Identification of MRP2 as a targetable factor limiting oxaliplatin accumulation and response in gastrointestinal cancer

Oxaliplatin is important for the clinical treatment of colorectal cancer and other gastrointestinal malignancies, but tumour resistance is limiting. Several oxaliplatin transporters were previously identified but their relative contributions to determining oxaliplatin tumour responses and gastrointestinal tumour cell sensitivity to oxaliplatin remains unclear. We studied clinical associations between tumour expression of oxaliplatin transporter candidate genes and patient response to oxaliplatin, then experimentally verified associations found with MRP2 in models of human gastrointestinal cancer. Among 18 oxaliplatin transporter candidate genes, MRP2 was the only one to be differentially expressed in the tumours of colorectal cancer patients who did or did not respond to FOLFOX chemotherapy. Over-expression of MRP2 (endogenously in HepG2 and PANC-1 cells, or induced by stable transfection of HEK293 cells) decreased oxaliplatin accumulation and cytotoxicity but those deficits were reversed by inhibition of MRP2 with myricetin or siRNA knockdown. Mice bearing subcutaneous HepG2 tumour xenografts were sensitised to oxaliplatin antitumour activity by concurrent myricetin treatment with little or no increase in toxicity. In conclusion, MRP2 limits oxaliplatin accumulation and response in human gastrointestinal cancer. Screening tumour MRP2 expression levels, to select patients for treatment with oxaliplatin-based chemotherapy alone or in combination with a MRP2 inhibitor, could improve treatment outcomes.

not respond to oxaliplatin chemotherapy. We searched the Oncomine transcriptome database for datasets of patients treated with oxaliplatin, who had tumour microarray gene expression profiling undertaken before treatment and annotation of their subsequent tumour response. Only one dataset was found, the Tsuji Colorectal dataset 29 (GDS4393 and GDS4396) comprising of 83 patients with metastatic colorectal cancer who had tumour microarray gene expression profiling before treatment with FOLFOX. Patients were stratified into FOLFOX responders (n = 42) or non-responders (n = 41). Differences between the two groups in the expression of reporters of each oxaliplatin transporter candidate gene (Table 1) were calculated. Only one of 18 oxaliplatin transporter candidate genes showed significantly different expression. MRP2 (ABCC2) was overexpressed (1.5-fold) in the tumours of patients who did not respond (P < 0.0001 Bonferroni post-test following Two-Way ANOVA; Twoway ANOVA Factors: Tumour response, P = 0.0025; Gene, P < 0.0001; Interaction, P = 0014) (Fig. 1). The tumour expression of the other 17 oxaliplatin transporter candidate genes showed no significant difference (<−0.4-to 0.8-fold) between responding and non-responding patients.
In vitro studies. In an isogenic pair of HEK293 cell lines, stable overexpression of MRP2 (HEK-MRP2 cells) decreased oxaliplatin accumulation and cytotoxicity but those deficits were reversed by inhibition of MRP2 with myricetin. Immunofluorescence confocal microscopy detected MRP2 protein localised to the plasma membranes of HEK-MRP2 cells but negligible immunoreactivity in parental HEK cells (HEK-P cells) ( Fig. 2A). Compared to HEK-P cells, HEK-MRP2 cells accumulated up to 3-fold less platinum during exposure to oxaliplatin for up to 2 hours (Fig. 2C,D). Concurrent exposure to oxaliplatin with myricetin increased the accumulation of platinum in HEK-MRP2 cells by about 3-fold (P < 0.001) but did not change the accumulation of platinum in HEK-P cells (Fig. 2E). IC 50 values for oxaliplatin-induced growth inhibition were increased in HEK-MRP2 cells by about 2-fold compared to HEK-P cells (Fig. 2B,F). Concurrent exposure to oxaliplatin with myricetin increased the sensitivity of HEK-MRP2 cells to oxaliplatin-induced growth inhibition but had no effect on HEK-P cells (Fig. 2F).
In vivo study. Mice bearing subcutaneous HepG2 tumour xenografts were sensitised to the in vivo antitumour activity of oxaliplatin by concurrent treatment with myricetin with little or no increase in toxicity. Nude mice (NIH-III) were implanted subcutaneously with HepG2 cells, which were demonstrated to have MRP2-mediated deficits in oxaliplatin accumulation in vitro. Tumour bearing mice were treated with oxaliplatin (3 mg/kg ip), myricetin (25 mg/kg iv) or their respective vehicles alone or in combination (n = 8/group), once weekly for five weeks, starting from when tumours measured about 200 mm 3 (range, 161 to 237 mm 3 ). The body weight of mice was reduced by oxaliplatin treatment when given alone or in combination with myricetin, compared to drug vehicle or myricetin alone (Repeated Measures Two-Way ANOVA Factors: Time, NS; Treatment group, P < 0.0001; Interaction, NS; Subjects, P < 0.0001) ( Fig. 5A and Supplementary Table 1). Growth of HepG2 tumour xenografts was slowed by treatment with oxaliplatin combined with myricetin compared to Figure 1. Differential expression of oxaliplatin transporter candidate genes in colorectal tumours of patients who responded or did not respond to FOLFOX chemotherapy. Data from the Tsuji Colorectal dataset 29 from Oncomine included 83 colorectal cancer patients who had tumour microarray gene expression analysis prior to FOLFOX chemotherapy then stratification by tumour response. Bars and error bars are means and standard errors of differences in the expression of individual oxaliplatin transporter candidate genes between responding and non-responding patients. Positive differences indicate higher expression in non-responders whereas negative differences indicate higher expression in responders. The asterisk is a P value (***P < 0.0001) from a Bonferroni post-test that followed a Two-way ANOVA (Two-way ANOVA Factors: Tumour response, P = 0.0025; Gene, P < 0.0001; Interaction, P = 0014). ABCC2/MRP2 was over-expressed in colorectal tumours from patients who did not respond to FOLFOX chemotherapy compared to responders. Other oxaliplatin transporter candidate genes were expressed similarly in both responding and non-responding patients. Asterisks are P values (***P < 0.001; ****P < 0.0001) for differences with each of the other three groups from Bonferroni post-tests that followed a Two-way ANOVA. Bars represent means and standard errors of the mean. HEK-MRP2 cells accumulated less platinum and were less sensitive to oxaliplatin-induced growth inhibition than control HEK-P cells. Myricetin increased the accumulation and cytotoxicity of oxaliplatin in HEK-MRP2 cells but not in HEK-P cells.  Table 3). The time taken to reach the endpoint of euthanasia was prolonged by treatment with oxaliplatin combined with myricetin (median, 47 days) compared to drug vehicle alone (median, 33.5 days; P = 0.0968), oxaliplatin alone (median, 28 days; P = 0.0003) or myricetin alone (median, 17 days; P = 0.0181)(P value for comparison of all groups = 0.0032) ( Fig. 5D and Supplementary Table 3).

Discussion
Oxaliplatin and its combination regimens are of major importance in the clinical treatment of colorectal cancer and other gastrointestinal malignancies, but some patients fail to respond to treatment. National and international clinical practice guidelines 10,11 now recommend oxaliplatin-based chemotherapy as a preferred regimen for the treatment of colon, gastric, oesophageal, pancreatic, rectal and other types of gastrointestinal cancer based on robust evidence from multiple randomised controlled trials [2][3][4][5][6][7][8][9] . However, in some patients the best tumour response is progressive disease and in others tumour responses are short-lived. Currently it is not possible to prospectively identify individual patients who are destined to fail oxaliplatin therapy prior to their exposure to treatment and observation of their unsatisfactory clinical outcomes. Moreover, there is an urgent need to identify alternative treatments for patients who will have such poor outcomes from standard oxaliplatin-based chemotherapy.
In the current study, we demonstrated a significant clinical association between tumour expression of the oxaliplatin transporter MRP2 and lack of responsiveness of colorectal cancer patients to standard oxaliplatin-based chemotherapy (FOLFOX). In contrast, tumour expression of 17 other oxaliplatin transporter candidate genes showed no significant association with clinical response to oxaliplatin in this dataset. Unfortunately, only one dataset was available in Oncomine of gastrointestinal cancer patients with tumour gene expression profiling undertaken before treatment with oxaliplatin together with annotation of their subsequent tumour response. Moreover, in this dataset, only steady state MRP2 mRNA levels were measured, not MRP2 protein level or function, which may be altered by posttranslational modification, such as by glycosylation 30,31 . However, these data we now report add to an increasing number of independent studies reporting clinical associations between gastrointestinal tumour expression of MRP2 and non-responsiveness to chemotherapy with oxaliplatin or cisplatin 23,[32][33][34][35] . Taken together, these findings suggest that MRP2 may cause clinical resistance to oxaliplatin in gastrointestinal cancer patients, but ultimate proof of a causal association will require generation and interrogation of other tumour gene expression clinical datasets along with prospective clinical trials evaluating the inhibition of MRP2 in combination with oxaliplatin in patients whose tumours overexpress MRP2.
Our in vitro experimental studies reported here provided insights into mechanisms underlying poor responses to oxaliplatin in gastrointestinal cancer. The mechanism identified here involves MRP2 actively transporting oxaliplatin and/or other cytotoxic platinum species derived from oxaliplatin, resulting in their extrusion from tumour cells, net reduction in cellular platinum accumulation and decreased inhibitory effects of oxaliplatin on cell growth. Previously, we demonstrated that MRP2 actively transports oxaliplatin-derived platinum in MRP2-expressing membrane vesicles 36 . Intact oxaliplatin and/or its anionic monochloro oxalate ring-opened early degradation product were identified in these studies as likely substrates for MRP2. Moreover, the active transport of oxaliplatin was shown to be inhibited by myricetin 36 . In the current study, we extended these findings to show lower accumulation of platinum and sensitivity to oxaliplatin-induced growth inhibition in HEK293 cells stably transfected to overexpress MRP2, and in human gastrointestinal tumour lines (HepG2 and PANC-1) with endogenous overexpression of MRP2. Inhibition of MRP2 by myricetin or siRNA knockdown increased the accumulation of platinum and sensitivity to oxaliplatin-induced growth inhibition in these MRP2 overexpressing tumour lines. None of the four colorectal cancer cell lines we studied displayed low oxaliplatin accumulation due to MRP2 overexpression but a hepatocellular (HepG2) 37 and a pancreatic carcinoma (PANC-1) line did. This created an incomplete match in the data presented here between the clinical-gene expression dataset of colorectal cancer patients (Fig. 1) and our experimental work that mainly focussed on the HepG2 37 and PANC-1 lines that displayed MRP2-mediated transport-resistance to oxaliplatin in vitro (Fig. 3). This incomplete match in the presented data was further compounded by no tumour gene expression clinical datasets of hepatocellular or pancreatic carcinoma patients with annotated tumour response to oxaliplatin being available in Oncomine. However, previous studies have found increased MRP2 expression and activity in oxaliplatin-resistant colorectal cancer lines in vitro, and that inhibition of MRP2 expression and activity by genetic knockdown, myricetin or other inhibitors, can reverse MRP2-mediated accumulation deficits and resistance to oxaliplatin in colorectal cancer [24][25][26] . This study also provided in vivo experimental proof of the therapeutic concept that targeting MRP2 could sensitise MRP2-overexpressing human gastrointestinal tumours to the anti-tumour activity of oxaliplatin without increased toxicity. In mice bearing subcutaneous xenografts of the HepG2 tumour line, which had shown MRP2-mediated deficits in oxaliplatin accumulation in vitro, oxaliplatin treatment given alone reduced body weight but had no significant antitumour activity. In contrast, when combined with myricetin, oxaliplatin treatment slowed tumour growth and extended the survival of animals without significant increase in toxicity. However, toxicity was measured only by body weight and the major clinical toxicity of oxaliplatin, neurotoxicity could have been increased by inhibition of MRP2. Further studies will be required to determine if MRP2 inhibition or genetic knockout increases oxaliplatin neurotoxicity, for example, using histomorphometric endpoints (eg. dorsal root ganglion neuron cell body size) of oxaliplatin neurotoxicity 38,39 . While other mechanisms cannot be ruled out 26,[40][41][42] , it is reasonable on the basis of the findings reported here to hypothesise that myricetin sensitises tumours to oxaliplatin antitumour activity by inhibiting oxaliplatin transport mediated by MRP2. In another recent study of an animal model of human gastrointestinal cancer, myricetin enhanced the in vivo antitumour activity of 5-fluorouracil 43 , which is often given with oxaliplatin in combination regimens such as FOLFOX. These findings now set the scene for further exploration of the therapeutic concept of targeting MRP2 with myricetin or other inhibitors to enhance the antitumour activity of oxaliplatin-based combination regimens, both experimentally and in the clinic.
In conclusion, MRP2 limits oxaliplatin accumulation and responses in gastrointestinal cancer. As tumour MRP2 expression is both readily measurable and potentially targetable in gastrointestinal cancer patients, screening MRP2 tumour expression to select patients for treatment with oxaliplatin alone or in combination with an MRP2 inhibitor could improve their outcomes.

Materials and Methods
Oncomine dataset and analysis. Log2 median-centred intensity values for the tumour expression of eighteen oxaliplatin transporter candidate genes from 83 patients in the Tsuji Colorectal dataset (GDS4393 and GDS4396) 29 were exported from Oncomine. Patients were stratified according to their response to FOLFOX chemotherapy. Differences between the responder (n = 42) and non-responder (n = 41) groups were analysed by Two-way ANOVA (Two-way ANOVA factors: transporter gene expression and tumour response to FOLFOX) followed by Bonferroni post-tests for each candidate gene. Data were assessed visually in graphs of absolute differences and standard errors of differences in mean log2 median-centred intensity values for each gene between the responder and non-responder groups. MRP2 Fluorescence Immunocytochemistry. Fluorescent immunocytochemistry was undertaken to detect cellular MRP2 protein expression in the HEK-293 sublines using a primary antibody, anti-MRP2 (1:100; ab3373 from Abcam), and a secondary antibody, Alexa Fluor 594-labeled anti-mouse IgG (Invitrogen, Carlsbad, CA, USA). Cells were cultured in the chamber slides until they reached the confluence of 60-70%. The growth medium was then replaced with pre-warmed PBS which was left for 5 min followed by fixation of cells with 4% paraformaldehyde for 15 min at room temperature. Then, cells were permeabilised with 0.2% Triton X-100 in PBS for 15 min. Blocking of the non-specific protein binding was achieved by incubating cells with blocking solution (PBS containing 0.2% Triton X-100, 3% goat serum, and 2% bovine serum albumin) for 1 h at room temperature. After blocking, cells were incubated with the primary antibody, anti-MRP2 (ab3373 from Abcam) prepared in immunobuffer (PBS containing 0.2% Triton X-100 and 3% goat serum) with the dilution ratio of 1:100 for overnight at 4 °C. After that, the cells were washed for 3-4 times with PBS containing 0.2% Triton X-100 followed by incubation of cells with the secondary antibody (Alexa Fluor 594-labeled anti-mouse IgG, Invitrogen, Carlsbad, CA, USA) diluted in the immunobuffer (dilution ratio was 1:500) for 3 h at 4 °C with the cover to protect from light. Cells were then washed with PBS and cover-slipped with VECTASHIELD Mounting Medium with DAPI (6-diamidino-2-phenylindole) to stain nucleic acid (Vector Laboratories, Burlingame, CA, USA). Images were captured using an Olympus FV1000 confocal laser scanning microscope (Olympus Inc., Tokyo) attached to a Nikon digital camera and then analysed using Nikon Elipse Net (Nikon, Melville, NY, USA) and ImageJ software (National Institutes of Health, Bethesda, MD, USA).

Platinum accumulation.
Inductively coupled plasma mass spectrometry (ICP-MS) was used to measure the cellular content of platinum after exposure to oxaliplatin. For these experiments, cells were seeded at 250,000 cells per well (350,000 cells per well for HepG2 and PANC-1) in collagen-coated six-well plates, and grown in the normal growth medium until around 80% confluent. Cells were then incubated with oxaliplatin at designated concentrations and treatment durations. At the end of the treatment period, cells were washed three times with 1 mL of ice-cold PBS (phosphate buffered saline) then dried in room air for 30 min. Then 330 μL 70% nitric acid was added to each well and the plate gently shaken at room temperature for 2 hours to digest the cells. A 300 μL aliquot of the digest was then added to a 96-well plate to determine protein content using a modified tyrosine nitration assay 44 . After the protein assay, aliquots were transferred to 5 mL screwtop vials for further digestion in 70% nitric acid at room temperature overnight before heating at 95 °C for 2 hours. Digests were diluted in milliQ water containing thallium 50 ppb as an internal standard before ICP-MS analysis using a Varian 820MS ICP-MS (Agilent Technologies Inc., Santa Clara, CA, USA) at LabPLUS (Auckland, New Zealand). Optimisation of signal/ noise was carried prior to each run using a tuning solution of platinum and thallium. Platinum isotopes (m/z 194 and 195) and thallium (m/z 205) were monitored during the run. Platinum counts for samples and standards were normalized for thallium internal standard counts. The platinum content of each experimental sample was calculated from the platinum to thallium count ratios and a calibration curve made from standard solutions at desired platinum concentrations made up in the same matrix as the samples. Quality control samples at three different concentrations were included for determing the accuracy, precision and reliability of each run. An ICP-MS analysis run was accepted if the standard curve was linear, and quality control and calibration standards were ±15% of their nominal value and replicate quality control samples had % coefficient of variation of ≤15%. The limit of detection and lower limit of quantification were 0.3 ppb and 1 ppb, respectively.
Cell growth inhibition. To evaluate the sensitivity of different cell lines to oxaliplatin-induced cell growth inhibition, cells were exposed to oxaliplatin at different concentrations for 2 h followed by measurement of viable cells with MTT assay 45 . Cells were seeded at 5000 cells/well in a collagen-coated 96-well plates then allowed to attach to the well surface over 24 h incubation in normal drug-free growth medium. After attachment, cells were exposed to oxaliplatin at different concentrations for the designated time at 37 °C with 5% CO 2 /95% air. When myricetin was used, cells were exposed to myricetin for 30 min immediately prior to and during the 2 h exposure to oxaliplatin. To terminate drug exposure, the medium containing drug was removed and replaced with the drug-free growth medium. Cells were then allowed to grow for 72 hours under usual incubation conditions before the MTT assay was performed. Percent cell viability at different drug concentrations was calculated from 540 nm absorbance values normalised to the mean absorbance of control untreated cells. IC-50 values for oxaliplatin-induced growth inhibition were determined from the data using nonlinear regression in GraphPad Prism version 6.   47,48 , the expression of the five housekeeping genes varied widely between the different cell lines and were unsuitable for use as controls, and expression level Ct values were normalised by the mean expression of all genes of each cell line and PCR run, then multiplied by −1 so that positive values indicated higher expression. Cell lines were stratified into two groups according to whether they had high or low accumulation of platinum after exposure to oxaliplatin. Differences in transporter gene expression between the groups of cell lines were analysed by Two-way ANOVA (Two-way ANOVA factors: transporter gene expression and platinum accumulation) followed by Bonferroni post-tests for each candidate gene. Data were assessed visually in graphs of absolute differences and standard errors of differences in the expression of each gene between the high and low platinum accumulation groups of cell lines. Animal experiments. Animals were housed in a temperature-controlled environment with access to food and water ad libitum, and were acclimatized to handling prior to the experiments. Pilot studies were undertaken to establish the treatment protocol prior to the definitive experiment. The pilot studies showed similar trends as shown in the final experiment. For the definitive experiment, 32 NIH-III nude mice weighing 18 to 21 g were injected subcutaneously into the right flank with 10 7 of HepG2 cells in three batches of 10 or 11 animals each. When tumours measured approximately 200 mm 3 (range, 161 to 237 mm 3 ), animals were randomly allocated to one of four treatment groups (control, oxaliplatin alone, myricetin alone or oxaliplatin plus myricetin). Accordingly, animals were treated with oxaliplatin (3 mg/kg ip), myricetin (25 mg/kg iv) or matching drug vehicle, alone or in combination once weekly for up to five weeks. Oxaliplatin was given 30 mins after myricetin or its vehicle. Animals were monitored during the treatment period twice each week by general inspection, and measurement of body weight and perpendicular tumour dimensions using calipers. Animals were euthanased if they had poor general condition, greater than 15% body weight reduction or had tumours greater than 20 mm in average diameter or 10% of body weight, with ulceration or restricting movement. Body weight was expressed as percentage of baseline. Tumour volumes were calculated using the formula π/6*w*l 2 , where w = the longest diameter and l = the shortest diameter. Change in body weight and tumour volume from baseline was assessed visually by plotting mean and standard error of the mean versus time, and analysed by Two-way Repeated Measures ANOVA with Bonferroni post-tests. Time to event endpoints (quadrupling of tumour volume and euthanasia) were assessed by Kaplan-Meier plots and analysis by Gehan-Breslow-Wilcoxon test without adjustment for multiple testing. All animal procedures were approved by and performed in compliance with ethical guidelines of the Animal Ethics Committee of the University of Auckland.
Data analysis. Data were analysed using GraphPad Prism 6.2 software (San Diego, CA, USA). In vitro experiments were replicated at least three times. The significance of differences between means was assessed by ANOVA and post-tests with correction for repeated testing, as appropriate for the data. For repeated measures ANOVA, missing values (3 out of a total of 64 values (<5%)) were replaced by imputation. Differences in time to event (survival) curves were analysed using the Gehan-Breslow-Wilcoxon test without adjustment for multiple testing. A P < 0.05 was regarded as statistically significant for all statistical tests.

Data Availability
The datasets generated during and analysed during the current study are available from the corresponding author on reasonable request.