Substrate Specificity of Human Protein Arginine Methyltransferase 7 (PRMT7)

Background: Mammalian PRMT7 has been implicated in multiple biological processes, but its in vivo substrates have not been identified. Results: Mutagenesis studies indicate key residues that affect the unique substrate preference of PRMT7. Conclusion: Acidic residues within the double E loop confer specificity to PRMT7. Significance: Understanding how PRMT7 recognizes its substrates will enhance our knowledge of its physiological role. Protein arginine methyltransferase 7 (PRMT7) methylates arginine residues on various protein substrates and is involved in DNA transcription, RNA splicing, DNA repair, cell differentiation, and metastasis. The substrate sequences it recognizes in vivo and the enzymatic mechanism behind it, however, remain to be explored. Here we characterize methylation catalyzed by a bacterially expressed GST-tagged human PRMT7 fusion protein with a broad range of peptide and protein substrates. After confirming its type III activity generating only ω-NG-monomethylarginine and its distinct substrate specificity for RXR motifs surrounded by basic residues, we performed site-directed mutagenesis studies on this enzyme, revealing that two acidic residues within the double E loop, Asp-147 and Glu-149, modulate the substrate preference. Furthermore, altering a single acidic residue, Glu-478, on the C-terminal domain to glutamine nearly abolished the activity of the enzyme. Additionally, we demonstrate that PRMT7 has unusual temperature dependence and salt tolerance. These results provide a biochemical foundation to understanding the broad biological functions of PRMT7 in health and disease.

PRMTs have been characterized as either type I enzymes that transfer up to two methyl groups from S-adenosyl-L-methionine (AdoMet) to the same terminal nitrogen atom forming -N G -monomethylarginine (MMA) and -N G ,N G -asymmetric dimethylarginine (ADMA) (PRMT1, -2, -3, -4, -6, and -8) or type II enzymes that transfer up to two methyl groups to different terminal nitrogen atoms, forming MMA and -N G ,NЈ Gsymmetric dimethylarginine (SDMA) (PRMT5) (1)(2)(3)(4). However, PRMT7 has been found to produce MMA residues only, which defines it as the first type III PRMT (5)(6)(7)(8)(9). Recent work utilizing an insect cell-expressed mouse PRMT7 demonstrated a unique substrate specificity for RXR motifs among multiple basic residues (8). No other PRMT has been shown to have such narrow sequence specificity. These results led to the question of how the substrate selectivity of PRMT7 is restricted.
The human PRMT7 gene encodes a protein containing ancestrally duplicated methyltransferase domains (1,5). Although the advantage of duplicating the catalytic domain is unknown, it has been shown that deletion of either domain abolishes its activity (5). A number of studies have indicated that PRMT7 expression in mammalian cells is regulated and can affect many biological processes, including pluripotency (10 -12). Consistent with this role, PRMT7 knockdown stimulates neuronal differentiation (13). Genetic variation of PRMT7 may lead to increased sensitivity of cell lines to etoposide, a chemotherapeutic agent that targets topoisomerase II (14,15). In addition, down-regulation of PRMT7 sensitizes tumor cells to the topoisomerase I inhibitor camptothecin (16) and modulates cell response to DNA-damaging agents (17,18). The role of PRMT7 has also been implicated in male germ line imprinting (19), Sm ribonucleoprotein methylation, and small nuclear ribonucleoprotein biogenesis (20) as well as breast cancer metastasis (21,22). Thus, PRMT7 appears to participate in broad cellular processes under normal and disease conditions and is of therapeutic interest.
Given the extensive biological involvement of PRMT7, it is particularly important to get a clear understanding of its fundamental catalytic mechanism and cellular substrate specificity. Although the insect cell-expressed mouse PRMT7 showed higher catalytic activity than the bacterially expressed GST-tagged human PRMT7 (7,8), site-directed mutagenesis and protein expression and purification are facilitated in the bacterial system. In this work, we used the bacterial system to determine which residues of human PRMT7 are important for substrate recognition. We determined the substrate specificity with multiple peptide and protein substrates and then performed mutagenesis and kinetic analysis to probe the roles of specific residues of PRMT7 in the methylation reaction. We were able to demonstrate that two acidic residues within the catalytic double E loop are essential for its substrate specificity. Additionally, we revealed that mammalian PRMT7 is relatively tolerant to low temperature but is very sensitive to high temperature and salt. These findings will broaden our understanding of the reaction catalyzed by PRMT7 and set directions for future studies of its cellular functions.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-Human PRMT7 was subcloned into a pGEX-2T vector and expressed in Escherichia coli BL21 Star (DE3) cells (Invitrogen, C601003) as a GST fusion protein (7). The enzyme was purified using a glutathione-Sepharose affinity chromatography method modified from that described previously (7). Cells containing GST-PRMT7 plasmid were grown to an optical density at 600 nm of 0.6 -0.8, and protein expression was induced with 0.4 mM isopropyl-Dthiogalactopyranoside at 16°C overnight. The cells were lysed with sonication in a phosphate-buffered saline solution (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 2 mM KH 2 PO 4 , pH 7.4) containing 1 mM phenylmethylsulfonyl fluoride. The cell lysate was centrifuged for 50 min at 23,000 ϫ g at 4°C, and the supernatant containing GST-PRMT7 was loaded to glutathione-Sepharose 4B beads (Amersham Biosciences) according to the manufacturer's instructions. After washing with the phosphate-buffered saline solution, the bound protein was eluted with an elution buffer containing 30 mM glutathione, 50 mM HEPES, 120 mM NaCl, and 5% glycerol (pH 8.0). Glutathione in the eluted protein solution was reduced by 10 -30-fold by adding fresh elution buffer without glutathione and reconcentrating using an Amicon centrifugation filter. Protein was quantified by a Lowry assay after trichloroacetic acid precipitation and stored at Ϫ80°C as 50-l aliquots. GST-GAR was expressed in E. coli BL21 Star (DE3) cells and purified with glutathione-Sepharose 4B affinity chromatography as described previously (7).
Amino Acid Analysis of Protein and Peptide Substrates by High Resolution Cation Exchange Chromatography-Substrates were methylated with the indicated amount of GST-PRMT7 wild-type or mutant protein and 0.7 M [ 3 H]AdoMet in a final reaction volume of 60 l. The reactions were incubated for 18 -20 h at 23°C in 50 mM potassium HEPES (pH 7.5), 10 mM NaCl, and 1 mM DTT. Control reactions were performed in the absence of substrate or enzyme or in the presence of the catalytically inactive enzyme mutant. When GST-GAR protein and P-SmD3 peptide were used as substrates, 60 l of 25% (w/v) trichloroacetic acid with 20 g of the carrier protein bovine serum albumin was added and incubated at 23°C for 30 min prior to centrifugation at 4000 ϫ g for an additional 30 min. The pellet was air-dried after washing with Ϫ20°C acetone. When smaller peptides were used as substrates, 3-4 l of 25% (w/v) trichloroacetic acid was used to lower the pH of the reaction mixture, and the peptides were isolated using OMIX C18 Zip-Tip pipette tips (Agilent Technologies) followed by vacuum centrifugation to remove the elution buffer (trifluoroacetic acid/acetonitrile/H 2 O (0.1:50:50)). Acid hydrolysis of all samples was performed in vacuo at 110°C for 20 h with 50 l of 6 N HCl. After removal of the HCl by vacuum centrifugation, samples were dissolved in 50 l of water and mixed with 1 mol each of -MMA (acetate salt; Sigma, M7033), SDMA (di-(p-hydroxyazobenzene)-pЈ-sulfonate salt; Sigma, D0390), and ADMA (hydrochloride salt; Sigma, D4268) as internal standards. High resolution cation exchange chromatography was then performed as described previously (8). The positions of the amino acid standards were determined using 570-nm absorbance after ninhydrin assay with 50 l of column fractions (8). Radioactivity in 950 l of each fraction was determined with liquid scintillation as described (8) but using the average of three 3-min counting cycles.
Fluorography of Methylated GST-GAR after SDS-PAGE-Methyl transfer reactions were set up with [ 3 H]AdoMet, GST-GAR, and GST-PRMT7 wild-type or mutant protein as described above for the amino acid analysis. For the time course, 10 g of GST-GAR was preincubated at 23°C for 5 h in the reaction buffer before the addition of the enzyme and [ 3 H]AdoMet. For the temperature dependence assays, reactions were set up using GST-GAR (10 g, 4.8 M final concentration), 0.8 M GST-PRMT7, and 0.7 M [ 3 H]AdoMet and were incubated for 16 h. In both cases, the reactions were quenched by the addition of SDS loading buffer and separated on 12.6% Tris-glycine gels, stained with Coomassie Blue for 1 h, destained overnight, treated in autoradiography-enhancing buffer (EN 3 HANCE, PerkinElmer, 6NE9701) for 1 h and vacuum-dried. Dried gels were then exposed to autoradiography film (Denville Scientific Inc., E3012) for 3-4 days at Ϫ80°C (Denville Scientific Inc., E3012). After exposure to film, ImageJ densitometric analysis was done on scanned images of the film and quantified as relative activity.

GST-PRMT7 Specifically Recognizes RXR Motifs Surrounded by Multiple Basic Residues in Peptide Sequences Based on Histone and Nonhistone
Proteins-Full-length human PRMT7 was expressed and purified as a GST fusion protein from E. coli to avoid contamination from other PRMTs (27). Using a P81 filter paper assay as described under "Experimental Procedures," we optimized reaction conditions (temperature, pH, and salt) where product formation was linear over a 24-h period ( Fig. 1). Importantly, at the 24 h point, only 6.3% of the initial AdoMet and only 1.3% of the initial peptide were converted to products, indicating that initial velocity conditions prevailed. To characterize the recognition sites for GST-PRMT7, methylation assays were first carried out with synthetic peptides containing the histone H2B repression domain (HBR, residues 27-34 (KKRKRSRK) (28)) that previously was observed to be methylated by insect cell-expressed PRMT7 (8). Peptides consisting of residues 23-37 were prepared with Arg-29, Arg-31, and Arg-33 as in the wild-type sequence or with one or two Arg 3 Lys substitutions at these sites ( Table 1). The methylation products of these peptides were analyzed with high resolution amino acid analysis (Fig. 2). The wild-type peptide H2B(23-37) bearing the three arginines in the RXRXR sequence was well methylated by GST-PRMT7, producing only MMA ( Fig. 2A). Peptides R29K and R33K, in which either the first or third arginine was mutated but where the RXR motif was retained remained good substrates with peak MMA radioactivity compared with the unmodified peptide of ϳ33 and ϳ6%, respectively (Fig. 2, B and D). However, peptide R31K with an RXXXR motif was rel-   atively poorly methylated by GST-PRMT7, giving a peak radioactivity of only about 0.4% (Fig. 2C). Furthermore, when only a single arginine was present, as in peptides R29K/R31K, R29K/ R33K, and R31K/R33K, less than 0.3% of the control radioactivity was seen (Fig. 2, E-G).
We also analyzed peptide substrates derived from the N-terminal 1-21, 1-8, or 14 -22 residues of human histone H4 (Table 1). Peptide H4(1-21) was shown to be a good substrate for GST-PRMT7, with a single MMA peak (Fig. 3A). The replacement of Arg-3, the target site of PRMT1 and PRMT5 (23, 24, 26, 29 -32), by either lysine or MMA in the peptide did not greatly decrease MMA formation or result in SDMA or ADMA formation (Fig. 3, B and C). Analysis of shorter peptides containing residues 14 -22 of H4 and its R17K and/or R19K derivatives revealed that the loss of either the Arg-17 or Arg-19 residue markedly decreased MMA formation (Fig. 3, D-F). Finally, peptides containing residues 1-8 of H4 and residues 1-7 of H3 that lack the RXR motif were analyzed; little increase in MMA formation was seen over the small amount in the nosubstrate control that reflects enzyme automethylation (Fig. 3, G-I). We did, however, find a role for the N-terminal residues in facilitating PRMT7-substrate interactions, because MMA formation in the H4(14 -22) peptide was some 250-fold less than in the H4(1-21) peptide (Fig. 3, A and D).
We also searched for substrates for GST-PRMT7 with similar basic RXR motifs that were not derived from histones. Three peptides, P-TAT bearing a RKKRRQRRR sequence, P-SmD3 bearing RGRGK and RGRGR sequences, and P-RPL3 bearing a HRGLRK sequence, were analyzed (Table 1 and Fig. 4). Each of these peptides was recognized by GST-PRMT7 and only formed MMA. With P-TAT, the incorporation of 3 H-methyl groups approached the incorporation seen with the histone H2B(23-37) peptide ( Fig. 2A), validating the preference of PRMT7 for the RXR motif in a high density of basic residues. P-SmD3 incorporated about 20% of the radioactivity seen with P-TAT, suggesting the importance of basic residues adjacent to the methylation sites. Finally, the P-RPL3 peptide only incorporated about 5% of the radioactivity of P-TAT. Here the insertion of a single residue between the pair of arginine residues appears to decrease PRMT7 recognition. Although we have not identified the precise sites of methylation in these peptides, all of them contain either RXR motifs or RXXR motifs in basic contexts. Taken together, the results shown in Figs. 2-4 suggest that bacterially expressed GST-human PRMT7 specifically recognizes pairs of arginine residues separated by one residue in a highly basic context to form only the monomethylated derivatives. This specificity is similar to that previously seen with the insect cell-expressed mouse PRMT7 enzyme (8).
Identification of Residues in PRMT7 Involved in Recognition of Basic RXR Substrates-Because the specific targeting of a basic RXR site by PRMT7 is unique among PRMT family members, we next tried to identify the residues in the enzyme important for substrate recognition. In all PRMT enzymes, the region following methyltransferase motif II has been associated with substrate binding (1, 2, 33) (Fig. 5). Here, PRMTs contain two highly conserved glutamate residues in a "double E" loop that directly bind the target arginine residue. We were intrigued to note that human PRMT7 has an aspartate residue (Asp-147) as the third residue following the initial glutamate in a position where all other PRMTs (except for PRMT9) contain a glycine residue. In addition, PRMT7 is the only human PRMT with an acidic residue (Glu-149) as the fifth residue following the initial glutamate of the double E loop. From the crystal structures of mouse and Caenorhabditis elegans PRMT7 (34, 35), Asp-147 and Glu-149 would be expected to face the substrate binding cavity. Interestingly, trypanosome PRMT7, which contains Gly and Met instead of Asp and Glu at these two positions (Fig. 5), methylates histone H4 at the non-RXR Arg-3 site, indicating that it may have a different substrate specificity from the human or mouse PRMT7 (9). However, it is clear that all of these species are type III PRMTs that only produce MMA as a product. We thus hypothesized that Asp-147 and Glu-149 might be important for targeting basic RXR sequences. Therefore, D147G and E149M mutations were made, changing the former into the conserved residue of the other PRMTs and the latter into the corresponding residue in TbPRMT7. We also constructed enzymes where residues 145-147 (LFD) in the double E loop of GST-PRMT7 were mutated to the conserved WMG sequence present in the type I PRMT1, -2, -3, -6, and -8 enzymes or mutated to the LLG sequence in the type II PRMT5 enzyme (Fig. 5). The C-terminal domain of human PRMT7 is not well conserved, with only one glutamate (Glu-478) present in the residues corresponding to the double E loop (Fig. 5, red box). To test the importance of this glutamate residue, a E478Q mutation was introduced. Finally, LC-MS/MS analysis on insect cell-expressed mouse PRMT7 (8) indicated possible phosphorylation sites on residues Thr-203 and Ser-204, which are conserved in human PRMT7. To determine whether these modifications might regulate the activity of the enzyme, T203E and S203E mutations were made to mimic the phosphorylated Thr-203 and Ser-204 residues.
Amino acid analysis was carried out on all of the GST-PRMT7 mutants with the H2B(23-37) peptide as a substrate. As indicated in Fig. 6, all the mutations lowered the methyl group incorporation in H2B (23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37) to different levels compared with the wild-type enzyme (Fig. 6H), although none of them altered the type III methyltransferase activity because only MMA product was observed. The radioactivity in MMA generated by GST-D147G and GST-E149M with the peptide (Fig. 6, A and B) was ϳ6 and ϳ17% of that by the wild-type enzyme, respectively. However, the enzyme automethylation in these mutants in the absence of peptide was 4-and 3-fold higher when compared with that of the wild type (Fig. 6H, blue  line). Thus, these two mutants were more active with auto-  Table 1) were methylated by GST-PRMT7 (0.48 M) with [ 3 H]AdoMet (0.7 M) at 23°C for 18 h and detected by amino acid analysis as described in the legend to Fig. 2, except that trichloroacetic acid precipitation was used with the P-SmD3 substrate. A-C, PRMT7-catalyzed methylation products of peptides derived from SmD3, RPL3, and TAT proteins, respectively. methylation but less active with the H2B(23-37) peptide containing the HBR domain, giving a clue that Asp-147 and Glu-149 might be essential in regulating the substrate specificity. Although the T203E and T204E mutants showed similar levels of product accumulation to the wild-type enzyme (Fig. 6, C and  D), the E478Q mutant surprisingly displayed only Ͻ1% of radioactivity incorporation to H2B(23-37) (Fig. 6E), demonstrating the importance of the acidic residue Glu-478 in the region corresponding to the double E loop in the C-terminal domain. Little or no activity was observed for the LFD(145-147)WMG and the FD (146 -147)LG mutants designed to mimic the double E loops of the type I and type II enzymes, respectively (Figs. 5 and 6 (F and G)).
Substrate Specificity Changes Determined by Kinetic Analysis of GST-PRMT7 Mutants with Peptide Substrates-We then performed detailed kinetic characterization on the wild-type GST-PRMT7 and mutants GST-D147G, GST-E149M, and GST-S204E using different peptide substrates. Under our conditions, product formation catalyzed by GST-PRMT7 is linear for 24 h at 23°C (Fig. 1). In these experiments, we incubated the reactions for 5 h, a time at which substrate consumption would be expected to be less than 2%. The concentration of peptide substrate was varied, whereas the [ 14 C]AdoMet was fixed at 15 M for the kinetic analyses. As shown in Fig. 7, the Michaelis-Menten curves showed similar patterns for H2B(23-37) and P-TAT, which both contain very basic RXR target sites. The D147G and E149M mutants showed very little activity toward these two peptides compared with the wild type. However, the catalytic pattern of the wild-type and mutant enzymes for the P-SmD3 substrate containing the less basic RGRG repeats was distinct. Although the P-SmD3 peptide was a markedly poorer substrate than H2B(23-37) and P-TAT, the relative activities of the D147G and E149M mutations were much greater with the P-SmD3 substrate. These results suggest that Asp-147 and Glu-149 are particularly important in recognizing the basic residues surrounding the target site. Interestingly, the activity of the E149M enzyme was greater than that of the D147G enzyme with the more basic H2B(23-37) and P-TAT substrates but lower with the less basic P-SmD3 substrate. The k cat and K m values for all of the mutants tested are shown in Table 2, where the k cat /K m value best reflects the overall substrate specificity. Compared with wild-type GST-PRMT7, the k cat /K m values are 23-and 38-fold lower for GST-D147G and GST-E149M in  Table 2. Each assay was performed in duplicate, and the error bars indicate the range. The solid lines represent the curves from the fitted k cat and K m values shown in Table 2. Data are shown for the wild-type enzyme (red), the D147G mutant (blue), and the E149M mutant (green).

TABLE 2
Kinetic data for selected GST-PRMT7 mutants with different substrates P81 paper assays were performed with varying concentrations of each substrate (typically 0 -200 M) incubated with 15 M [ 14 C]AdoMet and 0.5 or 0.75 M GST-PRMT7 or its mutant in the reaction buffer (50 mM potassium HEPES (pH 7.5), 10 mM NaCl, and 1 mM DTT) in a total volume of 30 l at 23°C. After 5 h, 25 l of the reaction mixture was spotted onto the P81 filter paper, which was washed and dried as described under "Experimental Procedures." The radioactivity on the P81 paper was detected by liquid scintillation as described. The initial velocity as a function of substrate concentration was fit to the Michaelis-Menten equation with the Prism 6 program, which gave the k cat and K m and their S.E. values as shown below.

Enzyme
Substrate   Fig. 2A and Fig. 3I. In H, the reaction was incubated for 20 h but under conditions otherwise identical to those shown in A-G, where the incubation was 18 h.
methylating H2B (23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37) and 52-and 28-fold lower for these two mutants in methylating P-TAT, whereas they are only 2and 3-fold lower for them in methylating P-SmD3. These data indicate that Asp-147 and Glu-149 are responsible for the selectivity of PRMT7 for RXR residues in basic sequence contexts. Looking at the H2B(23-37) substrate only, GST-D147G The reactions were allowed to proceed at 23°C for 20 h before quenching by trichloroacetic acid precipitation. Incubations in the absence of GST-PRMT7 or GST-GAR were used as controls. After acid hydrolysis, the methylated amino acid derivatives were analyzed by high resolution cation exchange chromatography together with standards of ADMA, SDMA, and MMA as described under "Experimental Procedures." 3 H radioactivity (solid lines) and ninhydrin absorbance of the methylated arginine standards (dashed lines; elution positions indicated) were detected with liquid scintillation counting and 570-nm absorbance, respectively. Red, wild-type GST-PRMT7 with GST-GAR; green, wild-type GST-PRMT7 alone; blue, LDIG(69 -72)AAAA mutant enzyme with GST-GAR; purple, GST-GAR alone. A2, magnification of the radioactivity scale to show PRMT7 automethylation (green), the residual activity of LDIG(69 -72)AAAA (blue), and the absence of ADMA and SDMA in the reaction products (red). A3, premethylated GST-GAR generates only MMA in a PRMT7-catalyzed reaction. GST-GAR (6 g) was preincubated with unlabeled AdoMet (125 ⌴ final) and GST-PRMT7 (3 g) in 40 l of reaction buffer at 23°C for 24 h to enrich for MMA. Next, the unlabeled AdoMet concentration was lowered to 0.05 M by adding fresh reaction buffer, and the sample was concentrated using an Amicon Ultra centrifugal filter (Millipore). Then [ 3 H]AdoMet (0.7 M) and extra GST-PRMT7 (2.4 g) were added into the reaction mixture, which was incubated at 23°C for another 20 h before acid hydrolysis and cation exchange chromatography. The radioactivity of 3 H-labeled protein residue was plotted in comparison with the standards. B, time course of GST-GAR methylation by double E loop GST-PRMT7 mutants. B1, fluorography of 3 H-methylated GST-GAR after reaction with wild type or mutant GST-PRMT7 as described under "Experimental Procedures." Dried gels were exposed to autoradiography film for 4 days at Ϫ80°C. Reactions were run in duplicate on separate gels, one of which is shown here. B2, quantification of GST-GAR methylation catalyzed by wild type GST-PRMT7 (red), D147G (blue), and E149M (green). Data were acquired from the densitometric analysis of the autoradiography film and are shown as activity relative to that at the 10-h time point for the wild-type enzyme. Error bars, range of duplicate experiments.
showed a 32-fold smaller k cat and a similar K m compared with the wild type, indicating that Asp-147 in the double E loop is mainly involved in catalysis; GST-E149M showed a 9-fold smaller k cat and a 4-fold larger K m compared with the wild type, indicating that Glu-149 in the double E loop is involved in both catalysis and substrate binding. As for GST-S204E, the k cat /K m values for all three substrates are generally similar to the wild type because the k cat and K m for individual peptide are both decreased. It is possible that the potential phosphorylation on Ser-204 might increase the substrate binding affinity. Specificity of GST-PRMT7 Mutants for the Protein Substrate GST-GAR-To confirm the alteration of substrate specificity for the GST-D147G and GST-E149M PRMT7 mutant enzymes, we also analyzed the substrate GST-GAR, consisting of residues 1-148 of human fibrillarin (7,36) and containing similar RGRG repeats as in the P-SmD3 peptide ( Table 1). We first performed amino acid analysis on the reaction products. Fig. 8A shows that wild-type GST-PRMT7 generated only [ 3 H]MMA on GST-GAR (Fig. 8A1). A small amount of [ 3 H]MMA automethylation was found for the enzyme-alone control (about 1.4% of that with GST-GAR) (Fig. 8A2, green line); less than 0.1% of [ 3 H]MMA was detected with the AdoMet binding-deficient mutant enzyme GST-LDIG(69 -72)AAAA and GST-GAR (Fig. 8A2, blue line). In addition, to verify if PRMT7 adds a second methyl group to monomethylated arginine residues, GST-GAR was premethylated with unlabeled AdoMet to enrich for MMA, and then [ 3 H]AdoMet was added for detection of methylated product. Again, only the [ 3 H]MMA peak was observed with no detectable [ 3 H]SDMA or [ 3 H]ADMA formation (Fig. 8A3). These results confirm that GST-PRMT7 is unable to transfer a second methyl group to the arginine site where one methyl group is already present.
Next, SDS-PAGE and fluorography were performed to monitor the time courses of GST-GAR methylation catalyzed by wild-type GST-PRMT7 and mutants GST-D147G and GST-E149M (Fig. 8B). The density of methylated GST-GAR bands on the fluoroimage (Fig. 8B1) was quantified, and product formation indicated by the relative radioactivity was linear for up to 10 h with the wild-type and mutant enzymes (Fig. 8B2). Importantly, the activities of the mutant enzymes were similar to that of the wild-type enzyme; GST-D147G showed, in fact, a higher catalytic activity. These results stand in marked contrast to those obtained above with the H2B(23-37) and P-TAT peptides, where very little activity was seen in the mutants compared with the wild type (Fig. 7, A and C). Therefore, the substrate specificity change is confirmed here: mutation of Asp-147 and Glu-149 on the double E loop does not greatly affect the specificity of PRMT7 for RGRG-containing GST-GAR but dramatically decreases its selectivity for substrates rich in basic residues.
Unusual Dependence of PRMT7 Activity on Temperature and Salt Concentration-During the reaction condition optimization, we found that PRMT7 had an unusual temperature dependence. When we tested the temperature effect on bacterially expressed human GST-PRMT7 with H2B(23-37) as a substrate in the P81 paper assay (Fig. 9A), the peak activity was found between 15 and 25°C. The enzyme remained ϳ50% active when the temperature was lowered to 5°C, whereas its The methylated products were separated with 12.6% SDS-PAGE, treated with EN 3 HANCE buffer, vacuum-dried, and exposed to autoradiography film for 3 days at Ϫ80°C. Densitometric analysis was done on scanned images of the film, and quantification of the data is shown as activity relative to that at 18°C. In each case, the assays were done in duplicate, and the error bars show the range of the data. activity dropped dramatically above 30°C. These results were surprising for a mammalian enzyme presumably working at body temperatures of about 37°C. To confirm this phenomenon, we also analyzed the temperature effect on the insectexpressed mouse PRMT7 (8) with H2B(23-37) as a substrate in the P81 paper assay (Fig. 9B) and the bacterially expressed human GST-PRMT7 with GST-GAR as a substrate in the fluorography assay (Fig. 9C). In both of these cases, a similar temperature dependence was found. These results indicate that PRMT7 is an unusual mammalian enzyme with significant activity at sub-physiological temperatures.

DISCUSSION
We have confirmed the type III activity of bacterially expressed GST-tagged human PRMT7 on basic RXR sequence motifs under our optimized reaction conditions. Product formation was significantly improved as we adjusted several conditions from our previous study (7), particularly the reaction temperature. The type III activity of GST-PRMT7 forming only MMA product has been confirmed with both GST-GAR protein derived from fibrillarin and various peptides derived from histone H4, histone H2B, splicing protein SmD3, ribosomal protein RPL3, and HIV-1 TAT protein as well as for PRMT7 automethylation. These results are consistent with previous reports on human and mouse PRMT7 (7,8) as well as studies on the trypanosome PRMT7 (6,9). Although PRMT7 may work synergistically with type I or type II PRMTs or other modulators in vivo, PRMT7 by itself only generates MMA residues.
We find that effective substrates for GST-tagged human PRMT7 consist of sequences with at least two closely spaced arginine residues with adjacent basic residues. Out of all of the peptides tested (Table 1), H2B(23-27) bearing a KKRKRSRK sequence and P-TAT bearing an RKKRRQRRR sequence are the best PRMT7 substrates, probably because they contain a high density of positive charges and multiple arginine residues, which may be preferably recognized by the extra two acidic residues (Asp-147 and Glu-149) in the double E loop (Fig. 7). H4(1-21) bearing a KRHRK sequence, P-SmD3 bearing RGRGK and RGRGR sequences, and P-RPL3 bearing an HRGLRK sequence are moderate substrates for PRMT7, suggesting the essential role of basic residue density. The H2B and H4 mutant peptides, where RXR sequences are modified to RXK or KXR sequences, are poor substrates. This demonstrates the importance of at least two arginine residues as the recognition site, which is also supported by the poor methylation of H4(1-8) and H3(1-7) sequences with a single arginine residue. These results are consistent with and extend those found earlier for the insect-expressed mouse enzyme (8).
Through site-directed mutagenesis, we show that the two additional acidic residues (Asp-147 and Glu-149) in the double E loop may be crucial elements in the substrate specificity of PRMT7 (Fig. 7). Enzymes where these residues are substituted with glycine and methionine, respectively, demonstrate dramatically decreased activity toward the H2B(23-37) and P-TAT substrates with high basic residue density but only moderately affected activity toward the P-SmD3 and GST-GAR substrates with relatively low basic residue density. It thus appears that the D147G and E149M mutations compromise the selectivity of PRMT7 for very basic RXR sites and that these two unique acidic residues in the PRMT7 double E loop are probably determinants for the preferable recognition of basic residues adjacent to or near the methylatable arginine residues.
We have also prepared PRMT7 with mutations in the double E loop designed to mimic conserved residues in type I and type II PRMTs (LFD to WMG and FD to LG, respectively). However, we found that neither of the mutated proteins displayed methyl transfer activity. Although the acidic residues do play an important role in conferring substrate specificity, mutating multiple residues in the double E loop destroys the ability of PRMT7 to methylate substrates. All of these mutations are within the catalytic N-terminal domain. PRMT7 also has an ancestrally duplicated C-terminal domain where many of the residues important for binding AdoMet or substrates have been altered. For example, only one of the glutamate residues in the corresponding region of the double E loop is present in the C-terminal domain (Glu-478). However, this domain is still required for enzyme activity (5), and in this study, we have shown that the E478Q mutation reduces the activity of PRMT7 to Ͻ1% of the wild-type activity. Although this second domain in PRMT7 is poorly conserved in sequence (5), it is certainly not redundant for substrate binding or catalysis.
In this study, we found special unique features of PRMT7 catalysis. First, the enzyme is more active at temperatures below the physiological mammalian value of 37°C and in fact can perform catalysis even at 5°C. Second, NaCl, EDTA, and some common metal ions were found to have a relatively strong inhibitory effect on PRMT7. These phenomena have not been observed for other members of the mammalian PRMT family. The unusual sensitivity to temperature may indicate that PRMT7 may be activated by cold stress to tissues.
Unfortunately, no in vivo substrates for PRMT7 have yet been unambiguously identified. Our results here, however, indicate that PRMT7-targeting sequences are often nucleotide binding sequences that compact and stabilize the structure of DNA/RNA, such as the HBR domain on histone H2B (28, 37), the RKKRRQRRR region (residues 49 -57) on HIV-1 TAT pro-tein (38), and the KRHRK region on histone H4 (39). The HBR domain at the end of the H2B N-terminal tail closely interacts with neighboring nucleosomal DNA and represses gene transcription (28,37). In yeast, it was found that HBR is necessary and sufficient for repressing ϳ8.6% of genes, many of which are located near telomeres or involved in the metabolism of vitamins and carbohydrates (28). No in vivo evidence has been found for posttranslational modifications on the HBR domain, although in vitro data are very clear that insect-expressed mouse PRMT7 monomethylates Arg-29, Arg-31, and Arg-33 in this region (8). HIV-1 TAT residues 49 -57 bear a nuclear localization signal and RNA-binding motif. This region has been reported to be asymmetrically dimethylated on Arg-52 and Arg-53 by PRMT6, leading to weaker interaction with the viral RNA and down-regulation of the transactivation capacity of TAT (40,41). Our study suggests that arginine monomethylation may also occur in this region. Posttranslational modifications catalyzed by PRMT7 at these arginine and lysine-rich regions, although poorly characterized, may play an essential role in modulating their interactions with DNA/RNA. More efforts are needed to validate these possible methylation sites and the biological consequences of their modification in vivo.