The Mitochondrial Disruptor Devimistat (CPI-613) Synergizes with Genotoxic Anticancer Drugs in Colorectal Cancer Therapy in a Bim-Dependent Manner

Colorectal cancer is among the most prevalent types of cancer worldwide and exhibits an increasing incidence and mortality rate in low- and middle-income countries (1). Not only are elderly age groups affected, but also an elevated number of cases is observed in young adults (2). Several risk factors are associated with colorectal cancer including nutrition (3), lack of physical activity (4), as well as genetic predisposition (5). The treatment of colorectal cancer includes surgical resection of the primary tumor, generally in combination with chemotherapy and radiation, depending on the stage and localization of the disease (6). Drugs usually applied are the antimetabolite 5-fluorouracil (5-FU) and folinic acid (leucovorin) combined with another DNA-damaging agent [irinotecan (IT) or oxaliplatin] used in a specific regimen (FOLFOX or FOLFIRI; ref. 6). Additionally, highly specific treatment with antiangiogenic drugs (e.g., bevacizumab) or epidermal growth factor receptor inhibitors (e.g., cetuximab) emerged in the last decade as targeted options for first- or second-line therapy of colorectal cancer, respectively (7).

More recently, a new class of anticancer drugs was developed, derived from the endogenous enzymatic cofactor α-lipoic acid (LA), which plays a crucial role in mitochondrial energy production and was shown to have cytotoxic effects in several types of cancer cells (8). Devimistat (CPI-613) is a derivative of LA, containing two benzyl rings covalently bound to the sulfur atoms of LA, rendering it non-redox active (9). Devimistat acts as an inhibitor of the mitochondrial enzymes pyruvate dehydrogenase (PDH) and α-ketoglutarate dehydrogenase (KGDH), thereby interfering with the altered energy metabolism of tumor cells (10). The mechanism of enzymatic inhibition differs between both enzymes: PDH activity is impaired indirectly due to induction of its negative regulator pyruvate dehydrogenase kinase, whereas the enzyme KGDH is inactivated by a devimistat-induced reactive oxygen species (ROS) burst (9, 11).

The cytotoxic activity of devimistat has been demonstrated in various cancer cell lines, including mainly pancreatic, kidney, lung, breast, ovarian, and prostate cancer cells, whereas moderate impact on untransformed cells was observed (9, 11, 12). The underlying mechanism of cytotoxicity included induction of apoptosis as well as caspase-independent cell death pathways, as revealed in pancreatic and lung cancer cells (11, 12). As an inhibitor of mitochondrial key enzymes, devimistat was found to target the altered energy metabolism of tumor cells: low serum and glucose conditions as well as carnosine, an inhibitor of glycolytic ATP production, synergistically enhanced the cytotoxicity of devimistat (9, 13). Several years ago, devimistat entered clinical testing and progressed especially for the treatment of acute myeloid leukemia and pancreatic cancer (14, 15). Clinical studies demonstrated beneficial effects by combining devimistat with standard chemotherapeutics such as antimetabolites, indicated by an increased overall and progression-free survival, patient performance, and response rate (14–17). Yet, the underlying mechanisms of this interplay are only partially understood.

In our study, we aimed to analyze the potential synergistic effects of devimistat with 5-FU and IT in the treatment of colorectal cancer in vitro and in vivo. To this end, we screened a panel of established human colorectal cancer cell lines including patient-derived short-term cultures and normal human colonic epithelial cells (HCEC) for their sensitivity toward devimistat and the chemotherapeutics 5-FU and IT. Tumor selectivity of devimistat was further addressed using murine intestinal organoids and intestinal tumoroids. The mode of action of devimistat in colorectal cancer cells was detailed by cell death measurements, assessment of mitochondrial function, and genotoxicity studies. The potential synergism of devimistat with chemotherapeutics was then investigated in colorectal cancer cells using Combenefit modeling and Chou–Talalay analysis. Underlying mechanisms such as apoptosis, cell-cycle regulation, and genotoxicity were analyzed by various methods such as qPCR, western blot detection, flow cytometry, and genetic knockdown experiments. Finally, these findings were translated into two xenograft mouse models and supported by tumor xenograft analysis.

Cell lines were cultured under sterile conditions at 37°C and 5% CO₂ in humidified atmosphere. The human colorectal cancer cell lines included HCT116-p53^+/+, HCT116-p53^−/− (both provided by Prof. Bert Vogelstein, John Hopkins University, Baltimore in 2012), Caco-2, DLD-1, HT29, RKO, SW48, SW480, and LS174T (all obtained from the Institute of Toxicology, University Medical Center Mainz between 2012 and 2015). Cells were reauthenticated by the p53 status, their differential response to genotoxic anticancer drugs such as 5-FU and IT (see Table 1) as well as their typical cell morphology. The primary patient-derived cell lines HROC60 and HROC278 were established as described previously (18). Cells were grown in corresponding medium (DMEM: HCT116, RKO, DLD-1; RPMI: SW48, HT29, SW480; IMDM: LS174T; DMEM/Ham F-12 1:1: HROC60, HROC278: MEM: Caco-2) supplemented with 10% fetal calf serum and 1% penicillin/streptomycin. Colorectal cancer cells were routinely maintained in culture for 20–25 passages. Media and supplements were obtained from Gibco Life Technologies or Pan-Biotech. Nontransformed HCEC (1CT) were kindly provided by Prof. Jerry W. Shay (Department of Cell Biology, UT Southwestern Medical Center, Dallas) in 2015 and reauthenticated by their p53 status and their typical cell morphology. Cells were isolated from normal human colonic biopsies and immortalized by CDK4- and hTERT-transfection as described (19). HCEC were grown in a nitrogen incubator with reduced oxygen levels (7% O₂) and 5% CO₂ at 37°C in 4:1 DMEM GlutaMax/Medium 199 (Thermo Fisher Scientific) with supplements as reported previously (20). HCEC were usually cultured for 8–10 passages at maximum. Cell lines were mycoplasma negative as routinely demonstrated by PCR using the VenorGeM Classic kit (Minerva Biolabs) and immunofluorescence microscopy with nuclear staining.

Knockdown of Bim was performed using siGENOME SMARTpool siRNA obtained from Dharmacon. Nonsense, scrambled RNA served as control. Transfections were conducted as reported (20), and successful knockdown was confirmed by western blot analysis as detailed below.

Devimistat (CPI-613; Hycultec GmbH) was dissolved in dimethylsulfoxide (DMSO) and stock solutions (125 mmol/L) were stored at −20°C. 5-FU (medac) and IT (Fresenius Kabi) were obtained from the pharmacy of the University Medical Center Mainz and prepared as solution in 0.9% saline solution stored at −20°C. In the combination regimen, cytostatic drug treatment was performed 2 hours after treatment with devimistat.

Preparation of total RNA, transcription into cDNA, and quantitative gene-expression analysis by real-time PCR was performed as described (21) using the primers listed in Supplementary Table S1. In three independent experiments, qPCR was conducted in technical duplicates with a CFX96 Real-Time PCR Detection System (Bio-Rad). The subsequent analysis was performed using CFX Manager Software. Nontranscribed controls were included in each run. Gene-expression levels were normalized to GAPDH as well as ACTB, and the solvent control was set to one.

Protein extracts were generated from either cell culture experiments or xenograft mouse studies (22). Whole-cell lysis was performed in RIPA buffer containing 25 mmol/L Tris-HCl pH 8.0, 2 mmol/L EDTA, 150 mmol/L NaCl, 1% NP-40, 0.5% Na-deoxycholate, 0.1% SDS supplemented with cOmplete protease inhibitor cocktail (Roche Diagnostics), 1 mmol/L Na₃VO₄, 2 mmol/L NaF, and 1 mmol/L PMSF for 15 minutes on ice, followed by sonication and centrifugation. Protein extracts of tumor xenografts were obtained by homogenization of snap-frozen isolated tumor tissue in RIPA buffer with slight modifications. Clarified extracts were determined for their protein content using the Bradford assay.

Western blot analysis was essentially conducted as described previously (23), and proteins were detected using a c300 chemiluminescence imager (Azure Biosystems). The used primary and secondary antibodies are detailed in Supplementary Table S2.

The In-Cell Western (ICW) analysis of γH2AX as DNA strand break marker was performed as previously described (24). Samples were analyzed using an Odyssey Infrared Imaging Scanner (LI-COR Biosciences) by simultaneous measurement of γH2AX and DNA staining. Statistical analysis was performed using Image Studio Lite (Version 5.2; LI-COR Biosciences) by determining the x-fold increase in γH2AX signal normalized to the DNA content per well in relation to the control.

Quantitative analysis of ROS levels was performed using the probe CM-H₂DCFDA (Invitrogen) as described (25). Staining was performed for 30 minutes at 37°C in the dark in phenol red- and serum-free medium, and samples were measured by flow cytometry using a BD Canto II and BD FACS Diva software 6.0 (BD Biosciences).

The alkaline Comet assay and the Fpg-modified alkaline Comet assay were essentially performed as described (26). Samples were analyzed by microscopy, and at least 50 cells per sample were scored using Comet IV software (Perceptive Instruments Ltd.). Olive tail moment (OTM) was used for the quantification of DNA damage.

MitoTracker Orange CMTMRos (Thermo Fisher Scientific) staining was used to visualize the relative amount of intact mitochondria. Cells were seeded on cover slips and treated as indicated. Upon the desired incubation time, cells were stained using 250 nmol/L MitoTracker for 15 minutes and then fixed with 4% paraformaldehyde. Following mounting using VectaShield with DAPI (Vector Labs), cells were analyzed using a Zeiss Axio Observer 7 microscope equipped with an LSM900 confocal laser scanner (Zeiss) as reported (24). Images were processed using ZEN software (Zeiss) and ImageJ software (NIH).

Cell death induction was measured using Annexin V/PI staining and flow-cytometric analysis as reported (27). Cells were incubated as indicated, harvested, and washed with PBS. Cells were then stained with Annexin V (Miltenyi Biotec) for 15 minutes on ice in binding buffer (5% dye; 10 mmol/L HEPES pH 7.4, 140 mmol/L NaCl, 2.5 mmol/L CaCl₂, and 0.1% BSA) followed by the addition of propidium iodide (2% dye in binding buffer). Cells were analyzed on a BD Canto II and the corresponding BD FACS Diva software 6.0 (BD Biosciences), which was used for gating of living cells (Annexin V/PI double negative), early apoptotic cells (Annexin V-positive, PI-negative), and late apoptotic/necrotic cells (Annexin V/PI double positive).

Cell-cycle distribution was assessed as reported previously (28). Cells were treated and incubated for various time periods as indicated. Cells were then harvested, washed with PBS, and incubated in 80% ethanol at −20°C overnight. Cells were washed with PBS, incubated with RNase A (Sigma) for 1 hour, and stained by propidium iodide (Sigma). Finally, cells were analyzed by flow cytometry using a BD Canto II and the corresponding BD FACS Diva software 6.0 (BD Biosciences).

Cell viability was determined by the CellTiter-Glo Luminescent Cell Viability Assay Kit (Promega). According to the manufacturer's protocol, cells were seeded in white 96-well plates, treated as indicated, and incubated for 72 hours. The medium was then exchanged for medium containing 50% CellTiter-Glo Luminescent Solution and analyzed with the luminometer Fluoroskan Ascent FL (Thermo Scientific). In order to determine the combination index (CI) by the Chou–Talalay method (29), experiments were performed as explained previously using CompuSyn software (CompuSyn Inc.; ref. 30). Simultaneously, Combenefit analysis was undertaken as described elsewhere (31).

Mitochondrial membrane potential was analyzed using the dye JC-1 (5′,6,6′-tetrachloro-1,1′,3,3′-tetraethyl-benzamidazolylcarbocyanine iodide; Sigma) as reported (27). Valinomycin served as positive control. Staining was performed according to the manufacturer's protocol of the Mitochondrial Staining Kit (Sigma). Monomeric cytosolic JC-1 in the cytoplasm exhibits green fluorescence emission (530 nm), whereas aggregated JC-1 in intact mitochondria exhibits red fluorescence emission (585 nm). This is monitored by flow cytometry using a BD Canto II and BD FACS Diva software 6.0 (BD Biosciences).

Mitochondrial dysfunction of colorectal cancer cells upon treatment with devimistat was assessed in Seahorse XF96 Cell Culture Microplates (Agilent). In brief, medium was replaced with Seahorse assay medium (glucose-free DMEM supplemented with 2 mmol/L glutamine, 1 mmol/L pyruvate, and 25 mmol/L glucose) at the day of measurement. Oxygen consumption rate (OCR) was measured with Seahorse XF396 Analyzer (Agilent). Mitochondrial ATP production was defined as the difference between basal OCR reading and the OCR following the injection of 2 μmol/L oligomycin. Maximal respiration was defined as the difference in OCR after injection of 1 μmol/L CCCP, which uncouples the electron transport, and OCR after injection of a mixture of 2.5 μmol/L rotenone/antimycin A (Anti A), inhibiting complex I and III.

Murine intestinal organoids and tumor organoids were generated and maintained as described previously (20, 32). Viability of organoids following treatment with increasing devimistat concentrations was assessed using the MTS assay and absorbance measurements in a Sirius HT reader (MWG Biotech). Solvent-treated organoids were defined as 100% viable. Furthermore, organoids were analyzed for morphologic changes and cell death induction using PI/Hoechst 33342 costaining and live-cell microscopy as reported (33).

Male Balb/c^nu/nu mice were purchased from Janvier Laboratories and maintained at 25°C and 65% humidified atmosphere at the animal facility of the University Medical Center Mainz or the Justus Liebig University Gießen. Mice were housed at a 12 hours day and night cycle with food and drinking water supply ad libitum. Following acclimatization, mice were inoculated subcutaneously with either HCT116 or HT29 cells (5 × 10⁶ and 10⁶ cells/flank) into the right and left flank each as reported previously (34). After 7 days of xenograft tumor growth, animals were randomized into four treatment groups and received tumor therapy for up to four consecutive weeks. Animals received i.p. injections with either devimistat (25 mg/kg bw, twice per week, dissolved in 1 mol/L triethanolamine, injected with 5% dextrose in water) or IT (40 mg/kg bw, once per week, dissolved in and injected with 0.9% NaCl) or the combination of both or the corresponding vehicle control. Three times per week, body weight was monitored and tumor diameters (length, width, and volume) were measured using a sliding caliper. At the end of the experiments, animals were sacrificed by cervical dislocation. Tumor xenografts were isolated, snap-frozen in liquid nitrogen, and stored at −80°C until further analysis.

All animal experiments were approved by the government of Hessia or Rhineland-Palatinate and the Animal Care and Use Committee at the JLU Gießen and the JGU Mainz. All animal studies were performed according to German federal law and the guidelines for the protection of animals.

Experiments were performed independently at least three times, except otherwise stated. Results from representative experiments are shown. Values underwent Grubb test to exclude outliers and are displayed as mean + standard error of the mean (SEM) using the GraphPad Prism 8.0 Software (GraphPad Software Inc.). Statistical analysis (Student t test, log-rank test, two-way-ANOVA) was performed as indicated in figure legends, and statistical significance was defined as P < 0.05.

The data generated in this study are available within the article and its supplementary data files.

Initially, we screened a panel of colorectal cancer cell lines including patient-derived short-term cultures for their sensitivity toward devimistat and established anticancer drugs used in colorectal cancer therapy, 5-FU and IT (Table 1). The molecular colorectal cancer subtypes and genetic features clearly affected the cytotoxicity of both 5-FU and IT, which were generally more potent in colorectal cancer cells with microsatellite instability (MSI) and wild-type p53. In contrast, microsatellite stable (MSS) colorectal cancer cells with mutated p53 displayed higher IC₅₀ values and were thus more resistant. Patient-derived short-term cultures (HROC60, HROC278) exhibit a high resistance particularly to 5-FU treatment. Interestingly, devimistat showed comparable toxicity in all tested colorectal cancer cells independent of the molecular subtype. Furthermore, nonmalignant HCECs were less vulnerable to devimistat, whereas 5-FU and IT caused strong toxicity. In contrast to HCEC, the viability of both HT29 and HCT116 colorectal cancer cells was strongly reduced upon treatment with the same devimistat concentration (Supplementary Fig. S1A).

In order to analyze the impact of p53 on cytotoxicity induced by devimistat, isogenic p53-proficient and -deficient HCT116 cells were used and exposed to devimistat. Our findings clearly showed a p53-independent cytotoxicity of devimistat (Fig. 1A). Further experiments provided evidence that devimistat exerts its cytotoxicity not only in proliferating colorectal cancer cells, but also in colorectal cancer cells pretreated with 5-FU or IT (Fig. 1B; Supplementary Fig. S1B and S1C). In opposition to that, both anticancer drugs lost their cytotoxic effects in arrested colorectal cancer cells rechallenged with the drugs. Flow cytometry–based cell-cycle analysis in HCT116 and HT29 cells revealed a reduction of cells in S- and G₂–M-phase and an induction of cells in sub-G₁, which was more prominent in HCT116 cells (Fig. 1C; Supplementary Fig. S1D–S1F). In line with these findings, devimistat caused cell death as shown by Annexin V/PI staining (Fig. 1D; Supplementary Fig. S1G). Finally, the cytotoxicity of devimistat was analyzed in murine intestinal organoids and intestinal tumoroids. Devimistat had almost no effects on normal intestinal organoids as attested by immunofluorescence microscopy and MTS viability assay (Fig. 1E and F). In contrast, devimistat caused a concentration-dependent decrease in viability together with severe morphologic changes and cell death in tumoroids (Fig. 1G and H). Taken together, devimistat was cytotoxic in a plethora of colorectal cancer cell lines with similar potency independent of the molecular colorectal cancer subtype as well as in intestinal tumoroids. Importantly, devimistat showed moderate to low toxicity in normal HCEC as well as normal intestinal organoids, and was active in pretreated and arrested cells, making it an attractive drug candidate for colorectal cancer therapy.

Because devimistat was reported to target enzymes involved in the mitochondrial TCA cycle and energy metabolism, metabolic measurements of OCRs were carried out in HCT116 and HT29 cells. Using the Cell Mito Stress Test kit (Seahorse assay) with oligomycin, CCCP and a combination of rotenone (Rot) and antimycin A (Anti A), oxidative phosphorylation (OXPHOS) was analyzed in the presence of increasing devimistat concentrations. Although 100 μmol/L devimistat had no effect on oxygen consumption rate, 150 μmol/L devimistat strongly impaired basal (oligomycin), ATP production-related (CCCP) and maximal (Rot/Anti A) respiration in both colorectal cancer cell lines (Fig. 2A and B; Supplementary Fig. S2A and S2B). This was preceded by a dose-dependent burst of ROS triggered by devimistat, which was shown by H₂DCFDA staining and also observed under hypoxic conditions (Fig. 2C and D; Supplementary Fig. S2C and S2D). Concomitantly, changes in mitochondrial membrane potential (MMP) were assessed by JC-1 dye. This staining revealed an elevated number of cells with JC-1 monomers upon devimistat treatment and, thus, indicated loss of MMP (Fig. 2E and F). A similar effect was observed with the positive control rotenone and with the ionophore valinomycin, which is known to depolarize mitochondria (Fig. 2E and F). As a next step, mitochondrial integrity was analyzed using MitoTracker Orange staining and confocal microscopy. This dye only accumulates in mitochondria with intact MMP and thus directly reflects intact mitochondria (35). Devimistat treatment caused a reduction of intact mitochondria in HCT116 cells, which was slightly more pronounced in cells exposed to the mitochondrial complex I inhibitor rotenone (Fig. 2G and H). In summary, devimistat increases ROS generation and disrupts MMP in colorectal cancer cell lines, thereby leading to mitochondrial dysfunction and impairment of OXPHOS.

Next, we wished to know whether devimistat is genotoxic similar to the established anticancer drugs 5-FU and IT. To this end, HCT116 cells were incubated with increasing concentrations of devimistat, and the DNA damage marker phosphorylated histone 2AX (γH2AX) was studied by ICW analysis. No γH2AX formation was detected after devimistat treatment, whereas 5-FU clearly increased γH2AX levels (Fig. 3A; Supplementary Fig. S3A). Furthermore, an alkaline Comet assay was performed, which reveals apurinic/apyrimidinic (AP) sites, DNA single-strand breaks and DNA double-strand breaks. However, no DNA damage formation was observed following devimistat treatment (Fig. 3B; Supplementary Fig. S3B). These experiments were repeated in HT29 cells, confirming lack of genotoxicity for devimistat (Supplementary Fig. S3C–S3F). Because we observed ROS formation by devimistat treatment (compare Fig. 2C), we also used a Fpg-modified Comet assay that allows for the detection of oxidative DNA damage. Devimistat caused a moderate, but not statistically significant, increase at the highest concentration tested (200 μmol/L; Fig. 3C and D). The positive control tBOOH induced both DNA strand breaks and oxidative DNA damage as revealed by the Fpg-modified Comet assay (Fig. 3C and D). In conclusion, our experiments demonstrated a nongenotoxic mode of action for devimistat, which is an important advantage over genotoxic anticancer drugs in terms of damage to healthy tissue.

Due to its nongenotoxic mitochondria-targeted mode of action, we hypothesized that devimistat should prove valuable in combination regimens with the DNA-damaging anticancer drugs 5-FU and IT. Therefore, mono- and combination treatments were performed in HCT116 cells, and viability was assessed by ATP assays. CIs were calculated according to the Chou-Talalay method (29), revealing a clear synergism of devimistat with both anticancer drugs (Supplementary Table S3). In order to determine the optimal concentration for combination therapy, a wide range of concentrations was tested followed by Combenefit modeling. This analysis also demonstrated synergy for devimistat at a concentration between 150 and 200 μmol/L with nearly all tested concentrations of 5-FU and IT (Fig. 4A). The same set of experiments was conducted in HT29 cells. Although Chou–Talalay analysis displayed clear synergism for devimistat with 5-FU and a moderate antagonism for its combination with IT, Combenefit modeling revealed synergistic effects for all combination regimens (Supplementary Table S3; Supplementary Fig. S4A).

Subsequently, we wanted to detail the mechanisms underlying the synergism between devimistat and the genotoxic anticancer drugs. HCT116 cells were exposed to devimistat, 5-FU, and IT or combinations thereof followed by cell death analysis with Annexin V/PI staining. As expected, all single treatments caused cell death to a certain degree (Fig. 4B; Supplementary Fig. S4B). However, the combination of devimistat with either 5-FU or IT strongly increased cell death in a synergistic manner (Fig. 4B). Cell death induction was also moderately augmented in HT29 cells treated with the combination regimens as compared with the single treatments (Supplementary Fig. S4C and S4D). Next, the expression levels of antiapoptotic and proapoptotic genes in HCT116 cells were determined by qPCR. Interestingly, the combination devimistat/5-FU as well as the combination devimistat/IT resulted in decreased expression of the antiapoptotic genes Bcl-X_L and Survivin (Fig. 4C). Proapoptotic gene expression was induced by the genotoxic anticancer drugs 5-FU and IT, but was hardly modulated by addition of devimistat except for FASR expression levels that were augmented in the combination devimistat/IT (Supplementary Fig. S5A and S5B). Interestingly, we observed that devimistat in a combination regimen with 5-FU or IT strongly increased Bim protein levels in HCT116 cells (Fig. 4D), although no effects were revealed on gene-expression level (Supplementary Fig. S5A). A similar upregulation of Bim protein was also detected in HT29 cells in the combination treatments (Supplementary Fig. S5C). Using p53-proficient and -deficient HCT116 cells, we found that Bim induction was even higher in cells without p53 as demonstrated by western blot analysis (Fig. 4E and F). It should also be noted that devimistat reduced basal p53 levels and blocked IT-triggered p53 accumulation in p53-proficient HCT116 cells (Fig. 4E; Supplementary Fig. S5D). To detail the contribution of Bim to the observed synergism, a siRNA-mediated knockdown was performed in HCT116 cells and confirmed by western blot analysis upon treatment with the drugs or combination for 48 hours (Fig. 4G). At the same time, cell death induction was analyzed by Annexin V/PI staining, showing a significantly reduced cell death level in the devimistat/IT combination regimen upon Bim knockdown (Fig. 4H). Finally, the impact of the combination regimens on the expression of cyclins and important CDK-cyclin inhibitors was analyzed. The experiments revealed a strong reduction of CCNB1 (cyclin B1) and CCNH (cyclin H) expression in all combinations as well as a suppression of CCND1 (cyclin D1) and CCND2 (cyclin D2) in the regimen with IT and devimistat (Supplementary Fig. S5E and S5F). The expression levels of the CDK-cyclin inhibitors p14, p16, and p21 were unaffected by the combination treatment (Supplementary Fig. S5G). Taken together, our data demonstrate that devimistat synergizes with IT and 5-FU in colorectal cancer cells mainly by increasing cell death induction, which is mediated by induction of the BH3-only protein Bim.

Finally, these results were translated into an in vivo setting with two different colorectal cancer xenograft mouse models focusing on the promising combination of devimistat and IT. To this end, Balb/c^nu/nu animals were inoculated subcutaneously with either HCT116 or HT29 cells, were randomized into four groups (control, devimistat, IT and devimistat/IT) and treated as depicted in Fig. 5A. In HCT116 xenografts, single treatment with devimistat or IT suppressed tumor growth (Fig. 5B), resulting in a significant increase in survival rates (Fig. 5C). Intriguingly, the strongest inhibition of tumor growth was observed in animals receiving the combination devimistat/IT (Fig. 5B). In agreement with this finding, this regimen displayed the best overall survival rates (Fig. 5C). In HT29 xenografts, both devimistat and IT retarded tumor growth on their own (Fig. 5D). The best tumor growth suppression was achieved with the combination regimen (Fig. 5D), although the growth retardation was not as strong as in HCT116 cells. Remarkably, only the devimistat/IT combination resulted in significantly increased overall survival (Fig. 5E). It should further be mentioned that devimistat single treatment exhibited no general toxicity as reflected by normal weight gain comparable to the control group (Supplementary Fig. S6A and B). IT displayed moderate toxicity starting after 14 days of therapy, which was, however, not further increased by the addition of devimistat in the combination regimen (Supplementary Fig. S6A and B). Finally, the DNA damage markers p53 and γH2AX as well as the BH3-only protein Bim were assessed in tumor homogenates using western blot analysis. IT treatment caused p53 induction and γH2AX formation in both colorectal cancer xenograft models, which was suppressed in the combination regimen particularly in HCT116 xenografts (Fig. 5F; Supplementary Fig. S6C and S6D, and S6F). Furthermore, both devimistat and IT treatment increased Bim levels, which was potentiated in the combination regimen especially in HCT116 xenografts (Fig. 5G; Supplementary Fig. S6C and S6E). In conclusion, these findings provide evidence that devimistat also synergizes with IT in colorectal cancer in vivo, resulting in better therapy response and improved overall survival. Notably, the combination therapy was not associated with a higher toxicity than the single treatments, but caused even reduced DNA damage levels and triggered Bim induction in tumor xenograft tissue. The key pathways involved in the synergism between devimistat and the anticancer drugs are summarized in a model shown in Fig. 5H.

This study demonstrates that the mitochondrial targeted drug devimistat is cytotoxic with similar efficacy in a broad panel of established colorectal cancer cell lines and patient-derived short-term cultures independent of their genetic and epigenetic alterations. This is a clear advantage over the mother compound LA, which is also active in various colorectal cancer cell lines, but displays IC₅₀ values differing up to 10-fold (8, 27). 5-FU is the gold standard in colorectal cancer chemotherapy, whose activity is linked to the mismatch-repair (MMR) status. Patients with MMR-deficient tumors and high MSI did not show a benefit from adjuvant 5-FU chemotherapy (36), which was confirmed in MMR-deficient colorectal cancer cells exposed to 5-FU (37). Genetic features also determine the response and adverse effects induced by SN-38, the active metabolite of IT (38). Colorectal cancer patients with a deficiency in uridine diphosphate glucuronosyltransferase 1A1 exhibit insufficient hepatic glucuronidation and subsequent elimination of SN-38, thereby causing severe side effects after high IT doses (39). In contrast, neither MSI nor chromosomal instability affected the susceptibility of colorectal cancer cells toward devimistat. Furthermore, the devimistat-triggered cytotoxicity is independent of p53 as revealed by our studies in isogenic HCT116 cell lines. This is an important finding, because p53 mutations occur frequently in human cancer cells and affect their sensitivity toward DNA-damaging chemotherapeutics, as illustrated by an increased resistance of p53-deficient colorectal cancer cells to 5-FU (30, 40).

Our cytotoxicity data further show that nonmalignant HCEC are less susceptible toward the cytotoxic effects of devimistat, whereas the genotoxic anticancer drugs 5-FU and IT exerted a similar degree of cytotoxicity in colorectal cancer cells and HCEC. In support of these findings, devimistat killed intestinal tumoroids in a concentration-dependent manner, but hardly affected the viability of normal intestinal organoids. The reduced cytotoxicity in noncancerous cells was also observed in normal lung, breast, and kidney cells, which were all more resistant toward devimistat than the respective tumor cell lines (9). This tumor-selective mode of action is very likely attributable to differences in the metabolism between normal and cancer cells as highlighted recently (10). Nevertheless, further studies are required to elucidate the precise molecular mechanisms. Another asset of devimistat is the fact that devimistat is able to eliminate colorectal cancer cells independent of ongoing cell proliferation as attested by our experiments with 5-FU or IT arrested colorectal cancer cells. Although devimistat was still active in this setting, treatment with 5-FU or IT displayed no cytotoxicity at all. This fits to the notion that the antimetabolite 5-FU and topoisomerase I poisons like IT require cancer cell proliferation to exert their cytotoxicity (38, 41).

Furthermore, we were able to show that devimistat targets mitochondria and strongly impairs OXPHOS in colorectal cancer cell lines, which is preceded by ROS formation and decreased MMP. This is in agreement with other studies performed in H460 lung cancer cells, in which devimistat caused a loss of MMP (9) and a mitochondrial ROS burst with oxidation of the mitochondrial peroxiredoxin 3 isoform (11). The underlying mechanism by which the non-redox active devimistat promotes ROS formation is not fully understood, but was attributed to the E3 dihydrolipoamide dehydrogenase subunit of KGDH as primary source (11). A concentration-dependent inhibition of mitochondrial respiration was also detected in leukemia cell lines (42), which were even more susceptible than the colorectal cancer cell lines tested herein. Our genotoxicity assays conducted with devimistat in two colorectal cancer cell lines displayed lack of genotoxicity, except for a moderate, but statistically insignificant increase in oxidative DNA damage at high concentrations (i.e., 200 μmol/L devimistat). This might be a consequence of the early ROS burst triggered by devimistat and the compromised mitochondria, resulting in nuclear DNA damage. However, this genotoxic effect is rather small compared with the potent DNA-damaging properties of 5-FU or IT, as attested in the γH2AX ICW. A similar lack of genotoxicity was previously reported for the mother compound LA (27).

Based upon their different mode of actions, we hypothesized that the combination of the metabolic disruptor devimistat with the genotoxic agents 5-FU and IT will result in synergistic colorectal cancer cell death. The synergism of the combination regimens was demonstrated in colorectal cancer cells by both Chou-Talalay analysis and Combenefit analysis. Increased apoptotic cell death was identified as a major mechanism for the observed synergism. In the combination regimen with 5-FU or IT, devimistat caused a downregulation of the antiapoptotic genes Bcl-X_L and Survivin. Targeting Survivin by small-molecule inhibitors has recently been shown to sensitize p53-mutated colorectal cancer cells to IT treatment (43), highlighting Survivin as critical target in colorectal cancer. High expression levels of Bcl-X_L are known to contribute to the resistance of colorectal cancer cells towards 5-FU treatment (44). Furthermore, our study revealed a pronounced upregulation of the proapoptotic protein Bim in combination regimens of 5-FU/IT with devimistat. This was not only observed in colorectal cancer cells but also found in tumor xenograft tissue. siRNA-mediated downregulation of Bim significantly reduced the cell death induction by devimistat and IT, thereby highlighting its important function for the observed synergism. Bim is a BH3-only protein that antagonizes antiapoptotic proteins of the Bcl2 family, such as Bcl-2 and Mcl-1, and thereby promotes mitochondrial apoptosis (45). Because no changes in gene expression were detected in our experiments, the observed upregulation of Bim is likely attributable to its posttranslational regulation. Phosphorylation of Bim by JNK and p38 MAPK is known to prevent Bim degradation and promote its proapoptotic activity (45). Interestingly, JNK-mediated Bim activation was shown in liver cancer cells after exposure to the antineoplastic agent doxorubicin, which was potentiated by combined treatment with the death ligand TRAIL (46). Also in glioblastoma cells, JNK/c-Jun—mediated induction of Bim enhanced apoptosis triggered by the alkylating anticancer drugs temozolomide and nimustine (47).

Moreover, we found a pronounced downregulation of different cyclins, especially cyclin B1, cyclin H1, and cyclin D1/D2, in colorectal cancer cells challenged with the combination regimen. While Cyclin B1 is involved in G₂–M transition, Cyclin D1 and D2 regulate the progression from G₁ to S-phase (48). Cyclin H is part of the CDK-activating kinase, which activates CDKs, including CDK4 and CDK6, to promote cell-cycle progression and is involved in transcription as subunit of TFIIH (49). The reduced gene expression levels of certain cyclins in the combination group therefore suggest an impaired cell-cycle progression. Interestingly, a previous study conducted in pancreatic cancer cells reported a devimistat-triggered downregulation of several cell-cycle regulating genes, including cyclin B1, cyclin D3, and cyclin E1 (50).

Finally, the antitumor activity of devimistat was demonstrated in two different human colorectal cancer xenograft mouse models (HCT116—MSI; HT29—MSS). This extends the findings of a previous study, which reported growth inhibition of human lung cancer xenografts by devimistat (9). In a human pancreatic cancer xenograft model, devimistat was found to be even more potent than the antimetabolite gemcitabine, which is the mainstay in pancreatic cancer chemotherapy (50). In the present work, the therapeutic effects of devimistat were comparable to that of IT and clearly prolonged survival. With regard to its antitumor activity, devimistat is also superior to that of the parental compound LA, which weakly retarded growth of HCT116 tumor xenografts (34). Importantly, we showed that devimistat synergizes with IT in both colorectal cancer xenografts, resulting in prolonged survival and enhanced therapeutic efficacy. Notably, this is the first preclinical study revealing a synergism between the mitochondrial disruptor devimistat and a genotoxic anticancer drug in vivo. In view of these positive results, clinical studies evaluating the benefit of devimistat in combination with anticancer drugs commonly used in colorectal cancer chemotherapy, such as 5-FU, IT, or oxaliplatin, are eagerly awaited. There is already an ongoing phase I study dealing with devimistat in combination with 5-FU for the treatment of patients with metastatic colorectal cancer (NCT02232152), which will be completed in the next year.

In conclusion, our study showed that devimistat is a promising building block in colorectal cancer chemotherapy and synergizes with antineoplastic agents in vitro and in vivo. By targeting mitochondrial metabolism in colorectal cancer cells independent of their genetic and epigenetic aberrations, devimistat complements the genotoxic mode of action of 5-FU and IT. This represents a novel and efficient combination strategy to fight colorectal cancer, which is still a fatal disease for many patients.

M. Huber reports grants from DFG during the conduct of the study. J. Fahrer reports grants from Wilhelm Sander Foundation and German Research Foundation during the conduct of the study. No disclosures were reported by the other authors.

C. Arnold: Conceptualization, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. P. Demuth: Formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. N. Seiwert: Methodology, writing–review and editing. S. Wittmann: Formal analysis, investigation, methodology, writing–review and editing. K. Boengler: Methodology, writing–review and editing. B. Rasenberger: Formal analysis and investigation. M. Christmann: Formal analysis, validation, methodology, writing–review and editing. M. Huber: Formal analysis, validation, methodology, writing–review and editing. T. Brunner: Resources, methodology, writing–review and editing. M. Linnebacher: Methodology, writing–review and editing. J. Fahrer: Conceptualization, formal analysis, supervision, funding acquisition, validation, visualization, writing–original draft, writing–review and editing.

J. Fahrer was supported by the Wilhelm Sander Foundation (grant numbers 2016.039.1 and 2016.039.2) and the German Research Foundation (DFG; INST 248/331-1 FUGG). M. Christmann received funding from the DFG (SFB 1361, project-ID 393547839, subproject 05). M. Huber was supported by the DFG (HU1824/5-2, HU1824/7-1, HU1824/9-1). We thank Svenja Stroh and Georg Nagel (both Department of Toxicology, University Medical Center Mainz, Germany) as well as Beate von Derschau (Rudolf Buchheim Institute of Pharmacology, Justus Liebig University Giessen, Germany) for excellent technical assistance and Dennis Vogel (Institute for Medical Microbiology and Hospital Hygiene, University of Marburg, Germany) for performing Seahorse measurements. We are grateful to Prof. Bert Vogelstein (John Hopkins University, Baltimore) for providing HCT116 cells and to Prof. Jerry W. Shay (University of Texas Southwestern Medical Center, Dallas) for providing HCEC. Fpg enzyme was a kind gift of Prof. Bernd Epe (University of Mainz, Germany).

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.