Metabolic and Physiologic Imaging Biomarkers of the Tumor Microenvironment Predict Treatment Outcome with Radiation or a Hypoxia-Activated Prodrug in Mice

Pancreatic ductal adenocarcinoma (PDAC), which constitutes 95% of pancreatic cancers, is one of the most aggressive and fatal cancers, with a 5-year survival rate of 5% (1). As PDAC is so insidious, more than 80% of PDAC cases are already advanced and metastatic when diagnosed, beyond the stage capable of surgical resection (2).

PDAC is commonly classified as a hypovascular and hypoxic tumor (3). However, quantitative and noninvasive methods of evaluating oxygenation of this cancer are limited. The only study, to the best of our knowledge, of quantitative oxygen measurement of pancreatic tumors was reported by Koong and colleagues using the invasive Eppendorf electrode method (4). The partial oxygen pressure (pO₂) in tumors across the 7 patients in this study ranged from 0 to 5.3 mm Hg, which is significantly hypoxic compared with the values for a normal pancreas (9.3–92.7 mm Hg; ref. 4). On the other hand, substantial interpatient heterogeneity of the extent of tumor hypoxia was observed. Matsuo and colleagues reported the results of a larger-scale, but not quantitative, study of oxygen levels in PDAC (5). In this study, expression of a hypoxia marker, hypoxia-inducible factor 1α (HIF1α), in PDAC tissues was measured histochemically, and nearly half of the tumors showed nondetectable or low levels of HIF1α expression. The HIF1α level was significantly (P = 0.035) correlated with poor prognosis. Metran-Nascente and colleagues reported striking interpatient variation in the hypoxic fraction of tumors in patients with pancreatic cancer using a 2-nitroimidazole-based PET tracer, ¹⁸F-fluoroazomycin arabinoside (6). In their study, the hypoxic fraction ranged widely, from <5% to >50%, and 5 of 20 evaluable patients did not exhibit any detectable tumor hypoxia.

These reports suggest that not all human pancreatic tumors are hypoxic. Tumor response to treatment in individual patients and the resulting overall survival rate can vary, depending either directly on a tumor's oxygenation status or indirectly on the downstream hypoxia signaling induced by the characteristic hypoxic and acidotic microenvironments. A priori knowledge of these microenvironmental features allows the appropriate choice between antiproliferative therapies, such as chemotherapeutic, radiation-based modalities for well-oxygenated tumors, and hypoxia-activated prodrugs for hypoxic tumors.

Electron paramagnetic resonance imaging (EPRI) is a spectroscopic imaging technique similar to MRI, which is available in preclinical study but not yet applied for humans. EPRI detects paramagnetic species that have unpaired electrons such as transition metal complexes and free radicals. With the availability of trityl radical probes as in vivo–compatible paramagnetic tracers, EPRI is being explored for mapping tissue oxygen levels in live animals (7, 8). The collisional interaction between the oxygen-sensitive tracer and dissolved paramagnetic oxygen leads to a broadening of the spectral line width of the tracer, which linearly correlates with local oxygen concentration, providing quantitative imaging capability of EPRI in determining tissue pO₂. EPRI has capabilities to provide quantitative maps of tumor pO₂ in preclinical studies (7, 8), and being evaluated for clinical implementation. On the other hand, hyperpolarized ¹³C MRI monitoring of pyruvate metabolism via oxidative phosphorylation or by aerobic glycolysis is a novel metabolic imaging technique that can potentially provide biochemical information related to tumor oxygenation and is already clinically implemented in humans (9). Cancer cells within hypoxic tumor regions survive by promoting aerobic glycolysis for energy production, where pyruvate is metabolized to lactate by lactate dehydrogenase (LDH). Although the underlying molecular mechanisms of the shift in energy metabolism are complex, diverse, and dependent on the kinds of genomic mutations in individual tumors, the resulting alteration of pyruvate metabolism offers opportunities for noninvasively characterizing tumor physiology and predicting tumor response to particular treatments.

We recently reported that three PDAC xenografts (Hs766t, MiaPaCa2, and Su.86.86) showed quite different response to a hypoxia-activated prodrug TH-302 and the different treatment efficacy was attributed to the difference in tumor oxygenation status, which can be artificially modified by a transient hypoxia induction agent (10). Based on the previous observation, we investigated if two noninvasive imaging biomarkers, pO₂ measured by EPRI and [1-¹³C] pyruvate metabolism rate measured by hyperpolarized ¹³C MRI, could capture the difference in oxygenation status in three human PDAC xenografts and predict their response to three representative pancreatic cancer treatments, including a first-line chemotherapy, a hypoxia-activated prodrug, and fractionated radiotherapy.

Female athymic nude mice were supplied by the Frederick Cancer Research Center, Animal Production (Frederick, MD). Three human PDAC cell lines were used for tumor model development. Su.86.86 and Hs766t cells were obtained from Threshold Pharmaceuticals, and MiaPaCa2 cells were obtained from the ATCC. All cell lines were tested in May 2013 and authenticated by IDEXX RADIL using a panel of microsatellite markers. PDAC xenografts were established by subcutaneously injecting 2 × 10⁶ cells into the hind legs of mice. Tumor volumes were measured every 2 to 4 days by T₂-weighted MR images and calculated with the formula: (length) × (width) × (height) × π/6.

Tumor-bearing mice were treated with either (i) 150 mg/kg of gemcitabine (LC Laboratories) i.p. on days 0 and 2, (ii) fractionated X-irradiation 3 Gy × 5 days (X-RAD 320, Precision X-ray Inc.), or (iii) 80 mg/kg of TH-302 (Threshold Pharmaceuticals) i.p. daily for 5 days. In combination therapies, chemotherapeutic drugs, same doses as monotherapy, were given 30 minutes after X-irradiation, or TH-302 was given 30 minutes after gemcitabine administration. The survival data for control and TH-302 groups in survival analysis were duplicates of our previously published study (10).

Technical details of the EPR scanner and oxygen image reconstruction were described previously (11). Parallel coil resonators tuned to 300 MHz were used for EPRI. After the animal was placed in the resonator, an oxygen-sensitive paramagnetic trityl radical probe, OX063 (1.125 mmol/kg bolus), was injected intravenously under isoflurane anesthesia. The free induction decay (FID) signals were collected following the radiofrequency excitation pulses (65 ns) with a nested looping of the x, y, and z gradients, and each time point in the FID underwent phase modulation, enabling 3D spatial encoding. Because FIDs last for 1 to 5 μs, it is possible to generate a series of T₂* maps, i.e., EPR line width maps, which linearly correlate with the local concentration of oxygen if the concentration of Ox063 is low enough to avoid the contribution of self line broadening, and allow pixel-wise estimation of tissue pO₂. The repetition time was 8.0 μs. The number of averages was 4,000. After EPRI measurement, anatomic T₂-weighted MR images were collected with a 7T or 1T scanner (Bruker BioSpin MRI GmbH). Because pyruvate injection transiently affects tumor oxygenation, EPRI measurement is always conducted prior to ¹³C MRI study of hyperpolarized pyruvate metabolism when both scans were run on the same day.

Details of the hyperpolarization procedure were reported previously (12, 13). Briefly, samples of [1-¹³C] pyruvic acid (30 μL) containing 15 mmol/L Ox063 and 2.5 mmol/L of the gadolinium chelate ProHance (Bracco Diagnostics) were polarized in a Hypersense DNP polarizer (Oxford Instruments). After 40 to 60 minutes, the hyperpolarized sample was rapidly dissolved in 4.5 mL of a superheated HEPES-based alkaline buffer. NaOH was added to the dissolution buffer to pH 7.4 after mixture with [1-¹³C] pyruvic acid. Hyperpolarized [1-¹³C] pyruvate solution (12 μL/g body weight) with a ¹³C polarization of 20% to 25% was intravenously injected through a catheter placed in the tail vein of each mouse.

Hyperpolarized ¹³C MRI studies were performed on a 4.7T scanner (Bruker Bio-Spin MRI GmbH) using a 17-mm home-built ¹³C solenoid coil placed inside of a saddle coil tuned to ¹H frequency. Dynamic spectroscopic images were acquired every 6 seconds after the start of pyruvate injection from a 32 × 32 mm field of view in a 8-mm coronal slice through the tumor, with matrix size of 12 × 12, spectral width of 6010 Hz, repetition time of 41.6 ms, and a 250-μs Gaussian excitation pulse with a flip angle of 10°. The total time required to acquire 40 serial images was 4 minutes.

Tumor tissues were excised 1 hour after intravenous injection of pimonidazole (60 mg/kg, Hypoxyprobe Inc.) and fixed with 4% paraformaldehyde overnight. The tumor tissues were frozen with Tissue-Tek O.C.T. compound (Sakura Finetek USA Inc.) in cold ethanol and 10-μm thick sections were obtained. The tissue sections were stained for pimonidazole (rabbit, Hypoxyprobe-1 Omni Kit, Hypoxyprobe Inc.), CD31 (rabbit, NBP1-49805, Novus Biologicals), carbonic anhydrase IX (rabbit, NB100-417, Novus Biologicals), and LDH-A (rabbit, NBP1-48336, Novus Biologicals). Horseradish peroxidase-conjugated secondary antibodies and 3,3′ diaminobenzidine were used to detect primary antibodies. The stained slides were scanned using a BZ-9000 microscope (Keyence), and the immunostain-positive area was quantified using ImageJ software (downloaded from https://imagej.nih.gov/ij/).

When tumor size reached >600 mm³, tumor tissues were excised, immediately frozen in liquid nitrogen, and stored at 80°C until analysis. A total of 116 metabolites involved in glycolysis, the pentose phosphate pathway, the tricarboxylic acid cycle, the urea cycle, and polyamine, creatine, purine, glutathione, nicotinamide, choline, and amino acid metabolism were analyzed using CE-TOF and QqQ mass spectrometry (Carcinoscope Package, Human Metabolome Technologies, Inc.).

All of the results are expressed as the means ± SD. The differences in the means of groups were determined by the Student two-tailed t test. The survival rate was estimated with the Kaplan–Meier survival analysis, where survival rate was defined as tumor volume increasing by <2.5-fold from the original size. The differences between control group and other groups are assessed by the log-rank test. The minimum level of significance was set at P <0.05.

All of the animal experiments were performed in compliance with the Guide for the Care and Use of Laboratory Animal Resources (National Research Council, 1996) and were approved by the National Cancer Institute Animal Care and Use Committee (NCI-CCR-ACUC, Bethesda, MD, protocol nos. RBB-153, -155, and -159).

Three xenografts of human PDAC (Hs766t, MiaPaCa2, and Su.86.86) were resected for histochemical analysis to investigate if there is any difference in the levels of markers related to tumor oxygenation and metabolism in these tumors. Representative images from two different tumors in each tumor type are shown in Fig. 1. Tumor vascularization was evaluated with immunostaining of vascular marker CD31. The fraction of CD31-positive area in Su.86.86 tumors (22.5% ± 8.5%) was higher than that of Hs766t (16.4% ± 7.1%) and MiaPaCa2 (14.4% ± 12.0%) tumors, suggesting that Su.86.86 tumors are well vascularized compared with the other two PDAC tumors. Immunostaining of an exogenous hypoxia marker, pimonidazole (34.5% ± 7.9%, 20.9% ± 10.6%, and 15.9% ± 8.2% for Hs766t, MiaPaCa2, and Su.86.86, respectively), and endogenous hypoxia markers CA-IX (34.1% ± 15.1%, 33.1% ± 18.7%, and 13.9% ± 7.3%) and LDH-A (65.6% ± 13.8%, 55.0% ± 18.8%, and 12.2% ± 5.9%), whose expression is controlled by HIFs, also suggested that Hs766t and MiaPaCa2 tumors were more hypoxic than Su.86.86 tumors (Fig. 1).

In order to further investigate the metabolic differences in the three PDAC tumors, mass spectroscopy–based metabolome analysis was conducted. Among 116 metabolites detected, the concentrations of representative metabolites related to energy metabolism are shown in Fig. 2. Concentrations of 3-phosphoglycerate and pyruvate were highest in Hs766t (15.2 ± 4.2 and 42.0 ± 6.7 nmol/g, respectively, expressed as mean ± SEM), intermediate in MiaPaCa2, (4.3 ± 1.5 and 26.0 ± 8.6 nmol/g) and lowest in Su.86.86 tumors (2.4 ± 0.4 and 18.0 ± 2.8 nmol/g), implying an increased glucose uptake in Hs766t tumors. Higher adenylate energy charge in Hs766t tumors compared with other two tumor lines can also be explained by the upregulated glycolysis in Hs766t tumors. Citrate concentration and NADH/NAD ratio were higher in Su.86.86 tumors (593 ± 55 nmol/g and 0.049 ± 0.0069, respectively) compared with Hs766t (411 ± 41 nmol/g and 0.020 ± 0.0011) and MiaPaCa2 (380 ± 12 nmol/g and 0.029 ± 0.0035) tumors, whereas lactate concentration was higher in Hs766t and MiaPaCa2 tumors than in Su.86.86 tumors (14,400 ± 860, 10,300 ± 2890, and 6,250 ± 504 mmol/g, respectively). These results suggest that mitochondrial oxidative phosphorylation is downregulated and pyruvate is metabolized to generate more lactate in Hs766t and MiaPaCa2 tumors. Lower concentrations of glutamate and total glutathione in Hs766t tumors are consistent with the decreased metabolic flux of the tricarboxylic cycle. Interestingly, substantially decreased concentrations of fumarate and malate were detected in MiaPaCa2 tumors, whereas no difference was observed in upstream succinate levels among the three PDCA tumors, suggesting impairment of succinate dehydrogenase (SDH) activity in MiaPaCa2 tumors. Immunoblotting of SDH isoforms revealed decreased SDH expression in MiaPaCa2 tumors (Supplementary Fig. S1), which may contribute to upregulated glycolysis.

To investigate the tumor oxygenation levels among the three pancreatic tumors examined, EPR oxygen imaging studies were conducted. T₂-weighted MR images of the centers of tumors and corresponding 2D slices extracted from 3D EPR oxygen image data set of the PDAC tumors are shown in Fig. 3A and B, respectively. Tumor areas were determined using the T₂-weighted MR images. Quantitative oxygen analysis was performed on the whole tumor area in the 3D oxygen images and histograms of pO₂ distribution of representative tumors are shown in Fig. 3C. The median pO₂ values of the tumors studied were 9.1 ± 1.7, 11.1 ± 2.2, and 17.6 ± 2.6 mm Hg, and the corresponding hypoxic tumor fractions (defined as <10 mmHg of oxygen) were 57.7% ± 13.3%, 46.6% ± 7.4%, and 27.3% ± 13.5% for Hs766t, MiaPaCa2, and Su.86.86 tumors, respectively (Fig. 3D and E). Supplementary Fig. S2 shows the representative images of whole mount immunostaining of pimonidazole. The results of quantitative oxygen imaging were consistent with the results of immunostaining and the metabolome analysis, confirming that the PDAC xenografts studied have significant variation in the level of tumor oxygenation. Median pO₂ and HF10 of individual PDAC tumors in EPR oxygen imaging study were listed in Supplementary Table S1.

To better understand the metabolic profile of the three PDAC tumor xenograft models, hyperpolarized ¹³C MRI using [1-13C] pyruvate was performed. After the tail-vain injection of hyperpolarized [1-¹³C] pyruvate (0.1 mg/g body weight), pyruvate-to-lactate conversion in the tumor region was observed in ¹³C MRI scans acquired every 6 seconds (Fig. 4A). The pyruvate and lactate signals were plotted, and the conversion rate constant (k_pyr-lac) was calculated based on the plot. Figure 4B shows how the conversion rate constant k_pyr-lac varies across the tumor area and between tumor types. From the data, it can be seen that Hs766t tumors showed the highest k_pyr-lac whereas Su.86.86 tumors showed the lowest. As expected, the k_pyr-lac was negatively correlated to the tumor pO₂ levels (Fig. 4C). Lactate to pyruvate ratio and k_pyr-lac of individual PDAC tumors in hyperpolarized ¹³C MRI study were listed in Supplementary Table S2.

To further examine the relationship between the oxygenation/aerobic glycolysis profiles and therapeutic effect of anticancer therapies, treatment responses to monotherapies and combination therapies of gemcitabine, TH-302, and fractionated radiotherapy on the three PDAC xenograft models were examined in vivo by measuring the tumor size after the treatments. The data for control and TH-302 monotherapy groups in survival analysis were from our previously published study (10). The doubling time of untreated control tumors were 7.0 ± 1.0, 8.0 ± 2.0, and 8.0 ± 1.2 days for Hs766t, MiaPaca2, and Su.86.86 tumors, respectively. TH-302 was used as a hypoxia-targeting therapy and fractionated radiotherapy was used as an oxygen-dependent therapy. Gemcitabine was also examined as a first-line drug for PDAC treatment in clinical settings. The results are presented as Kaplan–Meier survival curves, in which an endpoint was set at a tumor volume of 2.5-fold increase from the original size. Figure 5 shows the survival curves of the tumor-bearing mice treated with the monotherapies for three PDAC tumors. Although all tumors had similar survival time without treatment, the responses to the three treatments were significantly different. Supplementary Figure S3 shows the plots of tumor growth delay versus HF10, pO₂, and lac/pyr ratio.

Table 1 reports the tumor growth delay produced by monotherapies and combination therapies. Fractionated radiotherapy improved the survival time the most in oxygenated Su.86.86 tumors (18.0 ± 2.7 days), while TH-302 was most effective in hypoxic Hs766t tumors (25.0 ± 7.7 days). Gemcitabine displayed treatment responses unrelated to the tumor oxygenation. Hs766t tumors were most sensitive to gemcitabine (20.0 ± 3.5 days) and other two tumors also showed moderate treatment responses (2.7 ± 0.4 and 4.7 ± 0.4 days in MiaPaca2 and Su.86.86 tumors, respectively). Because the Hs755t tumor exhibited the most hypoxia among the three tumor types evaluated (see Fig. 3), we conducted an in vitro experiment to determine if Hs766t cells were exceptionally sensitive to gemcitabine under hypoxic conditions (<10 ppm oxygen) compared with aerobic conditions. The study revealed a 1.8-fold enhancement in cell killing from gemcitabine under hypoxic treatment compared with aerobic conditions (Supplementary Fig. S4). The reason for the Hs766t tumor cell enhancement in vitro and in vivo will require further study.

Figure 6 and Supplementary Table S3 show the synergistic effect of combination therapies calculated by the following equation: growth delay (A+B)/[{growth delay (A)} + {growth delay (B)}]. All combinations of the three therapies showed more than additive effects, presumably due to their different mechanisms of action. The combination of oxygen-dependent radiotherapy and hypoxia-targeting TH-302 therapy was effective in Hs766t and MiaPaCa2 tumors that were relatively hypoxic, while less hypoxic Su.86.86 tumors showed less treatment benefit, suggesting TH-302 was not effective against Su.86.86 tumors, even when combined with radiotherapy, which is known to induce temporary hypoxia but the timing and duration of radiation induced transient hypoxia is largely different dependent on the dose of radiation and tumor types (14–16). The combination therapies containing gemcitabine treatment showed unexpected results. When combined with gemcitabine, oxygen-dependent radiotherapy showed the strongest synergistic effect on the most hypoxic tumor (Hs766t), suggesting a strong radiosensitizing effect by gemcitabine.

In the present study, we used two novel molecular imaging techniques, EPR-based pO₂ imaging and hyperpolarized ¹³C MRI, to characterize the tumor microenvironment and examine the ability of physiologic and metabolic status to reliably predict the response to three cancer treatment modalities. EPRI can provide a quantitative pO₂ measurement in vivo. As the treatment outcome of radiotherapy or HAPs is known to be closely related to pO₂ in the tumor, oximetry by EPRI is considered to be a useful imaging assessment tool to predict the response to such treatments. However, due to the limitations of EPRI such as specific absorption rate and lower penetration depth of microwave, it has not been applied in clinical setting. On the other hand, ¹³C MRI of pyruvate is a clinically available imaging technique that provides in vivo metabolic information. Exogenously administered hyperpolarized pyruvate can be delivered into cancer cells where the conversion of pyruvate to either lactate by the enzyme LDH, or acetyl-CoA and downstream intermediates through oxidative phosphorylation, generates characteristic spectral signatures. The lactate:pyruvate ratio (L/P) indicates local LDH activity that is usually enhanced in cancer cells and can also be enhanced in hypoxic environments. Thus, monitoring the L/P ratio can be considered as a means to examining indirectly the pO₂ status of the tumor.

The three PDAC xenograft models were first examined using histologic and metabolomic analysis. Histologic analysis revealed that Hs766t tumors were hypoxic, hypovascular, and glycolytic, whereas Su.86.86 tumors were less hypoxic, vascular rich, and less glycolytic. MiaPaCa2 tumors showed an intermediate profile (Fig. 1). Metabolomic analyses confirmed the glycolytic profiling of these tumors by showing elevated levels of steady-state metabolites of consistent with glycolysis in Hs766t and MiaPaca2 tumors (Fig. 2). Such characterization of PDACs was also noninvasively performed by EPR oximetry and hyperpolarized ¹³C MRI of pyruvate. The oxygenation profiles of the three PDAC models obtained from EPR oximetry were consistent with the pimonidazole-positive area in histologic analysis (Figs. 1C and 3B and E), and the metabolic profiles obtained from k_pyr-lac in hyperpolarized ¹³C MRI (Fig. 4C) were also consistent with the LDH-positive area in histologic analysis (Fig. 1D). The negative correlation between pO₂ and glycolytic efficiency in PDACs revealed by these imaging modalities suggests a metabolic adaptation represented by activation of HIFs that are known to regulate the expression of more than 800 genes. However, the contribution from anaerobic glycolysis to the lactate:pyruvate ratio cannot be separated. LDH A and pyruvate dehydrogenase kinase 1 are known to be upregulated by HIF and enhance pyruvate to lactate flux.

The tumor growth experiments showed that the most hypoxic tumor type, Hs766t, showed the longest growth delay (25 ± 7.7 days) upon treatment with the hypoxia-activated prodrug TH-302 and shortest delay (6.3 ± 2.7 days) after radiation, whereas the least hypoxic tumor type, Su.86.86, showed the shortest delay (0.7 ± 0.6 days) with TH-302 treatment and the longest delay (18.0 ± 2.7 days) after radiation (Fig. 5; Table 1). This indicates that metabolic profiling by hyperpolarized ¹³C MRI as well as tumor oximetry by EPR imaging can potentially serve molecular imaging modalities to predict treatment efficacy of radiation and TH-302 therapies. Gemcitabine monotherapy unexpectedly showed the highest treatment response in Hs766t, the most hypoxic tumor type, although hypoxic tumors are, in general, considered to poorly respond to antiproliferative chemotherapies (17, 18). Previous reports have identified proline and glutamine as negatively correlated markers and aspartate, hydroxyproline, creatine, and creatinine as positively correlated markers for gemcitabine resistance in two other human cancer cell lines (19). In our metabolomic data, the levels of such markers were similar between MiaPaCa2 tumors and Hs766t tumors and did not predict the differential response to gemcitabine. Further investigations are required to establish whether alterations of metabolites are observed in gemcitabine-resistant PDACs.

Gemcitabine is an antimetabolite agent that is widely used to treat PDAC tumors. Gemcitabine is a nucleoside analogue in which the hydrogen atoms on the 2′ carbon of deoxycytidine are replaced by fluorine atoms. Gemcitabine is known to masquerade as a cytidine and get incorporated into new DNA strands, leading to apoptotic cell death. Radiotherapy is known to cause single- or double-strand breaks in DNA and generate reactive oxygen species (ROS) that inhibit DNA repair. Thus, oxygenated regions in tumors are susceptible to radiotherapy and hypoxic regions are not. On the other hand, TH-302 was developed to selectively target these less perfused hypoxic regions. TH-302 contains a nitroimidazole moiety covalently linked to a bromo-isophosphoramide (Br-IPM) moiety; it undergoes 1-electron reduction followed by unimolecular fragmentation to release the DNA alkylating Br-IPM at low concentrations of oxygen (20). In vitro studies showed a ∼1,000-fold lower IC₅₀ for cellular cytotoxicity in cells under hypoxia than cells under normoxic conditions (21).

The treatment responses to combination therapy were also compared with the imaging results. It should be noted that the nitroimidazole moiety of TH-302 is well known to enhance cancer cells to radiotherapy (10, 20) and gemcitabine is reported to enhance the effect of radiotherapy by inhibiting homologous recombination repair even at a low dose (21). As reported in several previous studies (20, 22, 23), the combination of radiation and TH-302 resulted in better outcomes than monotherapy in all three PDAC tumor types, presumably by targeting both hypoxic and nonhypoxic regions. The radiosensitizing effect of TH-302 might also potentiate the antitumor effects. All gemcitabine-containing combination regimens showed more than additive effects. Interestingly, a higher synergistic effect of radiation + gemcitabine combination therapy was observed in hypoxic Hs766t tumors than in less hypoxic Su.86.86 tumors (Supplementary Table S1). This was thought to be the result of the radiosensitizing effect of gemcitabine in Hs766t tumors (21), which is supported by the fact that this combination therapy was not as synergistic in Su.86.86 tumors as in Hs766t tumors. Su.86.86 tumors might be too inherently radiosensitive to benefit from radiosensitization effects of gemcitabine. This result suggested that radiotherapy can be a powerful adjuvant therapy, even for radiation-resistant tumors with large hypoxic regions, when added to gemcitabine treatment. Of the three combination therapies investigated, only the combination of gemcitabine and TH-302 showed large synergistic effects in all three PDAC tumor types. This supports the results of a clinical trial investigating this combination therapy and reinforces the rationale for it (24).

The results of in vivo tumor growth experiments indicated the potential use of metabolic imaging by hyperpolarized ¹³C MRI to assess hypoxia (25). Of note, we measured pyruvate metabolism in a single slice through the center of tumor to improve time resolution and accurately track the kinetic profile. Multisliced imaging may be preferable if taking into account the possible heterogeneity of tumor physiology. The enhanced glycolysis can potentially predict the oxygenation profile of a cancer and its response to oxygen-dependent or hypoxia-targeting treatments. This will expand the clinical application of metabolic profiling by hyperpolarized ¹³C MRI and lead to better treatment strategies. Several anticancer therapies were reported to modulate the metabolic profile of tumors, leading to a more accurate assessment of therapeutic effects. Further investigation is required to determine whether examining data from pretreated and posttreated cancers leads to even more accurate treatment selection.

In this study, we characterized the tumor microenvironment of three different PDAC xenografts in mice on a metabolic and physiologic basis and obtained imaging biomarkers that were predictive of the differential response to various treatments. The imaging results were consistent with the metabolomic and histologic information obtained from these three xenografts. The imaging data provided useful insight into predicting the treatment efficacy of gemcitabine, TH-302, and radiation, both as monotherapies and combination therapies, in three types of PDAC tumor xenograft models.

R.J. Gillies has ownership interest (including patents) in and is a consultant/advisory board member for HealthMyne Inc. and Helix Biopharma. C.P. Hart is a stockholder of Molecular Templates, the current developer of TH-302. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Matsumoto, C.P. Hart, R.J. Gillies, J.B. Mitchell, M.C. Krishna

Development of methodology: S. Matsumoto, R.J. Gillies, M.C. Krishna

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Matsumoto, K. Saita, Y. Takakusagi, J.P. Munasinghe, N. Devasahayam

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Matsumoto, S. Kishimoto, Y. Takakusagi, R.J. Gillies

Writing, review, and/or revision of the manuscript: S. Matsumoto, S. Kishimoto, C.P. Hart, R.J. Gillies, J.B. Mitchell, M.C. Krishna

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Kishimoto

Study supervision: M.C. Krishna

This study was mainly supported by the intramural research program of the NCI/NIH (S. Matsumoto, S. Kishimoto, K. Saito, Y. Takakusagi, N. Devasahayam, J.B. Mitchell, and M.C. Krishna). This study was partially supported by JST PREST (S. Matsumoto), the Toyota Physical and Chemical Research Institute (S. Matsumoto), and the Takeda Science Foundation (S. Matsumoto). We thank Melissa Stauffer for editing the manuscript.

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.