A Kinome Screen Identifies Checkpoint Kinase 1 (CHK1) as a Sensitizer for RRM1-Dependent Gemcitabine Efficacy

Gemcitabine is among the most efficacious and widely used antimetabolite agents. Its molecular targets are ribonucleotide reductase M1 (RRM1) and elongating DNA. Acquired and de novo resistance as a result of RRM1 overexpression are major obstacles to therapeutic efficacy. We deployed a synthetic lethality screen to investigate if knockdown of 87 selected protein kinases by siRNA could overcome RRM1-dependent gemcitabine resistance in high and low RRM1-expressing model systems. The models included genetically RRM1-modified lung and breast cancer cell lines, cell lines with gemcitabine-induced RRM1 overexpression, and a series of naturally gemcitabine-resistant cell lines. Lead molecular targets were validated by determination of differential gemcitabine activity using cell lines with and without target knock down, and by assessing synergistic activity between gemcitabine and an inhibitor of the lead target. CHK1 was identified has the kinase with the most significant and robust interaction, and it was validated using AZD7762, a small-molecule ATP-competitive inhibitor of CHK1 activation. Synergism between CHK1 inhibition and RRM1-dependent gemcitabine efficacy was observed in cells with high RRM1 levels, while antagonism was observed in cells with low RRM1 levels. In addition, four cell lines with natural gemcitabine resistance demonstrated improved gemcitabine efficacy after CHK1 inhibition. In tumor specimens from 187 patients with non-small-cell lung cancer, total CHK1 and RRM1 in situ protein levels were significantly (p = 0.003) and inversely correlated. We conclude that inhibition of CHK1 may have its greatest clinical utility in malignancies where gemcitabine resistance is a result of elevated RRM1 levels. We also conclude that CHK1 inhibition in tumors with low RRM1 levels may be detrimental to gemcitabine efficacy.


Introduction
Gemcitabine is one of the most widely used agents with proven efficacy in non-small-cell lung cancer (NSCLC), breast cancer, pancreatic cancer, and bladder cancer among others [1,2,3,4]. De novo and acquired resistance, however, limit its utility to a subset of patients and constrain efficacy to a period of time measured in months.
Gemcitabine's mechanism of action has been extensively studied, and the two major molecular mechanisms are inhibition of deoxynucleotide synthesis and blockade of DNA synthesis [5,6,7,8]. The former is achieved through covalent binding to and inactivation of the regulatory subunit of ribonucleotide reductase M1 (RRM1). The latter is achieved through integration into the nascent DNA, which results in an elongation block.
In cell lines and animal models, emergence of gemcitabine resistance coincides with increased RRM1 expression [9,10], and gemcitabine efficacy can be influenced directly through modulation of RRM1 levels [11,12]. RRM1 has been identified as the major determinant of gemcitabine efficacy in patients treated with this agent. High levels of tumoral RRM1 are predictive of treatment resistance while low levels are associated with treatment efficacy [11,12,13,14]. In addition, limited data suggest that RRM1 levels can rise in patients receiving gemcitabine-based therapy [15].
Increasing gemcitabine efficacy in patients with de novo resistance and reversing or delaying acquired resistance would be major therapeutic advances. Our prior investigations had identified a link between RRM1 and signal transduction pathways [16,17] as well as G2 cell cycle arrest [18]. Here, we investigated the impact of selected kinases involved in cellular proliferation and cell cycle control on the RRM1-modulated gemcitabine efficacy using synthetic lethal screens. We identified CHK1 as a major target capable of reversing gemcitabine resistance.

Knockdown of Selected Protein Kinases Impacts RRM1dependent Cell Viability
To assess the interaction of selected kinases with RRM1, and ultimately a combined gene modification on gemcitabine efficacy, a synthetic lethality screen was performed using transfection of siRNAs specific to each of 87 different receptor and non-receptor tyrosine and serine/threonine kinases critical for cell cycle, shape, proliferation, and differentiation control. Non-targeting siRNA was used as a control, and cell viability was used as a read-out.
Screens were performed in 96-well plates using two replicates with siGENOME SMARTpool siRNA in each of the four sublines of the H23 and MCF7 stably RRM1-modified model systems (H23-R1, H23-shR1, H23-Ct, H23-wt, and MCF7-R1, MCF7-shR1, MCF7-Ct, MCF7-wt) without gemcitabine. Differential cell viability was calculated comparing values obtained from high (R1) and low (shR1) RRM1 expressers (Figure 1). In the H23 system, knockdown of the genes CHK1, FLT4, KIT, MET, and TEC had a significantly (p#0.01) greater lethality if RRM1 expression was high, and BLK had a significantly greater lethality if RRM1 expression was low. In the MCF7 system, knockdown of the genes BTK, CHK1, and EPHB2 had a significantly (p#0.01) greater lethality if RRM1 expression was high, and ERBB4 and RYK had a significantly (p#0.01) greater lethality if RRM1 expression was low. In addition, in both cell line systems knockdown of the gene KDR had a significantly (p#0.05) greater lethality if RRM1 expression was high.
We attempted to verify these results with ON-TARGETplus SMARTpool siRNA using four separate replicate experiments. Only the genes CHK1, KDR, and TEC yielded similar results with significantly (p#0.05) greater lethality if RRM1 expression was high in both cell line systems. BLK yielded similar results for the H23 system only with significantly (p#0.05) greater lethality if RRM1 expression was low.

Knockdown of Selected Kinases Modulates RRM1dependent Gemcitabine Efficacy
BLK, CHK1, KDR, KIT, MET, and TEC were chosen to study interactions with RRM1-dependent gemcitabine efficacy. BLK, KDR, KIT, and MET knockdown did not appreciably impact gemcitabine efficacy, and both had no interaction with RRM1 in the H23 model (Table 1); i.e., there was no differential activity in high versus low RRM1 expressing cells (IC50 ratios of 0.79, 1.10, 0.85, and 1.02 respectively for high vs. low RRM1). CHK1 knockdown improved gemcitabine efficacy (1.8-fold improvement in IC50 concentration in H23 and 2.2-fold improvement in MCF7; Table 1), and it displayed a clear interaction with RRM1 in both models. The gemcitabine IC50 improved at least 3.8-fold in cells with high RRM1 expression, and it deteriorated 1.4 to 3.8-fold in cells with low RRM1 expression (Table 1; Figure 2). CHK1 knockdown also substantially improved gemcitabine efficacy in four naturally gemcitabine-resistant NSCLC cell lines (H125, H650, H1648, H2122 [19]) with a shift of the dose-response curve to the left in addition to an increase in the maximally achievable cytotoxicity (IC50 ratios of 0.02 to 0.76; Figure 3; Table 1). Finally, we generated gemcitabine resistant clones from cell lines H23 and H1299 in vitro by prolonged drug exposure. In these clones, gemcitabine resistance was associated with a 20-fold increase in RRM1 mRNA and protein expression; and RRM1 knockdown with ON-TARGETplus SMARTpool siRNA restored gemcitabine sensitivity (IC50 ratios: 0.15 for H23-G-C8 and H23-G-C23, 0.19 for H1299-G-C2 and 0.09 for H1299-G-C18). In this model system, CHK1 knockdown reversed gemcitabine resistance and resulted in a near 4-fold improvement in drug efficacy ( Figure 3; Table 1).
TEC knockdown minimally improved gemcitabine efficacy (1.5fold improvement in IC50 concentration) and displayed an interaction with RRM1 (Table 1; Figure 2). TEC knockdown slightly improved gemcitabine efficacy in H23 with high RRM1 expression and reduced efficacy in cells with low RRM1 expression. This interaction was not observed in the MCF7 model. We further investigated the impact of TEC knockdown on gemcitabine efficacy in the four NSCLC cell lines with wild-type resistance to gemcitabine, and no evidence for altered gemcitabine sensitivity was observed in these cell lines (Table 1).

A Small-molecule CHK1 Inhibitor Increases Gemcitabine Efficiency
We used AZD7762, a small-molecule ATP-competitive inhibitor of CHK1 activation [20], to study its impact on gemcitabine efficacy. The single-agent IC50 for AZD7762 was determined over a concentration range of 10 pM to 100 mM using 4-day exposure in the four gemcitabine-resistant clones of cell lines H23 and H1299, and it was 505 nM for H23-G-C8, 497 nM for H23-G-C23, 2.5 mM for H1299-G-C2, and 1.8 mM for H1299-G-C18. Cells were then exposed to increasing concentrations (10 pM-100 mM) of gemcitabine in the presence or absence of 150 nM or 300 nM AZD7762. All four clones showed an AZD7762 dose dependent increase in gemcitabine sensitivity ( Figure 4). Two additional wild-type NSCLC cell lines (H1648, H2122) were investigated for efficacy modulation using the combination index (CI) methodology [21]. We observed clear synergism between both drugs in these gemcitabine-resistant model systems (Table 2).
We also used AZD6474 (vandetanib), an oral inhibitor of KDR (VEGFR), EGFR, and RET [22], to study its impact on gemcitabine efficacy. The single-agent IC50 in the H23 and MCF7 model systems was between 10 and 17 mM without evidence for an RRM1-dependent activity differential.

Knockdown of CHK1 and Gene Expression
To gain insight into the mechanism of the CHK1-mediated gemcitabine sensitization, we determined mRNA and protein expression levels for the genes RRM1 and RRM2 in the four naturally gemcitabine-resistant NSCLC cell lines and the four in vitro-induced gemcitabine-resistant clones 24 h after transfection. As depicted in Figure 5, transfection of cells with 30 nM CHK1 siRNA resulted in a 90% or greater reduction of CHK1 mRNA and protein and a 20-80% reduction in RRM1 and RRM2 mRNA and protein expression levels. To reduce concentration-dependent off-target effects, 10 nM concentrations of siRNA were used and the associated gene expression reductions were similar to those observed with 30 nM concentrations. We also assessed the expression levels of the genes PTEN and MCM7, and no changes were observed. Treatment of cells with 50-300 nM AZD7762 for up to 48 h did not result in a reduction of RRM1, RRM2, PTEN, or CHK1 levels.
In contrast, RRM1 knockdown with 10 nM or 30 nM RRM1 siRNA, which resulted in a 90% reduction of RRM1 protein levels, did not result in a reduction of CHK1 or RRM2 levels, but it did reduce PTEN levels as previously reported [16].
We found no significant (p.0.05) associations of the in situ levels of these proteins with the clinical parameters tumor histology (categorized as adenocarcinoma, squamous cell carcinoma, or large cell carcinoma), tumor size (categorized as #2 cm, .2 to #3 cm, .3 to #5 cm, .5 to #7 cm, and .7 cm), smoking status (categorized a never smoker if the patient had smoked less than 100 cigarettes during life time, former smokers if the patient had quit for at least 1 year, and current smoker for all others), and gender (table 3).

Discussion
Systemic therapy, targeted at disrupting DNA synthesis, continues to be the mainstay of anti-cancer therapy, and it is likely to remain a cornerstone for years to come. Antimetabolites are a class of DNA-targeted agents, and gemcitabine is a widely Figure 1. Synthetic lethality of 87 kinases as a function of RRM1 expression. Two cell line systems, H23 (blue bars) and MCF7 (red bars), with stably increased or reduced RRM1 expression were transfected with gene-specific siRNAs and viability was scored. Bars to the left indicate greater lethality in high compared to low RRM1expressing cells. Bars to the right indicate greater lethality in low compared to high RRM1-expressing cells. Bars below the x-axis indicate one (wide bar) or two (narrow bar) standard deviations in the lethality scores for the two cell line systems respectively. doi:10.1371/journal.pone.0058091.g001 used class representative for treatment of solid malignancies [1,2,3,4]. RRM1 is a validated, predictive marker of gemcitabine efficacy [11,12,13,14], which provides an opportunity for the identification of molecules that critically interact with RRM1 to maintain cell proliferation and viability [16,17,18]. To improve de novo therapeutic efficacy and attenuate the emergence of resistance, kinases are particularly attractive molecules because of their role in proliferation and viability and because many are druggable targets with available inhibitors in clinical development [25]. Synthetic lethality screening is a well-suited method for the identification of molecules that critically interact to maintain viability [26]. Originally, synthetic lethality was the identification of two genes whose simultaneous loss-of-function resulted in cell death in yeast as a model organism, while loss of either gene alone was compatible with life [27]. However, in an expanded application, two molecules can be referred to as ''synthetically lethal'' if functional attenuation of two molecules synergistically impacts on a desired outcome. We thus deployed this method to assess the impact of attenuation by siRNA technology of a series of kinases on gemcitabine efficacy in the context of RRM1 levels.
We deployed a stringent algorithm that consisted of multiple stages of siRNA knockdowns in a series of gemcitabine-resistant cell line models, including cell lines naturally resistant, with gemcitabine-induced resistance, and with genetically-induced resistance, to identify the most promising kinase target. Among the 87 kinases screened, we identified CHK1 as the lead kinase with consistent interaction with RRM1-dependent gemcitabine efficacy, and this synthetic lethality interaction was validated using AZD7762, a small-molecule ATP-competitive inhibitor of CHK1 activation [20].
Our identification of CHK1 as a potent modulator of RRM1dependent gemcitabine efficacy is not surprising. The serine/ threonine kinase CHK1 is a key effector in response to DNA damage, which triggers arrest of G2 cell cycle progression and facilitates DNA damage repair [28,29]. It becomes activated by the phosphatidylinositol 3-kinase-like serine/threonine kinases ATR (ataxia telangiectasia response) and ATM (ataxia telangiec- Figure 3. CHK1 knockdown can overcome gemcitabine resistance. Four naturally gemcitabine-resistant NSCLC cell lines (H125, H650, H1648, H2122) and four gemcitabine-resistant clones generated in vitro (H23-G-C8, H23-G-C23, H1299-G-C2, H1299-G-C18) were exposed to gemcitabine for 5 days. Data points are the mean of four replicates, and bars depict the standard error. There is a shift of the dose-response curve to the left in addition to an increase in the maximally achievable cytotoxicity. doi:10.1371/journal.pone.0058091.g003 tasia mutated) through phosphorylation of Ser-317 and Ser-345 [28] followed by autophosphorylation of Ser-296 [30]. It has recently been identified as a lead therapeutic target in neuroblastoma [31], and prior work by the AstraZeneca research and development team identified CHK1 inhibition by AZD7762 as contributing to efficacy of DNA-damaging agents [20].
We observed a consistent decrease in RRM1 expression levels with CHK1 knockdown, which may partially explain the improved gemcitabine efficacy. However, we also observed a decrease in RRM2 levels, and published reports had demonstrated that gemcitabine efficacy is unrelated to RRM2 levels [9,10,15]. However, other mechanisms may also contribute to the observed CHK1 interaction with RRM1-dependent gemcitabine efficacy. Our prior work had identified delayed G2 cell cycle progression as a consequence of RRM1 overexpression [18], and the results presented here demonstrate a higher degree of synergism between CHK1 inhibition and RRM1 in cells with high levels of RRM1 expression compared to those with low RRM1 expression.
We had previously identified p53 inactivation as contributing to gemcitabine resistance in a panel of 26 NSCLC cell lines [19]. In addition, p53 knockdown in p53 expressing cell lines increased and p53 restoration in p53-null cells decreased gemcitabine resistance. Notably, all four cell lines with naturally occurring gemcitabine resistance (IC50 above 1 mM) used here have nonfunctional p53 alterations (H125, truncation N239*; H650, missense K164N; H1648, truncation D42*; H2122, missense C176F and an additional missense Q16L with no predicted  These results are consistent with earlier reports suggesting that abrogation of the G2 checkpoint may be particularly effective in cells with impaired p53 function [32,33,34].
In conclusion, we identified CHK1 inhibition as a major modulator of RRM1-dependent gemcitabine efficacy using a synthetic lethality screen with 87 protein kinases. The impact of CHK1 inhibition is particularly pronounced in cells resistant to gemcitabine as a result of high RRM1 expression levels. This implies that the combination of a CHK1 inhibitor and gemcitabine is likely to be most effective in patients with high tumoral levels of RRM1. Our quantitative in situ determination of RRM1 and total CHK1 protein levels demonstrated that CHK1 levels tend to decrease as RRM1 levels increase, which suggests that CHK1 inhibition may be best achieved in patients with RRM1-dependent gemcitabine resistance. Clinical trials utilizing both agents may provide an opportunity to assess this interaction in patients with solid malignancies and may afford the opportunity for patient stratification based on RRM1 levels.
Gemcitabine (Eli Lilly) resistance was generated by treating H23 and H1299 cells with 1 nM gemcitabine for two weeks. Surviving cells were then exposed to increasing doses until cells survived drug concentrations ten-fold higher than the original gemcitabine IC50 (8 nM for H23, 24 nM for H1299 [19]). Clonal sublines of resistant cells were generated from single colonies (H23-G-C8 and C23; H1299-G-C2 and C18), and resistance was confirmed using a cell viability assay as described below. RRM1 expression was assessed by Western blotting.
Thiophene carboxamide urea (AZD7762), an ATP-competitive inhibitor of checkpoint kinase 1 (CHK1), and vandetanib (AZD6474), an inhibitor of KDR, EGFR, and RET activation, were obtained from Synthesis Med Chem Pty Ltd, Australia. They were dissolved in 0.1% DMSO, and all experiments using these agents included controls with 0.1% DMSO alone.

Cell Viability Assay and Synthetic Lethality Screening
Cells were plated in 96-well plates for 16 h and then treated with increasing concentrations of gemcitabine (10 pM-100 mM) for 4 to 7 days in the presence or absence of 150 nM or 300 nM AZD7762. Dose response was determined using a luminescent cell viability assay (CellTiter-Glo, Promega, Madison WI) according to the manufacturer's instructions. Assay data were transferred to a spreadsheet program to calculate the percent viability relative to the replicate control cell wells that did not receive drug. The halfmaximal inhibitory concentrations; i.e., IC 50 values, were calculated using SigmaPlot dose-response curves (Systat Software).
For drug combination experiments, drugs were used at fixed dose ratios (i.e. 50:1, 1:10, 1:250), and results were analyzed for synergistic, additive, or antagonistic effects using the combination index (CI) method [21]. Briefly, the dose-effect curve for each drug alone was determined based on experimental observations using the median-effect principle and compared to the effect achieved with a combination of two drugs to derive a CI value. The method involves plotting dose-effect curves for each agent and their combination using the median-effect equation Gene specific knockdowns were done using the Dharmacon siGENOME SMARTpool and ON-TARGETplus SMARTpool libraries (Dharmacon RNAi Technologies). Non-targeting siRNA was used as a control in all experiments. Transfections were done with Lipofectamine RNAiMAX (Invitrogen) following the manufacturer's instructions using 30 nM siRNA concentrations unless indicated otherwise. Drugs were added 24 h after siRNA transfection.

RNA Isolation, cDNA Synthesis, and Quantitative Realtime PCR
Total RNA was isolated from cells 24 h after siRNA transfection, unless indicated otherwise, using a commercially available method (RNeasy, Qiagen) and reverse transcribed to generate cDNA (QuantiTect, Qiagen). A fluorescence-based real-time detection method (Taqman, Applied Biosystems) was used for quantification of gene expression. The primers and probes used for CHK1 and 18s-rRNA were purchased from ABI, and those for RRM1 and RRM2 have been published [19]. Reaction conditions and gene expression value calculations using 18s-rRNA as reference gene were as previously described [19]. Western Blot Analysis Cells were lysed (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5% NP-40, 10 mM glycerol, 5 mM EDTA, 0.5% sodium deoxycholate) in the presence of protease and phosphatase inhibitors (1 mM phenylmethylsulfonyl fluoride, 1 mg/mL aprotinin, 1 mg/mL leupeptin, 1 mM sodium orthovanadate). Proteins were separated in 8 to 10% SDS-polyacrylamide gels and blotted onto nitrocellulose membranes (Biorad). Filters were incubated with polyclonal antisera or monoclonal antibodies against b-actin, RRM1, RRM2, PTEN, MCM7 (Santa Cruz Biotechnology), and Chk1 (Cell Signaling).

Statistical Analysis
The average value of three TMA spot replicates was calculated for each patient and target.
Associations among parameters with continuous values were calculated using the Spearman rank correlation coefficient. Associations between continuous and categorical values were calculated using the Wilcoxon rank-sum test for variables with two categories and the Kruskal-Wallis test for variables with three or more categories.