Histone H2A and H4 N-terminal Tails Are Positioned by the MEP50 WD Repeat Protein for Efficient Methylation by the PRMT5 Arginine Methyltransferase*

Background: PRMT5-MEP50 is an arginine methyltransferase with significant roles in development and cancer. Results: MEP50 binds to the histone fold domain and is essential for the efficient use of SAM by PRMT5. Conclusion: MEP50 is essential for methylation of histones H4 and H2A by PRMT5. Significance: The mechanism of histone methylation by PRMT5-MEP50 provides novel insight into methyltransferase mechanisms and therapeutic development. The protein arginine methyltransferase PRMT5 is complexed with the WD repeat protein MEP50 (also known as Wdr77 or androgen coactivator p44) in vertebrates in a tetramer of heterodimers. MEP50 is hypothesized to be required for protein substrate recruitment to the catalytic domain of PRMT5. Here we demonstrate that the cross-dimer MEP50 is paired with its cognate PRMT5 molecule to promote histone methylation. We employed qualitative methylation assays and a novel ultrasensitive continuous assay to measure enzyme kinetics. We demonstrate that neither full-length human PRMT5 nor the Xenopus laevis PRMT5 catalytic domain has appreciable protein methyltransferase activity. We show that histones H4 and H3 bind PRMT5-MEP50 more efficiently compared with histone H2A(1–20) and H4(1–20) peptides. Histone binding is mediated through histone fold interactions as determined by competition experiments and by high density histone peptide array interaction studies. Nucleosomes are not a substrate for PRMT5-MEP50, consistent with the primary mode of interaction via the histone fold of H3-H4, obscured by DNA in the nucleosome. Mutation of a conserved arginine (Arg-42) on the MEP50 insertion loop impaired the PRMT5-MEP50 enzymatic efficiency by increasing its histone substrate Km, comparable with that of Caenorhabditis elegans PRMT5. We show that PRMT5-MEP50 prefers unmethylated substrates, consistent with a distributive model for dimethylation and suggesting discrete biological roles for mono- and dimethylarginine-modified proteins. We propose a model in which MEP50 and PRMT5 simultaneously engage the protein substrate, orienting its targeted arginine to the catalytic site.

The protein arginine methyltransferase PRMT5 is complexed with the WD repeat protein MEP50 (also known as Wdr77 or androgen coactivator p44) in vertebrates in a tetramer of heterodimers. MEP50 is hypothesized to be required for protein substrate recruitment to the catalytic domain of PRMT5. Here we demonstrate that the cross-dimer MEP50 is paired with its cognate PRMT5 molecule to promote histone methylation. We employed qualitative methylation assays and a novel ultrasensitive continuous assay to measure enzyme kinetics. We demonstrate that neither full-length human PRMT5 nor the Xenopus laevis PRMT5 catalytic domain has appreciable protein methyltransferase activity. We show that histones H4 and H3 bind PRMT5-MEP50 more efficiently compared with histone H2A(1-20) and H4(1-20) peptides. Histone binding is mediated through histone fold interactions as determined by competition experiments and by high density histone peptide array interaction studies. Nucleosomes are not a substrate for PRMT5-MEP50, consistent with the primary mode of interaction via the histone fold of H3-H4, obscured by DNA in the nucleosome. Mutation of a conserved arginine (Arg-42) on the MEP50 insertion loop impaired the PRMT5-MEP50 enzymatic efficiency by increasing its histone substrate K m , comparable with that of Caenorhabditis elegans PRMT5. We show that PRMT5-MEP50 prefers unmethylated substrates, consistent with a distributive model for dimethylation and suggesting discrete biological roles for mono-and dimethylarginine-modified proteins. We propose a model in which MEP50 and PRMT5 simultaneously engage the protein substrate, orienting its targeted arginine to the catalytic site.
Because PRMT5 methylates multiple histone and non-histone arginines in vivo, a long-standing question is how PRMT5 selects among these many sites and multiple protein targets. Our previous work using crystallography, qualitative binding studies, and electron microscopy reconstruction suggested that substrate recruitment was primarily mediated by the MEP50 subunit. The structure of Xenopus laevis PRMT5-MEP50 that we solved (22) and the human PRMT5-MEP50 structure solved by Emtage and colleagues (23) revealed a buried and poorly accessible catalytic site. These observations are consistent with a conserved and significant stringency in substrate selection.
MEP50 is a seven-bladed WD repeat protein that is unusually acidic. We previously hypothesized that MEP50 served as a substrate presenter, similar to the initial hypothesis for the WD repeat protein Wdr5 (24) and RbAp48 (25) in binding histones H3 and H4, respectively. Our crystal structure of the XlPRMT5-MEP50⅐SAH complex and our electron microscopy reconstruction of the complex bound to its substrate nucleoplasmin supported this hypothesis. We and others previously showed that human PRMT5 is inactive in the absence of MEP50, consistent with a required role in binding substrate.
Quantitative enzymatic analysis of methyltransferases requires ultrasensitive techniques due to their low turnover. Previously described kinetic assays are either not commercially available, have low throughput, or provide signals too low to be of use for slow enzymes such as PRMT5. Therefore, we employ a novel ultrasensitive coupled continuous assay to measure kinetic parameters of other methyltransferases introduced previously by our colleagues (26).
Here, we test the function of MEP50 in promoting PRMT5 histone methyltransferase activity. We employ structural analysis as well as qualitative and quantitative methylation assays to measure enzymatic activity and binding affinity for histones to PRMT5-MEP50. We show that mutation within the MEP50 insertion finger impaired kinetic parameters of both histone and SAM substrates. Our results support the concept that MEP50 interacts directly with histones and the N-terminal domain of PRMT5 but may also contribute to active site remodeling in the C-terminal domain of PRMT5 for efficient methyl transfer. Our computational modeling revealed multiple modes of substrate-enzyme interaction consistent with our experimental data. These studies support the essential function of MEP50 in binding histone fold and presenting histone tail substrate to the active site of PRMT5 cross-dimer for efficient methylation.

EXPERIMENTAL PROCEDURES
Reagents-Chemicals and reagents were obtained from Sigma, Fisher, or Research Products International. [ 3 H]Methyland [ 14 C]methyl-SAM were purchased from PerkinElmer Life Sciences. SAM and SAH were purified by HPLC (Luna C18(2), Phenomenex), desalted, concentrated, and stored at Ϫ80°C (26). Reducing agent tris(hydroxypropyl)-phosphine was from Novagen. ATP detection was achieved using the ATPLite 1Step system (PerkinElmer Life Sciences); molecular biology grade water, with additional charcoal treatment, was used for all enzymatic assays. DNase I was from New England Biolabs.
Filter Binding Methyltransferase Assay and Fluorography-Assays were performed as described (22). Briefly, substrate protein was incubated with PRMT5-MEP50 enzyme at the indicated concentrations in 15 l of reaction buffer (20 mM Tris, pH 8.0, 10 mM DTT, protease inhibitors, 0.5 M [ 3 H]methyl-SAM of specific activity 78.2 Ci mmol Ϫ1 ) for 20 min at 30°C. The reaction mixture was spotted on P81 filter paper, washed with sodium carbonate buffer (0.1 M, pH 8.5), air-dried, and analyzed via scintillation counter (Wallac Winspectral 1414 LSC). Alternatively, the reaction mixture was run on an SDS-polyacrylamide gel, stained and imaged, soaked in Amplify (NAMP100, GE Healthcare), dried, and exposed to film.
Nucleosome Methyltransferase Assay-The assay was performed with 2 g of mononucleosomes treated with 2 units of DNase I or reaction buffer alone for 1.5 h at 30°C prior to the addition of XlPRMT5-MEP50 (300 nM final concentration) in the presence of [ 3 H]methyl-SAM (0.3 M final concentration; specific activity 78.2 Ci mmol Ϫ1 ). Similar experiments using recombinant H2A and H4 (2.5 M) were run in parallel. After 20 min, reaction mixtures were run on an SDS-polyacrylamide gel, stained and imaged, soaked in Amplify (NAMP100, GE Healthcare), dried, and exposed to film.
FLAG Pull-down-Anti-FLAG M2 antibody-coupled agarose beads (20 l) were incubated with equimolar amounts (50 nM) of XlMEP50 (wild type or mutants) and FLAG-tagged HsPRMT5 in 250 l of TBS (150 mM NaCl, Tris-HCl, pH 8.0) at 4°C for 2 h. As a negative control, XlMEP50 was incubated under the same conditions without HsPRMT5. The suspension was then transferred to a Mini-spin column (USA Scientific, Inc.) and centrifuged (30 s, 500 ϫ g). After intense washing with TBS, the beads were boiled for 10 min in SDS loading dye for elution. The elution samples were analyzed via SDS-PAGE and Western blot with ␣-PRMT5 (Millipore) and ␣-MEP50 antibodies.
Luciferase-based Coupled Enzymatic Assays-Kinetic parameters were determined at 25°C by monitoring luminescence using a SpectraMax L instrument configured with two photomultipliers (photo counting mode; Molecular Devices) in a 96-well half-area flat bottom plate (Corning, Inc., catalog no. 3992). Luminescence was measured in relative light units. Briefly, 40 l of buffer B2 (containing 20 -200 nM PRMT5 enzymes) was mixed with SAM substrate (10 l of 125 M solution, final saturating concentration of SAM at 25 M; higher saturating concentrations are required for mutants with higher SAM K m ). Under our experimental conditions, the typical Michaelis-Menten model is not suited to characterize the behavior of PRMT5 from both X. laevis and C. elegans. In vitro, these enzymes have poor activity, with turnover rates (k cat ) as low as 20 h Ϫ1 , and the use of high nanomolar concentration of methyltransferase is generally required. During the early stage of substrate titration (SAM or peptide substrates), enzyme and substrate concentrations have the same order of magnitude. Therefore, a significant fraction of total substrate is bound to the enzyme (i.e. enzyme-substrate complex). The Morrison kinetic model accounts for this early substrate depletion (Equation 1). Therefore, initial rates were plotted against substrate concentrations and fitted to this kinetic model to yield corresponding K m and k cat values (29), where [E] T and [S] T are the total concentration of enzyme and substrate, respectively.
Competition Experiments between Histone H2A and Other Histones/Peptides for Binding onto XlPRMT5-MEP50 or CePRMT5-Experiments were performed in a 100-l total volume with 50 mM MOPS, pH 7.0, 100 nM XlPRMT5-MEP50 or 100 nM CePRMT5, 20 microunits of SeMTAN, histone H2A with concentration kept at 2.0 M, increasing amounts of histone/peptide substrates (0 -12.6 M final concentration), and 1 mM tris(hydroxypropyl)-phosphine. Reactions were initiated by adding 80 l of the above samples to PCR tubes already loaded with 20 l of SAM reagent (25 M final [ 3 H]methyl-SAM concentration; 10 Ci/reaction). After 15 min at 25°C, TFA (3 l, 10% (v/v)) was added to quench the reactions. Samples were kept at Ϫ80°C before processing by HPLC. Methyl transfer onto histone H2A was then quantified by liquid scintillation counting and plotted against matching concentrations of histone/peptide competitors.
Binding Affinity of Histone H2A for XlMEP50 and XlPRMT5-MEP50 -XlMEP50 was freshly prepared; a gel filtration was used as the last purification step (25 mM ADA, 100 mM NaCl, 1 mM ␤-mercaptoethanol, pH 6.5). Samples were prepared as described for the luciferase-based assay, with XlPRMT5-MEP50 and histone H2A concentration set at 100 nM and 2.0 M, respectively. Exogenous XlMEP50 was added (0 -20 M), and the composition of samples was kept constant (adjusted with gel filtration buffer: 25 mM ADA, 100 mM NaCl, 1 mM ␤-mercaptoethanol, pH 6. Considering the binding between exogenous MEP50 and full-length histone, the following expressions can be written,

Mechanisms of PRMT5-MEP50 Histone Methylation
Considering the methyl transfer reaction, the following relationships can be written.
and by analogy, Peptide Array Binding Studies-A library of 20-mer peptides spanning the sequences of histones was generated with or without known modifications and in various combinations (sequences available at JPT Peptide Technologies GmbH Web site), as described previously (22). For binding studies, FLAG-HsPRMT5 and XlMEP50 were preincubated in equimolar amounts (66.5 nM) in KCl/HEPES buffer for 30 min at room temperature to form a complex. The complex was applied to individual histone code peptide microarrays for 1 h at 30°C. For detection of binding events, the microarrays were incubated with anti-FLAG mouse monoclonal antibody (Pierce, catalog no. MA1-91878) or anti-MEP50 rabbit polyclonal antibody. Upon washing and incubation with fluorescently labeled anti-body (DyLight649 anti-mouse IgG (Thermo, catalog no. 35515) or DL649 anti-rabbit IgG (Pierce, catalog no. 35565)), the microarrays were washed again and dried. Incubation with primary and secondary antibody alone was used as a control. Each microarray was scanned using GenePix Autoloader 4200AL (Molecular Devices; pixel size, 10 m). Signal intensity was evaluated using GenepixPro software (Molecular Devices). Further evaluation and representation of results was performed using the R statistical programming system (version 2.11.1).
Structural Alignments, Binding Site Predictions, and Rigid Body Docking-All structural figures were visualized using VMD version 1.9.1 (30). Structural alignments were performed using the STAMP and Multiseq alignment tools within VMD (31,32). Rigid body docking was performed at the ClusPro 2.0 server (33,34).
ClusPro 2.0 Docking-Docking of H2A-H2B and H3-H4 histone dimers was done using the following attractive residues. Residue numbers are listed as Protein Data Bank chain letterresidue number in PRMT5, MEP50, and histones used for attraction in rigid body docking in ClusPro 2.0. Chain Y is the cross-dimer MEP50 from chain Q PRMT5. Chains C and D are H2A and H2B with the N-terminal tails removed, and chain A and B are H3 and H4 with the N-terminal tails removed. The residues below were assigned attractive forces in ClusPro. For H2A-H2B dimer docking: receptor attraction,

RESULTS
Our previous work defined two distinct relative classes of MEP50 molecules within the PRMT5-MEP50 tetramer: the directly bound MEP50 and the cross-dimer MEP50 (22). Additionally, both our solved X. laevis PRMT5-MEP50 structure and the solved human complex structure (23) demonstrated unique N-and C-terminal domains of PRMT5 simply connected by a loop. Based on its structural organization and residue conservation and our electron microscopy images of XlPRMT5-MEP50 complexed with substrate, we hypothesized that the cross-dimer MEP50 is responsible for organizing substrate for the PRMT5 catalytic domain.
Structural Arrangement of PRMT5 and MEP50 -To test the conserved structural relationships between MEP50 and the distinct N-and C-terminal domains of PRMT5, we performed C␣-carbon backbone alignments between the X. laevis and human PRMT5-MEP50 structures (Protein Data Bank codes 4G56 and 4GQB, respectively) (Fig. 1A) using the STAMP alignment tool in VMD MultiSeq (31,32). First we aligned only the entire PRMT5 molecule (as indicated by the black box in Fig. 1B and the schematic in Fig. 1A) and plotted the per-residue C␣ RMSD values (Å) for PRMT5 (purple) and both MEP50 molecules (directly bound (pink) and cross-dimer (blue)) ( Fig.  1B, top row). PRMT5 itself was only modestly well aligned, with many regions containing an RMSD of Ͼ2 Å. The directly bound MEP50 was induced into modest alignment by PRMT5, whereas the cross-dimer MEP50 was not aligned well at all. When we aligned the PRMT5 N-terminal domain that directly contacts MEP50, we observed tight alignment (RMSD Ͻ2 Å) for the PRMT5 domain as well as the directly bound MEP50 (Fig. 1B, second row).
When we aligned the C-terminal domain of PRMT5 that does not directly contact either MEP50, we observed no alignment of the directly bound MEP50 but strong alignment of the cross-dimer MEP50 (RMSD Ͻ2 Å; Fig. 1B, third row). Despite the different crystal forms of the solved structures, which could possibly lead to packing artifacts, this relationship is strikingly apparent. Therefore, we concluded that a conserved relationship between the catalytic C-terminal domain of PRMT5 and the cross-dimer but non-contacting MEP50 exists. This relationship is unlikely to be allosteric due to the multiple domain boundaries and intermolecule contacts that would be required to communicate this information.
XlPRMT5 Catalytic Domain Is Inactive-We and others have previously demonstrated that PRMT5 alone has exceedingly modest histone methyltransferase activity in the absence of MEP50 (22,23). Because we showed in Fig. 1 that a functional relationship probably exists between the catalytic domain of PRMT5 and the cross-dimer MEP50, we produced recombinant XlPRMT5 catalytic domain (PRMT5⌬N, residues 291-633) to test if the catalytic domain alone was active. This truncated protein was soluble in E. coli and sedimented primarily as monomer in analytical ultracentrifugation ( Fig. 2A) (data not shown). We then measured XlPRMT5(291-633) enzymatic activity using the filter-binding methyltransferase assay on multiple known peptide and protein substrates of PRMT5-MEP50 in the absence or presence of MEP50 (Fig. 2B). XlPRMT5(291-633) did not exhibit any activity on any of these substrates, whereas intact full-length HsPRMT5 did recover activity with XlMEP50 added in trans. A 1-year exposure of a fluorogram did reveal very low levels of histone methyltransferase activity by full-length HsPRMT5 in the absence of XlMEP50, compared with robust activity of the XlPRMT5-  MEP50 complex or XlMEP50 added in trans to HsPRMT5 (Fig.  2C). We concluded from these experiments that only the fulllength PRMT5-MEP50 complex is able to efficiently methylate its substrates.
To demonstrate the strict requirement for MEP50 in promoting PRMT5 histone methyltransferase activity, we titrated XlMEP50 with constant HsPRMT5 and measured histone H2A methylation in the filter binding assay. We observed a MEP50 dose-dependent increase in activity maximal at the 4:4 stoichiometry with PRMT5, consistent with a required role for MEP50 in organizing substrate (Fig. 2D).
Histone Affinities for XlPRMT5-MEP50 -Our results to this point implicate MEP50 in its essential function in promoting PRMT5 methyltransferase activity. However, in vivo, PRMT5-MEP50 has distinct protein substrate selections in different contexts. In particular, H4 Arg-3 methylation is primarily observed on chromatin, whereas H2A Arg-3 methylation is observed on soluble, cytoplasmic histones (35)(36)(37). Soluble H2A is heterodimerized with H2B but typically not complexed with H3-H4 in the cytoplasm or nucleus when in solution.
Chromatin-bound histones are all found within the same nucleosome embedded in DNA.
The response observed with fluorography during film exposure is non-linear. To overcome the drawbacks of this technique, [ 3 H]methyl transfer was quantified by liquid scintillation counting. We performed similar competition experiments and separated histones by reversed-phase HPLC (Fig. 3C) (38). Our results confirmed the highly efficient histone H4 competition with histone H2A for Arg-3 methylation (Fig. 3D). Furthermore, histone tail peptide H4(1-20) inhibited methyl transfer toward H2A but to a lesser extent (i.e. 25% at 14 M). With  D). F, FLAG-HsPRMT5-XlMEP50 complex was incubated on ultrahigh density histone peptide arrays. PRMT5-MEP50 binding on the histone peptide scan data was extracted, and relative binding levels were plotted as a heat map, with no to low signal as white to light yellow, and high relative binding was plotted in red. Histone amino acid sequence numbers are represented at the top of the plots. The histone fold domain is indicated as a gray box, and the substrate residue Arg-3 (R3) on both H2A and H4 is indicated. G, relief of methyltransferase activity through the addition of MEP50. In this model used to determine the affinities of histone H2A for MEP50, P/M*⅐H represents the PRMT5-MEP50⅐histone complex, where only the histone fold is bound to the MEP50 presenter. P*/M*⅐H represents the PRMT5-MEP50⅐histone complex, where the histone fold is bound to the MEP50 presenter and the histone tail is bound to the enzyme active site. M*⅐H represents the complex between histone and exogenous MEP50. Each step is characterized by a constant (i.e. k 1 , k Ϫ1 , k 2 , k Ϫ2 , k cat , k 3 , and k Ϫ3 ). G, transferase reaction catalyzed by XlPRMT5-MEP50 was followed continuously at pH 7.7 using the luciferase-based assay with histone H2A as substrate (H2A fixed at 2 M; see "Experimental Procedures"); XlMEP50 was added to the reactions (0 -20 M), and resulting transferase activities were recorded. Exogenous XlMEP50 competes with XlPRMT5-MEP50 for H2A binding, and methyl transfer is inhibited with increasing concentrations of XlMEP50. A dissociation constant (K d ) of H2A for both exogenous XlMEP50 (KЈ d H2A:MEP50) and PRMT5-associated MEP50 (K d H2A:complex) was determined.   (42), so we tested the ability of histone H3 to displace the activity toward H2A (Fig. 3D). PRMT5-MEP50 was unable to methylate histone H3 (no incorporation of 3 H-methyl at pH 7.0 when using 25 M SAM and 100 nM enzyme). However, we observed a strong competition between histones H3 and H2A. Although both histones H4 and H3 display similar competition for PRMT5-MEP50, the highest concentrations of H3 were insufficient to achieve complete inhibition of PRMT5 methylation of histone H2A. Our results suggest the presence of specific histone fold binding regions for the XlPRMT5-MEP50 complex.
CePRMT5 Displays Affinity for the Histone Fold Domain of H4 without Assistance from a MEP50 Binding Partner-We tested the ability of CePRMT5 to bind the histone fold domain using similar competition experiments with a fixed concentration of H2A and increasing concentration of histone/peptide competitors. Although CePRMT5 does not associate with a MEP50 homologue, we did observe strong displacement of H2A by the H4 histone fold domain in comparison with the H4(1-20) peptide (Fig. 3E). Likewise, the non-substrate H3 competed efficiently against H2A for binding onto CePRMT5. These data suggest that both CePRMT5 and XlPRMT5-MEP50 enzymes may have different binding mechanisms for their protein substrates.
Histone Peptide Array Interaction Studies-To further test our hypothesis that XlPRMT5-MEP50 may bind histones H3 and H4 through their histone fold, which would explain the competition of H2A activity by these histones, we employed an extremely high density histone peptide array containing peptides covering the entire sequence of the core histones. We incubated FLAG-HsPRMT5 complexed with XlMEP50 on the array and probed with anti-FLAG and anti-MEP50 antibodies. We extracted relative binding data from these assays and plotted the signal onto the core histone sequence (Fig. 3F). Strik-ingly, the highest binding signals were obtained on histone fold or C-terminal peptides of histones H3 and H4. We also determined the influence of histone post-translational modifications on complex binding. In particular, pronounced loss of H4 C-terminal tail binding by HsPRMT5-XlMEP50 was observed upon phosphorylation of residue Tyr-98 (data not shown). Although this modification has not yet been observed in vivo, follow-up mass spectrometry and targeted bindings studies may specifically analyze this modification.
Nucleosomes Are Not Substrates for PRMT5-MEP50 -We previously showed that PRMT5-MEP50 is incapable of methylation of nucleosomes in vitro (22). To conclusively test this inability to methylate the typical cellular state of histones, we incubated PRMT5-MEP50 with H2A, H4, and recombinant mononucleosomes in the absence or presence of DNase I with identical buffer conditions (Fig. 4A). Mononucleosomes were not substrates of PRMT5-MEP50 compared with its significant methylation of free H2A and H4. After digestion of the DNA by DNase I, PRMT5-MEP50 did have H4 and H2A methyltrans-  4). B, schematic representation of the nucleosome core particle (Protein Data Bank code 1KX5), with H3 shown in blue, H4 in green, H2A in red, and H2B in yellow. The targeted H4 Arg-3 is shown (gray coloring and arrow), and the region on H3-H4 corresponding to the strongest sites of interaction on the peptide array is indicated.
ferase activity, confirming that DNA in the nucleosome directly inhibits the enzyme activity.
We used the results from our peptide array studies and mapped the domains of high binding for HsPRMT5-XlMEP50 onto the nucleosome core structure. We found that these peptides contained residues on the lateral surface of H3-H4. Binding on this surface would probably position the targeted H4 Arg-3 toward the PRMT5 catalytic site (Fig. 4B). Tight binding on this face would also orient the H2A Arg-3 away from the catalytic site, consistent with the greater activity directed toward H4 in the competition assays above. Because the highest binding regions of the octamer for HsPRMT5-XlMEP50 lie on or adjacent to the DNA superhelical path on the nucleosome, these observations may explain the absence of methyltransferase activity toward nucleosomes.
Function of the MEP50 Insertion Finger-To test our hypothesis that MEP50 orients and presents substrate to the PRMT5 catalytic domain, we inspected the structures to determine potential residues to mutate. The X. laevis and human PRMT5-MEP50 structures both contain a conserved MEP50 insertion finger that extends from the cross-dimer MEP50 and ends ϳ10 Å above the catalytic domain of PRMT5 (Fig. 5A and boxed  APRIL 10, 2015 • VOLUME 290 • NUMBER 15 enlarged view). This insertion finger is directly on the surface that we initially hypothesized to be responsible for substrate presentation (22). X. laevis MEP50 Arg-42 is 100% conserved among vertebrate MEP50 proteins (i.e. Arg-52 in human MEP50), is at the terminus of the finger, and is ϳ3.4 Å away from Glu-403 of the cross-dimer PRMT5, potentially forming a salt bridge to stabilize the finger (Fig. 5A, boxed enlarged view). We mutated this residue to glutamate or glutamine and produced PRMT5-MEP50 R42E and PRMT5-MEP50 R42Q protein complexes and MEP50 R42E and MEP50 R42Q alone (Fig. 5, B and  C). The PRMT5-MEP50 R42E and PRMT5-MEP50 R42Q complexes were purified from insect cells in the same manner as the wild-type complex. MEP50 R42E sedimented as a monomer in an analytical ultracentrifuge (data not shown). To confirm that MEP50 mutants interacted normally with PRMT5, we incubated the proteins with FLAG-tagged HsPRMT5, and both were enriched on anti-FLAG resin just like wild-type MEP50 (Fig. 5D).

Mechanisms of PRMT5-MEP50 Histone Methylation
To determine the ability or inability of MEP50 mutants to promote PRMT5 methyltransferase activity, we first used the filter binding assay. On multiple known peptide and protein substrates, MEP50 R42E only promoted low levels of PRMT5 activity (Fig. 5E). In contrast, the MEP50 R42Q -containing complex, without the charge reversal that would disrupt the putative salt bridge, exhibited activity similar to that of the wild-type complex. These data support the hypothesis that cross-dimer interaction of MEP50 and the PRMT5 catalytic domain through the MEP50 insertion loop is important for methyltransferase activity.
The filter binding assays are subject to some substantial error and do not report kinetic parameters in the end point readout that we employed. Therefore, we used the luciferase-based coupled assay to determine K m and k cat of the XlPRMT5-MEP50 and mutant complexes. Our kinetic parameters for PRMT5-MEP50, the MEP50 R42E , and MEP50 R42Q complexes are shown in Fig. 6, A and B, and in Table 1.
To gain further insight into the possible interactions between MEP50 and its cross-dimer PRMT5, we determined the kinetic parameters for the SAM substrate using wild-type and MEP50 Arg-42 mutant complexes at saturating concentrations of H4(1-20) peptide substrate (Fig. 6B). Surprisingly, the Arg to Glu mutation had the highest impact on the SAM K m , with values increasing from 3.3 to 27 M when compared with the wild-type MEP50. The Arg to Gln mutation only raised the K m for the methyl donor by a factor of 1.6 (Fig. 6C, bottom gray bar   graph). To reduce potential interactions of peptide substrate with MEP50, we then measured kinetic parameters for SAM using the short peptide substrate H4(1-7) (Fig. 6B). The PRMT5-MEP50 R42E complex displayed a dramatic loss of affinity for SAM with a K m 23-fold higher than that of wild-type complex ( Table 1). The SAM substrate only binds to the PRMT5 active site. Therefore, our kinetic results suggest that the MEP50 insertion loop is critical for histone binding and tail orientation and is also important for configuration of catalytically efficient PRMT5.
Kinetic Parameters of XlPRMT5-MEP50 Reveal Substrate Preferences-Our complete set of histone substrate kinetic parameters is shown in Table 1 and in Fig. 6, A and B, as a semilog scatterplot of K m versus k cat (higher efficiency in the plot shown in the top left, lower efficiency in the bottom right). The enzyme exhibited slow turnover with all substrates, on the order of 10 -50 h Ϫ1 . The most efficient substrates were the 1-20 and 1-21 histone peptide tails from H4 and from H2A and H2A.X-F (also known as H2A.X.3), with catalytic efficiencies ranging from 2.9 ϫ 10 4 to 1.3 ϫ 10 5 M Ϫ1 s Ϫ1 . These kinetic parameters for peptide substrates are reminiscent of results reported by Thompson and co-workers (42) using HsPRMT5-MEP50, although differences in N-terminal functionalization (amine versus acetyl) may account for subtle discrepancies (Table 1). Full-length histone H4 and nucleoplasmin were also reasonably efficient substrates. The short H4(1-7) peptide was a very poor substrate, consistent with our hypothesized role of MEP50 in organizing substrate.
Finally, we tested a monomethylated H2A.X-F R3me1 peptide in this assay to determine whether PRMT5-MEP50 is likely to be processive or distributive in its catalysis of dimethylation. The R3me1 peptide was ϳ20-fold less efficiently methylated, with the majority of this effect embedded in the K m , consistent with poor substrate binding. This result is reminiscent of the previously implicated distributive mechanism of catalysis to the dimethyl state (43,44). Additionally, we tested H2A and H4 mono-and dimethylation over time using specific antibodies. The accumulation of me2s depended on saturation of the me1 state in this assay. Although the use of antibodies is not quantitative, our results are probably incompatible with a processive methylation model and suggest a distributive model (Fig. 6D).
Prediction of Histone Binding Sites onto XlPRMT5-MEP50 Using Computational Docking-To determine potential interaction surfaces and residues on XlPRMT5-MEP50, we used the SPPIDER (45) and PredUs (46) prediction algorithms and mapped the sites identified by both (Fig. 7A). Then we docked tailless H2A-H2B dimers and H3-H4 tetramers (from Protein Data Bank entry 1KX5) with XlPRMT5-MEP50 in ClusPro 2.0 (33, 34) using experimental and predicted interaction constraints on PRMT5, MEP50, and the histones. These constraints included attractive forces on the histone regions identified on the peptide array binding studies and attractive forces on MEP50 Arg-42 and the PRMT5-MEP50 SPPIDER-and Pre-dUs-predicted interaction sites. Sample output models for putative H2A-H2B and H3-H4 interactions are shown in Fig. 7, B and C, respectively. We conclude that our experimental and predicted binding domains are consistent with our MEP50-dependent model for histone substrate recognition.

DISCUSSION
Based on our published structure, we previously proposed the "cross-dimer" model for MEP50 presentation of substrate to PRMT5. Here we used structural analysis, biochemistry, and enzymology to test our model of MEP50-dependent histone recognition and methylation by its coordinated PRMT5. These multiple approaches provided solid support for our hypothesis that MEP50 critically enhances the histone substrate methylation and will help guide future studies to uncover specific mechanisms of recognition of other PRMT5 target proteins.

PRMT5-MEP50 Structural and Enzymatic Conservation-
The structural conservation of the arrangement of MEP50 in the complex with PRMT5 gave us initial support for our model due to the clear coordination of the cross-dimer MEP50 from the catalytic domain of PRMT5. We gleaned further understanding of the role of MEP50 through structural alignment of the X. laevis PRMT5-MEP50 structure with the C. elegans PRMT5 in the absence of MEP50. We aligned the XlPRMT5 N-terminal domain, contacting XlMEP50, and the cross dimer XlPRMT5 C-terminal catalytic domain with the CePRMT5 chain and did not observe any regions of poor alignment, con- Kinetic parameters for the various tested substrates (histone H4, histone peptides, and SAM) are plotted, with the k cat (h Ϫ1 ) on the y axis and the K m (nM; logarithmic scale) on the x axis. Highest enzymatic efficiencies are obtained with substrates found in the top left quadrant, whereas low enzymatic efficiencies are obtained with substrates found in the opposite bottom right quadrant. Arrows indicate the loss (squared values) of enzymatic efficiency upon arginine monomethylation (purple) or upon mutation of MEP50 residue Arg-42 to glutamic acid (red) and to glutamine (green). For reference, enzymatic behavior of CePRMT5 is represented in blue. A, representation of kinetic parameters for histone substrates using saturating concentration of SAM. B, representation of kinetic parameters for SAM substrate using saturating concentration of histone substrates. C, impact of XlMEP50 R42Q and XlMEP50 R42E on catalytic turnover (k cat ; pink bars) and substrates' affinities (K m ; gray bars) for both peptide and SAM substrates. The decrease of methyl transfer is represented as a percentage of wild-type k cat , whereas the loss of affinity is given as -fold increase of wild-type K m . D, histones H2A or H4 were incubated with XlPRMT5-MEP50 and SAM. Reactions were stopped at 0, 1, 5, 10, and 15 min with the addition of SDS-polyacrylamide gel loading buffer and heating to 100°C. Reaction products were immunoblotted with monomethylarginine (R3me1)-or symmetric dimethylarginine (R3me2s)-specific antibodies. sistent with the absence of substantial allostery upon MEP50 binding.
Considering the relatively buried catalytic site in XlPRMT5 and its tetrameric nature, it was formally possible that MEP50 was necessary to organize substrate to overcome the entropic cost for substrate binding to PRMT5. To test this possibility, we produced the catalytic C-terminal domain of XlPRMT5 alone with the expectation that it would have robust and promiscuous activity. Strikingly, the catalytic domain did not have any appreciable histone methyltransferase activity, suggesting that the entire assembly is necessary. XlMEP50 added to the catalytic domain reaction did not stimulate activity, as expected because MEP50 only binds to the missing PRMT5 N-terminal domain. The catalytic domain alone does not dimerize due to the head-to-tail arrangement of PRMT5.
Human and X. laevis PRMT5 are highly homologous; as we previously showed, HsPRMT5 methyltransferase activity is stimulated by XlMEP50 (22). Here, we demonstrated that HsPRMT5 does have very modest histone methyltransferase activity, with a signal on the fluorogram appearing after a year of film exposure, compared with the significantly stronger activity observed after the addition of MEP50. Using our continuous luciferase-coupled assay, we did not observe any activity above background for HsPRMT5, so we are unable to assign kinetic parameters. The addition of XlMEP50 also gave a quantitative increase in methyltransferase activity when titrated into the assay, consistent with multiple active PRMT5 molecules within the tetramer.
Substantial support for the role of MEP50 in histone recognition and catalysis by PRMT5 was provided by our continuous coupled assay. In agreement with Wang et al. (43), we show that the K m for CePRMT5 methylation of H4 tail peptide is about 650 times higher than that observed for XlPRMT5-MEP50, and the catalytic efficiency for this peptide (ϳ150 M Ϫ1 s Ϫ1 ) is roughly equivalent to the loss of efficiency that we observe upon mutation of residue Arg-42 from the XlMEP50 insertion loop (ϳ45-75-fold). These observations that protein substrate binding, represented by substrate K m , is strongly dependent on the presence of MEP50 provide significant support for our hypothesis that MEP50 organizes substrate for PRMT5. Furthermore, the H4(1-7) short peptide would not be anticipated to be bound by MEP50, and therefore it should be poorly methylated by XlPRMT5-MEP50. Indeed, that is what we observed because this very short peptide exhibited a 184 M K m , the highest value among all substrates assayed.
The MEP50-Histone Fold Interaction Orients the Substrate for Methylation-Competition studies between H2A and fulllength histone/peptide H4 clearly highlighted the importance of histone fold domain for substrate recognition by PRMT5-MEP50. Intriguingly, the binding region for histone H2A may strongly overlap with some binding sites for histone H4, thus inhibiting methyl transfer onto H2A Arg-3. However, H3 is unable to fully abrogate the formation of H2A R3me1, so histones H2A and H3 may only share partial binding sites on the PRMT5-MEP50 complex. According to the strong competitive effects observed for both histone H3 and H4, the more physiologically relevant H3-H4 complex may be a substrate for XlPRMT5-MEP50 under specific, yet undetermined, experimental conditions.
Our histone peptide array studies are also consistent with the PRMT5-MEP50 preference for binding that we observed for full-length H3 and H4 compared with H2A. Binding studies on these ultrahigh density arrays will be generally useful for probing mechanisms of histone recognition by all histone-acting enzymes. Furthermore, our quantitative determination of binding affinity between full-length H2A and XlPRMT5-MEP50 (K d ϭ 1.78 Ϯ 0.06 M; Fig. 3H) is in good agreement with peptide array and kinetic results in which the H2A histone tail accounts for most of the binding onto the enzyme complex.
Recent publications have used kinetic and mass spectrometry analysis to model the reaction mechanism of PRMT5 and mutated PRMT1 in the generation of mono-and symmetric dimethylarginine (43,44,47). We independently confirmed this distributive model of progression to dimethylation through our demonstration that a synthesized monomethylated histone peptide has a 15-fold higher K m than its equivalent unmethylated peptide. We did not observe any substantive difference in the k cat between these two peptide substrates; these observations are probably incompatible with processive methylation and are therefore consistent with observations by Wang et al. (43) of a distributive mode of progression to dimethylation with CePRMT5 on monomethylated substrates. Furthermore, our highly specific R3me1 and R3me2s antibodies confirmed that the dimethylation state did not appear until saturating levels of R3me1 (Fig. 6D). Overall, these data and our prior observations of monomethylarginine in vivo point to the probably signifi- FIGURE 7. Prediction of histone binding sites onto the PRMT5-MEP50 complex. A, the predicted interacting residues on the cross-dimer pair of XlPRMT5-MEP50 was determined using the SPPIDER and PredUs algorithms and mapped onto the structure, as shown in yellow. Shown are docking of H2A-H2B dimer (orange/yellow) (B) and H3-H4 dimer (blue/green) (C) onto XlPRMT5-MEP50 (PRMT5 monomer (purple), its directly bound MEP50 molecule (pink), and the cross-dimer MEP50 (blue)), using ClusPro with attractive forces as determined by the peptide array and predictions in A.
cant, but unexplored, biological role of monomethylation. Future studies will be needed to identify this biological role and potential monomethylarginine effector proteins.
PRMT5-MEP50 Exhibits Specific Histone Binding and Methyl Transfer-An enigma in the PRMT5 literature is that histone H3 and nucleosomal H4 substrates are well documented to be methylated by PRMT5 in vivo (40,41,48,49), yet H3 is not a substrate in vitro. Here, we confirmed that PRMT5-MEP50 does not methylate histones in recombinant nucleosomes. The results of our histone-peptide array study and the mapping of the primary interaction domain to the histone fold of H3 and H4 are entirely consistent with the inability of PRMT5 to methylate nucleosomes and provide an example of a novel use of these arrays. Cross-talk from other post-translational modifications may explain this discrepancy, as suggested by observations showing that lysine acetylation stimulates PRMT5-MEP50 activity (50). Histones acetylated or otherwise modified in vivo may loosen histone-DNA contacts and may allow enough interaction to promote PRMT5 activity. We previously tested the enzyme complex's ability to methylate HeLa or hyperacetylated HeLa nucleosomes, but it was still unable to methylate these modified nucleosomes. Alternatively, other documented PRMT5 protein cofactors, such as COPR5 (15), Menin/Men1 (16), RioK1 (12), or ATP-dependent remodeling factors (18), may promote nucleosomal activity.
The MEP50 Insertion Finger Is Critical for the Complex's Substrate Binding and Activity-The MEP50 insertion finger directed over the cross-dimer PRMT5 catalytic domain is the WD40 repeat protein's most unique feature. We previously hypothesized that this insertion finger participates in organizing substrate for catalysis; alternatively, it may function to allosterically activate the PRMT5 catalytic domain, possibly mediated through the putative salt bridge between MEP50 Arg-42 and PRMT5 Glu-403. Our mutagenesis of MEP50 Arg-42 to glutamic acid led to a dramatic loss of catalytic efficiency for both SAM and peptide substrates with the reversal of charge (R42E). However, we observed a different behavior of the more conservative MEP50 R42Q complex, with kinetic parameters for SAM substrate nearly identical to those of the wild-type XlPRMT5-MEP50. This loss of activity in the R42E mutant was primarily observed in a higher K m , with only modest k cat effects (32-40% of wild-type k cat ; Fig. 6C), consistent with our hypothesis that histone/peptide substrate binding is dictated by MEP50 through intact positioning of the insertion finger. To our surprise, the K m for SAM substrate (directly bound to the C-terminal PRMT5 active site) was ϳ8-fold higher upon R42E mutation, suggesting that MEP50 may indeed have some small but direct influence on catalysis, possibly mediated through the putative salt bridge to the PRMT5 catalytic domain.
RbAp46, a WD repeat protein and MEP50 analog that participates in multiple histone acetyltransferase complexes, was shown to bind histone H4 on its side face, as in our model for MEP50 binding (25). RbAp46 residues involved in H4 binding were isostructural with our predicted histone interaction domain on MEP50, providing convergent evolutionary support for our hypothesis. Furthermore, the H4 residues involved in binding RbAp46 were in the ␣1 helix of the histone fold (residues 24 -41), consistent with the peptide array binding studies and providing support for the necessity of the histone fold interactions in the increased efficiency of full-length histone substrate methylation by PRMT5-MEP50.
We combined all of our direct observations and propose the model shown in Fig. 8. MEP50 is a presenter that 1) binds histones through their histone fold domain and 2) orients histone tail substrates toward the PRMT5 cross-dimer active site for efficient arginine methylation. We anticipate that this model will guide studies on other histone methyltransferase complexes as well as provide insight for future drug design.