Research ResourceBiochemistry

Proteomics profiling of arginine methylation defines PRMT5 substrate specificity

See allHide authors and affiliations

Science Signaling  02 Apr 2019:
Vol. 12, Issue 575, eaat8388
DOI: 10.1126/scisignal.aat8388

Profiling PRMT5

The enzyme PRMT5 methylates arginine residues in target proteins to modify their function. PRMT5 functions in various cellular processes and is implicated in cancer. Musiani et al. used advanced and modified proteomics to explore the substrate specificity of PRMT5 and identified a preferred target site (a glycine-sandwiched arginine), as well as previously unknown substrates of the enzyme. This resource therefore offers clues to the function of PRMT5 in normal physiology and potential ways to target it in disease.


Protein arginine methyltransferases (PRMTs) catalyze arginine methylation on both chromatin-bound and cytoplasmic proteins. Accumulating evidence supports the involvement of PRMT5, the major type II PRMT, in cell survival and differentiation pathways that are important during development and in tumorigenesis. PRMT5 is an attractive drug target in various cancers, and inhibitors are currently in oncological clinical trials. Nonetheless, given the complex biology of PRMT5 and its multiple nonhistone substrates, it is paramount to fully characterize these dynamic changes in methylation and to link them to the observed anticancer effects to fully understand the functions of PRMT5 and the consequences of its inhibition. Here, we used a newly established pipeline coupling stable isotope labeling with amino acids in cell culture (SILAC) with immunoenriched methyl peptides to globally profile arginine monomethylation and symmetric dimethylation after PRMT5 inhibition by a selective inhibitor. We adopted heavy methyl SILAC as an orthogonal validation method to reduce the false discovery rate. Through in vitro methylation assays, we validated a set of PRMT5 targets identified by mass spectrometry and provided previously unknown mechanistic insights into the preference of the enzyme to methylate arginine sandwiched between two neighboring glycines (a Gly-Arg-Gly, or “GRG,” sequence). Our analysis led to the identification of previously unknown PRMT5 substrates, thus both providing insight into the global effects of PRMT5 and its inhibition in live cells, beyond chromatin, and refining our knowledge of its substrate specificity.


There is a growing interest in protein arginine (R) methylation in cell biology, leading to a better appreciation of its role across a broad range of cellular processes. In eukaryotes, this widespread posttranslational modification is catalyzed by protein arginine methyltransferases (PRMTs), which are involved in transcriptional and posttranscriptional regulation of gene expression, mRNA processing, protein translation, and intracellular signaling, both during development and in tumorigenesis (1). The PRMT family is conserved from yeast to humans. PRMTs are classified as types I, II, III, and IV based on the differential addition of methyl marks on arginine residues. All types can catalyze the formation of ω-NG-monomethylarginine (ω-NG-MMA, hereafter “MMA”) on guanidine nitrogen atom of arginine residues. Type I enzymes [PRMT1, PRMT2, PRMT3, PRMT4 (also known as CARM1), PRMT6, and PRMT8] can additionally catalyze asymmetric ω-NG, NG-dimethylarginine (ω-NG, NG-ADMA, hereafter “ADMA”), by adding a second methyl group to the same nitrogen atom. Type II enzymes (PRMT5 and PRMT9) can catalyze the formation of symmetric ω-NG, N′G-dimethylarginine (ω-NG, N′G-SDMA, hereafter “SDMA”), with a second methyl group added on the other terminal nitrogen atom. Type III PRMT activity is restricted to generating MMA on substrates. PRMT7 is currently the only type III methyltransferase identified so far (2, 3).

PRMT1 and PRMT5 are the predominant type I and II arginine methyltransferases, respectively. They are able to methylate histone and nonhistone substrates, thus regulating a wide range of biological processes. With regard to substrate specificity, they are known to methylate proteins containing arginine- and glycine-rich motifs (RGG/RG) (4). PRMT5 methylates histones H2AR3, H3R2, H3R8, and H4R3 and affects transcriptional regulation (46). PRMT5 also methylates several nonhistone proteins, such as the RNA binding Sm proteins (7), the 40S ribosomal protein S10 RPS10 (8), and the transcription factor E2F1 (9). The PRMT5-dependent arginine methylation on Sm proteins facilitates their interaction with the SMN complex and mediates efficient pre-mRNA splicing (7). Loss of PRMT5 in mouse neural stem/progenitor cells (NPCs) leads to the reduced methylation of Sm protein, selective retention of introns, and skipping of exons with weak donor sites (10). During malignant transformation, PRMT5 acts as an oncogene. Consistently, PRMT5 depletion or enzymatic inhibition leads to reduced cellular proliferation, whereas its overexpression leads to hyperproliferation (1113). Increased expression of PRMT5 is associated with multiple cancer types, including gastric, colorectal, and lung cancer, as well as lymphoma and leukemia (12, 14). These studies highlight the important functions of PRMT5, which are almost exclusively linked to its catalytic activity (10). Therefore, PRMT5 is emerging as an attractive target for cancer treatment, and intense research has been recently devoted to designing selective and potent small-molecule inhibitors targeting its catalytic activity (15, 16).

One outstanding question in the field is which are the relevant substrates that, in the absence of PRMT5-mediated methylation, cause the observed phenotypes. To fill this gap, and to better understand the consequences of PRMT5 inhibition on the cellular methyl proteome, we took advantage of a recently described potent and selective PRMT5 inhibitor (GSK591) (17), with the aim to systematically identify previously unknown PRMT5 targets.

Protein methylation is, however, a challenging posttranslational modification to characterize. Mass spectrometry (MS)–based analysis of methylation has been hindered—until recently—by two major issues: the first is the lack of efficient biochemical strategies for the enrichment of methyl peptides that are substoichiometric to the unmodified counterpart. Second, several amino acid substitutions are isobaric to methylation, hindering the confident identification of in vivo (live cell) methylation at single-site resolution (18).

The development of antibody-based strategies for the enrichment of R-methyl peptides has made it possible to annotate several hundreds of R-methylation sites in human cells (18). On the other hand, methods based on the isotope labeling of methyl groups, such as heavy methyl stable isotope labeling with amino acids in cell culture (hmSILAC) (19) and iso-methyl SILAC (20), have shown their efficacy as orthogonal strategies to increase the confident identification of methyl sites by MS, sensibly reducing the very high false-positive rate typically associated to label-free approaches through the efficient discrimination of enzyme-mediated modification from artifacts (21).

In the current study, we took advantage of an analytical platform that couples hmSILAC and standard SILAC with the biochemical separation and affinity enrichment of R-methyl peptides, followed by MS, to acquire a comprehensive profiling of the PRMT5-dependent arginine methyl proteome that includes an accurate measurement of both MMA and SDMA. ADMA was not assessed. This in-depth and high-quality analysis revealed a more prominent decrease of SDMA, mirrored by the up-regulation of MMA on the same sites, a molecular response that could not be detected in a previous extensive methyl proteome profiling in response to PRMT inhibition, where only monomethylation was assessed (18). It also led to the identification of novel PRMT5 targets, all involved in RNA posttranscriptional processing. In vitro methylation assays on a subset of identified PRMT5 targets led to the refinement of the consensus motif targeted by PRMT5, with the substrate arginine often surrounded by glycines. This is different from PRMT1 that is more permissive regarding the surrounding amino acids in the targeted motif.


Global profiling of PRMT5 substrates at single-site resolution by quantitative LC-MS/MS

To globally profile PRMT5 substrates in live cells, we used an established SILAC-based analytical platform, using pan-methyl type-specific antibodies to enrich peptides carrying MMA and SDMA from total extracts of HeLa cells, treated with either the PRMT5 inhibitor GSK591, or its inactive structural analog SGC20969 (Fig. 1A). Our experimental design capitalized on the proven advantage in methyl peptide identification offered by the separation of tryptic peptides through offline high-pH reversed-phase liquid chromatography (HpH RPLC) before methyl peptide immunoenrichment (fig. S1A). This separation strategy, followed by fraction concatenation, was previously shown to be orthogonal to the low-pH nano-RPLC that is directly coupled with the mass spectrometer and has emerged as the gold standard to increase sample heterogeneity and circumvent the bias toward abundant polypeptides (22). In the “forward” experimental setup, untreated HeLa cells were light labeled and GSK591-treated cells were heavy labeled (Fig. 1A). The efficiency of the PRMT5 inhibition was evaluated by Western blot analysis that confirmed the reduction of SDMA upon cell treatment with GSK591 (Fig. 1B).

Fig. 1 Global profiling of PRMT5 substrates by SILAC coupled with high-resolution MS.

(A) Schematic representation of the SILAC-based experimental strategy used for global detection of PRMT5 substrates. Enrichment of MMA- and SDMA-carrying peptides was followed by quantitative LC-MS/MS analysis. GSK591 was used for pharmacological inhibition of PRMT5, whereas the inactive compound SGC20969 was used as a negative control. (B) Western blot analysis of MMA, SDMA, and ADMA in HeLa cells, treated as indicated. Vinculin was used as a loading control (n = 2). DMSO, dimethyl sulfoxide. (C) Scatter plot of the log2-transformed protein-normalized methyl peptide SILAC ratios (PRMT5i/control), quantified in both the forward (FWD) and reverse (REV). Methyl peptides up- and down-regulated by GSK591 in both replicates were determined by applying a μ ± 3σ cutoff, calculated on the unmodified peptide distributions (see Materials and Methods), displayed on the respective axes.

The experiment was carried out in two biological replicates, in forward and reverse experimental setups, in which SILAC labels were swapped among the two conditions. Upon MS acquisition and processing of the data through the MaxQuant algorithm, we robustly quantified in both biological replicates 958 R-methyl peptides, with 741 R-methyl sites including 690 monomethylations and 138 dimethylations, with 92 arginines that were identified as both mono- and dimethylated (Andromeda score ≥ 25 and posttranslational modification (PTM) localization probability ≥ 0.75) (Fig. 1C and table S1, spreadsheets 1, 2, and 2b). Fewer DMA-containing peptides than MMA peptides were quantified, either as a consequence of differing antibody efficiency or the difference in abundance of these two types of methylation in live cells.

In addition to its well-documented role as a splicing modulator (10, 13), PRMT5 functions as both a transcriptional activator and a repressor (5, 23). Consequently, we reasoned that pharmacological inhibition of this enzyme might also affect global protein expression. We thus analyzed the proteomes from untreated and treated cells (that were used as input for methyl peptide enrichment) and detected a very minor effect of GSK591 at the proteome level, with only 7 (0.2%) and 54 (1.4%) of 3839 proteins identified in both channels being up- and down-regulated, respectively (fig. S1B and table S1, spreadsheet 3). Despite the minor effect observed in protein expression, we still took this variation into account and used the SILAC protein ratios as normalization factor for each corresponding methyl peptide change.

The analysis of normalized methyl peptide SILAC ratios revealed more prominent down-regulation of R-methylation, with 86 (12.6%) methyl sites decreased and 56 (8.1%) up-regulated in both replicates, based on cutoffs calculated as described for the data in Fig. 1 (Fig. 2A and table S1, spreadsheet 2). To determine whether the observed changes were statistically significant based on the two biological replicates, we applied a moderated t test through the limma algorithm (24, 25), as previously done in similar studies (26), and found that 84% of the GSK591-responding methyl sites were statistically significant (fig. S1C). The dimethylated peptides were particularly enriched in the subset down-regulated by GSK591 (27.9%), as compared to the up-regulated one (21.4%) or the unchanging subset (20.7%), in line with the expected decrease of DMA following the inhibition of a type II PRMT. However, the relatively small subset of R-methyl peptides regulated by PRMT5 inhibition indicated that the R-methyl proteome was not markedly perturbed by the blockade of the major type II enzyme, which is in line with the response extent measured in a similar MS study, where PRMT5 was depleted and pharmacologically inhibited with a different compound, but only monomethylation was assessed (18).

Fig. 2 PRMT5 protein targets are RNA processing factors methylated within GRG sequences.

(A) The histogram indicates the number of sites carrying MMA (darker) and DMA (lighter) within each group of methyl sites differentially responding to GSK591: unchanging (gray), down-regulated (blue), and up-regulated (orange) (n = 2). (B) Bar graph displaying the opposite trend of response to GSK591 between DMA (orange) and MMA (green) for the peptides carrying the indicated methylated R sites, as quantified in both forward and reverse experiments and expressed as log2-transformed SILAC ratios (PRMT5i/control) (n = 2). (C) Motif analysis performed with iceLogo shows the overrepresentation of glycine residues at positions −1 and + 1 with respect to the methylated R in the peptides regulated by GSK591, meaning PRMT5 targets (foreground). The complete dataset of R-methyl peptides identified was used as background (P < 0.05; P value of the glycines in the +1 and −1 positions, <0.00125) (n = 2). (D) Interaction analysis of the proteins carrying GSK591-regulated R-methyl peptides generated using the STRING database and visualized with Cytoscape. RNA splicing factors are highlighted in yellow (n = 2). (E) Assessment of whether specific GO features are enriched in the group of regulated peptides. Methylated proteins were tested against the human proteome for significantly overrepresented Molecular Function (light blue), Cellular Components (cyan), and Biochemical Processes (dark blue) terms using David software (P < 0.05) (n = 2).

Focusing on the modulated peptides, we noticed that an increase of MMA on various sites corresponded with a reduction in DMA at the same residues. This suggests that the paradoxical increase of MMA after PRMT5 inhibition could actually result from the partial loss of methylation at PRMT5 target sites. For instance, MMA and DMA of eukaryotic translation initiation factor 4 γ3 (EIF4G3) at Arg881, constitutive coactivator of PPARγ-like protein 1 (FAM120A) at Arg886, heterogeneous nuclear ribonucleoprotein H1 (hnRNPH1) at Arg217 and Arg224, RPS10 at Arg158 and Arg160, and zinc finger protein 326 (ZNF326) at Arg175 were increased and decreased, respectively, in cells treated with GSK591 (Fig. 2B). This finding, achieved at the single-site resolution by MS, was confirmed globally by Western blot using pan-SDMA and pan-MMA antibodies (Fig. 1B) and was somehow reminiscent of the previously reported widespread increase of MMA that paralleled the ADMA decrease upon loss of PRMT1 enzyme (27).

For 270 R-methyl peptides quantified in both biological replicates, we could calculate the respective protein SILAC ratio in either only one (156) or none of the input samples (114); hence, such peptides lacking the protein normalization need further validation to determine whether their change occurs at the modification or protein level (table S1, spreadsheet 2b). Nevertheless, the pattern of regulation of this group of modified peptides was comparable to the normalized ones, with 29 sites (12.4%) being down-regulated and only 15 (6.4%) being increased by PRMT5 inhibitor (fig. S1D).

Overall, using a high-resolution MS-based quantitative strategy, we compiled the first combined profile of global MMA and DMA changes in response to PRMT5 inhibition, with a single-site resolution. Beside the identification of new PRMT5 targets, our data unraveled a complex pattern of response consisting in a down-regulation of DMA, partially associated with increased MMA at the same sites.

Targeting of RNA processing factors by PRMT5 within RGG domains

To establish whether the methyl sites changing upon GSK-591 treatment display the enrichment of specific local sequence contexts around the modified arginine, we analyzed them with the iceLogo web server (28) using the whole R-methyl peptide dataset as background. The motif analysis revealed that the regulated methyl peptides showed significant enrichment for glycine residues at positions −1 and +1 from the modified arginine (Fig. 2C). Noteworthy, the consensus motif emerging from the protein extract input before immunoenrichment confirmed the presence of G in position −1, whereas G at position +1 was not significantly enriched (fig. S1E). This confirmed that the GR motif was genuinely associated to the activity of PRMT5, whereas the use of pan-methyl antibodies to enrich for R-methyl peptides might have introduced a slight bias for G at position +1 (see also Fig. 5 and discussion for further elucidation). Furthermore, logo analysis showed that proline (P) at position −1 disfavors methylation by PRMT5, consistently with the PR sites being preferentially methylated by PRMT4/CARM1 (Fig. 2C) (29). The “GR” sequence belongs to the family of arginine- and glycine-rich motifs (RGG/RG), whose asymmetrical and symmetrical arginine dimethylation affects protein-DNA and protein-protein interactions (1, 30). Hence, it is very likely that these RGG domains carry both ADMA and SDMA in human cells, making the selective immunoenrichment of one dimethylation form over the other an indispensable step before LC-MS.

Having established that 134 peptides showed the enrichment of PRMT5 consensus sequence and carry methyl sites changing in dependence of GSK591 treatment, we investigated their parent proteins: the peptides mapped on 62 proteins (table S1, spreadsheet 6), some of which were previously described as PRMT5 substrates [such as RPS10, hnRNPA1, ZNF326, DEAH (Asp-Glu-Ala-His) box helicase 9 (DHX9), hnRNPH1, and splicing factor proline- and glutamine-rich protein (SFPQ)]. Target proteins mainly belonged to the class of RNA binding proteins (RBPs), with established roles in the regulation of RNA splicing (Fig. 2D) (3032). In addition, we found PRMT5-dependent methylation sites (such as Arg158/160 on RPS10 and Arg881 on EIF4G3) that lost SDMA upon GSK591 treatment and were involved in translation, in line with the emerging role for SDMA in the regulation of protein synthesis (33). These included the eukaryotic translation initiation factor EIF4G3, which is involved in mRNA cap recognition and transport to ribosome. Gene Ontology (GO) analysis of the regulated methyl proteins over the nonresponding ones confirmed a substantial enrichment for nuclear RBPs involved in various steps of RNA processing (Fig. 2E and table S1, spreadsheet 7). For example, among the GSK591-regulated proteins, we found RNA binding motif protein 3 (RBM3), involved in RNA splicing in stress conditions (9), RBM27, which has a poorly characterized RNA recognition motif (RRM), and RBM47, whose RNA chaperone function was shown to be crucial for the splicing of a subset of mRNAs, thus determining their expression (34).

Orthogonal validation of GSK591-regulated methyl peptides by hmSILAC

Despite the expansion of annotated methyl proteomes in human cells, global analysis of methylation by MS suffers from high false discovery rate (FDR), because methylation can be easily mis-assigned, due to the existence of various amino acid substitutions isobaric to this modification, or to chemical methylation of various residues arising during sample preparation. This awareness has highlighted the need for orthogonal validation strategies in MS to discriminate artifacts from genuine methylation events in live cells (1921, 35). To validate our quantitative data, we adopted the hmSILAC approach, first established by the Mann group in 2004, which is based on growing cells in medium containing 13CD3-labeled methionine, to mark methyl groups with heavy isotopes (Fig. 3A). Methionine is metabolically converted into S-adenosyl-l-methionine (AdoMet), the sole cellular methyl group donor; hence, heavy methyl groups will be added on newly synthesized proteins upon enzymatic reaction of protein methylation. When light and heavy cells are mixed in a 1:1 ratio, each peptide bearing either a methionine or a methyl group will be detected at MS1 level as a doublet, consisting of light and heavy forms separated by a delta mass, predictable and diagnostic for the number of incorporated methyl groups. The presence of the doublet allows discriminating enzymatic methylation from amino acid substitutions and/or chemical artifacts; thus, the hmSILAC approach has emerged as the gold standard to minimize FDR in global analysis of protein methylation by MS (21). We generated a repository of R-methyl peptides from three hmSILAC-labeled human cancer cell lines (NB4, SK-OV-3, and HeLa cells), using pan-methyl antibodies specific for MMA, SDMA, and ADMA, according to the experimental workflow (Fig. 3A).

Fig. 3 Orthogonal validation of PRMT5 targets by intersection with the hmSILAC methyl proteome.

(A) Schematic representation of the strategy adopted to validate the dynamic SILAC experiments through the intersection with a high-confidence hmSILAC-based dataset of R-methyl peptides, upon immunoenrichment of MMA, SDMA, and ADMA peptides from protein extracts of three different hmSILAC-labeled cell lines, before LC-MS/MS. The hmSILAC doublets were searched in the MaxQuant output tables using the in-house pipeline hmSEEKER (see also fig. S2) and then matched to the SILAC methyl peptides through the hmLINKER script. (B) Validation of the SILAC dynamic dataset (left), the nonresponding methylated peptides (top right), and the GSK591-regulated methyl peptides (bottom right) by intersection with the hmSILAC dataset. The percentages of validated and nonvalidated methyl peptides are displayed in light and dark blue, respectively (n = 2).

The automated and confident identification of hmSILAC doublets from raw MS data was achieved through an in-house developed algorithm called hmSEEKER that, starting from the MaxQuant-executed methyl peptide search output, automatically associates the light and heavy forms and retrieves hmSILAC doublets (fig. S2 and table S2) (36). This tool has been designed to reconstruct hmSILAC pairs, also when the identification of either the light or the heavy peak is not available.

We then validated the SILAC-based PRMT5-dependent methyl proteome by intersecting it with our high-confidence repository (Fig. 3A) and found that 77% of the SILAC methyl peptides were validated by the presence of hmSILAC pairs (Fig. 3B and table S3). Whereas most methylations were confirmed as genuine enzymatic ones, the 23% nonconfirmed highlighted a potential high FDR associated to our SILAC-based dynamic analysis. We reasoned that false-positive methylations would not be affected by the inhibitor treatment and hence might fall in the unchanging peptide subset. Accordingly, the fraction of unchanging peptides found in the hmSILAC repository was similar (75%), whereas the intersection with the significantly changing methyl peptides increased to 84% (Fig. 3B and table S3). Although this result seems to confirm a higher degree of confidence associated to the methyl peptides dynamically regulated by the PRMT5 inhibitor, we cannot exclude that the imperfect match between the two datasets may be also, in part, due to the incomplete coverage of the hmSILAC repository.

Validation with in vitro methylation assays and importance of neighboring residues for PRMT5 methylation

The crystal structure of human PRMT5 has been solved when in complex with its binding partner methylosome protein 50 (WDR77, also known as MEP50), the S-adenosylmethionine (SAM) analog A9145C, and a histone H4 peptide (37). MEP50 is known to stimulate PRMT5 catalytic activity (37, 38). We first verified these findings by expressing human PRMT5, alone and in combination with MEP50 in insect cells (fig. S3A), and tested different amounts of PRMT5 for in vitro methylation assays. The PRMT5-MEP50 complex showed a higher methyltransferase activity than PRMT5 alone (fig. S3B) and was then purified by affinity and gel filtration chromatography (fig. S3C). The catalytic activity of PRMT5-MEP50 was further tested on recombinant histones H2A, H2B, H3, and H4 (fig. S3D). As previously demonstrated, purified PRMT5-MEP50 can methylate H2A and H4 efficiently and H3 weakly (38).

hnRNPA1, hnRNPH1, hnRNPK, SFPQ, KHDRBS1 (KH RNA binding domain-containing, signal transduction–associated 1), and CNBP (cellular nucleic acid–binding protein) were selected for further in vitro validation of MS data. Although these proteins were previously described as PRMT1 substrates (30, 3942), we found their methylation differentially regulated upon GSK591 treatment with the exception of hnRNPK (table S1, spreadsheets 2 to 6). Methylation assays on these recombinant proteins proved that they could be directly methylated in vitro by both PRMT5-MEP50 (Fig. 4A) and PRMT1 (Fig. 4B), but not by PRMT4 (Fig. 4C). Nonetheless, we observed relative differences in selectivity between PRMT1 and PRMT5. Specifically, a glutathione S-transferase (GST)–tagged SFPQ peptide composed of amino acids 1 to 57 of SFPQ [hereafter “SFPQ (1–57)”] could be methylated by both PRMT5 and PRMT1, whereas hnRNPH1 and SFPQ (298–707) were better substrates for PRMT5. In addition, whereas hnRNPH1 could be methylated by both PRMT5 and PRMT1, a hnRNPH1 mutant in which the arginines MS-identified as PRMT5 targets were mutated into lysines could no longer be methylated by PRMT5, and PRMT1 methylation was unaffected. This again emphasized the differences in selectivity between PRMT5 and PRMT1 (fig. S4, A and B). To test for a functional readout of the PRMT5-mediated methylation of hnRNPH1, we first depleted endogenous hnRNPH1 by short hairpin RNA (shRNA) and concurrently overexpressed either a wild-type hnRNPH1 or a R217K/R224K mutated version (fig. S4C). Cells rapidly stopped proliferating upon hnRNPH1 depletion, and only wild-type, but not mutant, hnRNPH1 was able to partially restore their growth (fig. S4D). Overall, our experiments confirmed the essential role of PRMT5 targets in cell proliferation and highlight the potential regulatory effect of PRMT5 methylation on their function.

Fig. 4 Biochemical validation of PRMT5 substrates.

(A) In vitro radioactive methylation assay on recombinant GST-tagged hnRNPA1, hnRNPH1, hnRNPK, SFPQ (157), SFPQ (298–707), KHDRBS1 (81–443), and CNBP proteins by purified PRMT5-MEP50. (B) As in (A), but the methylation assay was carried out using recombinant PRMT1 alone. (C) As in (A), but methylation reaction was carried out using recombinant PRMT4/CARM1. (D) Mutagenesis analysis on CNBP RGG/RG motif, achieved by mutating arginines to lysines at positions R25, R27, R32, and R34. The RGG/RG motif is highlighted in purple, and the zinc finger is highlighted in light blue. WT, wild type. (E) In vitro radioactive methylation assays by PRMT5-MEP50 on the arginine mutant CNBP proteins. The Coomassie gel of the recombinant proteins is shown at the bottom panel. (F) As in (E), but methylation assay was carried out using recombinant PRMT1. (G) PRMT5 methylation of Arg32 and Arg34 after mutation of Ser31. Mutant CNBP proteins CNBP-G26A and CNBP-G26A/S31G were used as substrates in in vitro radioactive methylation assay by PRMT5-MEP50. A Coomassie gel of the recombinant proteins is shown at the bottom. All autoradiography and Coomassie gels (A to C and E to G) are representative of three independent experiments.

As an additional validation step, synthetic unmodified and pre-methylated peptides from these six targets were used in in vitro methylation assays to further confirm the identified peptide sequences (fig. S4, E and F). All the unmodified peptides were actively methylated by PRMT5-MEP50, whereas the pre-methylated peptides could not be modified by PRMT5, with the exception of the KHDRBS1-ADMA peptide. The latter result could be explained by the presence of one residual unmodified R at the N terminus of the KHDRBS1-ADMA peptide (fig. S4F).

Differential methylation of RGG/RG motifs by PRMT5 and PRMT1

CNBP is an RNA and DNA binding protein involved in vertebrate craniofacial development (43). CNBP was reported to contain ADMA and SDMA and to be a substrate of PRMT1 (41). R-methylation of CNBP has been implicated in modulating its RNA binding capabilities (41). CNBP contains one RGG/RG motif with four arginine residues (Arg25, Arg27, Arg32, and Arg34) and seven zinc fingers (Fig. 4D). To further confirm our findings, we mutated the four arginine residues to lysine and purified the resultant mutant CNBP proteins. We then incubated the wild-type and mutant CNBP proteins with recombinant PRMT5-MEP50 complex (Fig. 4E), or PRMT1 alone (Fig. 4F), in the presence of radioactive SAM. Compared to wild type, the methylation activity of PRMT5 on CNBP-R25K and CNBP-R27K was attenuated. This was not the case for CNBP-R32K and CNBP-R34K. When both Arg25 and Arg27 were mutated to lysine, the methylation signal was inhibited, indicating that PRMT5 methylated Arg25 and Arg27 (Fig. 4E). These results suggested that PRMT5 preferentially targets Arg25 and Arg27, rather than Arg32 and Arg34. Next, we tried to determine why Arg32 and Arg34 are not ideal methylation sites for PRMT5. Under the hypothesis that Ser31 may block the methylation at these two sites, we first attenuated the methylation of Arg25 and Arg27 by mutating Gly26 to alanine, which inhibited the arginine methylation by PRMT5. Next, we mutated Ser31 to glycine. This CNBP-G26A/S31G mutant was efficiently methylated by PRMT5 (Fig. 4G), which suggested that the arginine methylation by PRMT5 is sequence dependent. Together, these results indicated that PRMT5 has more stringent substrate specificity than PRMT1.

Importance of glycine residues before and after the target arginine site for methylation by PRMT5

PRMT1 is known to methylate substrates beyond the RGG motif (44). However, PRMT5 activity has only been characterized on histone H4 peptide with mutations on the positively charged residues, but not on the amino acids before or after the target residue (H4 Arg3, or H4R3). Therefore, we felt that it was important to investigate the substrate selectivity of PRMT5 on nonhistone proteins. Given the PRMT5-selective methylation of CNBP Arg25 and Arg27, we wanted to have a clean system to assess the impact of the upstream and downstream amino acids on arginine methylation. To have such a system, we mutated the full-length CNBP Arg27 to glycine (R27G). This construct was methylated by PRMT5 exclusively on Arg25 and results in the methylation signal being reduced by half, when compared to wild-type CNBP, which was methylated on Arg25 and Arg27 (Fig. 5, A and B, and fig. S5, A and B). Gly24 on CNBP was further mutated into different amino acids and subjected to in vitro methylation by PRMT5. When Gly24 was replaced by polar, hydrophobic, or charged amino acids, PRMT5 methylation activity on Arg25 was strongly attenuated (Fig. 5A). This indicated that an arginine positioned after any amino acid other than glycine is a poor substrate for PRMT5. The human PRMT5-MEP50 structure provides the molecular basis of PRMT5 substrate binding: Gly2 on the histone H4 peptide precedes the target arginine site, whose main-chain carbonyl group forms a hydrogen bond with the main-chain amino group of PRMT5 Phe580. Gly2 is in close proximity to Phe580; therefore, the substitution of this glycine residue to any other amino acid creates a physical hindrance with Phe580 and blocks the formation of the hydrogen bond (Fig. 5C).

Fig. 5 Substrate specificity of PRMT5 on Gly24mutant CNBP proteins.

(A) Arg27 (R27) on CNBP was first mutated to glycine. Gly24 was further mutated to different amino acids on the CNBP R27G construct. Mutant CNBP proteins were expressed as GST fusion protein, purified, and subjected to in vitro radioactive methylation assay by PRMT5-MEP50. Quantification of each radioactive signal was normalized to the signal of CNBP R27G. Data are representative of three independent experiments. (B) Substrate specificity of PRMT5 on Gly26 mutant CNBP proteins. Gly26 was substituted to different amino acids on the CNBP R27G construct. Mutant CNBP proteins were subjected to in vitro radioactive methylation assay with PRMT5. Quantification of each radioactive signal was normalized to the signal of CNBP R27G. Data are representative of three independent experiments. (C) Hydrogen bond is formed between Gly2 of histone H4 and Phe580 of PRMT5. Histone H4 peptide and Phe580 on PRMT5 are shown as stick representation. PRMT5 SAM-binding domain is shown in red, β-barrel domain in yellow, TIM domain in light pink, SAM analog A9145C in green, H4 peptide in magenta, and MEP50 in cyan. (D) Proposed model illustrating the interaction between GRK motif and PRMT5. The overall structure of PRMT5-MEP50 complex (Protein Data Bank: 4GQB) is shown on top. Hydrogen bond is formed between Ser1 of the H4 peptide and Gln309 of PRMT5 (left bottom panel). The bottom right panel models the substitution of Gly4 to lysine residue on H4 peptide. Histone H4 peptide, Tyr307, and Gln309 on PRMT5 are shown as stick representation. PRMT5 SAM-binding domain is shown in red, β-barrel domain in yellow, TIM domain in light pink, SAM analog A9145C in green, H4 peptide in magenta, and MEP50 in cyan.

Next, we constructed a series of mutants at position 26 to assess the impact of the amino acid side chain following the target arginine (Arg25). As expected, most of Gly26 mutant CNBP proteins were poor substrates for PRMT5, except for G26Y and G26K (Fig. 5B and fig. S5B). To understand this selectivity, we modeled a mutated Gly4 on the substrate H4 peptide and proposed the following mechanism: The main-chain carbonyl group of Ser1 of H4 forms a hydrogen bond with the side-chain amino group of Gln309 of PRMT5 (Fig. 5D). Mutation of Gly4 to lysine in H4 may lead to the translocation of Gln309 residue on PRMT5 and break this hydrogen bond interaction. However, the side chain of lysine in this position may form a potential hydrogen bond with main-chain carbonyl group of Tyr307 (Fig. 5D). The same situation may happen on a G4Y mutant, whose side chain contains a hydroxyl group that can be involved in hydrogen bond formation. In contrast, replacement of Gly4 with a hydrophobic residue on histone H4 would not lead to the formation of a hydrogen bond with Tyr307. Overall, these experiments confirmed the importance of glycine at position −1 for the efficient methylation by PRMT5 (Fig. 2C). The requirement for the glycine at position +1 may be less stringent, because the enzyme seems to tolerate the presence of other amino acids that can fit the catalytic pocket and form alternative transient interactions, favoring the methylation reaction.

Discrimination between ADMA and SDMA by inspection of hmSILAC methyl peptide MS2 spectra

Previous studies on histones indicate that SDMA and ADMA can decorate the same arginine site, where they elicit different, and even opposite, functional outcomes; prototypic examples are H4 Arg3 and H3 Arg2. For the former, symmetric dimethylation by PRMT5 (H4R3me2s) leads to transcriptional repression, whereas asymmetric dimethylation (H4R3me2a) by PRMT1 is linked with gene activation (45); the opposite is true for the latter, with PRMT5-mediated methylation (H3R2me2s) leading to gene activation, whereas PRMT6-mediated H3R2me2a is associated with heterochromatinization and silencing (5, 4648).

The marked overrepresentation of glycine residues in the positions surrounding the modified arginines identified by SILAC-MS upon PRMT5 inhibition confirmed that PRMT5 preferentially targets glycine-enriched sequences, which are anyway also typically targeted by PRMT1, the major type I PRMT responsible for 90% of ADMA in human cells (49). Hence, in light of the possible competition between PRMT1 and PRMT5 for the same sites on nonhistone proteins, several RGG sequences are likely present as both ADMA and SDMA: these two modifications are isobaric and thus undistinguishable in the survey scan (MS1). However, different fragmentations of the side chain of ADMA and SDMA are generated in the MS2 spectra, enabling discrimination between these two dimethylation variants. Specifically, in the presence of SDMA, the neutral loss of a monomethylamine (NH2CH3, 31.04 Da) was generated and could be detected in MS2 (Fig. 6A), whereas the neutral loss of dimethylamine [NH(CH3)2, 45.05 Da] was detectable when ADMA is present (Fig. 6B) (50).

Fig. 6 Unambiguous assignment of SDMA by MS2 spectrum inspection.

(A) Left: Schematic representation of how neutral loss of monomethylamine arises from MS/MS fragmentation of SDMA. Right: MS2 spectra of the indicated hmSILAC peptide in both heavy and light forms are shown: Ions b2, b3, b4, b6, and b7 of the light peptide show a neutral loss of 31.04 Da (light monomethylamine), whereas the corresponding ions of the heavy counterpart display a neutral loss of 35.06 Da (heavy monomethylamine), as expected from the stable isotope labeling of the methyl group. The methyl peptide was enriched using the SDMA type–specific antibody. Full MS (MS1) of the corresponding hmSILAC doublet is shown in the inset. The difference in mass to charge ratio (Δm/z) given by the presence of the isotopically labeled dimethylated group is 8/3 = 2.68 Thomson (Th). The MaxQuant-based search for neutral losses was performed twice (n = 2). MW, molecular weight. (B) Schematic representation showing how neutral loss of dimethylamine arises from MS/MS (MS2) fragmentation of ADMA (left). MS2 spectra of the indicated hmSILAC peptide in both heavy and light forms are shown (right): Ions y9, y10, and y11 of the light peptide show a neutral loss of 45.06 Da (light dimethylamine), whereas the corresponding ions of the heavy counterpart show a neutral loss of 53.10 Da (heavy dimethylamine), due to the stable isotope-labeled methylation. The methyl peptide was enriched using the ADMA type–specific antibody. MS1 of the corresponding hmSILAC doublet is shown in the inset. The Δm/z of the isotopically labeled dimethylated group is 8/3 = 2.68 Th. The MaxQuant-based search for neutral losses was performed twice (n = 2). (C) Bar graph showing the number of R-methyl peptides displaying the neutral loss of monomethylamine/dimethylamine in the subsets of the GSK591-responding or not regulated. The MaxQuant-based search for neutral losses was performed twice (n = 2).

We searched for the neutral losses of monomethylamine and dimethylamine in the hmSILAC dataset of dimethylated peptides enriched using antibodies to SDMA and ADMA, respectively. The identification at the MS2 level of the fragment ions resulting from the neutral loss of both light and heavy mono- and dimethylamine confirmed that they were genuine methylation events in live cells (Fig. 6, A and B, and data file S1). Furthermore, because PRMT5 targets preferentially GRG motifs that are often present as highly repetitive sequences, an average-length methyl peptide might harbor more than one dimethylated Arg (R) site, which trypsin typically does not cleave; as a matter of fact, the inspection of the MS2 spectra of a subset of multimethylated peptides enriched by the SDMA-specific antibody revealed the specific loss of two monomethylamines, confirming the co-occurrence of two SDMAs on the same peptide (fig. S6).

On the basis of this result, we searched for the neutral loss of monomethylamine and dimethylamine also in the dynamic SILAC-based dataset of methyl peptides affinity-enriched with the antibody to SDMA. Notably, we identified the loss of monomethylamine in the MS2 spectra of all significantly down-regulated R-methyl peptides, confirming that they were true SDMA-bearing peptides (Fig. 6C, data file S2, and table S1, spreadsheet 8). Conversely, the peptides displaying loss of dimethylamine, and not monomethylamine, were all found in the group of the ones unchanging upon PRMT5 inhibition (Fig. 6C, data file S3, and table S1, spreadsheet 9).

The detection of peptides bearing both dimethylamine and monomethylamine neutral losses suggests possible cross-talk between PRMT1 and PRMT5, which may impinge on the same residues. In-depth inspection of their MS/MS fragmentation spectra showed, however, that most of them contain multiple sites that can be methylated. It was thus not possible to determine unambiguously whether the two modifications occurred on exactly the same residue. Because only three peptides of this group—annotated in the hmSILAC dataset and mapping to proteins G3BP1, TFEB, and CHTOP, respectively—displayed symmetric and asymmetric dimethylation on the same arginine (data file S4), we could not extrapolate a motif specific for the shared sites and distinct from the SDMA- or ADMA-only sites. Our data indicate that the search for dimethylamine and monomethylamine neutral losses is a unique tool for the annotation of PRMT-specific methyl proteome datasets.


In this study, we globally profiled protein arginine mono- and dimethylation in human cancer cells upon pharmacological inhibition of PRMT5, demonstrating the preferential modulation of arginine residues within the GR motif. In particular, PRMT5 impairment led to a decrease of SDMA at some sites, which was paralleled by an increase of MMA on the same residues.

Loss of PRMT1 in mouse embryonic fibroblasts (MEFs) has already been shown to reduce ADMA and markedly increase the global MMA and SDMA in cells (27). Therefore, loss of SDMA resulted in accumulation of MMA at the same sites, suggesting a direct methylation of these substrates by PRMT5. However, we cannot rule out that other PRMTs may actively participate in the up-regulation of MMA when PRMT5 is blocked, nor we can calculate the contribution of site-specific activity of potential demethylases in this equation. Furthermore, it is noteworthy that the pattern of MMA variation in response to GSK591 included sites that showed a concomitant decrease of MMA and SDMA. Most probably, these sites either are particularly sensitive to PRMT5 inhibition or have a high demethylation kinetics.

Our analysis highlighted the predictive value of a glycine before, and to a minor extent after, the target arginine (GR or GRG) for the identification of bona fide PRMT5 targets. We want to stress that a glycine following the target arginine seemed to favor in vitro methylation by PRMT5 over more hydrophobic or polar amino acids (Fig. 5B). However, given the design of our experiment, which included an enrichment step of R-methylated peptides using a polyclonal antibody raised using peptides that presented a G+1, we cannot exclude a certain degree of bias in our target prediction logo (Fig. 2C).

To unambiguously distinguish between asymmetrically and symmetrically DMAs, we combined the methyl peptides immunoenrichment with both the high quality ensured by the use of hmSILAC validation and the manual inspection of MS2 spectra for identification of the diagnostic ions derived from ADMA and SDMA neutral losses (50). The detection of predicted neutral losses of monomethylamine and dimethylamine in the MS2 spectra of affinity-enriched peptides confirmed the selectivity of the pan-methyl antibodies used and the reliability of our PRMT5 target analysis. Because the anti-SDMA antibody had never been used before for peptides immunoenrichment before MS, this result is particularly relevant because, by proving the efficacy of this reagent for global MS studies, it paves the way to PRMT5 target profiling in different model systems or functional states.

To establish whether some of the identified methylated proteins were direct PRMT5 targets, we validated CNBP, hnRNPA1, hnRNPK, hnRNPH1, SFPQ, and KHDRBS1 by in vitro methylation assays and investigated the substrate specificity of PRMT5 on CNBP. In vitro methylation assays on arginine mutant CNBP and hnRNPH1 proteins indicated that PRMT5 differed from PRMT1 in substrate specificity. In addition, the substrate specificity of PRMT5 on the Gly24 and Gly26 mutant CNBP proteins indicated that the glycine residue before, and to a lesser extent after, the target arginine was critically important for methylation. Besides that, the observation that substitution of Gly26 to tyrosine or lysine enables PRMT5 methylation is an interesting finding that can help to computationally predict previously unknown PRMT5 targets in different cellular contexts.

Together, our results set the foundation for the amino acid sequence–based prediction of putative PRMT5 target site methylation among more than 1000 human proteins that contain the RGG/RG motif (51). This is of paramount importance now that several selective and potent PRMT5 inhibitors have been developed, and one (GSK3326595) is currently being tested in patients with non-Hodgkin’s lymphomas and solid tumors (NCT02783300). The outstanding question is which of the many PRMT5 targets should be traced in each cancer to monitor and predict response. We believe that our study, which on purpose was conducted not in an experimental setting of complete PRMT5 deletion but rather using a drug at IC50 (half maximal inhibitory concentration) concentrations, addresses this important question. Specifically, our data highlight the relevance of targets involved in RNA processing downstream of PRMT5 enzymatic inhibition.


Cell culture conditions and reagents

HeLa cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS; Life Technologies), 1% glutamine, penicillin (100 U/ml), and streptomycin (100 mg/ml) and cultured as a suspension culture in roller bottles at 37°C. The medium was supplemented with 25 mM Hepes (pH 7.5) to compensate for the lack of controlled CO2 percentage. GSK591 (synonyms: GSK3203591, EPZ015866) and its inactive structural analog were obtained from the Structural Genomic Consortium (SGC, Toronto) and used at 50 nM for 6 days.

SILAC and hmSILAC cell culture conditions

For SILAC, HeLa cells were grown in “light” and “heavy” SILAC DMEM (Thermo Fisher Scientific, #88364), supplemented with either l-arginine, l-lysine, or their heavy isotope counterparts l-arginine (13C6, 99%; 15N4, 99%; Sigma-Aldrich, 608033) and l-lysine (13C6, 99%; 15N2, 99%; Sigma-Aldrich, 608041). Arginine and lysine were added at the concentration of 84 and 146 mg/liter, respectively. The SILAC media were further supplemented with 10% dialyzed FBS (Gibco, Life Technologies, 26400-044), 1% glutamine, penicillin (100 U/ml), streptomycin (100 mg/ml), and 10 mM Hepes (pH 7.5).

For hmSILAC, HeLa and NB4 cells were cultured in light and heavy hmSILAC DMEM media (Carlo Erba reagents, custom-made), whereas the ovarian cancer cell line SK-OV-3 was grown in hmSILAC RPMI media. Both custom-made media were supplemented with l-arginine (Sigma-Aldrich, A6969), l-lysine (Sigma-Aldrich, L8662), and either l-[13CD3]methionine (Met-4, heavy, Sigma-Aldrich, 299154) or l-[12CH3]methionine (Met-0, light, Sigma-Aldrich, M5308), respectively. The concentration of l-methionine was 30 mg/liter. The hmSILAC media were then supplemented with 10% dialyzed FBS (Gibco, Life Technologies, 26400-044), 1% glutamine, penicillin (100 U/ml), streptomycin (100 mg/ml), and 10 mM Hepes (pH 7.5).

Both SILAC and hmSILAC HeLa cells were grown in the respective heavy isotopes containing media for at least nine replication cycles to ensure full incorporation of heavy amino acids, with careful monitoring of their growth rate, viability, and overall morphology, to guarantee that the normal physiology was preserved.

Western blot analysis

For whole-cell lysis, harvested cell culture pellets were rinsed with phosphate-buffered saline (PBS), lysed in SDS-containing buffer [4% SDS, 20% glycerol, 0.1 M tris-HCl (pH 7.5)], and sonicated. Protein lysates were centrifuged at maximum speed for 10 min. Protein extracts were quantified using the BCA Protein Assay Kit (Pierce, 23225), and equal amounts of protein were resolved by SDS–polyacrylamide gel electrophoresis (PAGE) electrophoresis and blotted onto the transfer membrane (Immobilon-P, Merck Millipore, IPVH00010). Membrane blocking (10% bovine serum albumin/tris-buffered saline and 0.2% Tween 20 for 30 min at room temperature) was followed by incubation with the appropriate primary antibodies and horseradish peroxidase–conjugated secondary antibodies (Cell Signaling Technology). Proteins were detected by enhanced chemiluminescence (Thermo Fisher Scientific). The following primary antibodies were used: Anti-monomethylated arginine (MeR4-100, Cell Signaling Technology, #8015, 1:1000 dilution), anti-symmetric dimethylated arginine (Cell Signaling Technology, #13222, 1:1000 dilution), and anti-asymmetric dimethylated arginine (Cell Signaling Technology, #13522, 1:1000 dilution) were purchased from Cell Signaling Technology; anti-vinculin (06-866, 1:10,000 dilution) was purchased from Merck Millipore.

Arginine methyl peptide separation and enrichment before LC-MS/MS analysis

Equal numbers of light- and heavy-labeled HeLa cells differentially treated were mixed in a 1:1 ratio, pelleted, and washed twice with PBS. Cell pellets were lysed in urea lysis buffer [9 M urea, 20 mM Hepes (pH 8.0)] supplemented with 1× Roche proteases and phosphatase inhibitors, sonicated, and cleared by ultracentrifugation (20,000g for 15 min at 15°C). For in-solution digestion, 50 mg of proteins was reduced by adding 4.5 mM dithiothreitol (DTT) (Sigma-Aldrich) for 30 min at 55°C, alkylated with 5.5 mM iodoacetamide [10% (v/v) for 15 min at room temperature in the dark; Sigma-Aldrich], and digested overnight with sequencing-grade trypsin [1:100 (w/w); Promega] after a fourfold dilution in 25 mM ammonium bicarbonate solution. Protease digestion was terminated by the addition of trifluoroacetic acid (TFA) to adjust pH < 3. Precipitated material was removed by centrifugation for 15 min at 1780g at room temperature. Peptides were purified using reversed-phase Sep-Pak C18 cartridges (Waters, Milford, MA) and eluted off the Sep-Pak with 40% acetonitrile (ACN) with a subsequent step of removal of ACN by 48 hours of lyophilization. Lyophilized peptides were dissolved in 25 mM ammonium hydroxide (NH4OH) and subsequently offline fractionated by high-pH fractionation using a Phenomenex Jupiter C12 4 μm Proteo 90 Å, LC column 250 × 4.6 mm, on an ÄKTA-FPLC (fast protein liquid chromatography) system (GE Healthcare) operating at 1 ml/min. Buffer A was 25 mM NH4OH, and buffer B was 25 mM NH4OH in 90% ACN. Fractions were collected using a collector in a 96-deep well plate at 1-min intervals. Samples were initially loaded onto the column at 1 ml/min for 3 min, after which the fractionation gradient was as follows: 5% B to 30% B in 60 min, 30% B to 60% B in 2 min and ramped to 70% B for 3 min. At this point, fraction collection was halted, and the gradient was held at 100% B for 5 min before being ramped back to 5% B, where the column was then washed. The 60 collected fractions were concatenated to 14 throughout each experiment. After lyophilization, each fraction was dissolved in 250 μl of 1× immunoaffinity purification buffer (IAP buffer, #9993, Cell Signaling Technology) and subjected to two consecutive steps of methyl-R peptide enrichment using the SDMA antibody-conjugated beads (PTMScan [sdme-R] Kit #13563, Cell Signaling Technology) and MMA antibody-conjugated beads (PTMScan Mono-Methyl Arginine Motif [mme-RG] Kit #12235, Cell Signaling Technology) following the manufacturer’s instruction. After peptide incubation with the antibody-conjugated beads for 2 hours at 4°C, the immunoprecipitates were washed twice in ice-cold IAP buffer followed by three washes in water; then, bound methyl peptides were eluted with 2 × 50 μl of 0.15% TFA. Peptide eluates were desalted on reversed-phase C18 StageTips, as described previously (52), and subjected to a second round of trypsin digestion before nano–LC-MS/MS analysis.

Nano–LC-MS/MS analysis

Peptide mixtures were analyzed by online nanoflow LC-MS/MS (nLC-MS/MS) using an EASY-nLC 1200 (Thermo Fisher Scientific) connected to a Q Exactive instrument (Thermo Fisher Scientific) through a nanoelectrospray ion source. The nano-LC system was operated in one column setup with a 50-cm analytical column (75-μm inner diameter and 350-μm outer diameter) packed with C18 resin (EASY-Spray PepMap RSLC C18 75 μm inner diameter, 2 μm particles; Thermo Fisher Scientific) configuration. Solvent A was 0.1% formic acid (FA), and solvent B was 0.1% FA in 80% ACN. Samples were injected in an aqueous 0.1% TFA solution at a flow rate of 500 nl/min. SILAC and hmSILAC immunoenriched methyl peptides were separated with a gradient of 5 to 30% solvent B over 120 min, followed by a gradient of 30 to 60% for 10 min and 60 to 95% over 5 min at a flow rate of 250 nl/min in the EASY-nLC 1200 system. The Q Exactive was operated in the data-dependent mode to automatically switch between full-scan MS and MS/MS acquisition. Survey full-scan MS spectra (from m/z 300 to 1650) were analyzed in the Orbitrap detector with resolution R = 60,000 at m/z 200. The 10 most intense peptide ions with charge states ≥2 were sequentially isolated and fragmented by higher-energy collision dissociation with a normalized collision energy setting of 28%. The maximum allowed ion accumulation times were 20 ms for full scans and 50 ms for MS/MS, and the target value for MSMS was set to 1 × 106. The dynamic exclusion time was set to 20 s.

Data analysis of SILAC arginine methyl peptides

Acquired raw data were analyzed with the integrated MaxQuant software v1.6.1.0 using the Andromeda search engine (53, 54). In MaxQuant, the estimated FDR of all peptide identifications was set to a maximum of 1%. The main search was performed with a mass tolerance of 7 parts per million (ppm). Enzyme specificity was set to trypsin/P. A maximum of three missed cleavages was permitted, and the minimum peptide length was fixed at seven amino acids. Carbamidomethylation of cysteine was set as a fixed modification. The 2018_07 (18 July 2018) version of the UniProt sequence was used for peptide identification.

To assign and quantify SILAC methyl peptides, all raw data were processed, indicating N-terminal acetylation, methionine oxidation, mono-methyl-K/R, and di-methyl-K/R as variable modifications. The MaxQuant evidence.txt output file was then filtered: Potential contaminants and reverse sequences were removed, and methyl peptides were required to have an Andromeda score ≥ 25 and a PTM localization probability ≥ 0.75. For the methyl peptides quantified more than once, the median SILAC ratio was calculated. Last, methyl peptide SILAC ratios were normalized on the respective protein SILAC ratios extracted from the proteinGroups.txt MaxQuant output file. These were calculated using unmodified peptides in the “input” experiment. To define up- or down-regulated methyl peptides by GSK591, we used mean (μ) and SD (σ) based on the distribution of the unmodified peptide SILAC ratios calculated separately in the forward and reverse experiments and applied a μ ± 3σ cutoff to the distributions of the modified peptides of the respective replicate. To assess which of the regulated peptides were statistically significant, we used the limma algorithm in the “DEP” R package (24, 25) and applied the following filters: log2 fold change >0.6 (fold change >1.5 on a linear scale) and adjusted P value [corrected with Benjamini-Hochberg method (55)] <0.05.

hmSEEKER: A Perl-based pipeline for high-confidence assignment of methyl peptides from hmSILAC data

To assign hmSILAC peptide sequences, we defined new modifications in MaxQuant with the mass increment and residue specificities corresponding to heavy monomethylation (mono-methyl4-K/R) and dimethylation (di-methyl4-K/R). In addition, we defined new modifications for heavy methionine (Met4) and oxidized heavy methionine (OxMet4). To reduce the search complexity, raw data were analyzed twice with the following sets of variable modifications: (i) N-terminal acetylation, Met4, OxMet4, oxidation, mono-methyl-K/R, mono-methyl4-K/R; (ii) N-terminal acetylation, Met4, OxMet4, oxidation, di-methyl-K/R, di-methyl4-K/R.

Identification of high-confidence methyl sites was carried out with an in-house developed, Perl-based pipeline, named hmSEEKER (36). hmSEEKER identifies doublets of heavy and light hmSILAC peptides from MaxQuant output tables. However, during data-dependent acquisition, the choice of peptide ions to fragment is stochastic, so there are no guarantees that the heavy and light forms will both be sequenced. Hence, we adopted a strategy to retrieve heavy/light counterparts that are not identified by MS/MS. Briefly, the doublet search step is performed as follows: Information regarding each peptide from the evidence.txt MaxQuant output file is stored in an indexed table. Similarly, MS1 peak information is extracted from the msmsScans.txt file and stored in a second index. Each peptide from the first index is then associated to its corresponding peak in the second one, in which the search of corresponding heavy/light counterpart is performed. This retrieves a counterpart and builds an H/L methyl peptide pair even when one of the two is not MS/MS-sequenced but only detected as a peak in the MS1 spectrum. The doublets search is based on the assumption that heavy and light counterparts of the same modified peptide co-elute from the chromatographic column and differ for a highly specific, unambiguously associated delta mass, which we can measure at high resolution in MS1. We used hmSEEKER to automatically filter the MaxQuant evidence.txt file and remove contaminants and reverse sequences, as well as peptides carrying simultaneously light and heavy modifications. To increase the confidence of our findings, the remaining peptides were further filtered to remove any peptide with Andromeda score < 25 or Andromeda delta score < 12 and any methylation with a PTM localization probability < 0.75. Heavy and light methyl peptide pairs were considered true positive when the difference between calculated and expected mass shift was <2 ppm and the difference between their retention times was <30 s.

Use of hmLINKER to intersect the SILAC methyl proteome with the hmSILAC dataset

Validation of the methylated peptide identified in the SILAC experiments through the hmSILAC identifications was achieved by using hmLINKER, another in-house developed bioinformatic tool that compares the sequences of the peptides in the SILAC dataset to those in the hmSILAC repository (hmSEEKER output). If a match is not found at the sequence level, then the peptide is not immediately discarded, but a second round of match attempt is performed using a 31–amino acid sequence window centered on each modification site.

Assignment of symmetric and asymmetric arginine methylation

To search for symmetric and asymmetric arginine methylations within the hmSILAC peptide sequences, we defined new specifications in MaxQuant with the mass difference and residue specificities corresponding to neutral loss of light monomethylamine and dimethylamine [NH2CH3, 31.0421 Da; NH(CH3)2, 45.0578 Da] and loss of heavy monomethylamine and dimethylamine [NH2(13CD3), 35.0641 Da; NH(13CD3)2, 53.1018 Da] to the following variable modifications: di-methyl-K/R and di-methyl4-K/R. In addition, we added heavy methionine (Met4) and oxidized heavy methionine (OxMet4) as variable modifications. To search for symmetric and asymmetric arginine methylations within the SILAC dataset, we defined new specifications in MaxQuant with the mass difference and residue specificities corresponding to neutral loss of light monomethylamine and dimethylamine [NH2CH3, 31.0421 Da; NH(CH3)2, 45.0578 Da] and loss of heavy monomethylamine and dimethylamine [15NH2CH3, 32.0392 Da; 15NH(CH3)2, 46.0548 Da] to the following variable modification: di-methyl-R. Then, MaxQuant viewer was used for manual inspection of the MS2 spectra of the identified dimethylated peptides.

Motif analysis of R-methyl peptides

Motif analysis of R-methyl sites was performed using the iceLogo web application (28), which allowed the visualization of statistically significant enrichment variations between the set of GSK591-regulated sequences and the unchanging methyl peptides used as a background set. P value threshold was set to 0.05.

Functional analysis of GSK591-responding R-methylated proteins

We used GOrilla to assess whether any GO features are enriched in the GSK591-responding methylated proteins: These were tested against the bulk of the identified methylated proteins for statistically significantly overrepresented GO molecular function, biochemical processes, and cellular component terms (56). The analysis was performed, indicating a P value cutoff of 0.001. We built up the protein network of GSK591-responding methylated proteins using the STRING database (57) and visualized it with Cytoscape (58).

Expression and purification of the PRMT5-MEP50 complex

Full-length human PRMT5 (GenBank accession number NM_001039619.2) and MEP50 (GenBank accession number NM_024102.2) were cloned into the baculovirus transfer vector pFB-LIC-Bse. Baculovirus expression was performed as previously described (59). After coexpression of PRMT5 and MEP50, the Sf9 cells were harvested and the pellets were stored at −80°C until use. Frozen insect cell pellets of PRMT5-MEP50 were resuspended in lysis buffer [50 mM Hepes (pH 7.5), 250 mM NaCl, 10% glycerol, 10 mM imidazole, 0.5 mM tris(2-carboxyethyl)phosphine–HCl (TCEP), 0.1% Triton X-100, 1× protease inhibitors (Roche)]. The cells were lysed by sonication for 5 min and clarified by centrifugation. The supernatant was applied to 2 ml of nickel beads (Qiagen) and incubated for 2 hours. Nickel beads were washed three times with wash buffer [50 mM Hepes (pH 7.5), 250 mM NaCl, 10% glycerol, 30 mM imidazole, 0.5 mM TCEP] and eluted in 5 ml of elution buffer [50 mM Hepes (pH 7.5), 250 mM NaCl, 10% glycerol, 250 mM imidazole, 0.5 mM TCEP]. The eluted protein was run on SDS-PAGE and applied to a HiPrep 26/60 Sephacryl S-300 column that had been equilibrated with gel filtration buffer [25 mM Hepes (pH 7.5), 150 mM NaCl, 10% glycerol, 2 mM DTT]. The fractions containing PRMT5-MEP50 complex were pooled, concentrated, and frozen at −80°C.

Expression and purification of histone proteins

Plasmids encoding human histones H2A, H2B, H3, and H4 were obtained from K. Luger. Plasmids were transformed into Escherichia coli BL21 Star (DE3). Bacteria were cultured in 2× TY medium at 37°C, 200 rpm until the OD600 (optical density at 600 nm) reached 0.4, and the protein expression was induced with isopropyl-β-d-thiogalactopyranoside (IPTG) (Golden Biotechnology) at a final concentration of 0.4 mM for 3 hours. Bacteria were pelleted down by centrifugation and resuspended in wash buffer [50 mM tris-HCl (pH 7.5), 100 mM NaCl, 1 mM benzamidine, 1 mM β-mercaptoethanol]. The histone purification protocol was established as previously described (60). In brief, cell pellets were lysed and inclusion bodies were dissolved in 7 M guanidine HCl. The denatured histones were purified by size exclusion and ion exchange, and refolded in refolding buffer [2 M NaCl, 10 mM tris-HCl (pH 7.5), 1 mM EDTA, 10 mM β-mercaptoethanol]. The refolded histones were further dialyzed to histone storage buffer [300 mM NaCl, 20 mM tris-HCl (pH 7.5), 1 mM EDTA, 1 mM DTT].

Expression and purification of PRMT5 substrates, PRMT1 and PRMT4

Human HNRNPA1, HNRNPH1, HNRNPK, SFPQ, KHDRBS1, and CNBP genes were amplified from 293T complementary DNA and further cloned into the pGEX-6P-1 vector (GE Healthcare Life Sciences). Human PRMT1 and mouse PRMT4 genes were also cloned into the pGEX-6P-1 vector. All of these plasmids were transformed into E. coli BL21 Star (DE3). Bacteria were grown in LB medium at 37°C, 200 rpm until OD600 0.6 and induced with 0.4 mM IPTG at 25°C for 5 hours. Cells were harvested by centrifugation and resuspended in 50 mM tris-HCl (pH 8.0), 150 mM NaCl, and 1 mM DTT and further lysed by sonication. The cell lysate was clarified by centrifugation, and the supernatant was filtered through 0.4-μm syringe filter and applied to 5-ml GSTrap FF column (GE Healthcare) on ÄKTAxpress system for affinity purification. GST-tagged proteins were eluted with 50 mM tris-HCl (pH 8.0) and 10 mM reduced glutathione and frozen at −80°C.

Site-directed mutagenesis and cell culture experiments

Mutant CNBP and hnRNPH1 constructs were generated using a KOD Hot Start DNA Polymerase kit (Merck Millipore) and further confirmed by DNA sequencing. All mutations were carried in the context of full-length CNBP (UniProtKB; P62633) and hnRNPH1 (UniProt; P31943). The full sequence of hnRNPH1 was cloned into pMX-IRES-blasticidin overexpression retroviral vector (Cell Biolabs, #RTV-016, a gift from Kaldis Lab). The mutagenesis of the residues R217K and R224K was generated using a KOD Hot Start DNA Polymerase kit (Merck Millipore) and further confirmed by DNA sequencing. Retroviral particles were produced using Phoenix-AMPHO (ATCC CRL-3213) packaging cell line as previously described (10). Likewise, lentiviral particles for hnRNPH1 knockdown were similarly produced using the plasmid pLKO_TRC005 TRCN0000236086 (Sigma-Aldrich), targeting the 3′ untranslated region sequence TTGCCCTTTGCCACGTTAAAT. HeLa cells were transduced for 48 hours with concentrated virus for hnRNPH1 overexpression and subsequently selected with blasticidin (20 μg/ml) for 48 hours. Cells were then transduced with the lentiviral particles for knockdown. After 48 hours, cells were selected with puromycin (1 μg/ml) for 24 hours and counted with trypan blue after 2 days. The following primers were used for quantitative polymerase chain reaction validation of knockdown and overexpression: endogenous hnRNPH1, AGCAGCATGAGTGGATACGA (forward) and GCTAGACAACCCTCCCAT (reverse); overall hnRNPH1, GTGGAGAGGGATTCGTGGT (forward) and GGTCTGCCTTCTCTGGTGT (reverse). To statistically analyze these knockdown data in fig. S4D, generalized linear models with gamma distribution were performed to test the change in ratios. P values of less than 0.05 were considered to indicate statistical significance. These tests, and all statistical analyses, were performed using SAS software, version 9.4 (SAS Institute, Cary, NC).

Structure and dynamics of the PRMT5-GGRxG complex

Before the molecular dynamics (MD) simulation, PRMT5 was centered and solvated in cubic box using TIP3P water. Then, energy minimization was performed in Gromacs 4.5 (61) using the all-atom Amber99sb FF (62) with steepest descent algorithm, which allows fast convergence to the local energy minima. Next, the equilibration of the system was performed in two steps: 100-ps canonical ensemble (NVT) and 100-ps isothermal-isobaric ensemble (NPT) simulation with harmonic restraints (force constant, 1 kJ mol−1 K−1) applied to the backbone atoms of the protein. Furthermore, the production MD runs were carried out for 100 ns with the NPT ensemble without any restraints.

During all simulations, a 2-fs integration time step was used. The temperature was maintained at 300 K with the V-rescale algorithm, and the pressure was coupled at 1 bar by isotropic pressure coupling using the Parrinello-Rahman algorithm (time constant, 10 ps; isothermal compressibility of water, 4.5 × 10−5 bar−1). Long-range electrostatics was calculated by fourth-order particle mesh Ewald algorithm with a grid spacing of 0.16 nm. All bonds were constrained using a fourth-order P-LINCS algorithm. Electrostatic and van der Waals interactions were cut off at 1 nm, and a dispersion correction was applied to account for it. Periodic boundary conditions were implemented in three dimensions. Initial atomic velocities, before NVT equilibration, were obtained from Maxwell’s distribution at 310 K. Trajectories were analyzed by Gromacs tools, Pymol, VMD, and custom Bash shell and Python ( scripts using NumPy ( and Matplotlib libraries.

In vitro radioactive methylation assay

Recombinant proteins were incubated with 17.3 μM S-adenosyl-l-[methyl-14C]methionine (PerkinElmer) and enzyme in methylation buffer [50 mM tris-HCl (pH 8.0), 20 mM KCl, 5 mM MgCl2, 2 mM DTT, 10% glycerol] at 30°C for 2 hours. The samples were run on SDS-PAGE, dried, and exposed to a phosphor screen for 3 days.


Fig. S1. Profiling of protein expression and arginine methylation in response to GSK591.

Fig. S2. Schematic representation of hmSEEKER workflow.

Fig. S3. PRMT5-MEP50 complex is more active than PRMT5.

Fig. S4. Validation of PRMT5-MEP50 substrates by in vitro radioactive methylation assay on synthetic peptides.

Fig. S5. Substrate specificity of PRMT5 on G24 mutant CNBP proteins.

Fig. S6. MS2 spectrum of the hmSILAC-labeled multimethylated peptide of 40S ribosomal proteins S10, carrying two SDMA.

Data file S1. Neutral loss analysis of hmSILAC dimethyl peptides.

Data file S2. MS2 spectra of all methyl peptides down-regulated by GSK591.

Data file S3. MS2 spectra of all methyl peptides not responding to GSK591.

Data file S4. MS2 spectra of hmSILAC methyl peptides bearing ADMA and SDMA on the same arginine residue.

Table S1. The dynamic R-methyl proteome in response to GSK591 treatment.

Table S2. hmSILAC R-methyl peptides used for the orthogonal validation of the SILAC experiment.

Table S3. Nonredundant lists of SILAC-quantified R-methyl peptides.


Acknowledgments: We thank the SGC (Toronto, Canada) for providing the PRMT5 small-molecule inhibitor GSK591 and the SGC20969 inactive compound. We also thank the Kaldis Lab for providing the pMX-IRES-blasticidin overexpression retroviral vector (Cell Biolabs, #RTV-016). Funding: Research activity in the T.B. laboratory was supported by grants from the Italian Association for Cancer Research (AIRC [15741]), the Italian Ministry of Health (GR-2011-02347880), and CNR-EPIGEN flagship project. D.M. was supported by the CNR-EPIGEN flagship project, the Fondazione Umberto Veronesi (FUV) fellowship, and the Fondazione Istituto Europeo di Oncologia (FIEO) fellowship. L.W. and T.T. were supported by the A*STAR SINGA fellowship. J.B. was supported by the A*STAR returning scholar funds. E.G. acknowledged funding support from the NRF2016-CRP001-103 CRP award and from the RNA Biology Center at the Cancer Science Institute (CSI) of Singapore, National University of Singapore, and funding by the Singapore Ministry of Education’s Tier 3 grants (grant number MOE2014-T3-1-006). Author contributions: D.M. performed the proteomics experiments. J.B., L.W., T.T., H.Z., and M.R.I. performed some of the biochemical experiments. D.M., E.M., and A.C. analyzed the proteomics data. U.G. and R.C.R. performed the crystallographic modeling. U.O. supervised the statistical analysis. D.M., E.G., and T.B. wrote the manuscript. E.G. and T.B. designed the research project, supervised the experiments, and managed the scientific collaboration among the two groups. T.B. focused on the proteomics part, whereas E.G. supervised the biochemical follow-up experiments. All authors read and approved the final manuscript. Competing interests: E.G. has received research funding from Eli Lilly and Prelude Therapeutics, has served as a consultant for Prelude Therapeutics and SKBP, has served on advisory board for LION TCR and Janssen, and is a co-founder of ImmuNOA Pte. Ltd. Data and materials availability: The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (63) partner repository with the dataset identifier PXD009070. All other data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.
View Abstract

Stay Connected to Science Signaling

Navigate This Article