Preclinical Benefit of Hypoxia-Activated Intra-arterial Therapy with Evofosfamide in Liver Cancer Free

Hypoxia is one of the most common and frequent physiologic alteration of solid tumors, including hepatocellular carcinoma, and has been strongly correlated with disease progression and resistance to chemotherapy (1, 2). In addition, tumor hypoxia has been shown to promote the emergence of a more aggressive and metastatic cancer phenotype (1, 3) which demonstrates increased resistance to apoptosis (4), stimulates angiogenesis (1), and favors cancer growth by altering cell metabolism (5). Overall, tumor hypoxia contributes to a poor clinical prognosis (6). As a result, a variety of hypoxia-targeted therapies have been developed including hypoxia-inducible factor 1 (HIF-1α) targeting, hypoxia-selective gene therapy, genetic engineering of anaerobic bacteria and hypoxia-activated prodrugs (HAP; ref. 2).

Among these different hypoxia-targeted strategies, HAPs are unique in that they are selectively activated under low oxygenation conditions, and remain as nontoxic prodrug during normoxic or aerobic conditions. This distinctive characteristic renders HAPs as potentially very effective in the setting of transarterial chemoembolization (TACE). Indeed, the embolic effect of TACE increases the hypoxic microenvironment of solid tumors and provides an attractive setting for selective activation of bioreductive prodrugs under low tissue oxygen state. Moreover, TACE allows for the locoregional delivery of high-doses of chemotherapy that have the potential to reach distal tumor regions where hypoxic cells reside in a pharmacologic sanctuary (7, 8). Furthermore, the use of HAPs in TACE could help mitigate detrimental effects triggered by the embolization with the activation of HIF-1α–dependent pathways, such as subsequent gene overexpression of VEGF (9, 10) and hypoxia-induced chemoresistance (11).

Evofosfamide (formerly called TH-302), is a 2-nitroimidazole triggered HAP of the cytotoxin bromo-isophosphoramide mustard (12). Evofosfamide has shown broad-spectrum anticancer efficacy in multiple human cell lines and xenograft models (13–18) and is currently undergoing clinical evaluation (19–21).

The purpose of this study was to evaluate the safety and efficacy of hypoxia-activated intra-arterial therapy (HAIAT) with evofosfamide in an orthotopic rabbit model of liver cancer.

Adult female New Zealand white rabbits (Millbrook Breeding Labs) weighing 3.9–5.7 kg were used in accordance with institutional guidelines under approved animal care and use committee protocols from the Johns Hopkins University (Baltimore, MD). The rabbit was chosen as it is considered the animal model of choice for intra-arterial therapies to the liver (22). Food and water was allowed ad libitum. VX2 tumor chunks from previous experiments were injected into the hind legs of carrier rabbit and tumor was grown for 2 weeks. Each carrier was used to supply recipients for tumor implantation into the left liver lobe of the experimental rabbits as detailed in previous publication (23). The tumors were allowed to grow in the livers for 13–14 days (24).

Animals were randomized into 4 hepatic intra-arterial treatment regimens and received an infusion of either: (i) normal saline (5 mL of 0.9% NaCl; control group; n = 7); (ii) evofosfamide (Evo group; n = 7); (iii) an emulsion of doxorubicin (Sigma Aldrich) and lipiodol (Lipiodol Ultra-Fluide, Laboratoire Guerbet) followed by embolization with 100–300 μm bland beads [Embospheres, Merit Medical Systems; conventional TACE (cTACE) group; n = 7); (iv) a combination of doxorubicin/lipiodol emulsion and evofosfamide followed by embolization with 100–300 μm bland beads (cTACE + Evo group; n = 7). Animals underwent multidetector computed tomography (MDCT) 24 hours pretreatment and then at 7 and 14 days after the interventional treatment. Comprehensive blood work (complete liver, hematologic, and renal panels) was obtained at baseline, 1, 2, 7, and 14 days after treatment. Daily clinical monitoring was performed. Three animals in each group were randomly sacrificed 2 days after treatment for histopathologic and immunohistochemical analysis. The rest of the animals were sacrificed on day 14 after the last imaging assessment.

Evofosfamide was formulated in saline at 5 mg/mL and filtered through a 0.2-μm filter before animal dosing. Evofosfamide was prepared by study members who did not participate in the embolization procedures. A treatment dose of 7.7 mg/kg was used on the basis of the FDA guidelines for dose translation based on body surface area and previous studies (14, 19–21, 25). Doxorubicin powder was reconstituted with saline immediately before intra-arterial injection, for a final concentration of 10 mg/mL. This dose calculation was based on the most common clinically administered doses of doxorubicin in the setting of cTACE (50 mg, for a ∼70 kg person; ref. 26). The doxorubicin solution was mixed to obtain a homogenous clear red solution. In a separate syringe, an equal-to-doxorubicin-solution volume of lipiodol was withdrawn. The two solutions were mixed according to the push-and-pull method using a metallic stopcock. Emulsion (0.6–0.8 mL; total doxorubicin dose: 3–4 mg) was administered (cTACE and cTACE + Evo groups) followed by a volume of up to 0.1–0.2 mL of 100–300 μm beads mixed with an equal volume of nonionic contrast medium (Visipaque 320, GE Healthcare). In the cTACE + Evo group, the emulsion was injected following a sequential protocol. First a small volume of the emulsion (0.1–0.2 mL) was injected and pushed with saline to occlude distal tumor vessels and significantly slow down the arterial circulation. This was followed by a slow infusion of evofosfamide over a 10-minute period after what the rest of the emulsion was administered completed by bland embolization. A sequential protocol was chosen to increase tumor hypoxic microenvironment due to distal occlusion of the tumor feeding arteries to enhance evofosfamide activation and lead to more efficient targeting of tumor hypoxia. Moreover, this sandwich injection technique that involves embolization before and after evofosfamide infusion may allow for a better activation of evofosfamide that is trapped in the embolized tissue to enhance antitumor efficacy. At the time of the embolization procedure, once the microcatheter was in the treatment position, the interventional radiologists were given the embolization material. At that time, the interventional radiologists could guess the actual treatment that was administered on the basis of visual assessment (e.g., saline vs. lipiodol + doxorubicin) and based on injection protocols. The interventional radiologists could not guess the actual treatment based on visual assessment or injection protocols when saline or evofosfamide alone were administered (both saline and evofosfamide alone are translucent).

The interventional procedure was performed as reported previously (23, 24). Briefly, animals were preanesthetized with inhalant isoflurane (5%) on oxygen using a cone mask and anesthesia was then maintained via inhalant isoflurane (1.5%–3%) on oxygen. Surgical cut down was done to gain access into the common femoral artery followed by the placement of a 3-French vascular sheath (Cook, Inc.). A 2.1/1.7-French microcatheter (Echelon 10, ev3 Endovascular, Inc.) was manipulated into the celiac axis, after which a celiac arteriogram was done to delineate the blood supply to the liver. The left hepatic artery was then selectively catheterized; this was occasionally performed with the aid of a steerable guide wire (0.014 in. Transcend wire; Boston Scientific Oncology). The tumor was visualized on digital subtraction angiography as a region of hypervascular blush located in the left lobe of the liver. With the microcatheter in this position, the different treatment regimens were carefully administered under real-time fluoroscopy to prevent nontarget delivery. The endpoint of embolization was absence of forward flow (for cTACE and cTACE + Evo groups). Upon completion of the treatment, the microcatheter was removed, the common femoral artery was ligated, and the surgical cut down was closed in one layer using absorbable suture material. Analgesic buprenorphine (0.02–0.05 mg/kg) was administered intramuscularly.

Blood samples were collected from a marginal ear vein at baseline, 1, 2, 7, and 14 days after treatment to determine liver function. Potential hematologic (white blood cell, neutrophil, lymphocyte, monocyte, eosinophil, and basophil counts and complete erythrocyte panel) or renal (creatinine, blood urea nitrogen, Na⁺, K⁺, Cl⁻) toxicities were evaluated as well using standard enzymatic reaction.

Preanesthetized rabbits were scanned at baseline (24 hours before treatment), and at 7 and 14 days following treatment. Images were obtained with a 320-detector-row cardiac MDCT scanner (Aquillon One, Toshiba). As the coverage in the Z-direction was of 16 cm (320-detector-rows) the upper abdomen of the animals that includes the whole liver could be acquired with one gantry rotation with the table of the scanner remaining static. The large field of view and the fast image acquisition (scan time of 350 msec/360°) eliminates reconstruction artifacts allowing for high image quality (27). Unenhanced and contrast-enhanced scans were performed by intravenous injection of 1.5 mL/kg isoosmolar contrast medium (Isovue-370) at 1 mL/second followed by a saline flush. Other imaging parameters were: 120 kVp, 350 mA, 0.5 mm slice thickness, 0.5 mm collimator, FOV 22 × 22 cm.

A semiautomatic 3D quantitative software (Medisys, Philips Research) was used to segment the tumors on the arterial phase of the MDCT scans and obtain the entire tumor volumes. The arterial phase was chosen due to the known hypervascular pattern of the tumors. This software uses non-Euclidean geometry and theory of radial basis functions which allows for highly precise segmentation of 3D objects with multiple shapes as described previously (ref. 28; Supplementary Fig. S1).

At 2 days or at the study endpoint (14 days after treatment), animals were sacrificed under deep anesthesia by intravenous injection of thiopental (100 mg/kg). To investigate the significance of the hypoxic fraction, pimonidazole (Hypoxyprobe, Inc.) was intraperitoneally injected at 50 mg/kg two hours before sacrifice in 3 rabbits per group at 2 and 14 days after the interventional treatment (29). Rabbit livers were immediately harvested. The liver with the tumor was cut into 3-mm thick slices. Peritumoral treated and contralateral untreated liver parenchyma was also collected. Tissues were immediately placed in 10% buffered formalin. Subsequent complete necropsy (except brain) was done on all animals. Formalin-fixed tissue was paraffin-embedded and sectioned at 4 μm. Tissues slices were stained with hematoxylin and eosin (H&E) or processed for IHC. IHC was performed on the specimens collected 2 days after treatment as early time points have shown to give optimal biomarkers expression to evaluate response to therapy (30, 31).

Tumor sections were evaluated for the following targets: HIF-1α (Thermo Scientific, Inc.), VEGF (Dako, Agilent Technologies, Inc.), γ-H2AX (Santa Cruz Biotechnology, Inc.), Annexin V (Santa Cruz Biotechnology), caspase-3 (Santa Cruz Biotechnology), Ki-67 (Dako), glyceraldehyde-3-phospahte dehydrogenase (GAPDH; Santa Cruz Biotechnology), monocarboxylate transporter 4 (MCT4; Santa Cruz Biotechnology), lactate dehydrogenase (LDH; Santa Cruz Biotechnology), and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) using the TUNEL Apoptosis Detection Kit (Millipore). Specimens were deparaffinized using xylene and rehydrated using a descending ethanol dilution series. After washing with deionized water, samples were permeabilized in boiling retrieval solution containing citrate (Dako) for 40 minutes at 95°C. Specimens were cooled down to room temperature and incubated with 100 μL of Peroxidase Quenching Solution (Invitrogen) for 5 minutes, and incubated with 100 μL of blocking solution (Invitrogen) for 20 minutes. Incubation with primary antibodies (HIF-1α, 1:50; VEGF, 1:100; γ-H2AX, 1:50; Annexin V, 1:50; caspase-3, 1:50; Ki-67, 1:200; GAPDH, 1:500; MCT4, 1:50; LDH, 1:50; in PBS) occurred at room temperature in a moist chamber for 45–60 minutes. Mouse anti-goat [Santa Cruz Biotechnology; or goat anti-mouse (Bio-Rad) where so appropriate] IgG-HRP–conjugated secondary antibody was added to the samples for 30 minutes in a moist chamber. 3,3′-diaminobenzidine (DAB; 26.5 μL) chromogen was mixed well in the dark with 1 mL of DAB substrate buffer and 100 μL were added to each specimen for 5 minutes. Hematoxylin was used as a counterstain. Incubation steps were followed by washing with distilled water and twice with PBS for 2 minutes each. Samples were sealed using antifade mounting agent (ProLong Gold Antifade, Life Technologies) and covered with a coverslip.

All histologic slides (H&E, pimonidazole, and other molecular targets) were digitized at 20× magnification using an Aperio ScanScope AT slide scanner (Leica Biosystems Inc.). Manual segmentation and image analysis was performed using the Aperio ImageScope viewer.

H&E–stained slides were used for tumor morphology and quantification of tumor necrosis. Pathologic tumor necrosis was quantified in the whole tumor. An experienced liver pathologist (C.F. Brayton) who was blinded to all clinical and radiologic data manually outlined in contiguous slides the whole tumor area and area of necrosis for each whole-slide image (Fig. 1A; ref. 32). The necrotic fraction of each tumor was calculated as the percentage of the tumor necrotic area in the entire tumor.

To obtain objective and quantitative immunohistochemical analysis of tumor tissues, a color deconvolution image analysis algorithm was used as described in previous works (33–36). Briefly, the Color Deconvolution v9 and ImageScope software (Aperio) were used to individually calibrate each staining in the disease-relevant areas (tumor tissues) that had been manually annotated. This process involved iterative adjustments of the deconvolution algorithm with various intensity thresholds for positivity. The resultant colormap was then compared with stained tissues and validated by an experienced pathologist (T.C. Cornish) who was blinded to all clinical and radiologic data to obtain concordance (Fig. 1B). The metrics used for staining quantification were the percentage of positive pixels and the total tumor area, which was defined as the cumulative total area of combined positive and negative pixels. The hypoxic fraction (HF) was calculated in the viable tumor tissue with exclusion of tumor necrotic areas (HF = pimonidazole-positive area/viable tumor area). The hypoxic compartment (HC) was defined as the percentage of pimonidazole-positive area in the whole tumor area (16). HIF-1α, VEGF, γ-H2A.X, Annexin V, caspase-3, GAPDH, MCT4, LDH positivity was evaluated in the whole tumor area. The Aperio nuclear V9 algorithm was used to quantitatively identify Ki-67–positive cells as described previously (33, 37).

Confocal fluorescence microscopy (LSM 700, Zeiss; EC Plan-Neofluar 40x/1.3 NA Oil lens DIC M27) was performed at 40× magnification for the TUNEL-stained tissue sections. A total of 5–10 fields were captured per section. Semiquantitative image analysis was performed using the Bio-Formats Importer plugin of ImageJ (NIH, Bethesda, MD).

Targeted and contralateral untreated liver specimens were evaluated on H&E staining by an experienced pathologist (C.F. Brayton). The severity of liver injury, if any, was evaluated using a score adapted from Suzuki and colleagues (38) and Kleiner and colleagues (39). The scoring system comprised of 8 histologic features that were assessed semiquantitatively: hepatocellular necrosis (0–4), vacuolization (0–4), ballooning (0–2), lobular inflammation (0–4), steatosis (0–3), fibrosis (0–4), biliary hyperplasia (0–4), and granulation (0–4).

Animals were randomly assigned to treatment groups using randomization online software (GraphPad Software). Experiment data were expressed as mean ± SEM. Measurements were compared among all four experimental groups using the Kruskal–Wallis test. For this and the other tests, significance levels were calculated, when possible, with exact methods based on the permutation distribution of the treatment labels, as these do not rely on normal approximations, to increase validity of the findings. Specifically, the slopes of the trajectories of log tumor volume over time (exponential growth) were estimated by Generalized Estimating Equations (40), and were compared across experimental groups using the exact distribution of the estimated slopes. The correlation of HF and HC with response to therapy across treatment groups was evaluated with the Pearson correlation coefficient r, and its significance was assessed in the distribution of r obtained by permuting the order of one of the two variables being correlated. Measurements between treated and untreated lobes across animals within treatment groups were compared with the exact distribution of the t test obtained by permuting the treated and untreated values within each animal. A two-tailed P value of less than 0.05 was considered statistically significant. A power analysis determined the sample size n = 7/group was sufficient to detect with high probability the substantial treatment effects we had aimed to see in the study. For example, when comparing the volume trajectories for the 4 rabbits with trajectory data in cTACE to the 4 rabbits with trajectory in cTACE + Evo, the average slopes between the two groups differed by 4.4 SDs (effect size 4.4), and the power for detecting such difference with the exact test was greater than 97%. The power was still > 80% for effect sizes > 3. Statistical analysis was performed with two software packages (R: A Language and Environment for Statistical Computing, R Core Team, R Foundation for Statistical Computing, 2013 and GraphPad Prism version 6.00, GraphPad Software).

The safety profile of evofosfamide was determined by clinical evaluation, body weight measurements, serial blood work, and complete necropsy. Two animals from the control group were moribund at the study endpoint due to high tumor burden. Animals tolerated evofosfamide well with no apparent drug-induced toxicity determined by daily clinical assessment. There was no significant difference in body weight change among the groups. All treatment groups significantly increased plasma alanine aminotransferase and aspartate aminotransferase compared with vehicle control. The cTACE + Evo group had a similar profile in liver enzyme elevation compared with cTACE, although it was higher and more sustained. However, this elevation was only statistically significant compared with cTACE at day 7 for alanine aminotransferase. All the tested liver enzymes returned to normal by day 14. Total bilirubin levels were slightly higher with cTACE + Evo but not statistically significant. The synthetic function of the liver remained unchanged throughout the experiment as shown by albumin levels (Supplementary Fig. S2). Similar elevations of white blood cells were observed in all groups following the interventional procedure. Neither hematologic toxicity nor abnormal renal function was observed when evofosfamide (7.7 mg/kg) was intra-arterially delivered either alone or in combination with cTACE (data not shown). As a target organ, the liver exhibited morphologic change by H&E staining and hypoxia change by pimonidazole staining. In normal nontumoral liver, there was a gradient of hypoxia radiating outwards from portal triads, but after the embolization treatment (cTACE and cTACE + Evo groups), hypoxic extent and intensity was dramatically increased (Supplementary Fig. S3A). Besides the targeted liver, there was no other organ abnormalities noted during necropsy (Supplementary Fig. S3B).

The extent of hepatocellular damage was investigated with a dedicated liver tissue scoring system. Evofosfamide and cTACE demonstrated similar toxicity when analyzing the liver parenchyma around the tumors, supporting the fact that evofosfamide is activated in less oxygenated tumor areas and its released effector diffuses locoregionally beyond hypoxic areas, consistent with a bystander effect. Most importantly, similar scores were obtained between cTACE and those of cTACE + Evo. However upon analyzing the subgroup of animals at day 14, lobular inflammation was more pronounced in the combination therapy (average 2.5 points higher; P = 0.028; Table 1).

The anticancer efficacy of evofosfamide was evaluated using quantitative volumetric image analysis. Compared with the control group, evofosfamide-treated tumors were significantly smaller at 7 and 14 days (4.6 ± 0.7 vs. 1.5 ± 0.5 cm³ and 15.9 ± 3.2 vs. 2.8 ± 0.9 cm³, respectively). At day 14, both chemoembolization groups showed smaller volumes compared with the evofosfamide group. Importantly, cTACE + Evo demonstrated significantly smaller volumes than cTACE alone (0.31 ± 0.08 vs. 0.8 ± 0.07 cm³, respectively; Fig. 2A). At day 14, evofosfamide achieved significantly reduced tumor growth rate compared with control and similar tumor growth rate to cTACE (Fig. 2B). cTACE + Evo outperformed the other treatment groups by achieving not only significantly lower tumor growth rate but also tumor shrinkage (Fig. 2B). By permutation-based linear regression assuming dependent observations, tumor growth slopes between cTACE (+0.04) and cTACE + Evo (−0.11) were significantly different (P = 0.027). cTACE + Evo achieved a reduction in tumor volume of 50% by day 9, whereas the calculated tumor volume doubling time in the cTACE group was 25 days (Fig. 2C).

Pathologic tumor necrosis was quantified by slide-by-slide histosegmentation to investigate the entire tumors. When compared with vehicle, histopathologic analysis performed at day 2 showed increased tumor necrosis in the treatment groups (87.5%±12.1%, 97.5%±1.1% and 97.5%±0.8% with evofosfamide, cTACE, cTACE + Evo, respectively; Fig. 3A). At day 14, the combination therapy with cTACE + Evo demonstrated significantly more necrotic fraction compared with vehicle, evofosfamide alone, and cTACE (Fig. 3B).

Tumor hypoxic regions were detected by pimonidazole staining and quantified by color deconvolution algorithm. The HF and the HC changed significantly post intra-arterial therapy (R² = 76%, P = 0.002 and R² = 95%, P = 0.005, respectively; Fig. 4). To further characterize this significance, we examined which subgrouping showed the next highest R². Tumors treated with cTACE + Evo had a significantly lower HF compared with evofosfamide alone and cTACE groups (P = 0.037; Fig. 4B). Regarding the magnitude of hypoxia in the whole tumor, evofosfamide alone and cTACE + Evo demonstrated a significantly lower HC compared with the vehicle (P = 0.037) and a clear trend compared with cTACE (P = 0.07), indicating antitumor efficacy of hypoxia targeting by evofosfamide (Fig. 4C). To further investigate the relationship between tumor hypoxia and treatment response, a correlation analysis was performed using exact permutations distribution to increase the validity of the findings. A significant negative correlation was found across treatment groups between the HF and the magnitude of the necrotic fraction at day 2 (Pearson r = −0.753; P = 0.004). As well, the decrease in the HC showed a strong negative correlation with tumor necrosis (Pearson r = −0.826; P = 0.001).

At day 14 post intra-arterial therapy, no difference in the HF was noted (P = 0.11; Fig. 4E). However, the HC was significantly different between the groups (R² = 81%, P = 0.026). The HC of cTACE + Evo remained significantly lower compared with the other groups. Interestingly, the evofosfamide group showed a significantly higher HC compared with the other groups (Fig. 4F).

The phosphorylation of the histone H2A.X (γ-H2A.X) variant was investigated to evaluate whether evofosfamide produces DNA crosslinking damage. γ-H2AX was significantly upregulated in the tumors recovered at 2 days after therapy (R² = 85%, P < 0.001). The chemoembolization groups (cTACE and cTACE + Evo) demonstrated a significant increase expression of γ-H2AX compared with control and Evo groups. This expression was significantly higher for cTACE + Evo compared with the evofosfamide and cTACE groups (Fig. 5B). Similarly, a marked increase of the apoptotic biomarkers Annexin V and caspase-3 was achieved by evofosfamide and to a higher extent by the chemoembolization groups. The combination therapy demonstrated a significant increase in caspase-3 expression compared with the evofosfamide and cTACE groups (Fig. 5C and D). Importantly, all tumors in the cTACE + Evo group showed higher expression of γ-H2AX, Annexin V, and caspase-3 compared with the tumors treated with cTACE. Following treatment with evofosfamide and cTACE + Evo, expression of TUNEL was increased significantly above control levels (Fig. 5E). Compared with cTACE, cTACE + Evo showed more advanced apoptotic stage with nuclear fragmentation resulting in semiquantitative decreased pixel intensity (Fig. 5E). Cell proliferation decreased in all treatment groups compared with controls. This reduction in cell proliferation was significantly more pronounced in tumors treated with evofosfamide and cTACE + Evo compared with control and more importantly compared with cTACE alone (Fig. 5F).

The upregulation of HIF1α was significantly increased in the chemoembolization groups compared with controls indicating increase in the hypoxic microenvironment resulting from TACE. To investigate tumor glycolysis in the setting of HAIAT with procedure induced ischemia/hypoxia, key indicators of glycolysis such as LDH, GAPDH, and, the lactate transporter, MCT4 (41) were evaluated. Marked increase in the expression of MCT4 and LDH were observed in the cTACE and cTACE + Evo groups (Fig. 6).

The main finding of this study is that hepatic HAIAT with evofosfamide as a single agent and more importantly in combination with cTACE achieved strong anticancer effects in the rabbit tumor liver model with minimal toxicity. Our data demonstrate a significant correlation between hypoxia-targeting and antitumor activity further supporting the anticancer potential of the locoregional delivery of the HAP technology.

Tumor hypoxia constitutes one of the main Achilles heel of anticancer therapy. Thus, the development of therapeutic strategies such as HAPs to target tumor hypoxia is of utmost clinical relevance and importance. Evofosfamide has demonstrated anticancer efficacy in vitro and in xenograft models (13–18) and is currently undergoing clinical evaluation as a systemic chemotherapy agent (19–21). Here we investigated the locoregional delivery of evofosfamide in the setting of TACE by characterizing its safety and toxicity profile and quantifying its antitumor activity using different mechanistic approaches.

In clinical practice, determination of liver function after locoregional therapy is critical for treatment toxicity evaluation and patient selection to identify candidates who would benefit from repeated therapy. Evofosfamide alone showed a slight increase in liver enzymes with rapid return to baseline within 2 days of therapy. As expected, a transient increase of liver enzymes was observed in the chemoembolization groups reaching a peak 1–2 days after treatment. cTACE + Evo showed a similar, although more pronounced, rise in liver enzymes when compared with cTACE. However, only alanine aminotransferase levels were significantly higher at day 7. Moreover, the histologic toxicity scores were similar between cTACE + Evo and cTACE alone, with the exception of lobular inflammation being more pronounced at day 14 in the combination therapy group. In the clinic, a majority of TACE procedures are performed in patients with liver cirrhosis. Although the rabbit constitutes the animal model of choice to investigate intra-arterial therapies to the liver, baseline liver function is normal and no cirrhosis is present (22). Thus, the impact on liver function of the addition of evofosfamide to cTACE in the setting of liver cirrhosis is yet to be assessed. Of note, no hematologic toxicity was observed, whereas such toxicity constitutes a known limitation of systemic therapy with evofosfamide (19–21). Taken together, these results demonstrate that evofosfamide alone and in combination with cTACE has a favorable toxicity profile with no compromise of liver metabolic and synthetic functions.

Using quantitative imaging and histopathologic analysis of the whole tumor, we demonstrated that cTACE + Evo had a stronger anticancer efficacy than cTACE alone. cTACE was able to achieve a good local control of the tumors with low tumor volumes and tumor growth rate. However the addition of evofosfamide to cTACE enhanced antitumor efficacy. Specifically, cTACE + Evo outperformed the other treatment groups by not only resulting in significantly smaller tumor volumes and lower tumor growth rates but also actual tumor shrinkage. Of note, cTACE + Evo achieved consistent results in all treated animals at all observed time points when compared with cTACE and evofosfamide alone, as illustrated by the consistently high necrotic fraction. This added efficacy of the combination therapy over cTACE alone could be beneficial in patients to achieve better local tumor control rate and improve current outcomes. In addition, the correlation analysis demonstrated a significant negative correlation between the necrotic fraction and the magnitude of the HF and the HC, further validating the antitumor activity of evofosfamide and underlining the relationship between tumor hypoxia targeting and treatment response. Interestingly, the hypoxic environment of control tumors decreased over time. It is possible that the development of tumor vessels was unable to meet the perfusion demand imposed by a rapidly growing and expanding tumor such as the VX2. As a result, oxygen and nutrients deprivation induced extensive tumor necrotic areas. An unexpected result was observed in the evofosfamide group with increased HC at day 14. This may be explained by evofosfamide potentially interfering with HIF-1α–dependent pathways such as VEGF-related angiogenesis with subsequent decreased vessel recruitment (42). In addition, once activated, evofosfamide may have locoregionally diffused and damaged distal tumor vessels contributing to a subsequent increase in the hypoxic tumor microenvironment (43). While these findings deserve further investigations, they also evidence an opportunity for repeated therapy.

To further characterize the antitumor efficacy of evofosfamide, we quantitatively evaluated the expression of γ-H2AX, Annexin V, caspase-3, TUNEL, and Ki-67. The ability of evofosfamide to induce cross-linking DNA damage has also been reported in a murine model with the prodrug being delivered systemically (16). The expression of γ-H2AX was mainly seen near hypoxic areas. Our data demonstrated that evofosfamide alone or in combination therapy with Lipiodol/doxorubicin-based chemoembolization not only achieved strong antitumor activity in tumor hypoxia but also in normoxic tumor regions located close to blood vessels. These findings suggest strong activation of the prodrug and diffusion to nearby tumor tissues consistent with a bystander effect (15, 44). When evofosfamide was given alone, this phenomenon was probably the main trigger for cell death while selective catheterization with decreased forward blood flow during drug delivery could have enhanced the activation of the bioreductive prodrug. Moreover, the generation of superoxides by evofosfamide could have contributed to part of the observed γ-H2AX expression. Indeed, Meng and colleagues (15) demonstrated that superoxide is generated under normoxic conditions by the redox cycling of evofosfamide through its one-electron reduction to the radical anion with subsequent reoxidization by oxygen back to the original prodrug configuration (15). The authors showed that a 100X-higher concentration of evofosfamide was necessary under normoxia compared with hypoxic conditions to achieve similar H2AX expression (15). As intra-arterially delivered evofosfamide reaches much higher concentration than evofosfamide given systemically, this could have contributed in part to the γ-H2AX expression in our study. Furthermore, similar to γ-H2AX, evofosfamide also induced the expression of apoptotic markers with a concomitant decrease in cell proliferation substantiating the efficacy of locoregional delivery.

Iatrogenic increase of hypoxic tumor microenvironment (i.e., by the embolization), as demonstrated here by HIF-1α upregulation and also previously reported (10, 45–49), together with increased reliance on anaerobic glycolysis by tumor cells (50), as shown by increased expression of MCT4 and LDH, provide a mechanistic explanation of the added anticancer efficacy of cTACE + Evo over cTACE and further validate TACE as an ideal setting for selective targeting of tumor hypoxia. Increased expression of HIF-1α and VEGF observed in the evofosfamide group could potentially be explained by decreased forward blood flow around the microcatheter during superselective catheterization with consecutive increase in tumor hypoxia.

Anticancer drugs must access and diffuse into the extravascular space to reach the tumor microenvironment and cancer cells. To achieve efficacy, therapeutic drug concentrations must be reached. Doxorubicin is the most widely used cytotoxic drug in TACE. When given systemically, doxorubicin has a poor penetration through cancer tissue due to its avid binding to DNA and sequestration in endosomes by perivascular cells (51, 52). However, intra-arterial administration of doxorubicin with TACE demonstrated improved drug penetration into tumor tissue compared with systemic therapy (53). Evofosfamide showed the ability to enhance the activity of commonly used systemic chemotherapeutics agents, notably doxorubicin, in xenograft models (14, 44). In patients with soft tissue sarcoma, doxorubicin combined with evofosfamide achieved a better outcome than doxorubicin alone (19). Here we demonstrated that evofosfamide combined with cTACE achieved better anticancer efficacy than cTACE supporting the added benefit of combining doxorubicin and evofosfamide as reported in preclinical and clinical studies (14, 19, 44).

Because cancer cells are located at greater distances from the microvasculature than normal cells, the efficacy of hypoxic cytotoxins may be decreased by their limited diffusion into the extravascular tumor tissue to reach the hypoxic cancer cells (8, 54). Several hypoxia-activated prodrugs have progressed to clinical trials; however, none has yet been granted approval for clinical use, reflecting some challenges in the use of these novel drugs. Indeed hypoxia-independent cytotoxicity, low diffusion, and short plasma half-life are some of the limitations of the bioreductive drug technology (54–56). Evofosfamide has shown an optimized profile with good drug penetrability and hypoxia-selective cytotoxicity with no evidence of hypoxia-independent targeting (15).

Taken together, our preclinical data support the hypothesis that tumor hypoxia can be an exploitable target for liver tumors by HAIAT. Hepatic HAIAT holds promise by several methodologic strengths: (i) it allows for locoregional delivery of high concentration of the hypoxia-activated prodrug that could otherwise not be achieved by systemic delivery, potentially overcoming excessive metabolic consumption [diffusion-limited (chronic) hypoxia; ref. 54]; (ii) ischemia induced by the therapeutic embolization offers an ideal mechanistic setting for hypoxia-activated prodrug; (iii) the tumor hypoxic microenvironment can be modified perprocedure by selective catheterization and sequential injection protocols like those used in this study or using devices such as balloon occlusion catheters in future research (e.g., temporary occlusion of the tumor feeding artery before drug delivery to intentionally increase tumor hypoxia) which potentially could enhance drug efficacy [perfusion-limited (acute) hypoxia] and lead to more efficient targeting of hypoxia.

In conclusion, preclinical benefit of HAIAT with evofosfamide was observed in liver cancer. HAIAT is a promising step toward locoregional targeting of tumor hypoxia. The favorable toxicity profile and strong anticancer effects of evofosfamide in combination with cTACE pave the way towards clinical trials in patient with liver cancer.

J. Chapiro reports receiving commercial research grants from Philips Healthcare. C.P. Hart holds ownership interest (including patents) in Threshold Pharmaceuticals. J.-F. Geschwind reports receiving commercial research grants from Boston Scientific, BTG, Guerbet Healthcare, Philips Healthcare, and Threshold Pharmaceuticals; and is a consultant/advisory board member for Boston Scientific, BTG, Guerbet Healthcare, Philips Healthcare, Prescience, Terumo, and Threshold Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: R. Duran, J. Chapiro, M. Lin, J.D. Sun, C.P. Hart

Development of methodology: R. Duran, S. Mirpour, V. Pekurovsky, T.C. Cornish, J. Chapiro, M. Lin, J.D. Sun

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Duran, S. Mirpour, V. Pekurovsky, C.F. Brayton, T.C. Cornish, B. Gorodetski, J. Chapiro, R.E. Schernthaner, M. Lin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Duran, S. Ganapathy-Kanniappan, C.F. Brayton, T.C. Cornish, B. Gorodetski, J. Chapiro, C. Frangakis, M. Lin, J.D. Sun

Writing, review, and/or revision of the manuscript: R. Duran, V. Pekurovsky, S. Ganapathy-Kanniappan, C.F. Brayton, T.C. Cornish, B. Gorodetski, J. Chapiro, R.E. Schernthaner, C. Frangakis, M. Lin, J.D. Sun, C.P. Hart

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Duran, V. Pekurovsky, S. Ganapathy-Kanniappan, C.F. Brayton, B. Gorodetski, R.E. Schernthaner, M. Lin

Study supervision: R. Duran, S. Ganapathy-Kanniappan, M. Lin, J.-F. Geschwind

Other (surgical procedures (tumor implantation): J. Reyes

This study was funded by NIH/NCI R01 CA160771 and Threshold Pharmaceuticals, Inc.

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.