Abstract
PR-104, currently in clinical trial, is converted systemically to the dinitrobenzamide nitrogen mustard prodrug PR-104A, which is reduced selectively in hypoxic cells to cytotoxic hydroxylamine (PR-104H) and amine (PR-104M) metabolites. Here, we evaluate the roles of this reductive metabolism, and DNA interstrand cross-links (ICL), in the hypoxic and aerobic cytotoxicity of PR-104. Using a panel of 9 human tumor cell lines, cytotoxicity was determined by clonogenic assay after a 2-hour aerobic or hypoxic exposure to PR-104A. PR-104H and PR-104M were determined by high performance liquid chromatography/mass spectrometry, and ICL with the alkaline comet assay. Under hypoxia, the relationship between ICL and cell killing was similar between cell lines. Under aerobic conditions, there was a similar relationship between ICL and cytotoxicity, except in lines with very low rates of aerobic reduction of PR-104A (A2780, C33A, H1299), which showed an ICL-independent mechanism of PR-104A cytotoxicity. Despite this, in xenografts from the same lines, the frequency of PR-104–induced ICL correlated with clonogenic cell killing (r2 = 0.747) with greatest activity in the fast aerobic metabolizers. In addition, changing levels of hypoxia in SiHa tumors modified both ICL frequency and tumor growth delay in parallel. We conclude that both aerobic and hypoxic nitroreduction of PR-104A contribute to the monotherapy antitumor activity of PR-104 in human tumor xenografts, and that ICL are responsible for its antitumor activity and represent a broadly applicable biomarker for tumor cell killing by this novel prodrug. [Cancer Res 2009;69(9):3884–91]
Introduction
The abnormal microvasculature in tumors gives rise to spatially and temporally disorganized blood flow, resulting in regions of severe hypoxia (1, 2). The difference in oxygenation between tumors and normal tissues makes hypoxia a theoretically attractive therapeutic target, particularly when hypoxic cells are refractory to radiotherapy and many chemotherapy drugs, and also contribute to tumor progression through multiple mechanisms (3, 4).
Attempts to target tumor hypoxia currently focus on agents that exploit the molecular phenotype of hypoxic cells, especially stabilized HIF-1 and its transcriptional targets (5), and on hypoxia-activated prodrugs (HAP, also called hypoxic cytotoxins or bioreductive drugs). The latter are reduced enzymatically to DNA-reactive cytotoxins via pathways that are inhibited by molecular oxygen (3, 6–8). A complicating factor in targeting HIF-1 is that its activity is subject to complex oxygen-independent regulation (9–11). In contrast, many HAP can act as direct sensors of oxygen that oxidizes the initial free radical reduction product to regenerate the parent prodrug, thus confining toxicity to hypoxic cells (12). Although many HIF-1 inhibitors and HAP are in development, none have yet been registered for clinical use. The HAP that is most advanced in clinical trial, tirapazamine, recently failed its primary end point in a pivotal registration trial in combination with chemoradiation for advanced head and neck cancer (13), although improved locoregional control was shown in a subset of patients with more hypoxic tumors as assessed by positron emission tomography imaging (14). This observation is consistent with a large body of evidence that hypoxia limits outcome in radiotherapy of head and neck tumors (15–17), but the clinical experience with tirapazamine points to the need for HAP with superior therapeutic activity.
The 3,5-dinitrobenzamide mustard PR-104A is a new HAP that is selectively reduced to reactive cytotoxic nitrogen mustards (hydroxylamine PR-104H and amine PR-104M) under hypoxia (18); chemical structures are shown in Supplementary Fig. S1. PR-104A is delivered by administration of a water-soluble phosphate ester, PR-104, which shows greater antitumor activity than tirapazamine in preclinical models (18, 19) and has recently completed phase I trial (20). PR-104A penetrates into hypoxic regions of tumors more efficiently than tirapazamine because it is less rapidly reduced in moderately hypoxic tissue (19), which may also avoid toxicities of tirapazamine that arise through physiologic hypoxia in normal tissue (21, 22). In common with other dinitrobenzamide mustards (23), reduced metabolites of PR-104A are able to diffuse from cells in which they are formed (18) and provide a bystander effect that seems to contribute to preclinical antitumor activity of PR-104 as monotherapy and in combination with radiation (18, 19).
Hypoxic tumor cells are often noncycling or slowly cycling (24–26), making them resistant to many anticancer agents. A key objective in the design of PR-104 was therefore to generate a long-lived DNA lesion capable of killing hypoxic cells when they reoxygenate and resume cycling. We therefore sought to exploit the potency and longevity of nitrogen mustard DNA interstrand cross-links (ICL) that are difficult for mammalian cells to repair, requiring collaboration between proteins of the nucleotide excision repair, mismatch repair, homologous recombination repair, translesion synthesis, and Fanconi's anemia pathways (27, 28). The first stage in repair of nitrogen mustard ICL (“unhooking”) has a half-time of ∼1 to 2 days (29, 30) and full repair is presumed to be even slower (27). Reflecting this slow repair, nitrogen mustards show little loss of cytotoxic potency in noncycling cells (31, 32).
To generate such a lesion in hypoxic cells, we used the reduction of an electron-withdrawing nitro group (in PR-104A) to an electron-donating hydroxylamine or amine, which acts as an “electronic switch” that activates a prepositioned latent nitrogen mustard moiety (33–35). A preliminary study showed the formation of ICL from PR-104A, selectively in hypoxic cells, and the subsequent ser-139 phosphorylation of histone H2AX, which is presumably in part due to DNA replication fork arrest (18). However, whether ICLs are the sole (or major) mechanism of PR-104A cytotoxicity, in either aerobic or hypoxic cells, is unknown.
Our two key objectives here were to clarify the mechanism of action of PR-104 and assess whether ICL represent a reliable surrogate for its antitumor activity. To this end, we examine relationships between metabolism of PR-104A to its reduced metabolites (by mass spectrometry), frequency of DNA ICL (by alkaline comet assay), and cytotoxicity (by clonogenic assay) in a panel of nine human tumor cell lines under aerobic and hypoxic conditions. We also evaluate DNA cross-linking and clonogenic cell killing in the corresponding tumor xenografts, and test whether changing the level of hypoxia in SiHa tumors influences ICL and antitumor activity.
Materials and Methods
Compounds. PR-104 was supplied as its clinical formulation by Proacta, Inc. PR-104A was synthesized as described (36), stored at −20°C, and stock solutions (200 mmol/L in DMSO) were stored at −80°C. PR-104H, prepared by zinc dust reduction of PR-104A (18), was purified by preparative high performance liquid chromatography and stored as acetonitrile stock solutions (1.67 mmol/L) at −80°C. Tetradeuterated (d4) PR-104H was synthesized from d4-PR-104A (37). All PR-104 compounds had a purity of >95% by high performance liquid chromatography based on absorbance at 254 nm. Chlorambucil (Sigma) stock solutions (200 mmol/L in DMSO) were stored at −80°C.
Cell lines. Cell lines were passaged in α minimal essential medium supplemented with 5% fetal bovine serum without antibiotics. Cells were obtained from American Type Culture Collection except for gifts of H1299 and C33A (Onyx Pharmaceuticals), HT29 (Dr. D. Ross, University of Colorado, Boulder, CO), A2780 (European Collection of Cell Cultures), and SiHa (Dr. D. Cowan, Ontario Cancer Institute, Toronto, Canada). The SiHa cell line overexpressing NADPH:cytochrome P450 oxidoreductase, SiHaCYPOR, has been described (38). Cells were used within 3 mo of thawing frozen stocks confirmed to be Mycoplasma free by PCR-ELISA (Roche Diagnostics).
Clonogenic assays. Cell killing was assessed by clonogenic assay after 2 h of drug exposure under aerobic or hypoxic conditions as described (38), except that 2 × 106 cells were exposed in 3 mL medium/P60 dish. Surviving fraction was calculated as the ratio of colonies from treated/control plates.
Detection of DNA cross-linking (alkaline comet assay). Effects on DNA breakage induced by cobalt-60 γ-irradiation (10 Gy) was assayed using the alkaline comet assay at the University of Auckland (Figs. 1, 3, and 6) as previously described (18). One hundred images on duplicate slides were analyzed to determine mean tail moment. DNA cross-linking index was calculated as the ratio (A−B)/A where A = (mean tail moment of irradiated non–drug-treated cells) − (mean tail moment of unirradiated non–drug-treated cells) and B = (mean tail moment of irradiated drug-treated cells) − (mean tail moment of unirradiated drug treated cells). Studies reported in Supplementary Fig. S4 and Fig. 5B were undertaken at Stanford University using a variant (39) of the above comet assay, which had lower sensitivity to ICL.
A, time course of DNA cross-links, determined using the alkaline comet assay, in aerobic SiHa cells after a 1-h exposure to PR-104A or chlorambucil (each 80 μmol/L) under hypoxia. Values are mean ± range for duplicate cultures in the same experiment. B, C, and D, clonogenic survival and DNA cross-linking in SiHa cells after a 2-h exposure to PR-104A under hypoxic (closed symbols) of aerobic (open symbols) conditions. Values are for three experiments, represented by different symbol shapes. Lines, linear regressions with 95% confidence limits. B, surviving fraction immediately after PR-104A exposure. The hypoxic cytotoxicity ratio (HCR) is the mean separation of the regression lines. C, DNA cross-linking index 24 h after PR-104A exposure of the same cultures. D, relationship between DNA cross-linking index and cell killing in the same SiHa cell populations. Dashed line, regression through aerobic values; solid line, through hypoxic values.
A, time course of DNA cross-links, determined using the alkaline comet assay, in aerobic SiHa cells after a 1-h exposure to PR-104A or chlorambucil (each 80 μmol/L) under hypoxia. Values are mean ± range for duplicate cultures in the same experiment. B, C, and D, clonogenic survival and DNA cross-linking in SiHa cells after a 2-h exposure to PR-104A under hypoxic (closed symbols) of aerobic (open symbols) conditions. Values are for three experiments, represented by different symbol shapes. Lines, linear regressions with 95% confidence limits. B, surviving fraction immediately after PR-104A exposure. The hypoxic cytotoxicity ratio (HCR) is the mean separation of the regression lines. C, DNA cross-linking index 24 h after PR-104A exposure of the same cultures. D, relationship between DNA cross-linking index and cell killing in the same SiHa cell populations. Dashed line, regression through aerobic values; solid line, through hypoxic values.
Determination of PR-104A metabolites by liquid chromatography tandem mass spectrometry. Cells were exposed to PR-104A under aerobic and hypoxic conditions as for clonogenic assays above, except that 5 × 105 cells were plated per well in 24-well trays, in 0.35 mL medium. After attachment for 2 h, 50 μL prewarmed medium without (controls) or with PR-104A (final concentration, 100 μmol/L) was added to triplicate cultures for each cell line, which were incubated for 60 min. Plates were removed from the aerobic or anoxic incubator onto ice, and the medium was transferred to chilled microcentrifuge tubes. The cell monolayer was extracted with 0.8 mL methanol containing 1.5 μmol/L d4-PR-104H internal standard, added to the medium samples, mixed, and stored at −80°C for liquid chromatography tandem mass spectrometry analysis. Details of the analytic method are given in the legend to Supplementary Fig. S3.
Clonogenic killing and DNA cross-linking in human tumor xenografts. Female homozygous nude (CD1-Foxn1nu) mice were bred by the Vernon Jansen Unit (University of Auckland). Animal studies were approved by the University of Auckland Animal Ethics Committee (approval R279). Tumors were grown s.c from 107 cells in 100 μL α minimal essential medium. Mice were randomized to treatment groups (5 per group) when tumors reached 8 to 10 mm mean diameter and dosed i.p. with 348 mg/kg (600 μmol/kg) PR-104 at 0.01 mL/g body weight. Twenty-four hours later, tumors were excised, dissociated enzymatically (18), and clonogenic cell survival and ICL were quantified as above.
Tumor growth inhibition and DNA cross-linking in SiHa xenografts. Tumor growth delay experiments were done with 150 to 200 mm3 SiHa tumors treated with 250 mg/kg PR-104 weekly for 3 wk with each exposure being under 1 of 3 oxygen breathing conditions. Immediately after PR-104, mice were transferred to chambers where they breathed carbogen (95% oxygen, 5% CO2), air, or 10% oxygen for 2 h. This is sufficient time for PR-104 clearance (18). Tumor volumes were measured with calipers thrice per week. Six hours after the first PR-104 dose, fine needle aspirates were drawn from each tumor using a 20 cc syringe with a 23 gauge needle containing 2 to 3 mL cold calcium- and magnesium-free PBS. Cells were counted, diluted to 3 × 104/mL, and irradiated on ice (5 Gy) for measurement of ICL by comet assay. In parallel experiments, hypoxia in tumors under these three oxygen-breathing conditions was assessed using the hypoxia marker pimonidazole (100 mg/kg; Hypoxyprobe-1 kit; Chemicon Int). Percentages of pimonidazole labeled cells were quantitated using point scoring under ×150 magnification.
Results
ICL from PR-104A and chlorambucil are repaired with similar kinetics. DNA ICL frequencies in SiHa cells, determined using the alkaline comet assay, showed that ICL followed a similar time course for both PR-104A and chlorambucil after a 1-hour hypoxic exposure, increasing in the first 4 hours after treatment before declining with a half-life of ∼2 days (Fig. 1A).
ICL correlate with PR-104A clonogenic cell killing in aerobic and hypoxic SiHa cells. Consistent with previous reports (18, 19), PR-104A was selectively toxic to SiHa cells under hypoxia; dose-response curves for clonogenic cell killing showed a potency differential between hypoxic and aerobic exposure (hypoxic cytotoxicity ratio) of 22 (Fig. 1B). Dose-response relationships for DNA ICL also showed ∼20-fold higher frequencies under hypoxia (Fig. 1C), and the relationship between clonogenic cell killing and ICL 24 hours after PR-104A exposure was similar under aerobic and hypoxic conditions with ∼3 logs of cell kill at a DNA cross-link index of 0.5 (Fig. 1D). In contrast, the dose response for chlorambucil-induced ICL and cell killing (Supplementary Fig. S2) showed only minor differences between aerobic and hypoxic exposure (1.2-fold for cell killing; ∼1.7-fold for ICL); again a DNA cross-linking index of 0.5 corresponded to ∼3 logs of cell killing under both conditions.
PR-104A ICL correlate with clonogenic cell killing in a panel of human tumor cell lines under hypoxia. We then asked whether this same relationship held across a panel of nine human tumor cell lines. Clonogenic survival curves showed little difference in PR-104A sensitivity between cell lines under hypoxia, with only a 3-fold range in concentrations required for 10% survival (C10; Fig. 2). Under aerobic conditions, there was a larger 18-fold range in C10 values, resulting in a ∼10-fold variation in hypoxic cytotoxicity ratio values (Fig. 2). For 4 of these cell lines, we followed the time course of ICL after a 2-hours hypoxic exposure to PR-104A (Fig. 3A). This showed broadly similar kinetics in each cell line with rapid ICL formation and slow repair. At 24 hours, residual ICLs were readily measurable and this time was selected for comparison of all 9 cell lines. The dose response for ICL was similar in all cell lines except C33A and 22Rv1, which required 2- to 3-fold lower concentrations for a cross-linking index of 0.5 (Fig. 3B). The relationship between ICL and clonogenic cell killing under hypoxia (Fig. 3B) was similar to that for SiHa, and was indistinguishable for all 9 cell lines with an r2 for the cytotoxicity/ICL correlation of 0.74 (P < 0.001).
Selective cytotoxicity of PR-104A against a panel of 9 human tumor cell lines determined by clonogenic assay after exposure to PR-104A for 2 h under aerobic or hypoxic conditions. C10 value is the interpolated drug concentration required for a surviving fraction of 10%. Columns, mean for replicates from two (H1299, HCT116, HT29, and H460) or three (SiHa) independent experiments; bars, SE. The hypoxic cytotoxicity ratio, shown above each cell line, is the ratio of C10 values under aerobic and hypoxic conditions.
Selective cytotoxicity of PR-104A against a panel of 9 human tumor cell lines determined by clonogenic assay after exposure to PR-104A for 2 h under aerobic or hypoxic conditions. C10 value is the interpolated drug concentration required for a surviving fraction of 10%. Columns, mean for replicates from two (H1299, HCT116, HT29, and H460) or three (SiHa) independent experiments; bars, SE. The hypoxic cytotoxicity ratio, shown above each cell line, is the ratio of C10 values under aerobic and hypoxic conditions.
DNA ICLing in multiple human tumor cell lines after 2 h of exposure to PR-104A. A, kinetics of formation and removal of DNA cross-links in aerobic tumor cell lines after hypoxic exposure to PR-104A at 40 μmol/L. Dashed line, result for SiHa cells redrawn from Fig. 1A (80 μmol/L for 1 h). B and C, dose response for ICL, 24 h after hypoxic (B) or aerobic (C) exposure to PR-104A (left) and relationship to clonogenic cell survival determined immediately after drug exposure fitted by first order regression through all data with 95% confidence intervals (right). Dotted ellipse, encompasses values from three cell lines (A2780, C33A, and H1299) with no significant DNA cross-linking at PR-104A concentrations causing extensive clonogenic cell killing (C).
DNA ICLing in multiple human tumor cell lines after 2 h of exposure to PR-104A. A, kinetics of formation and removal of DNA cross-links in aerobic tumor cell lines after hypoxic exposure to PR-104A at 40 μmol/L. Dashed line, result for SiHa cells redrawn from Fig. 1A (80 μmol/L for 1 h). B and C, dose response for ICL, 24 h after hypoxic (B) or aerobic (C) exposure to PR-104A (left) and relationship to clonogenic cell survival determined immediately after drug exposure fitted by first order regression through all data with 95% confidence intervals (right). Dotted ellipse, encompasses values from three cell lines (A2780, C33A, and H1299) with no significant DNA cross-linking at PR-104A concentrations causing extensive clonogenic cell killing (C).
Aerobic cytotoxicity of PR-104A is not mediated by ICL in all cell lines. Under aerobic conditions, higher PR-104A concentrations were required to form ICL than under hypoxia, but much greater differences between cell lines were now evident (Fig. 3C). Cross-linking frequencies were high in 5 of the lines but were lower in HCT116 and not detectable in A2780, C33A, or H1299 cells at up to 600 μmol/L PR-104A. Plotting the cross-linking index against clonogenic cell killing identified these three cell lines as outliers, with substantial cell killing in the absence of significant cross-linking (Fig. 3C,, points within the dotted oval). Excluding these three cell lines, there was again a strong correlation between the DNA cross-linking index and clonogenic cell killing (r2 = 0.57; P < 0.001); across all 9 cell lines (Fig. 3C), the correlation was still significant (P < 0.001) but weaker than under hypoxia with the r2 value decreasing to 0.34. These observations suggested that aerobic cytotoxicity of PR-104A is mediated by ICL in some tumor cell lines but is due to a different mechanism in A2780, C33A and H1299.
Metabolism of PR-104A to its reduced metabolites by tumor cell lines. To examine this further, we measured the metabolism of PR-104A to its cytotoxic reduced metabolites, PR-104H and PR-104M, which are considered responsible for the cross-linking activity of PR-104A (18). To do this, we developed a triple quadruple mass spectrometry assay using tetradeuterated (d4) PR-104H as an internal standard. Difficulties in synthesizing authentic PR-104M required calibrating the mass spectrometry response by reducing 3H-PR-104A with hypoxic SiHaCYPOR cells that overexpress the one-electron reductase NADPH:cytochrome P450 oxidoreductase and generate relatively high levels of PR-104M (38). Comparison of radioactivity and absorbance of the PR-104A, PR-104H, and PR-104M high performance liquid chromatography peaks (Supplementary Fig. S3A) showed that molar extinction coefficients of all three species are indistinguishable at 250 nm (Supplementary Fig. S3B). Comparing the ion counts for PR-104H and PR-104M from these fractions showed that PR-104M was detected with 1.73-fold higher efficiency than PR-104H under the assay conditions used (Supplementary Fig. S3C), making it possible to quantify PR-104M using the d0/d4 PR-104H calibration curve. The assay gave a linear response (Supplementary Fig. S3D) over at least 3 orders of magnitude from the lower limit of quantitation (corresponding to 1 nmol/L in the cell culture).
We used this assay to determine steady-state concentrations of PR-104H and PR-104M at 1 hour. Concentrations of both reduced metabolites were similar to each other under anoxia (Fig. 4A), with highest values in SiHa and lowest in A2780. Although this 13-fold difference was significant by ANOVA (P = 0.003), there was no correlation between reduced metabolites (PR-104H, PR-104M, or total) and cytotoxicity (C10 values; P > 0.05; r2 = 0.02), suggesting differences in intrinsic sensitivity to the activated nitrogen mustards are a larger determinant. Thus, under anoxia, ICL determined with the comet assay provides a more reliable biomarker for PR-104A cytotoxicity than levels of active metabolites.
Metabolism of PR-104A to its cytotoxic metabolites (PR-104H and PR-104M) in 9 human tumor cell lines under hypoxic (A) and aerobic (B) conditions, assayed by liquid chromatography tandem mass spectrometry after a 1-h exposure to 100 μmol/L PR-104A. Values are for methanol-extracted whole cultures (5 × 105 cells in 0.4 mL); columns, means for two to four independent experiments with three replicates per experiment; bars, SE.
Metabolism of PR-104A to its cytotoxic metabolites (PR-104H and PR-104M) in 9 human tumor cell lines under hypoxic (A) and aerobic (B) conditions, assayed by liquid chromatography tandem mass spectrometry after a 1-h exposure to 100 μmol/L PR-104A. Values are for methanol-extracted whole cultures (5 × 105 cells in 0.4 mL); columns, means for two to four independent experiments with three replicates per experiment; bars, SE.
Oxygen suppressed levels of reduced metabolites in all cell lines, but much larger differences were now evident (Fig. 4B) with a ∼137-fold range between A2780 and the fastest metabolizer, A549. Notably, the four cell lines with the slowest rate of aerobic reduction of PR-104A (A2780, C33A, H1299, and HCT116) were the lines with low levels of ICL (Fig. 3C), although the linear regression between reduced metabolites and cytotoxicity under aerobic conditions did not reach significance (P = 0.09; r2 = 0.34).
PR-104 forms hypoxia-dependent ICL in SiHa tumor xenografts. We then tested whether the alkaline comet assay could detect DNA cross-links in human tumor xenografts under conditions providing monotherapy antitumor activity. After treatment of mice with PR-104 at 175 mg/kg, ICL were maximal at 4 to 6 hours, consistent with kinetics in vitro, although subsequent unhooking of cross-links seemed to be more rapid (Supplementary Fig. S4A). This might reflect faster elimination of dead cells with ICL in tumors than in culture. PR-104 showed a linear dose response for ICL formation up to at least 450 mg/kg in SiHa tumors when assayed 6 hours after treatment (Fig. S4B).
To modify the level of hypoxia in SiHa tumors, we placed mice in 95% O2/5% CO2 (carbogen), air, or 10% O2/90% N2 for 2 hours after dosing with the hypoxia marker pimonidazole and assessed its binding in tumors by immunohistochemistry. Estimated hypoxic fractions were 9.4%, 16.1%, and 45.9% for mice breathing carbogen, air, and 10% O2, respectively (Supplementary Table S1). These gas mixtures were then used for 2 hours after each dose of PR-104 on a weekly ×3 schedule. Hyperoxic gas breathing suppressed monotherapy activity of PR-104, whereas the hypoxic gas mixture increased its activity (Fig. 5A). We determined ICL frequency in fine needle aspirate samples from the same tumors, 6 hours after the first dose of PR-104, and in a second experiment in which separate mice were used to determine tumor growth delay and ICL (Fig. 5B). This showed a strong correlation between ICL frequency and tumor growth delay, and confirmed that ICL formation depends on the level of hypoxia in this tumor model.
ICL in vivo predict the sensitivity of human tumor xenografts to PR-104. A, effect of hypoxia on inhibition of SiHa tumor growth by PR-104. Nude mice with 150 to 200 mm3 SiHa xenografts were treated with 250 mg/kg PR-104 weekly for 3 wk with the animals breathing 10% O2, air, or 95% oxygen/5% CO2 (carbogen) for 2 h immediately after each PR-104 injection. Mice breathing carbogen also received nicotinamide (500 mg/kg, i.p.) 15 min before PR-104. Points, mean for five mice per group, plotted from the first day of treatment; bars, SE. B, tumor growth delay after three treatments with 250 mg/kg PR-104 compared with DNA cross-link index measured by fine needle aspirate after the first PR-104 treatment. Open symbols, PR-104 on days 0, 7, and 14, with cross-links and growth delay measured on the same tumors as in A. Closed symbols, PR-104 on days 0, 4, and 8 with cross-links and growth delay measured on different tumors. Points, mean of two to five mice per group; bars, SE.
ICL in vivo predict the sensitivity of human tumor xenografts to PR-104. A, effect of hypoxia on inhibition of SiHa tumor growth by PR-104. Nude mice with 150 to 200 mm3 SiHa xenografts were treated with 250 mg/kg PR-104 weekly for 3 wk with the animals breathing 10% O2, air, or 95% oxygen/5% CO2 (carbogen) for 2 h immediately after each PR-104 injection. Mice breathing carbogen also received nicotinamide (500 mg/kg, i.p.) 15 min before PR-104. Points, mean for five mice per group, plotted from the first day of treatment; bars, SE. B, tumor growth delay after three treatments with 250 mg/kg PR-104 compared with DNA cross-link index measured by fine needle aspirate after the first PR-104 treatment. Open symbols, PR-104 on days 0, 7, and 14, with cross-links and growth delay measured on the same tumors as in A. Closed symbols, PR-104 on days 0, 4, and 8 with cross-links and growth delay measured on different tumors. Points, mean of two to five mice per group; bars, SE.
PR-104 cross-links correlate with clonogenic cell killing in xenografts. The above results suggest that ICL may provide a pharmacodynamic biomarker of PR-104 activity in tumors. However, in vitro studies showed that ICL are not universally predictive of cytotoxicity under aerobic conditions across all cell lines. To test whether ICL provide a reliable pharmacodynamic biomarker across tumor types in vivo, we determined ICL and clonogenic cell killing in the same 9 cell lines, grown as xenografts in nude mice, 24 hours after a single i.p. dose of PR-104. This showed a strong correlation (r2 = 0.75; P < 0.001) between ICL and clonogenic cell killing across all 9 cell lines (Fig. 6).
Relationship between DNA cross-links and clonogenic cell killing in nine human tumor cell lines grown as s.c. xenografts. Tumors (mean diameter, 8–10 mm) were assayed 24 h after a single i.p. dose of PR-104 at 350 mg/kg. Each point is for a single tumor. Solid line, first order regression fitted to the combined data; dashed lines represent the 95% confidence intervals.
Relationship between DNA cross-links and clonogenic cell killing in nine human tumor cell lines grown as s.c. xenografts. Tumors (mean diameter, 8–10 mm) were assayed 24 h after a single i.p. dose of PR-104 at 350 mg/kg. Each point is for a single tumor. Solid line, first order regression fitted to the combined data; dashed lines represent the 95% confidence intervals.
Discussion
The present study shows that PR-104A induces ICL in a hypoxia-selective manner in all nine human tumor cell lines investigated, and that ICL correlate with clonogenic cell killing under hypoxia across this cell line panel (Fig. 3B). This is strong evidence that ICL formation is the dominant mechanism of PR-104A cytotoxicity in hypoxic tumor cells, as intended in its design. Our findings are consistent with the earlier demonstration that Chinese hamster ovary cells with mutant XPF (defective in the unhooking of ICL) show marked (∼25-fold) hypersensitivity to the reduced metabolite PR-104H (18).
Coupled with evidence for replication arrest at PR-104A–derived ICL, as shown by γH2AX (18, 40), this suggests that PR-104A disrupts the DNA replication fork and thus shares a common downstream mechanism with many other cytotoxins (topoisomerase poisons, chain-terminating antimetabolites, and cross-linking agents such as platins, mitomycin C, and nitrogen mustards). Given the critical importance of these agents in cancer therapy, the replication fork is arguably the most important molecular target in oncology at present (28, 41), and has clinically validated (if incompletely understood) tumor selectivity that seems to reside in the alterations in cell cycle progression, replication stress, and defects in the resolution of stalled replication forks by homologous recombination in tumor cells (41, 42). In fact, recent studies suggest that chronic hypoxia suppresses translation of key proteins required for homologous recombination, resulting in sensitization to ICL and radiation (43). However, cells that are only transiently hypoxic may be more important in tumor progression and therapeutic resistance (9). PR-104 provides a strategy for exploiting the sensitivity of chronically hypoxic cells to DNA cross-links but also superimposes a hypoxia-selective release mechanism that ensures highly cytotoxic levels of DNA cross-links in both acutely and chronically hypoxic cells. This selective prodrug activation is clearly the main contributor to selective hypoxic cell killing by PR-104A when cell cultures are made acutely hypoxic as shown by much weaker hypoxic selectivity of chlorambucil (Supplementary Fig. S2) and by the demonstration that PR-104H lacks hypoxic selectivity under similar conditions.3
Y. Gu, A.V. Patterson, G.J. Atwell et al. Roles of DNA repair and reductase activity in the cytotoxicity of the hypoxia-activated dinitrobenzamide mustard PR-104A. Mol Cancer Ther, submitted.
PR-104A is metabolized to PR-104H at a much lower rate in aerobic than hypoxic SiHa and MDA-231 cells (18, 38). A surprising finding in the present study was that aerobic bioreductive metabolism is highly variable between human tumor cell lines (Fig. 4B); we have recently identified the enzyme responsible for this metabolism as aldo-keto reductase 1C3 (44). Cell lines lacking this two-electron reductase (A2780, C33A, H1299, and HCT116) show little or no cross-linking after aerobic exposure to PR-104A, whereas substantial cytotoxicity is still observed (Fig. 3C). These results clearly establish that cross-linking by PR-104A is due to its bioreductive metabolism but also show that there is a second (as yet unidentified) mechanism of cytotoxicity in cell lines lacking the aerobic reductase. Whatever the mechanism of cross-link–independent aerobic cytotoxicity in culture, it seems not to contribute significantly in tumors at pharmacologically relevant exposures given that we found a strong correlation between ICL and killing across all cell lines when treated with PR-104 as xenografts (Fig. 6). This validates ICL frequency as a surrogate end point for clonogenic cell killing in tumors. The alkaline comet assay is well-validated for quantifying ICL (45–47). Notably, in the present study, we show that ICL frequency is a better biomarker than PR-104H/M metabolite levels in culture, suggesting that the downstream marker (cross-links) is likely to be more useful than tumor metabolite levels in support of clinical development. In addition, the much longer half-life of nitrogen mustard cross-links than initial reactive metabolites makes ICL frequency more practical as a response biomarker.
It is noteworthy that PR-104 showed broad spectrum activity against human tumor xenografts as monotherapy after treatment at just 45% of the maximum tolerated dose (Fig. 6). Monotherapy activity was also observed for SiHa tumors using growth delay end points (Fig. 5A) as previously described (18). This activity would not be expected for a HAP that only eliminates the hypoxic subpopulation. We have recently shown, using flow cytometry to quantify pimonidazole binding, that the hypoxic fraction of the tumors in the xenograft panel used in the present study is in the range 10% to 62% (data not shown), which is too low to account for the monotherapy activity of PR-104 (>90% kill) for 22Rv1, A549, SiHa, and H460 tumors in Fig. 6. This raises the critical question as to how PR-104 kills aerobic tumor cells; the evidence from the present study suggests this reflects both aerobic nitroreduction to PR-104H/M and a bystander effect due to their diffusion from hypoxic zones.
The key observation suggesting a role for aerobic nitroreduction is that the most sensitive tumors (SiHa, A549, H460, and 22Rv1) were those with high rates of aerobic metabolism in culture. The relationship between aerobic metabolism in cell lines (Fig. 4B) and monotherapy activity in the present study (Fig. 6), although significant, is weak (P = 0.032; r2 = 0.505 for first order regression of log PR-104H+M against log cell kill), as is also the case for aerobic cytotoxicity in culture (C10 values in Fig. 2) versus log cell kill in tumors (P = 0.042; r2 = 0.470). This leaves open the possibility that hypoxic metabolism also plays a part in monotherapy activity of PR-104.
The present study provides evidence that hypoxia is also involved in the monotherapy activity of PR-104. Thus suppressing hypoxia in SiHa tumors using carbogen/nicotinamide after PR-104 inhibited antitumor activity and ICL formation, whereas increasing tumor hypoxia using 10% oxygen enhanced antitumor activity and ICLs (Fig. 5). Even under conditions of hypoxic breathing, monotherapy activity seems too great to be accounted for by killing of the 46% of cells that are positive for pimonidazole binding (Supplementary Table S1), suggesting the existence of a bystander effect. This is consistent with a previous pharmacokinetic/pharmacodynamic modeling study that suggested tumor cell killing in SiHa xenografts to be too great to explain without a bystander effect (19). Further studies will be required to clarify the relative contribution of these two pathways (aerobic activation and hypoxic activation with a bystander effect) to the activity of PR-104 in both monotherapy and combination settings.
Disclosure of Potential Conflicts of Interest
W.R. Wilson, J.M. Brown, and A.V. Patterson are stockholders in and consultants to Proacta, Inc. The other authors disclosed no potential conflicts of interest.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
Acknowledgments
Grant support: Tertiary Education Commission Top Achiever Doctoral Scholarship (R.S. Singleton), Health Research Council of New Zealand grant 08/103 (D.M. Ferry, S.M. Pullen, A.V. Patterson, and W.R. Wilson), Proacta, Inc., (C.P. Guise), and NIH grants P01 CA67166 and R01 CA-118202 (J.M. Brown, M.J. Dorie).
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.
We thank Dr. B.G. Siim for teaching the comet assay, G.J. Atwell and Prof. W.A. Denny for provision of PR-104A, and K. Patel for synthesis of PR-104H.