Clinical relevance and therapeutic predictive ability of hypoxia biomarkers in head and neck cancer tumour models

Tumour hypoxia promotes poor patient outcomes, with particularly strong evidence for head and neck squamous cell carcinoma (HNSCC). To effectively target hypoxia, therapies require selection biomarkers and preclinical models that can accurately model tumour hypoxia. We established 20 patient-derived xenograft (PDX) and cell line-derived xenograft (CDX) models of HNSCC that we characterised for their fidelity to represent clinical HNSCC in gene expression, hypoxia status and proliferation and that were evaluated for their sensitivity to hypoxia-activated prodrugs (HAPs). PDX models showed greater fidelity in gene expression to clinical HNSCC than cell lines, as did CDX models relative to their paired cell lines. PDX models were significantly more hypoxic than CDX models, as assessed by hypoxia gene signatures and pimonidazole immunohistochemistry, and showed similar hypoxia gene expression to clinical HNSCC tumours. Hypoxia or proliferation status alone could not determine HAP sensitivity across our 20 HNSCC and two non-HNSCC tumour models by either tumour growth inhibition or killing of hypoxia cells in an ex vivo clonogenic assay. In summary, our tumour models provide clinically relevant HNSCC models that are suitable for evaluating hypoxia-targeting therapies; however, additional biomarkers to hypoxia are required to accurately predict drug sensitivity.

Clonal dynamics limits detection of selection in tumour xenograft CRISPR/Cas9 screens

Transplantable in vivo CRISPR/Cas9 knockout screens, in which cells are edited in vitro and inoculated into mice to form tumours, allow evaluation of gene function in a cancer model that incorporates the multicellular interactions of the tumour microenvironment. To improve our understanding of the key parameters for success with this method, we investigated the choice of cell line, mouse host, tumour harvesting timepoint and guide RNA (gRNA) library size. We found that high gRNA (80-95%) representation was maintained in a HCT116 subline transduced with the GeCKOv2 whole-genome gRNA library and transplanted into NSG mice when tumours were harvested at early (14 d) but not late time points (38-43 d). The decreased representation in older tumours was accompanied by large increases in variance in gRNA read counts, with notable expansion of a small number of random clones in each sample. The variable clonal dynamics resulted in a high level of 'noise' that limited the detection of gRNA-based selection. Using simulated datasets derived from our experimental data, we show that considerable reductions in count variance would be achieved with smaller library sizes. Based on our findings, we suggest a pathway to rationally design adequately powered in vivo CRISPR screens for successful evaluation of gene function.

Microglial transcriptome analysis in the rNLS8 mouse model of TDP-43 proteinopathy reveals discrete expression profiles associated with neurodegenerative progression and recovery

The microglial reaction is a hallmark of neurodegenerative conditions, and elements thereof may exert differential effects on disease progression, either worsening or ameliorating severity. In amyotrophic lateral sclerosis (ALS), a syndrome characterized by cytoplasmic aggregation of TDP-43 protein and atrophy of motor neurons in the cortex and spinal cord, the transcriptomic signatures of microglia during disease progression are incompletely understood. Here, we performed longitudinal RNAseq analysis of cortical and spinal cord microglia from rNLS8 mice, in which doxycycline-regulatable expression of human TDP-43 (hTDP-43) in the cytoplasm of neurons recapitulates many features of ALS. Transgene suppression in rNLS8 mice leads to functional, anatomical and electrophysiological resolution that is dependent on a microglial reaction that is concurrent with recovery rather than disease onset. We identified basal differences between the gene expression profiles of microglia dependent on localization in spinal cord or cortex. Microglia subjected to chronic hTDP-43 overexpression demonstrated transcriptomic changes in both locations. We noted strong upregulation of Apoe, Axl, Cd63, Clec7a, Csf1, Cst7, Igf1, Itgax, Lgals3, Lilrb4, Lpl and Spp1 during late disease and recovery. Importantly, we identified a distinct suite of differentially expressed genes associated with each phase of disease progression and recovery. Differentially expressed genes were associated with chemotaxis, phagocytosis, inflammation, and production of neuroprotective factors. These data provide new insights into the microglial reaction in TDP-43 proteinopathy. Genes differentially expressed during progression and recovery may provide insight into a unique instance in which the microglial reaction promotes functional recovery after neuronal insult.

Abstract 5853: Identification of TSC1 and TSC2 as potential determinants of sensitivity to trastuzumab emtansine

Trastuzumab emtansine (T-DM1, Kadcyla) is an antibody-drug conjugate used in the treatment of HER2-positive breast cancer. However, its use is limited by acquired and intrinsic resistance, the mechanisms of which are not well understood. Further knowledge of T-DM1 resistance may provide new combination strategies or therapeutic targets to overcome resistance or new predictive biomarkers to identify the patients most likely to benefit from T-DM1 therapy. To discover genes responsible for T-DM1 sensitivity and resistance in an unbiased manner, we have conducted CRISPR/Cas9 functional genomics screens by a two-stage process. Firstly, we performed whole genome screens in MDA-MB-361 and MDA-MB-453 cells transduced with Cas9 and the GeCKOv2 lentiviral library that were exposed to T-DM1 and its effector DM1 for 8-13 weeks. Gene knockouts enriched or depleted in response to T-DM1 or DM1 treatment in either cell line were identified by sequencing of genomic DNA and differentially expressed genes by RNA sequencing, revealing 599 candidate genes of T-DM1 sensitivity. For high-throughput validation of the 599 genes, we developed a custom library of 2539 guide RNAs (gRNAs) to target these 599 genes, plus non-targeting controls. Cas9-expressing MDA-MB-361 cells were transduced with the custom library and exposed to T-DM1 for 28 days. MAGeCK analysis of gRNA sequencing revealed 11 genes that were significantly enriched and one gene that was significantly depleted at a false discovery rate (FDR) of <0.1. Two of the top hits in the secondary screen were two genes whose loss is known to promote T-DM1 resistance: ERBB2 (HER2) and SLC46A3 (P<8 × 10−5; FDR <0.007). Other top hits were TSC1 and TSC2 (P<3 × 10−6; FDR= 0.0004); which are both tumor suppressor genes and negative regulators of mTOR complex 1 (mTORC1). For subsequent validation, we have generated TSC2 knockout cell pools, which were more resistant to T-DM1 than wildtype cells in a competition growth assay. Knockout clones have been isolated and are being tested for T-DM1 resistance. Since mTOR inhibitors can phenocopy TSC1 and TSC2 by inhibiting mTORC1 activity, we have also evaluated T-DM1 in combination with the mTOR inhibitor KU-0063794 in a sulforhodamine B assay in MDA-MB-361 and MDA-MB-453 cells. Each agent potently inhibited cell proliferation and demonstrated synergistic anti-proliferative activity in combination. Together, our results suggest that TSC1 and TSC2 knockout may promote T-DM1 resistance and that targeting mTOR may be an effective strategy to overcome T-DM1 resistance.

Citation Format: Francis W. Hunter, Barbara A. Lipert, Kyla N. Siemens, Aziza Khan, Hilary R. Barker, Troy W. Ketela, William R. Wilson, Tet-Woo Lee, Stephen M. Jamieson. Identification of TSC1 and TSC2 as potential determinants of sensitivity to trastuzumab emtansine [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 5853.

Mechanisms of resistance to trastuzumab emtansine (T-DM1) in HER2-positive breast cancer

The HER2-targeted antibody-drug conjugate trastuzumab emtansine (T-DM1) is approved for the treatment of metastatic, HER2-positive breast cancer after prior trastuzumab and taxane therapy, and has also demonstrated efficacy in the adjuvant setting in incomplete responders to neoadjuvant therapy. Despite its objective activity, intrinsic and acquired resistance to T-DM1 remains a major clinical challenge. T-DM1 mediates its activity in a number of ways, encompassing HER2 signalling blockade, Fc-mediated immune response and payload-mediated microtubule poisoning. Resistance mechanisms relating to each of these features have been demonstrated, and we outline the findings of these studies in this review. In our overview of the substantial literature on T-DM1 activity and resistance, we conclude that the T-DM1 resistance mechanisms most strongly supported by the experimental data relate to dysfunctional intracellular metabolism of the construct and subversion of DM1-mediated cell killing. Loss of dependence on signalling initiated by HER2-HER2 homodimers is not substantiated as a resistance mechanism by clinical or experimental studies, and the impact of EGFR expression and tumour immunological status requires further investigation. These findings are instructive with respect to strategies that might overcome T-DM1 resistance, including the use of second-generation anti-HER2 antibody-drug conjugates that deploy alternative linker-payload chemistries.

Hypoxia-selective radiosensitisation by SN38023, a bioreductive prodrug of DNA-dependent protein kinase inhibitor IC87361

DNA-dependent protein kinase (DNA-PK) plays a key role in repair of radiation-induced DNA double strand breaks (DSB) by non-homologous end-joining. DNA-PK inhibitors (DNA-PKi) are therefore efficient radiosensitisers, but normal tissue radiosensitisation represents a risk for their use in radiation oncology. Here we describe a novel prodrug, SN38023, that is metabolised to a potent DNA-PKi (IC87361) selectively in radioresistant hypoxic cells. DNA-PK inhibitory potency of SN38023 was 24-fold lower than IC87361 in cell-free assays, consistent with molecular modelling studies suggesting that SN38023 is unable to occupy one of the predicted DNA-PK binding modes of IC87361. One-electron reduction of the prodrug by radiolysis of anoxic formate solutions, and by metabolic reduction in anoxic HCT116/POR cells that overexpress cytochrome P450 oxidoreductase (POR), generated IC87361 efficiently as assessed by LC-MS. SN38023 inhibited radiation-induced Ser2056 autophosphorylation of DNA-PK catalytic subunit and radiosensitised HCT116/POR and UT-SCC-54C cells selectively under anoxia. SN38023 was an effective radiosensitiser in anoxic HCT116 spheroids, demonstrating potential for penetration into hypoxic tumour tissue, but in spheroid co-cultures of high-POR and POR-null cells it showed no evidence of bystander effects resulting from local diffusion of IC87361. Pharmacokinetics of IC87361 and SN38023 at maximum achievable doses in NIH-III mice demonstrated sub-optimal exposure of UT-SCC-54C tumour xenografts and did not provide significant tumour radiosensitisation. In conclusion, SN38023 has potential for exploiting hypoxia for selective delivery of a potent DNA-PKi to the most radioresistant subpopulation of cells in tumours. However, prodrugs providing improved systemic pharmacokinetics and that release DNA-PKi that elicit bystander effects are needed to maximise therapeutic utility.

Radiosensitization of head and neck squamous cell carcinoma lines by DNA-PK inhibitors is more effective than PARP-1 inhibition and is enhanced by SLFN11 and hypoxia

Background and purpose: Poly(ADP-ribose)polymerase-1 (PARP1) and DNA-dependent protein kinase (DNA-PK) play key roles in the repair of radiation-induced DNA double strand breaks, but it is unclear which is the preferred therapeutic target in radiotherapy. Here we compare small molecule inhibitors of both as radiosensitizers of head and neck squamous cell carcinoma (HNSCC) cell lines.Methods: Two PARP1 inhibitors (olaparib, veliparib) and two DNA-PK inhibitors (KU57788, IC87361) were tested in 14 HNSCC cell lines and two non-tumorigenic lines (HEK-293 and WI-38/Va-13), with drug exposure for 6 or 24 h post-irradiation, using regrowth assays. For three lines (UT-SCC-54C, -74B, -76B), radiosensitization was also assessed by clonogenic assay under oxia and acute (6 h) anoxia, and for 54C cells under chronic hypoxia (0.2% O₂ for 48 h). Relationships between sensitizer enhancement ratios (SER) and gene expression, assessed by RNA sequencing, were evaluated.Results: The inhibitors were minimally cytotoxic in the absence of radiation, with 74B and 54C cells the most sensitive to both olaparib and KU57788. Median SER values for each inhibitor at 1.1 µM were 1.12 (range 1.02-1.24) for olaparib, 1.08 (1.04-1.13) for veliparib, 1.35 (1.10-1.64) for IC87361 and 1.77 (1.41-2.38) for KU57788. The higher SER values for the DNA-PK inhibitors were observed with all cell lines (except HEK-293) and all concentrations tested and were confirmed by clonogenic assay. Radiosensitization by the DNA-PK inhibitors correlated with expression of SLFN11 mRNA. Radiosensitization by IC87361 and olaparib was significantly enhanced under acute anoxia and chronic hypoxia.Conclusions: The DNA-PK inhibitors KU57788 and IC87361 are more effective radiosensitizers than the PARP-1 inhibitors olaparib and veliparib at non-cytotoxic concentrations in HNSCC cell cultures and their activity is enhanced by SLFN11 and hypoxia.

Impact of Tumour Hypoxia on Evofosfamide Sensitivity in Head and Neck Squamous Cell Carcinoma Patient-Derived Xenograft Models

Tumour hypoxia is a marker of poor prognosis and failure of chemoradiotherapy in head and neck squamous cell carcinoma (HNSCC), providing a strategy for therapeutic intervention in this setting. To evaluate the utility of the hypoxia-activated prodrug evofosfamide (TH-302) in HNSCC, we established ten early passage patient-derived xenograft (PDX) models of HNSCC that were characterised by their histopathology, hypoxia status, gene expression, and sensitivity to evofosfamide. All PDX models closely resembled the histology of the patient tumours they were derived from. Pimonidazole-positive tumour hypoxic fractions ranged from 1.7-7.9% in line with reported HNSCC clinical values, while mRNA expression of the Toustrup hypoxia gene signature showed close correlations between PDX and matched patient tumours, together suggesting the PDX models may accurately model clinical tumour hypoxia. Evofosfamide as a single agent (50 mg/kg IP, qd × 5 for three weeks) demonstrated antitumour efficacy that was variable across the PDX models, ranging from complete regressions in one p16-positive PDX model to lack of significant activity in the three most resistant models. Despite all PDX models showing evidence of tumour hypoxia, and hypoxia being essential for activation of evofosfamide, the antitumour activity of evofosfamide only weakly correlated with tumour hypoxia status determined by pimonidazole immunohistochemistry. Other candidate evofosfamide sensitivity genes-MKI67, POR, and SLFN11-did not strongly influence evofosfamide sensitivity in univariate analyses, although a weak significant relationship with MKI67 was observed, while SLFN11 expression was lost in PDX tumours. Overall, these data confirm that evofosfamide has antitumour activity in clinically-relevant PDX tumour models of HNSCC and support further clinical evaluation of this drug in HNSCC patients. Further research is required to identify those factors that, alongside hypoxia, can influence sensitivity to evofosfamide and could act as predictive biomarkers to support its use in precision medicine therapy of HNSCC.

Functional CRISPR and shRNA Screens Identify Involvement of Mitochondrial Electron Transport in the Activation of Evofosfamide

Evofosfamide (TH-302) is a hypoxia-activated DNA-crosslinking prodrug currently in clinical development for cancer therapy. Oxygen-sensitive activation of evofosfamide depends on one-electron reduction, yet the reductases that catalyze this process in tumors are unknown. We used RNA sequencing, whole-genome CRISPR knockout, and reductase-focused short hairpin RNA screens to interrogate modifiers of evofosfamide activation in cancer cell lines. Involvement of mitochondrial electron transport in the activation of evofosfamide and the related nitroaromatic compounds EF5 and FSL-61 was investigated using 143B ρ ⁰ (ρ zero) cells devoid of mitochondrial DNA and biochemical assays in UT-SCC-74B cells. The potency of evofosfamide in 30 genetically diverse cancer cell lines correlated with the expression of genes involved in mitochondrial electron transfer. A whole-genome CRISPR screen in KBM-7 cells identified the DNA damage-response factors SLX4IP, C10orf90 (FATS), and SLFN11, in addition to the key regulator of mitochondrial function, YME1L1, and several complex I constituents as modifiers of evofosfamide sensitivity. A reductase-focused shRNA screen in UT-SCC-74B cells similarly identified mitochondrial respiratory chain factors. Surprisingly, 143B ρ ⁰ cells showed enhanced evofosfamide activation and sensitivity but had global transcriptional changes, including increased expression of nonmitochondrial flavoreductases. In UT-SCC-74B cells, evofosfamide oxidized cytochromes a, b, and c and inhibited respiration at complexes I, II, and IV without quenching reactive oxygen species production. Our results suggest that the mitochondrial electron transport chain contributes to evofosfamide activation and that predicting evofosfamide sensitivity in patients by measuring the expression of canonical bioreductive enzymes such as cytochrome P450 oxidoreductase is likely to be futile.

Evofosfamide for the treatment of human papillomavirus-negative head and neck squamous cell carcinoma

Evofosfamide (TH-302) is a clinical-stage hypoxia-activated prodrug of a DNA-crosslinking nitrogen mustard that has potential utility for human papillomavirus (HPV) negative head and neck squamous cell carcinoma (HNSCC), in which tumor hypoxia limits treatment outcome. We report the preclinical efficacy, target engagement, preliminary predictive biomarkers and initial clinical activity of evofosfamide for HPV-negative HNSCC. Evofosfamide was assessed in 22 genomically characterized cell lines and 7 cell line-derived xenograft (CDX), patient-derived xenograft (PDX), orthotopic, and syngeneic tumor models. Biomarker analysis used RNA sequencing, whole-exome sequencing, and whole-genome CRISPR knockout screens. Five advanced/metastatic HNSCC patients received evofosfamide monotherapy (480 mg/m2 qw × 3 each month) in a phase 2 study. Evofosfamide was potent and highly selective for hypoxic HNSCC cells. Proliferative rate was a predominant evofosfamide sensitivity determinant and a proliferation metagene correlated with activity in CDX models. Evofosfamide showed efficacy as monotherapy and with radiotherapy in PDX models, augmented CTLA-4 blockade in syngeneic tumors, and reduced hypoxia in nodes disseminated from an orthotopic model. Of 5 advanced HNSCC patients treated with evofosfamide, 2 showed partial responses while 3 had stable disease. In conclusion, evofosfamide shows promising efficacy in aggressive HPV-negative HNSCC, with predictive biomarkers in development to support further clinical evaluation in this indication.

Cellular pharmacology of evofosfamide (TH-302): A critical re-evaluation of its bystander effects

Evofosfamide (TH-302) is a clinical-stage hypoxia-activated prodrug with proven efficacy against hypoxic cells in preclinical tumour models. TH-302 is designed to release the DNA crosslinking agent bromo-isophosphoramide mustard (Br-IPM) when reduced in hypoxic tissue. Br-IPM is considered to diffuse locally from hypoxic regions, eliciting additional tumour cell killing, but the latter 'bystander effect' has not been demonstrated directly. Previous studies with multicellular co-cultures that included cells expressing the E. coli nitroreductase NfsA as a model TH-302 reductase have provided clear evidence of a bystander effect (which we confirm in the present study). However, NfsA is an oxygen-insensitive two-electron reductase that is not expected to generate the nitro radical intermediate that has been demonstrated to fragment to release Br-IPM. Here, we use mass spectrometry methods to characterise TH-302 metabolites generated by one-electron reduction (steady-state radiolysis by ionising radiation and cellular metabolism under hypoxia, including HCT116 cells that overexpress P450 oxidoreductase, POR) or by NfsA expressed in HCT116 cells under oxic conditions, and investigate the stability and cytotoxicity of these products. Br-IPM is shown to have very low cytotoxic potency when added to extracellular culture medium and to be rapidly converted to other hydrophilic products including dichloro-isophosphoramide mustard (IPM). Only traces of Br-IPM or IPM were detected in the extracellular medium when generated by cellular metabolism of TH-302. We identify, in NfsA-expressing cells, the hydroxylamine metabolite of TH-302, and downstream products resulting from rearrangement or hydration of the imidazole ring, and demonstrate that formation of these candidate bystander effect mediators is suppressed by hypoxia. This characterisation of the cellular pharmacology of TH-302 implies that bystander effects from hypoxic activation of TH-302 are unlikely to contribute to its anticancer activity.

Antagonism in effectiveness of evofosfamide and doxorubicin through intermolecular electron transfer

Hypoxic cells pose a problem in anticancer chemotherapy, in which often drugs require oxygen as an electron acceptor to bring about the death of actively cycling cells. Bioreductive anticancer drugs, which are selectively activated in the hypoxic regions of tumours through enzymatic one-electron reduction, are being developed for combination with chemotherapy-, radiotherapy- and immunotherapy-containing regimens to kill treatment-resistant hypoxic cells. The most clinically-advanced bioreductive drug, evofosfamide (TH-302), which acts by releasing a DNA-crosslinking mustard, failed to extend overall survival in combination with doxorubicin, a topoisomerase II inhibitor, for advanced soft tissue sarcoma in a pivotal clinical trial. However, the reasons for the lack of additive efficacy with this combination are unknown. Here, we show that the radical anion of evofosfamide undergoes electron transfer to doxorubicin in kinetic competition to fragmentation of the radical anion, thus suppressing the release the cytotoxic mustard. This electron transfer process may account, at least in part, for the lack of overall survival improvement in the recent clinical trial. This study underlines the need to consider both redox and electron transfer chemistry when combining bioreductive prodrugs with other redox-active drugs in cancer treatment.

Abstract 169: Preclinical efficacy and sensitivity determinants of evofosfamide in molecularly defined models of head and neck squamous cell carcinoma

Tumor hypoxia is prevalent in head and neck squamous cell carcinoma (HNSCC), where it limits radiotherapy outcomes. Hypoxia-activated prodrugs (HAPs) have been developed to target hypoxic regions of tumors. These agents undergo oxygen-sensitive reductive activation, thereby delivering cytotoxic species within hypoxic cells. This study investigated the efficacy and sensitivity determinants of the clinical-stage HAP evofosfamide (TH-302) using molecularly-characterized models of HNSCC. We deployed a collection of 27 HPV-negative HNSCC cell lines derived from lesions of varying TNM stages and primary, nodal or recurrent sites. The collection was characterized for gene expression by RNA-seq, from which somatic variants were also called. Their transcriptomic features were investigated in the context of pan-cancer TCGA data by hierarchical clustering. The potency and hypoxic selectivity of 3 HAPs - evofosfamide, PR-104A and SN30000 - were assessed by antiproliferative assay in 22 lines and compared to bromo-isophosphoramide mustard (Br-IPM), cisplatin and 5-FU. The antitumor activity of evofosfamide (50 mg/kg qdx5 for 2-3 cycles with or without a single 10 Gy dose of radiation on day 5 of cycle 1) was evaluated in HNSCC xenografts in addition to a PDX isolated from an SCC of the glottic larynx. The hypoxic fraction at baseline and after 5 days of treatment was quantified by pimonidazole staining. Genetic modifiers of sensitivity to evofosfamide and its cytotoxic metabolite Br-IPM were explored through whole-genome CRISPR-Cas9 screens using the GeCKO v2 library. High-throughput screens with a custom shRNA pool were performed in one HNSCC and two pancreatic ductal adenocarcinoma cell lines to identify reductases responsible for the activation of evofosfamide in hypoxic cells. Evofosfamide was more potent and more selective for hypoxic HNSCC cells in vitro than PR-104A or SN30000. Cell line sensitivity to evofosfamide was correlated with Br-IPM and cisplatin but not with PR-104A, SN30000 or 5-FU, indicating distinct sensitivity determinants. Evidence of antitumor activity with evofosfamide was observed in vivo. CRISPR screens identified potential evofosfamide sensitivity genes that were reproducibly enriched following drug exposure. Reductase-focused RNA interference screens defined a cluster of sensitivity genes that mapped to mitochondrial electron transport, whereas shRNA’s targeted against presumed activating enzymes such as POR were not enriched. Concentration-dependent oxidation of cytochrome a and decreased respiration was observed in cells exposed to evofosfamide, suggesting reduction by mitochondrial complexes. This study provides a rationale for the clinical evaluation of evofosfamide with radiotherapy in genetically defined subsets of HNSCC patients.

Citation Format: Francis W. Hunter, Avik Shome, Dan Li, Way W. Wong, Peter Tsai, Nooriyah Poonawala, Purvi M. Kakadiya, Troy M. Ketelä, Maria K. Kondratyev, Courtney R. Lynch, Tet-Woo Lee, Khanh B. Tran, Jules B. Devaux, Rachel Zussman, Cho R. Hong, Dennis Kee, Andrew M. Macann, Anthony J. Hickey, Stefan K. Bohlander, Cristin G. Print, William R. Wilson, Bradly G. Wouters, Stephen M. Jamieson. Preclinical efficacy and sensitivity determinants of evofosfamide in molecularly defined models of head and neck squamous cell carcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 169. doi:10.1158/1538-7445.AM2017-169

Hypoxia-activated prodrugs: paths forward in the era of personalised medicine

Tumour hypoxia has been pursued as a cancer drug target for over 30 years, most notably using bioreductive (hypoxia-activated) prodrugs that target antineoplastic agents to low-oxygen tumour compartments. Despite compelling evidence linking hypoxia with treatment resistance and adverse prognosis, a number of such prodrugs have recently failed to demonstrate efficacy in pivotal clinical trials; an outcome that demands reflection on the discovery and development of these compounds. In this review, we discuss a clear disconnect between the pathobiology of tumour hypoxia, the pharmacology of hypoxia-activated prodrugs and the manner in which they have been taken into clinical development. Hypoxia-activated prodrugs have been evaluated in the manner of broad-spectrum cytotoxic agents, yet a growing body of evidence suggests that their activity is likely to be dependent on the coincidence of tumour hypoxia, expression of specific prodrug-activating reductases and intrinsic sensitivity of malignant clones to the cytotoxic effector. Hypoxia itself is highly variable between and within individual tumours and is not treatment-limiting in all cancer subtypes. Defining predictive biomarkers for hypoxia-activated prodrugs and overcoming the technical challenges of assaying them in clinical settings will be essential to deploying these agents in the era of personalised cancer medicine.

Identification of P450 Oxidoreductase as a Major Determinant of Sensitivity to Hypoxia-Activated Prodrugs

Hypoxia is a prevalent feature of many tumors contributing to disease progression and treatment resistance, and therefore constitutes an attractive therapeutic target. Several hypoxia-activated prodrugs (HAP) have been developed, including the phase III candidate TH-302 (evofosfamide) and the preclinical agent SN30000, which is an optimized analogue of the well-studied HAP tirapazamine. Experience with this therapeutic class highlights an urgent need to identify biomarkers of HAP sensitivity, including enzymes responsible for prodrug activation during hypoxia. Using genome-scale shRNA screens and a high-representation library enriched for oxidoreductases, we identified the flavoprotein P450 (cytochrome) oxidoreductase (POR) as the predominant determinant of sensitivity to SN30000 in three different genetic backgrounds. No other genes consistently modified SN30000 sensitivity, even within a POR-negative background. Knockdown or genetic knockout of POR reduced SN30000 reductive metabolism and clonogenic cell death and similarly reduced sensitivity to TH-302 under hypoxia. A retrospective evaluation of head and neck squamous cell carcinomas showed heterogeneous POR expression and suggested a possible relationship between human papillomavirus status and HAP sensitivity. Taken together, our study identifies POR as a potential predictive biomarker of HAP sensitivity that should be explored during the clinical development of SN30000, TH-302, and other hypoxia-directed agents.

Dual targeting of hypoxia and homologous recombination repair dysfunction in triple-negative breast cancer

Triple-negative breast cancer (TNBC) is an aggressive malignancy with poor clinical outcome and few validated drug targets. Two prevalent features of TNBC, tumor hypoxia and derangement of homologous recombination (HR) repair, are potentially exploitable for therapy. This study investigated whether hypoxia-activated prodrugs (HAP) of DNA-damaging cytotoxins may inhibit growth of TNBC by simultaneously addressing these two targets. We measured in vitro activity of HAP of DNA breakers (tirapazamine, SN30000) and alkylators (TH-302, PR-104, SN30548) in TNBC cell lines and isogenic models, and related this to measures of HR repair and expression of prodrug-activating enzymes. Antitumor activity of HAP was examined in isogenic BRCA2-knockout xenograft models and compared with platinum chemotherapy. All five HAP selectively inhibited growth of TNBC cell lines under hypoxia. Sensitivity to HAP was not strongly associated with BRCA1 genotype. However, HAP sensitivity was enhanced by suppression of HR (assessed by radiation-induced RAD51 focus formation) when BRCA1 and PALB2 were knocked down in a common (MDA-MB-231) background. Furthermore, knockout of BRCA2 markedly sensitized DLD-1 cells to the clinical nitrogen mustard prodrugs TH-302 and PR-104 and significantly augmented sterilization of clonogens by these agents in xenografts, both as monotherapy and in combination with radiotherapy, but had less effect on activity of the benzotriazine di-N-oxide SN30000. PR-104 monotherapy was more effective than cisplatin at inhibiting growth of BRCA2-knockout tumors at equitoxic doses. This study demonstrates the potential for HAP of nitrogen mustards to simultaneously exploit hypoxia and HR defects in tumors, with translational implications for TNBC and other HR-deficient malignancies.

The flavoprotein FOXRED2 reductively activates nitro-chloromethylbenzindolines and other hypoxia-targeting prodrugs

The nitro-chloromethylbenzindoline prodrug SN29428 has been rationally designed to target tumour hypoxia. SN29428 is metabolised to a DNA minor groove alkylator via oxygen-sensitive reductive activation initiated by unknown one-electron reductases. The present study sought to identify reductases capable of activating SN29428 in tumours. Expression of candidate reductases in cell lines was modulated using forced expression and, for P450 (cytochrome) oxidoreductase (POR), by zinc finger nuclease-mediated gene knockout. Affymetrix microarray mRNA expression of flavoreductases was correlated with SN29428 activation in a panel of 23 cancer cell lines. Reductive activation and cytotoxicity of prodrugs were measured using mass spectrometry and antiproliferative assays, respectively. SN29428 activation under hypoxia was strongly attenuated by the pan-flavoprotein inhibitor diphenyliodonium, but less so by knockout of POR suggesting other flavoreductases contribute. Forced expression of 5-methyltetrahydrofolate-homocysteine methyltransferase reductase (MTRR), as well as POR, increased activation of SN29428 in hypoxic HCT 116 cells. SN29428 activation strongly correlated with expression of POR and also FAD-dependent oxidoreductase domain containing 2 (FOXRED2), in cancer cell lines. This association persisted after removing the effect of POR enzyme activity using first-order partial correlation. Forced expression of FOXRED2 increased SN29428 activation and cytotoxicity in hypoxic HEK293 cells and also increased activation of hypoxia-targeted prodrugs PR-104A, tirapazamine and SN30000, and increased cytotoxicity of the clinical-stage prodrug TH-302. Thus this study has identified three flavoreductases capable of enzymatically activating SN29428, one of which (FOXRED2) has not previously been implicated in xenobiotic metabolism. These results will inform future development of biomarkers predictive of SN29428 sensitivity.

Homologous recombination repair-dependent cytotoxicity of the benzotriazine di-N-oxide CEN-209: comparison with other hypoxia-activated prodrugs

CEN-209 (SN30000) is a second-generation benzotriazine di-N-oxide currently in advanced preclinical development as a hypoxia-activated prodrug (HAP). Herein we describe the DNA repair-, hypoxia- and one-electron reductase-dependence of CEN-209 cytotoxicity. We deployed mutant CHO cell lines to generate DNA repair profiles for CEN-209, and compared the profiles with those for other HAPs. Hypoxic selectivity of CEN-209 was significantly greater than PR-104A and the nitro-chloromethylbenzindoline (nCBI/SN29428) and comparable to tirapazamine and TH-302. CEN-209 was selective for homologous recombination (HR) repair-deficient cells (Rad51d⁻/⁻), but less so than nitrogen mustard prodrugs TH-302 and PR-104A. Further, DNA repair profiles for CEN-209 differed under oxic and hypoxic conditions, with oxic cytotoxicity more dependent on HR. This feature was conserved across all three members of the benzotriazine di-N-oxide class examined (tirapazamine, CEN-209 and CEN-309/SN29751). Enhancing one-electron reduction of CEN-209 by forced expression of a soluble form of NADPH:cytochrome P450 oxidoreductase (sPOR) increased CEN-209 cytotoxicity more markedly under oxic than hypoxic conditions. Comparison of oxygen consumption, H₂O₂ production and metabolism of CEN-209 to the corresponding 1-oxide and nor-oxide reduced metabolites suggested that enhanced oxic cytotoxicity in cells with high one-electron reductase activity is due to futile redox cycling. This study supports the hypothesis that both oxic and hypoxic cell killing by CEN-209 is mechanistically analogous to tirapazamine and is dependent on oxidative DNA damage repaired via multiple pathways. However, HAPs that generate DNA interstrand cross-links, such as TH-302 and PR-104, may be more suitable than benzotriazine di-N-oxides for exploiting reported HR repair defects in hypoxic tumour cells.