|Home | About | Journals | Submit | Contact Us | Français|
Protein Arginine Methyltransferases (PRMTs) catalyze the posttranslational methylation of arginine using S–adenosyl–methionine (SAM) as a methyl–donor. The PRMT family is widely expressed and has been implicated in biological functions such as RNA splicing, transcriptional control, signal transduction, and DNA repair. Therefore, specific inhibitors of individual PRMTs have potentially significant research and therapeutic value. In particular, PRMT1 is responsible for >85% of arginine methyltransferase activity, but currently available inhibitors of PRMT1 lack specificity, efficacy, and bioavailability. To address this limitation, we developed a high–throughput screening assay for PRMT1 that utilizes a hyper–reactive cysteine within the active–site, which is lacking in almost all other PRMTs. This assay, which monitors the kinetics of the fluorescence polarization signal increase upon PRMT1 labeling by a rhodamine–containing cysteine–reactive probe, successfully identified two novel inhibitors selective for PRMT1 over other SAM–dependent methyltransferases.
Arginine methylation is a posttranslational modification with crucial functions in RNA splicing, transcriptional control, signal transduction, and DNA repair(1-3). Methylation of arginine is catalyzed by a family of enzymes called the Protein Arginine Methyltransferases (PRMTs)(1-3). Arginine methylation of proteins was first discovered in 1968, but only three decades later was the first PRMT family member, PRMT1, cloned and initially characterized (4, 5). Currently, the PRMT family is known to contain at least eleven enzymes that all use S–adenosylmethionine (SAM) as a methyl donor(1-3, 5-13), and are divided into three classes based on methylation activity: Type I PRMTs (PRMT1, 3, 4, 6, and 8) catalyze the formation of asymmetric ω–NG,NG–dimethylarginine (5-9, 14), Type II PRMTs (PRMT5) form symmetric ω–NG,N’G–dimethylarginine(6, 10, 12, 14), and Type III PRMTs (PRMT7) form ω-NG-monomethylarginine(15).
PRMT1, a ubiquitously expressed Type I enzyme, is responsible for over 85% of the arginine methylation occurring in the cell and has a wide range of substrates involved in transcription, chromatin modification, and signal transduction(1-3, 16). PRMT1 has a preference for arginine residues flanked by glycines in what are termed glycine–arginine–rich regions(14, 17). Seven isoforms of PRMT1 have been cloned which contain a conserved enzymatic core but vary in their amino–termini(18). PRMT1 deficient mice are embryonically lethal and derived mouse embryonic fibroblasts (MEFs) lacking PRMT1 contain defects such as DNA damage, polyploidy, and cell cycle progression delays(19). Therefore, a specific inhibitor would be valuable to further study the biological functions of PRMT1.
Many current inhibitors for PRMT1 are SAM analogs, such as sinefungin and S–adenosylhomocysteine (SAH), but they lack specificity and inhibit the activity of all methyltransferases that use SAM as a methyl donor, including lysine methyltransferases and the entire family of PRMTs(7, 8, 12, 20). Additional non–nucleoside compounds, including AMI–1 and AMI–408, have been discovered that retain more specificity to PRMT inhibition over lysine methyltransferases, but they still lack specificity across Type I PRMT members(7, 21). Other indirect inhibitors, including MDL 28,842 and adenosine dialdehyde (Adox) (7, 8, 12, 20), have been developed to block the production of SAM by inhibiting SAH hydrolase, but these compounds are obviously not compatible with evaluating the specific function of PRMT1 in a context where multiple methyltransferases are present. Recently, inhibitors of PRMT1 consisting of modified substrate peptides have been developed (22, 23), although these still cross-react with other PRMTs, especially PRMT6.
Competitive activity–based protein profiling (ABPP) has emerged as a powerful platform to identify inhibitors of enzymes from multiple mechanistic classes, including serine hydrolases, oxidoreductases, kinases, and protein arginine deiminases(24-29). Briefly, activity–based probes typically contain a reactive chemical moiety that covalently labels the active site residues of enzymes, and a tag (e.g., a fluorophore, biotin, or click handle) to enable identification or enrichment of the labeled enzymes (30, 31). Competitive ABPP involves assaying the ability of a compound to block the labeling of an enzyme by an activity–based probe, and is usually monitored using SDS–PAGE gels or mass–spectrometry (32, 33). Unfortunately, these methods are not amenable to screening more than a few hundred compounds (34). However, a recently developed version of competitive ABPP in which the interaction between a fluorescently–labeled probe and an enzyme is monitored by the resulting increase in fluorescence polarization signal (FluoPol–ABPP) is compatible with high–throughput screening (HTS) of large compound libraries (24). Here, we describe the development of a FluoPol–ABPP assay for PRMT1.
Unfortunately, classical activity–based probes do not yet exist for the PRMT enzyme family. Recently, however, Thompson and colleagues reported the use of a fluorescently–conjugated chloroacetamide bearing a histone 4 tail substrate analog as a probe capable of covalently interacting with PRMT1, though the target residue of this molecule is unknown(23). Further, a previous screen for hyper–reactive cysteines in cell lysates identified cysteine 101 (Cys101) in PRMT1(35). Interestingly, the homologous cysteine is absent in all other PRMTs except PRMT8 (Supplemental Figure 1). We hypothesized that this residue could form the basis of a screen involving the use of a “non–directed” activity–based probe that generally reacts with specific types of nucleophilic residues in proteins rather than a particular enzyme class (36-38). Here, we report that a thiol–reactive maleimide probe conjugated to AlexaFluor488 covalently binds to Cys101 of PRMT1 and generates a strong fluorescence polarization signal. We adapted this interaction into a FluoPol–ABPP format and screened the Maybridge Hitfinder Collection for compounds that block the fluorescence polarization signal. Using this screen, we identified several novel and selective inhibitors for PRMT1.
PRMT1 contains a hyper–reactive cysteine (Cys101) within its active site (Figure 1a)(35). Cys101 directly contacts SAM during catalysis of arginine methylation, although is not essential for activity(35). In order to establish an HTS assay to screen for PRMT1 inhibitors, we reasoned that a “non–directed” activity–based probe that specifically labels hyper–reactive cysteine residues in proteins could provide a basis to monitor active PRMT1 protein. To evaluate this premise, we incubated the cysteine–reactive maleimide conjugated to AlexaFluor488 (maleimide–AF488) (5nM) with purified wild–type or cysteine 101 alanine (C101A)–mutant PRMT1 (4μM). After 45 minutes, the reactions were quenched, separated by SDS–PAGE, and quantified by in gel–fluorescence scanning (Figure 1b). The wild–type enzyme encouragingly gave a much stronger fluorescent signal by gel. Specificity of the signal from the maleimide probe binding to PRMT1 was determined by pre–incubating PRMT1 or the C101A mutant (4μM) with unlabeled maleimide prior to addition of the fluorescently–labeled probe (5nM). As predicted, free maleimide competed with labeled maleimide for enzyme binding (Figure 1c).
As an initial test whether this assay could detect PRMT1 inhibition, we added nonspecific methyltransferase inhibitors (100μM) to either PRMT1 or PRMT1–C101A (4μM) prior to incubation with maleimide–AF488. The SAM analog sinefungin reduced the fluorescence signal of only the wild–type enzyme (Figure 1d). 4–Hydroxynonenal (HNE), previously shown to reduce labeling of wild–type but not PRMT1–C101A by a thiol–reactive iodoacetamide probe(35), encouragingly also reduced maleimide–AF488 binding in this assay (Figure 1d). However, the small molecule inhibitor AMI–408 interfered with probe binding to both PRMT1 and PRMT1–C101A (Figure 1d), indicating AMI–408 may interact with PRMT1 independent of C101.
We next monitored the fluorescence polarization signal upon incubation of maleimide–AF488 with the PRMT1 proteins, and again observed a significantly stronger signal with the wild–type enzyme compared with the C101A mutant (Figure 2a). This indicates that the hyper–reactive cysteine 101 is indeed labeled by the probe and that this signal should be adaptable to a FluoPol–ABPP format. However, since maleimide–AF488 reactivity with PRMT1–C101A was still detected by fluorescence polarization (Figure 2a), we next titrated the enzyme to determine an optimal protein concentration to maximize the signal–to–noise ratio. We discovered that the fluorescence polarization signal generated by the interaction of maleimide–AF488 with PRMT1-C101A was reduced to similar levels as detected by probe alone at an enzyme concentration of 0.4μM. Importantly, the wild–type enzyme still gave a robust signal at this concentration (Figure 2a). To determine the assay quality at these concentrations, we next performed reactions using automated liquid handlers and monitored polarization over 2 hours (Figure 2b). We observed that the Z’–factor increases over time and plateaus around 0.8 at 1 h, indicating that the assay is compatible with HTS (Table 1). Further, the assay is largely insensitive to DMSO, as little change in polarization was observed up to concentrations of 2% (Figure 2c). Finally, we observed that sinefungin inhibited the fluorescence polarization increase by wild–type PRMT1 in a dose dependent manner (Figure 2d), thus potential reversible inhibitors would not be out-competed by the covalent binding of the maleimide probe, as has been shown previously in other ABPP screens(24). Therefore, this FluoPol–ABPP assay was deemed to provide a high–quality, simple format to screen for compounds that occlude the active–site of PRMT1.
We applied the FluoPol–ABPP PRMT1 assay to screen the 16,000 compounds in the Maybridge Hitfinder Collection. The activity of the compounds (at 10μM) was determined by the percent of inhibition of PRMT1 polarization compared to DMSO–treated wild–type enzyme (0% inhibition) and PRMT1–C101A (100% inhibition) (Figure 3). 789 (4.9%) compounds screened had greater than 50% inhibition. Although this hit rate number is higher than what is typically observed in high–throughput screens, it is similar to rates observed with enzymes previously known to be sensitive to inhibition by generally thiol–reactive compounds (24). The addition of the detergent F-127 prevents limits false positives that act as aggregators(39, 40). As a counter–screen to identify false positives, PRMT1 HTS results were compared to two screens using the identical maleimide–AF488 probe with different proteins containing active-cysteines to which maleimide-AF488 binds(41). This counter–screen identified false positives, cross–reactive compounds, and/or thiol-reactive compounds, which were discarded from further analysis. The remaining 98 compounds (0.6% of total) were re–tested in triplicate to confirm inhibition. From these compounds, the top 19 inhibitors with greater than 50% inhibition were selected for further screening.
We next tested these 19 compounds using an orthogonal in vitro methylation assay in which a radiolabeled SAM substrate is used for the methylation of PRMT substrates by PRMT1. Of these 19 compounds, all but four at 100μM (2729237, 2745438, 2810812, and 2815040) showed detectable inhibition of PRMT1 activity (Figure 4a and Supporting Information 2b). Four of the top compounds, 2806087, 2811408, 2818500, and 5380390, were selected for further characterization, along with two negative controls, 2797621 and 2729237 (Figure 4b). To confirm their activity in the methylation assay, titration curves these compounds were generated for inhibition of PRMT1. As seen in Figure 4c, the strongest inhibition was observed with 5380390 and 2818500, with apparent IC50 values of 23 μM and 11 μM respectively. The other four compounds tested had limited activity below the 100μM used in Figure 4a. Additionally, as the activity of PRMT1 is reported to be affected by amino-terminal tags (42), we performed titration experiments using an untagged PRMT1 and found similar IC50s (data not shown). The activity of the two hits, 5380390 and 2818500, is unlikely to be due to a covalent interaction between the compounds and the substrate, since Histone4 lacks cysteine residues. Therefore, novel PRMT1 inhibitors were discovered using our PRMT1 HTS assay.
While a determination of their mechanism of action must await further studies, it is interesting to note that four of the five compounds identified contain electrophilic groups and both of the high–affinity inhibitors (5380390 and 2818500) are nitroalkenes. It is therefore expected that these compounds compete with the maleimide probe for modification of Cys101. Consistent with this hypothesis, these two compounds had no effect on the methylation activities of both CARM1, a Type I PRMT family member that does not contain a cysteine homologous to PRMT1 Cys101 in the SAM binding site (Figure 5a), and the SAM–dependent lysine methyltransferase Set7/9 (Figure 5b). Contrariwise, 5380390 and 2818500 did inhibit the activity of PRMT8 as well as PRMT1, the only two Type I PRMTs that contain a SAM–binding cysteine residue. The IC50 values obtained from the dose-response studies of Figure 4c reflect only the apparent inhibitory power of these molecules under the experimental conditions employed. Since they are likely to be covalent inhibitors, their potency can be affected by several factors that we have not yet explored, including their intrinsic electrophilicity, ability to access the Cys101 site, and rates of competitive decomposition by attack of solvent or other nucleophiles. The possibility also remains that these two compounds have the potential to cross–react with other cysteine–dependent proteins outside the PRMT class. The possibility will be further investigated using competitive ABPP on whole cell lysates with cysteine–reactive probes(35).
In conclusion, we have developed a high–throughput fluorescence polarization assay to screen for novel PRMT1 inhibitors based on a specific SAM–binding cysteine absent in other SAM–dependent methyltransferases except PRMT8. Using this screen, we have indentified two novel inhibitors specific for PRMT1 and PRMT8 over other SAM–dependent methyltransferases. We anticipate that these inhibitors may serve as lead compounds to probe cellular functions of PRMT1, and more generally that the assay described herein can be readily adapted to screen much larger compound libraries to identify new lead inhibitors of this important enzyme.
AlexaFluor488–C5–maleimide was purchased from Invitrogen (#A10254). Sinefungin (#567051) and 4–Hydroxynonal (#393204) were purchased from EMD Chemicals. AMI–408 was a generous gift from M.G. Finn and colleagues(21). Recombinant proteins were acquired as follows: CARM1 from Upstate (#14–575), Set7/9 from Enzo Life Sciences (#ALX–201–178), and Histone4 and Histone3.1 from New England Biolabs (#M2504S and #M2503S, respectively).
Full–length PRMT1 cDNA and PRMT1–C101A subcloned into the pET45b(+) vector (Novagen) was a gift from the Cravatt laboratory(35). His–tagged PRMT8 was generated by subcloning full–length human PRMT8 cDNA into pET29a(+) vector (Novagen). For recombinant expression, OverExpress C41(DE3) pLysS Chemically Competent Cells (Lucigen) were transformed with the above constructs and grown on LB media + antibiotic agar plates. Selected colonies were grown in 2X YT media to OD600=0.8 and induced with 0.25mM IPTG for 4 hours at 37°C. Cell pellets were resuspended in Lysis Buffer (50mM NaH2PO4, 300mM NaCl, 10mM imidazole) supplemented with 1mg ml−1 lysozyme (Sigma), sonicated, and centrifuged at 17,000g for 30 minutes. The clarified lysates were collected and incubated overnight with Ni–NTA slurry (Qiagen) at 4°C. The Ni–NTA beads were collected at 1000g for 5 minutes, washed twice in Lysis Buffer + 30mM imidazole for 15 minutes. Protein was eluted with Lysis Buffer + 300mM imidazole 3 times for 1hour at 4°C. PD–10 columns (GE Healthcare) were used to size–exclude contaminants and exchange the buffer to PBS, and protein–containing fractions were concentrated by Amicon centrifugal filter (Millipore). Protein concentration was determined by BCA kit (Pierce). The purification yielded >95% purity (by SDS–PAGE).
In a 384–well Greiner Low Volumes plate (#788076/1V), 10μl of PRMT1 or PRMT1–C101A (0.44μM final unless otherwise specified) in assay buffer (100mM HEPES pH 7.0, 100mM NaCl, 0.05% Pluronic F–127 (Invitrogen)) was added to each well by BioRAPTR FRD. Compounds were then dispensed by Pintool (50nl; final concentration 10μM) to test wells and DMSO to control wells. The plates were incubated at room temperature for 30 minutes. Then 10μl of maleimide–AF488 (5nM final unless otherwise specified) in assay buffer was added to all wells, the plates were centrifuged briefly and incubated at room temperature for another 60 minutes. Plates were then read on an Envision microplate reader (PerkinElmer, Turku, Finland) using a FITC FP filter set (excitation=480nm, emission=535nm) and a FITC Dual enhanced FP dichroic mirror.
Data for the counter-screens came from maleimide-AF488 binding experiments to the protein mitoNEET, which contains reactive-cysteines (P. Jennings and S. Brown, personal communication). Incubation time of library compounds and concentrations of compounds of counter-screen were identical to PRMT1 HTS.
Gel–based ABPP experiments were performed similarly as FluoPol–ABPP assays with the following exceptions: Total reaction volume for each sample was 30μl. After the final room temperature incubation, the reactions were quenched with 5μl 6X SDS–PAGE loading buffer and boiled for 5 minutes at 95°C. Samples were separated by SDS–PAGE and fluorescence was imaged on a flat–bed fluorescence scanner (Hitashi). Subsequently, the gels were stained for 30 minutes with Coomassie Stain (0.25% Brilliant Blue, 50% MeOH, 10% acetic acid) and distained overnight in 50% MeOH, 10% acetic acid. Gels were dried and stored at room temperature.
Recombinant protein (0.4μM unless otherwise stated) and compounds were incubated for 60min at room temperature in 20mM Tris–HCl pH8, 200mM NaCl and 0.4mM EDTA. Subsequently, 1μg of substrate protein (H4 or H3.1) and 2μl S–adenosyl–[methyl–3H]methionine (2μCi) (Perkin Elmer) were added for an additional 60min at room temperature. Reactions were stopped with 6X SDS Loading Buffer, boiled for 5 minutes at 95°C, and separated by SDS–PAGE. Gels were stained by Coomassie as above and subsequently incubated with Amplify (GE Healthcare #NAMP100) for 30 minutes. After drying, gels were exposed to HyBlot CL film at −80°C.
Compounds are from the Maybridge Hitfinder Collection version 4.
P. Jennings, UCSD, for counter-screen data. Assay Development and HTS were supported by grant 5U54MH084512–04 from the NIH to the Scripps Molecular Libraries Screening Center. Other funding sources: MBCD & KAM: NIH/NIAID 5R01AI067460-05 & NIH/NIGMS 5R01GM85117-04; DAB & BFC: NIH CA087660; MGF: NIH/NIAID 1R56 AI090832.
Supporting Information Available: This material is available free of charge via the Internet at http://pubs.acs.org.