.09±0.21   1.14±0.11  
PRMT5   17   3.32±0.39   n.d.   1.58±0.12   –  
PRMT6   24   n.d.   3.10±0.17   –   1.59±0.09  
PRMT7   11   0.65±0.07   0.97±0.11   7.82±0.64   5.82±0.74  
PRMT8   –   n.d.   n.d.   n.d.   n.d.  
EGFP   –   n.d.   n.d.   n.d.   n.d.  
βGal-NLS   19   n.d.   2.46±0.11   –   1.94±0.09  
Protein n Cytoplasm t/2 ± s.e.m. (seconds) Nucleus t/2 ± s.e.m. (seconds) Cytoplasm D ± s.e.m. (μm2/second) Nucleus D ± s.e.m. (μm2/second)
A. Analysis in untreated HEK293 cells       
PRMT1   20   2.75±0.21   4.61±0.42   1.83±0.12   1.20±0.12  
PRMT2   17   0.97±0.08   3.51±0.33   5.28±0.44   1.56±0.16  
PRMT3   15   0.90±0.08   0.95±0.11   5.53±0.44   5.55±0.52  
PRMT4   15   1.61±0.13   2.63±0.29   3.12±0.21   1.98±0.17  
PRMT5   12   3.59±0.36   n.d.   1.44±0.15   –  
PRMT6   16   n.d.   3.94±0.13   –   1.18±0.04  
PRMT7   20   0.58±0.07   1.04±0.08   9.21±0.64   4.99±0.38  
PRMT8   16   24.78±2.09*  n.d.   0.21±0.03*  n.d.  
EGFP   8   0.23±0.07   0.56±0.08   22.18±3.14   9.59±1.20  
βGal-NLS   19   n.d.   2.68±0.14   –   1.80±0.09  
B. Analysis in HEK293 cells treated with 15 μM oxidized adenosine for 48 hours       
PRMT1   24   2.61±0.16   4.01±0.29   1.93±0.13   1.26±0.07  
PRMT2   21   1.67±0.10   5.35±0.35   2.98±0.19   0.92±0.05  
PRMT3   13   1.28±0.11   2.10±0.22   3.86±0.27   2.72±0.36  
PRMT4   18   1.61±0.12   4.69±0.42   3.09±0.21   1.14±0.11  
PRMT5   17   3.32±0.39   n.d.   1.58±0.12   –  
PRMT6   24   n.d.   3.10±0.17   –   1.59±0.09  
PRMT7   11   0.65±0.07   0.97±0.11   7.82±0.64   5.82±0.74  
PRMT8   –   n.d.   n.d.   n.d.   n.d.  
EGFP   –   n.d.   n.d.   n.d.   n.d.  
βGal-NLS   19   n.d.   2.46±0.11   –   1.94±0.09  

Recovery curves shown in Fig. 6 were fitted to a modified exponential association function as described in Materials and Methods, to allow determination of the half-time (t/2) of fluorescence recovery. Diffusion coefficients D were calculated from t/2 as described in Materials and Methods. t/2 and D are given as mean ± s.e.m. separately for cytoplasmic and nuclear fractions of individual PRMTs. PRMT8 was measured at the plasma membrane; the value is given in the `Cytoplasm' column for easier arrangement, and is marked with an asterisk. Between 12 and 22 cells (n) were analyzed for each PRMT. Monomeric EGFP (28 kDa) and tetrameric β-galactosidase–EGFP (600 kDa) were analyzed for reference

Fig. 4.

Subcellular distribution of PRMT1 and PRMT4 show cell-type specificity. Cell lines stably expressing PRMT1-GFP or PRMT4-GFP were created from HEK293 (human embronic kidney), U2OS (human osteosarcoma), HeLa (human cervix carcinoma) and MCF-7 (human breast adenocarcinoma) cells. Fields of typical cells for each construct and cell line are shown. Scale bar: 50 μm.

Fig. 4.

Subcellular distribution of PRMT1 and PRMT4 show cell-type specificity. Cell lines stably expressing PRMT1-GFP or PRMT4-GFP were created from HEK293 (human embronic kidney), U2OS (human osteosarcoma), HeLa (human cervix carcinoma) and MCF-7 (human breast adenocarcinoma) cells. Fields of typical cells for each construct and cell line are shown. Scale bar: 50 μm.

Fig. 5.

Photobleaching experiments reveal in vivo dynamics of PRMTs. Selected images from confocal FRAP experiments on HEK293 cells that stably express PRMT1-GFP to PRMT8-GFP. Typical cells are shown before and directly after photobleaching, and at distinct time points during fluorescence recovery. Bleaching was performed using a rectangular region spanning both the cytoplasm and the nucleus.

Fig. 5.

Photobleaching experiments reveal in vivo dynamics of PRMTs. Selected images from confocal FRAP experiments on HEK293 cells that stably express PRMT1-GFP to PRMT8-GFP. Typical cells are shown before and directly after photobleaching, and at distinct time points during fluorescence recovery. Bleaching was performed using a rectangular region spanning both the cytoplasm and the nucleus.

Mobile fractions of PRMT1 to PRMT7 were between 94% and 98%, comparable with GFP alone, demonstrating that the enzymes behave as completely soluble (i.e. not structurally bound) proteins in both the cytoplasm and the nucleus. For PRMT8, the mobile fraction in the membrane was 74%, in line with other membrane proteins (Bates et al., 2006). Interestingly, the mobility of individual members of the PRMT family was vastly different (Table 1). The most mobile enzymes were PRMT3 and PRMT7, with recovery half-times of 0.97±0.08 and 0.58±0.07 seconds in the cytoplasm and 0.95±0.11 and 1.04±0.08 seconds in the nucleus, respectively. Only these two members of the PRMT family behave, in live cells, like monomers or small oligomers when compared with a control protein, a tetrameric β-galactosidase-GFP construct with a molecular mass of approximately 600 kDa. Of the remaining PRMTs, which all show diffusion characteristics of high molecular weight complexes in these measurements, PRMT1 and PRMT4 (and PRMT5, but these data do not reflect the endogenous protein, see above) were least mobile in the cytoplasm, and PRMT1 and PRMT6 were least mobile in the nucleus. For all PRMTs that have both a cytoplasmic and a nuclear population, the nuclear population was less mobile than the cytoplasmic population, as expected, because of the known higher intranuclear viscosity (Wachsmuth et al., 2000) (and unpublished measurements from our laboratory). The only notable exception was PRMT3, which did not differ significantly between nucleus and cytoplasm, indicating that PRMT3 is slowed down by transient interactions with cytoplasmic proteins.

Identical in vivo mobility assays were repeated in cells that had been incubated with the methylation inhibitor periodate-oxidized adenosine (Adox) (Fig. 6; Table 1B), because we had recently found that this treatment leads to an immobile fraction of 26% of PRMT1 in the nucleus (Herrmann et al., 2005). This effect was reproduced here. In addition, the nuclear but not the cytoplasmic fraction of PRMT3 showed a similar effect, although to a weaker extent, leading to 12% immobile PRMT3 upon Adox treatment. The mobile fractions of all other PRMT family members were not affected by Adox treatment. Control experiments published earlier, using hnRNP-C and a β-galactosidase model protein showed that Adox treatment does not measurably affect intranuclear viscosity (Herrmann et al., 2005). We found, however, that diffusion of several PRMTs was considerably altered by Adox treatment (Table 1B). In particular, diffusion of nuclear PRMT2, PRMT3 and CARM1/PRMT4 was significantly (P<0.001) reduced, whereas a subtle, yet statistically significant (P<0.001), increase in the diffusion of nuclear PRMT6 was observed. In the cytoplasm, the mobilities of PRMT2, PRMT3 and PRMT7 were also decreased (P<0.001). For PRMT1, the difference was not significant, neither in the nucleus (P=0.055) nor in the cytoplasm (P=0.019). Also, no significant difference was found for cytoplasmic PRMT4 (P=0.685) and cytoplasmic PRMT5 (P=0.012). These results show that PRMT family members react differently to the accumulation of unmethylated substrates, and also that individual PRMTs are differently affected in different subcellular locations, probably reflecting the presence of cognate substrates with which transient interactions can occur.

Fig. 6.

Quantitative analysis of FRAP experiments. Mobility of PRMTs was quantified individually for the nuclear (A) and the cytoplasmic (B) fraction of the proteins with or without oxidized adenosine (Adox) treatment. Note that PRMT5 lacks a nuclear fraction, and PRMT6 lacks a cytoplasmic fraction. The recovery curve of membrane bound PRMT8 is shown separately (C). Each recovery curve represents the mean of 12-22 individual cells.

Fig. 6.

Quantitative analysis of FRAP experiments. Mobility of PRMTs was quantified individually for the nuclear (A) and the cytoplasmic (B) fraction of the proteins with or without oxidized adenosine (Adox) treatment. Note that PRMT5 lacks a nuclear fraction, and PRMT6 lacks a cytoplasmic fraction. The recovery curve of membrane bound PRMT8 is shown separately (C). Each recovery curve represents the mean of 12-22 individual cells.

PRMT complex size determined by FCS

In vivo mobility measurement of the eight PRMT family members demonstrated that, except for PRMT3 and PRMT7, all PRMTs have diffusion characteristics of multiprotein complexes. In these experiments, mobility is affected by many factors, including transient interactions with substrates or other binding partners, size and shape of the enzymatically active oligomer, and intracellular viscosities. To complement the in vivo measurements, we performed fluorescence correlation spectroscopy (FCS) measurements for all PRMTs in diluted ex-vivo cell extracts. Under these conditions, the effects of medium viscosity and transient interactions should be minimized or absent, allowing for measurement of diffusion characteristics of the enzymes in stable complexes. To this end, extracts from cells expressing individual PRMT-GFP fusion proteins were prepared, and diluted with isotonic buffer to 1-5 molecules per confocal volume of measurement. Identical experiments were performed with extracts from untreated cells and cells treated with Adox. Auto-correlation curves were obtained (Fig. 7) and quantified (Table 2) for PRMT1 to PRMT7, and for GFP alone as a calibration standard. All auto-correlation curves could be successfully fitted with a one-component algorithm, directly yielding a value for τd, which represents the characteristic diffusion time of the molecules through the measuring space. As described in the Materials and Methods, τd was used to determine the diffusion coefficient, and to estimate the hydrodynamic radius and the apparent molecular mass of the measured enzymes. Results show that Adox treatment does not affect the diffusion of any PRMT under these conditions (Table 2A), revealing that the effects of Adox on in vivo mobility shown in Table 1B do not reflect changes in the stable enzyme complexes, but rather in their transient interaction with substrates. These results are also in excellent agreement with those from glycerol gradient centrifugation and in vivo mobility measurements in untreated (non-inhibited) cells. They show that PRMT1 and PRMT4 are present in high molecular mass complexes, and that PRMT5-GFP forms large, presumably non-physiological aggregates. By contrast, FCS data for PRMT2 are compatible with the presence of a dimer, and data for PRMT3 suggest the existence of a tetramer (see Discussion). Clearly, PRMT6 and PRMT7 both behave as monomeric proteins under these conditions.

Table 2.

Quantitative diffusion analysis of fluorescence correlation spectroscopy experiments

Protein – Adox τ ± s.e.m. (μseconds) + Adox τ ± s.e.m. (μseconds)
A. Comparison of diffusion times (t) in extracts from untreated or Adox-treated cells       
PRMT1   272.6±6.5   278.4±14.7     
PRMT2   122.3±3.2   137.0±1.3     
PRMT3   167.8±6.2   159.3±7.0     
PRMT4   261.0±3.0   267.3±8.6     
PRMT5   641.2±33.5   710.0±14.0     
PRMT6   100.1±0.6   121.0±4.2     
PRMT7   116.8±2.5   108.5±2.0     
EGFP   70.3±0.9   n.d.     
Protein   τ ± s.e.m. (μseconds)  D ± s.e.m. (μm2/second)  rH ± s.e.m. (nm)   Approx. mass ± s.e.m. (kDa)   Mass monomer (kDa)  
B. Measured parameters for each PRMT in untreated cells       
PRMT1   272.6±6.5   23.3±0.6   8.3±0.2   1743±140   72  
PRMT2   122.3±3.2   51.8±1.5   3.7±0.1   157±14   78  
PRMT3   167.8±6.2   37.7±1.5   5.1±0.2   419±52   96  
PRMT4   261.0±3.0   24.2±0.2   7.9±0.1   1519±60   90  
PRMT5   641.2±33.5   10.0±0.6   19.5±1.1   23389±3729   100  
PRMT6   100.1±0.6   63.2±0.5   3.0±0.1   85±2   72  
PRMT7   116.8±2.5   54.2±1.3   3.6±0.1   137±10   115  
EGFP   70.3±0.9   90.2±1.1   2.1±0.1   29±1   30  
Protein – Adox τ ± s.e.m. (μseconds) + Adox τ ± s.e.m. (μseconds)
A. Comparison of diffusion times (t) in extracts from untreated or Adox-treated cells       
PRMT1   272.6±6.5   278.4±14.7     
PRMT2   122.3±3.2   137.0±1.3     
PRMT3   167.8±6.2   159.3±7.0     
PRMT4   261.0±3.0   267.3±8.6     
PRMT5   641.2±33.5   710.0±14.0     
PRMT6   100.1±0.6   121.0±4.2     
PRMT7   116.8±2.5   108.5±2.0     
EGFP   70.3±0.9   n.d.     
Protein   τ ± s.e.m. (μseconds)  D ± s.e.m. (μm2/second)  rH ± s.e.m. (nm)   Approx. mass ± s.e.m. (kDa)   Mass monomer (kDa)  
B. Measured parameters for each PRMT in untreated cells       
PRMT1   272.6±6.5   23.3±0.6   8.3±0.2   1743±140   72  
PRMT2   122.3±3.2   51.8±1.5   3.7±0.1   157±14   78  
PRMT3   167.8±6.2   37.7±1.5   5.1±0.2   419±52   96  
PRMT4   261.0±3.0   24.2±0.2   7.9±0.1   1519±60   90  
PRMT5   641.2±33.5   10.0±0.6   19.5±1.1   23389±3729   100  
PRMT6   100.1±0.6   63.2±0.5   3.0±0.1   85±2   72  
PRMT7   116.8±2.5   54.2±1.3   3.6±0.1   137±10   115  
EGFP   70.3±0.9   90.2±1.1   2.1±0.1   29±1   30  

Autocorrelation curves shown in Fig. 7 were fitted with the Zeiss FCS software module to determine the diffusion time t of individual PRMTs. All values are given as mean ± s.e.m. Using t values from untreated cells, the diffusion coefficients D, hydrodynamic radii rH and approximate mass of the molecules were calculated as described in the Materials and Methods. The monomeric molecular mass of each PRMT is given for reference. Note that PRMT6 and PRMT7 (and the control protein EGFP) have diffusion characteristics of monomeric proteins, PRMT2 behaves like a dimer, PRMT3 like a tetramer, and PRMT1 and PRMT4 are present in large complexes. PRMT5 appears to aggregate or associate with very large structures

Taken together, our results provide the first comprehensive comparison of all eight canonical human PRMTs with regard to localization, intracellular mobility in live cells and the size of the active complexes. They show that the enzymes all differ in their properties, and therefore apparently perform distinct, non-redundant functions in the cell.

Discussion

The family of protein arginine methyltransferases, with their eight canonical members in humans, has been functionally implicated in a wide variety of cellular processes. Here, we performed the first full-family comparison of the PRMT enzymes, because a profound and quantitative knowledge of their properties is essential to understand the functional diversity in the PRMT family.

We show that GFP fusion proteins can be used as in vivo tools for studies of intracellular localization and mobility of individual members of the PRMT family. Except for PRMT5-GFP, which appears to form large (presumably non-physiological) aggregates, fusions of the seven other members of the PRMT family behave like their endogenous counterparts. Thus, although we cannot completely rule out the idea that the GFP tag affects some properties of the tagged enzymes, such as the substrate specificity (Pawlak et al., 2002), we are confident that the fusion proteins well reflect the behavior of the endogenous proteins. Microscopic live-cell analysis demonstrated that PRMTs differ in their subcellular localization, mostly but not entirely in line with earlier data obtained for the individual PRMTs. For example, PRMT6 was found exclusively in the cell nucleus, owing to its nuclear localization signal (Frankel et al., 2002; Herrmann et al., 2005) and PRMT8 was found at the plasma membrane as a result of its N-terminal myristoylation (Lee et al., 2005a). PRMT5 was the only PRMT that we found to be completely cytoplasmic in our cells. This was compatible with earlier data from immunofluorescence experiments (Rho et al., 2001), but at variance with studies that also show functions for PRMT5 in the nucleus (Ancelin et al., 2006; Lacroix et al., 2008; Pal et al., 2003). As noted above, however, we assume that PRMT5-GFP is aggregated (see supplementary material Fig. S2) and therefore does not reflect the behavior of the endogenous protein. The remaining PRMT1, PRMT2, PRMT3, PRMT4 and PRMT7 are found in the cytoplasm as well as in the nucleus, although the relative amounts in these compartments differ significantly. The localization of PRMT1 in our experiments was especially interesting, because the laboratory of Jocelyn Coté had recently identified a functional nuclear export signal in the N-terminus of the splicing variant used in this article (Goulet et al., 2007). In fact, splicing variant 2 of PRMT1 (PRMT1v2), was localized predominantly in the cytoplasm in the HEK293 cells we originally investigated. This was also in line with earlier data from our laboratory showing that PRMT1 is approximately six- to eightfold more abundant in the cytoplasm than in the nucleus, but that this ratio is variable in response to the methylation status of the cell (Herrmann et al., 2005). More recently, we found that a catalytically inactive PRMT1 accumulates further in the nucleus (Herrmann and Fackelmayer, 2009). This suggests that the presence of a nuclear export signal alone is not sufficient to completely understand the behavior of PRMT1v2. In fact, the situation is more complex, because cells from other cell types revealed striking differences in the localization of PRMT1v2, after creating stable cell lines with the identical expression construct. We found that, despite the presence of the nuclear export signal, PRMT1v2 is predominantly nuclear in stably transfected HeLa and MCF-7 cells, and in a fraction of U2OS cells. This is compatible with cell-type-specific regulation of the intracellular shuttling of PRMT1 in and out of the nucleus. It is interesting to note that CARM1/PRMT4 behaves in a similar way, being a predominantly cytoplasmic protein in HEK293 cells, but a mainly nuclear protein in MCF-7 and HeLa cells. More work will be necessary to investigate this shuttling in detail, because it might provide a better understanding of the regulation of substrate proteins, and might link to pathophysiological consequences of its misregulation (Goulet et al., 2007). When the intracellular mobility of PRMT1 and CARM1/PRMT4 was investigated by FRAP analysis, we found that both proteins showed diffusion characteristics of large multiprotein complexes. These complexes probably consist of multiple copies of the same enzyme, which has previously been found to be a prerequisite for enzymatic activity (Higashimoto et al., 2007; Lim et al., 2005; Zhang and Cheng, 2003; Zhang et al., 2000), and of transiently interacting proteins that might act either as substrates or regulators of activity (Cote et al., 2003; Herrmann et al., 2004; Lin et al., 1996; Xu et al., 2004). Photobleaching experiments show that complexes containing PRMT1 or CARM1/PRMT4 differ in their mobility, both in the nucleus and the cytoplasm, suggesting that at least the bulk of the two enzymes do not interact with each other. This is most pronounced when methylation of substrates is inhibited. Under these conditions, PRMT1 gains an immobile fraction in the nucleus of live cells because it is tightly tethered to unmethylated substrates (see Herrmann et al., 2005), but PRMT4 is not affected. In addition, glycerol gradient centrifugation (Fig. 2) shows that, although both enzymes sediment in overlapping fractions, the peaks are found in different fractions, with PRMT4 in slightly smaller complexes. Fluorescent correlation spectroscopy in ex vivo extracts confirms these different complex sizes, although the differences are not statistically significant. Interestingly, immunoprecipitation experiments suggested a possible interaction between PRMT1 and CARM1/PRMT4, because they could be co-immunoprecipitated to a certain degree (data not shown). It is conceivable that these enzymes can join forces in large multiprotein complexes to provide methylation activity with non-overlapping substrate specificity under some conditions. This would explain the finding that PRMT1 and CARM1/PRMT4 cooperate in transcriptional activation (An et al., 2004; Kleinschmidt et al., 2008), and that both are necessary to affect target gene expression.

Fig. 7.

Diffusion characteristics of PRMT complexes investigated by fluorescence correlation spectroscopy. Comparison of normalized auto correlation functions of diluted total cell extracts from cells stably expressing GFP fusion proteins. Cells were treated with oxidized adenosine (Adox) (B) 2 days prior to harvesting, or left untreated (A). Note that PRMT1, PRMT4 and PRMT5 show diffusion characteristics of high molecular weight complexes. No significant difference between auto correlation functions in –Adox and +Adox samples was observed.

Fig. 7.

Diffusion characteristics of PRMT complexes investigated by fluorescence correlation spectroscopy. Comparison of normalized auto correlation functions of diluted total cell extracts from cells stably expressing GFP fusion proteins. Cells were treated with oxidized adenosine (Adox) (B) 2 days prior to harvesting, or left untreated (A). Note that PRMT1, PRMT4 and PRMT5 show diffusion characteristics of high molecular weight complexes. No significant difference between auto correlation functions in –Adox and +Adox samples was observed.

PRMT2 was predominantly found in the nucleus, where it forms a granular pattern and displays regions of enrichment that resemble nuclear speckles. Ongoing work in our laboratory is investigating the potential role of PRMT2 in RNA maturation; we would like to note that we have also, for the first time, found weak enzymatic activity of PRMT2 (supplementary material Fig. S3C). Photobleaching revealed that PRMT2 diffuses more slowly in the nucleus than expected for its size and diffusion characteristics in vitro, where it behaves like a dimer. This suggests that it forms transient interactions with nuclear proteins that reduce its mobility. A similar effect was observed for PRMT6, which sediments as a free protein and behaves like a monomer in fluorescence correlation spectroscopy, but displays an intranuclear mobility between those of the multiprotein complexes containing PRMT1 or CARM1/PRMT4. Interestingly, inhibition of methylation by Adox slightly increases the mobility of PRMT6, in contrast to all other PRMTs that were either unaffected within the limits of statistical significance (PRMT1 and 7), or decreased (PRMT2, CARM1/PRMT4, and most significantly, PRMT3). PRMT3 is known to methylate both nuclear and cytoplasmic proteins, such as the nuclear poly(A)-binding protein (PABPN1) (Fronz et al., 2008; Smith et al., 1999) and, most prominently, the cytoplasmic ribosomal protein S2 (rpS2) (Bachand and Silver, 2004; Choi et al., 2008; Swiercz et al., 2007; Swiercz et al., 2005). The localization in both compartments was therefore not a surprise. However, photobleaching experiments revealed that PRMT3 is the only PRMT with almost identical diffusion coefficients in the nucleus and the cytoplasm. This is unusual for a protein, because the viscosity of the nucleus and the cytoplasm differs by a factor of two to three (Wachsmuth et al., 2000), which is usually reflected in the diffusion coefficient of a protein present in both compartments (see for example, the mobility of GFP or PRMT7, Table 1A). The comparatively slow diffusion of PRMT3 in the cytoplasm apparently arises from interactions of the enzyme with substrates that are in large complexes, presumably with rpS2 in ribosomes (Choi et al., 2008; Swiercz et al., 2007). This interaction is not significantly affected when methylation is inhibited, suggesting that methylation of rpS2 (or other substrates) is not essential for binding. However, the nuclear fraction of PRMT3 gains an immobile fraction under Adox treatment, which is similar to PRMT1, but not as pronounced (Fig. 3) (Herrmann et al., 2005). This suggests that nuclear PRMT3 interacts with entities of low mobility, possibly complexes containing pre-mRNA and poly(A)-binding protein or nuclear substructures, such as chromatin. Complementary in vitro experiments demonstrated that human PRMT3 behaves like a monomer or a small oligomer in glycerol gradients, in agreement with gel filtration results reported by Tang and colleagues (Tang et al., 1998). Quantitative analysis by fluorescence correlation spectroscopy revealed diffusion characteristics compatible with PRMT3 forming a tetramer in solution. Alternatively, PRMT3 could be in a stable complex with another protein that affects its diffusion. In this context, it is interesting to note that neither our gradient centrifugations nor the fluorescence correlation spectroscopy measurements revealed the presence of human PRMT3 bound to ribosomal subunits, as previously described for PRMT3 from fission yeast (Bachand and Silver, 2004) and mouse (Swiercz et al., 2007). We obtained the same results for the endogenous and for the stably expressed GFP fusion protein, suggesting that the interaction of PRMT3 with ribosomes might be weaker in human cells. Interestingly, while we were preparing this paper, Choi and colleagues (Choi et al., 2008) published identical results for PRMT3, also from human HEK293 cells. Thus, the binding of human PRMT3 might be strong enough to be responsible for the slow mobility of PRMT3 in the cytoplasm of HEK293 cells (see above), but might dissociate when cells are disrupted.

PRMT7 is predominantly cytoplasmic in the HEK293 cells investigated here, and is a very rapidly diffusing protein in vivo, similar to PRMT3. Its diffusion characteristics in live cells are typical for a protein with no or very few interactions with substrate proteins, and they are not altered when methylation is inhibited. Fluorescence correlation spectroscopy shows that the protein is monomeric in solution. However, because PRMT7 has two methyltransferase domains (see Fig. 1), the monomeric protein might mimic oligomerization and behave as an intramolecular dimer. This is in line with data from Miranda and colleagues who demonstrated that both methyltransferase domains are required for functionality (Miranda et al., 2004). In our experiments, we did not detect enzymatic activity of PRMT7, suggesting that either the specific activity of the expressed enzyme is too low (as is the case for PRMT3) (Tang et al., 1998), or that the substrate is missing or too dilute in the total cell extracts used for the activity assay. It is possible that substrates of PRMT7 are also missing or too dilute in the HEK293 cells, in contrast to PRMT1 and PRMT4 for example, which have many substrates present at high concentrations, so that photobleaching experiments measure only the free enzyme.

As stated above, our combined results provide novel and, for the first time, quantitative data about the dynamics of PRMT family members in human cells. They show that PRMTs have non-overlapping properties, and are therefore probably involved in different cellular processes. The data presented here provide a firm ground for future experiments, which are certainly needed to better understand the physiological roles of arginine methylation and to find potential links to consequences of its misregulation.

Materials and Methods

Cell culture and transfection

Human embryonic kidney cells (HEK293), human osteosarcoma cells (U2OS), or human cervix carcinoma cells (HeLa) were cultivated on plastic dishes in DMEM with 10% fetal calf serum in a humidified atmosphere containing 5% CO2, and were split 1:5 every second day. Human breast carcinoma cells (MCF-7) were cultured identically, but in medium containing additional insulin at a final concentration of 10 μg/ml. Cells were transfected using polyethylenimine (Boussif et al., 1995) with expression vectors encoding EGFP fusions of PRMT1 to PRMT8, respectively; stably expressing cell lines were created by selection with G418 for 4-6 weeks and, after selection, expressing cells were obtained by fluorescence-activated cell sorting of GFP-positive cells. For FRAP analysis and microscopy, we used cell populations rather than single cell clones, to rule out clonal variation or other potential problems inherent in the work with single cell clones. The generation of the GFP-PRMT1 to GFP-PRMT6 constructs (Frankel et al., 2002) and GFP-PRMT8 construct (Lee et al., 2005a) has been described previously. Human cDNA was used to amplify PRMT7 by PCR. The PCR product was digested with BglII and EcoRI, and subcloned into pEGFP-C1.

For inhibition of methylation and for preparation of hypomethylated cell extracts, cell culture medium was supplemented with 15 μM of periodate-oxidized adenosine (adenosine-2′-3′-dialdehyde, Sigma A7154), and cells were cultured for additional 48 hours before they were analyzed or harvested. Conditions of Adox treatment had been empirically determined and thoroughly controlled as described (Herrmann et al., 2005).

Live-cell microscopy

For live-cell microscopy, cells were split onto 35 mm culture dishes with a glass bottom (Mattek) and analyzed 2 days after splitting either untreated or treated with Adox as described above. For the analysis, cells were placed on the heated stage of a Zeiss Meta 510 confocal laser-scanning microscope and images were acquired from typical cells using a ×63 oil-immersion lens and 488 nm excitation at 1% laser intensity.

Fluorescence recovery after photobleaching (FRAP)

Photobleaching was performed as described previously (Herrmann et al., 2005; Mearini and Fackelmayer, 2006). Approximately 20 cells were analyzed for each PRMT, and mean pixel intensities in the bleached region of interest (ROI) and a non-bleached reference region were determined with the freeware image analysis software ImageJ by Wayne Rasband (http://rsb.info.nih.gov/ij/). Raw recovery data were imported into Microsoft Excel for normalization with pixel intensities from the reference region and the initial fluorescence intensity directly after bleaching. Non-linear regression was performed with Prism4 (Graphpad) to determine the mobile fraction (Mf) and halftime of recovery (t/2). All PRMT recovery curves could consistently be fitted very well with the Weibull function:
\[\ F(t)=F_{\mathrm{max}}^{{\ast}}[1{\ }-{\ }\mathrm{exp}(-K^{{\ast}}t^s)],\ \]
where F(t) represents the fluorescence at time t, Fmax represents the asymptote of the ROI fluorescence recovery curve (equal to Mf), K represents the rate constant of recovery, and s represents the Weibull `shape parameter'. This parameter describes a change of the exponential recovery rate with time; for a constant recovery rate, s equals 1 and the Weibull function converts to the one-phase exponential association function:
\[\ F(t)=F_{\mathrm{max}}^{{\ast}}[1{\ }-{\ }\mathrm{exp}(-K^{{\ast}}t)],\ \]
For our experimental setup, a value of 0.74 for s yielded a significantly better curve fitting than the one-phase exponential association function (see supplementary material Fig. S1). Diffusion coefficients D (expressed as μm2/second) were then calculated from t/2 using:
\[\ D={\omega}^{2}{\ }{/}{\ }(4^{{\ast}}{\tau}),\ \]
where τ represents t/2 in seconds, and ω represents the height of the bleaching region in μm (Axelrod et al., 1976).

Fluorescence correlation spectroscopy

Cell extracts were prepared by swelling cells stably expressing EGFP fusion proteins in hypotonic buffer (20 mM Tris-HCl, pH 7.4) for 5 minutes, and homogenization in a Dounce homogenizer (Wachsmuth et al., 2000). The extract was cleared by centrifugation for 5 minutes at 6000 g, adjusted to isotonic conditions (20 mM Tris-HCl, pH 7.4, 130 mM NaCl) and diluted in the same buffer to 1-5 molecules per confocal volume of measurement.

The FCS measurements were performed on the Zeiss LSM 510/ConfoCor 2 combination. EGFP was excited with the 488 nm line from an argon-ion laser (1% laser intensity) and focused into the sample with a water immersion C-Apochromat ×40 lens objective (Zeiss). The excitation and emission light were separated by a dichroic beam splitter (HFT 488/633). The fluorescence was detected by an avalanche photodiode after being filtered through a longpass filter >505 nm. The pinhole was set to 50 μm. Between three and six independent measurements with 100 autocorrelation traces (10 second measuring time) were performed for each investigated protein. The data were analyzed using the Zeiss ConfoCor2 software. The autocorrelation traces were fitted with a model describing Brownian motion of one or two distinct species in three dimensions (autocorrelation function). This curve fit resulted in values for τd, which represent the characteristic diffusion time of the molecules, and allow calculation of the translational diffusion coefficient D by using the equation:
\[\ D={\omega}^{2}{\ }{/}{\ }(4^{{\ast}}{\tau}_{\mathrm{d}},\ \]
where ω is the equatorial radius of the confocal measuring volume and is obtained from calibration measurements using Rhodamine-6-green (Rh6G). In our experimental setup, ω was determined to be 159 nm.
Approximate values for the hydrodynamic radii were calculated, assuming a globular shape of the molecules, using the Stokes-Einstein relation:
\[\ r_{\mathrm{H}}=k^{{\ast}}T{\ }{/}{\ }(D^{{\ast}}6{\pi}^{{\ast}}{\eta}),\ \]
where rH represents the hydrodynamic radius of the molecule, k represents the Boltzmann constant, T represents the temperature (310 K), and η represents the viscosity of the solution which was experimentally determined from measurements with EGFP (1.18 mPa seconds). The hydrodynamic radius allows estimating the molecular mass of the protein by using equation:
\[\ m=4{\pi}^{{\ast}}{\rho}^{{\ast}}N_{\mathrm{A}}^{{\ast}}r_{\mathrm{H}}^{3},\ \]
where ρ represents the mean density of the protein (set to 1.2 g/cm3) and NA represents Avogadro's number.

Western blotting

Western blotting was performed as described previously (Herrmann et al., 2004) using commercially available antibodies against GFP (Roche, Applied Science), hnRNP-C (ab10294, Abcam) or PRMT1 to PRMT7 (07-404, 07-256, 07-405; 07-639, Upstate; ab3667, Abcam; IMG-496, IMG-506, Imgenex). For detection, HRP-conjugated secondary antibodies (Sigma) and ECL chemiluminescent reagent (Amersham Biosciences) were used.

In vitro methylation assay

Semiconfluent to confluent cells were harvested by scraping off the culture dishes with a rubber policeman after two washes with PBS, and collected by centrifugation (200 g, 5 minutes). Extract was prepared by resuspending cells in lysis buffer [1.5× PBS, 1% Triton X-100 and protease inhibitor cocktail Complete, EDTA-free (Roche)], incubating on ice for 5 minutes, and clearing by centrifugation for 15 minutes in an Eppendorf microcentrifuge at full speed. Extract from 1-5×106 cells was supplemented with 5 μg antibodies against GFP (Roche, Applied Science), incubated for 2 hours at 4°C, and immune complexes were collected by incubation with protein-G-Sepharose for an additional 1 hour. Immune complexes were washed thoroughly six times and the activity of precipitated PRMTs was measured by radioactive methylation assays exactly as described earlier (Herrmann et al., 2004). Briefly, immunoprecipitated PRMT (on 20 μl Protein-G-Sepharose beads) were combined with heat-inactivated extract from hypomethylated cells and 5 μCi S-adenosyl-L-[methyl-3H]methionine (Amersham Biosciences TRK865, specific activity 2.96 TBq/mmol), and reactions were incubated for 2 hours at 37°C. The reaction mixture was then resolved by SDS-PAGE and radioactively labeled proteins were visualized by fluorography with enHance reagent (NEN Dupont).

Glycerol gradient centrifugation

Native total cell extract was prepared as described for in vitro methylation assay, and layered over a 10-30% glycerol gradient (in lysis buffer) in a Beckman SW41 tube. The gradient was centrifuged at 4°C for 20 hours at 30,000 r.p.m. and was fractionated in 0.8 ml aliquots from the top. Aliquots of each fraction were resolved by SDS page, and proteins were detected by western blotting.

We wish to thank Arne Düsedau from the Heinrich-Pette-Institute, Hamburg, Germany, for expert help with cell sorting and Magdalena Hyzy, Gdansk, for pioneering work with PRMT2. This work was supported by grant Fa376/3-1 of the Deutsche Forschungsgemeinschaft (DFG) to F.O.F. and a short-term fellowship from the Schering foundation to Magdalena Hyzy.

References

An, W., Kim, J. and Roeder, R. G. (
2004
). Ordered cooperative functions of PRMT1, p300, and CARM1 in transcriptional activation by p53.
Cell
117
,
735
-748.
Ancelin, K., Lange, U. C., Hajkova, P., Schneider, R., Bannister, A. J., Kouzarides, T. and Surani, M. A. (
2006
). Blimp1 associates with Prmt5 and directs histone arginine methylation in mouse germ cells.
Nat. Cell Biol.
8
,
623
-630.
Axelrod, D., Koppel, D. E., Schlessinger, J., Elson, E. and Webb, W. W. (
1976
). Mobility measurement by analysis of fluorescence photobleaching recovery kinetics.
Biophys. J.
16
,
1055
-1069.
Bachand, F. and Silver, P. A. (
2004
). PRMT3 is a ribosomal protein methyltransferase that affects the cellular levels of ribosomal subunits.
EMBO J.
23
,
2641
-2650.
Balint, B. L., Szanto, A., Madi, A., Bauer, U. M., Gabor, P., Benko, S., Puskas, L. G., Davies, P. J. and Nagy, L. (
2005
). Arginine methylation provides epigenetic transcription memory for retinoid-induced differentiation in myeloid cells.
Mol. Cell. Biol.
25
,
5648
-5663.
Bates, I. R., Wiseman, P. W. and Hanrahan, J. W. (
2006
). Investigating membrane protein dynamics in living cells.
Biochem. Cell Biol.
84
,
825
-831.
Bedford, M. T. (
2007
). Arginine methylation at a glance.
J. Cell Sci.
120
,
4243
-4246.
Bedford, M. T. and Richard, S. (
2005
). Arginine methylation an emerging regulator of protein function.
Mol. Cell
18
,
263
-272.
Berthet, C., Guehenneux, F., Revol, V., Samarut, C., Lukaszewicz, A., Dehay, C., Dumontet, C., Magaud, J. P. and Rouault, J. P. (
2002
). Interaction of PRMT1 with BTG/TOB proteins in cell signalling: molecular analysis and functional aspects.
Genes Cells
7
,
29
-39.
Blanchet, F., Schurter, B. T. and Acuto, O. (
2006
). Protein arginine methylation in lymphocyte signaling.
Curr. Opin. Immunol.
18
,
321
-328.
Boisvert, F. M., Cote, J., Boulanger, M. C., Cleroux, P., Bachand, F., Autexier, C. and Richard, S. (
2002
). Symmetrical dimethylarginine methylation is required for the localization of SMN in Cajal bodies and pre-mRNA splicing.
J. Cell Biol.
159
,
957
-969.
Boussif, O., Lezoualc'h, F., Zanta, M. A., Mergny, M. D., Scherman, D., Demeneix, B. and Behr, J. P. (
1995
). A versatile vector for gene and oligonucleotide transfer into cells in culture and in vivo: polyethylenimine.
Proc. Natl. Acad. Sci. USA
92
,
7297
-7301.
Branscombe, T. L., Frankel, A., Lee, J. H., Cook, J. R., Yang, Z., Pestka, S. and Clarke, S. (
2001
). PRMT5 (Janus kinase-binding protein 1) catalyzes the formation of symmetric dimethylarginine residues in proteins.
J. Biol. Chem.
276
,
32971
-32976.
Chen, D., Ma, H., Hong, H., Koh, S. S., Huang, S. M., Schurter, B. T., Aswad, D. W. and Stallcup, M. R. (
1999
). Regulation of transcription by a protein methyltransferase.
Science
284
,
2174
-2177.
Cheng, D., Cote, J., Shaaban, S. and Bedford, M. T. (
2007
). The arginine methyltransferase CARM1 regulates the coupling of transcription and mRNA processing.
Mol. Cell
25
,
71
-83.
Choi, S., Jung, C. R., Kim, J. Y. and Im, D. S. (
2008
). PRMT3 inhibits ubiquitination of ribosomal protein S2 and together forms an active enzyme complex.
Biochim. Biophys. Acta
1780
,
1062
-1069.
Cook, J. R., Lee, J. H., Yang, Z. H., Krause, C. D., Herth, N., Hoffmann, R. and Pestka, S. (
2006
). FBXO11/PRMT9, a new protein arginine methyltransferase, symmetrically dimethylates arginine residues.
Biochem. Biophys. Res. Commun.
342
,
472
-481.
Cote, J., Boisvert, F. M., Boulanger, M. C., Bedford, M. T. and Richard, S. (
2003
). Sam68 RNA binding protein is an in vivo substrate for protein arginine N-methyltransferase 1.
Mol. Biol. Cell
14
,
274
-287.
Fackelmayer, F. O. (
2005a
). A stable proteinaceous structure in the territory of inactive X chromosomes.
J. Biol. Chem.
280
,
1720
-1723.
Fackelmayer, F. O. (
2005b
). Protein arginine methyltransferases: guardians of the Arg?
Trends Biochem. Sci.
30
,
666
-671.
Fielenbach, N., Guardavaccaro, D., Neubert, K., Chan, T., Li, D., Feng, Q., Hutter, H., Pagano, M. and Antebi, A. (
2007
). DRE-1: an evolutionarily conserved F box protein that regulates C. elegans developmental age.
Dev. Cell
12
,
443
-455.
Frankel, A., Yadav, N., Lee, J., Branscombe, T. L., Clarke, S. and Bedford, M. T. (
2002
). The novel human protein arginine N-methyltransferase PRMT6 is a nuclear enzyme displaying unique substrate specificity.
J. Biol. Chem.
277
,
3537
-3543.
Fronz, K., Otto, S., Kolbel, K., Kuhn, U., Friedrich, H., Schierhorn, A., Beck-Sickinger, A. G., Ostareck-Lederer, A. and Wahle, E. (
2008
). Promiscuous modification of the nuclear poly(A) binding protein by multiple protein arginine methyl transferases does not affect the aggregation behavior.
J. Biol. Chem.
283
,
20408
-20420.
Goulet, I., Gauvin, G., Boisvenue, S. and Cote, J. (
2007
). Alternative splicing yields protein arginine methyltransferase 1 isoforms with distinct activity, substrate specificity, and subcellular localization.
J. Biol. Chem.
282
,
33009
-33021.
Herrmann, F. and Fackelmayer, F. O. (
2009
). Nucleo-cytoplasmic shuttling of protein arginine methyltransferase 1 (PRMT1) requires enzymatic activity.
Genes to Cells
DOI: .
Herrmann, F., Bossert, M., Schwander, A., Akgun, E. and Fackelmayer, F. O. (
2004
). Arginine methylation of scaffold attachment factor A by heterogeneous nuclear ribonucleoprotein particle-associated PRMT1.
J. Biol. Chem.
279
,
48774
-48779.
Herrmann, F., Lee, J., Bedford, M. T. and Fackelmayer, F. O. (
2005
). Dynamics of human protein arginine methyltransferase 1 (PRMT1) in vivo.
J. Biol. Chem.
280
,
38005
-38010.
Higashimoto, K., Kuhn, P., Desai, D., Cheng, X. and Xu, W. (
2007
). Phosphorylation-mediated inactivation of coactivator-associated arginine methyltransferase 1.
Proc. Natl. Acad. Sci. USA
104
,
12318
-12323.
Iberg, A. N., Espejo, A., Cheng, D., Kim, D., Michaud-Levesque, J., Richard, S. and Bedford, M. T. (
2008
). Arginine methylation of the histone H3 tail impedes effector binding.
J. Biol. Chem.
283
,
3006
-3010.
Katsanis, N., Yaspo, M. L. and Fisher, E. M. (
1997
). Identification and mapping of a novel human gene, HRMT1L1, homologous to the rat protein arginine N-methyltransferase 1 (PRMT1) gene.
Mamm. Genome
8
,
526
-529.
Kim, J., Lee, J., Yadav, N., Wu, Q., Carter, C., Richard, S., Richie, E. and Bedford, M. T. (
2004
). Loss of CARM1 results in hypomethylation of thymocyte cyclic AMP-regulated phosphoprotein and deregulated early T cell development.
J. Biol. Chem.
279
,
25339
-25344.
Kleinschmidt, M. A., Streubel, G., Samans, B., Krause, M. and Bauer, U. M. (
2008
). The protein arginine methyltransferases CARM1 and PRMT1 cooperate in gene regulation.
Nucleic Acids Res.
36
,
3202
-3213.
Lacroix, M., Messaoudi, S. E., Rodier, G., Le Cam, A., Sardet, C. and Fabbrizio, E. (
2008
). The histone-binding protein COPR5 is required for nuclear functions of the protein arginine methyltransferase PRMT5.
EMBO Rep.
9
,
452
-458.
Lake, A. N. and Bedford, M. T. (
2007
). Protein methylation and DNA repair.
Mutat. Res.
618
,
91
-101.
Lee, J., Sayegh, J., Daniel, J., Clarke, S. and Bedford, M. T. (
2005a
). PRMT8, a new membrane-bound tissue-specific member of the protein arginine methyltransferase family.
J. Biol. Chem.
280
,
32890
-32896.
Lee, J. H., Cook, J. R., Yang, Z. H., Mirochnitchenko, O., Gunderson, S. I., Felix, A. M., Herth, N., Hoffmann, R. and Pestka, S. (
2005b
). PRMT7, a new protein arginine methyltransferase that synthesizes symmetric dimethylarginine.
J. Biol. Chem.
280
,
3656
-3664.
Lim, Y., Kwon, Y. H., Won, N. H., Min, B. H., Park, I. S., Paik, W. K. and Kim, S. (
2005
). Multimerization of expressed protein-arginine methyltransferases during the growth and differentiation of rat liver.
Biochim. Biophys. Acta
1723
,
240
-247.
Lin, W. J., Gary, J. D., Yang, M. C., Clarke, S. and Herschman, H. R. (
1996
). The mammalian immediate-early TIS21 protein and the leukemia-associated BTG1 protein interact with a protein-arginine N-methyltransferase.
J. Biol. Chem.
271
,
15034
-15044.
Mearini, G. and Fackelmayer, F. O. (
2006
). Local chromatin mobility is independent of transcriptional activity.
Cell Cycle
5
,
1989
-1995.
Meyer, R., Wolf, S. S. and Obendorf, M. (
2007
). PRMT2, a member of the protein arginine methyltransferase family, is a coactivator of the androgen receptor.
J. Steroid Biochem. Mol. Biol.
107
,
1
-14.
Miranda, T. B., Miranda, M., Frankel, A. and Clarke, S. (
2004
). PRMT7 is a member of the protein arginine methyltransferase family with a distinct substrate specificity.
J. Biol. Chem.
279
,
22902
-22907.
Mowen, K. A., Schurter, B. T., Fathman, J. W., David, M. and Glimcher, L. H. (
2004
). Arginine methylation of NIP45 modulates cytokine gene expression in effector T lymphocytes.
Mol. Cell
15
,
559
-571.
Niu, L., Lu, F., Pei, Y., Liu, C. and Cao, X. (
2007
). Regulation of flowering time by the protein arginine methyltransferase AtPRMT10.
EMBO Rep.
8
,
1190
-1195.
Pal, S., Yun, R., Datta, A., Lacomis, L., Erdjument-Bromage, H., Kumar, J., Tempst, P. and Sif, S. (
2003
). mSin3A/histone deacetylase 2- and PRMT5-containing Brg1 complex is involved in transcriptional repression of the Myc target gene cad.
Mol. Cell. Biol.
23
,
7475
-7487.
Pawlak, M. R., Scherer, C. A., Chen, J., Roshon, M. J. and Ruley, H. E. (
2000
). Arginine N-methyltransferase 1 is required for early postimplantation mouse development, but cells deficient in the enzyme are viable.
Mol. Cell. Biol.
20
,
4859
-4869.
Pawlak, M. R., Banik-Maiti, S., Pietenpol, J. A. and Ruley, H. E. (
2002
). Protein arginine methyltransferase I: substrate specificity and role in hnRNP assembly.
J. Cell Biochem.
87
,
394
-407.
Pollack, B. P., Kotenko, S. V., He, W., Izotova, L. S., Barnoski, B. L. and Pestka, S. (
1999
). The human homologue of the yeast proteins Skb1 and Hsl7p interacts with Jak kinases and contains protein methyltransferase activity.
J. Biol. Chem.
274
,
31531
-31542.
Qi, C., Chang, J., Zhu, Y., Yeldandi, A. V., Rao, S. M. and Zhu, Y. J. (
2002
). Identification of protein arginine methyltransferase 2 as a coactivator for estrogen receptor alpha.
J. Biol. Chem.
277
,
28624
-28630.
Rho, J., Choi, S., Seong, Y. R., Cho, W. K., Kim, S. H. and Im, D. S. (
2001
). Prmt5, which forms distinct homo-oligomers, is a member of the protein-arginine methyltransferase family.
J. Biol. Chem.
276
,
11393
-11401.
Ruthenburg, A. J., Li, H., Patel, D. J. and Allis, C. D. (
2007
). Multivalent engagement of chromatin modifications by linked binding modules.
Nat. Rev. Mol. Cell. Biol.
8
,
983
-994.
Sayegh, J., Webb, K., Cheng, D., Bedford, M. T. and Clarke, S. G. (
2007
). Regulation of protein arginine methyltransferase 8 (PRMT8) activity by its N-terminal domain.
J. Biol. Chem.
282
,
36444
-36453.
Scebba, F., De Bastiani, M., Bernacchia, G., Andreucci, A., Galli, A. and Pitto, L. (
2007
). PRMT11: a new Arabidopsis MBD7 protein partner with arginine methyltransferase activity.
Plant J.
52
,
210
-222.
Shen, E. C., Henry, M. F., Weiss, V. H., Valentini, S. R., Silver, P. A. and Lee, M. S. (
1998
). Arginine methylation facilitates the nuclear export of hnRNP proteins.
Genes Dev.
12
,
679
-691.
Smith, J. J., Rucknagel, K. P., Schierhorn, A., Tang, J., Nemeth, A., Linder, M., Herschman, H. R. and Wahle, E. (
1999
). Unusual sites of arginine methylation in Poly(A)-binding protein II and in vitro methylation by protein arginine methyltransferases PRMT1 and PRMT3.
J. Biol. Chem.
274
,
13229
-13234.
Swiercz, R., Person, M. D. and Bedford, M. T. (
2005
). Ribosomal protein S2 is a substrate for mammalian PRMT3 (protein arginine methyltransferase 3).
Biochem. J.
386
,
85
-91.
Swiercz, R., Cheng, D., Kim, D. and Bedford, M. T. (
2007
). Ribosomal protein rpS2 is hypomethylated in PRMT3-deficient mice.
J. Biol. Chem.
282
,
16917
-16923.
Taneda, T., Miyata, S., Kousaka, A., Inoue, K., Koyama, Y., Mori, Y. and Tohyama, M. (
2007
). Specific regional distribution of protein arginine methyltransferase 8 (PRMT8) in the mouse brain.
Brain Res.
1155
,
1
-9.
Tang, J., Gary, J. D., Clarke, S. and Herschman, H. R. (
1998
). PRMT 3, a type I protein arginine N-methyltransferase that differs from PRMT1 in its oligomerization, subcellular localization, substrate specificity, and regulation.
J. Biol. Chem.
273
,
16935
-16945.
Wachsmuth, M., Waldeck, W. and Langowski, J. (
2000
). Anomalous diffusion of fluorescent probes inside living cell nuclei investigated by spatially-resolved fluorescence correlation spectroscopy.
J. Mol. Biol.
298
,
677
-689.
Wysocka, J., Allis, C. D. and Coonrod, S. (
2006
). Histone arginine methylation and its dynamic regulation.
Front. Biosci.
11
,
344
-355.
Xu, W., Cho, H., Kadam, S., Banayo, E. M., Anderson, S., Yates, J. R., 3rd, Emerson, B. M. and Evans, R. M. (
2004
). A methylation-mediator complex in hormone signaling.
Genes Dev.
18
,
144
-156.
Yadav, N., Lee, J., Kim, J., Shen, J., Hu, M. C., Aldaz, C. M. and Bedford, M. T. (
2003
). Specific protein methylation defects and gene expression perturbations in coactivator-associated arginine methyltransferase 1-deficient mice.
Proc. Natl. Acad. Sci. USA
100
,
6464
-6468.
Yanagida, M., Hayano, T., Yamauchi, Y., Shinkawa, T., Natsume, T., Isobe, T. and Takahashi, N. (
2004
). Human fibrillarin forms a sub-complex with splicing factor 2-associated p32, protein arginine methyltransferases, and tubulins alpha 3 and beta 1 that is independent of its association with preribosomal ribonucleoprotein complexes.
J. Biol. Chem.
279
,
1607
-1614.
Zhang, X. and Cheng, X. (
2003
). Structure of the predominant protein arginine methyltransferase PRMT1 and analysis of its binding to substrate peptides.
Structure
11
,
509
-520.
Zhang, X., Zhou, L. and Cheng, X. (
2000
). Crystal structure of the conserved core of protein arginine methyltransferase PRMT3.
EMBO J.
19
,
3509
-3519.

Supplementary information