Mitogen-activated protein kinase phosphatase 1 (MKP-1) is a tyrosine phosphatase superfamily member that dephosphorylates and inactivates cardinal mitogen-activated protein kinase (MAPK) substrates, such as p38, c-Jun NH2-terminal kinase, and extracellular signal-regulated kinase. Although these MAPK substrates regulate many essential cellular processes associated with human diseases, few pharmacological inhibitors have been described. The lack of readily available selective MKP-1 inhibitors has severely limited interrogation of its biological role and was one rationale for using a recently described tricyclic pyrrole-2-carboxamide library in our screening efforts. In this report we demonstrate the pharmacological richness of the pyrrole carboxamide library by the finding that 10 of 172 members inhibited human MKP-1. Two of the pyrrole carboxamides, PSI2106 and MDF2085, were especially notable in vitro inhibitors of recombinant human MKP-1 enzyme activity with IC50 values of 8.0 ± 0.9 and 8.3 ± 0.8 μM, respectively. Both showed some selectivity for MKP-1 over the closely related phosphatases MKP-3, Cdc25B, VHR, and PTP1B. Computational examination of the surface properties near the catalytic site revealed that the phosphatases studied differ significantly in their electrostatic potential at the substrate binding site. The compounds inhibited MKP-1 reversibly but displayed mixed kinetics. Phosphatase inhibition was retained in the presence of physiologically relevant concentrations of glutathione. Molecular docking studies suggested that PSI2106 may interact with His229 and Phe299 on MKP-1. These results reveal the power of using a small focused library for identifying pharmacological probes.
Many contemporary drug discovery efforts exploit either diversity-oriented chemical libraries, which are usually designed to have maximum differences in the core structure, or focused chemical libraries, which are fundamentally designed to emulate the structural properties of previously defined “drug-like” core skeletons with variable appended moieties (Stockwell, 2004). Both approaches are complementary. The University of Pittsburgh's Center for Chemical Methodologies and Library Development has generated novel chemical libraries containing elements that occupy poorly populated chemical space. One subset that was recently described was based on a tricyclic pyrrole-2-carboxamide scaffold (Werner et al., 2006). This library was of particular pharmacological interest because the pyrrole carboxamide skeleton is unique, has not been well explored, and is a core unit in many marine natural products with experimental antitumor (agelastatin A and yatakemycin), antihistamine (dispacamide A), antiviral (sceptrin), or anti-bacterial (storniamide A) activity (Okano et al., 2006; Werner et al., 2006). In the current study we provide the first documentation of the biological activity of members in this novel pyrrole carboxamide library, which was synthesized using a sequence of three reactions: Pauson-Khand cycloaddition and Stetter and Paal-Knorr condensations (Brummond et al., 2005; Werner et al., 2006).
It is now widely recognized that mammalian cells use mitogen-activated protein kinases (MAPKs) as distal effectors for many extracellular growth factors, stress detectors, and drug sensors. MAPKs are activated by intracellular tyrosine kinases and inactivated by mitogen-activated protein kinase phosphatases (MKPs), which are dual specificity phosphatases (DSPs) capable of dephosphorylating both phosphotyrosine and phosphothreonine on the same protein substrate. The most promiscuous of the 11 human MKPs is MKP-1, which can dephosphorylate p38, c-Jun NH2-terminal kinase and extracellular signaling-related protein kinase (Erk) 1/2 in vitro. Surprisingly, the precise determinants of its substrate specificity remain uncertain (Wu and Bennett, 2005), although it regulates cytokine and mitogen response. The crystal or solution structure for MKP-1 is not available to guide any rational inhibitor design. MKP-1 expression is elevated in prostate, breast, gastric, and renal cancer (Magi-Galluzzi et al., 1997; Liao et al., 2003; Wang et al., 2003) and is correlated with decreased progression-free survival (Denkert et al., 2002). Moreover, a reduction in MKP-1 expression by antisense decreases the tumorigenecity of pancreatic cancer cells (Liao et al., 2003). Induction of MKP-1 leads to resistance to therapeutically useful ionizing radiation (Nyati et al., 2006). Consequently, the availability of potent and selective inhibitors of MKP-1 would be quite desirable both as biological reagents to deconstruct the roles of the MKP family members and as possible lead structures for future cancer-directed therapies. Unfortunately, MKP-1 appears to be a challenging pharmacological target for small molecule inhibitors. The limited number of known small molecule MKP-1 inhibitors are either not potent or not selective (Pathak et al., 2002; Vogt et al., 2005; Lazo et al., 2006; Arnold et al., 2007). Therefore, we were inspired to examine the pyrrole carboxamide library from the Center for Chemical Methodologies and Library Development because its members are bisfunctional and are reminiscent of potent inhibitors of other protein tyrosine phosphatases that are postulated to bind to the active site and an adjacent subpocket. Moreover, the library has never been systematically biologically evaluated, and its members contain a core structure found in many marine natural products with pharmacological activity. Interestingly, we found that 10 of the 172 pyrrole carboxamide library members inhibited human MKP-1 and that 2 of these are potent and selective. We used contemporary computational methodologies to speculate on the sites of interactions and to guide future synthetic efforts.
Materials and Methods
Chemical and Reagents. The synthesis of the tricyclic pyrrole-2-carboxamide library has been published previously (Werner et al., 2006). Only 172 of the previously published compounds were available in sufficient quantities to permit biological analysis. Epitope-tagged His6Cdc25B2, His6MKP-1, and His6MKP-3 were expressed in Escherichia coli and purified by Ni-NTA as described previously (Lazo et al., 2001). Recombinant VHR and PTP1B were obtained from Biomol (Plymouth Meeting, PA). A549 cells were obtained from American Type Culture Collection (Manassas, VA), and growth inhibition was determined with Alamar Blue (CellTiter-Blue; Promega, Madison, WI). Unless otherwise indicated, all other reagents were obtained from Sigma-Aldrich (St. Louis, MO).
Enzyme Assays. Enzyme activities in the absence and presence of small molecule inhibitors were measured using the artificial substrate 3-O-methylfluorescein phosphate (OMFP) at concentrations equal to the Km of each enzyme and at the optimal pH for individual enzyme activity as described previously (Lazo et al., 2001). We first evaluated the 172 compounds in the pyrrole carboxamide library for inhibition of MKP-1 at 10 μM in a 96-well high-throughput screening format. The standard assay condition contained 40 μM OMPF in an assay buffer comprising 30 mM Tris-HCl (pH 7.0), 75 mM NaCl, 1 mM EDTA, 0.33% bovine serum albumin and 1 mM dithiothreitol. Fluorescence emission from the product was measured after a 60-min incubation period at ambient temperature with a multiwell plate reader (Cytofluor II, excitation filter, 485 nm/20 nm bandwidth; emission filter, 530 nm/30 nm bandwidth; Applied Biosystems, Foster City, CA). Compounds that inhibited MKP-1 by >50% in the initial screen were reassayed for IC50 values, which were determined from two independent experiments using samples of eight concentrations ranging from 0.78 to 100 μM. Data were analyzed with Prism 3.0 (GraphPad Software, Inc., San Diego, CA). Unbiased assignments for the best-fit model were determined using Enzyme Kinetics Module 1.0 (SPSS Inc., Chicago, IL). For the reversibility studies, we used a previously described dilution method (Brisson et al., 2005). Briefly, the enzyme and 90 μM PSI2106, MDF2085, or vehicle control were preincubated for 0, 5, or 20 min and then diluted 20-fold with incubation buffer. The OMFP substrate was added, and the remaining enzyme activity was determined and compared with the vehicle control sample. For the reductant studies, 1 to 25 mM glutathione was added during the incubation period.
Inhibition of Erk dephosphorylation was determined by incubating recombinant His6-tagged MKP-1 (125 ng) with tyrosine- and threonine-phosphorylated Erk2 (10 ng; New England Biolabs, Ipswich, MA) in a final reaction mixture (15 μl) containing 30 mM Tris-HCl, pH 7.0, 75 mM NaCl, 0.67 mM EDTA, 1 mM dithiothreitol, and 0.033% bovine serum albumin. PSI2106 or MDF2085 (6.3–50 μM), sanguinarine (50 μM), or dimethyl sulfoxide vehicle control was added, and after incubation for 60 min, Erk dephosphorylation was determined by Western blotting using 10% Tris-glycine gels and a monoclonal phospho-p44/42 MAPK antibody (catalog no. 91065; Cell Signaling Technology Inc., Danvers, MA) at 1:1000 dilution.
Molecular Modeling and Docking Studies. The active catalytic domain structures of MKP-1 and MKP-3 were modeled on the basis of that of MKP-5 deposited in the Protein Data Bank (Protein Data Bank code: 1ZZW) (Jeong et al., 2006). Homology modeling was possible, given that the target and template structures had sufficiently high sequence identity: MKP-1 and MKP-5 have 44% amino acid sequence identity (56% identity around the active site) and MKP-3 and MKP-5 have 48% sequence identity (66% identity around the active site). Models were generated using MODELLER 8v2 (Sali and Blundell, 1993) and refined using AMBER force field implemented in SYBYL 7.2 (Tripos, Inc., St. Louis, MO). The small molecule inhibitors examined (Table 1) were docked onto model structures, as well as known Cdc25B and VHR structures, using a two-step protocol. First, by using AutoDock 3.05 (Morris et al., 1998), an unbiased docking was performed in which the whole catalytic domains were targeted to identify potential binding sites. Then, by using GOLD 3.1 (Jones et al., 1997), flexible side-chain docking was performed to characterize optimal binding poses of the inhibitors. The electrostatic potentials of the solvent-exposed surfaces of the four DSPs were calculated using APBS software (Baker et al., 2001).
An initial analysis of the pyrrole carboxamide library revealed that 22 of the 172 members inhibited MKP-1 in vitro with IC50 values of <30 μM (Table 2), which was a remarkably high positive rate from a chemical library and documented the benefits of the focused library approach. A further analysis of the 10 compounds with IC50 values for MKP-1 of <10 μM indicated that several were selective for this phosphatase compared with other related phosphatases, such as MKP-3, Cdc25B, PTP1B, or VHR (Tables 3 and 4). Of particular note were PSI2106 and MDF2085, which had IC50 values of 8.0 ± 0.9 and 8.3 ± 0.8, respectively. All of the 10 inhibitors possess groups at the para position of the aryl ring that are inductively electron withdrawing (R1 = OMe or F). Both pyrrolidine and N,N-dimethyl substitutions were acceptable in the R2 position, whereas a wide variety of heterocyclic moieties were found on the R3 position. The pyrrole carboxamide library contained 41 building blocks for the R3 position; these included aliphatic as well as aromatic and heteroaromatic functionalities and lipophilic as well as hydrophilic ones. Interestingly, all inhibitors contained an aromatic or heteroaromatic ring system in this position, and all of them were substituted in the ortho and/or meta position. Although a variety of substituents in R1 and R2 positions were tolerated, a more specific substitution pattern in the R3 position was required.
The physicochemical properties of the 172 pyrrole carboxamide library members were evaluated previously (Werner et al., 2006). Using an identical computational analysis, we identified only 1 of the 10 inhibitors, namely MDF2050, that was significantly different from the mean values of the overall population for molecular mass and molecular volume (Table 4). All 10 compounds shared similar computational attributes with previously defined “drug-like” compounds with the exception of the calculated partition coefficient (log P), aqueous solubility (log S) and serum protein binding (log Khsa). All three of these calculated properties were frequently violated in this pyrrole carboxamide library (Werner et al., 2006).
Some protein tyrosine phosphatases appear to be highly sensitive to oxidants, such as H2O2 or quinones, because of the low pKa of the catalytic cysteine (Brisson et al., 2005). To exclude the possibility that inhibition by the pyrrole carboxamides was secondary to simple oxidation, we evaluated the reversibility of enzyme inhibition in the presence or absence of the physiological reductant GSH. Both PSI2106 and MDF2085 retained significant inhibition in the presence of 1 to 25 mM GSH (Fig. 1). This lack of a marked increase in IC50 values with increasing amounts of reducing agent suggested that the inhibitors were not acting simply through oxidation-reduction mechanisms.
Because of the initial positive evaluation of PSI2106 and MDF2085, we examined the reversibility and kinetics of inhibition with the small molecule substrate OMFP. Reversibility of inhibition was determined using a dilution method described previously (Brisson et al., 2005). As indicated in Fig. 2A, preoccupation with either pyrrole carboxamide for up to 20 min did not alter the percent inhibition of MKP-1. We next formally determined the kinetics of inhibition and found that both PSI2106 and MDF2085 had calculated Ki values below 10 μM with mixed kinetics, exhibiting both competitive and noncompetitive properties (Fig. 2, B and C). These results would be consistent with the two compounds interacting at sites near or outside of the catalytic domain and could reflect the bisfunctional nature of the compounds and the possible presence of a second subpocket common to some tyrosine phosphatases. Because some MKPs assume an altered conformation and are activated after binding to their cognate substrate (Camps et al., 1998; Zhou et al., 2001), we examined the ability of PSI2106 and MDF2085 to inhibit the dephosphorylation of the phosphorylated Erk (Fig. 3). PSI2106 caused a concentration-dependent inhibition of Erk dephosphorylation but was not as effective as sanguinarine, a previously reported inhibitor of MKP-1 (Vogt et al., 2005). MDF2085 also inhibited the dephosphorylation of the protein substrate, but it was less effective than PSI2106.
To assist in understanding the potential specificity PSI2106, we examined the surface properties of the catalytic domains of the four phosphatases (Fig. 4). MKP-1 and MKP-3 displayed similar surface properties in the neighborhood of the active site. In particular, there was a concave region surrounded by hydrophobic residues on both structures, where the hydrophobic groups of the compounds tended to be positioned. Cdc25B and VHR, on the other hand, exhibited a significantly more polar/charged character near the catalytic site. The Cdc25B active site was essentially basic, whereas VHR showed mostly acidic and some hydrophobic character. We then conducted unbiased molecular docking simulations in which the 10 active compounds were docked without any predefined binding site to the four protein phosphatases using AutoDock. Thirty docking runs were performed for both enantiomers of each compound, leading to a total of 600 poses for each phosphatase. The centroids of all poses are displayed in Fig. 4 to identify potential binding sites where the small molecules cluster. Notably, at least three potential binding sites were predicted for the 10 active compounds, with one of them being near the catalytic cysteine, and are seen enclosed in the blue ellipse.
Docking poses were found to cluster at hydrophobic regions and large cavities on the surface. Clusters were determined using an agglomerative clustering scheme and are colored differently. The poses closest to catalytic sites are shown in yellow. These are proposed to form the bound conformations that can potentially exhibit competitive inhibition. Based on the comparative analysis of these binding surfaces in the four DSPs, we speculated that the greater inhibitory action for MKP-1 and MKP-3 of the 10 compounds might be associated with the hydrophobic nature of the surface near their active site. The centers of the clusters closest to the catalytic site were further used as the center of biased search space using GOLD. The recurrent features observed in the docking poses of PSI2106 from these biased docking simulations were interactions of aromatic moieties of PSI2106 with the aromatic side chains of residues His229 and Phe299, in addition to a number of hydrophobic contacts involving residues Ala260 and Ile262 and the β-carbons of Ser263 and Asn298 illustrated in Fig. 5 for MKP-1. Comparable interactions were observed with MKP-3. Cdc25B and VHR, on the other hand, lacked this type of hydrophobic interactions. Apparently, shape complementarity, rather than the specificity of interactions, played a major role in selecting the binding site near the catalytic cysteine.
Unlike the serine/threonine phosphatases, which have large numbers of potent small molecule inhibitors, mostly natural products, the tyrosine phosphatases and the DSPs in particular have few inhibitors. Pentamidine (Pathak et al., 2002) and sanguinarine (Vogt et al., 2005) have been reported to be in vitro inhibitors of MKP-1, but they are not selective and sanguinarine is not very potent. The benzofuran NU-126 showed some selectivity for MKP-1 compared with MKP-3, which supported the theoretical argument that differences in the amino acid composition would support selective inhibitors (Lazo et al., 2006). Nonetheless, NU-126 was not very potent with an IC50 value for MKP-1 of almost 30 μM (Lazo et al., 2006). The newly identified pyrrole carboxamides are structurally unique from previously known MKP-1 inhibitors. PSI2106 and MDF2085, which have IC50 values below 10 μM, showed >5-fold selectivity for human MKP-1 over MKP-3 and a much greater preference for MKP-1 over Cdc25B, PTP1B, or VHR. Their reversible inhibition was consistent with a covalent adduct- or redox-independent inhibitory mechanism, which separates these compounds from other reported inhibitors of DSPs.
The identification of 10 active compounds in the discovery library of 172 suggested that the pyrrole carboxamide pharmacophore could be an attractive chemical platform for future analog development. It is interesting to compare this active compound ratio with that seen with a more traditional focused library. We recently found 100 active compounds after screening 65,239 diverse small molecules from the National Institutes of Health repository (http://pubchem.ncbi.n-lm.nih.gov/assay/assay.cgi?aid=374), which was a positive ratio that was 40-fold less than that seen with the pyrrole carboxamide library. Furthermore, the pyrrole carboxamide core structure was not identified in the National Institutes of Health diverse small molecule library, indicating the value of our focused library. The ability of two of the library compounds to inhibit the dephosphorylation of phosphorylated Erk also suggested that they could be useful. It should be noted, however, that the compounds appear to be much better inhibitors of the dephosphorylation of OMFP than of the protein substrate phosphorylated Erk. This may reflect the altered MKP conformation that is thought to occur after binding to their cognate substrate (Camps et al., 1998; Zhou et al., 2001), which is probably not recapitulated by OMFP. We have attempted to evaluate the ability of PSI2106 and MDF2085 to inhibit MKP-1 in cells using previously described cell-based assays (Vogt et al., 2005), including one that has been modified using a tetracycline-regulated promoter and will be described elsewhere. We did not, however, observe any evidence for MKP-1 inhibition. We considered the possibility that this lack of cellular activity might be due to high serum protein binding properties that were predicted by QikProp (Table 4), but we have observed significant growth inhibition of A549 cells cultured in the presence of 10% fetal bovine serum with PSI2106 (IC50 = 5.3 ± 2.8 μM; n = 2) and MDF2085 (IC50 = 4.4 ± 1.2 μM; n = 2). Thus, both compounds have biological activity in the presence of serum. We have not yet linked A549 cell growth inhibition to cellular MKP-1 inhibition, however, and we have not seen growth inhibition with HeLa cells under similar conditions. Additional mechanistic studies of these pyrrole carboxamides are warranted.
Because of the highly conserved protein tyrosine phosphatase active site, it was originally thought that selective inhibitors would not be obtainable. A strategy has been proposed, however, in which a substrate might bind to the active site, the phospho-Tyr-binding site, and then to a peripheral site. Indeed, some protein tyrosine phosphatase inhibitors do possess two pharmacophore subunits tethered together via a conformationally mobile linker (Puius et al., 1997). It is enticing to think that some of these 10 compounds are active because they contain two pharmacologically interesting subunits: a pyrrole carboxamide and a substituted proline. Documentation of this theory, however, will require additional studies, which it is hoped will be aided by the molecular modeling already conducted.
In summary, we believe these results illustrate the potential of the pyrrole carboxamide library for future pharmacological studies. Selective and potent inhibitors of MKP-1 would be valuable reagents to deconstruct the cellular role of this phosphatase in inflammation and neoplasia and might provide the chemical foundation for the development of new anticancer agents.
We appreciate our constructive discussions with members of the University of Pittsburgh Center for Chemical Methodologies and Library Development, the Drug Discovery Institute, and the Pittsburgh Molecular Library Screening Center, in particular Drs. Peter Wipf, Elizabeth R. Sharlow, Paul A. Johnston, and Andreas Vogt.
This work was supported by National Institutes of Health Grants MH074411, GM067982, and CA052995 and by the Fiske Drug Discovery Fund.
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
ABBREVIATIONS: MAPK, mitogen-activated protein kinase; MKP-1, mitogen-activated protein kinase phosphatase-1; DSP, dual specificity phosphatase; Erk, extracellular signaling-related protein kinase; OMFP, 3-O-methylfluorescein phosphate; GSH, glutathione; NU-126, 2-((E))-2-(5-cyanobenzofuran-2-yl)vinyl)-1H-indole-6-carbonitrile.
- Received March 3, 2007.
- Accepted May 29, 2007.
- The American Society for Pharmacology and Experimental Therapeutics