Using virtual screening and biological assays, we have identified and characterized a new class of compounds that selectively inhibit the proliferation of cancer cells containing constitutively active kinase signalling pathways by targeting kinase regulation of oncogenic transcription factors.
The mitogen-activated protein kinase (MAPK) family members, which include the extracellular-signal-regulated kinases 1 and 2 (ERK1/2), c-Jun N-terminal kinases (JNKs), p38 MAPKs and ERK5, are major regulators of cell proliferation, apoptosis, differentiation, migration and inflammation in response to a variety of intra- and extra-cellular signals [1,2]. MAPK proteins regulate cellular functions through phosphorylation of a diverse number of substrates. In particular, the ERK1/2 proteins have been implicated in the phosphorylation of well over 100 substrates, and stringent control over interactions of ERK1/2 with substrate proteins that allow efficient phosphate transfer is essential for cellular functions [3–5]. Unregulated activation of the ERK1/2 pathway is often observed in a variety of cancers, which contributes to uncontrolled cell proliferation, survival and resistance to anti-cancer drugs [6,7]. Thus, a better understanding of the mechanisms regulating ERK1/2 interactions with substrates may provide information on new approaches to selectively inhibit substrates involved in cancer cell proliferation and survival. Although several selective ATP-competitive and catalytic site inhibitors of ERK1/2 have been developed [8–10], these compounds have not advanced to the clinic and do not allow examination of select ERK functions that are dependent on interactions with specific substrate proteins.
Many of the ERK1/2 substrate and interacting proteins, including the Elk-1 transcription factor, p90 ribosomal S6 kinase-1 (RSK-1), the caspase 9 protease and the hematopoietic protein tyrosine phosphatase (hePTP), contain a DEJL (docking site for ERK and JNK, LXL) motif or D-domain that is involved in kinase recognition [11–14]. The D-domain consists of basic residues followed by a hydrophobic LL motif and interacts with acidic and hydrophobic regions in the C-terminus of ERK1/2, referred to as the common docking (CD) domain or D-recruitment site (DRS) [13–15]. A second docking domain, known as the F-site or DEF (docking site for ERK, FF) motif, has been identified on several ERK1/2 substrates, including Elk-1 and c-Fos transcription factors, A Raf kinase, the kinase suppressor of Ras-1 (KSR-1) scaffold protein and nuclear pore proteins such as NUP153 and NUP214 [16–18]. The F-site is typically separated from the phosphorylation site by six to ten amino acids, whereas the D-domain may be located 20 amino acids further from the phosphorylation site to accommodate the spatially separated hydrophobic interactions.
The DRS on ERK2 includes residues Asp316 and Asp319, adjacent hydrophobic amino acids and ED domain residues (Glu160/Asp161 for p38α MAPK and Thr157/Thr158 for ERK2) that facilitate selective interactions between D-domain-containing substrates and MAPKs . F-site-containing substrates interact with hydrophobic regions on the ERK2 F-recruitment site (FRS) and include residues Leu198, Tyr231, Leu232, Leu235 and Tyr261 [20,21]. Other MAPKs, including p38α, may also utilize a FRS-like binding motif during substrate recognition [22,23]. Beyond the DRS and FRS, experimental studies using ERK2 mutants have implicated other residues that may be important for ERK interactions with substrates or other regulatory proteins . Indeed, ERK2 and other MAPKs have a unique insert in the kinase homology region that may regulate its interactions with upstream activating MEK (MAPK/ERK kinase) proteins . Given the large number of substrates regulated by ERK1/2 proteins, it is likely that additional docking sites on ERK1/2 proteins will be identified and shown to regulate specific protein–protein interactions.
In previous studies, we used computer-aided drug design (CADD) to identify low-molecular-mass compounds that inhibit ERK1/2-mediated phosphorylation of substrate proteins in an ATP-independent manner and inhibit cancer cell proliferation in in vitro and in vivo models [26,27]. This approach targeted the DRS region of ERK2, such that the identified inhibitors containing a thiazolidinedione scaffold that may selectively regulate distinct ERK2 signalling functions. From those efforts, several compounds were identified that inhibit phosphorylation of ERK1/2 substrates, including D-domain-containing substrates RSK-1 and caspase 9 and selectively inhibit cancer cell lines containing constitutively active ERK1/2 signalling [28–30].
The present study reports the identification of a new class of small molecules, based on a thienyl benzenesulfonate scaffold, which are putative inhibitors of substrates that interact with ERK1/2 through the FRS. The compounds were initially identified using virtual database screening followed by experimental assays. Validation of the biological activity of the scaffold was then achieved by designing analogues using the site identification by ligand competitive saturation (SILCS) approach [31,32], chemical synthesis and experimental evidence demonstrating inhibition of the expression of F-site-containing immediate early genes (IEGs) that comprise the activator protein-1 (AP-1) transcription factor complex. Correspondingly, melanoma cells containing mutated BRaf and active ERK1/2 signalling, including cells that are inherently resistant to currently available kinase inhibitors used in the clinic, were selectively inhibited by the new compounds. These results provide support for the feasibility of developing low-molecular-mass compounds that can selectively inhibit ERK1/2-mediated signalling functions that drive the proliferation and survival of cancer cells containing constitutively active ERK1/2 signalling.
Initial test compounds were obtained from Chembridge Corporation. Lead compound analogues were synthesized using reagent grade chemicals and solvents purchased from Sigma–Aldrich, Alfa Aesar, Oakwood Chemicals and TCI America. 1H- and 13C-NMR spectra were recorded on Varian INOVA 400 MHz and Varian INOVA 500 MHz NMR spectrometers at 25°C. Chemical shifts are reported in parts per million (ppm). The residual solvent peak was used as an internal reference. The mass spectra were obtained on an Electrospray TOF (ESI–TOF) mass spectrometer (Bruker amaZon X). Prior to biological testing, final compounds were confirmed to be >95% pure by HPLC chromatography. The MEK1/2 inhibitor U0126 was purchased from EMD Millipore and the ATP-dependent pyrazolylpyrrole ERK2 inhibitor  and p38α/β MAPK inhibitor SB239063 were purchased from Santa Cruz Biotechnology. Unless indicated, all other chemicals listed below were purchased from Sigma–Aldrich and/or Fisher Scientific. HeLa cervical carcinoma cells, Jurkat T-cell leukaemia cells, A375 melanoma and mutated BRaf inhibitor-resistant RPMI7951 melanoma cells were purchased from the A.T.C.C. Additional details are provided in the Supplementary Methods. Antibodies recognizing total c-Fos, FosB, Fra1, c-Myc and c-Jun were purchased from Cell Signaling Technology. The phospho-specific ERK1/2 (pThr183/pTyr185) and α-tubulin antibodies were purchased from Sigma. Phosphorylation-specific antibodies against c-Fos (pSer32, pThr232, pThr325 or pSer374), RSK-1 (pThr573 or pSer380), MEK1/2 (pSer217/pSer221), EGFR (epidermal growth factor (EGF) receptor; pTyr1068), cRaf-1 (pSer289, pSer296, pSer301) or Elk-1 (pSer383) were purchased from Cell Signaling Technology. Antibodies against total early growth response-1 (Egr1), RSK-1, ERK2 and MEK1 were purchased from Santa Cruz Biotechnology.
Protein expression and purification
His6-tagged ERK2 wild-type was expressed in Escherichia coli and purified as described previously . See the Supplementary Methods for more details.
Analysis of protein expression levels, activity and cell proliferation were done as previously described [28,29]. Cells were washed with ice-cold PBS (pH 7.2; Invitrogen) and protein lysates were collected with 2× SDS/PAGE sample buffer (4% SDS, 5.7 M β-mercaptoethanol, 0.2 M Tris/HCl, pH 6.8, 20% glycerol and, 5 mM EDTA). Proteins were separated by SDS/PAGE, analysed by immunoblotting and detected using ECL (GE Healthcare).
In vitro kinase assays
Assays were done with 10 nM of recombinant active forms of ERK2 or ERK5 (Abcam) and substrates including 100 nM c-Fos (Abcam) or ∼0.2 μg/reaction GST–RSK1 (residues 386–752) or using previously described buffer conditions . The p38α MAPK (0.43 μg) was activated with 75 ng of recombinant active MKK6 (S207E, S211E mutations; Life Technologies) and included recombinant ATF2 (2 μg) as the substrate. Compounds (0–60 μM) were pre-incubated with ERK or p38α MAPK proteins in the absence of substrates for 30 min at room temperature. Kinase assays were initiated with the addition of substrate and active MKK6 for p38α reactions, plus 100 μM ATP. Reactions were stopped after 30 min with an equal volume of 2× SDS/PAGE sample buffer and the proteins were analysed by immunoblotting for specific substrate phosphorylation sites and quantified by densitometry. Catalytic site inhibitors SCH772984 (Selleck Chemicals) and SB239063 were used as positive controls for ERK2 and p38α inhibition respectively.
Cell proliferation and apoptosis assays
Cells were seeded at 25 000 cells/well in 96-well plates, cultured overnight and treated for 24 h with the indicated dose of compounds. Cell viability or caspase 3/7 activity was measured according to the manufacturer's instructions using the fluorescent Cell Titer Blue Assay or luminescent Caspase 3/7 Glo Assay, respectively (Promega).
AP-1/SRE promoter luciferase assays
HeLa or A375 cells were seeded in 24-well plates (40 000 cells/well) and incubated for 18 h to achieve ∼60–70% confluence. Cells were transfected with AP-1 [pAP1(PMA)-TA-Luc; Clontech] or the serum-response element (SRE; pGL4.33-SRE; Promega) luciferase reporter plasmids (250 ng/well) using Lipofectamine™ (Invitrogen). After 16 h, cells were pre-treated with 1–100 μM compounds as indicated for 20 min, followed by stimulation with 25 ng/ml EGF or 200 nM PMA for 4.5 h. In some experiments, cells were transfected with the pAP1(PMA)-TA-Luc promoter and a constitutively active MEK1 mutant cDNA were treated with various amounts of compounds during the last 4 h of incubation. The luciferase activity in the cell extracts was determined with a Dual Luciferase Assay System (Promega) according to the manufacturer's instructions. Human embryonic kidney (HEK) 293 cells (20 000 cells/well) were plated in white walled 96-well plates (Nunclon Delta Surface; Thermo Scientific) and then transfected with the luciferase reporter plasmids indicated above. Cells were treated with EGF or PMA for 5 h in the presence or absence of compounds (1–100 μM) and luminescence was measured using Bright-Glo Substrate (Promega). Luciferase activities were monitored with a Lumat LB 9507 luminometer (Berthold Technology) and data were normalized to the amount of protein in each sample.
Molecular dynamics (MD) simulations were performed using the programs Chemistry at HARvard Macromolecular Mechanics (CHARMM)  and NAnoscale Molecular Dynamics (NAMD)  with the CHARMM22/CMAP additive force field [36,37] with the Transferable Intermolecular Potential with 3 interaction sites (TIP3P) water model . Simulations initiated with the 3D structures of ERK2 in both the phosphorylated (active)  and unphosphorylated (inactive)  states. SILCS simulations for the FRS (site 5) of ERK2 were initiated the unphosphorylated ERK2 structure and included 1 M benzene and 1 M propane, as previously described . Binding site identification  and database screening was performed on diverse conformations from the simulations. Screening was performed in two steps against a virtual database of more than 1.5 million low-molecular-mass commercially available compounds. Final selection of the 150 compounds for experimental assay involved maximizing chemical and structural diversity, via fingerprint-based clustering [42,43] with the program Molecular Operating Environment (MOE) (Chemical Computing Group Inc.) as well as considering their potential bioavailability . The purity of compounds shown to have biological activity was verified by MS.
Prediction of ligand modifications were performed using ligand grid free energy (LGFE) scores based on the SILCS methodology . SILCS simulations were initiated from the 2.3.2–ERK2 crystal structure determined in the present work. Visualization of the SILCS FragMaps identified the region beyond the benzene ring of 2.3.2 as suitable for non-polar moieties. Subsequently, the analogues of 2.3.2 were modelled on to ERK2 starting from the 2.3.2–ERK2 crystal structure. Relative LGFE scores with respect to 2.3.2 were then determined and are presented in the Results’ section. Additional details of all the computational methods are included in the Supplementary Methods.
In silico approaches to identify lead compounds
The initial objective of these studies was to use in silico modelling and experimental assays to identify low-molecular-mass compounds that selectively inhibit ERK2 interactions with substrate proteins by targeting unique docking sites and do not interfere with ATP binding. Computational approaches were first applied to identify putative small-molecule-binding sites located on regions in the C-terminus of ERK2 previously implicated via mutagenesis studies to be important for substrate binding, as described in the Supplementary Online Data. The five sites targeted, with the residues shown on the structure of ERK2, include the DRS , which was previously targeted in our laboratory , the FRS known to be involved in interactions with ERK1/2 substrates containing the F-site or DEF motifs  and the focus of the present study and three putative sites that span the region between the DRS and FRS (Supplementary Table S1; Supplementary Figure S1). Database screening was then performed against these sites in which 50 000 compounds were selected following the initial round of docking with a second round of screening yielding 1000 compounds. Details of the screening approach, including the use of score normalization based on molecular mass , are given in the Supplementary Online Data (Supplementary Tables S2–S4; Supplementary Figure S2). Final compound selection for experimental assays from the 1000 compounds emphasized chemical and structural diversity to maximize the identification of active compounds . Diversity was attained using fingerprint-based similarity clustering to group compounds into structurally similar clusters from which individual molecules were selected. This selection process also included bioavailability considerations .
Biological and structural evaluation of lead compounds
From the in silico screen, we focused on a group of seven structurally diverse compounds (Figure 1A) and their effects on ERK1/2-mediated phosphorylation and signalling events. Using lysates from EGF-treated HeLa cells, 100 μM compound 2.3.2 inhibited ERK1/2-mediated phosphorylation of the transcription factor Elk-1, which contains both a D-domain and an F-site (Figure 1B). Compounds 2.3.3, 2.3.4 and 2.3.5 also inhibited Elk-1 phosphorylation, but to a lesser extent (Figure 1B). Importantly, none of the compounds caused a corresponding inhibition of ERK1/2 activation, as measured by phosphorylation of the activation site residues, suggesting the compound's mechanism of action involved targeting signalling downstream of ERK1/2. Whereas 2.3.2 inhibited ERK1/2-mediated Elk-1 phosphorylation, it appeared to be less potent at inhibiting ERK1/2-mediated phosphorylation of the D-domain-containing substrate RSK-1 (Figure 1C). Further support that 2.3.2 was acting on ERK1/2-mediated events was demonstrated by the lack of effects on EGF-induced phosphorylation of MEK1/2, which interact with ERK1/2 through their D-domains  or autophosphorylation of the EGFR (Figure 1C).
Selective inhibition of ERK-mediated phosphorylation of substrates by compounds
To provide additional evidence for the binding site of 2.3.2, crystallographic studies were undertaken and crystals of 2.3.2 bound to the inactive form of ERK2 were obtained (Supplementary Table S5). Upon inspection of the model that included protein and 46 water molecules, additional density was identified in a pocket vacated by Phe181/Leu182 located below the activation loop that includes residues Thr183 and Tyr185 (Supplementary Figure S3). Given that the densities of these two residues are well resolved, it was predicted that the elongated density filling this region was representative of 2.3.2 (Supplementary Figure S3). The irregular shape of the electron-density fragment was accounted for by assuming that 2.3.2 binds in at least two orientations. In one of the modelled conformations, 2.3.2 appears to make hydrogen bonds to the backbone amides of Phe181 and Leu182; however, this should be taken with caution given the disordered and partially occupied nature of the ligand in the binding site. Thus, the data suggest that 2.3.2 is the first molecule identified to interact with ERK2 in the FRS region involved with regulating interactions with substrates containing an F-site. Although the role of the FRS region in co-ordinating protein–protein interactions may be unique to ERK2, it is possible that other MAPKs utilize this region for determining substrate recognition. For example, a comparable FRS region has recently been identified on p38α/β MAPKs  as well as the kinase domain of ERK5 . However, sequence alignment suggests that the ERK2 amino acids predicted to interact with 2.3.2 have differences with the corresponding residues on other major MAPKs, which could allow for FRS selectivity with small molecules (Supplementary Figure S4).
It has to be emphasized that the observed difference omit electron density (i.e. Fo−Fc map calculated in the absence of ligand) is by itself not conclusive of 2.3.2 binding to the FRS region of ERK2. However, the additional density observed with 2.3.2 was not observed for the ERK2 crystals obtained without ligand soaking. This indicates that the electron density is probably originating from 2.3.2 binding and not that of other components of the crystallization buffer (e.g. non-detergent sulfobetaines (NDSB)-256). Furthermore, it is clearly possible to place other molecules that were present in crystallization buffer (including NDSB-256) into the rather featureless electron density. Additional complication arises from the proximity of the crystal contact to the putative binding site, which may result in a binding mode that is somewhat distorted compared with what happens in solution. Thus, the crystallographic evidence presented can only suggest the possibility of the 2.3.2 binding in the detected location on the protein surface and does not unequivocally prove it. Therefore, additional support for 2.3.2 and chemically related analogues regulating ERK1/2 functions through the FRS comes from further computational and biochemical studies described in the following sections.
Optimization of lead compound by SILCS
Chemically similar analogues were obtained from SILCS 3D FragMaps obtained from SILCS simulations initiated from the putative 2.3.2–ERK2 crystal structure are shown overlaid on the crystal structure obtained (Figure 2A). The FragMaps represent the regions to which different type of functional groups bind and include contributions from protein flexibility, desolvation and functional group protein interactions . Visual analysis revealed the overlap of the aliphatic and aromatic SILCS FragMaps with the phenyl moiety of 2.3.2, with those FragMaps extending beyond the crystallographic position of the phenyl ring. This indicated that the addition of non-polar moieties to this phenyl ring may improve affinity and subsequent potency. Synthetically accessible chemical modifications of the phenyl ring were identified, modelled on to 2.3.2 bound to ERK2, and the relative SILCS LGFE scores were calculated and presented as a change in LGFE scores (∆LGFE) relative to 2.3.2 and suggested that hydrophobic modifications may increase activity (Figure 2B). Accordingly, along with 2.3.2, which is from now on referred to as SF-3-026, analogues containing the p-CH3 (SF-3-027), biphenyl (SF-3-029) or naphthyl (SF-3-030) group were synthesized for evaluation in subsequent biochemical and cell-based assays.
SILCS identified analogues of 2.3.2 inhibit Elk-1 phosphorylation
Lead compounds inhibit ERK1/2-mediated IEG promoter activity
ERK1/2 phosphorylation of Elk-1 regulates the SRE that drives the transcription of IEGs, such as c-fos, which is an F-site containing ERK1/2 substrate . Initial testing of SF-3-026 and analogues at 50 μM in EGF-treated HeLa cells showed preferential inhibition of the phosphorylation of Elk-1 as compared with the D-domain-containing substrate RSK-1 (Figure 2C). In addition, the phosphorylation of ERK1/2 or MEK1/2 proteins was not affected by the compounds at the dose tested (Figure 2D). It was next determined whether the compounds could affect the transcription factor functions of Elk-1 or AP-1 proteins. HeLa cells expressing luciferase constructs driven by the SRE or AP-1 promoters, which are regulated by Elk-1 or Fos family proteins, respectively, were treated with SF-3-026 and analogues in the context of EGF stimulation. EGF alone caused a more than 20-fold increase in SRE- or AP-1-mediated luciferase activity, which could be inhibited by ∼90% with 10 μM of the MEK1/2 inhibitor, U0126 (results not shown). Treatment with the test compounds caused a dose-dependent inhibition of the SRE and AP-1 promoters (Figures 3A and 3B). As predicted by SILCS, SF-3-029 and SF-3-030 were the most potent inhibitors especially with the AP-1 promoter where IC50 values were ∼10 μM (Figure 3B). To further demonstrate the targeting of ERK1/2 mediated transcription, AP-1 promoter activity was measured in HeLa cells expressing an active MEK1 mutant, which only activates ERK1/2. As shown in Figure 3(C), SF-3-026 and its analogues caused dose-dependent inhibition of AP-1 promoter activity with SF-3-029 and SF-3-030 with IC50 values around 5–10 μM as compared with SF-3-026 and SF-3-027 with IC50 values of ∼40 μM. As controls, MEK1-induced AP-1 promoter activity was inhibited by >80% with 10 μM U0126 or a previously reported ATP-competitive pyrazolylpyrrole ERK inhibitor  (results not shown). We further evaluated AP-1 or SRE promoter activity in HEK293 cells stimulated with EGF or PMA. As shown in Table 1, the compounds were more potent at inhibiting PMA-induced promoter activity compared with EGF and SF-3-030 was the most potent lending further support of the SILCS computational predictions.
Compound SF-3-026 and analogues inhibit SRE and AP-1 transcription factor function
|AP-1 (PMA)||AP-1 (EGF)||SRE (PMA)||SRE (EGF)|
|AP-1 (PMA)||AP-1 (EGF)||SRE (PMA)||SRE (EGF)|
Lead compounds inhibit EGF or PMA-induced expression of IEG proteins
HeLa cells treated with EGF show a transient phosphorylation of Elk-1, c-Fos and Fra 1, followed by robust expression of c-Fos, FosB, FosB2 and Fra 1 after 30–60 min (Figure 4A). Although SF-3-026 and analogues had little effect on the initial phosphorylation of c-Fos and Fra 1 as was evident by gel shifts (Figure 4B), the robust induction of Fos family proteins by EGF at 1 h was inhibited in a dose-dependent manner with SF-3-030 (Figure 4C). Inhibition of c-Fos phosphorylation at Ser32 was observed prior to loss of protein, but phosphorylation of Ser374, which is the reported priming site for other phosphorylation events on c-Fos , was unaffected (Figure 4C). SF-3-030 also inhibited EGF-mediated expression of other IEGs including c-Myc, but was less effective at inhibiting c-Jun and Egr 1 protein induction (Figure 4C). These data suggest that SF-3-030 can selectively inhibit EGF-mediated induction of IEG proteins.
SF-3-026 and analogues inhibit EGF-induced expression of IEGs
SF-3-030 also inhibited IEG expression in cells treated with the phorbol ester PMA. Like EGF, PMA caused a transient phosphorylation of Elk-1 followed by increased expression of c-Fos, FosB and Fra 1 after 60–90 min (Figure 5A). At a dose of 40 μM, SF-3-030 inhibited c-Fos phosphorylation at Thr232 and Ser32, which may be involved in transcription regulation, but, similar to EGF treatment, had no effect on phosphorylation of Ser374 (Figure 5B). The expression of total FosB, FosB2, Fra1 and F-site-containing c-Myc was also decreased at the 40 μM dose (Figure 5B). The inhibition of c-Fos phosphorylation at Thr232 and Ser32 by SF-3-030 corresponded to a loss of FosB and Fra 1 expression, suggesting that c-Fos is transcriptionally inactive (Figure 5B). At high doses (70 and 100 μM), SF-3-030 increased ERK1/2 phosphorylation (Figure 5B), which indicated potential off-target toxicity or inhibition of negative feedback mechanisms on MEK1/2 and Raf-1 [50,51]. The PMA-mediated IEG phosphorylation and expression was further evaluated at doses of less than 40 μM. As shown in Figure 5(C), a dose-dependent inhibition of c-Fos phosphorylation of Thr232 by SF-3-030 corresponded to decreased FosB, FosB2 and Fra1 expression and supported impaired c-Fos transcription activity. SF-3-030 at low doses inhibited the negative feedback phosphorylation sites of Raf-1 (Figure 5C), which appeared to increase with higher doses in a manner that correlated with increased ERK1/2 phosphorylation (Figure 5B). A Raf, but not Raf-1 or BRaf, has a consensus F-site  and the specific docking interactions between Raf-1 and ERK1/2 have not been defined.
SF-3-026 and analogues inhibit PMA-induced expression of Fos family IEGs
The effects of SF-3-030 on the expression of Fos proteins appeared to be selective as ERK1/2-mediated phosphorylation of RSK-1 or the induction of other IEGs, such as Egr1 and c-Jun, were less affected (Figure 5D). SF-3-030 also had minimal effects on anisomycin-mediated p38 MAPK signalling (Figure 5E), including phosphorylation of p38 MAPK substrates ATF2 and MK2. The phosphorylation of ATF2 at Thr71 may also be achieved by ERK1/2 , which could explain the lack of effects of the p38 inhibitor SB239063 and some inhibition by SF-3-030 at doses >20 μM (Figure 5E). Together, these data provide additional support that SF-3-030 preferentially affects ERK1/2-mediated signalling events.
SF-3-030 selectively inhibits ERK2-mediated phosphorylation of c-Fos in vitro
The effects of SF-3-030 on ERK-mediate phosphorylation of c-Fos were evaluated using in vitro kinases assays. ERK5 was used for comparison as it has been reported to phosphorylate c-Fos at Thr232 and Ser32 and regulate protein stability and transcription factor function , although others suggest that ERK1/2 is the primary regulator of c-Fos phosphorylation and activity . As shown in Figure 6(A), active ERK2 was the primary regulator of c-Fos as was evident by phosphorylation at the F-site-dependent Thr325 site, Ser374 and Thr232, as well as the c-Fos gel shift, which did not occur in the presence of active ERK5 (Figure 6A). Ser32 phosphorylation increased to a lesser degree with ERK2 as a result of high basal phosphorylation at this site, which was probably due to c-Fos expression and purification from eukaryotic cells (Figure 6A). Whereas SF-3-030 inhibited ERK2-mediated phosphorylation of Thr325 in a dose-dependent manner, it had little effect on the phosphorylation of Ser374, Thr232 or Ser32 (Figures 6A and 6B). However, SCH772984, a catalytic site inhibitor of ERK1/2, inhibited phosphorylation of all c-Fos sites (Figure 6A). SF-3-030 did not affect the modest induction of Ser374 phosphorylation by ERK5, but it did appear to cause a slight decrease in ERK5-mediated phosphorylation of Thr232 (Figure 6A), which is consistent with the cell data in Figure 5. Similarly, SF-3-030 had no apparent effects on ERK2 phosphorylation of the D-domain-containing substrate RSK1 or p38α-MAPK-mediated phosphorylation of the F-site-containing ATF2 protein, further supporting selective inhibition of ERK-mediated phosphorylation events by SF-3-030 (Figure 6C).
SF-3-030 selectively inhibits ERK2-mediated c-Fos phosphorylation in an in vitro kinase assay
Lead compounds inhibit AP-1 promoter activity in melanoma cells with constitutively active ERK1/2
The effects of SF-3-026 and analogues were next tested in A375 melanoma cells containing an activating valine to glutamate mutation on residue 600 (V600Q) of BRaf and constitutive ERK1/2 activation . Similar to the EGF- or PMA-stimulated HeLa and HEK293 cells, SF-3-029 and SF-3-030 were more potent inhibitors of AP-1 activity in A375 cells (Figure 7A). Given that SF-3-030 appeared to be the most effective at inhibiting the basal expression of Fos family proteins in A375 cells (Figure 7B), the dose-dependent inhibition effects of SF-3-030 on the expression of AP-1 proteins and other IEGs were further examined. SF-3-030 showed a dose-dependent inhibition of c-Fos, FosB, FosB2, Fra1 and c-Myc, but Egr1 expression was unaffected (Figure 7C). SF-3-030 also caused an increase in phosphorylated ERK1/2 (Figure 7C), which others have observed in A375 cells treated with an ATP-competitive ERK1/2 inhibitor, VTX-11e, and attributed to effects on negative feedback pathways . However, MEK1/2 phosphorylation is not increased with SF-3-030 (Figure 7C), suggesting that the effects on ERK1/2 negative feedback by SF-3-030 in cells expressing mutated BRaf are different from those in HeLa cells treated with EGF or PMA (Figures 4 and 5). The MEK1/2 inhibitor U0126 blocked Fra1 phosphorylation involved in transcription regulation (results not shown) but did not block the expression of Fra1 as was observed with other Fos proteins and c-Jun (Figures 7B and 7C). Additional selectivity of the compounds was shown where SF-3-026 and analogues did not affect basal expression of the D-domain-containing c-Jun protein or the phosphorylation of MEK1/2 (Figure 7D). As shown in Figure 7(C), all compounds at 50 μM increased phosphorylation of ERK1/2 and the D-domain substrate RSK1 (Figure 7D).
SF-3-026 and analogues inhibit Fos family proteins in melanoma cells
Lead compounds selectively inhibit proliferation of melanoma cells with constitutively active ERK1/2
It was next determined whether SF-3-026 and analogues affected the proliferation of A375 and RPMI7951 melanoma cells driven by BRaf (V600Q) mutation. RPMI7951 cells are heterozygous for the BRaf (V600Q) mutation but are resistant to BRaf inhibitors due to overexpression of MAP3K8 that re-activates ERK1/2 signalling [56,57]. Both melanoma cells lines were treated with various doses of SF-3-026 or analogues and cell viability was compared with HeLa cervical carcinoma and Jurkat T-cell leukaemia cells, which are p53-defective but contain no known activating mutations in the ERK1/2 pathway . Compound SF-3-026 and its analogues showed a dose-dependent inhibition of each cell line; however, the A375 and RPMI7951 melanoma cells were more sensitive to the test compounds with GI50 values in the low micromolar range and severalfold lower than the HeLa or Jurkat cells even when adjusted for individual cell growth rates (Table 2). As a control, the BRaf inhibitor PLX4032 was effective against the drug-sensitive A375 cells but not the RPMI7951, HeLa or Jurkat cells (Table 2).
To test whether the inhibition of cell proliferation was due to induction of apoptotic markers, A375 and HeLa cells were treated with various doses of SF-3-030 and analogues, and cell viability along with caspase 3/7 activity was measured. Treatment with compounds caused a dose-dependent increase in caspase 3/7 activity in A375 cells but had little effect on caspase 3/7 activity in HeLa cells (Figure 8A).
Lead compounds induce caspase 3/7 activity in A375 melanoma cells
Lastly, the expression of AP-1 proteins was evaluated in the four cell lines to evaluate whether differences in protein expression could explain why the test compounds selectively inhibit the proliferation of melanoma cells with active ERK1/2 signalling. As expected, phosphorylated and active ERK1/2 was elevated in the A375 and the drug-resistant RPMI7951 melanoma cells as compared with the HeLa or Jurkat cells (Figure 8B). When normalized to α-tubulin expression, all AP-1 proteins evaluated were expressed in A375 cells, which also contained the highest basal expression levels of Fra1 and FosB proteins as compared with the other cells (Figure 8B). Fra1 expression was also found in RPMI7951 cells but not in HeLa or Jurkat cells. Interestingly, HeLa cells showed the highest level of basal c-Fos expression, which could explain their increased sensitivity to growth inhibition when treated with the compounds compared with the Jurkat cells that have undetectable levels of the AP-1 proteins (Figure 8B). However, these data indicate that selective targeting of AP-1 heterodimer partners consisting of Fra1 by the lead compounds may play a role in sensitizing melanoma cells containing mutated BRaf and active ERK1/2 signalling to growth inhibition and apoptosis.
The findings of the present study describe a thienyl benzenesulfonate scaffold that can target ERK1/2 signalling pathways and selectively inhibit the expression of Fos family proteins involved in AP-1-regulated transcription. A variety of cancer types have been associated with dysregulated AP-1 transcription factor components [59,60]. Thus, inhibition of AP-1 functions may sensitize cancer cells with activated ERK1/2 signalling to growth inhibition and be used to augment the effects of existing anti-cancer drugs [61,62]. Targeted inhibition of the ERK1/2 signalling has received much attention given the presence of activating mutations in receptor tyrosine kinases, Ras G-proteins and BRaf [63–66]. However, therapies targeting mutated BRaf or MEK1/2 proteins invariably lead to drug resistance due to alternative mechanisms to activate ERK1/2 and recent efforts to directly inhibit ERK1/2 proteins are being tested [10,67,68]. Current small-molecule kinase inhibitors being used or evaluated in the clinic act by competing with ATP binding or targeting allosteric sites in the catalytic domain that causes complete kinase inhibition. A frequent cause of resistance to kinase inhibitors is the development of mutations in the ATP binding and catalytic site, which subsequently reduce inhibitor interactions and efficacy .
The present study provides evidence for a new class of compounds that selectively inhibit ERK1/2 signalling through F-site-containing substrates, including Fos family proteins and c-Myc, but have limited effects on D-domain-containing substrates, such as c-Jun or RSK1 (Figures 4, 5 and 7). In addition, other IEGs that contain undefined docking domains and are regulated by ERK1/2, such as Egr1, were also unaffected by the lead compounds (Figures 4, 5 and 7). Although the structural studies (Supplementary Online Data) and in vitro kinase assays (Figure 6) suggest the lead compounds target the FRS on ERK2, future studies will need to evaluate whether other proteins are targeted in cells. It was particularly encouraging to observe selective inhibition of the F-site-dependent Thr325 phosphorylation site with SF-3-030 in the in vitro kinase assays. Unfortunately, we were unable to detect Thr325 phosphorylation with endogenous c-Fos from cell lysates. Thus, validation that the test compounds inhibit F-site-mediated phosphorylation in cells remains to be obtained. Nonetheless, other phosphorylation sites on c-Fos (Ser32 and Thr232) and Fra1 (Ser265) that were inhibited by the MEK1/2 inhibitor U0126 are also inhibited by SF-3-030 prior to loss of total c-Fos and Fra1 protein (Figures 4, 5 and 7).
The data from the present study suggest that targeted inhibition of F-site-containing ERK1/2 substrates and AP-1-mediated transcription can selectively sensitize melanoma cells to growth inhibition and cell death. This response appears to correlate with the levels of activated ERK1/2 in the cells and is not related to the cell growth rates. It should be noted that the compound's mechanism of action may be different with different types of melanoma. As shown in Table 2, the GI50s values for the compounds are similar for A375 and RPMI7951 cells, despite different expression levels of AP-1 proteins (Figure 8). However, the expression of Fra1 is common to both of these melanoma cell lines and future studies will be aimed at determining whether Fra1 expression is elevated in other ERK1/2-driven cancer cells and whether Fra1 inhibition is sufficient to sensitize these cells to growth inhibition and induction of apoptosis. Given that the compounds show selective inhibition of ERK1/2-mediated signalling events, we propose that specific targeting of ERK1/2 signalling functions could be an effective approach to inhibit the survival of ERK1/2-addicted cancer cells in a way that reduces off-target toxicity and the potential to develop resistance to these selective agents.
There is significant interest in developing new approaches to target protein–protein interactions to improve specificity and efficacy in treating cancers with constitutive ERK1/2 pathway activity due to Ras and Raf mutations . Although the present study supports the identification of new molecules that disrupt ERK1/2-dependent signalling events, other targets of these compounds are likely to be identified. There are only a limited number of studies reporting the discovery of small-molecule inhibitors of MAPK substrate-docking sites. An earlier report using deuterium-exchange MS described the identification of substrate-selective inhibitors of the p38α MAPK as a potential approach to improve regulation of inflammatory responses . These studies described a small-molecule inhibitor that targeted the active site of p38α MAPK and caused perturbations in residues in the DRS region of this kinase, which allowed the inhibitor to discriminate between the D-domain-containing substrates MK2 and ATF2. However, others suggested the differential effects of this compound on MK2 and ATF2 phosphorylation were a function of substrate:kinase stoichiometry and not due to the substrate selectivity of the compound . A previous study indicated that the natural product zuonin A can target the DRS on JNK MAPK and inhibit phosphorylation of c-Jun . Similarly, BI-78D3 was a small molecule identified to compete with a D-domain-containing peptide found on the JNK-interacting protein-1 (JIP1) and inhibits interactions with JNK1 . Other screening approaches have identified novel compounds that bind to exposed allosteric sites on the JNK or p38 MAPK surface and are ATP-independent . Although it is not known whether these compounds have differential effects on substrate proteins, this represents an alternative approach for identifying novel kinase inhibitors with improved selectivity over inhibitors that target highly conserved ATP-binding sites.
Although the present results indicate selective inhibition of the phosphorylation of substrate proteins by the studied inhibitors, selectivity is not absolute and may be stimulus dependent. This is evident in the results showing inhibition of phosphorylation or expression of the D-domain substrates RSK-1 and c-Jun at higher concentrations of test compound following treatment with EGF (Figures 1, 2 and 4). However, RSK-1 and c-Jun phosphorylation or expression was unaffected by the compounds in cells treated with PMA or in melanoma cells containing BRaf mutations (Figures 5 and 7). This highlights the complexity of the signalling pathways and the potential for differential effects of inhibitors that have a common target. As an example, it has been reported that catalytic site inhibitors of MEK have differences in efficacy and effects on ERK1/2 signalling depending on whether the cells contain BRaf or KRas mutations . We have also observed partial specificity with putative DRS-targeted inhibitors, which inhibit ERK1/2-mediated caspase 9 and RSK-1 phosphorylation and sensitize transformed cells to undergo apoptosis . Two factors may contribute to this lack of absolute specificity. First, the interaction between ERK2 and substrate may involve a relatively large portion of the docking domain such that interactions with more than one of the spatially adjacent sites may occur. Although the structure of a full substrate protein bound to ERK2 is not available, structures of D-domain peptides from haematopoietic protein tyrosine phosphatase and MAPK phosphatase-3 are available [77–79]. These peptides bind primarily to the DRS, although the C-terminal regions of the peptides extend towards other putative binding sites. Such additional interactions may be anticipated to lead to compounds binding to a given site on the docking domains to partially block interactions of substrate proteins not primarily interacting with those sites.
In the context of substrate interactions with the FRS on ERK1/2, conformational changes associated with the active phosphorylated form of ERK2 were reported to promote interactions with F-site substrates . Recently, studies using D-domain- or F-site-specific peptides reported that there was little communication between the DRS and FRS on ERK2 when one site was occupied with a peptide . Nonetheless, occupancy at either site with a substrate protein may lead to conformational perturbations on either the structural and/or dynamical levels that regulate the formation of signalling complexes.
In summary, a combination of computational and experimental methods has been used to identify a new class of low-molecular-mass compounds that regulate the expression and function of F-site-containing ERK1/2 substrates involved in AP-1-mediated transcription and cancer cell proliferation. The availability of these compounds will facilitate investigations into the biological function of ERK1/2-mediated signalling using chemical biology and potentially lead to the development of novel therapeutic agents. In addition, the successful identification of ATP-independent substrate-specific inhibitors of ERK1/2 is anticipated to open the door for similar ligand design strategies targeting protein–protein interactions in systems where a given protein has multiple interacting partners.
computer-aided drug design
docking site for ERK, FXF
epidermal growth factor
epidermal growth factor receptor
1, early growth response-1
human embryonic kidney
immediate early gene
c-Jun N-terminal kinase
ligand grid free energy
mitogen-activated protein kinase
p90 ribosomal S6 kinase-1
site identification by ligand competitive saturation
Paul Shapiro and Alexander MacKerell, Jr, conceived and designed the studies and wrote the paper. Ramin Samadani and Jun Zhang performed most of the experiments. Amanda Brophy assisted with the experimental analyses. Taiji Oashi, Deva Priyakumar and Prabhu Raman performed the CADD analysis. Franz St John and Edwin Pozharski helped with setting of the X-ray crystallography studies and structural analysis. Kwan-Young Jung and Steven Fletcher synthesized SF-3-026 and analogue compounds.
Portions of this research were carried out at the Stanford Synchrotron Radiation Lightsource, a Directorate of SLAC National Accelerator Laboratory and an Office of Science User Facility operated for the U.S. Department of Energy Office of Science by Stanford University. The SSRL Structural Molecular Biology Program is supported by the DOE Office of Biological and Environmental Research and by the National Institutes of Health, National Institute of General Medical Sciences (including grant number P41GM103393). The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official views of NIGMS, NCRR or NIH.
This work was supported by the National Institutes of Health [grant numbers R01 CA120215 (to P.S. and A.D.M.) and F31 GM100693 (to R.S.)]; and the University of Maryland Computer-Aided Drug Design Center.
These authors contributed equally to this work.
The structure of MAPK ERK2 complexed with 1,1-dioxo-2,3-dihydrothiophen-3-yl benzenesulfonate has been deposited in the PDB under accession code 3QYI.