As a critical signaling node, ERK1/2 are attractive drug targets, particularly in tumors driven by activation of the MAPK pathway. Utility of targeting the MAPK pathway has been demonstrated by clinical responses to inhibitors of MEK1/2 or RAF kinases in some mutant BRAF-activated malignancies. Unlike tumors with mutations in BRAF, those with mutations in KRAS (>30% of all cancers and >90% of certain cancer types) are generally not responsive to inhibitors of MEK1/2 or RAF. Here, a covalent ERK1/2 inhibitor, CC-90003, was characterized and shown to be active in preclinical models of KRAS-mutant tumors. A unique occupancy assay was used to understand the mechanism of resistance in a KRAS-mutant patient-derived xenograft (PDX) model of colorectal cancer. Finally, combination of CC-90003 with docetaxel achieved full tumor regression and prevented tumor regrowth after cessation of treatment in a PDX model of lung cancer. This effect corresponded to changes in a stemness gene network, revealing a potential effect on tumor stem cell reprograming.
Here, a covalent ERK1/2 inhibitor (CC-90003) is demonstrated to have preclinical efficacy in models of KRAS-mutant tumors, which present a therapeutic challenge for currently available therapies.
The MAPK signal transduction cascade, consisting of RAS/RAF/MEK/ERK, is activated in a variety of cancers including melanoma, colorectal cancer, pancreatic adenocarcinoma (PDAC), and lung cancer (1). Critical nodes of this pathway, BRAF, MEK1/2, and ERK1/2, have been extensively explored as potential therapeutic targets (1). Although the response rate of patients with V600E BRAF–mutant metastatic melanoma to Vemurafenib was approximately 80% (ClinicalTrials.gov identifier: NCT00405587; ref. 2), the response rate of patients with V600E BRAF–mutant colorectal cancer was approximately 5% (ref. 3 and ClinicalTrials.gov identifier: NCT00405587). Subsequent studies have shown that resistance to BRAF inhibition in BRAF-mutant colorectal cancer is largely attributable to EGFR-mediated reactivation of signaling through the MAPK pathway (4, 5). Despite robust initial responses in BRAF-mutant melanoma patients with prolonged and deep inhibition of the target (6), response is limited to 2 to 18 months (7).
Similarly, encouraging preclinical results were achieved with MEK1/2 inhibitors, yet clinical responses to single-agent MEK inhibition have been limited by restoration of ERK activation via pathway feedback loops or pre-existing/newly acquired mutations in the MAPK pathway (8, 9). Importantly, all signaling through the MAPK pathway converges on ERK1/2, the ultimate signaling effectors of the cascade. Accordingly, BRAF-mutant melanoma cell lines that develop resistance to BRAF inhibition and KRAS-mutant colorectal cancer cells that have acquired resistance to MEK inhibition still respond to ERK inhibition (10, 11).
Tumors with RAS mutations have presented a particular therapeutic challenge. KRAS mutations, which account for approximately 85% of all RAS mutations in cancer, are found in approximately 98%, 32%, and 52 % frequency in PDAC, lung adenocarcinoma, and colorectal cancer KRAS mutations, respectively (12). In the absence of effective direct inhibitors of KRAS itself, inhibitors of BRAF and MEK have been tested and found to be generally ineffective, at least in part, due to incomplete inhibition of the pathway (8, 13), as well as dose-limiting toxicities of MEK inhibitors alone and in combination with other pathway inhibitors (14). A central node downstream of KRAS, RAF and MEK, and ERK1/2 phosphorylates substrates, which stimulate cell proliferation and inhibit apoptosis. ERK is thus an attractive therapeutic target in tumors with activation of the MAPK pathway (refs. 11, 15–17 and ClinicalTrials.gov Identifier: NCT01781429).
The present study characterizes a covalent ERK1/2 inhibitor, CC-90003, in terms of potency, specificity, and in vitro and in vivo activity in xenograft and patient-derived xenograft (PDX) models of PDAC, lung cancer, and colorectal cancer. A novel target-binding assay was used to uncover mechanisms of response and resistance. Finally, this study demonstrates that CC-90003 alone and in combination with docetaxel regulates a stemness gene network in a KRAS-mutant lung cancer PDX model, and combination treatment in vivo induces tumor regression and prevents tumor regrowth, following discontinuation of treatment.
Materials and Methods
All cell lines used in this study were obtained from the American Type Culture Collection (ATCC) and were used for experiments within 15 passages of authentication. Cell propagation was in accordance with the ATCC guidelines. Cell media and components were supplied by the ATCC.
Antibodies for Western blot and Meso Scale Discovery (MSD) assay were as follows: primary antibodies were acquired from Cell Signaling Technology, Santa Cruz Biotechnology, Millipore, Abcam, and BD Transduction Laboratories (as listed in the Supplementary Information). The secondary anti-mouse and anti-rabbit antibodies, used for Western blotting conjugated with fluorescent probes, were obtained from LI-COR Technologies. Secondary antibodies used in MSD assays, conjugated to Sulfotag, were purchased from Meso Scale Discovery.
CC-90003 was internally synthesized. All other kinase inhibitors were purchased from Selleckchem and Chemietek and dissolved in dimethyl sulfoxide (DMSO) of 99.9% high-performance liquid chromatography grade (Sigma-Aldrich). The final dilution was performed in the treatment medium. For in vitro experiments, the cells were treated as specified in each figure.
Kinase selectivity profiling was carried out by ActiveX (full method in Supplementary Materials).
Mass spectrometry assays
Intact protein kinases were incubated for 1 hour in an excess of CC-90003 to protein at room temperature. After incubation, each sample was diluted with 10 μL of 0.2% TFA prior to desalting using micro C4 ZipTips and spotted directly onto the MALDI target plate using sinapic acid as the desorption matrix (10 mg/mL in 0.1% TFA:acetonitrile 50:50) for mass spectrometric analysis on an ABSciex 4800 MALDI TOF-TOF fitted with a CovalX HM2 detector. Either human transferrin or albumin was used as an internal standard. The centroid mass of the target protein in the control sample is compared with the centroid mass of the target protein incubated with compound. A shift in the centroid mass of the treated protein compared with the untreated protein is divided by the molecular weight of the compound. The HM2 detector of the MALDI was used with the following settings: HV1, 2.7 kV; HV2, 20 kV; 20 ns bin size; and the laser intensity set to a constant 6,000 units.
Assessment of cell proliferation
The cells were plated at a density of 3,000 cells/well in 90 μL of growth media on 96-well clear bottom black-well plates (Corning; Cat# 3904) and incubated overnight under standard cell culture growth conditions at 37°C with 5% CO2. The following day, one plate for each cell line was used for “Day 0” cell growth control readout, whereas others were treated with 9-point 3-fold dilutions of one compound or combination of compounds, and a DMSO control. Each concentration was tested in triplicate. Cell viability was measured 72 hours later according to CellTiter Glo (Promega) reagent according to the manufacturer's specifications. CellTiter Glo reagent signal was read on the spectramax L luminescence detector, and data were processed using Excel and Prism software.
PDX ex vivo culturing in soft agar (3D tumor clonogenic assays)
Dissociated patient-derived tumor cells were grown in multilayer soft-agar growth conditions (details in Supplementary Materials). Twenty-four hours after seeding, test compounds were added into the culture medium at 10 concentrations ranging between 10 and 0 μmol/L, as above, for the duration of the experiment. Every dish included six untreated control wells and drug-treated groups in duplicate at 10 concentrations. Sunitinib at a toxic concentration of 100 μmol/L was used as positive control for inhibition of colony formation. Cultures are incubated at 37°C and 7.5% CO2 in a humidified atmosphere for 8 to 13 days and monitored closely for colony growth using an inverted microscope. Within this period, ex vivo tumor growth leads to the formation of colonies with a diameter of >50 μm. At the time of maximum colony formation, the cells are stained with 2-(4-iodophenyl)-3-(4-nitrophenyl)-5-phenyltetrazolium chloride (1 mg/mL, 100 μL/well), and colony counts are performed with an automatic image analysis system. Sigmoidal concentration–response curves are fitted to the data points using 4-parameter nonlinear curve fit using Oncotest Warehouse Software.
HCT-116 study female athymic nude mice were inoculated subcutaneously with 5 × 106 HCT-116 cancer cells. Twelve days after implant, when the tumors were in the range of 108 to 126 mm3, the mice were randomized and treated once or twice daily orally with vehicle (5% DMSO/15% Solutol/80% PBS) or CC-90003 (free base, Lot #10). Paclitaxel was used as a positive control.
PDX TCA and in vivo studies
PDX models were obtained from patient biopsies and grown in NMRI nu/nu mice or cryopreserved until initiation of studies. In vitro soft-agar colony formation studies (tumor clonogenic assay) as well as in vivo studies of PDX tumors grown in immunocompromised mice were conducted by Oncotest GMBH/Charles River Discovery, as described in Supplemental Materials and Methods.
Nanostring stem cell assay
RNA was extracted using an RNeasy mini kit (Qiagen). Note that 70 ng total RNA was used in each assay using Nanostring Human Stem Cell panel. Data analysis and scoring of significant hits was performed using nCounter software.
LXFL 1674 cells (150,000 cells/well) were plated in 6-well ultra-low attachment plates (Corning #3471) using MammoCult medium (Stemcell Technologies; #05620), supplemented with hydrocortisone (Stemcell Technologies; #07925) and heparin (Stemcell Technologies; #07980) in the presence or absence of compounds, as indicated in the respective experiments. Once colonies/tumor spheres have formed, representative images of the unstained cultures were taken using Olympus CK40 microscope. Then, spheres will be collected and dissociated with trypsin, followed by washing once with complete media. One fourth of the cells were replated in 6-well ultra-low attachment plates using MammoCult medium for secondary sphere formation, and the rest of the cells (three fourth) were used for Pharmacodynamic (PD) marker assessment. The procedure was repeated for 5 passages.
Gene expression analysis was based on Affymetrix HG-U133 Plus 2.0 platform. mRNA signal values from the CEL files were extracted, and quality analyses were carried out using the "R" statistical computing environment and associated modules from Bioconductor. Jetset was used to select the best probe set for each gene and performed subsequent analysis on the selected probe sets rather than the full array (18). Differential gene expression analysis was performed using the limma package. Copy-number data were based on the Affymetrix Genome-Wide Human SNP array v6.0. Quality control was performed using the Affymetrix Genotyping Console v4.1 (Affymetrix), and samples with a contrast QC above 0.4 and MPAD below 0.35 were excluded from further analysis. Copy-number variations were calculated using the Affymetrix GTC v4.1 and PICNIC (Cancer Genome Project from the Welcome Trust Sanger Institute, http://www.sanger.ac.uk/resources/software/picnic/) software suites. The authors are unable to publish the mutation data pertaining to the proprietary PDX models derived by the third party per legal agreement.
The following software were used: IPA, Prism, Excel, and nCounter.
CC-90003 in a potent and selective covalent ERK1/2 inhibitor
The structural design of CC-90003 was based on the analysis of ERK1 and ERK2 in relation to the entire kinome. The ERK2 active site with CC-90003 bound was modeled (Fig. 1A). The targeted cysteine for covalent binding with CC-90003 (Cys183 of ERK1 and Cys164 of ERK2) is located in the A-loop of the ATP binding site. The covalent binding was confirmed by Mass Spectrometry (Supplementary Fig. S1). A structurally equivalent cysteine is present in 48 of 512 human protein kinases. Further analysis of these 48 kinases showed that kinase selectivity could be improved by applying a substitution on the ligand that excludes two third of these 48 kinases that have large amino acids (such as phenylalanine and tyrosine) at the “dimple” position, adjacent to the hinge, while still allowing the interactions with small amino acids (such as leucine in ERK1 and ERK2 in 15 of the 48 kinases) at this position. Cellular kinase screening confirmed the anticipated kinase selectivity (Fig. 1B).
In biochemical, cellular, and mass spectrometry assays of 347 kinases, CC-90003 was found to strongly inhibit kinase activities of ERK1 and ERK2 with IC50s in the 10 to 20 nmol/L range and had good kinase selectivity. In a 258-kinase biochemical assay panel, significant inhibition of 213 kinases (<50% inhibition), moderate inhibition of 28 kinases (50%–80% inhibition), and >80% inhibition of 17 kinases by CC-90003 were found. In an ActivX cellular kinase screening using A375 BRAF V600E–mutant melanoma cell line, only 5 of 194 kinases (ERK1, ERK2, MKK4, MKK6, and FAK) were inhibited by >80% at 1 μmol/L of CC-90003 (Fig. 1B and C; Supplementary Table S1). At the same concentration, no significant inhibition (<14%) was found in a Cerep panel of 40 nonkinase enzymes and receptors (data not shown). Through our iterative analyses, only 3 kinases, in addition to ERK1/2, were inhibited in cells at biologically relevant concentrations: KDR, FLT3, and PDGFRa (Fig. 1C).
CC-90003 inhibits growth of KRAS-mutant cell line models in vitro and in vivo
Activity of CC-90003 was tested on 240 cancer cell lines in a 3-day proliferation assay and demonstrated potent in vitro antiproliferative activity on cancer cells of various tumor types. As expected, bioinformatics analyses indicated that tumors with BRAF mutations were particularly sensitive to CC-90003. Of the 27 BRAF-mutant cancer cell lines tested, 25 (93%) demonstrated sensitivity to CC-90003 inhibition (GI50 < 1 μmol/L; Supplementary Table S2). CC-90003 was also active against 28 of 37 (76%) KRAS-mutant cancer cell lines, consistent with prior observations with MEK and other ERK inhibitors (19). In many, but not all cases, CC-90003 had cytotoxic effects in KRAS-mutant PDAC, lung cancer, and colorectal cancer cell lines, as indicated by values < 0 (i.e., fewer cells at 3 days than at time zero) and Caspase 3/7 activation (Supplementary Fig. S2). CC-90003 did not significantly inhibit proliferation of normal lung fibroblasts or bronchial epithelial cells (data not shown). CC-90003 modulation of PD markers, including pRSK, Dusp4, pS6RP, and pERK, was assessed at multiple time points and doses in A375 (BRAF mut) and HCT-116 (KRAS mut) cell lines and compared with the effects of other ERK and upstream MAPK inhibitors. Single administration of CC-90003 potently brought down Dusp4 levels and inhibited phosphorylation of ERK for 24 hours (Supplementary Fig. S3).
In order to understand mechanism of action and response versus resistance in several models, the KRAS G13D-mutant colorectal cancer cell line HCT-116 was treated with a range of concentrations of CC-90003 and other ERK inhibitors, GDC-0994, or Vertex 11e. CC-90003 was more potent than GDC-0994 in decreasing growth of HCT-116 cells; unlike either ERK inhibitor at concentrations up to 10 μmol/L, CC-90003 induced cell death starting at 1 μmol/L (Fig. 2C). In addition, we compared CC-90003 potency with that of two published ERK inhibitors GDC-0994 and BVD-523 in a panel of 6 KRAS- and BRAF-mutant lung cancer cell lines. CC-90003 demonstrated potency that was superior or comparable with that of BVD-523 and a significant improvement over GDC-0994 in each of the lines tested (Supplementary Fig. S4).
In in vivo studies of an HCT-116 xenograft model, CC-90003 was well tolerated at a range of doses (12.5 mg b.i.d.–100 mg qd), although doses of 50 mpk b.i.d. and 75 mpk b.i.d. group caused mortality by days 6 to 18 of study (data now shown). Both dosing schedules (qd and b.i.d.) led to tumor growth inhibition (TGI; Fig. 2D). Because 50 mpk qd was the minimally efficacious and tolerated dose in this xenograft model (65% TGI), this dose/schedule was used for subsequent in vivo studies. The corresponding TGI values and PK parameter AUC are shown in Supplementary Table S3.
CC-90003 decreases colony formation ex vivo and inhibits tumor growth in vivo of three KRAS-mutant PDX models
To further assess activity of CC-90003 in PDXs, we evaluated response of 84 KRAS-mutant PDX models of PDAC, non-small cell lung cancer, or colorectal cancer in a 3D colony formation assay (Supplementary Fig. S2). This assay was chosen based on reports that response of KRAS-mutant tumor cells to MAPK inhibitors in 3D may correlate more closely with in vivo efficacy (16, 20–22). Significantly, 18 of 84 models responded with absolute IC50 values below 1 μmol/L, which is comparable with exposures at clinically (and preclinically) tolerable doses (Supplementary Fig. S5; ref. 23).
Three PDX models with sensitivity to CC-90003 in vitro [PDAC model PAXF-2059 (IC50 0.621 μmol/L), lung cancer model LXFA-983 (IC50 0.8 μmol/L), and colorectal cancer model CXF-243 (IC50 0.586 μmol/L)] were tested for in vivo response to CC-90003 (Fig. 3A). CC-90003 (50 mg/kg, qd) caused tumor stasis in models of lung cancer (LXFA-983) and PDAC (PAXF-2059), with minimal body weight change (Supplementary Fig. S6). A higher dose of CC-90003 (100 mg/kg, qd) did not show significantly better efficacy. The finding that the colorectal cancer model (CXF-243) was less responsive than the other two models tested provided an opportunity to evaluate the biology of in vivo response and potential resistance to CC-90003 (Fig. 3B).
Disconnect between ERK occupancy and efficacy of CC-90003 in a colorectal cancer PDX model reveals potential mechanism of resistance to ERK inhibition via increased signaling through the MAPK axis and engagement of parallel signaling pathways
To assess CC-90003 inhibition of ERK and downstream effects on MAPK signaling, we assessed ERK occupancy, using a novel assay, created to measure direct target modification by CC-90003. Data are plotted as percent free ERK or the fraction of ERK which is not bound by drug. The amount of free ERK in tumor samples from in vivo studies collected 4 hours after the last dose was reduced to 31.3% in lung, 23.6 % in pancreatic, and 14.2 % in colorectal PDX models; which did not track with efficacy of CC-90003 in the CXF-243 model (Fig. 3C).
To investigate this disconnect, we evaluated whole-exome sequencing and gene expression arrays (Affymetrix) of the three PDX models. Although mutation data were not instructive with regard to the relative resistance of the CXF-243 model (data not shown), gene expression levels suggested increased activation of the MAPK and other pathways (data not shown). Several known downstream effectors of MAPK as well as JNK pathways, including MEK1/2, were differentially regulated in the CXF-243 model. Ingenuity Pathway Analysis (IPA) showed four activated gene networks, which involved MAPK and JNK/Jun signaling nodes (Fig. 4A, full version Supplementary Fig. S7). Two additional colorectal cancer PDX models, CXF-975 and CXF-1034, which were intrinsically resistant to CC-90003 in ex vivo assays, also shared a similar gene expression and activated gene network (Supplementary Fig. S8).
In addition to ERK 1/2, JNK/JUN and MSK signaling axes can regulate cell survival and proliferation, via activation of transcription factors ATF-1 and CREB (24, 25). Thus, MEK overexpression and flux through converging JNK and MSK pathways may contribute to the decreased efficacy of CC-90003 in vivo in CXF-243 model.
Western blot analyses of PDX tumors collected 4 hours after the last dose showed increased expression of EGFR, MEK, and RSK proteins in CXF-243 relative to other models tested (Fig. 4B, quantification; Supplementary Fig. S10). As expected, pEGFR and pMEK levels were increased after CC-90003 treatment of LXFA-983 and CXF-243 tumors. Although CC-90003 decreased phosphorylation of ERK (pERK) in PDAC PAXF-2059 and lung LXFA-983 PDX models, pERK levels were increased in CXF-243 after treatment with CC-90003. CC-90003 decreased pRSK in all models; however, remaining levels of pRSK were higher in CXF-243 than in the other two PDX models. These data demonstrate that the signaling flux through the MAPK pathway is heightened in the CXF-243 model at baseline so that CC-90003 is unable to sufficiently inhibit the pathway.
Baseline MSK and pCREB/pATF were higher in CXF-243, and total and phospho-cJun were increased after CC-90003 treatment in CXF-243 but not the other PDX models. These increases in JNK/Jun and MSK signaling may explain maintained function of common downstream effectors, which promote cell survival and proliferation in CXF-243 tumor cells, despite CC-90003 engagement of ERK.
Taken together, our data reveal a novel mechanism of resistance to ERK inhibition in a KRAS-mutant colorectal cancer PDX model (mechanism summary, Fig. 4C).
To test whether inhibition of JNK and or/upstream nodes in the MAPK pathways that were specifically hyperactivated in the CXF-243 model may be effective at overcoming ERK inhibitor resistance, an ex vivo treatment of CXF-243 in a 3D ultra-low attachment setting was conducted using pharmacologic inhibitors of EGFR and JNK signaling, Afatinib, and JNK-IN-8, respectively. As shown in Supplementary Fig. S9A, the levels of respective PD markers were effectively modulated by each drug as well as the combination of the inhibitors. Addition of Afatinib downregulated MAPK signaling, and an additive effect was observed on the viability of the cells. Although JNK-IN-8 successfully downregulated the levels of p-Jun, the inhibitor had no single-agent effect on the viability of the PDX, and combination with CC-90003 treatment did not show any advantage over CC-90003 alone (Supplementary Fig. S9B). We thus conducted an in vivo study combining CC-90003 and Afatinib in CXF-243 model. Addition of JNK inhibitor was not merited by the ex vivo data, and a triple combination was expected not to be tolerable. Afatinib alone showed no efficacy; however, combination with CC-90003 produced a moderate additive effect (Supplementary Fig. S9C). The treatment continued till day 35; however, multiple animals had to be taken off study in multiple groups due to tumor burden; thus, the data are shown through the last day on which all animals were on study. The PD markers were assessed 4 hours after last dose on day 35 (Supplementary Fig. S9D).
Our data indicate that although combining treatments within EGFR–ERK axis shows some efficacy in this ERK inhibitor resistance model, targeting JNK may not be a sufficient strategy and thus finding ways of targeting MSK1, which is constitutively active and currently has no specific tools for pharmacologic inhibition or degradation, may be key.
Combination of CC-90003 and docetaxel induces full regression and prevents regrowth of KRAS-mutant lung PDX model through modulation of a stemness program
Preclinical and clinical data suggest that the combination of MAPK inhibition upstream of ERK with standard chemotherapeutic agents that induce microtubule stabilization or DNA damage leads to increased efficacy and prolonged survival (26, 27). Sequential administration of the MEK inhibitor, Selumetinib, preceding or following docetaxel treatment, in a xenograft model, has shown differential effects resulting from the order of administration, presumably arising from effects on cell-cycle progression: G1 versus mitotic arrest, respectively (26). In addition, metronomic treatment with Vemurafenib has been shown to prolong response in treatment of a xenograft model bearing V600E BRAF–mutant melanoma (28). The importance of combinations and scheduling was shown recently for a triplet treatment (BRAF, MEK, ERK inhibitors) of BRAF-mutant tumors (16, 29). We have thus explored two CC-90003 scheduling paradigms: continuous treatment with docetaxel coadministered on days 0 and 7, inducing continuous pressure on MAPK signaling and cell-cycle progression, versus intermittent CC-90003 treatment (4 days on, 3 days off) with docetaxel being dosed on days 5 and 12, hypothetically causing cells to synchronize in G1 cell-cycle arrest, followed by induction of docetaxel-induced mitotic arrest. A rapidly growing lung PDX (median doubling time 4.3 days) resistant to MAPK inhibition alone (data not shown) was chosen to explore potential benefits of this combination treatment in a model that may reflect the biology of a late-stage highly proliferative tumor.
Combinations of CC-90003 and docetaxel were well tolerated with minimal to no effect on body weight in either dosing regimen (Supplementary Fig. S11). The continuous coadministration of CC-90003 with docetaxel on days 0 and 7 suppressed tumor growth following the first dose and rapidly induced tumor regressions reaching full remission at day 18 (Fig. 5A). Intermittent combination dosing induced tumor stasis by day 14 and by day 18 regression, which was less profound than the continuous treatment regimen (maximum 27.7% of starting tumor volume on day 25). This effect was superior to other single agent treatments or combinations. The single-agent docetaxel treatment on days 0 and 7 also resulted in tumor regressions, which reached their maximum at day 25. Both single-agent CC-90003 dosing schedules induced minimal tumor growth suppression, demonstrating that scheduling paradigms do not affect single-agent efficacy of CC-90003 in this model.
Although four tumors in the intermittent combination or single-agent docetaxel arms started to regrow immediately following treatment discontinuation (Fig. 5A, Supplementary Fig. S12), no measurable regrowth of tumors was found among animals in the constant dosing combination arm. These results indicate that continuous inhibition of MAPK pathway signaling in combination with mitotic arrest caused by docetaxel is more efficacious than an intermittent dosing paradigm, and that this combination has the potential for inducing durable and profound tumor regression in fast growing KRAS-mutant models.
Because tumor regrowth following treatment has been linked to the presence of treatment resistant cancer stem cells (30) and MAPK signaling has been previously linked to maintenance of the stem cell compartment in colorectal cancer and other tumors (31–33), we next investigated the expression of “stemness” genes in tumor cells following single-agent CC-90003 treatment or in combination with docetaxel. Because no tumors were available for analysis from the concomitant combination arm of the study, only tumors from the intermittent combination cohort could be analyzed. Four genes associated with stemness—CCND2, DX1, CXCL12, and ALDH1A1—were significantly downregulated by both CC-90003 treatment regimens, compared with vehicle control (Fig. 5B). Expression of these genes returned to baseline levels in tumors that regrew after combination treatment was stopped. The pattern was reversed for WNT5A gene expression. Unbiased IPA analysis of gene expression changes identified a single signaling network, which was connected to ERK/MAPK signaling (Fig. 5C). These results suggest that CC-90003 may act on the transcriptional program of stemness within the treated tumor, and that regrowth following treatment discontinuation is associated with outgrowth of a clone within the cell population that is able to overcome CC-90003–induced changes in stemness.
To further understand the molecular underpinnings of this response, we used an in vitro assay based on MammoCult media that preferentially allow growth of stem cell–derived colonies from a heterogeneous cell population, and thus allow us to explore effects of CC-90003, docetaxel, and the combination of these treatments on the stem cell component of the PDX model. Cells derived in culture from LXFA-1674 PDX were cultured in MammoCult medium -/+ CC-90003, docetaxel, or a combination of the two, in nonadherent conditions. Cells were imaged at day 7, collected and dissociated from the colonies, and split with 25% of each population being replated and 75% used for marker analysis (workflow summary, Fig. 6A).
Docetaxel alone had little/no effect on sphere formation. In contrast, colony size and density were decreased by CC-90003 at 1 μmol/L, the physiologically relevant dose corresponding to Cmax in blood at 50 mg/kg in in vivo treatment and in patients treated with the MTD of 120 mg/day (23). No material could be collected at this dose level following passage 2, indicating elimination of stem cells in the cell population. In subsequent passages (up to 5 total), lower concentrations of CC-90003 also suppressed colony growth. Addition of docetaxel allowed for effects of CC-90003 at lower doses to be demonstrated after 2 passages. Representative images from passages 1 to 3 are shown in Fig. 6B.
Next, we assessed samples collected from each treatment condition for 5 passages by Nanostring Human Stem Cell assay, as previously described. The RNA yield from samples treated with 1 μmol/L CC-90003 was too low, and thus these were excluded from analysis. Analysis of genes significantly modulated >2x by either treatment revealed a group of 36 genes modulated at passage 1 (Fig. 7A). Effects of CC-90003 at each concentration were closely paralleled by those of the same dose of CC-90003 combined with docetaxel, thus indicating that the overall impact on stemness gene expression in this assay was attributable to ERK inhibition. Starting at passage 2, the group of significantly regulated genes was reduced to 16, and at subsequent passages, the number of genes sequentially increased to 33, indicating a possible early effect, followed by loss of a cell population and emergence of another subclone within the model system. In later passages, a subgroup of genes in which changes were more significant in the higher dose combination treatment group compared with CC-90003 and docetaxel treatment alone were observed (Fig. 7A, red outline), indicating a potential additive or synergistic effect of the combination.
Each group of genes was analyzed via IPA, and the top signaling networks observed at each passage contained ERK as one of its key regulators (Fig. 7B). This result supports validity of the analysis, which was agnostic of the CC-90003 drug target. Notch signaling was scored as one of the top canonical pathways for all 5 passages, the engagement of which was predicted by IPA (Supplementary Fig. S13). In addition, passages 1 and 2 also predicted engagement of “human stem cell pluripotency” and “role of nanog in mammalian embryonic stem cell pluripotency”, whereas instead “regulation of epithelial- mesenchymal transition pathway” was scored higher for subsequent passages, attesting to possible clonality effect in the population over time.
Taken together, our findings suggest that ERK inhibition is responsible for viability and signaling effects on stem cell subpopulation in KRAS-mutant lung cancer PDX model with possible benefit of cotreatment with docetaxel, as self-renewal potential of stem cells comes under additional selective pressure at later passages in the MammoCult system, modeling clonal selection that occurs in tumors during long term treatment in vivo. These data indicate that in a patient population refractory to taxane treatment or MAPK inhibition alone, a combination strategy that takes advantage of the observed effect on tumor stemness may be beneficial to prolong responses and potentially prevent tumor regrowth.
Aberrant activation of the MAPK signaling pathway occurs in many cancer types. Among these, mutations in KRAS, occurring in lung cancer, PDAC, and colorectal cancer, have proven particularly challenging to treat. In part, the challenges are due to the complexity of KRAS signaling, the presence of feedback loops, and signaling interplay among RAF isoforms. Despite this complexity, MAPK signaling converges on a single regulatory node, ERK.
The present study demonstrates that the covalent ERK1/2 inhibitor CC-90003 effectively binds to and shuts off signaling through its target in KRAS-mutant xenografts and PDX models, both in vitro and in vivo. Covalent binding of CC-90003 to its target allows for specific, potent, and durable inhibition of ERK, which has been shown to be essential for other MAPK pathway inhibitors in order to successfully exert their function in tumor tissues with activating mutations in this signaling axis. CC-90003 was well tolerated in the preclinical models at the dose levels necessary to achieve extensive ERK inhibition, thus allowing for detailed exploration of the biological underpinnings of ERK inhibitor response and resistance in models that represent KRAS-mutant cancer subtypes.
MAPK signaling can be hyperactivated via multiple mechanisms, including MEK amplification and increased EGFR expression. In addition, modulation of downstream effectors responsible for cell survival and proliferation can be achieved through signaling by contributing pathways via JNK/JUN and MSK. In the later signaling context, as in the CXF-243 model, 85% inhibition of ERK by CC-90003 was not sufficient to fully inhibit tumor growth. Identification of additional signaling inputs contributing to this mode of resistance demonstrates possible value in evaluating baseline levels of cooperating pathways and has implications for potential clinical utility of combining MAPK and other pathway inhibitors (8, 9). However, our experiments also indicate that in this model of resistance, combinations of ERK and JNK inhibitors may not be sufficient to suppress tumor growth, but rather multiple treatments within the MAPK pathway combined with new strategies to target MSK1 and/or transcriptional activators directly downstream may be required.
Although mechanisms of resistance to MAPK inhibitors have been previously described in BRAF-mutant malignancies and generally occur through reactivation of the pathway via amplification and mutation events (34), mechanisms of resistance in KRAS-mutant forms of cancer are less well understood. The multifactorial resistance mechanism described here may represent the complexity of resistance to MAPK inhibitors that occurs in KRAS-mutant colorectal cancer. Engagement of parallel pathways, such as JNK/JUN axis, which has been previously shown to play a role in evolution of melanoma resistance to BRAF inhibition (8, 9), may shield downstream effectors connected to MAPK pathway from upstream inhibition, even at the convergence point of ERK. Although a recent study indicates that MSK1 expression correlates with colorectal cancer aggressiveness and poor prognosis, the role of MSK1 as well as other parallel inputs in MAPK inhibitor resistance in this disease has not been extensively explored to date (35). Thus, combination treatments that include inhibitors within MAPK pathway as well as parallel signaling axes may be necessary to fully address the complexity of resistance in KRAS-mutant malignancies, such as those represented by the CXF-243 KRAS-mutant colorectal cancer PDX herein.
Lastly, we demonstrate that a combination of CC-90003 with docetaxel in a fast-growing KRAS-mutant lung cancer model, known to be poorly responsive to MAPK inhibition, was able to cause full tumor regression with no recurrence over a 52-day period, when administered concomitantly at the start of treatment, followed by continuation of CC-90003 alone, suggesting a possible induction plus maintenance strategy in the clinic. The lack of tumor regrowth after concomitant combination treatment withdrawal suggested effects on stemness, a hypothesis supported by additional functional and gene expression studies in vitro. The role of ERK in modulating normal as well as tumor stem cell signaling and proliferation has been previously described (31–33); however, inhibition of tumor stemness has remained challenging in in vivo models and patients. The ability to suppress tumor stem cell signaling is particularly important for the combination of MAPK inhibitors with docetaxel and other standard of care taxanes, as Notch signaling and other stemness-associated pathways have been previously implicated in docetaxel resistance (36–38). Thus, the ability of CC-90003 to modulate stemness genes may be key to preventing acquired resistance to taxanes during long-term treatment. Our data also indicate that continuous inhibitory pressure on the MAPK pathway when combined with docetaxel is superior to intermittent CC-90003 therapy in inducing tumor regression and preventing tumor regrowth, consistent with the recent proposal of a fitness barrier (16, 29).
In the clinical setting, identification of patients with stem-cell expression signatures may require baseline profiling of primary tumor tissue and derived cultures. Similar hypotheses addressing EGFR inhibitor resistance have recently gained interest, and efforts are underway to identify diverse evolutionary mechanisms employed by tumors to circumvent response in this signaling context (39–41). The present study supports the rationale for extending such efforts to identification of expression subtypes that may be actionable for personalized ERK inhibitor combinations (39–41).
Taken together, our findings demonstrate that targeting ERK as a convergence node of the MAPK pathway is a successful strategy to inhibit growth of some KRAS-mutant cancer models both in vitro and in vivo. Furthermore, using a covalent, potent, and selective ERK inhibitor, we have dissected the molecular underpinnings of a potential novel mechanism of resistance to single-agent ERK inhibition via increased signaling through the MAPK pathway as well as activation of additional pathways, converging on downstream effector substrates. In addition, we have demonstrated efficient tumor regression induced by CC-90003 and docetaxel coadministration in an ERKi refractory lung cancer PDX model, and proposed a mechanistic basis for combination strategies to overcome resistance and potentially target the cancer stem cell compartment to prevent tumor regrowth.
Disclosure of Potential Conflicts of Interest
I. Aronchik has an ownership interest (including stock, patents, etc.) in Patents and Stocks. L. Qiao, L. Beebe, and T. Shi have an ownership interest (including stock, patents, etc.) in Celgene K. Mavrommatis is a consultant/advisory board member for Discitis Dx. G.L. Bray was Executive Director, Translational Development, at, and has an ownership interest (including stock, patents, etc.) in, Celgene Corp. No potential conflicts of interest were disclosed by the other authors.
Conception and design: I. Aronchik, L. Qiao, L. Beebe, E.H. Filvaroff
Development of methodology: I. Aronchik, Y. Dai, M. Labenski, C. Barnes
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I. Aronchik, Y. Dai, C. Barnes, W. Elis
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): I. Aronchik, Y. Dai, C. Barnes, L. Qiao, L. Beebe, W. Elis, T. Shi, K. Mavrommatis, E.H. Filvaroff
Writing, review, and/or revision of the manuscript: I. Aronchik, L. Qiao, L. Beebe, G.L. Bray, E.H. Filvaroff
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Dai, T. Jones, L. Qiao, M. Malek, W. Elis, K. Mavrommatis
Study supervision: I. Aronchik, L. Beebe, E.H. Filvaroff
Other (performed experimental work): M. Malek
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.