Crystal structure of the human PRMT5:MEP50 complex

Stephen Antonysamya,1, Zahid Bondayb, Robert M. Campbellb, Brandon Doyleb, Zhanna Druzinaa, Tarun Gheyia, Bomie Hanb, Louis N. Jungheimb, Yuewei Qianb, Charles Raucha, Marijane Russella, J. Michael Saudera, Stephen R. Wassermanc, Kenneth Weicherta, Francis S. Willardb, Aiping Zhanga, and Spencer Emtagea,1

aLilly Biotechnology Center, Eli Lilly and Company, San Diego, CA 92121; bLilly Research Laboratories, Eli Lilly and Company, Indianapolis, IN 46285; and cLilly Research Laboratories Collaborative Access Team, Eli Lilly and Company, Advanced Photon Source, Argonne National Laboratory, Argonne, IL 60439

Edited* by Wayne A. Hendrickson, Columbia University, New York, NY, and approved September 20, 2012 (received for review June 12, 2012)

Protein methyltransferases (PRMTs) play important roles RAS to ERK signaling pathway through methylation of RAF in several cellular processes, including signaling, regulation, (10), regulates ribosome biogenesis through methylation and transport of proteins and nucleic acids, to impact growth, of ribosomal S10 (RPS10) (11), and plays an essential role differentiation, proliferation, and development. PRMT5 symmetri- in survival through regulation of eIF4E expression and cally di-methylates the two-terminal ω-guanidino nitrogens of ar- p53 translation. PRMT5 in association with MEP50 methylates ginine residues on substrate proteins. PRMT5 acts as part of a mul- cytosolic H2A to repress differentiation in ES cells (12). timeric complex in concert with a variety of partner proteins that During germ-cell development, the PRMT5 complex with tran- regulate its function and specificity. A core component of these scriptional repressor Blimp1 translocates from the nucleus to the complexes is the WD40 protein MEP50/WDR77/p44, which medi- cytoplasm at embryonic day 11.5 (E11.5), concomitantly with the ates interactions with binding partners and substrates. We have up-regulation of likely target genes (13). Recent data indicate that PRMT5 and MEP50 are components of the Grg4 complex, es- determined the crystal structure of human PRMT5 in complex with sential for its mediation of transcriptional repression (14). PRMT5 MEP50 (methylosome protein 50), bound to an S-adenosylmethio- has also been shown to inhibit the tumor suppressive function of nine analog and a peptide substrate derived from histone H4. The PDCD4 (15). The methyltransferase activity of PRMT5 can be structure of the surprising hetero-octameric complex reveals the controlled by of either MEP50 or of PRMT5 β close interaction between the seven-bladed -propeller MEP50 itself. Cyclin D1/CDK4 phosphorylates Thr5 on MEP50, activating and the N-terminal domain of PRMT5, and delineates the structural the methyltransferase activity of PRMT5 and resulting in pro- elements of substrate recognition. longed survival of tumor cells (16). In contrast, oncogenic mutants of Jak2 (V617F, K539L) phosphorylate tyrosine residues 297, 304, epigenetics | protein-protein complex | A9145C and 306 of PRMT5, disrupt association with MEP50 and down- regulate its methyltransferase activity on histone substrates (17). osttranslational methylation of lysine and arginine residues MEP50 was initially identified as a WD40 repeat protein that Pby protein lysine methyltransferases and protein arginine associates with PRMT5 and as an integral component of the 20S methyltransferases (PRMTs) alters the activity and interactions protein methyltransferase complex, termed the methylosome of substrate proteins, with crucial consequences to diverse cellular (18), and independently as an androgen receptor (p44) functions (1–3). Histone methylation is an epigenetic mark that that is overexpressed in prostate cancer cells (19). WD40 pro- plays a vital role in normal cell function, and whose dysregulation teins are known to play vital roles in various cellular networks – is associated with several diseases (4). (20, 21). These proteins function as protein protein and pro- The PRMT family of methyltransferases belongs to the largest tein–DNA interaction platforms and also as recognition mod- fi class (class I) of S-adenosylmethionine (AdoMet)-dependent meth- ules of posttranslational modi cations (22, 23). Recent studies fi yltransferase , responsible for the transfer of a methyl group on MEP50 identi ed the presence of two nuclear exclusion from AdoMet to the arginine side-chains of histones and other signals and three nuclear localization signals that control its proteins. PRMTs are further subdivided into type I, type II, type subcellular localization and its function as a transcriptional III, and type IV enzymes based on their patterns of arginine cofactor of androgen receptor during prostate development and methylation. Eleven human PRMTs have been identified to date tumorigenesis (24). MEP50 serves as a coactivator of both an- (5), and they all methylate the terminal guanidino nitrogen atoms drogen receptor and estrogen receptor in ovarian cells, and of arginine residues. Type I PRMT enzymes (PRMT1, -2, -3, -4, mediates hormonal effects during ovarian tumorigenesis (25). -6, and -8) generate ω-NG-monomethyl and ω-NG,NG-asymmetric MEP50 has been shown to interact with SUZ12 (a component di-methyl , whereas PRMT5 is a type II PRMT that cat- of the PRC2/EED-EZH2 complexes), and to selectively bind alyzes the formation of ω-NG-monomethyl and ω-NG,N′G-sym- H2A, and has been postulated to be an adapter protein between PRMT5 and its substrates (26). MEP50 and PRMT5 have been metric di-methyl arginine residues. PRMT7 was initially thought to fi have type II activity, but recent evidence suggests that it may be identi ed as components of the FCP1 complex, suggesting a link a type III that is only able to monomethylate substrates to form ω-NG-monomethyl arginine (6). A type IV enzyme that ca- talyses the formation of δ-N-methyl arginine has been identified in Author contributions: S.A., Z.D., T.G., B.H., C.R., M.R., J.M.S., F.S.W., and S.E. designed yeast (7). All PRMTs share the highly conserved methyltransferase research; S.A., B.D., Z.D., T.G., B.H., C.R., M.R., J.M.S., S.R.W., K.W., F.S.W., and A.Z. performed research; L.N.J. and Y.Q. contributed new reagents/analytic tools; S.A., B.D., catalytic domain, and several PRMTs contain additional domains Z.D., T.G., B.H., C.R., M.R., S.R.W., F.S.W., A.Z., and S.E. analyzed data; and S.A., Z.B., R.M.C., that modulate their activity and specificity. PRMT2, PRMT3, and B.D., Z.D., B.H., J.M.S., F.S.W., and S.E. wrote the paper. fi PRMT9 contain SH3, zinc nger, and TRP2 domains, respectively, Conflict of interest statement: This research was funded by Lilly Research Laboratories and PRMT5 contains a largely uncharacterized N-terminal region. and all the authors are employees of Eli Lilly and Company. In contrast to type I PRMTs, PRMT5 functions as part of various *This Direct Submission article had a prearranged editor. high molecular weight protein complexes that invariably contain Freely available online through the PNAS open access option. the WD-repeat–containing protein MEP50 (methylosome protein Data deposition: The atomic coordinates and structure factors have been deposited in the 50). PRMT5 associates with other cellular proteins in a context- Research Collaboratory for Structural Bioinformatics , www.rscb.org dependent manner (Fig. S1), enabling the methylation of a myriad (PDB ID code 4GQB). of cytoplasmic and nuclear substrates, including Sm proteins, 1To whom correspondence may be addressed. E-mail: [email protected] or , p53, histones H2A, H3, and H4, SPT5, and MBD2, and [email protected]. thereby plays a role in RNA processing, chromatin remodeling, This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. and control of gene expression (1, 8, 9). PRMT5 modulates the 1073/pnas.1209814109/-/DCSupplemental.

17960–17965 | PNAS | October 30, 2012 | vol. 109 | no. 44 www.pnas.org/cgi/doi/10.1073/pnas.1209814109 Downloaded by guest on September 30, 2021 between transcription elongation and splicing (27, 28). PRMT5 that eluted as a dimer in gel filtration columns, but had a pro- together with MEP50 is thought to constitute a core complex, pensity to aggregate and did not crystallize. which either binds pICln to methylate Sm proteins, or binds Coexpression of PRMT5 with MEP50 in insect cells increased Riok1 to methylate nucleolin (29, 30). the yield of soluble protein, and dramatically improved the ho- Protein methyltransferases are being actively pursued as drug mogeneity of the purified product, as judged by gel filtration (Fig. targets for various types of cancer where enhanced levels of 1 A–C and Fig. S2). In contrast to the homo-dimer observed from PRMT5 have been observed (4, 31–33). Despite the vital role PRMT5 alone, coexpression with MEP50 produced a tight com- played by PRMT5 in diverse cellular processes, and its potential plex that eluted from preparative gel filtration columns in the 400- as a drug target, there has been scant structural information on to 500-kDa molecular weight range. To additionally characterize PRMT5 or its interaction with MEP50. We present here the this complex, we determined the molecular weight by sedimenta- complex of human PRMT5 with MEP50 analyzed by chroma- tion velocity analytical centrifugation (Fig. 1D). The most abun- tography, sedimentation analysis, enzymology, and X-ray crystal- dant species (65% by mass) had an estimated molecular weight of lography. We have elucidated the crystal structure of the complex 435 kDa, consistent with a PRMT5:MEP50 complex containing bound to an AdoMet analog A9145C (34) and a substrate peptide four molecules each of PRMT5 and MEP50. The second most from histone H4. As we were preparing this article, the crystal abundant species (12.5% by mass), with an estimated molecular structure of PRMT5 from Caenorhabditis elegans, which shares weight of 720 kDa, potentially consists of higher-order complexes a sequence identity of 31% with human PRMT5, was published of hetero-octamer, or alternative PRMT5:MEP50 complexes. (35). In contrast to the dimeric C. elegans PRMT5 structure, our There was negligible material smaller than 435 kDa, demonstrat- work reveals that human PRMT5 binds MEP50 to form a het- ing the absence of free PRMT5 or MEP50 under these conditions. ero-octameric complex of molecular weight ∼450 kDa. This Samples were also routinely analyzed by LC-MS to identify post- (PRMT5)4(MEP50)4 module is likely to be the core structural translational modifications and a fraction of MEP50 was found to unit that interacts with partner proteins to form the plethora of be phosphorylated on Thr5 (16) (Fig. S3). multisubunit complexes with discrete specificities and functions. PRMT5:MEP50 Complex Is More Active than Dimeric PRMT5. Samples Results of PRMT5 and PRMT5:MEP50 were analyzed for methyltrans- Production and Characterization of the PRMT5:MEP50 Complex. Ini- ferase activity using a peptide derived from histone H4 (residues tial attempts to isolate and crystallize the C-terminal catalytic 1–21) as a substrate. Consistent with its description as a type II domain of human PRMT5 were unsuccessful. Although it was arginine methyltransferase, both PRMT5 and PRMT5:MEP50 possible to express and purify soluble constructs from Escherichia were able to generate di-methylated H4 peptide product (Fig. 1 E coli and insect cells, the protein was biochemically inactive and and F). However, production of the di-methylated species de- did not bind AdoMet or AdoHcy as determined by surface plas- pended critically on generation of the monomethylated peptide. mon resonance analysis. Expression of full-length PRMT5 in The concentration of monomethylated peptide produced by insect cells yielded a soluble and biochemically active protein PRMT5:MEP50 reached ∼70 nM between 3 and 5 h (Fig. 1F),

Fig. 1. PRMT5 forms a tight complex with MEP50 when coexpressed in insect cells, and the two proteins copurify on an anti-FLAG affinity column followed by size-exclusion chromatography. (A) SDS-PAGE gel of the purified PRMT5:MEP50 complex. (B) Analytical gel filtration chromatogram of the PRMT5 (in black) along with molecular weight (MW) reference standard BSA (in red, MW 66.7 kDa). PRMT5 elutes with a similar retention time as the BSA dimer (133.4 kDa, 15.6 min), suggesting that PRMT5 is a dimer in solution. Intrinsic tryptophan fluorescence was used to detect PRMT5, because PRMT5 had a tendency to

aggregate on concentration, resulting in low absorbance (A280). (C) Analytical gel filtration chromatogram of the PRMT5:MEP50 complex (in black) along with MW reference standards thyroglobulin (in red, MW 669 kDa) and β-amylase (in blue, MW 200 kDa). (D) Continuous distributions c(M), for two PRMT5:MEP50 protein samples. Samples were sedimented at 24,000 rpm, 20 °C, and data were obtained using absorbance at 280 nm. The PRMT5:MEP50 complex has an estimated mass of 435 kDa. (E) Nonprocessive methylation of substrate by PRMT5 and (F) PRMT5:MEP50 proteins. Ten nanomolars of isolated PRMT5 protein

and PRMT5:MEP50 protein complex were incubated with 10 μM AdoMet and 100 nM of H4 peptide for indicated times. Unmethylated (Me0), mono- BIOCHEMISTRY methylated (Me1), and dimethylated (Me2) H4 peptides were quantified by mass spectrometry.

Antonysamy et al. PNAS | October 30, 2012 | vol. 109 | no. 44 | 17961 Downloaded by guest on September 30, 2021 significantly exceeding the enzyme concentration of 10 nM, and generated by the crystallographic D2 symmetry. Almost the en- production of di-methylated peptide could not be observed until tire PRMT5 polypeptide (residues 13–637) was ordered in the the concentration of the monomethylated peptide exceeded that crystal structure, and residues 21–329 of MEP50 were ordered, of the unmethylated substrate. This result demonstrates that with the exception of loops 208–211 and 245–246. PRMT5 has a nonprocessive enzymatic mechanism for peptide PRMT5 adopts a two-domain structure, with the N-terminal substrates, in contrast to the partially processive mechanism de- domain (residues 13–292) adopting a TIM barrel structure, which scribed for PRMT1 (36). Lack of processivity for the PRMT5 or makes extensive charged and hydrophobic interactions with the PRMT5:MEP50 complex was observed at all tested concen- C-terminal catalytic domains of adjacent monomers. The PRMT5 trations of the peptide (0.01–10 μM) and AdoMet (0.1–10 μM). molecules form a tetramer at the center of the complex, with MEP50 Although there was some variability in the specific activity molecules decorating the outer surface and interacting solely with between different batches of enzyme preparations, the PRMT5: the TIM barrel domains of PRMT5 to form the PRMT54:MEP504 MEP50 complex consistently had a higher level of methyltrans- hetero-octamer. The dimerization domain in human PRMT5 is ferase activity compared with PRMT5 alone under similar ex- much smaller than in C. elegans (Fig. S6), and it adopts a very perimental conditions (Figs. 1 E and F). This apparent difference different fold. Residues 488–494 in this region form a short loop, in enzyme activity was attributed to a large difference in affinities with Arg488 and Asp491 from one monomer making salt bridges for the peptide and AdoMet between the two forms of the en- with Asp491 and Arg488 of an adjacent monomer (Fig. 2B). zyme. The Km values of PRMT5 for the peptide and AdoMet These loops from one pair of monomers pack against the equiv- were approximately 50- and 10-times higher, respectively, than alent loops of an adjacent pair of monomers, which are similarly those of the PRMT5:MEP50 complex, indicating a positive al- linked by Arg-Asp salt bridges. Apart from this direct interaction losteric effect of MEP50 on the cofactor and substrate binding between the catalytic domains of the four PRMT5 molecules in affinity of PRMT5 (Table 1). Binding of the cofactor or the sub- the hetero-octamer, the bulk of the interactions between the strate peptide does not appear to affect binding of the other be- monomers are mediated by extensive contacts between the N- cause Km values of the AdoMet or the peptide were largely terminal TIM barrels and C-terminal catalytic domains of adja- unaffected by variation in the concentration of the other (Fig. S4). cent monomers (Fig. 2C and Fig. S7). The TIM barrel fold −1 The kcat value of the PRMT5:MEP50 complex was 20–25 h at adopted by the N-terminal region was not predicted based on room temperature and a higher value was estimated for PRMT5, analyses of the primary sequence. although accurate determination of this value for PRMT5 was The interaction surface between the TIM barrel domain and fi the catalytic domain within the same monomer (buried surface dif cult because of low enzyme activity and high Km values. 2 Screening of AdoMet analogs using a scintillation proximity assay area of 821 Å ) is comparable to the interactions between the identified A9145C (34) as a potent inhibitor of PRMT5 with an monomers of the tetramer (buried surface areas of 713 and 608 Å2). The quaternary structure of one of the two dimers that make IC50 of 35 nM (Fig. S5). up the human PRMT5 tetramer is conserved with the C. elegans Overall Structure of the PRMT5:MEP50 Complex. PRMT5:MEP50 structure (Fig. S7), although the putative dimerization domains fi was cocrystallized in complex with the AdoMet analog A9145C exhibit signi cant variability in length, conformation, and inter- and an N-terminally acetylated substrate peptide from histone actions between the human and C. elegans structures. Analysis of H4 (residues 1–21); X-ray diffraction data were collected at the this dimer interface in the human PRMT5:MEP50 structure re- ... Advanced Photon Source, Argonne, IL. Attempts at Molecular veals that several of the hydrogen bonds (Asp70:OD1 Thr400: ...... Replacement using homology models prepared from type I meth- OG1, Asp70:OD2 Arg368:NH2, Tyr116:O Ser321:N) at this yltransferases and WD40 domains were unsuccessful, leading us interface are preserved in the C. elegans structure, and the iden- to determine the structure by selenomethionine (Se-Met) MAD tities of these residues are conserved across species. In contrast, phasing (Fig. S3). The structure revealed that the PRMT5 com- the residues forming hydrogen bonds across the second dimer in- ∼ terface in the human PRMT5 tetramer (Arg101:NH1...Asp531: plex with MEP50 forms a 453 kDa hetero-octamer in agree- ...... ment with the sedimentation and chromatography data (Fig. 2A). OD1, Arg101:NH2 Asn533:O, Asn133:OD1 Trp603:NE1) are The crystallographic asymmetric unit contains one molecule not conserved in the C. elegans sequence. The extensive and close each of PRMT5 and MEP50, with the hetero-octamer being interactions between the TIM barrel and methyltransferase do- mains across monomers in the human PRMT5 structure suggest that these interactions are the primary driver for oligomerization.

Table 1. Kinetic parameters (Km and kcat values) for PRMT5 and fi Structure and Interactions of MEP50. MEP50 adopts the WD40 PRMT5:MEP50 for xed concentrations of AdoMet or substrate β peptide -propeller structure with seven blades (Fig. 2C), six of which are composed of four β-strands, with the last β-sheet containing only Parameters for AdoMet or H4 PRMT5:MEP50 PRMT5 three ordered strands; the first 12 and last 13 residues are dis- ordered and not visible in the electron density. The structure Kinetic parameters for H4 peptide at different [AdoMet] “ ” μ σ μ — therefore lacks the velcro closure seen in the majority of WD40 1 M Km ( M), M 0.049 (0.010) β μ σ μ domains, where the three strands from the last blade form a - 10 M Km ( M), M 0.022 (0.004) 4.2 (2.2) sheet with the strand that precedes the first blade (22, 37). The μ σ > 10 M kcat ( M), 1/h 20 (1) 30 (20) sequence motifs that characterize the WD40 domain are scarce in Kinetic parameters for AdoMet at different [H4 peptide] MEP50, with only two of the seven blades containing the WD μ σ μ — 0.1 M Km ( M), M 1.1 (0.03) motif at the end of the third strand. One of the other blades has μ σ μ — 0.2 M Km ( M), M 1.1 (0.3) a WE motif, and the remaining blades have a Val, Leu, Phe, or μ σ μ 1 M Km ( M), M 1.0 (0.1) 10.2 (0.3) His residue instead of Trp at this position. The Gly-His dipeptide μ σ > — 1 M kcat ( M), 1/h 25 (2) 18 ( ) that occurs at the end of the outer strand is absent in all seven blades of MEP50, although five of the blades contain a Kinetic parameters (Km and kcat values) for PRMT5:MEP50 and PRMT5 for fixed concentrations of AdoMet or substrate peptide as indicated on the far residue preceded by a nonglycine residue at this position. The aspartic acid residue that usually occurs in the loop between the left columns. Incubation was carried out at room temperature with 1 nM of fi enzyme for 15–60 min (PRMT5:MEP50) or 2 nM of enzyme for 2–3 h (PRMT5), second and third strand is absent in the rst blade but present in the six other blades. The top of the domain, defined as the surface during which each enzyme gave linear product formation with time. kcat value for PRMT5 is estimated as a lower limit because the concentrations formed by the loops that link the outermost strand of one blade of AdoMet (10 μM) or the peptide (1 μM) were insufficient to saturate the to the innermost strand of the next blade (20), is oriented toward

enzyme, considering the measured values of the Km. SEM is given in paren- and interacts closely with PRMT5. This surface is the primary theses (σM). —, data not available. interaction and recognition surface for the majority of WD40

17962 | www.pnas.org/cgi/doi/10.1073/pnas.1209814109 Antonysamy et al. Downloaded by guest on September 30, 2021 between the side-chains of Phe133 and Phe263. In the human PRMT5:MEP50 complex, the top surface of MEP50 is used for recognition of the TIM barrel of PRMT5. In addition, MEP50 lacks the arginine recognition motif of WDR5, with Ser129/ Gly260 replacing the “phenylalanine clamp” (Phe133/Phe263) that binds the guanidyl side-chain in WDR5. The loop following the β2 strand of the TIM barrel domain of PRMT5 traces the top surface of MEP50 and the Arg49 side-chain interacts with Asp99 of MEP50. Because the top surface and one side of MEP50 are fully engaged in binding to PRMT5, MEP50 likely uses the bot- tom and the rest of the circumference of the barrel to interact with and recruit partners and substrates.

Catalytic Domain and Salient Features of the PRMT5 Cofactor Binding Site. The catalytic domain adopts the canonical arginine methyl- transferase tertiary structure similar to the type I PRMTs, with an AdoMet binding domain containing the nucleotide binding , followed by a β-sandwich domain involved in substrate binding. PRMT5 lacks the YFxxY motif seen in type I PRMTs (40), and this region, which is involved in binding both cofactor and substrate, adopts a very different conformation below the AdoMet binding site in PRMT5. In addition, the conserved THW motif of the type I PRMTs is FSW in human PRMT5. Al- though the tryptophan residue adopts a similar conformation as the type I PRMTs, the threonine in type I PRMTs packs against the tryptophan, whereas the equivalent Phe577 of the FSW motif in PRMT5 is oriented toward the solvent, pi-stacking against Phe300 in the linker between the catalytic and TIM barrel domains. There is clear density for the bound AdoMet analog A9145C, which binds in a similar orientation to the closely related methyltransferase inhibitor sinefungin and AdoHcy observed in type I PRMT structures (Fig. 3A). However, there are significant differences in the composition and nature of the binding inter- actions of the cofactor analog. In all known type I PRMT structures the adenine ring of AdoHcy makes a pair of hydrogen bonds to the protein, accepting a proton from the main-chain Fig. 2. (A) Structure of the human PRMT5:MEP50 hetero-octameric complex. amino group, and donating a proton to the acidic residue that PRMT5 monomers-1, -2, -3, and -4 are colored green, blue, wheat, and yellow, follows. In PRMT5, although the adenine ring accepts a proton respectively. MEP50 molecules in red decorate the outer surface of the mol- from the main-chain amino of Met420, the succeeding residue is ecule, interacting solely with the N-terminal TIM barrel domains of PRMT5. (B) Arg421 (a basic residue), which instead stabilizes Asp419 (the The dimerization domains of monomers-1 and -2 interact directly with each other at the center of the PRMT5:MEP50 hetero-octamer through salt-bridges. residue preceding Met420), and it is Asp419 that accepts a pro- (C) The dimer formed by PRMT5 monomers-1 and -2 in the human PRMT5 : ton from the adenine. The hydroxyl groups of the ribose moiety 4 are stabilized by hydrogen bonds to the side-chains of Glu392 MEP504 complex. This interaction is conserved with the C. elegans PRMT5 structure. Highlighted in orange is the linker between the N-terminal TIM and Tyr324. The side-chain of Glu392 interacts with both barrel and C-terminal methyltransferase catalytic domain, with the Tyrosine hydroxyls, and is conserved in the other known PRMT struc- residues (Y297, Y304, and Y307) that are phosphorylated by Jak2 (V617H) tures, although it is an aspartate in PRMT3, instead of a gluta- shown as sticks. The bound AdoMet analog (A9145C) is shown in black, and mate. The Tyr324 residue in PRMT5 that hydrogen bonds to the the H4 peptide in magenta with substrate residue Arg3 in sticks. ribose is variable among the PRMTs; it is a histidine in PRMT1 and PRMT3, and a glutamine in PRMT4. The carboxylate at the end of the ligand makes a split hydrogen bond to the domains, although they can also interact with partner proteins side-chain of Tyr334 in PRMT5. This residue is not conserved in through the circumference and the bottom of the barrel. The the other PRMTs, and is a threonine in PRMT1, PRMT3, and segment between β2andβ3 of the PRMT5 TIM barrel wraps PRMT4. The preceding residue, Lys333, which hydrogen bonds around the side of the MEP50 domain, interacting closely with to the carboxylate of the cofactor analog, is an arginine in all of the outer surface of the third and fourth blade. The MEP50 the other PRMTs. Despite the differences in protein–ligand molecule interacts solely with the N-terminal TIM barrel domain interactions between PRMT5 and type I PRMTs, the confor- of PRMT5, making extensive charged and Van der Waals inter- mation of the AdoMet analog is similar to what has been actions, burying an accessible surface area of 2,027 Å2. With such observed in the of type I arginine methyltransferases. a large interaction surface, and a calculated free-energy gain of 13.6 kcal/mol (38), MEP50 is very tightly bound to PRMT5. Peptide Complex and the Substrate Binding Pocket. Our structure A comparison of the MEP50 structure with structures of WDR5 has clear density for the first eight residues (SGRGKGGK) of the (another WD40 protein and a component of many histone bound peptide (Fig. 3B and Fig. S8). The histone H4-derived methyltransferase complexes) reveals distinct differences in the substrate peptide binds in a groove on the surface of the β-barrel functional residues involved in peptide recognition in WDR5, domain, inserting the arginine side-chain (Arg3 of the peptide) despite a conserved overall structure. WDR5 binds both unme- through a narrow tunnel formed by Leu312, Phe327, and Trp579 thylated and lysine mono-, di-, and trimethylated histone H3K4 to access the active site. The peptide residues that flank the argi- peptides, and has recently been shown to differentiate between nine form a sharp β- at the neck of the tunnel stabilized by symmetric and asymmetric arginine di-methylation by specifi- a hydrogen bond between the main-chain carbonyl of the Ser1 and cally recognizing H3R2me2s (39). Peptides bind WDR5 on the the amino group of Gly4; the Ser1 carbonyl makes an additional

top surface of the β-propeller domain, inserting the side-chain of hydrogen bond to the Gln309 side-chain of PRMT5. The bulk of BIOCHEMISTRY Arg2 into the central channel where the guanidine group packs the interactions between the substrate peptide and PRMT5 are

Antonysamy et al. PNAS | October 30, 2012 | vol. 109 | no. 44 | 17963 Downloaded by guest on September 30, 2021 mediated by protein backbone interactions. The main-chain car- TIM barrel domains, composed of eight β/α-segments that bonyls of peptide residues Gly2 and Arg3 make hydrogen bonds to fold to form an internal β-barrel surrounded by helices, are one the main-chain amino groups of Phe580 and Leu312, respectively, of the most well-studied structural families (43). The low se- and the amino and carbonyl groups of peptide residue Lys5 makes quence similarity among TIM barrel proteins coupled with the hydrogen bonds to the carbonyl and amino groups of PRMT5 functional diversity of this structural class has led to extensive Ser310. The conformational and spatial restraints of this binding research on the potential role of convergent and divergent mode provide a rationale for the preference for glycine residues evolution in generating the ability to catalyze various reactions flanking the substrate arginine. using a conserved structural scaffold. TIM barrel proteins are This structure is unique in revealing the critical role in catal- typically enzymes involved in molecular or energy metabolism, ysis played by the highly conserved active site glutamate residues and additional studies are warranted to elucidate whether the Glu435 and Glu444 of the so-called double-E loop (41). Each of PRMT5 TIM barrel domain has any catalytic activity. In TIM the two glutamate residues form a pair of salt bridges with the barrel enzymes, the loops linking the C-terminal ends of the guanidine side-chain of the substrate arginine (H4R3) with the β-strands to the helical segments contribute the residues that are ω-NG nitrogen atom poised for methyl transfer (Fig. 3B). These involved in catalysis, metal-ligation, and phosphate-binding. The glutamate residues are likely involved in de-protonating and equivalent site of the PRMT5 TIM barrel is oriented away from activating the ω-NG nitrogen atom. The phenylalanine residue the methyltransferase catalytic domain and interacts intimately (Phe327) that has been shown to play a role in specifying sym- with the MEP50 molecule. metric di-methylation of PRMT5 (35) pi-stacks against the side Comparison of the human PRMT5 monomer to the dimeric C. of the guanidyl group orienting the substrate arginine for methyl- elegans structure suggests that the C. elegans monomer has pos- transfer. Significantly, two of the three tyrosine residues that are sibly been misinterpreted as being composed of the equivalent of phosphorylated by Jak2 are involved in substrate binding. Tyr304 the TIM barrel domain from one polypeptide chain and the cat- packs against the acetylated N-terminus of the substrate peptide, alytic domain from another chain (Fig. S7). This misinterpretation and the hydroxyl group of Tyr307 hydrogen bonds to main-chain may have been precipitated by the fact that the region between carbonyl and amino groups of substrate residues Gly6 and Lys8, the TIM barrel and the catalytic domain is disordered in the C. respectively. elegans structure. In the human structure there is no ambiguity, because the equivalent segment between the TIM barrel and Discussion catalytic domain (residues 293–320) is fully ordered. The preser- The hetero-octameric structure of the PRMT54:MEP504 meth- vation of residues and interactions across this dimer interface yltransferase complex presented here likely represents the core between the human and C. elegans structures suggests that this unit that associates with different binding partners, in a context- dimer is a conserved quaternary structure of PRMT5 across dependent manner, to form larger multicomponent complexes species, and likely to be the structure adopted by human PRMT5 that specifically methylate a diverse set of substrates in both the in the absence of MEP50. In contrast, the residues at the second cytoplasm and the nucleus. The interactions between the PRMT5 dimer-interaction surface in the human PRMT5:MEP50 structure monomers in the complex are very different from the conserved are not conserved across species, and the formation of the tetra- dimer interaction observed in the structures of type I PRMTs. mer is brought about by the binding of MEP50. A sequence search In the type I PRMT dimers, the “dimerization arm” from the did not identify a MEP50 ortholog in C. elegans, suggesting that β-barrel domain of each monomer interacts with the AdoMet the nematode PRMT5 may have a different mechanism of binding domain of the second monomer in a head-to-tail orien- recruiting binding partners and substrates. Contrary to expecta- tation, and this interaction has been shown to be necessary for tion, the structure of the human PRMT5:MEP50 complex re- cofactor binding and activity (41, 42). In contrast, the corre- vealed that MEP50 does not bind at any of the PRMT5 sponding regions in human PRMT5 interact directly with each oligomerization interfaces but binds on the outer surface of the other at the center of the oligomer. Apart from this interaction tetrameric PRMT5 core, interacting solely with the TIM barrel between catalytic domain residues, the bulk of the interactions domain. Therefore, the binding of MEP50 must introduce a subtle between the PRMT5 monomers are mediated by the N-terminal conformational change that favors the formation of PRMT5 tet- TIM barrel domains. The PRMT5 TIM barrel domain therefore ramers, and also lowers the Km for AdoMet and substrate peptide. plays dual structural roles, promoting oligomerization by inter- The observation from the structure that two of the tyrosine acting with the catalytic domain to form the PRMT5 tetramer, residues (Tyr304 and Tyr307) that are phosphorylated by onco- and binding specifically and tightly to MEP50 molecules. genic V617F Jak2 (17) to down-regulate the methyltransferase activity are involved in substrate binding, provides a rationale for the observed loss of activity upon phosphorylation, because phosphorylation would disrupt the substrate pocket. The occur- rence of these residues on a short helix and loop that is part of the linker between the TIM barrel and catalytic domains suggests a possible regulatory role for this segment. The observed lack of processivity in successive mono- and di- methylation by PRMT5 is due to the requirement of the mono- methylated substrate having to dissociate from the enzyme and bind again with the ω-N′G nitrogen atom poised for methylation. The partial processivity of type I PRMTs, such as PRMT1, is likely enabled by its retention of the monomethylated product in the active site, while allowing the methyl group on the arginine to rotate about the Cζ-ωNG bond and re-present the methylated Fig. 3. (A) Interactions of the AdoMet analog (A9145C in black) with the nitrogen atom for the second methylation in concert with the PRMT5 cofactor binding site in green. The histone H4 derived substrate peptide is shown in magenta, and the omit map of A9145C is contoured at replacement of AdoHcy by AdoMet in the cofactor binding site. 3σ in blue. (B) A close-up view of the PRMT5 active site. The methyl- In type II PRMTs, retention of the peptide in the active site for processive di-methylation would require larger rotations of side- transferase domain is depicted in green, the linker between the catalytic and ω TIM barrel domains in orange, the AdoMet analog (A9145C) in black, and chain torsion angles to interchange the positions of the nitrogen the bound substrate peptide in magenta. Highlighted are the interactions atoms, which would be precluded by the spatial constraints of the between the Arg3 guanidyl side-chain of the substrate peptide, the meth- active site. In the cellular environment a monomethylated histone yltransferase catalytic glutamate residues (Glu435 and Glu444) and A9145C. tail may remain in the vicinity for successive methylation, because The peptide omit map is contoured at 3σ in blue. the octameric histone complex with multiple substrates (H2A,

17964 | www.pnas.org/cgi/doi/10.1073/pnas.1209814109 Antonysamy et al. Downloaded by guest on September 30, 2021 H3, and H4) could simultaneously bind multiple active sites in to the methyltransferase, and reflective of the broader role of the PRMT54:MEP504 complex. The documented ability of the WD40 proteins in epigenetic pathways. PRMT5 F327M mutant to perform both symmetric and asym- The view revealed by this structure clearly delineates the in- metric di-methylation (35) is likely enabled by its capacity to teractions between the methyltransferase and the substrate pep- accommodate the methyl group of monomethylated substrate tide, and clarifies the roles of the active site residues in catalysis. and still present the appropriate nitrogen for both symmetric or Thus, in addition to advancing our understanding of this complex asymmetric di-methylation. The conformational flexibility of the target, this work also provides a valuable resource for structure methionine side-chain potentially allows sufficient compliance based drug design efforts on this emerging target class. in binding monomethylated arginine, so that it can present the appropriate terminal ω-NG nitrogen atom toward AdoMet. Methods The extensive and tight interaction between MEP50 and Full-length PRMT5 (NP_006100, residues 1–637) and MEP50 (NP_077007, resi- PRMT5 is consistent with the role of MEP50 as the primary dues 2–342) were coexpressed in insect cells and PRMT5:MEP50 complex binding partner of PRMT5, and is also evidenced by the presence purified by affinity followed by size-exclusion chromatography. Crystals were of MEP50 in several of the characterized PRMT5 complexes grown at 8 °C in sitting drops with protein incubated with peptide and A9145C. (Fig. S1). The increased stability and methyltransferase activity Detailed methods are provided in SI Methods and data and refinement sta- of the hetero-octameric PRMT5:MEP50 complex compared tistics are included in Table S1. with isolated PRMT5 lends further credence to the view that the PRMT5:MEP50 complex is the functional biological module. ACKNOWLEDGMENTS. We thank Guemalli Cardona for work on the enzyme inhibition measurements. Use of the Advanced Photon Source at The well-documented interaction of MEP50 with many of the Argonne National Laboratory was supported by the US Department of binding partners and substrates of PRMT5 is suggestive of an Energy, Office of Science, Office of Basic Energy Sciences, under Contract additional role for MEP50 in recruiting substrates and partners DE-AC02-06CH11357.

1. Bedford MT, Clarke SG (2009) Protein arginine methylation in mammals: Who, what, 23. Migliori V, Mapelli M, Guccione E (2012) On WD40 proteins: Propelling our knowl- and why. Mol Cell 33(1):1–13. edge of transcriptional control? Epigenetics 7(8):815–822. 2. Herrmann F, Pably P, Eckerich C, Bedford MT, Fackelmayer FO (2009) Human protein 24. Gu Z, Zhou L, Gao S, Wang Z (2011) Nuclear transport signals control cellular locali- arginine methyltransferases in vivo—Distinct properties of eight canonical members zation and function of androgen receptor cofactor p44/WDR77. PLoS ONE 6(7): of the PRMT family. J Cell Sci 122(Pt 5):667–677. e22395. 3. Martin C, Zhang Y (2005) The diverse functions of histone lysine methylation. Nat Rev 25. Ligr M, et al. (2011) Expression and function of androgen receptor coactivator p44/ Mol Cell Biol 6(11):838–849. Mep50/WDR77 in ovarian cancer. PLoS ONE 6(10):e26250. 4. Yost JM, Korboukh I, Liu F, Gao C, Jin J (2011) Targets in epigenetics: Inhibiting the 26. Furuno K, Masatsugu T, Sonoda M, Sasazuki T, Yamamoto K (2006) Association of methyl writers of the histone code. Curr Chem Genomics 5(Suppl 1):72–84. Polycomb group SUZ12 with WD-repeat protein MEP50 that binds to histone H2A 5. Wolf SS (2009) The protein arginine methyltransferase family: An update about selectively in vitro. Biochem Biophys Res Commun 345(3):1051–1058. function, new perspectives and the physiological role in humans. Cell Mol Life Sci 66 27. Licciardo P, et al. (2003) The FCP1 phosphatase interacts with RNA polymerase II and (13):2109–2121. with MEP50 a component of the methylosome complex involved in the assembly of 6. Zurita-Lopez CI, Sandberg T, Kelly R, Clarke SG (2012) Human protein arginine meth- snRNP. Nucleic Acids Res 31(3):999–1005. yltransferase 7 (PRMT7) is a type III enzyme forming ω-NG-monomethylated arginine 28. Amente S, et al. (2005) Identification of proteins interacting with the RNAPII FCP1 residues. J Biol Chem 287(11):7859–7870. phosphatase: FCP1 forms a complex with arginine methyltransferase PRMT5 and it is 7. Zobel-Thropp P, Gary JD, Clarke S (1998) delta-N-methylarginine is a novel post- a substrate for PRMT5-mediated methylation. FEBS Lett 579(3):683–689. translational modification of arginine residues in yeast proteins. J Biol Chem 273(45): 29. Pesiridis GS, Diamond E, Van Duyne GD (2009) Role of pICLn in methylation of Sm 29283–29286. proteins by PRMT5. J Biol Chem 284(32):21347–21359. 8. Karkhanis V, Hu YJ, Baiocchi RA, Imbalzano AN, Sif S (2011) Versatility of PRMT5-induced 30. Guderian G, et al. (2011) RioK1, a new interactor of protein arginine methyltrans- methylation in growth control and development. Trends Biochem Sci 36(12):633–641. ferase 5 (PRMT5), competes with pICln for binding and modulates PRMT5 complex 9. Friesen WJ, et al. (2001) The methylosome, a 20S complex containing JBP1 and pICln, composition and substrate specificity. J Biol Chem 286(3):1976–1986. produces dimethylarginine-modified Sm proteins. Mol Cell Biol 21(24):8289–8300. 31. Wang L, Pal S, Sif S (2008) Protein arginine methyltransferase 5 suppresses the tran- 10. Andreu-Pérez P, et al. (2011) Protein arginine methyltransferase 5 regulates ERK1/2 scription of the RB family of tumor suppressors in leukemia and lymphoma cells. Mol signal transduction amplitude and cell fate through CRAF. Sci Signal 4(190):ra58. Cell Biol 28(20):6262–6277. 11. Ren J, et al. (2010) Methylation of ribosomal protein S10 by protein-arginine meth- 32. Copeland RA, Solomon ME, Richon VM (2009) Protein methyltransferases as a target yltransferase 5 regulates ribosome biogenesis. J Biol Chem 285(17):12695–12705. class for drug discovery. Nat Rev Drug Discov 8(9):724–732. 12. Tee WW, et al. (2010) Prmt5 is essential for early mouse development and acts in the 33. Esteller M (2008) Epigenetics in cancer. N Engl J Med 358(11):1148–1159. cytoplasm to maintain ES cell pluripotency. Genes Dev 24(24):2772–2777. 34. Pugh CS, Borchardt RT, Stone HO (1978) Sinefungin, a potent inhibitor of virion 13. Ancelin K, et al. (2006) Blimp1 associates with Prmt5 and directs histone arginine mRNA(guanine-7-)-methyltransferase, mRNA(nucleoside-2′-)-methyltransferase, and methylation in mouse germ cells. Nat Cell Biol 8(6):623–630. viral multiplication. J Biol Chem 253(12):4075–4077. 14. Patel SR, Bhumbra SS, Paknikar RS, Dressler GR (2012) Epigenetic mechanisms of 35. Sun L, et al. (2011) Structural insights into protein arginine symmetric dimethylation Groucho/Grg/TLE mediated transcriptional repression. Mol Cell 45(2):185–195. by PRMT5. Proc Natl Acad Sci USA 108(51):20538–20543. 15. Powers MA, Fay MM, Factor RE, Welm AL, Ullman KS (2011) Protein arginine meth- 36. Obianyo O, Osborne TC, Thompson PR (2008) Kinetic mechanism of protein arginine yltransferase 5 accelerates tumor growth by arginine methylation of the tumor methyltransferase 1. Biochemistry 47(39):10420–10427. suppressor programmed cell death 4. Cancer Res 71(16):5579–5587. 37. Wall MA, et al. (1995) The structure of the G protein heterotrimer Gi alpha 1 beta 1 16. Aggarwal P, et al. (2010) Nuclear cyclin D1/CDK4 kinase regulates CUL4 expression gamma 2. Cell 83(6):1047–1058. and triggers neoplastic growth via activation of the PRMT5 methyltransferase. Cancer 38. Krissinel E, Henrick K (2007) Inference of macromolecular assemblies from crystalline Cell 18(4):329–340. state. J Mol Biol 372(3):774–797. 17. Liu F, et al. (2011) JAK2V617F-mediated phosphorylation of PRMT5 downregulates 39. Migliori V, et al. (2012) Symmetric dimethylation of H3R2 is a newly identified its methyltransferase activity and promotes myeloproliferation. 19(2): histone mark that supports euchromatin maintenance. Nat Struct Mol Biol 19(2): 283–294. 136–144. 18. Friesen WJ, et al. (2002) A novel WD repeat protein component of the methylosome 40. Troffer-Charlier N, Cura V, Hassenboehler P, Moras D, Cavarelli J (2007) Functional binds Sm proteins. J Biol Chem 277(10):8243–8247. insights from structures of coactivator-associated arginine methyltransferase 1 do- 19. Hosohata K, et al. (2003) Purification and identification of a novel complex which is mains. EMBO J 26(20):4391–4401. involved in androgen receptor-dependent transcription. Mol Cell Biol 23(19): 41. Zhang X, Cheng X (2003) Structure of the predominant protein arginine methyl- 7019–7029. transferase PRMT1 and analysis of its binding to substrate peptides. Structure 11(5): 20. Stirnimann CU, Petsalaki E, Russell RB, Müller CW (2010) WD40 proteins propel cel- 509–520. lular networks. Trends Biochem Sci 35(10):565–574. 42. Weiss VH, et al. (2000) The structure and oligomerization of the yeast arginine 21. Avdic V, et al. (2011) Structural and biochemical insights into MLL1 core complex methyltransferase, Hmt1. Nat Struct Biol 7(12):1165–1171. assembly. Structure 19(1):101–108. 43. Nagano N, Orengo CA, Thornton JM (2002) One fold with many functions: The 22. Xu C, Min J (2011) Structure and function of WD40 domain proteins. Protein Cell 2(3): evolutionary relationships between TIM barrel families based on their sequences, 202–214. structures and functions. J Mol Biol 321(5):741–765. BIOCHEMISTRY

Antonysamy et al. PNAS | October 30, 2012 | vol. 109 | no. 44 | 17965 Downloaded by guest on September 30, 2021