Discovery and Mechanistic Study of a Class of Protein Arginine Methylation Inhibitors

Protein arginine methylation regulates multiple biological processes such as chromatin remodeling and RNA splicing. Malfunction of protein arginine methyltransferases (PRMTs) is correlated with many human diseases. Thus, small molecule inhibitors of protein arginine methylation are of great potential for therapeutic development. Herein, we report a type of compound that blocks PRMT1-mediated arginine methylation at micromolar potency through a unique mechanism. Most of the discovered compounds bear naphthalene and sulfonate groups and are structurally different from typical PRMT substrates, for example, histone H4 and glycine- and arginine-rich sequences. To elucidate the molecular basis of inhibition, we conducted a variety of kinetic and biophysical assays. The combined data reveal that this type of naphthyl-sulfo (NS) molecule directly targets the substrates but not PRMTs for the observed inhibition. We also found that suramin effectively inhibited PRMT1 activity. These findings about novel PRMT inhibitors and their unique inhibition mechanism provide a new way for chemical regulation of protein arginine methylation.

Introduction

Among different post-translational modifications, arginine methylation is catalyzed by protein arginine methyltrans- ferases (PRMTs),a which utilize the cofactor S-adenosyl-L- methionine (AdoMet, SAM) as a methyl donor (Figure 1). Thus far, 11 PRMT members have been identified and are categorized into two major types, type I and type II, according to the substrate and product specificity.1-3 Type I enzymes (e.g., PRMT-1, -2, -3, -4, -6, and -8) catalyze the transfer of the methyl group from AdoMet to one of the terminal nitrogen atoms of the guanidino group of specific arginine residues in a protein substrate, resulting in ω-NG-monomethylarginine (MMA, L-NMMA) and ω-NG,NG-asymmetric dimethylar- ginine (ADMA) products.2,4-7 Type II enzymes (e.g., PRMT 5, 7, and 9) catalyze the formation of MMA and ω-NG,N0G- symmetric dimethylarginines (SDMA).7-10 The catalytic properties of PRMT-10 and -11 remain to be characterized. Protein arginine methylation is involved in a broad spectrum of biological processes, including transcriptional activation and repression, mRNA splicing, nuclear-cytoplasmic shut- tling, DNA repair, and signal transduction.11

In recent years, the significance of PRMTs in human diseases has been been increasingly recognized. As the pre- dominant PRMT member, PRMT1 is likely to be responsible for bulk protein arginine methylation in mammalian cells.Aberrant expression of spliced forms of PRMT1 has been observed in several tumor states, including breast cancer12,13 and colon cancer.14,15 Deregulation of the methylation of histone H4 at R3 (a major cellular substrate of PRMT1) is a suggestive marker of prostate cancer.16 PRMT1 was recently shown to be a component of mixed lineage leukemia (MLL) transcription complex, and the activity of PRMT1 is required for malignant transformation.17 PRMT4 [better known as co- activator-associated arginine methyltransferase 1 (CARM1)] is overexpressed in both aggressive prostate cancer and breast tumor.18,19 The type II PRMT member, PRMT5, is recruited to the promoters of tumor suppressor genes such as ST7 and NM23, and its overexpression was observed in a variety of lymphoma and leukemia cells,1,20 in gastric carcinoma,21 and in immortalized fibroblast cells.22 In cardiovascular disorders, PRMT activity is associated with the up-regulation of serum ADMA, which subsequently blocks NO production and causes many cardiovascular implications such as diabetes and hyper- tension.23-25 Furthermore, a number of viral proteins have been shown to be substrates of PRMTs, such as the herpes simplex virus 1 nuclear regulatory protein ICP27, the hepatitis C virus protein NS3, the Epstein-Barr virus nuclear antigen 2, adenovirus E1BAP5 and L4-100K, the HIV-1 proteins Rev and Tat, and the nucleocapsid protein.6,11 Together, these multiple lines of evidence point toward the extensive roles of PRMTs in human pathogenesis and suggest that PRMT inhi- bitors could be very useful research tools and have pharma- cological merits for disease intervention.

Given the essential roles that PRMTs play in normal biology and in disease, quite a few efforts have been invested in developing small molecule PRMT inhibitors both as che- mical genetic tools and as therapeutic agents.26-28 The first type of chemical inhibitors for PRMTs are analogues of the AdoMet cofactor, including S-adenosylhomocysteine (SAH), methylthioadenosine, and sinefungin.27 Because of structural AdoMet-consuming methyltransferases, such as DNA methyl-transferases, protein lysine methyltransferases, and O-methyl- transferases. In the past few years, several groups reported their work in a row on developing small molecule inhibitors specific for PRMTs. Most of these studies on PRMT inhibitor discovery adopted targeted or random screening approaches. The screening strategy is so widely accepted largely because of technical advancement in computer-aided drug design and in medium- and high-throughput inhibitor screening. Thus far, the disclosed small molecule PRMT inhibitors include 1 (AMI-1),29 stilbamidine and allantodapsone,30 the thioglycolic amide RM65,31 pyrazole amide compounds,32 and others.33-37 It is important to point out that past inhibitor development efforts were mostly focused on inhibition of the enzyme, presumably, through binding to the active site. This is also the case with the vast majority of enzyme inhibition work reported related to other enzymes so far. Because a single PRMT catalyzes the methylation of arginine residues in multiple proteins, inhibition of a particular PRMT inevitably results in bulky inhibition of methylations of a large number of proteins but not on a particular substrate sequence. Herein, we report our discovery of a class of inhibitors that modulate PRMT-mediated reaction through binding to the substrates, thus revealing a new way of chemical modulation of PRMT activities.

Results

Screening for New PRMT1 Inhibitors. In an effort to dis- cover potent and selective PRMT1 inhibitors, we conducted a virtual screening to search for novel PRMT1 inhibitors from the ChemBridge small molecule compound collection (about 0.4 million compounds) using the reported crystal structure of rat PRMT1.38 Briefly, the two-dimensional (2D) structures of individual compounds were first converted into three-dimensional (3D) structures by using the CONCORD program39 and then docked to the PRMT1 structure (PDB entry: 1OR8) using the DOCK 6 program.40 From this virtual screening, 50 compounds with high consensus scores were selected and experimentally tested for PRMT1 inhibi- tion (see Table SI-1 in the Supporting Information). In a typical radioactive experimental assay, a reaction mixture contained 0.1 μM recombinant His6x-PRMT1, 5 μM [14C]- labeled AdoMet, and 2 μM the amino-terminal tail peptide of histone H4, that is, H4(1-20), as the substrate, and was incubated at 30 °C in the presence or absence of 100 μM indi- vidual compounds. The retained fractional activity of PRMT1 was used as a parameter to evaluate the potency of the compounds in blocking PRMT1-mediated methylation. It is notable to mention that, for all of the enzymatic assays, the methylation reaction is maintained under initial conditions so that the reaction yields of the limiting substrates are lower than 10%, which is to ensure that the concentrations of AdoMet and peptide substrate do not decrease significantly over the time course of methylation reaction. The experi- mental assays, unfortunately, showed that the accuracy of the virtual screening was rather poor. Out of the 50 tested compounds, only one weak hit, 2 (#5252870, ChemBridge product ID), was found to show inhibition potency at an IC50 value of about 1 mM. Subsequently, we searched for structural analogues of 2 from the ChemBridge small mole- cule collection to examine if more potent inhibitors could be found. Thirty-one compounds that bear similar structures or functional groups with 2 were selected, and radioactive methylation assays were performed to evaluate this second set of compounds for inhibition of PRMT1 (Figure 2 and Figure SI-1 in the Supporting Information). From the test, nine compounds were identified to have inhibition activity against PRMT1, and their IC50 values ranged from 12 to 867 μM (Table 1). Interestingly, we noticed that some of the tested inhibitors bear great structural similarity to 1, a previously reported PRMT1 inhibitor by Bedford’s group.29 All of these compounds have rigid, planar, conjugated systems and contain one or more naphthalene aromatic rings. It is also recognized that most of them have negatively charged sulfonate groups and polar hydroxyl groups. Here- in, we name these compounds as 3 [naphthalene-sulfo deri- vative 1 (NS-1)], 4 (NS-2), 5 (NS-3), and so on because of their characteristics of possessing naphthalene and sulfo groups. We also compared the inhibition potencies of these inhibitors with several PRMT1 inhibitors recently reported in the literature, including 1, stilbamidine, and allantodap- sone. As shown in Table 1, stilbamidine (IC50 = 105.7 μM) exhibited comparable inhibition activity with 1 (IC50 = 137.1 μM). The inhibition potency of the other PRMT1 inhibitors, including allantodapsone and compounds 5756663 and 7280948 reported in the literature,30,33 is even weaker. For example, at a 2 mM concentration of these three compounds, still 45%, 87%, and 100% of PRMT1 activity retains, respec- tively. By contrast, the NS series of compounds identified here are quite stronger PRMT1 inhibitors. Four of these inhibitors, that is, 3, 4, 5, and 6 (NS-4), showed stronger potency than 1 and stilbamidine. In particular, 3 exhibited the best inhibition potency with an IC50 value of 12.7 μM, which is about 10-fold lower than that of 1 and stilbamidine under the same reaction conditions.

Kinetic Pattern of PRMT1 Inhibition by 3. To provide the biochemical basis of PRMT inhibition by this type of NS compound, we investigated the inhibition mechanism of 3, the most potent inhibitor in the series. First, we performed a fluorescence anisotropy binding assay to check if the inhi- bitor can compete with the PRMT1 substrates. In this experiment, a fluorescein-labeled amino-terminal H4 peptide, namely, H4(1-20)FL, with the sequence of Ac-SGRGKG- GKGDpr(FL)GKGGAKRHRK, was used as a substrate ligand for PRMT1 binding.41 The anisotropy of H4(1-20)FL increased upon binding to PRMT1, due to formation of a large macromolecular PRMT1-ligand complex (Figure 3). A Kd value of 0.49 ( 0.10 μM was calculated. The addition of 3 led to reversal of the anisotropy change, thus offering direct evidence that 3 is a competitive inhibitor versus the peptide substrate. A Ki value of 3 was deduced to be 1.71 ( 0.54 μM by fitting the titration data using the DynaFit program.42 Similar competitive binding between 3 and a fluorescently labeled glycine- and arginine-rich (GAR) substrate, R4FL, was also observed (Figure SI-2 in the Supporting In- formation). To further validate the results of competitive inhibition, we conducted steady-state kinetic characterization. The initial velocities of PRMT1 were measured at several selected concentrations of the inhibitor over a range of varied concentrations of one substrate while fixing the concentra- tion of the other. The data were plotted in the double reciprocal fromat with 1/velocity versus 1/concentration of the varied substrate (Figure 4). The kinetic inhibition data points were analyzed by fitting to the linear competitive or noncompetitive inhibition equations.43 As can be seen from the double-reciprocal plots, a series of straight lines inter- sected on the 1/velocity ordinate when the concentrations of H4(1-20) are varied, while the intersecting point moved to the western side of the ordinate when concentrations of AdoMet are varied. These data clearly demonstrate that 3 is competitive versus the peptide substrate and noncompeti- tive versus the methyl donor. This result is consistent with the fluorescent binding assay and is also in agreement with our previous report showing that 1 is competitive versus peptide substrates and noncompetitive versus AdoMet.41 In addi- tion, 6, another inhibitor in this class, also exhibited the same inhibition pattern (Figures SI-3 and SI-4 in the Supporting Information). These combined results support that these types of naphthalene-sulfo derivatives target PRMT1 by blocking the access of the substrate to the PRMT1 active site. Interestingly, the positively charged stilbamidine also showed a competitive binding pattern with respect to the peptide substrate (Figure SI-4 in the Supporting In- formation), which is consistent with a previous report.30

Selectivity of 3. It is of interest to examine whether the NS inhibitors target PRMT1 selectively or also inhibit other PRMT members. Using radioactive methylation assays, we measured inhibitory activities of 3, the most potent com- pound in this series, toward PRMT3, PRMT4/CARM1, and PRMT6 (Table 2). The inhibition potency of 3 for glu- tathione S-transferase (GST)-tagged PRMT1 is similar to that for His6x-PRMT1, with an IC50 of 12.9 ( 0.2 μM. PRMT3 was inhibited at a lower IC50 at 7.1 μM. At the concentration of 500 μM, 3 showed no apparent inhibitory effect to CARM1 (5 μM CARM1, 1 mM H3(1-31), and 30 μM SAM). The IC50 for PRMT6 is about 3-fold larger than that for PRMT1. It is worth mentioning that PRMT1 and PRMT3 share similar substrate specificity, both of which methylate H4 and GAR peptides. On the other hand, CARM1 exhibits distinct substrate specificity from PRMT1. CARM1 targets H3 but does not methylate GAR sequences.6 Therefore, the variation in inhibition potency is likely caused either by the differences between CARM1 and PRMT1 structures or by the distinct nature of the substrates used. Overall, 3 inhibits PRMT1 and PRMT3 stronger than CARM1 and PRMT6.

Compound 3 Directly Targets the Substrates but Not PRMT1. One paradox question is that the NS series of inhibitors bear none or little structural similarity to the arginine-containing substrates of PRMT1 such as histone H4 or GAR peptides. In particular, all of the reported PRMT1 substrates are rich in positive residues, that is, arginines and/or lysines. The crystal structure of PRMT1 also reveals several large acidic grooves present at the protein surface, which have been proposed to participate in the recognition of substrates.38 On the other hand, most of the NS compounds are negatively charged due to the existence of one or more sulfonate groups. Therefore, it seems difficult to envision that 3 and its analogues will bind to PRMT1 at the same site as that of substrates to explain the competitive nature between the inhibitor and the peptide substrates. To look into these paradox problems, we hypothesized that 3 and its analogue inhibitors might bind to the PRMT1 substrate directly and that the binding subsequently prevents the substrate from accessing to PRMT1. Under this scenario, the inhibitor-substrate interaction is likely facilitated by the electrostatic interaction between the negatively charged sul- fonate groups and the cationic arginine guanidino group, as well as the van der Waals interaction between the naphtha- lene ring and the hydrophobic side chains in the substrate. It is conceivable that such an interaction may shield the key motifs in the substrate so that they are not recognized by enzyme. To nail down the mechanistic details, we conducted several biophysical measurements to detect inhibitor- substrate and inhibitor-enzyme associations.

The interaction between the inhibitors and the PRMT1 substrate was confirmed by several experiments. First, when mixing high concentrations of H4(1-20) and 3 or 1 (e.g., 100 μM peptide and 200 μM inhibitor), a red or brown precipi- tate occurred immediately. We reasoned that the precipita- tion is a strong indication of the association between the H4 substrate and the inhibitors. The precipitate was pelleted by centrifugation, washed with water, dissolved in 5% TFA, and analyzed with matrix-assisted laser desorption/ioniza- tion mass spectrometry (MALDI-MS). Indeed, the mass spectra revealed a peak corresponding to the H4(1-20)-1 complex (Figure 5). We could not observe a peak for the H4(1-20)-3 complex on MS spectra, probably due to the instability of the complex and/or the strong anionic nature of 3.

Next, we examined UV-vis spectra of 3 in the presence of different concentrations of H4(1-20) or PRMT1 (Figure 6). Compound 3 has a maximum absorption at 498 nm with an extinction coefficient of 0.0182 μM-1 cm-1. When H4(1-20) was gradually added to a solution of 3 (the concentration of 3 was kept constant at 40 μM), the absorbance decreased dramatically. For instance, at 10 μM H4(1-20), the absorbance at 498 nm decreased to 25% of the original value. By con- trast, the presence of PRMT1 had a quite minor effect on the absorption of 3. At 10 μM PRMT1, the absorbance of 3 at 498 nm still retained 83%. These data directly pointed out that the interaction between 3 and H4(1-20) is much stron- ger than the 3-PRMT1 interaction (if there is any).

We then measured the fluorescence emission of H4- (1-20)FL in the presence of 3. Because the absorption band of 3 overlaps with the fluorescence emission peak of H4- (1-20)FL, it is anticipated that fluorescence energy transfer from fluorescein (donor) to the 3 (acceptor) will occur if the two molecules form a complex. Indeed, the addition of the inhibitor quenched the fluorescence emission of H4(1-20)FL (Figure 7). On the other hand, in the control experiment, the addition of the inhibitor to a fluorescein solution caused little change to its fluorescence spectra (after removing the inner filter effect). These data demonstrate that the interaction between the inhibitor and the H4(1-20)FL depends on the H4 peptide itself, instead of being caused by the attached fluorescein label. The Kd of 3 from the fluorescence binding measurement is 2.53 ( 0.64 μM, which is in a similar range as the value measured from the fluorescence anisotropy titra- tion. Furthermore, we determined the binding stoichiometry of 3 with H4(1-20)FL by using Job’s method (Figure 8). The intersecting point appears at 0.5 in the plot, suggesting that 3 binds to H4(1-20)FL at a 1:1 ratio.

We also attempted to study the 3-H4(1-20) interaction using circular dichroism (CD) spectra. In the spectra, H4(1- 20) showed a pattern of absorption that is correspondent with random coil structures (Figure SI-5 in the Supporting Information). Upon addition of 3, the absorption band for the random coil decreased, indicating that the inhibitor caused a structural change to the H4 peptide. However, upon continuous addition of 3, a red precipitate occurred that prevented further accurate quantitative analysis of the H4(1-20)-inhibitor interaction. Similar results were also observed for the interaction between 1 and H4(1-20) (data not shown).

In enzyme inhibition, if an inhibitor targets the active site of an enzyme, its inhibition potency will be largely deter- mined by the binding affinity between the inhibitor and the enzyme, with little interference by substrates. In our case, however, the experimental findings show that NS com- pounds target the substrate but not the enzyme. This suggests that the potency of these inhibitors in blocking PRMT1 activity will be greatly affected by the nature of the substrate used for the assay. To investigate whether this is true, we tested PRMT1-mediated methylation on H4(1-11) and the inhibition by 3. Indeed, 3 inhibited the methylation of H4(1-11) with an IC50 (i.e., 176.5 μM) significantly higher than that of H4(1-20) methylation (i.e., 12.7 μM). These data indicate that the inhibition caused by the NS molecules depends on the structural sequences of the substrates used and strongly validate the proposal that these inhibitors target substrate instead of enzyme. On the other hand, stilbamidine inhibited the methylation of H4(1-20) and H4(1-11) with similar Ki values, that is, 25.6 and 35.0 μM, respectively. Therefore, the inhibition mode of stilbamidine is different from that of NS compounds. Our data, in agree- ment with a previous study,30 support that stilbamidine targets the active site of PRMT1 to compete with substrates.
By now, there is no structural information available regarding the nature of the H4(1-20)-3 interaction. The potency difference of 3 in inhibiting H4(1-20) methylation versus H4(1-11) methylation suggests that the inhibitor targets both the N-terminal and the C-terminal residues in H4(1-20). The 1:1 stoichiometry of H4(1-20)-3 binding indicates that the interaction is quite specific. Electrostatic interaction is likely a strong factor, but other modes of interactions such as van der Waals and hydrogen bonding may also play significant roles. On the basis of the results of CD, fluorescence, and absorption spectral changes, it is quite possible that the secondary structure of the peptide substrate is altered upon inhibitor binding.

Compound 3 Inhibits the Activity of Histone Acetyltrans- ferase (HAT) p300. The finding that 3 and its analogues inhibit PRMT1 activity by targeting its substrates implicates that 3 may also inhibit other enzymes that utilize H4 or GAR peptides as substrate. To test this possibility, we measured the acetylation of H4(1-20) by p300, a well-known HAT that is able to acetylate the N-terminal tail of H4 at multiple sites.44 The measurement was carried out with 0.02 μM p300, 10 μM [14C]-acetyl CoA, and 10 μM H4(1-20), at varied concentrations of 3 or 1. Indeed, both compounds exhibited strong inhibition of p300-mediated H4 acetylation (Figure 9). The IC50 values of 3 and 1 for p300 inhibition were determined to be 21.3 ( 2.0 and 118.5 ( 6.5 μM, respectively. Notably, these IC50 values are very similar to those obtained for PRMT1 inhibition. These data again support our conclusion that the NS series of compounds binds to H4 directly, and the binding subsequently prevents H4 from being recognized by the H4-modifying enzymes, such as PRMT1 and p300. Thus, the inhibition caused by NS series compounds depends on the binding between the inhibitor and the substrate, irre- spective of enzyme targets. On the other hand, although stilbamidine is a competitive inhibitor in PRMT1-catalyzed methylation, it does not inhibit significantly H4 acetylation catalyzed by p300 (Figure 9d). For example, at 1 mM stilbamidine concentration, p300 still retained 84% of its HAT activity. Therefore, quite likely, stilbamidine targets the active site of PRMT1 to compete with the H4 substrate, which is in agreement with a previous study.30
Suramin Inhibits Arginine Methylation. The NS series of compounds has great structural similarity to suramin, a sym- metrical polyanionic aromatic urea containing two naphtha- lene and six sulfonate groups (Figure 1). We tested whether suramin could inhibit PRMT1-mediated methylation. In- deed, suramin strongly inhibited arginine methylation on both H4(1-20) and GAR substrate R4, with IC50 values of 5.3 and 1011 μM, respectively (Table 1). As a matter of fact, the inhibition potency of suramin is even stronger than that of 3, the most potent inhibitor identified from our screening search. Furthermore, suramin also inhibited the HAT activ- ity of p300 with an IC50 value of 13.7 μM (Figure 9c). Given their structural analogy and the similar properties in the inhibition of methylation and acetylation, suramin likely shares the same mechanism of inhibition with NS com- pounds, namely, suramin also targets H4 and GAR proteins and blocks PTMs on the substrates. These data revealed a previously unknown function of sumarin.

Discussion

The development of PRMT inhibitors represents an active research endeavor in the field of epigenetics. Because PRMT- mediated arginine methylation regulates nucleosomal remo- deling, gene expression, DNA repair, RNA processing and shuttling, and other cellular processes, PRMT inhibitors will be useful tools to dissect the function of protein arginine methylation and the pathways in which PRMTs participate. In this work, we initially attempted to discover PRMT1 inhibitors by using the approach of virtual screening to dock individual small molecules from the ChemBridge compound collection to the crystal structure of rat PRMT1, but the experimental tests only yielded one millimolar-potent inhibi- tor (i.e., 2) out of the 50 compounds that were predicted from the virtual screening. The reason for such a low efficiency of virtual screening is not clear at this time. One possible reason might be that the structure of PRMT1 used in the virtual screening was obtained under nonphysiological conditions (pH 4.7) and the structural conformation of PRMT1 may be different from that at pH 8.0 under which the biochemical assays were carried out. In this respect, better success rates have been observed by Jung and co-workers who performed virtual screening of PRMT1 inhibitors using homology hPRMT1 and A. nidulans RmtA models that were generated on the basis of the rat PRMT3 X-ray structure, which was more physiologically relevant.30,33

By searching and testing structural analogues of the weak hit 2, we identified several naphthalene-sulfo-containing PRMT1 inhibitors with potency in the micromolar range. Four of them, that is, 3, 4, 5, and 6, exhibited stronger potency than 1 and stilbamidine, two effective PRMT1 inhibitors reported previously. Compound 3 has the strongest potency, with an IC50 value of 12.7 μM in our assay using H4(1-20) as a substrate. It is notable that the IC50 values of inhibitors may vary depending on different experimental conditions used. For instance, the IC50 values of 1 and stilbamidine were measured to be 1.2 ( 0.5 and 56.9 ( 6.2 μM, respectively, in a previous study.30 In our inhibition assays, these values were determined as 137.1 ( 12.1 and 105.7 ( 0.7 μM (Table 1). The differences reflected the influence of assay conditions used. It is well-known that IC50 is a relative parameter for the evalua- tion of the inhibition potency of enzyme inhibitors, and its quantitative value can be affected dramatically by experimen- tal parameters such as assay formats, signal readouts, enzyme resources, the nature and concentration of substrates, reaction time, temperature, etc. It is worth pointing out that all of our measurements were carried out using simple radioactive methylation assays or fluorescent binding tests. The enzy- matic processes were analyzed under initial condition so that the change of substrate concentrations and the effect of product inhibition were negligible.

Our fluorescence binding and enzymatic methylation data showed that the tested NS compounds inhibited PRMT activity in a competitive manner against the peptide sub- strates, that is, H4(1-20) and the GAR peptide R4. This result is in agreement with our previous report showing that 1 competes with peptides but not with AdoMet in PRMT1 inhibition.41 Importantly, our experimental data, which in- clude steady-state kinetics, fluorescence anisotropy and inte- nsity titrations, MS, and UV-vis, strongly support that these NS compounds directly target the substrates but not the enzyme to achieve the observed competitive inhibition effect. Because the structures of the tested inhibitors all contain naphthalene and sulfonyl groups, these two motifs likely are the key components of the pharmacophores for this type of PRMT inhibitors. Our data showed that 3 and 1 inhibited the activity of PRMT1 regardless of H4(1-20) or R4 being used as substrates but with different potencies. This provides additional evidence supporting that these compounds target PRMT1 substrates, instead of the enzyme. The result is also in accord with a previous study showing that the methylation of Npl3, a protein rich in GAR motifs, was inhibited by 1.29

In our experiment, it was noted that with the same H3(1-31) peptide substrate, PRMT6 was inhibited to a larger degree than CARM1 in the presence of 3. Given that PRMT6 methylates R2 while CARM1 predominately methylates R17 of H3, it is likely that the difference in inhibition potency is related to the substrate site specificity of the two PRMT members; namely, 3 targets the amino acid residues that are adjacent to arginine 2 but relatively distal from arginine 17 in the substrate H3(1-31). In addition to histones, PRMT1 methylates a variety of proteins at GAR regions. In our experiments, we designed a typical GAR peptide, R4, and tested its methylation by PRMT1 as well as inhibition by the NS series of inhibitors. From Table 1, it can be seen that the inhibition potency of NS compounds toward GAR peptide

methylation is generally weaker than that of H4 methylation. The difference in inhibition potency caused by substrates further supports the proposal that the naphthalene-sulfo molecules target substrates but not enzymes for inhibition. The weaker potency of NS compounds in inhibiting R4 methylation is likely because multiple methylating arginine residues exist in the R4 sequence and binding of an inhibitor is not sufficient to block the methylation at the other sites. Thus far, no structural information is available about the molecular basis of NS-substrate interactions. It is possible that the anionic sulfonate substituents create electrostatic and hydro- gen-bonding interactions with the guanidino nitrogens of the target arginine residues in PRMT1 substrates, and the naph- thalene ring interacts with the arginine side chain through π-stacking and van der Waals contact. The other functional moieties of the inhibitors may fine-tune and further enhance the inhibition by interacting with peptide backbones and other residues in the substrates. It is worth noting that the inhibitor- substrate assoication seems quite specific and is not exclu- sively determined by electrostatic factors. For example, 9 (NS-7) and 10 (NS-8) contain two and three sulfonates, but their inhibition potencies are weaker than those of 6, which contains only one sulfonate group. The Job’s plot shows that 3 binds to H4(1-20)FL with a clear 1:1 stoichiometry, suggesting that the binding is quite specific. An effort is in progress to obtain cocrystal structures of 3 with H4(1-20) and GAR peptides. On the other hand, although demonstrated to inhibit PRMT1-mediated H4 methylation, stilbamidine did not in- hibit H4 acetylation catalyzed by p300. These distinct inhibi- tory properties suggest that the molecular basis of stilbami- dine for PRMT1 inhibition differs from that of 3 and 1; that is to say, stilbamidine directly targets the active site of PRMT1, but the NS series of compounds targets the substrates of PRMT1. From a structural point of view, the competitive nature of stilbamidine and PRMT1 substrates seems reason- able because the cationic formamidine group in stilbamidine is analogous to the guanidino group in arginine residues.

The mechanism by which the NS series of inhibitors (including 1) target the substrates rather than PRMT impli- cates that they may have unique applications in comparison to classic enzyme-targeting inhibitors. Previously, 1 has been shown in cellular assays to block PRMT1-mediated protein methylation and transactivation of luciferase reporter genes.29 Because the N-terminal tail of H4 undergoes a variety of post- translational modifications, for example, phosphorylation at the S1 site,45,46 arginine methylation at the R3 site,47 lysine acetylation at the K5, K8, K12, and K16,48-52 and lysine methylation at the K20 site, etc.,53-55 it is conceivable that the binding of the amino-terminal tail of H4 to 3, 1, or other NS series of compounds will influence many modifications occur- ring on the N-terminal tail of H4 in addition to arginine methylation, although the degree of influence on individual modifications may vary. Our data about the inhibition of p300- catalyzed H4 acetylation by both 3 and 1 clearly substantiate this potential. Interestingly, during the preparation of this paper, we noticed that Kundu and co-workers reported that TBBD (ellagic acid), a chemical component from the pome- granate extract, inhibited methyltransferase activity of PRMT4/CARM1 also in the substrate-targeting manner.56 TBBD was determined to target histone H3 near the arginine 17 site and has been used to elucidate the significance of H3R17 methylation in the p53-responsive gene expression pathway.

Organic compounds that contain naphthyl and sulfonate groups have been previously reported to have multiple biological impacts. For example, 1 and several analogue molecules that contain naphthyl, urea, and sulfonate groups were shown to inhibit the activity of HIV-1 reverse transcriptase (RT) by preventing the binding of DNA substrates to the enzyme, and structural similarities of those compounds to DNA base pairs were also conjectured therein.57 Suramin, a symmetrical polyanionic aromatic urea, also contains naphthalene-sulfo- nate groups. Suramin is an antiparasitic drug but also showed efficacy for the treatment of several kinds of carcinomas.58,59 Biochemically, suramin binds to and inhibits a variety of enzymes (e.g., thrombin,60 vaccinia virus complement control protein,61 and NAD+-dependent deacetylase SIRT562) and peptidic growth factors (e.g., epidermal growth factor63 and fibroblast growth factor64) (see ref 58 for a recent review). Our data clearly showed that suramin inhibits the activity of PRMT1 on both H4 and GAR substrates, thus disclosing a possible new biological function of suramin and its deriva- tives. Also, it would be intriguing to investigate whether the NS series of compounds studied here also have similarly broad biological effects.

One minor concern is that the NS compounds are nega- tively charged and may have low membrane permeability. This could limit their biological applications for PRMT functional study. It will be necessary to investigate how efficiently these NS compounds cross plasma membranes. Nevertheless, it is of note that many anionic compounds, such as 1 (a structural analogue in the NS series),29 ellagic acid,56 and anarcardic acid,65 have been shown to be cell permeable and have intracellular inhibition activities. Also, the issue of cell permeability could be improved by adding drug delivery vehicles or chemically modifying the compounds in pro-drug formats. In addition to in vivo applications, the NS inhibitors are valuable chemical probes for the mechanistic study of PRMTs and other histone-modifying enzymes in biochemical systems. For instance, they can be used to investigate sub- strate-enzyme interactions and downstream effector mod- ules that recognize arginine methylation marks in histones.

Conclusion

In conclusion, we have discovered a type of organic com- pounds containing naphthalene and sulfonyl pharmacophore components that inhibit PRMT activity in the micromolar range, whose inhibition mechanism is fundamentally distinct from the other PRMT inhibitors reported so far. The bio- chemical and biophysical data of representative compounds (e.g., 3, 6, and 1) show that these inhibitors are competitive versus PRMT1 substrates (e.g., H4 and GAR peptides) and noncompetitive versus the methyl donor. Detailed studies illustrate thatthey directly targetthe peptide substrates instead of PRMT1, and the binding subsequently blocks the recogni- tion of the substrates by enzyme, which is largely responsible for the observed PRMT1 inhibition effect. We also show that the antiparasitic drug suramin is also an effective arginine methylation inhibitor. These NS inhibitors will be useful chemical tools for the mechanistic study of arginine methyla- tion and other epigenetic modifications. Furthermore, illumi- nation of the inhibitory mechanism provides a new insight for understanding the pharmacological effect of these structurally unique molecules in biological systems.

Experimental Section

Materials. Fmoc-protected amino acids and solid phase resins were purchased from NovaBiochem. All of the screening compounds were obtained from ChemBridge Corp. Stilbami- dine and allantodapsone were obtained from the National Cancer Institute. Suramin was purchased from Acros Organics. For all of the effective inhibitors discovered and tested, including compounds 1, 3-11, suramin, and stilbamidine, their purities were confirmed to be g95% by analytical C18 reverse-phased high-performance liquid chromatography (HPLC). Radioactive AdoMet and acetyl-CoA were ordered from GE Healthcare and Perkin-Elmer. Other chemical reagents were purchased from Fisher, VWR, and Sigma, etc. Peptide Substrates. Peptides were synthesized using the stan- dard solid phase peptide synthesis (SPPS) protocols, purified with C-18 reversed phase HPLC, and confirmed with MALDI- MS as previously described.66 The sequence of the NH2-term- inal 20 aa peptide of histone H4, H4(1-20), is Ac-SGRGKG- GKGLGKGGAKRHRK. The sequence of the NH2-terminal
11 aa peptide of histone H4, H4(1-11), is Ac-SGRGKGGKGLG. The sequence of the GAR peptide, R4, is Ac-GG- RGGFGGRGGKGGRGGFGGRGGFG. The underlined Rs designate the methylation sites. The NH2-terminal tail 31 aa peptide of human H3, H3(1-31), was synthesized and used as a substrate in CARM1 and PRMT6 analyses. The sequence of H4(1-20)FL is Ac-SGRGKGGKGDpr(FL)GKGGAKRHRK.

Radioactive Methylation Assay. The inhibitory activities of small molecule compounds were tested using carbon 14-labeled radioactive methylation assays. The assays were carried out in 0.6 mL plastic tubes at 30 °C in a reaction volume of 30 μL. The reaction buffer contained 50 mM HEPES (pH 8.0), 0.5 mM dithiothreitol (DTT), 1 mM EDTA, and 50 mM NaCl. In a typical procedure, 2 μM peptide substrate, 5 μM [14C]-AdoMet, and varied concentrations of an inhibitor were preincubated in the reaction buffer for 5 min prior to the initiation by the addition of PRMT1 (0.1 μM final). After it was incubated for an appropriate period of time, the reaction was quenched by spreading the reaction mixture onto P81 filter paper discs (Whatman). The paper disk was washed with 1 L of 50 mM NaHCO3, and dried in air for 2 h. The amount of methylated products was quantified by liquid scintillation. The IC50 value is the concentration of inhibitor at which half of the maximal activity is reached. The Ki for stilbamidine was calculated from IC50 by using the equation: Ki = IC50/(1 + [S]/Km). Km was obtained by measuring the initial velocity of reaction at different concentrations of a particular substrate and fitting the kinetic data with the Michaelis-Menten equation. The inhibition patterns of 3 and 6 were determined by measuring initial velocities of PRMT1 at a range of varied concentrations of one substrate, a fixed concentration of the other substrate, and selected concentrations of the inhibitors. The data were dis- played in double reciprocal formats and fitted to competitive or noncompetitive kinetic equations.82

Inhibition of p300 Catalysis. Recombinant p300 HAT domain (1287-1666) was a gift from Dr. Philip Cole at Johns Hopkins University, and its expression was described in an earlier report.83 The enzymatic activity of p300 and its inhibition by 3 and other compounds were measured by radioactive acetylation assays. A reaction mixture of 10 μM H4(1-20), 10 μM [14C]- acetyl CoA, 20 nM p300, and increasing concentrations of the inhibitors were incubated in the reaction buffer [50 mM HEPES (pH 8.0), 50 mM NaCl, 0.5 mM DTT, and 1 mM EDTA] at 30 °C for 10 min, and the reaction was quenched by loading the mixture onto p81 filter paper. The radioactive products were quantified by liquid scintillation, and the fractional activity of p300 was plotted with respect to the concentration of individual inhibitors.

Fluorescent Binding Assay. The fluorescence intensity and anisotropy of fluorescein-labeled peptides were measured on a Fluoromax-4 spectrofluorometer (Horiba Jobin Yvon). The buffer was the same as that for the radioactive assay. The excitation wavelength and emission wavelength were selected at 498 and 524 nm, respectively. The competitive binding of small molecule compounds to the PRMT1-substrate solution was measured using the fluorescence anisotropy mode in similar manners as described previously.41 Typically, 0.2 μM H4(1- 20)FL and 2 μM PRMT1 were mixed, and increasing concen- trations of an inhibitor were added until the fluorescence anisotropy signals leveled off. The anisotropy values at 524 nm from several scans were plotted as a function of inhibitor concentration. Data were fitted with a competitive binding model using DynaFit program to calculate the Ki value.42,84 Also, fluorescence intensity changes of H4(1-20)FL at different con- centrations of 3 were measured to detect their interaction. A 0.2 μM concentration of H4(1-20)FL at 30 °C was titrated with increasing concentrations of 3 (0-28.2 μM). NHS-fluorescein was used as a control. The fluorescence intensities of both H4(1-20)FL and NHS-fluorescein were corrected to remove the inner filter effect that was caused by 3 absorption, and the data were plotted as a function of the concentration of 3. Job’s method was applied to determine the binding stoichiometry of 3 with H4(1-20)FL. A series of solutions with a fixed total amount (0.225 nmol) but varied ratios (0-9) of 3 and H4- (1-20)FL in the same reaction buffer were prepared. The fluorescence intensity of each sample at 524 nm was measured, which is related to the amount of binding complex. The fluor- escence intensity was divided by the molar fraction of H4- (1-20)FL and then plotted as a function of the molar fraction of 3.

UV-Vis Spectroscopy of 3 upon Titration with H4(1-20) or PRMT1. The UV-vis spectra of compound 3 (40 μM in the same reaction buffer) were acquired on a Shimadzu UV-1700 spectrophotometer, in the presence of different concentrations of H4-20 or PRMT1.CD Measurement. CD spectral changes of H4(1-20) upon the addition of different concentrations of 3 were measured on a Jasco J-810 spectropolarimeter. A 100 μM concentration of H4(1-20) was titrated with 0, 100, and 200 μM 3 in 10 mM Tris buffer (pH 7.4) in a 400 μL CD cuvette. The CD spectrum of each equilibrated sample was scanned (100 nm/min) with an accumulation of LLY-283 three times (Figure SI-5 in the Supporting Information).