Drug-Repositioning Screens Identify Triamterene as a Selective Drug for the Treatment of DNA Mismatch Repair Deficient Cells

Germline mutations in DNA mismatch repair (MMR) genes, including MLH1, MSH2, MSH6, and PMS2, can lead to lynch syndrome, an autosomal condition also known as hereditary non-polyposis colorectal cancer (HNPCC; ref.1). Patients with this condition have an 80% lifetime risk of developing colorectal cancer and a 60% lifetime risk of developing endometrial cancer. In addition, patients are also at an increased risk of developing other cancers such as small bowel, pancreatic, prostate, urinary tract, liver, kidney, and bile duct cancer. Defects in the MMR system can also occur as a result of somatic mutations or epigenetic silencing. Significantly, it is thought that 15% of all colorectal cancers and 30% of all endometrial cancers have loss of a functional MMR pathway (2, 3). Furthermore, mutations in the MMR gene MSH6 have been identified in 26% to 41% of temozolomide-resistant glioblastoma (GBM) patients and mediate temozolomide resistance (4–6). More recently, a number of studies have shown that a reduction in MMR protein levels, in particular MSH2 and MSH6, occurs upon GBM recurrence and that transcript levels of MMR genes are prognostic for patient survival after temozolomide treatment (6–8).

Synthetic lethality with loss of DNA repair proteins has previously been successfully exploited (9–13). To date, a number of studies have identified synthetic lethal interactions with specific MMR gene mutations or specific tumor types (9, 10, 14, 15). In this study, we carried out drug-repositioning compound screens in a panel of MMR-deficient cellular models from a range of different tumor types, to identify drugs that sensitize with MMR loss in general. We identified the potassium-sparing diuretic drug triamterene as a novel therapeutic agent in MMR-deficient tumor cells. Our data suggest that the selectivity of triamterene is based on its antifolate activity and is dependent on expression of the folate synthesis enzyme, thymidylate synthase. Taken together, our data reveal that triamterene is a promising novel therapeutic strategy for the treatment of MMR-deficient disease in a range of different tumor types.

The U251.TR3 GBM cell lines were a kind gift from Dr. David Louis (Massachusetts General Hospital, MA). In the original paper (5), the nomenclature for these cell lines was A172.TR3. In a subsequent correction to the paper, this nomenclature was updated to U251.TR3 (MSH6⁻; ref.16). We have STR profiled these cell lines and confirm they originate from U251 cells. The U251 (MMR⁺), MFE-280 (MMR⁺), MFE-296 (MLH1⁻), KLE (MMR⁺), AN3CA (MLH1⁻), HEC1B (MSH6⁻), RL95-2 (MSH2⁻, MSH6⁻, MSH3⁻), and ISHIKAWA (MLH1⁻) cell lines were purchased from ATCC. The colorectal DLD1 (MSH6⁻) and DLD1+Chr2 (MMR⁺) cell lines and endometrial HEC59 (MSH2⁻) cell lines were a kind gift from Dr. Thomas Kunkel (National Institute of Environmental Health Sciences, NIEHS). The human colon cancer cell lines HCT116 (MLH1⁻) and HCT116+Chr3 (MMR⁺) were a kind gift from Dr. Alan Clark (NIEHS). DLD1 and DLD1+Chr2 cells were grown in RPMI (Sigma-Aldrich), 10% FBS, and 1% penicillin–streptomycin at 37°C with 5% CO₂. All other cell lines were grown in DMEM (Sigma-Aldrich), 10% FBS, and 1% penicillin–streptomycin at 37°C with 5% CO₂. DLD1+Chr2 and HCT116+Chr3 cells were maintained under selective pressure of 400 μg/mL geneticin (G418 sulfate; Roche). U251.TR3 cells were maintained in 100 μmol/L TMZ (Santa Cruz). All cell lines were authenticated on the basis of STR profile, viability, morphologic inspection, and were routinely mycoplasma tested.

The FDA-approved compound library incorporating 1,170 drugs was purchased from Selleck Chemicals. Cells were plated in 96-well plates and treated with vehicle (0.01% DMSO) or the compound library (average compound concentration of the library in media was 10 μmol/L). After 4 days of incubation with the drug library, cell viability was assessed using the CellTiter Glo assay (Promega) according to the manufacturer's instructions. Luminescence readings from each well were log transformed and normalized according to the median signal on each plate and then standardized by use of a Z-score statistic, using the median absolute deviation to identify the variation in each screen. Z-scores were compared with identified compounds that cause selective loss of viability in MMR-deficient cells, in comparison with MMR-proficient cells. For validation experiments, cells were treated with increasing concentrations of triamterene, and cell viability using the CellTiter Glo assay was assayed after 5 days. Triamterene was purchased from Sigma-Aldrich. N-acetyl cysteine was purchased from Santa Cruz.

Validation of triamterene was performed by colony formation assays. Cells were seeded at various densities in 6‐well plates and exposed to the drug at the indicated concentrations. Cells were retreated every 4 days, whereby the drug-containing media were removed, and fresh drug-containing media were added. After 10 to 14 days, cells were fixed and stained with sulphorhodamine‐B (Sigma) and counted.

Following triamterene treatment, cells were fixed in 70% ice-cold ethanol and stained with 4% propidium iodide (PI) and 10% RNase A in PBS for cell-cycle analysis. The sample readout was performed on the BD LSRFortessa (Becton Dickinson), and the data were analysed using FlowJo (FlowJo LLC).

DLD1 and DLD1+Chr2 cell lines (1.6 × 10⁶ cells) resuspended in PBS, were injected subcutaneously into the right flank of adult (∼10 weeks old) male NOD-SCID mice (Charles-River Laboratories). Tumors were allowed to develop to a mean tumor diameter between 4 and 8 mm before treatment. Mice were then treated 3 times a week by gavage, with 25 mg/kg Triamterene or vehicle. Tumors were measured twice weekly. Mice were sacrificed in case of sickness or when the tumors reached 1.44 cm². All animal procedures were carried out as per the Animal Scientific Procedures Act 1986 under the Home Office approval licenses (PPL-70/7275 and PIL-70/23444).

Cell pellets were lysed in 20 mmol/L Tris (pH 8), 200 mmol/L NaCl, 1 mmol/L EDTA, 0.5% (v/v) NP40, 10% glycerol, supplemented with protease inhibitors. For Western blotting, lysates were electrophoresed on Novex precast gels (Invitrogen) and immunoblotted using the following antibodies: anti-MSH6 (#5424), anti-MSH2 (#2017), anti-MLH1 (#4256), anti-thymidylate synthase (#9045), β-Actin (#4970), purchased from Cell Signaling. The following antibodies were also used: anti-MSH3 (sc-11441; Santa Cruz) and anti β-tubulin (T8328; Sigma). This was followed by incubation with anti-IgG-horseradish peroxidase and chemiluminescent detection (Supersignal West Pico Chemiluminescent Substrate, Pierce). Immunoblotting for β-Actin and β-tubulin was performed as loading control.

For siRNA transfections, cells were transfected with individual siRNA oligos (Qiagen) using Lipofectamine RNAiMax (Invitrogen) according to the manufacturer's instructions. As a control for each experiment, cells were left untransfected or transfected with a nontargeting control siRNA and concurrently analyzed.

Cellular ROS was measured using DCFDA-Cellular Reactive Oxygen Species Detection Assay Kit (Abcam, ab113851) according to the manufacturer's instructions. Briefly, to detect ROS levels, non-fluorescent 2′,7′-dichlorofluorescein diacetate (DCFDA) is converted to fluorescent DCF upon ROS (H₂O₂, •OH, ONOO⁻ and •O₂⁻) induction. Cells were plated in clear bottom black 96-well plates and treated with triamterene for the indicated times. Cells were then treated with 20 μmol/L DCFDA or incubated in assay buffer as a negative control. After 30 minutes of incubation at 37°C, cells were washed with 1X PBS and incubated for 4 hours in fresh assay buffer at 37°C in a 5% CO₂ incubator. Fluorescence was measured, using the Wallac 1420 plate reader (PerkinElmer). Each assay condition was performed in duplicate, and cell viability was measured in replicate plates using the CellTiter-Glo assay. Fluorescence DCF values were normalized to the corresponding cell viability luminescence data.

A commercially available Comet assay kit from Cell Biolabs (STA-351) was used to measure levels of cellular DNA damage. The assays were performed according to the manufacturer's instructions. Briefly, 1 × 10⁵ cells were mixed with molten agarose. DNA from embedded cells was then denatured in an alkaline solution. Samples were electrophoresed in a horizontal chamber to separate intact DNA from damaged fragments. Following electrophoresis, samples were then stained with a Vista Green DNA dye and visualized by fluorescence microscopy. Cellular DNA damage is visualized as it migrates further than intact DNA and results in a comet tail shape. For assessment of the Comet assay, 50 comets were scored per condition and ImageJ was used to quantify the intensity and score the comets using the following calculation: Tail DNA % = 100 × Tail DNA intensity/cell DNA intensity.

A commercially available ELISA kit from Cell Biolabs (STA-320) was used to measure levels of 8‐oxodG. Although this 8-oxodG ELISA has potential shortcomings for the precise quantification of 8-oxodG levels, it allows an estimation of the change of 8-oxodG levels upon triamterene treatment. Genomic DNA was extracted using the QIamp DNA isolation kit (Qiagen), digested with nuclease P1, treated with calf intestinal phosphatase and denatured. To avoid artifactual production of 8-oxodG, we used a phenol-free method of DNA isolation, and DNA was completely digested. The assays were performed according to the manufacturer's instructions. Briefly, 10 to 15 μg DNA from untreated and treated cells or the 8‐oxodG standard (0.078–20 ng/mL) was incubated with an 8‐oxodG monoclonal antibody in an 8‐oxodG-precoated microtiter plate. The assay was normalized by using an equal amount of DNA for each sample. Standard curves were calculated with serial dilutions of 8‐oxodG standard to calculate reaction efficiency. Samples were assayed in triplicate.

Cells were seeded onto poly-lysine coated coverslips and treated with drugs as indicated. After 48-hour treatment, cells were fixed for 10 minutes with 4% PFA in PBS. Cells were then permeabilized with Triton, blocked for 1 hour at room temperature and subsequently incubated with γH2AX antibody (#05-636, Millipore) for 18 hours at 4°C. This was followed by incubation with anti-IgG-Alexa568 (#A11031, Invitrogen). Coverslips were then washed in 4% DAPI/1X PBS and mounted with ProLong gold antifade mounting solution (Invitrogen). Slides were imaged using a Zeiss LSM 510 confocal microscope. Per condition, a minimum of 300 cells were counted and quantified for γH2AX positive cells (>5 foci per nucleus).

Unless otherwise stated, data represent standard error of the mean of at least three independent experiments. The two-tailed paired Student t test was used to determine statistical significant with P < 0.05 regarded as significant. For confocal experiments, images are representative of at least three independent experiments, where a minimum of 300 cells were analyzed.

To identify compounds that can sensitize MMR-deficient cells, we screened a large panel of cell lines with a range of different MMR gene mutations from a number of tumor types. These included the MSH6-deficient colorectal cancer cell line DLD1 and its isogenic MSH6-proficient DLD1+Chr2 cell line (Fig. 1A), the previously characterized temozolomide-resistant MSH6-deficient U251.TR3 GBM cell line and the isogenic MSH6-proficient U251 cell line (Fig. 1B refs.5, 16) and a panel of endometrial cancer cell lines; KLE (MMR-proficient), MFE-280 (MMR-proficient), MFE-296 (MLH1-deficient), ISHIKAWA (MLH1-deficient), and HEC1B (MSH6-deficient; Fig. 1C).

Based on the concept of drug repositioning, of identifying previously approved compounds for new clinical indications, cells were screened in the presence of either vehicle (DMSO) or a compound library comprising 1,170 FDA-approved drugs. This approach aimed to identify compounds with previous unknown potential for repurposing as MMR-selective drugs. Analysis of our screens revealed that the potassium-sparing diuretic compound triamterene was a promising candidate for a new MMR-selective drug. Validation experiments revealed that, although the sensitization was variable depending on the MMR-defect, however when compared with the MMR-proficient cells, treatment with triamterene induced toxicity over a range of concentrations specifically in MMR-deficient cells (Fig. 1D–F). Triamterene also caused sensitivity in MMR-deficient cells, in comparison with MMR-proficient cells, in a clonogenic survival assay (Fig. 1G).

To further investigate this selectivity for MMR deficiency, we measured cell viability of the MLH1-deficient colorectal cancer cell line HCT116 and its isogenic matched-paired MLH1-proficient cell line, HCT116+Chr3, when treated with triamterene (Fig. 1H). Significantly, triamterene also induced selectivity in the MLH1-deficient HCT116 cells, but not in MLH1-proficient cells. Significantly, we observed selectivity in all MMR-deficient cell lines tested, regardless of MMR mutation or tumor type, while no significant effect was observed in MMR-proficient cells. This suggests that triamterene is selective with loss of MMR pathway function and may provide a novel therapeutic strategy in a wide range of cancers.

Previous studies have shown that triamterene can act as both an inhibitor of the epithelial sodium channel (ENaC; ref.17) and also as an antifolate (18). Therefore, to determine the mechanism underlying the observed triamterene-induced cytotoxicity with MMR deficiency, we performed siRNA-mediated depletion of a panel of ENaC isoforms (α, β, and γ) previously suggested to be inhibited by triamterene and analyzed cell viability (Fig. 2A). We observed no differential cytotoxicity upon targeting the ENaC isoforms, either alone or in combination, which suggests that the observed selectivity is not due to ENaC inhibition. We next determined whether triamterene-mediated selectivity was due to its antifolate activity. To this end, we measured the cell viability of MMR-deficient and proficient cells treated with folates (dihydrofolate and tetrahydrofolate) in addition to triamterene. We observed that the selectivity of triamterene for MMR deficiency could be rescued by the addition of folates (Fig. 2B and C), therefore suggesting that triamterene is selective through its antifolate activity.

To further analyze the antifolate effect of triamterene, we analyzed the requirement for thymidylate synthase in triamterene-mediated selectivity. Thymidylate synthase is the only de novo enzyme for dTMP synthesis. It catalyzes the reductive transfer of a methyl group from N⁵,N¹⁰-methylenetetrahydrofolate (CH₂-THF) to dUMP, forming dTMP and dihydrofolate (DHF). Our results suggest that thymidylate synthase protein expression is not altered upon Triamterene treatment (Fig. 2D). However, silencing thymidylate synthase by siRNA prevents triamterene-induced lethality in MMR-deficient cells (Fig. 2E and F; Supplementary Fig. S1A). These results suggest that thymidylate synthase expression is necessary for the triamterene-mediated selectivity in MMR-deficient cells. Furthermore, treatment with the clinically approved thymidylate synthase inhibitors, 5-FU and raltitrexed, also rescued the triamterene-induced cytotoxicity in MMR-deficient cells (Supplementary Fig. S1B). Taken together, our results suggest that triamterene-induced selectivity is due to the antifolate activity of triamterene and is dependent on thymidylate synthase expression.

It has previously been shown that folate starvation can increase ROS levels, leading to cellular oxidative stress (19). Our previous studies have shown that an increase in oxidative stress is synthetically lethal with MMR deficiency (9, 10, 14, 15). Therefore, we investigated whether triamterene can induce an increase in ROS levels due to folate inhibition in MMR-deficient and -proficient cells. To this end, we treated MMR-deficient and -proficient cells with increasing concentrations of triamterene and measured ROS levels (Fig. 3A and B). Our results show a greater increase in the level of ROS in triamterene-treated MMR-deficient cells, in comparison with MMR-proficient cells. To further investigate if this increase in ROS levels in MMR-deficient cells was responsible for triamterene selectivity, we treated cells with triamterene alone or in combination with the ROS scavenger, N-acetylcysteine (NAC; Fig. 3C). Our results demonstrate that the triamterene-induced selectivity in MMR-deficient cells can be rescued by addition of NAC, which suggests that increased ROS levels are, at least in part, the mechanism of toxicity upon triamterene treatment. Our data indicate the importance of thymidylate synthase expression in triamterene-induced selectivity. To further investigate this, we analyzed ROS levels upon thymidylate synthase silencing and triamterene treatment (Fig. 3D and E). Interestingly, we observed that silencing thymidylate synthase by siRNA prevents the triamterene-induced increase in ROS levels. These results suggest that thymidylate synthase is required for ROS accumulation, leading to triamterene cytotoxicity.

Our results suggest that MMR-deficient cells have reduced cellular viability upon triamterene treatment. To investigate the mechanism of this selectivity further, we stained cells, before and after triamterene treatment, with propidium iodide and measured cells by flow cytometry to determine which phase of the cell cycle they accumulated in after treatment (Fig. 4A). Interestingly, our results suggest that upon triamterene treatment, MMR-deficient cells arrest in the G₂–M phase of the cell cycle, corresponding to the reduced cellular viability we observed. An increase in ROS levels can lead to cellular DNA damage, including DNA single-strand breaks, alkali-labile sites, oxidative DNA damage and ultimately an increase in DNA double-strand breaks (DSB), thereby triggering cell-cycle arrest. To assess if triamterene induces DNA damage in MMR-deficient cells, we measured cellular DNA damage, including DNA fragmentation and DNA strand breaks using the comet assay (Fig. 4B). Treatment with triamterene resulted in an increase in DNA damage in the MMR-deficient cells only. To further investigate the type of DNA damage induced, we measured accumulation of the oxidative DNA lesion, 8-Oxo-2′-deoxyguanosine (8‐oxodG) using an ELISA assay (Fig. 4C). We observed a significant increase in 8‐oxodG DNA lesions in the triamterene-treated MMR-deficient cells only. To determine whether this oxidatively damaged DNA, resulted in DNA DSB triggering cell-cycle arrest, we measured γH2AX foci, a marker for DNA DSBs, by confocal microscopy (Fig. 5A–D). Treatment with triamterene induced an increase in γH2AX foci in MMR-deficient cells, which can be rescued by addition of the ROS scavenger, NAC. Taken together, our results suggest that triamterene treatment leads to an increase in ROS levels in MMR-deficient cells, which ultimately leads to an increase in 8-oxodG DNA lesions and DNA DSBs, resulting in cell-cycle arrest (Fig. 4D).

To examine the in vivo efficacy of triamterene, the MMR-deficient and MMR-proficient DLD1 and DLD1+Chr2, respectively, colorectal cancer cells, were injected subcutaneously into NOD-SCID mice. Xenografted mice were subjected to treatment 3 times a week with triamterene or vehicle (PBS). We observed that tumor growth from the MMR-deficient xenografts was significantly reduced (P = 0.01) by triamterene treatment when compared with vehicle (Fig. 6A). In addition, no difference in tumor growth was observed in the triamterene-treated MMR-proficient xenograft tumors when compared with vehicle, further validating triamterene as a compound that specifically targets cancer cells with deficient MMR pathways (Fig. 6B). Taken together, these in vivo observations further indicate that triamterene treatment has potential clinical utility in patients with MMR deficient, for which therapeutic options are scarce.

Through a compound screen, we identified the diuretic drug triamterene as an in vitro and in vivo selective compound with MMR deficiency. Collectively, our data argue that stratifying patients according to their MMR status would prove efficacious with regard to treatment with triamterene. Furthermore, our data suggest that levels of thymidylate synthase govern sensitivity to triamterene in MMR-deficient cells. Upon thymidylate synthase inhibition, treatment with triamterene was no longer selective, suggesting that thymidylate synthase levels determine the balance between triamterene resistance and sensitivity in these cells. Our data suggest that folate inhibition and thymidylate synthase are critical for the triamterene selectivity in MMR-deficient cells; however, we cannot exclude the fact that other pathways may also be influencing this effect, at least in part. One of the most intriguing findings from this work is the requirement for thymidylate synthase for the induction of ROS upon triamterene treatment. It is likely due to an activity of thymidylate synthase in the absence of sufficient levels of methyl donor CH2-THF, thus leading to ROS. It would be interesting to understand whether thymidylate synthase is regulating ROS levels through NAPDH oxidase complexes or providing protection against ROS production through the antioxidant response. Numerous studies have investigated ROS induction upon treatment with thymidylate synthase inhibitors, but, to our knowledge, no study has identified a role for thymidylate synthase in the regulation of ROS levels. Previously, antifolate agents targeting the folate metabolic enzyme dihydrofolate reductase (DHFR), such as methotrexate and pemetrexed, were identified as cytotoxic in MSH2-deficient cells, but not in other MMR-deficient cell lines (9, 20). Perhaps this is due to the generation of specific DNA lesions upon DHFR inhibition in MSH2-deficient cells, rather than a more general inhibition of folate metabolism upon triamterene treatment. Taken together, these data also suggest a potential difference in the folate metabolic pathway in MMR-deficient and -proficient cell lines.

An increase in ROS levels can lead to cellular DNA damage, including DNA fragmentation, oxidative DNA damage and ultimately an increase in DNA DSB, thereby triggering cell-cycle arrest. We observed a significant increase in cellular DNA damage, and more specifically 8-oxodG DNA lesions in the triamterene-treated MMR-deficient cells only. Our data suggest that these lesions, if incompletely repaired, can induce DNA DSBs, which ultimately result in the cell-cycle arrest and reduced cellular viability we observe.

We and others have focused largely on targeting loss of DNA repair in tumor cells. However, recent advances indicate that targeting both the tumor cell and its interaction with the immune microenvironment may significantly improve patient benefit. A recent phase II clinical trial in patients with deficiency in the MMR pathway indicated that MMR status predicted clinical benefit with the PD-1 inhibitor pembrolizumab (21). However, only 50% of MMR-deficient patients responded to pembrolizumab, suggesting that selectively targeting MMR-deficient tumor cells with for example triamterene, in combination with immune checkpoint inhibitors may increase therapeutic efficacy and may prove to be more clinically beneficial.

In this study, we exploit the concept of drug repurposing, which is the discovery of new indications for existing drugs, to identify novel selective drugs for the treatment of MMR-deficient tumors. Repositioning of drugs highlights an increasingly effective means of drug discovery. In addition to the reduced cost and time commonly associated with traditional drug discovery, the advantage of drug repositioning strategies is the fact that existing drugs have already been used in patients and, therefore, their toxicity and safety profiles are already established. Hence, drugs identified in drug repositioning approaches can enter clinical trials rapidly, thereby maximizing their potential benefit to patients. Here, we have shown for the first time that triamterene, originally developed as a diuretic, has antitumor activity and we provide evidence of its efficacy in a range of MMR-deficient tumor types.

No potential conflicts of interest were disclosed.

Conception and design: D. Guillotin, S.A. Martin

Development of methodology: D. Guillotin, S. Short, S. Marino

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Guillotin, P. Austin, R. Begum, M.O. Freitas, A. Merve, T. Brend, S. Short

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Guillotin, M.O. Freitas, S. Marino, S.A. Martin

Writing, review, and/or revision of the manuscript: D. Guillotin, T. Brend, S. Short, S. Marino, S.A. Martin

Study supervision: S.A. Martin

We thank Dr. D. Louis, Dr. T. Kunkel, and Dr. A. Clarke for the provision of cell lines. We thank J. Andow-Cleaver, T. Chaplin-Perkins, and H. Schmidt for their help with animal work. This research was supported by the BCI Flow Cytometry and Microscopy Facilities. We also thank Dr. P. S. Ribeiro and Dr. T. V. Sharp for critical reading of the manuscript. We also thank all members of the Martin lab for helpful discussions.

This work was supported by funding from the Barts and The London Charity (467/1505), Wellbeing of Women charity supported by Sanctuary Spa (RG1629), and Cancer Research UK (C16420/A18066).

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.