Selective Killing of Hypoxia-Inducible Factor-1–Active Cells Improves Survival in a Mouse Model of Invasive and Metastatic Pancreatic Cancer

Pancreatic cancer is one of the most lethal solid tumors of the gastrointestinal tract. The high mortality rate of pancreatic cancer is due to the high incidence of metastatic disease at the time of diagnosis, a fulminant clinical course, and the lack of adequate systemic therapies. Peritoneal and extrapancreatic seeding is observed during end-stage pancreatic cancer. Pancreatic cancer is characterized by extensive invasion into surrounding tissues and metastasis to distant organs, even at an early stage. The prognosis of patients is poor (1) and new systemic agents are urgently needed.

Intratumoral hypoxia is a major factor contributing to cancer progression (1–4). This situation is exacerbated by the high oxygen demands of rapidly proliferating tumor cells, poor lymphatic drainage resulting in high interstitial pressure, and shunting of blood through immature tumor vasculature (3). The evidence for hypoxia in pancreatic cancers includes their characteristic avascular appearance on computed tomography (5) and low intratumoral oxygen tension measurements (6). Furthermore, the expression of target genes of hypoxia-inducible factor-1 (HIF-1), the major transcription factor activated under hypoxic conditions, is elevated in pancreatic cancers (2, 7–11). They include proangiogenic factors such as vascular endothelial growth factor (VEGF), growth/survival factors such as insulin-like growth factor, and extracellular matrix remodeling proteins such as metalloproteinase-9 (12). Thus, HIF-1 activity may play a crucial role in malignant progression of pancreatic cancers.

Orthotopic implantation of human pancreatic cancer cell line SUIT-2 has been reported as a clinically relevant model showing an aggressive and malignant phenotype with liver metastasis, peritoneal dissemination, and ascites (13, 14). Although pathologic studies have shown the progression of pancreatic cancer in this model, the contribution of hypoxic/HIF-1–active tumor cells to the metastatic process has not been evaluated.

HIF-1 is a transcription factor composed of HIF-1α and HIF-1β subunits. Under normoxic conditions, HIF-1α protein is degraded by ubiquitination and proteasomal degradation following hydroxylation of proline residues in the oxygen-dependent degradation (ODD) domain (15). Under hypoxic conditions, HIF-1α protein accumulates and translocates to the nucleus where it heterodimerizes with HIF-1β to form an active transcription factor. More than 60 putative direct HIF-1 target genes have been identified and many of them are implicated in angiogenesis, invasion, and metastasis (12). Furthermore, HIF-1 has been recognized as a major resistance factor for anticancer therapies such as chemotherapy (16–19) and radiotherapy (20–22). Thus, HIF-1 activity is a hallmark of tumor malignancy.

We hypothesized, first, that the invasion and metastasis of pancreatic cancer is associated with hypoxia and increased HIF-1 activity and, second, that therapy targeting HIF-1–active cells would impair malignant progression. We developed and validated a fusion protein, TAT-ODD-procaspase-3 (TOP3), to specifically eradicate HIF-1–active hypoxic cells (20, 23–28). TOP3 consists of three domains. The TAT protein transduction domain (PTD) efficiently delivers the fusion protein into cells in vivo (29). The ODD domain contains a VHL-mediated protein destruction motif from the human HIF-1α protein that confers hypoxia-dependent stabilization to the fusion protein (25). The procaspase-3 domain is an inactive proenzyme form of human caspase-3 that is activated by preexisting apoptotic signals in hypoxic cells.

Here, we report a newly developed fusion protein, POP33, which has a novel PTD that mediates significantly higher protein transduction activity than TAT-PTD. We examined the activity of PP33 in a clinically relevant orthotopic model of pancreatic cancer. We serially monitored HIF-1–active cancer cells with a bioluminescence photon counting device and showed their involvement in local invasion and metastasis. POP33 significantly increased caspase-3 activity, suppressed peritoneal dissemination, and prolonged survival. These results indicate that HIF-1 activity plays a crucial role in malignant progression and support the use of drugs targeting HIF-1 for the treatment of pancreatic cancer.

Cell culture and isolation of SUIT-2/EF-Luc and SUIT-2/HRE-Luc cells. The human pancreatic cancer cell line SUIT-2 was purchased from the American Type Culture Collection. SUIT-2/EF-Luc and SUIT-2/HRE-Luc cells were isolated from a colony formed after transfection with plasmid pEF/Luc and pEF/HRE-Luc, respectively (24), by calcium phosphate method (30) and antibiotic selection. The cells were maintained at 37°C in 5% FCS-DMEM (Nacalai Tesque) supplemented with penicillin (100 units/mL) and streptomycin (100 μg/mL). Hypoxic cell cultures were done in 5% CO₂-1% O₂ in a multigas incubator (ASTEC).

Growth assay. SUIT-2/EF-Luc and SUIT-2/HRE-Luc cells (100/100 μL/well) were seeded in a 96-well plate and cultured at 20% or 1% O₂. The following day, 5 μL of SF cell counting reagent containing 2-(2-methoxy-4-nitrophenyl)-3-(4-nitrophenyl)-5-(2, 4-disulfophenyl)-2H-tetrazolium, monosodium salt (Nacalai Tesque) were added to each well, and 24 h later, the plates were monitored by a plate reader photometer (absorbance at 450 nm). The assay was repeated every day. The absorbance of each well was divided by the absorbance on day 1 to calculate the relative cell number. The experiments were done in triplicate and results were presented as mean relative absorbance ± SE. The experiment was repeated thrice.

Luciferase activity assay. Cells (10⁴) were seeded onto 24-well plate and cultured overnight and then further cultured in an ordinary CO₂ incubator or a hypoxic chamber BACLITE-1 (pO₂, >0.02%; Sheldon Manufacturing) for 18 h. The cells were lysed and their luciferase activity was measured as described previously (19).

Western blot analysis. SUIT-2/EF-Luc, SUIT-2/HRE-Luc, and parental SUIT-2 cells were cultured under hypoxic (1% O₂) or aerobic (20% O₂) conditions for 20 h and then cells were harvested and the cell lysates were analyzed by Western blotting as described previously (19). The experiment was repeated thrice.

Orthotopic transplantation of tumor cells and real-time monitoring of luciferase activity in vivo. Orthotopic transplantation of tumor cells was done as described previously (31). Briefly, SUIT-2/HRE-Luc or SUIT-2/EF-Luc cells (1 × 10⁵/25 μL) mixed with an equal volume of Geltrex (Invitrogen) were injected into the pancreas of 8-wk-old male nude mice (BALB/c-nu/nu; Japan SLC, Inc.). For the in vivo photon counting of bioluminescence, tumor-bearing mice were i.p. injected with 200 μL of d-luciferin solution (10 mg/mL in PBS; Promega Corp.). Acquisition and analysis of images were done with IVIS-200 (Xenogen Corp.) as described previously (24). As background photon counts/second in untreated nude mice (with d-luciferin injection) were <4,000, we considered >1 × 10⁴ photon counts/s a significant signal and showed images >1 × 10⁵ photon counts/s in the figures. The minimum and maximum photons/second of each figure are indicated in each rainbow bar scale.

Measurement of pO₂ in the abdominal cavity. Mice were anesthetized by 2.5% of isoflurane gas in oxygen flow (25 mL/s) and kept on a heating pad to maintain body temperature between 37°C and 38°C. A needle (20 G × 1^1/2, TERUMO) was inserted and fixed on the abdominal skin with a quick-drying glue. A modified PtiO₂ probe (LICOX CC1.R, 7-mm sensitive area length, 0.65-mm diameter; Integra NeuroSciences) was inserted into the abdominal cavity through the needle and connected to the LICOX CMP instrument (Integra NeuroSciences). The most outer code cover was fixed to the needle with a quick-drying glue so that the abdominal cavity was kept closed. The oxygen partial pressure was measured for >5 min and automatically recorded every 20 s while a needle-type electrode was moved throughout the abdominal cavity. The average oxygen concentration (mm Hg) ± SE was calculated from each record. Ascites samples were drawn from the abdominal cavity by puncturing with heparin-containing syringes and then immediately subjected to analysis (ABL 700 series, Radiometer) for oxygen tension.

ELISA assay for VEGF. Ascites fluid collected from the mice was left to stand at room temperature for 2 h and then centrifuged for 20 min at 2,000 × g. The supernatant was immediately assayed for VEGF concentration with the Quantikine mouse VEGF immunoassay kit (R&D Systems).

Preparation and analysis of recombinant EGFP proteins. Expression vector plasmids of PTD-EGFP fusion proteins encoding enhanced green florescent protein were constructed by inserting the cDNA fragments encoding PTD, which was prepared by annealing corresponding oligonucleotides, and EGFP, which was excised from pEF-EGFP (Promega), into plasmid pGEX-6P-3. Recombinant proteins were prepared as described previously (20) and suspended in PBS at a final concentration of 50 μmol/L. The cells were trypsinized and suspended in 300 μL of ice-cold PBS. The fluorescence intensity of the cells was determined by fluorescence-activated cell sorting using CellQuest (Becton Dickinson).

POP33 preparation and administration to mice. DNA sequences encoding PTD3 were prepared by annealing appropriate oligonucleotides and replaced the TAT-ptd sequence present in TOP3. POP33 fusion protein was prepared as described previously for TOP3 (23). POP33 suspended in 10 mmol/L Tris buffer (1 mg/mL) was i.p. injected into tumor-bearing mice (5 mg/kg).

In vivo imaging of caspase-3 activity. For preparation of Z-DEVD-aminoluciferin, 10 mg of VivoGlo Caspase-3/7 substrate (Promega) were dissolved in 67 μL of surfactant mix (70 μL of Tween 80, 640 μL of PEG 400, and 290 μL of N,N-dimethylacetamide) and then added to 100 μL of 5% dextrose in water. The mixture was vortexed to make a homogenous solution and peritoneally injected into tumor-bearing mice (100 μL/mouse). Ten minutes after Z-DEVD-aminoluciferin injection, the mice were imaged with IVIS.

Statistical analysis. The Kaplan-Meier method was used to determine survival curves, and comparisons between the curves were done by log-rank test. The statistical analyses were done using Statistical Package for the Social Sciences 16.0 for Windows and Excel. A P value of <0.05 was considered statistically significant.

Characterization of SUIT-2 pancreatic cancer cells with luciferase reporter genes. HIF-1α overexpression is one of the hallmarks of malignancy (32). We hypothesized that HIF-1–activity represents an important driving force for invasion and metastasis. To test this hypothesis, we did serial quantitative measurements of HIF-1 activity in xenografts derived from cancer cell lines stably transfected with plasmid p5HRE-luciferase (21), which expressed firefly luciferase under the control of a HIF-1–dependent promoter. We applied this system to a clinically relevant orthotopic model to determine the levels of HIF-1 activity during the progression of pancreatic cancer. We first established a stable human pancreatic cancer cell line, SUIT-2/HRE-Luc. The control cell line SUIT-2/EF-Luc constitutively expressed luciferase, and thus, the luciferase activity was proportional to the tumor mass (21). The growth (Fig. 1A) and HIF-1α expression (Fig. 1B) of these cell lines were similar. The luciferase activity in SUIT-2/HRE-Luc cells under hypoxic conditions was >100 times higher than under aerobic conditions (Fig. 1C). Conversely, the luciferase activity in SUIT-2/EF-Luc decreased as pO₂ decreased (Fig. 1C), reflecting the global inhibition of transcription and translation that occurs in response to hypoxia (33).

In vivo imaging of the progression of pancreatic cancer. The orthotopic model using SUIT-2 cells has been extensively studied and well characterized (13). SUIT-2 cells, which are c-Met positive, form invasive tumors with peritoneal dissemination and fulminant ascites in a reproducible manner (13). We transplanted SUIT-2/EF-Luc and SUIT-2/HRE-Luc cells to the pancreas and measured the intensity of the bioluminescent signal [photons/second/region of interest (ROI)]. The bioluminescent signals from SUIT-2/EF-Luc and SUIT-2/HRE-Luc xenografts reflect tumor mass and HIF-1 activity, respectively.

A preliminary experiment was done with mice bearing various size of xenografts and found that the photon counts of all xenografts peaked between 13 and 15 minutes after d-luciferin injection (Supplementary Fig. S1) Thus, in the subsequent experiments, we acquired images of the transplanted mice between 13 and 15 minutes after d-luciferin injection.

Representative bioluminescence images of HIF-1 activity during progression of the orthotopic pancreatic tumor model are shown in Fig. 2A. A significant signal over background was detected at the injection site 1 to 2 weeks after orthotopic transplantation; by the following week, it had spread to the abdomen; by the 3rd week, it had spread throughout the abdomen with peritoneal fluid accumulation; and by 4 weeks after initial signal detection, 100% of mice had died with massive ascites. Although the time course of the increases in tumor mass (SUIT-2/EF-Luc) and HIF-1 activity (SUIT-2/HRE-Luc) was similar, increases in HIF-1 activity were delayed about a week compared with increases in tumor mass (Fig. 2B and C), suggesting that the HIF-1 activity corresponded to the development of hypoxic regions in the growing cancers. Interestingly, the timing of increased bioluminescence associated with spread of cancer to the abdominal cavity, typically 11 to 13 days after transplantation, was the same for the tumor mass and HIF-1 activity (Fig. 2C), indicating that the tumor cells invading the abdominal cavity already had HIF-1 activity.

Peritoneal dissemination caused hypoxia in the abdominal cavity. To directly measure pO₂ in the abdominal cavity at different stages of tumor progression, a PtiO₂ probe was inserted into the abdominal cavity. Real-time monitoring of pO₂ throughout the abdominal cavity was done over 5 minutes. When tumor cells were localized to the transplant site and no ascites was detected (Fig. 3A, left), pO₂ in the abdomen was 86.4 ± 10.1 mm Hg, which was similar to the oxygen concentration in untreated mice (98.5 ± 9.5 mm Hg). On the other hand, when HIF-1–specific bioluminescence spread to the abdomen (∼5 × 10⁶ photon counts/s/ROI), the pO₂ was significantly decreased (42.5 ± 5.6 mm Hg). When the abdominal cavity was filled with ascites and a strong signal (∼1 × 10⁸ photon counts/s/ROI) was detected throughout the abdomen, pO₂ was <10 mm Hg. Ascites fluid accumulated significantly when the HIF-1–specific bioluminescent signal was >1 × 10⁷ photon counts/s/ROI. The pO₂ of ascites was also measured and was similar to the abdominal cavity (Fig. 3A). HIF-1 activity was reciprocally correlated with oxygen tension in both the abdominal cavity and ascites (Fig. 3A).

At the final stage of the pancreatic cancer, the ascites contained extremely high levels (between 1,000 and 2,200 pg/mL) of VEGF, a prototypical HIF-1 target gene product. The serum levels of VEGF in control mice were 113.2 ± 30.8 pg/mL. These results confirm that hypoxia induces HIF-1 activity in this model. Together with Fig. 2, these results suggest that HIF-1–active cancer cells invade locally and create hypoxic conditions during growth and dissemination throughout the abdominal cavity.

Pancreatic cancer cells with HIF-1 activity were also visualized by bioluminescent imaging postmortem. Mice with fulminant ascites were dissected 13 minutes after d-luciferin injection and HIF-1 activity of xenografts in the abdominal cavity was directly examined by ex vivo imaging (Fig. 3B, a-e). The average volume of ascites was 4.1 mL when mice died with fulminant ascites. Cells in the ascites, which were collected by centrifugation, also had high levels of luciferase activity (Fig. 3B, b). Bioluminescence was also detected in the intestines (Fig. 3B, a and c), primary tumors in the pancreas (Fig. 3B, d), and liver metastases (Fig. 3B, e). These data indicate that HIF-1–active tumor cells contribute to local invasion, metastasis, and peritoneal dissemination.

Construction of a novel hypoxia-targeting protein drug. We hypothesized that if HIF-1–active cells play a crucial role during the progression of pancreatic cancer, eradication of these cells would suppress invasion and metastasis. For eradication of HIF-1–active cells, we previously developed and validated a hypoxia-targeting prodrug, TOP3, which efficiently induced apoptosis and decreased HIF-1 activity in tumors (23, 24) and significantly enhanced radiotherapy (20). To improve efficacy, we constructed POP33, which contains a novel protein transduction domain, PTD3, consisting of polylysine and noncharged amino acids (Fig. 4A). To assess the membrane transduction function of PTD fusion proteins, PTD3 (present in POP33) and TAT-ptd (present in TOP3) were fused to EGFP and their membrane transduction activity was assessed with a fluorescence microscopy and flow cytometry. The fusion proteins had similar fluorescent intensity in vitro (Fig. 4B). SUIT-2/HRE-Luc cells were treated with PTD3-EGFP or TAT-ptd-EGFP for 30 minutes and washed extensively. The cells were then trypsinized and their fluorescent intensity was quantified by flow cytometry (Fig. 4C). The results indicate that PTD3 has more than five times higher cell-penetrating activity than TAT-PTD.

In vivo caspase-3 activity induced by POP33 treatment. Apoptosis induction by TOP3 was previously observed in vitro (23, 25) and in tumor sections by terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (24). To directly examine the ability of POP33 to increase caspase-3 activity in vivo, we used Z-DEVD-aminoluciferin (34), a prosubstrate that is converted to a luciferase substrate when cleaved by caspase-3. POP33 or buffer alone was i.p. injected into mice with various degrees of HIF-1–dependent bioluminescence (Fig. 5A, top). Six hours later, Z-DEVD-aminoluciferin was injected and caspase-3 activity was determined (Fig. 5A, bottom). Caspase-3 activity increased as HIF-1 activity increased in both control and POP33-treated mice [Fig. 5A (top) and B]. However, in the POP33-treated mice, higher caspase-3–dependent bioluminescence was detected compared with control mice with similar levels of HIF-1–dependent bioluminescence (Fig. 5B). These results indicate that the procaspase-3 domain in POP33 is cleaved, thereby increasing caspase-3 activity in HIF-1–active/hypoxic tumor cells.

POP33 suppressed peritoneal dissemination and improved survival. When total photon counts were ∼1 × 10⁵ photons/s/ROI, mice were randomized to control and treatment groups. Representative ex vivo images of tumors with 1 × 10⁵ and 1 × 10⁶ photons/s/ROI are shown in Supplementary Fig. S2. The mice in the treatment group (n = 8) were injected with 5 mg/kg POP33 every 4 days until the mouse expired or was euthanized. We treated the mice in this way because PTD-ODD-procaspase-3 does not affect well-oxygenized proliferating tumor cells, and thus, growing tumors make new hypoxic fractions and increase HIF-1 activity in the tumors 3 days after the treatment (20, 23, 24). HIF-1 activity in the xenografts of untreated mice rapidly increased, and all of the mice died by 24 days after the initial image detection [Fig. 6A (left graph) and B (top)]. On the other hand, in half of the mice treated with POP33, HIF-1 activity in the xenografts increased more slowly or stayed at low levels [Fig. 6A (right graph) and B (bottom)]. The survival of the POP33-treated mice was significantly extended (P < 0.05, log-rank analysis). The median survival time of the treated mice was 23 days compared with 13 days for untreated mice (Fig. 6C).

HIF-1α expression has been shown in the majority of human cancers, and there is a direct correlation between the degree of overexpression and patient mortality in many types of cancer (12, 32). In the current study, we visualized HIF-1 activity in an orthotopic pancreatic cancer model using the HIF-1–dependent expression of luciferase and in vivo bioluminescent imaging and showed a strong correlation between HIF-1 activity and tumor progression. These results are consistent with the hypothesis that “HIF-1 activity plays a crucial role in invasion and metastasis.” In this model, 100% of mice died with peritoneal dissemination and fulminant ascites. The abdominal cavity filled with ascites was severely hypoxic (<10 mm Hg) at the terminal stage. This result was consistent with the observation that the level of VEGF, which is a prototypical HIF-1 target gene product (35), was dramatically increased in ascites, as previously described for human pancreatic cancer specimens (36).

The bioluminescent signal in SUIT-2/EF-Luc–transplanted mice corresponded to the previously reported progression of SUIT-2 pancreatic cancer model (13). The appearance of the bioluminescent signal from SUIT-2/HRE-Luc xenografts (HIF-1 activity) was significantly delayed about a week compared with the constitutive signal from SUIT-2/EF-Luc xenografts (tumor mass). Because the growth rates of both cell lines were comparable in vitro, it is likely that both cell lines grow at similar rates in vivo and that the signal from SUIT-2/HRE xenografts reflects tumor hypoxia. Therefore, a hypoxic microenvironment must have developed about a week after a tumor mass formed at the transplant site. Furthermore, the images of HIF-1 activity and tumor mass (from day 13) appeared almost identical between these two cell lines, indicating that HIF-1–active/hypoxic cells were the major population during cancer progression. HIF-1 imaging showed the process of pancreatic cancer progression from local growth in the pancreas to invasion of the abdominal cavity and then peritoneal dissemination, which was associated with severe hypoxia. Taken together, these results indicate that HIF-1–active cancer cells play an important role in the malignant progression of pancreatic cancers.

The abdominal cavity was not hypoxic (pO₂, ∼85 mm Hg) in non–tumor-bearing mice. However, when HIF-1–active cells invaded the abdominal cavity, the local pO₂ decreased as the HIF-1 activity increased. Taken together with the bioluminescence imaging data, these results suggest that HIF-1–active cancer cells locally invaded, proliferated, and disseminated, creating a severely hypoxic environment.

Based on these findings, we have pursued strategies to target hypoxic pancreatic cancer cells. PTD-ODD-procaspase-3 fusion proteins must enter cells to induce apoptosis, and thus, PTD function is very important for their efficacy. To improve PTD function, we constructed PTD3, which consists of two polar amino acid sequences, a hydrophobic tryptophan-rich motif related to Pep-1 (37), and polylysine. PTD3 endowed the fusion protein with higher membrane-penetrating activity than TAT-PTD sequences. The precise mechanism of PTD3 membrane penetration is under investigation.

The significance of HIF-1 in pancreatic cancer has been suggested (8–11, 38). HIF-1–targeting prodrug POP33 successfully suppressed local invasion and thus peritoneal dissemination. Hypoxia activates caspase-3 in cells to some extent in vivo as well as in vitro (23–25). Here, we directly showed that it is also true in vivo by visualization of caspase-3 activity with a real-time imaging method; in untreated mice, cancer cells had some caspase-3 activity, which also increased as HIF-1 activity increased. PTD-ODD-procaspase-3 increased the caspase-3 activity significantly higher than the control (23–25) and enhanced apoptosis in vivo (23, 25), suppressing the tumor growth (24), metastasis (26), and local invasion of HIF-1–active cell in the pancreatic cancer. These results not only show that HIF-1 plays a crucial role in the progression of cancers but also suggest that eradication of HIF-1–active cells may improve the therapeutic outcome in pancreatic cancer.

No potential conflicts of interest were disclosed.

We thank Dr. Ohtsura Niwa (National Institute of Radiological Sciences) for discussion, Waka Kawaguchi for skilled technical assistance, Shigeaki Watanabe (Summit Pharmaceuticals International Corp.) for discussion and technical support of IVIS, and Tadaki Kido (Medical Agent Corp.) for PtiO₂ probe (LICOX CC1.R) operation.