Schedule-dependent potentiation of chemotherapy drugs by the hypoxia-activated prodrug SN30000

ABSTRACT

Hypoxia-activated prodrugs (HAPs) are hypothesized to improve the therapeutic index of chemotherapy drugs that are ineffective against tumor cells in hypoxic microenvironments. SN30000 (CEN-209) is a benzotriazine di-N-oxide HAP that potentiates radiotherapy in preclinical models, but its combination with chemotherapy has not been explored. Here we apply multiple models (monolayers, multicellular spheroids and tumor xenografts) to identify promising SN30000/chemotherapy combinations (with chemotherapy drugs before, during or after SN30000 exposure). SN30000, unlike doxorubicin, cisplatin, gemcitabine or paclitaxel, was more active against cells in spheroids than monolayers by clonogenic assay. Combinations of SN30000 and chemotherapy drugs in HCT116/GFP and SiHa spheroids demonstrated hypoxia-and schedule-dependent potentiation of gemcitabine or doxorubicin in growth inhibition and clonogenic assays. Co-administration with SN30000 suppressed clearance of gemcitabine in NIH-III mice, likely due to SN30000-induced hypothermia which also modulated extravascular transport of gemcitabine in tumor tissue as assessed from its diffusion through HCT116 multicellular layer cultures. Despite these systemic effects, the same schedules that gave therapeutic synergy in spheroids (SN30000 3 h before or during gemcitabine, but not gemcitabine 3 h before SN30000) enhanced growth delay of HCT116 xenografts without increasing host toxicity. Identification of hypoxic and S-phase cells by immunohistochemistry and flow cytometry established that hypoxic cells initially spared by gemcitabine subsequently reoxygenate and re-enter the cell cycle, and that this repopulation is prevented by SN30000 only when administered with or before gemcitabine. This illustrates the value of spheroids in modeling tumor microenvironment-dependent drug interactions, and the potential of HAPs for overcoming hypoxia-mediated drug resistance.

Introduction

Hypoxia is a prevalent characteristic of tumors^1,2 due to structurally and functionally inefficient microvasculature.^3,4 Hypoxic tumor cells are resistant to many chemotherapy drugs via multiple mechanisms,⁵ including limited extravascular transport into hypoxic zones,⁶ decreased proliferative rates because of oxygen and nutrient deprivation,⁷ and changes in gene expression driven by clonal selection or adaptation to hypoxia.^8,9 Although severely hypoxic cells in tumors have limited survival times,^10,11 hypoxic cells that survive chemotherapy can be ‘rescued’ by improved nutrient and oxygen supply following loss of more chemosensitive cells closer to blood vessels, thus enabling previously hypoxic cells to repopulate tumors.^11–13

Hypoxia-activated prodrugs (HAPs) are activated selectively under hypoxia^5,14,15 and therefore have potential for eliminating hypoxic tumor cells spared by chemotherapy. Such interactions are expected to be tumor-selective given that hypoxia is more prevalent and severe in tumors than in normal tissues.^1,4 There is evidence for therapeutic synergy between HAPs and chemotherapy in preclinical models^11,16,17 and in clinical trials,^18–20 although the latter activity has not been sufficient for HAPs to be approved for clinical use. Thus there is a need to identify optimal HAP/chemotherapy combinations, which is challenging to do efficiently because of the limitations of tissue culture models in representing the complexity of drug pharmacokinetics and pharmacodynamics in the tumor microenvironment.

Tumor spheroids capture important characteristics of tumor microenvironments, such as gradients of oxygen, nutrients, cell proliferation, and necrosis.^21–23 Hence, spheroids are seen as a tractable in vitro model for exploration of drug combinations, filling the gap between conventional monolayer cell culture and in vivo models; however few studies have used this approach to examine combinations with HAPs. In the present study, we explore the use of multicellular spheroids as a key step in expediting identification of HAP/chemotherapy combinations warranting critical evaluation against tumors.

The benzotriazine di-N-oxide SN30000 (CEN-209) is an analog of the well-studied HAP tirapazamine, with superior extravascular diffusion properties and activity against hypoxic cells in xenograft models.^24,25 Free radical metabolites of SN30000 are responsible for its cytotoxicity in hypoxic cells,²⁶ but these are too reactive to diffuse out of hypoxic zones.²⁷ This localization of killing to the cells that metabolically activate SN30000 (i.e. lack of a bystander effect) simplifies investigation of its pharmacokinetics/pharmacodynamics, making it an ideal model compound to investigate HAP/chemotherapy drug interactions. The spatial complementarity of SN30000 and radiation, which have almost mirror-image oxygen dependencies,²⁴ has been demonstrated,^24,28 but combination of SN30000 with chemotherapeutic drugs has not been investigated.

Here, we explore schedule- and hypoxia-dependent responses of HCT116 and SiHa spheroids and monolayers to SN30000 in combination with widely used representatives of the main classes of chemotherapy drugs, using both growth delay and clonogenic survival endpoints. The most promising combination (SN30000 + gemcitabine) was further tested using HCT116 xenografts, including evaluation of changes in the subpopulation of hypoxic cells and repopulation in response to the treatments.

Results

Sensitivity of monolayers and spheroids to SN30000 and chemotherapy drugs

Multicellular spheroids model important features of the tumor microenvironment that influence drug sensitivity, including hypoxia. However, when grown under (supraphysiologic) 20% O₂ as in the present study 4 day old HCT116 spheroids contain very few cells that are hypoxic enough to bind the hypoxia probe EF5, while transfer to 5% O₂ 3 h before EF5 exposure markedly increases the hypoxic fraction.²⁹ We showed that sensitivity to SN30000, assessed by clonogenic assay, increased progressively for 3 h after the transfer to 5% O_2, with no further change by 5 h (Figure S1). Thus, a 3 h equilibration time was used in all subsequent experiments. Under these conditions, SN30000 was more active against cells in spheroids than monolayers, while three chemotherapy drugs showed either lower (doxorubicin, gemcitabine) or unchanged (cisplatin) activity in spheroids (Figure 1A). The resistance of spheroid to doxorubicin at high concentrations is consistent with its limited ability to penetrate spheroids,^30–34 while resistance to gemcitabine may reflect the low S-phase fraction in 4-day old HCT116 spheroids²⁸.

Figure 1.

Clonogenic survival of HCT116 monolayers and spheroids exposed to SN30000 or clinical chemotherapeutic drugs. HCT116 monolayers (5 × 10⁴ cells/well) and spheroids (one spheroid/well) grown under 20% O₂ in 96-well plates were exposed to a range concentrations of SN30000 (A) for 2 h under 5% O₂ or cisplatin (B), doxorubicin (C), or gemcitabine (D) for 2 h under 20% O₂. Immediately after drug exposure, 2 wells of monolayers or 8 spheroids were pooled for each group and dissociated to single cell suspensions for clonogenic survival assay. Values are mean ± SEM for 2–4 replicate determinations of plating efficiency from each pool for spheroids, and from 4 pools for monolayers. The curves are fitted to the clonogenic cell kill data using monoexponential without minimum (A, B), bi-exponential (C, representing sensitive and resistant subpopulations, without minimum) and monoexponential (D, with minimum representing a completely resistant subpopulation).

SN30000 potentiates activity of some chemotherapy drugs in spheroids

To interrogate SN30000/drug interactions we then assessed clonogenic survival of HCT116/GFP and SiHa spheroids immediately after the end of drug treatment under 5% or 20% O₂ (Figure 2). Three different treatment schedules were compared: SN30000 before (SN-Drug) or after (Drug-SN) the chemotherapy drug with 1 h incubation in drug-free medium between the 2 h treatments, or addition of both drugs at the same time (SN+Drug). To provide similar (ca. 90%) clonogenic cell kill by SN30000, spheroids were exposed to 25 μM SN30000 under 5% O₂ or 100 μM SN30000 under 20% O₂. A marked supra-additive response was observed with simultaneous exposure to cisplatin and SN30000. In addition, despite only marginal activity of doxorubicin or gemcitabine alone, combination with SN30000 gave greater than additive responses in both HCT116 and SiHa spheroids. These interactions were greatest in more hypoxic spheroids (5% versus 20% O₂) and under 5% O₂ the interaction was greater with drug treatment before or during SN30000 exposure.

Figure 2.

Clonogenic survival of cells in spheroids in response to SN30000 and/or chemotherapeutic drugs under 5% and 20% O₂. HCT116/GFP (A) or SiHa (B) spheroids were exposed to SN30000 (25 µM under 5% O₂, or 100 µM under 20% O₂), chemotherapy drug (25 µM cisplatin, 5 µM doxorubicin, 10 µM gemcitabine), combinations of SN30000 before (SN-Drug) or after (Drug-SN) chemotherapy drugs with 1 h interval between, or simultaneously (SN+Drug). Each drug exposure was for 2 h. Immediately after the final drug treatment, 8 spheroids were pooled for each group and dissociated to single cell suspensions for clonogenic survival assay. The bold horizontal lines represent the predicted combined effect from adding the log cell kill for the individual drugs under 5%, (solid lines) or 20% O₂ (dashed lines). Values are mean ± SEM for 2–4 replicate determinations of plating efficiency from each pool.

Overall spheroid (or tumor) growth inhibition is influenced by inhibition of cell proliferation as well as sterilization of clonogens, so we also measured growth of spheroids after treatment using the same drugs and schedules as in the clonogenic survival assays. Firstly, we confirmed the stability of GFP expression in HCT116/GFP cells by comparing the GFP fluorescence intensity of cells cultured with or without puromycin selection for 2 weeks (Figure S2A). We then tested whether growth of HCT116/GFP spheroids, grown in the absence of puromycin, could be monitored with a fluorescence plate reader; this showed a linear relationship with volume up to at least 0.25 mm³ (Figure S2B). HCT116/GFP spheroid growth was therefore monitored using total fluorescence of spheroids. Exposure to 25 μM SN30000 alone under 5% O₂ induced a small growth delay (Figure 3A) while, as for the clonogenic assays, a higher concentration (100 μM) was required for similar activity in spheroids with a smaller hypoxic fraction under 20% O₂ (Figure 3B). Under 5% O₂ (Figure 3A) the potentiation of cisplatin by SN30000 was only observed when cells were exposed to both agents simultaneously, consistent with the spheroid clonogenic survival data. We also evaluated paclitaxel, which significantly inhibited growth of the spheroids (despite lack of clonogenic cell killing at this concentration, data not shown) with enhancement of activity by pre-treatment with SN30000. Spheroid growth inhibition by either doxorubicin or gemcitabine was strongly enhanced by co-treatment with SN30000 or by pre-treatment with SN30000. Results for spheroids exposed under 20% O₂ (Figure 3B) were qualitatively similar but enhancement of doxorubicin, gemcitabine and paclitaxel activity by SN30000 was generally less than under 5% O₂, despite the use of a 4-fold higher SN30000 concentration, which was again in agreement with the clonogenic survival data. This is consistent with more effective killing by SN30000 of cells in the center of spheroids that are resistant to these drugs (but not to cisplatin) when the hypoxic zone is increased.

Figure 3.

HCT116/GFP spheroid growth delay induced by SN30000 and chemotherapeutic drugs. HCT116/GFP spheroids were exposed to SN30000 (25 µM under 5% O₂, or 100 µM under 20% O₂, for 2 h) and chemotherapy drugs (25 µM cisplatin for 2 h, 5 µM doxorubicin for 2 h and 10 µM gemcitabine for 2 h) under 5% O₂ (A) and 20% O₂ (B), using the same schedules as in Figure 2. Immediately after the final drug exposure, spheroids were cultured in fresh medium in a standard incubator (20% O₂) and growth was monitored every second day by measuring fluorescence intensity of HCT116/GFP spheroids. Values are mean ± SEM, n = 16 spheroids. (C) Representative bright-field images of spheroids after treatment on day 4 with SN30000 and/or gemcitabine under 5% O₂ are illustrated. P values for statistically significant differences between gemcitabine only and the combination groups were determined with one-way ANOVA. * P < 0.05; ** P < 0.01; *** P < 0.001.

Bright field images of spheroids confirmed the results from GFP fluorescence as illustrated for SN30000/gemcitabine Figure 3C which shows that SN30000 alone had little effect, gemcitabine alone arrested spheroid growth until day 16 but SN30000 dosed before or at the same time as gemcitabine completely destroyed the spheroids with accumulation of cell debris at the bottom of the wells. Again, SN30000 after gemcitabine moderately delayed spheroid growth compared with gemcitabine alone but was less effective than the other two schedules.

The lower activity of SN30000 after gemcitabine suggested that gemcitabine may suppress hypoxia in HCT116/GFP spheroids. We therefore labeled hypoxic cells with PIMO before gemcitabine treatment and with EF5 3 h after gemcitabine treatment, which demonstrated increased PIMO+/EF5- cells after gemcitabine (controls 9.0 ± 0.8% vs treated 12.9 ± 1.6%, p < 0.05; Figure S3) suggesting that reoxygenation after gemcitabine may contribute to the observed schedule dependence.

Schedule-dependent spheroid growth delay of each combination was also tested in SiHa spheroids under 5 and 20% O₂ (Figure S4), providing results very similar to HCT116/GFP spheroids including marked hypoxia-dependent inhibition by SN30000 before or during gemcitabine treatment, although doxorubicin at the same concentration as used with HCT116/GFP spheroids (5 μM) was off-scale suggesting that its penetration into these relatively loosely packed spheroids³⁵ may be less of a limitation than in HCT116 spheroids. At a lower doxorubicin concentration, providing measurable growth inhibition, activity was slightly enhanced by SN30000 at 5% O₂ but not at 20% O₂ (Figure S5).

In vivo translation: implications of SN30000-induced hypothermia in mice

Although spheroids capture multiple characteristics of real tumors, they of course cannot represent all in vivo features. One potential confounding factor is the previously-reported dose dependent hypothermia induced in mice by SN30000³⁶ which has the potential to modify pharmacokinetics and pharmacodynamics in vivo. Therefore, we investigated the effect of SN30000 on the systemic (plasma) pharmacokinetics of gemcitabine. We confirmed that mouse body temperature rapidly decreased after i.p. dosing of SN30000 at 130 mg/kg, with a nadir of 31℃ at 30 min followed by gradual recovery over 3 h (Figure 4A). Gemcitabine (100 mg/kg) did not induce hypothermia or influence the hypothermic effect of SN30000 when dosed before or with SN30000 (Figure 4A).

Figure 4.

Implications of SN30000-induced hypothermia in mice. (A) Mouse body temperature response to the drug treatments. Mice were injected i.p. with saline, SN30000 (130 mg/kg), gemcitabine (100 mg/kg), SN30000 3 h before (SN-3hr-Gem) or after (Gem-3hr-SN) gemcitabine, or gemcitabine 5 min before SN30000 (Gem-5min-SN). Values are mean ± SEM for three mice. Plasma pharmacokinetics of gemcitabine (B), SN30000 and its 1-oxide metabolite (C) after dosing as in (A). Values are mean ± SEM, n ≥ 3 mice. *** P < .001 between co-administration and gemcitabine alone groups. (D) Diffusion of gemcitabine, (300 µM), urea and mannitol through HCT116 multicellular layers under 32℃ or 37℃. Values are mean ± SEM, n = 4 replicates. Mannitol curves are offset on the time axis for clarity. Flux of gemcitabine was significantly greater at 32℃ (P = .005) than at 37℃, while flux of urea was not significantly different at the 2 temperatures (P = .072).

These data suggest that when gemcitabine and SN30000 are dosed simultaneously, hypothermia could impact gemcitabine activity. We therefore investigated potential pharmacokinetic interactions in mice. For co-administration, gemcitabine was dosed 5 min before SN30000, rather than combining in the dosing solution. Co-administration substantially prolonged the plasma half-life of gemcitabine, presumably as a consequence of hypothermia (Figure 4B), while gemcitabine had little effect on the pharmacokinetics of SN30000 or reductive metabolism to its 1-oxide metabolite (Figure 4C). Plasma pharmacokinetic parameters for gemcitabine in the mice (Table S1), showed that exposure of xenografts to gemcitabine when administered 3 h before or after SN30000 (plasma AUC = 28 µM×h) was similar to that in the spheroid experiments (AUC = 20 µM×h), but the AUC was substantially increased when SN30000 and gemcitabine were co-administered (AUC = 66 µM×h).

To explore potential intratumoral effects of hypothermia, the diffusion of gemcitabine through HCT116 multicellular layer cultures was compared at 32℃ and 37℃(Figure 4D). This showed that penetration of gemcitabine is dramatically lower than for mannitol or urea. In addition, gemcitabine transport was substantially faster at 32℃ than at 37℃. The results suggest that the penetration distance of gemcitabine in 3D cell models is markedly constrained by its cellular uptake and metabolic trapping, and that this entrapment is also suppressed at the lower temperature. The hypothermic effect of SN30000 thus impacts both systemic and intratumoral pharmacokinetics of gemcitabine when they are co-administered.

SN30000 potentiates activity of gemcitabine in HCT116 tumor xenografts in a schedule-dependent manner

The two most promising combinations (SN30000 with gemcitabine or doxorubicin) in the spheroid models were investigated against HCT116 xenografts. The most favourable in vitro schedule for SN30000 and doxorubicin (co-administration) resulted in increased mouse toxicity relative to the individual drugs (Figure S6) so was not investigated further. A pilot experiment to investigate the most active gemcitabine schedule in spheroids (SN30000 3 h before gemcitabine), administered on day 0 and day 7, showed a promising tumor growth delay and survival benefit relative to gemcitabine alone (Figure S7). We then tested all the three schedules, again administered on days 0 and 7, in a separate experiment which demonstrated that both co-administration and SN30000 before gemcitabine enhanced tumor growth delay relative to gemcitabine alone while SN30000 3 h after gemcitabine did not (Figure 5A). The preferred schedules also gave significantly longer survival times (Figure 5B). These combinations of SN30000 and gemcitabine did not show additional host toxicity, as assessed by body weight loss, compared to each drug alone (Figure 5C).

Figure 5.

HCT116 tumor growth delay induced by SN30000 and gemcitabine. Mice bearing HCT116 xenografts were randomized when tumors reached 9 mm diameter and treated as described in the legend of Figure 4 except that dosing was repeated 7 days later (arrows in A). (A) Tumor volume and (B) percentages of mice with tumor volume less than 3-times the pre-treatment volume. P values for statistically significant differences between gemcitabine only and the combination groups were determined with the LogRank test. (C) Days for tumor to reach 3-times pre-treatment volume versus body weight loss at nadir. Values represent mean ± SEM for 6–8 mice. Weight loss in drug treated groups was not statistically significant (one-way ANOVA).

Hypoxic cells spared by gemcitabine re-oxygenate and repopulate tumors, and are eliminated by SN30000 in a schedule-dependent manner

To explore the mechanism of schedule-dependent activity of SN30000 and gemcitabine, hypoxic and S-phase cells in HCT116 tumors were compared before and after the treatments by immunohistochemistry (Figure 6A; single colour fluorescence images for these tumors are shown in Figure S8 and for replicate tumors in Figure S9). Mice were dosed with PIMO 2 h before the commencement of drug treatments, then 45 h later with EF5 plus EdU to label S-phase cells. Three hours after dosing with EF5 and EdU, in control tumors (Figure 6A1) there were limited numbers of previously hypoxic (PIMO+) cells still evident at the edge of some necrotic regions (arrowhead) and many newly hypoxic (EF5+) cells, indicating that most of the originally hypoxic cells in untreated tumors were eliminated by necrosis within two days. A decreasing gradient in the frequency of S-phase (EdU+) cells was found from oxygenated (EF5-) regions to hypoxic (EF5+) regions in untreated tumors (Figure 6A1). After treatment with SN30000 the distribution of hypoxic and S-phase cells were similar to controls (Figure 6A2). Importantly, 2 days after gemcitabine treatment the population of originally hypoxic cells was larger than in untreated tumors (Figure 6A3). Furthermore, S-phase cells were present in the originally hypoxic regions but not in well-oxygenated regions, with many cells showing co-localization of PIMO and EdU (Figure 6A3), indicating re-population by cells that were hypoxic at the time of gemcitabine treatment. Administration of SN30000 before gemcitabine (Figure 6A4) and co-administration (Figure 6A5), but not SN30000 after gemcitabine (Figure 6A6), eliminated these originally hypoxic cells that would otherwise have been rescued by gemcitabine, and also broadly inhibited the proliferation of both oxygenated and hypoxic cells.

Figure 6.

S-phase and hypoxic fractions of tumors in response to SN30000 and gemcitabine. Mice bearing HCT116 tumors (mean diameter 9 mm) were dosed i.p. with PIMO (60 mg/kg, to label initially hypoxic cells) 2 h before the drug treatments as described in the legend of Figure 4. 45 h after PIMO injection, mice were dosed i.p. with a second hypoxia marker, EF5 (60 mg/kg, to label newly hypoxic cells) and EdU (50 mg/kg, to label S-phase cells) 3 h before tumor excision. Half of each tumor sample was used for immunohistochemistry (A). The remainder was dissociated to single cell suspensions for ex-vivo clonogenic survival assay (B) and flow cytometry (C-G). (A) Merged false-colour images for PIMO (green), EF5 (gold), EdU (red) with H33342 nuclear counterstain (blue). (B) Clonogens per gram of tumor. (C) % newly hypoxic cells (EF5+) at 45–48 h; (D) % originally hypoxic cells (PIMO+) still present at 48 h; (E) % S-phase in well-oxygenated cells (EdU+, EF5-); (F) originally hypoxic cells that have survived and re-oxygenated by 48 h (PIMO+, EF5-) as a percentage of total cells; (G) originally hypoxic cells that are in S-phase at 48 h (PIMO+, EdU+) as a percentage of total PIMO+ cells; Values are mean ± SEM for 6–8 mice. * P < 0.05, ** P < 0.01.

Quantitation of hypoxic and S-phase cells by flow cytometry (Figure 6B-G) confirmed the above interpretation of the immunohistochemistry images. We first compared total cells and clonogens recovered when tumors were dissociated after treatment, which showed no significant differences between groups either for total cells (data not shown) or clonogen numbers (Figure 6B) although there was a trend to lower clonogen numbers in all gemcitabine-treated groups. In addition, there was no statistically significant difference between the recently hypoxic fraction (EF5+) of control (27.7 ± 3.8%) and treated tumors at the time of termination, although with a trend to lower values in the treatment groups (Figure 6C). Consistent with the immunohistochemistry observations, the population of originally hypoxic cells (PIMO+) in control tumors was very low (Figure 6D), as in SN30000 treated tumors presumably because most of the SN30000-targeted cells (cells that were hypoxic at the time of treatment) are destined to die spontaneously by necrosis anyway when the oxic population is not suppressed. However, the PIMO+ fraction was substantially increased two days after gemcitabine treatment (Figure 6D), although gemcitabine suppressed the proliferation of oxygenated (EdU+, EF5-) cells (Figure 6E). Furthermore, these surviving hypoxic cells had re-oxygenated (PIMO+, EF5-, Figure 6F) and the proportion of these PIMO+ cells that were in S-phase (PIMO+, EdU+) was also increased (Figure 6G). Co-administration, and SN30000 before (but not after) gemcitabine, decreased the number of these originally hypoxic cells (Figure 6D), and prevented their reoxygenation (Figure 6F) and repopulation (Figure 6G). Of note, SN30000 before gemcitabine further potentiated inhibition of the proliferation of oxygenated cells by gemcitabine (Figure 6E, P < 0.05), suggesting that there may be a synergistic interaction of SN30000 and gemcitabine even against relatively well oxygenated cells. To test this possibility, we evaluated effects of both drugs on proliferation of HCT116 and SiHa cells in monolayer cultures under 5% O₂, which demonstrated that SN30000 before gemcitabine substantially inhibited monolayer growth compared with gemcitabine alone or the other schedules (Figure S10).

We hypothesized that the lack of hypoxic cell elimination by SN30000 when it was dosed 3 h after gemcitabine may be because of rapid, short-term reoxygenation of tumors after gemcitabine treatment as demonstrated above in spheroids (Figure S3). We therefore tested reoxygenation of xenografts with or without gemcitabine treatment. There was a large variation of hypoxic fraction in both control and gemcitabine treated tumors, with no significant difference in reoxygenation (PIMO+, EF5-) or change of hypoxic fraction (EF5+/PIMO+) 3 h after gemcitabine (Figure S11).

Discussion

In this study, we demonstrate that gemcitabine, at a dose that suppresses proliferation in oxic regions of HCT116 tumors, paradoxically rescues hypoxic cells that are otherwise fated to undergo necrosis (Figure 6). We further demonstrate that the HAP SN30000 prevents this process and, presumably through this mechanism, increases the antitumor activity of gemcitabine without enhancing host toxicity (Figure 5).

These findings extend a previous report that the cells that resume cycling after gemcitabine treatment of HCT116 tumors are located in regions distant from functional blood vessels.¹³ The latter study suggested that this apparent gemcitabine resistance may result from limited penetration of the drug as well as slow proliferation in hypoxic regions. Poor penetration of gemcitabine has also been inferred from the limited penetration of ³H-gemcitabine through MGH-U1 and EMT6 multilayers based on measurement of total radioactivity.³⁷ Here, we report direct evaluation of multilayer penetration of gemcitabine, based on a compound-specific analytical method for the first time, confirming severely limited diffusion through HCT116 multicellular layers in vitro (Figure 4D). The finding that penetration of gemcitabine, but not urea or mannitol, was faster at 32°C than 37°C strongly suggests that metabolic trapping is the major impediment. These data suggest that the maximum penetration distance of gemcitabine at 37°C is in the order of 150 μm when the concentration in the donor compartment (representing blood plasma) is 300 μM, and that penetration limitations as well as a low S-phase fraction underlie the resistance of hypoxic cells to gemcitabine in spheroids and tumors.

While these factors likely contribute to the sparing of hypoxic cells, the mechanism(s) enabling repopulation from these cells are not entirely clear. We observed a decrease in hypoxia after treatment of spheroids with gemcitabine (Figure S3), consistent with previous observations of suppressed EF5 binding in gemcitabine-treated HCT116 spheroids.³⁸ However, after gemcitabine treatment of tumor-bearing mice we did not detect significant reoxygenation of hypoxic cells after 3 h (Figure S11) or 48 h (Figure 6C), in agreement with an earlier study using PIMO staining.¹³ Despite this, tumor growth inhibition along with suppression of the S-phase fraction in oxic regions, must enhance nutrient and/or oxygen delivery sufficiently to support reoxygenation and recruitment of hypoxic cells.

Our focus on gemcitabine/SN30000 interactions in xenografts resulted from a spheroid screen with representatives of the major chemotherapeutic classes (DNA crosslinkers, antimetabolites, topoisomerase, and microtubule poisons). Detecting useful chemotherapy/HAP interactions in vitro is challenging because cooperation between killing of oxic and hypoxic cells requires that both populations are present simultaneously, which is an important feature of spheroid models. The SN30000 interactions with gemcitabine and doxorubicin were particularly evident using spheroid growth inhibition rather than clonogenic assay endpoints, suggesting that reversible cell growth effects contribute substantially to growth delay and/or that slow washout of the drugs³¹ when spheroids are not dissociated after drug exposure alters apparent activity.

A notable feature of the gemcitabine/SN30000 interaction in spheroids was its marked schedule dependence, with loss of activity when gemcitabine was administered 3 h before SN30000 (Figure 3) which likely reflects the above gemcitabine-induced reoxygenation. Even though suppression of EF5 binding was modest, given the low O₂ concentration required for 50% inhibition of metabolism of nitroaromatic compounds (K_O2 ~ 0.1 µM), a small change of hypoxic fraction measured by these hypoxia probes may reflect a larger change in regions of intermediate hypoxia where SN30000 (K_O2, 1–3 µM) is activated. ²⁴

Perhaps surprisingly, this same schedule dependence was seen in vivo, despite the complications of the hypothermic effect of SN30000 (which is also induced by tirapazamine³⁶) which is the probable cause of the decrease in gemcitabine clearance when the drugs are co-administered (Figure 4B). This drop in temperature likely also modifies intratumoral pharmacokinetics, with an increase in penetration distance of gemcitabine because of slower metabolic entrapment in cells (Figure 4D). While these pharmacokinetic changes will enhance gemcitabine exposure, increased oxygen penetration at the lower temperature (potentially compromising SN30000 activity) and decreased metabolic activation of the prodrugs (phosphorylation of gemcitabine by deoxycytidine kinase and bioreduction of SN30000) may compensate for these changes.

Although our study has not determined all aspects of these complex interactions, effects of SN30000 on cell-intrinsic sensitivity to gemcitabine are also likely, as illustrated by enhanced antiproliferative activity of uniformly oxygenated monolayers with simultaneous exposure to both drugs (Figure S10). Other studies have also demonstrated time-dependent interactions between HAPs and cytotoxic drugs, including a synergistic interaction between tirapazamine and cisplatin ascribed to inhibition of repair of cisplatin-DNA interstrand cross-links.³⁹ In the present study, simultaneous treatment with SN30000 and cisplatin also enhanced activity in monolayers and spheroids, but was not investigated further as this interaction was not enhanced by increased hypoxia in the spheroid model (Figure 3).

The most widely investigated HAP in clinical studies, evofosfamide (TH-302), has been shown in preclinical studies to block repopulation by hypoxic cells in MCF7 and PC3 tumors after treatment with doxorubicin or paclitaxel.¹¹ TH-302 is generally more active when dosed before cytotoxic drugs in xenograft models, although in the combination with gemcitabine there was no significant time dependence of activity against H460 tumors but enhanced host toxicity with co-administration.⁴⁰ In a randomized Phase II trial of gemcitabine plus TH-302 against advanced pancreatic cancer, i.v. gemcitabine was commenced 2 h after the HAP,¹⁹ although it is not clear whether this scheduling is optimal. Timing of drug sequencing in combination settings is often poorly controlled in the clinic, and is an aspect that requires further study.

In conclusion, the schedule-dependent activity of SN30000 in combination with gemcitabine in HCT116 xenografts reflects their activity in spheroid models. This study emphasizes the complexity of drug interactions in the tumor microenvironment, and the value of multicellular spheroids in dissecting aspects of these interactions – particularly with HAPs that offer potential spatial complementarity with cytotoxic drugs. In addition, the ability of SN30000 (and TH-302) to prevent reoxygenation and repopulation from hypoxic tumor cells spared by chemotherapy suggest the potential application of HAPs in overcoming this important mechanism of drug resistance.

Materials and methods

Drugs and reagents

SN30000 was synthesized as reported²⁶ and stored at −20℃. Stock solutions of SN30000, cisplatin (Sigma-Aldrich) and paclitaxel (Sigma-Aldrich) were dissolved in saline and stored at −80℃. Doxorubicin hydrochloride (Actavis Ltd) and gemcitabine (Sandoz) solutions were stored in the dark at 4℃ and room temperature respectively. Pimonidazole (PIMO, Hypoxyprobe Inc), pentafluoroetanidazole (EF5, a gift from the National Cancer Institute, U.S.A.) and 5-ethynyl-2-deoxyuridine (EdU, Abcam) solutions were freshly dissolved in saline. FITC-conjugated anti-PIMO antibody (Hypoxyprobe Inc), CY5-conjugated Elk 3.51 anti-EF5 antibody (Dr CJ Koch, University of Pennsylvania, PA) and EdU click chemistry azides (Alexa fluor 488 azide, Alexa fluor 647 azide, tetramethyrhodamine azide, all from Thermo Fisher Scientific) were stored at −20℃. 16% paraformaldehyde (Thermo Scientific) was diluted in phosphate-buffered saline to 4% for fixing cells.

Cell lines and drug treatments

Human colorectal adenocarcinoma HCT116 and human cervical SiHa cell lines were from American Type Culture Collection and were authenticated by short tandem repeat profiling. A puromycin-resistant HCT116 cell line constitutively expressing copGFP (HCT116/GFP) was generated by electroporation (Neon^TM Transfection System, Invitrogen) with an empty piggyBac-CMV-MCS-EF1α-GreenPuro expression vector (PB513B-1, System Biosciences), with selection of a clone retaining high expression after withdrawal of 2 µM puromycin. Cells were passaged as monolayers in alpha minimum essential media (αMEM) (Gibco, Life Technologies) with 5% heat-inactivated fetal calf serum (FCS, Morgate Biotech), and 2 µM puromycin for HCT116/GFP. Cultures were exposed to drugs under either 20% O₂/5% CO₂ (Forma Series II water jacketed CO₂ incubator, Thermo Electron Corporation, ‘standard incubator’) or 5% O₂/5% CO₂ (Whitley H45 HEPA HypOxystation, Don Whitley Scientific Ltd).

Spheroid culture and measurement of spheroid area and fluorescence intensity

10³ HCT116, HCT116/GFP or SiHa cells in 20 µL αMEM with 10% FCS, 10³ Units/mL penicillin and 10³ µg/mL streptomycin were seeded in Corning 7007 low-attachment round-bottom 96-well plates (Sigma-Aldrich) to form spheroids in a standard incubator. After 24 h, each well was supplemented with 180 µL of the same meium and cultured for another 3 days before drug treatment. Spheroids were cultured with replacement of 100 µL of medium every second day from day 4.

The fluorescence intensity of HCT116/GFP spheroids was measured on an Enspire multimode plate reader (PerkinElmer) with excitation and emission wavelengths of 482 nm and 502 nm, respectively. An ImageJ plugin²⁹ was used to estimate the area from the 2D projected bright field spheroid images captured by a JuLI^TM stage Real-Time Cell History Recorder (NanoEnTek Inc.) using a 4× objective. The area, converted to a single diameter, was used to calculate the volume of an equivalent sphere.

Clonogenic survival assays

Clonogenic survival assays were performed as previously described.⁴¹ Briefly, single cell suspensions prepared from monolayers, spheroids (by dissociation with 0.05% trypsin/EDTA) or tumors (as described below) were serially diluted, plated in 60 mm dishes with 4.5 mL αMEM with 5% FCS and penicillin/streptomycin and colonies of >50 cells were counted after 10 days to calculate plating efficiency (PE, number of colonies/cells plated). Surviving fraction (SF) was calculated as PE (treated)/PE (control).

HCT116 xenograft model and drug treatment

All animal experiments were approved by The University of Auckland Animal Ethics Committee. Tumor xenografts were grown subcutaneously in the dorsal flank of female NIH-III nude mice (NIH-Lyst^bg-J Foxn1^nu Btk^xid) by inoculating 5 × 10⁶ HCT116 cells. When tumors reached a mean diameter of 9 mm, mice were randomized for intraperitoneal (i.p.) dosing with saline, SN30000 alone (130 mg/kg), gemcitabine alone (100 mg/kg), SN30000 3hr before (SN-3hr-Gem) or after (Gem-3hr-SN) gemcitabine, or gemcitabine 5 min before SN30000 (Gem-5min-SN), with a second dose one week later. Tumor length (L) and width (W), measured with callipers, and body weights were determined 3 times weekly until tumors reached a mean diameter of 18 mm. Tumor volume was calculated as $\mathrm{V}=\frac{\pi }{6}\ast L^2\ast W$.

Flow cytometry

Tumors from mice treated with PIMO (60 mg/kg), drugs, EF5 (60 mg/kg) and EdU (60 mg/kg) were excised and bisected. One half was dissociated using pronase (2.5 mg/mL), collagenase (1 mg/mL) and DNAase I (0.2 mg/mL, all from Sigma-Aldrich), in αMEM containing 10% FCS and penicillin/streptomycin. Up to 10⁵ cells/P60 dish were assayed for clonogenic survivors. The remaining cells were fixed in 4% paraformaldehyde at 4°C. For flow cytometry, cells were centrifuged and incubated in blocking solution (PBS with 0.3% Tween 20, 0.04 g/mL thimerosal and 0.125 g/mL sodium azide (PBStt) plus 20% fat-free milk and 1.5% lipid-free ablumin) for 20 min. To dual stain for PIMO and EF5 binding, cells were incubated in PBStt containing 150 µg/mL FITC-conjugated anti-PIMO antibody at 4°C overnight, followed with 6 h incubation at 4℃ in PBStt with 50 µg/mL CY5-conjugated Elk 3.52 anti-EF5 antibody. To dual stain EdU and PIMO or EF5, cells were incubated in PBStt containing either 150 µg/mL FITC-conjugated anti-PIMO antibody or 50 µg/mL CY5-conjugated Elk 3.52 anti-EF5 antibody at 4℃ overnight. Subsequently, EdU was stained with 10 µM Alexa fluor 647 azide for PIMO-labeled samples or 10 µM Alexa fluor 488 azide for EF5-labeled samples using click chemistry according to the manufacturer’s instructions. An Accuri-6 flow cytometer (B.D. Biosciences, San Jose, CA) was used to quantify FITC or Alexa fluor 488 in single cells with excitation and emission wavelengths of 488 nm and 525 nm, and to measure CY5 or Alexa fluor 647 azide using excitation and emission wavelengths of 640 nm and 665 nm. Control tumors, from mice not treated with probes or drugs, were used for thresholding.

Immunohistochemical staining

The other half of each tumor was fixed in 10% neutral buffered formalin and embedded in paraffin for sectioning. Sections were deparaffinized and rehydrated before heating in 10 mM citrate buffer (pH 6.5) in a 2100 Retriever pressure cooker (Aptum Biologics Ltd) for 2 h for antigen retrieval, followed by blocking in TBS-T (H₂O with 24.2 mg/mL Trizma base, 80 mg/mL sodium chloride and 0.1% Tween-20, pH = 7.6) containing 10% goat serum at 4℃ for 1 h. The sections were then immersed in 100 µL TBS-T containing 5% goat serum and 75 µg/mL CY5 conjugated Elk 3.52 anti-EF5 antibody at 4℃ overnight, followed by 50 µg/mL FITC-conjugated anti-PIMO antibody at 4℃ for 6 h. The sections were then exposed to click reaction cocktails with 10 µM tetramethylrhodamine azide at room temperature for 30 min before being counterstained with 100 µL 8 µM Hoechst 33342 (Sigma-Aldrich) for 10 min and mounted with Prolong Diamond Antifade Mountant (Invitrogen). The images were captured on a Zeiss Axio Imager Z2 fully motorized microscope.

Measurement of body temperature in mice

Rectal temperature was measured at intervals following drug treatment of tumor-bearing NIH-III mice using a BAT-12 digital microprobe thermometer (Physitemp Instruments).

Plasma pharmacokinetics of SN30000 and gemcitabine in mice

After i.p. dosing, blood was collected, chilled on ice in EDTA tubes (Falcon) and centrifuged (3000 g, 5 min). Plasma was stored at −80℃ for quantitation of SN30000 and gemcitabine by High-Performance Liquid Chromatography (HPLC).

Multicellular layer transport assay

HCT116 multicellular layers were cultured as described previously.⁴² Three-day-old MCLs were transferred to diffusion chambers⁴³ in a water bath at 37℃ or 32℃ and equilibrated with 95% O₂/5% CO₂ for 1 h in αMEM without FCS. Gemcitabine (final concentration 300 μM), together with ¹⁴C-urea (2.11 GBq/mmol) and ³H-mannitol (740 GBq/mmol, both from American Radiolabeled chemicals Inc.) as internal standards, were added to the donor compartment. The donor and receiver compartments were sampled (100 μL) at intervals. Radioactivity was measured in 25 μL samples with a liquid scintillation analyzer (Tri-Carb 2910TR, Perkin Elmer Inc.) to determine ³H-Mannitol and ¹⁴C-urea to confirm MCL integrity, as described previously.⁴² The remaining samples were frozen at −80℃ for determination of gemcitabine by high performance liquid chromatography (HPLC).

High performance liquid chromatography

Samples (50 µL) were mixed with 100 µL acetonitrile, centrifuged (16000 g, 3 min, 4℃). Supernatants were mixed 1:1 with 0.1% formic acid, and analyzed by HPLC as previously described for SN30000 and its 1-oxide metabolite.²⁹ Gemcitabine (retention time 6.9 min) was analyzed using a Hypercarb column (2.1 × 150 mm, 5 µm particle size) by photodiode array detection at 274 nm with a mobile phase of 0.1% formic acid in a linear gradient of acetonitrile/water over 10 min. Calibration curves for SN30000, its 1-oxide metabolite and gemcitabine were determined in each analysis under the same sample-handling conditions.

Funding Statement

This work was supported by grants from the Marsden Fund of the Royal Society of New Zealand [13-UOA-187, to KOH], the University of Auckland Faculty Research Development Fund [3706812 to KOH], the Health Research Council of New Zealand [14/538, to WRW] and China Scholarship Council to XM [Doctoral Scholarship].

Disclosure of Potential Conflicts of Interest

The authors declare no potential conflicts of interest.

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website.

Abbreviations

AUC	Area under the concentration-time curve
EdU	5-ethynyl-2ʹ-deoxyuridine
EF5	pentafluoroetanidazole
HAP	hypoxia-activated prodrug
PIMO	pimonidazole