Cancer cells often have deficiencies in cell-cycle control mechanisms and could be dependent on specific cell-cycle checkpoints to maintain viability. Because of the documented role of KRAS in driving replication stress, we targeted the checkpoint governing DNA replication using CHK1 kinase inhibitors in pancreatic ductal adenocarcinoma (PDAC) models and examined mechanisms of resistance.
Single-agent efficacy of CHK1 inhibition was investigated in established and primary PDAC lines. Drug screening was performed to identify cooperative agents. In vitro and in vivo studies were employed to interrogate combination treatment efficacy and mechanisms of resistance.
Many PDAC models evade single-agent inhibition through mechanisms that allow S-phase progression with CHK1 inhibited. Gene expression analysis revealed FOXM1 as a potential marker of CHK1 sensitivity and defined a form of pancreatic cancer with poor prognosis. Drug screen analysis identified WEE1 as a cooperative agent with CHK1 and was effective in cell culture. In vivo experiments validated the combination efficacy; however, resistance could evolve. Resistance was due to selection of a stable subclone from the original PDX tumor, which harbored high baseline replication stress. In vitro analysis revealed that gemcitabine could eliminate viability in the resistant models. The triplet regimen of gemcitabine, CHK1, and WEE1 inhibition provided strong disease control in all xenograft models interrogated.
These results demonstrate the therapeutic resiliency of pancreatic cancer and indicate that coordinately targeting cell-cycle checkpoints in concert with chemotherapy could be particularly efficacious.
Pancreatic ductal adenocarcinoma (PDAC) is an aggressive disease with limited therapeutic options. Cell-cycle checkpoints are perceived as an important target for cancer therapy due to oncogenic stress and deregulated proliferation in tumors. We investigated mechanisms of resistance to single-agent CHK1 inhibition across a wide panel of PDAC cell lines. Most PDAC models were relatively resistant to CHK1 inhibition, although a FOXM1 gene expression signature was associated with response. Drug screening revealed potent cooperation between CHK1 and WEE1 inhibitors. In spite of this potential efficacy, patient-derived xenograft (PDX) models could bypass this combination due to selection of a stable subclone. Perturbation of dNTP pools and replication block were effective in overcoming resistance. Subsequent triple combination therapy with CHK1 and WEE1 inhibitors in concert with gemcitabine was effective in inducing durable control of tumor growth in PDX models. Thus, our data reveal modes of single- and double-agent resistance and provide rationale for coordination of targeted therapies in PDAC.
Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal types of cancer, with an overall 5-year survival rate of approximately 10% (1–3). Although there has been substantial effort in determining an effective therapy for PDAC, most targeted regimens have not demonstrated clinical disease control beyond combination chemotherapy (4). While FOLFIRINOX (oxaliplatin, irinotecan, leucovorin, and 5-FU) and gemcitabine with nab-paclitaxel have increased survival, improvements have been relatively modest and the vast majority of patients succumb to the disease (5, 6). Thus, there is an urgent need for effective new therapies for the treatment of PDAC that consider the intrinsic therapeutic resistance.
Cancer cells harbor a wide spectrum of mutations that drive the cancer phenotype and support uncontrolled proliferation. In pancreatic cancer, activating KRAS mutations are considered the key event in disease etiology and are observed in >90% of cases (7–9). KRAS mutations drive deregulated proliferation and activate survival signals, but can also elicit replication stress, induction of reactive oxygen species, and promote oncogene-induced senescence (10–13). KRAS-induced replication stress is balanced by other aberrations that facilitate disease progression in pancreatic cancer, including loss of CDKN2A or mutation of TP53 (8, 9, 14, 15).
Cell-cycle checkpoint mechanisms play a key role in the tolerance of replication and mitotic dysregulation. Conventionally, cell-cycle checkpoints coordinate the pausing of DNA replication or mitosis with activating DNA repair processes to facilitate recovery and viability in the face of replication stress or under-replicated DNA (16–18). These highly conserved pathways, while discovered in yeast, are highly germane to cancer and are believed to enable the viability of cancer cells in the face of deregulated cellular proliferation (19). The CHK1 kinase plays a critical role in recovery from DNA replication stress by functioning downstream from ATR to coordinate inhibition of DNA replication and repair of strand breaks (20–22). In part, the inhibition of DNA replication is mediated by the suppression of CDK2 activity (23, 24). Another important negative regulator of CDK activity is WEE1 (25–27). While identified as a key regulator of mitotic progression via the control of CDK1 activity, emerging data in yeast and mammalian cells indicate that WEE1 also acts in S-phase to modulate DNA replication checkpoints (27, 28). As such, there is growing literature that the combination of CHK1 and WEE1 could represent an important treatment modality (29–31)
Because cell-cycle dysregulation is a hallmark function of cancer (32), and KRAS is strongly linked to oncogene induced replication stress (33), we interrogated the action of cell-cycle checkpoint inhibitors as therapeutic agents for pancreatic cancer in a wide panel of established and low-passage patient-derived models. The findings illustrate the challenge of treating pancreatic cancer and the importance of rationally designed combination therapy that minimize the selection of resistance.
Materials and Methods
Cell culture and chemicals
Established cell lines (PL45, PANC1, MiaPaca2, YAPC, BXPC3, CAPAN2) were purchased from ATCC and cultured in recommended media. Primary cell lines (EMC828, EMC7310, EMC3226, EMC226) were grown in keratinocyte serum-free (KSF) medium with 0.2 ng/mL EFG, 30 μg/mL bovine pituitary extract (Life Technologies, 10744019), and 2% FBS on collagen-coated (Millipore 08-115) plates. Cells were grown at 37°C with 5% CO2. The 4662 syngeneic model was provided by Dr. Robert Vonderheide's laboratory and cultured as described previously (34). Drugs used in this study were purchased from Selleck Chemicals and DMSO was used as control (Table 1).
|Name .||Target .|
|DNA replication machinery|
|Name .||Target .|
|DNA replication machinery|
ES-FUCCI, a gift from Pierre Neveu (Addgene plasmid #62451), was transfected to EMC226 cell line using Lipofectamine 3000 Reagent (Invitrogen LSL3000001) by manufacturer's protocol. Transfected cells were selected using Hygromycin and sorted with FACSAria.
Cells were seeded in 96-well, 12-well, or 6-well plates, and treated with drugs as indicated for 72 hours. Viability was assessed using CellTiter-Glo Reagent (Promega), and luminescence was read on a Biotek Synergy 2 plate reader. Crystal violet staining was performed after 72 hours on 6-well and 12-well plates. IncuCyte S3 Live-Cell Analysis System (Essen Biosciences) was utilized for live-cell analysis. Primary PDAC cells were labeled with H2B-GFP and seeded in 96-well collagen-coated plates. Cells were treated with DMSO, CHIR-124, and AZD7762 for 72 hours. Two-phase and fluorescent images per well were captured hourly for 72 hours at 10× magnification. Essen Bioscience software was used to quantify number of cells per well and normalized to DMSO-treated cell proliferation. Data were exported to Prism 7 (GraphPad) for statistical analyses. Each drug treatment was performed in quadruplicate and verified in three independent experiments.
Gene expression analysis
Mutations and copy number variations were assessed from whole-exome sequencing for the primary cell lines as described previously (35). Mutation data were collected for the established cell lines from the Cancer Cell Line Encyclopedia (CCLE). Note that PL45 did not have mutation data listed, thus KRAS and other gene-variant status was unable to be determined. Cell lines were ordered on the basis of CHK1 sensitivity data. Gene expression for the primary cell lines was determined by RNA sequencing (35), which was log2 transformed and normalized using the edgeR R package (36, 37). Gene expression for the established cell lines was obtained from the cancer cell line encyclopedia. Z-scores were calculated for the primary and established cell lines separately, then combined into a single data matrix. A correlation heatmap was generated for genes with correlation coefficients between the z-score expression and sensitivity (AUC values) greater than ±0.5 for either treatment. Genes in the FOXM1 gene signature were selected by using the correlation coefficient between cell line gene expression and sensitivity greater than ±0.4 for both the AZD7762 and CHIR-124 treatments, as well as a correlation coefficient greater than 0.4 between The Cancer Genome Atlas (TCGA) PDAC gene expression and TCGA FOXM1 expression. Hierarchical clustering using Manhattan distance was performed with z-score–centered gene expression and low, intermediate, and high FOXM1 signature expression groups were defined on the basis of the resulting clustering. The Kaplan–Meier plot was generated, comparing each of the defined groups and a log-rank P value was calculated. All clustering and survival analysis was performed in R.
Synergy scores using the Bliss independence model were calculated using the SynergyFinder Web application.
Parental and resistant EMC226 cell lines were cultured and provided to Roswell Park Pathology Network Services for spectral karyotyping (SKY) analysis. Fragile site analysis was performed for chromosome number and abnormalities.
Cells were reverse transfected with Dharmacon Human ON-TARGETplus siRNA: CHEK1 (L-003255-00-0005), WEE1 (L-005050-00-0005), FOXM1 (L-009762-00-0005), UBB (L-013382-00-0055), and nontargeting siRNA (D-001810-10-05). Transfection was done using Lipofectamine RNAiMax Transfection Reagent (Invitrogen, 13778150) according to manufacturer's protocol. Cells were then analyzed for viability using CellTiter-Glo. Immunoblot analysis was performed to confirm knock down.
Flow cytometry cell-cycle analysis
For BrdU analysis, cells were treated with CHIR-124 for 48 hours. Before harvesting, cells were pulsed with 5-bromo-2-deoxyuridine, (BrdU, Sigma) for 1 hour. Cells were trypsinized after BrdU pulse and fixed in 70% EtOH overnight in 4°C. Cell pellets were washed with IFA buffer (1× HEPES, 4% FBS, and 0.1% NaN3) and then with IFA + 0.5% Tween20. Pellets were incubated in FITC-conjugated anti-BrdU (BD Pharmingen) or FITC anti-H2AX (phospho-Ser139) for 1 hour at room temperature. Cells were resuspended in propidium iodide and RNAse before analysis in FACSCanto II flow cytometer. Cell-cycle analysis was performed with propidium iodide staining after indicated time of treatment. For γH2AX analysis, cells were treated for 48 hours and trypsinized. Cell pellets were fixed and permeabilized using fixation and permeabilization solution. Cells were then stained with anti-phospho H2A.X FITC conjugate and propidium iodide before analysis in flow cytometer. Analysis was done with FlowJo software. Annexin V staining was performed using Annexin V-FITC Apoptosis Detection Kit (Sigma) after 48 hours of treatment. Analysis was done with FlowJo software.
Western blot analysis
Cells were seeded in 60-mm plates and treated the next day with drug. After 48 hours, cells were collected and lysed with RIPA buffer (50 mmol/L Tris, 150 mmol/L NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, sodium orthovanadate, sodium fluoride, EDTA, and leupeptin). Protein concentration was determined by Bradford Protein Assay Dye (Bio-Rad) and 30 μg of protein per sample was resolved by SDS-PAGE. Gel was transferred to Immobilon-P membrane (Millipore). Blots were blocked with 5% milk for 1 hour at room temperature and incubated with primary antibody overnight at 4°C. Incubation in secondary antibody was at room temperature for one hour. Antibodies for specific antibodies detected were: from Cell Signaling Technology (PARP, CHK1, LC3B and CyclinB1, CDK2). Antibodies from Santa Cruz Biotechnology were GAPDH, B-Actin, CDK1, pCDK1 (Tyr15), Cyclin A, WEE1, and FOXM1. Abcam antibodies used were RPA32 (pT21) and RPA32.
Cells were seeded on glass coverslips and treated with drug (AZD7762, CHIR-124, MK1775) for 48 hours. Cells were washed in PBS and fixed in 4% paraformaldehyde. Then, they were permeabilized in 0.5% Triton X-100. For PCNA staining, cells were preextracted with CSK buffer (10 mmol/L HEPES, 300 mmol/L sucrose, 100 mmol/L NaCl, and 3 mmol/L MgCl2) and extraction buffer (50 mmol/L NaF, 0.1 mmol/L sodium orthovanadate, 1 mmol/L phenylmethylsulfonylfluoride, 0.5% Triton X-100, protease inhibitor). After permeabilization, cells were blocked in IF buffer (5% BSA, 0.4% NP40 in PBS) and incubated with primary antibody diluted in IF buffer for one hour in 37°C. Primary antibodies used in this study were: γH2AX (phospho-ser139) from Cell Signaling Technology, MCM7 (sc-9966) from Santa Cruz Biotechnology, anti-phospho-Histone H3 (ser10; 06-570) from Millipore, and RPA32 (pT21) from Abcam. Coverslips were washed in PBS and secondary antibody diluted in IF buffer was applied for one hour in 37°C. Coverslips were washed again after secondary and mounted on slides. Images were taken with either Leica Confocal Microscope at 63× magnification or fluorescence microscope at 40×. Quantification was performed using ImageJ.
In vivo xenograft studies
All animal studies were approved by Roswell Park Cancer Institute Institutional Animal Care and Use Committee. NSG mice were subcutaneously implanted with early passage patient-derived xenograft (PDX) tumor fragments. Mice were randomized to control, AZD7762 (AZD), MK1775 (MK), and combination (AZD and MK) groups when tumor volumes reached 150–200 mm3. In the control group, mice were administered with vehicle. AZD7762 group was administered AZD7762 (35 mg/kg) via intraperitoneal (i.p.) injection. Combination group was treated with MK1775 (30 mg/kg, gastric gavage) and AZD7762 (35 mg/kg, i.p.). AZD7762 was prepared in 11.3% 2-hydroxyproply-β-cyclodextrin (Sigma) and sterile saline. MK1775 was dissolved in 0.5% methylcellulose. Tumor size was measured every other day, and volume was calculated per the following equation: V = 0.5 × ([greatest diameter] × [shortest diameter]2). For single and combination treatment, mice were randomized to control, AZD7762, MK1775, and combination of AZD7762 and MK1775 (AZD + MK). For regiments with gemcitabine, mice were randomized to control, gemcitabine, combination of gemcitabine and AZD7762 (Gem + AZD) or gemcitabine and MK1775 (Gem + MK), and triple therapy of gemcitabine, MK1775, and AZD7762 (Gem + AZD + MK). Gemcitabine was prepared with sterile saline. Treatment lasted for 21 days, and tumors were harvested at end of treatment, or when tumor volumes reached 2,000 mm3. For orthotopic studies, 5 × 105 EMC226 or 4662 cells were injected into the pancreas of NSG and C57BL/6J mice, respectively. Mice were scanned with MRI for baseline volume at the imaging core facility at Roswell Park Comprehensive Cancer Center. Mice were then randomized to control or triple therapy (Gem + AZD + MK) treatment groups. Upon completion of treatment, tumors and major body organs (liver, lung, small intestine, and kidney) were harvested and processed for histologic evaluation. IHC for Ki67 was performed using Ki67 antibody (Thermo Scientific; catalog no. RM-9106-S1, 1/200 dilution) on DAKO Omnis autostainer. For the evaluation of drug toxicity, organs were formalin-fixed and paraffin-embedded. Tissue sections were stained with hematoxylin and eosin (H&E). In each organ, cell damage (e.g., cell ballooning, apoptosis, necrosis) and degree of inflammation were evaluated.
PDAC cells exhibit a range of responses to CHK1 inhibition
Targeting DNA replication checkpoints has emerged as a therapeutic approach that could have broad application to tumors that are proliferating under substantial oncogene-induced replication stress (38, 39). To investigate the effect of single-agent CHK1 inhibition, we subjected established PDAC cell lines (PL45, PANC1, MiaPaca2, BXPC3, YAPC, and CAPAN2) and primary PDAC cell lines (EMC3226, EMC226, EMC7310, EMC810, EMC828, EMC519), to two CHK1 inhibitors, CHIR-124 and AZD7762. Established cell lines were also treated with additional CHK1 inhibitor, prexasertib (LY2606368). We observed that both established and primary cell lines had variable sensitivities (Fig. 1; Supplementary Fig. S1). In the established lines, PL45 and MiaPaca2 were more sensitive, whereas PANC1 and YAPC lines were more resistant (Fig. 1A and B). To ensure that the sensitivity was not due to off-target effects, we utilized CHK1 siRNA and confirmed PL45 sensitivity and PANC1 resistance to CHK1 inhibition (Fig. 1C). Primary cell lines were also grouped into sensitive and resistant categories (Fig. 1D). To monitor effect on overall cell growth, primary cell lines were transfected with H2B-GFP and treated with CHIR-124 and AZD7762 for live-cell analysis (Fig. 1E). As expected, growth was inhibited in the sensitive lines EMC226 and EMC3226, whereas the resistant lines EMC828 and EMC7310 growth rates did not change with treatment. Thus, there is a diversity of response to CHK1 inhibition and understanding the basis of resistance could yield rational combination therapies.
CHK1 inhibition induces replication stress in sensitive lines
To understand the potential mechanism of resistance, we first examined the level of DNA damage upon CHK1 inhibition. Flow cytometry analysis of γH2AX, an indicator of DNA damage, was utilized to investigate differences between sensitive lines (PL45, EMC226) and resistant lines (PANC1, YAPC). We observed that there was greater induction of γH2AX in the sensitive lines compared with resistant (Fig. 2A). Immunofluorescence analysis also confirmed flow cytometry findings, where sensitive lines had a greater percentage of γH2AX-positive cells than resistant line PANC1 (Fig. 2B). Furthermore, annexin V staining indicated that the sensitive lines (PL45 and MiaPaca2) had greater percentage of apoptotic cells upon CHK1 inhibition than resistant (PANC1 and YAPC) lines (Fig. 2C). Although aberrations in nuclear morphology were observed in cells that were treated with CHK1 inhibitors, this event surprisingly did not correlate with cytotoxic response (Fig. 2D). To determine whether the levels of DNA damage were related to the proliferation of the cell lines, EdU incorporation levels were measured before and after treatment (Fig. 2E; Supplementary Fig. S2). Baseline levels of EdU incorporation were not associated with sensitivity; and in all cell lines, there was a similar inhibition of overall incorporation. Flow cytometry was used to specifically examine cell-cycle stages and DNA synthesis concordantly. Consistent with the EdU incorporation, the overall BrdU incorporation levels between sensitive and resistant lines did not differ. However, the amount of S-phase cells (DNA-content between 2N and 4N) compromised for BrdU incorporation was associated with sensitivity (Fig. 2F). To further explore the DNA replication defect, we performed a time-course treatment and probed for a marker of single-strand breaks, phosphorylated RPA (pRPA). Sensitive cell lines had induction of pRPA at earlier time points and at greater levels than resistant lines (Fig. 2G; Supplementary Fig. S2). Immunofluorescence analysis also revealed greater magnitude of pRPA levels in sensitive lines compared with resistant line (Fig. 2H; Supplementary Fig. S2). Together, the data suggest that sensitive lines are more susceptible to replication stress and DNA damage upon CHK1 inhibition, whereas the resistant cell models essentially do not require CHK1 to pass through S-phase.
FOXM1 is a potential marker of sensitivity
To determine whether the common mutations present in PDAC are associated with sensitivity to CHK1 inhibition, we evaluated the genetics of the cell lines used in the study (Fig. 3A). These data showed no significant relationship between KRAS, SMAD4, TP53, or CDKN2A genetic alterations and susceptibility to CHK1 inhibitor. Correlation analysis of gene expression data from the cell lines was used to investigate gene expression levels associated with response. Using AUC values of CHIR-124 and AZD7762, we first confirmed that there was a strong correlation between genes associated with the response to either drug indicating that the sensitivities were not unique to a particularly drug, but encompassed the class of CHK1 inhibitors (Fig. 3B). We found that multiple genes related to cell cycle and DNA damage repair were correlated with sensitivity (Fig. 3B; Supplementary Fig. S2; and Supplementary Data S1). One of these genes, FOXM1 is a known transcriptional driver for the expression of many of the genes associated with response (40). To determine the functional role of FOXM1 in sensitivity, FOXM1 was knocked down in PL45 and MiaPaca2 lines before treatment with CHK1 inhibitors CHIR-124 and prexasertib. FOXM1 knockdown significantly rescued sensitivity to CHK1 inhibitors (Fig. 3C; Supplementary Fig. S3). We also employed CDK4/6 inhibitor, palbociclib as it has been reported that CDK4/6 regulates the stability of FOXM1 (41). As expected, palbociclib treatment reduced levels of FOXM1 and pretreatment with palbociclib, before CHIR-124 and AZD7762 were added, reduced pRPA levels (Fig. 3D) and increased survival (Fig. 3E). We also interrogated FOXM1 expression levels in TCGA data and found that high expression of FOXM1 signature corresponded with poor overall survival (Fig. 3F). These data suggest that there is a proportion of PDAC cases with poor prognosis that could be particularly responsive to CHK1 inhibitors.
Drug screen reveals cooperation between WEE1 and CHK1 inhibition
Because only a subset of pancreatic cancer cells is sensitive to CHK1 inhibition at reasonably low doses that are clinically achievable, there is the need for combination with other agents to enhance efficacy. To define agents that cooperate with CHK1 inhibitors, we pretreated H2B-GFP–labeled cell lines with DMSO or CHIR-124 for 24 hours and performed a drug screen with over 300 agents (Supplementary Data S2) by live-cell analysis. We identified agents that were cooperative with CHIR-124 in various drug groups such as EGFR inhibitors. However, there was exceptionally potent cooperation with the WEE1 inhibitor, MK1775, present in the compound library (Fig. 4A). Further analysis of the growth rate in two different cell models confirmed that cell proliferation was significantly suppressed by combination treatment with CHIR-124 and MK1775 (Fig. 4B). To ensure that the combination would be effective across more PDAC models a small drug panel was screened for cooperation with CHK1 inhibition by viability analysis. We observed that MK1775 and CHIR-124 were cooperative (Fig. 4C; Supplementary Fig. S4), while among the relatively large spectrum of different chemotherapies and related compounds (i.e., BCL2 and ATM/ATR inhibitors), only gemcitabine elicited a low-dose interaction with CHIR-124. While single-agent MK1775 and CHIR-124 showed little effect, dose–response analysis between MK1775 and CHIR-124 validated potent synergistic low-dose effects in established lines PANC1 and PL45 (Fig. 4D; Supplementary Fig. S4). Synergy scores of MK1775 and CHIR-124 were calculated through the Bliss independence model in a range of primary cell lines to confirm results seen in established lines (Fig. 4E; Supplementary Fig. S4). These results indicate that WEE1 and CHK1 inhibition is a promising combination in PDAC cells.
WEE1 and CHK1 inhibition causes apoptosis through S-phase accumulation and DNA damage
To elucidate the underlying mechanism of cooperation, we first examined proliferation levels through EdU incorporation. Although single agent affected proliferation in some cell lines, inhibition of both WEE1 and CHK1 caused a highly significant decrease in EdU incorporation (Fig. 5A and B). We examined cell-cycle stage through flow cytometric analysis, we found that all tested cells (both sensitive and resistant to CHK1 inhibition) were arrested with an S-phase DNA content, in spite of lack of EdU incorporation with combination treatment (Fig. 5C; Supplementary Fig. S5). Comparison of active chromatin-associated MCM7 levels in single-agent and combination treatment revealed significant decrease upon combined CHK1 and WEE1 inhibition (Fig. 5D) indicating the DNA replication licensing is perturbed (31). In addition, we performed flow cytometry analysis of EMC226 cells transfected with the FUCCI (fluorescent, ubiquitination–based cell-cycle indicator) plasmid (42) to examine features of cell-cycle coordination (Supplementary Fig. S5). Combination treatment led to a population of cells that have neither APC-CDH1 (indicated by unchanged GFP levels) nor SCF-SKP2 (seen by increased RFP levels) activity as would typically occur at the exit of mitosis, although the cells have an S-phase DNA content. To investigate whether DNA damage was occurring in cells undergoing replication, cells were costained with γH2AX and EdU after treatment with single agents and combination. The lack of overlap in the γH2AX and EdU-positive cells indicates the damaged cells are not actively undergoing DNA replication (Fig. 5E). On the basis of these results, we examined whether the treated cells progressed through mitosis. Cells were stained for phosphorylated Ser10 Histone H3 (pH3), 24 hours and 48 hours after treatment with CHIR-124, MK1775, or combination. While control cells and cells treated with MK1775 or CHIR-124 single agent exhibit progression through anaphase (Fig. 5F and not shown), we found that veritably no cells treated with the combination progressed through anaphase or exhibited appropriately condensed chromatin (Fig. 5F). These data suggest that while cells are not actively replicating there are multiple elements of cell-cycle deregulation that are occurring in the cells arrested in S-phase. Further evaluation of DNA damage by immunofluorescence also revealed a dramatic increase in replication stress and strand breaks upon combination treatment (Fig. 5G; Supplementary Fig. S5). To understand the mechanism by which the cells were arrested, immunoblot analysis was performed on cells treated with single agent and combination. In addition, knockdown of CHK1 and WEE1 was performed to parallel single-agent and combination treatments. Consistent with the FUCCI results, we observed decrease in both CDK1 and Cyclin B1 associated with mitotic exit (Fig. 5H; Supplementary Fig. S5). Cell-cycle agents can induce cell death through multiple processes and have been associated with increased autophagy, apoptosis, and necroptosis (43). To determine the mode of death in combination treatments, markers of apoptosis (annexin V staining and PARP/cleaved PARP), autophagy (LC3I/II), and necroptosis were measured (Fig. 5H; Supplementary Fig. S6). Both cleaved PARP and annexin V were induced by combination treatment, as was the conversion of LC3 forms reflecting enhanced autophagic flux (Fig. 5H; Supplementary Fig. S6). Morphologic analysis failed to show evidence of necroptosis, while demonstrating membrane swelling and nuclear condensation indicative of apoptosis (Supplementary Fig. S6). Furthermore, utilizing inhibitors of autophagy (chloroquine), necroptosis (necrostatin), and apoptosis (Z-vad FMK) revealed that only inhibition of apoptosis could partially rescue viability (Fig. 5I). These findings confirm that significant DNA damage and replication stress upon combined CHK1 and WEE1 inhibition causes cells to undergo apoptosis.
WEE1 and CHK1 inhibition is effective in vivo; however, resistance can develop
To confirm the efficacy of the therapy in vivo, we employed PDX models that match three of the cell lines (EMC226, EMC3226, and EMC828) and subjected them to treatment with single agent and the combination for 21 days (Fig. 6A). To monitor toxicity, body weight was measured and there was no significant decrease in weight in single- or double-agent treatment groups (Supplementary Fig. S7). Two models showed a significant delay in tumor growth in response to combination treatment that significantly exceeded both single-agent treatments (Fig. 6B and C). Histologic examination of the tumors harvested at the end of treatment revealed increased necrotic regions in the tumors from combination therapy arm compared with single-agent arms (Fig. 6D). In the 226 PDX model, there was an initial response to treatment, which was rapidly overcome with tumor growth on treatment, paralleling single agent or control (Fig. 6E). To investigate the mechanism of resistance to combination treatment, we developed a resistant cell line, 226R, from a tumor that was progressing on therapy. We confirmed resistance by dose–response analysis of CHK1 and WEE1 inhibitor (Fig. 6F). Immunoblot analysis indicated that 226R harbored increased levels of baseline pRPA relative to the parental EMC226 line and had constant levels of CDK1 in spite of the reduction in Cyclin A with treatment (Fig. 6G). This suggests that the resistance can develop as the cells evolve mechanisms that allow them to withstand higher levels of the replication stress induced by the combination treatment. Spectral karyotyping (SKY) analysis of parental and resistant line showed a decrease in chromosome numbers as well as more homogeneous pattern of translocations relative to the parental sensitive line indicating selective pressure for a subclone of the PDX (Fig. 6H). Functionally, the levels of EdU incorporation were higher in 226R than the parental model (Fig. 6I; Supplementary Fig. S7), indicating that the resistant model has in place mechanisms to continue DNA replication even in the face of combined blockade of WEE1 and CHK1. Because we found gemcitabine also cooperated with CHK1 inhibition and has a distinct mechanism of action (i.e., perturbation of dNTP pools and replication block), we interrogated its ability to rescue the sensitivity to CHK1/WEE1 in the resistant cells. Cell viability analysis demonstrated that the triple combination was effective in the resistant cell line (Fig. 6J) and chromatin-bound MCM7 levels were effectively decreased upon triple treatment compared with combination (Fig. 6K). Thus, employing two different agents that cooperate with CHK1 inhibition could be particularly efficacious to limit evolution of resistance.
In vivo efficacy of targeting cell-cycle checkpoints and gemcitabine
To evaluate the efficacy of triple drug combination in vivo we utilized the same PDX models that were employed for double treatment therapy as well as additional orthotopic models (Fig. 7). Addition of gemcitabine to the treatment regimen (Fig. 7A) did not yield additional toxicities as seen by body weight (Fig. 7B). Histologic evaluation of major organs (lung, small intestine, liver, and kidney) did not reveal significant toxicities (Fig. 7C). Murine cell line 4662 and EMC226 line were used for the orthotopic models. Mice were imaged by MRI to ensure volumes were similar at baseline prior to randomization to treatment arms (Supplementary Fig. S8). The 4662 orthotopic model gave rise to metastatic disease (60% of mice), which was absent in the triple regimen group (0% of mice), which also showed robust control of tumor growth (Fig. 7D). The PDX 226 model also had significant decrease in tumor growth as determined by tumor weight at end of treatment (Fig. 7E). Upon treatment completion, tumor and small intestine were stained with Ki67 to examine impact on proliferation in tumor and in intestinal crypts that contain progenitor cells essential for self-renewing capacity of the intestinal mucosa (Fig. 7F). Staining revealed that there was minimal inhibition of proliferation in the intestine while Ki67 levels dramatically decreased in tumors. We also evaluated whether addition of gemcitabine to CHK1 and WEE1 inhibitors could minimize resistance observed in the 226 PDX model with dual (CHK1 and WEE1) treatment (Fig. 7G). Analysis of tumor tissue revealed more extensive zones of necrosis in the triple therapy arm compared with the single and double therapy (Fig. 7H). Triple combination treatment was evaluated in additional PDX models confirming that addition of gemcitabine to AZD7762 and MK1775 results in increased disease control (Fig. 7I and J). Therefore, while CHK1 and WEE1 inhibition is effective in vivo, resistance to this therapy can be minimized by agents that further enhance replication stress and perturb dNTP pools such as gemcitabine.
Pancreatic cancer is a therapy recalcitrant disease; therefore, understanding limitations of specific targeted strategies and defining effective combination therapies is particularly important (44, 45). KRAS drives multiple survival and proliferative processes that likely contribute to the difficulty in targeting effector pathways (e.g., MEK) to much effect (46). Because DNA replication stress is a downstream biological consequence of KRAS oncogene, we reasoned that targeting cell-cycle checkpoints could be particularly efficacious and perhaps represent a precision approach to pancreatic cancer treatment.
In multiple PDAC models, we found that CHK1 inhibition could elicit cytotoxic effect, although this response was highly heterogeneous. This heterogeneity was observed in both established and patient-derived models of pancreatic cancer. Mechanistically, the response to CHK1 inhibition is related to induction of replication failure and increased DNA damage, while resistance is associated with cells that have adapted means to pass through DNA synthesis with essentially little requirement for CHK1. This resistance suggests that such pancreatic cancer cells have evolved alternative mechanisms to bypass replication stress. While it has been proposed that TP53 mutation is associated with sensitivity to CHK1 inhibition (47–50), we observed that none of the common genetic alterations present in pancreatic cancer (e.g., TP53, CDKN2A, or SMAD4) were associated with response. Similarly, the gene expression level of CHK1 was not significantly associated with sensitivity (not shown). However, a gene expression signature related to FOXM1 correlated with sensitivity in the cell models. FOXM1 is a potent driver of cell-cycle progression suggesting that the hyperactive cell cycle is a determinant for the sensitivity to CHK1 inhibition (40). Interestingly, this same signature is associated with particularly poor prognosis in pancreatic cancer clinical cases, suggesting the CHK1 inhibition could have efficacy in a small fraction of very aggressive pancreatic cancers.
In spite of the findings related to single-agent CHK1 inhibition (47, 50, 51), we strove to define combination approaches that would be active at low dose and would provide a means to broaden the efficacy. Using both broad and focused drug screens, we found that CHK1 inhibition was particularly synergistic in combination with WEE1 inhibition. This combination was effective against all cell lines tested and effectively converted CHK1-resistant cells to sensitivity. This finding agrees with genetic data from yeast that CHK1 and WEE1 are synthetically lethal, and synergistic effects of this combination that have been observed in cancer cells (29–31, 52–54). Other studies have indicated the dual effect of CHK1 and WEE1 inhibition leads to a state that has been equated to mitotic catastrophe (29, 31). Because WEE1 inhibition enhances the replication stress in cells that are resistant to CHK1 inhibition, it strongly indicates that WEE1 plays a role in the bypass to single-agent CHK1 inhibition that is operable in many pancreatic cancer models (Fig. 5G). However, we also observe that the combination specifically leads to an S-phase arrest, with reduced expression of CDK1, Cyclin B1, and loss of MCM7 from chromatin. Likely, this reflects both the replication stress and specific effects of WEE1 on CDK activity (55). This population with S-phase content has SCF-SKP2 inhibited upon combined treatment with CHK1 and WEE1 inhibitors, suggesting that the cells likely lack requisite CDK activity and thus are trapped in a “replicative-catastrophe.” These cells ultimately die through what appears to be an apoptotic process (Supplementary Fig. S5). Consistent with the cell-based sensitivity, the combination of CHK1 and WEE1 was effective in PDX models; however, despite initial response, progression on therapy could occur.
That a potent therapy in vitro failed to achieve durable disease control underscores the clinical issue of therapeutic resistance that is commonly observed in pancreatic cancer. However, it also provided the opportunity to delineate the nature of the resistance. In the context of pancreatic cancer, there is the possibility that resistance is associated with the tumor microenvironment or lack of drug access (56, 57). This does not appear to be the case as cells derived from the resistant PDX exhibited intrinsic resistance in cell culture. Interestingly, the resistant cell line represented a more chromosomally homogeneous subclone compared with parental model. In addition, the resistant model has high baseline replication stress and thus has apparently evolved a mechanism to evade the dual inhibition of CHK1 and WEE1 presumably at the cost of having an inefficient S-phase. Because we found that gemcitabine also cooperates with CHK1 inhibition and could further interfere with S-phase due to mediating replication blocks and perturbation of dNTP pools, the triple regiment (gemcitabine, CHK1, and WEE1 inhibitors) was employed in the treatment of multiple distinct xenograft models. In all cases there was potent disease control that far exceeded any of the other treatments employed. Therefore, coordinately targeting replication stress and disrupting two major cell-cycle checkpoint pathways enables consistent control of tumor growth. In general, triplet therapies are associated with additional toxicities; however, in our study, we did not observe significant toxicity in the mice with the utilized doses and schedules and treatments appear to be highly tumor selective. Together, our study illustrates the challenge of mechanistically targeted therapies in pancreatic cancer, where multiple mechanisms of therapeutic resistance can emerge. In spite of these challenges, tumor-selective combination approaches that target cell-cycle checkpoints can be identified and have broad-acting activity across a number of disease models.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: S. Chung, A.K. Witkiewicz
Development of methodology: S. Chung, A.K. Witkiewicz, E.S. Knudsen
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Chung, A.K. Witkiewicz
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Chung, P. Vail, A.K. Witkiewicz, E.S. Knudsen
Writing, review, and/or revision of the manuscript: S. Chung, A.K. Witkiewicz, E.S. Knudsen
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E.S. Knudsen
Study supervision: A.K. Witkiewicz, E.S. Knudsen
The authors thank members of the Witkiewicz and Knudsen laboratories for technical assistance and discussion of study results. This study was supported by grants from the NIH. Services were provided by the Flow and Image Cytometry Core facility, Pathology Network Shared Resource, and Translational Imaging Shared Resource, which are supported by the Roswell Park Cancer Center (NCI P30CA16056).
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.