NRF2 Blockade Suppresses Colon Tumor Angiogenesis by Inhibiting Hypoxia-Induced Activation of HIF-1α

The use of molecular oxygen (O₂) as the final electron acceptor enables aerobic organisms to produce metabolic energy (ATP) with high efficiency. However, mitochondrial respiration can generate reactive oxygen species (ROS) such as superoxide (O₂⁻) and hydrogen peroxide (H₂O₂), which are primarily harmful to cellular micromolecules by nonspecifically reacting with nucleic acids, proteins, and lipids (1). Therefore, excess ROS has been implicated in various human diseases, including cancer, neurodegeneration, inflammatory damage, and aging (1, 2).

In response to the deleterious effects of excess ROS, cells need to develop a system to counteract ROS attack. As a major defense system, antioxidant proteins and phase II detoxifying enzymes contribute to the restoration of cellular redox balance by preventing ROS accumulation. Many antioxidant genes are coordinately regulated through the antioxidant response element (ARE) in their promoters, and the transcription factor nuclear factor-erythroid 2 (NRF2) is the master regulator of ARE-driven genes (3–6). A subset of genes encoding detoxifying enzymes [e.g., glutathione S-transferases (GST) and NAD(P)H: quinone oxidoreductase (NQO1)], thiol molecules and their regenerating enzymes [e.g., thioredoxin (TXN) and glutamate cysteine ligase (GCL)], stress-response proteins [e.g., heme oxygenase-1 (HO-1)], and direct ROS-removing enzymes [e.g., glutathione peroxidase (GPX)], are known targets of NRF2 regulation (7–10). NRF2 activity is mainly regulated by Kelch-like ECH-associated protein 1 (KEAP1), which is an adaptor protein for ubiquitination and subsequent degradation of NRF2 by the 26S proteasome under quiescent conditions (11–13). In the presence of high ROS levels, NRF2 is liberated from KEAP1 and translocates to the nucleus, where it transactivates ARE-driven gene expression with other bZIP proteins such as small MAFs. Extensive studies during the last decade have provided strong evidence that cells or organisms with the NRF2 defect are more susceptible to an array of oxidative insults (5, 6, 14–16). There is an emerging concern that several types of cancer cells overactivate NRF2 signaling, which results in the refractoriness of cancer cells to oxidative stress and chemotherapy (17, 18). Therefore, transient suppression of NRF2 expression in these cancer cells could diminish their chemoresistance by affecting the expression of NRF2 target genes (19).

Hypoxia-inducible factor-1 (HIF-1) is a key transcriptional regulator that senses O₂ homeostasis (20). In response to hypoxia, HIF upregulates the expression of many genes responsible for adaptation to hypoxia. HIF-1 is a heterodimeric complex of O₂-regulated subunit HIF-1α and constitutively expressed subunit HIF-1β (21). Under normal O₂ concentrations, prolyl hydroxylase domain proteins (PHD), which are enzymes that consume O₂, hydroxylate specific HIF-1α proline residues. Von Hippel-Lindau (pVHL) tumor suppressor protein binds hydroxylated HIF-1α, which in turn leads to ubiquitination and degradation of this protein. Under low O₂ concentrations, insufficient O₂ substrate prevents PHDs activation, which leads to the increase in the expression of HIF-1α target genes, including vascular endothelial growth factor (VEGF) (20, 22, 23). Often, HIF-1α is deregulated in tumors located in hypoxic microenvironments and plays an important role in angiogenesis, tumor progression, and resistance to chemo- and radiotherapies (24). Whereas, a recent study demonstrated that carbon monoxide produced by HO-1, a target gene of HIF-1α as well as NRF2, increases HIF-1α stability and enhances VEGF expression (25).

In the current study, we investigated the role of NRF2 in colon tumor angiogenesis. Colon cancer cells expressing NRF2-specific shRNA showed retarded tumor growth, diminished vascular formation, and low VEGF expression in mouse xenografts model. A chicken embryo chorioallantoic membrane (CAM) and endothelial tube formation assays showed that cancer cell-induced angiogenesis is suppressed by NRF2 inhibition. In a mechanistic investigation, hypoxia-inducible HIF-1α signaling is blocked in NRF2-inhibited cells. Exposure of NRF2 knockdown cancer cells to 1% O₂ maintained PHD-mediated HIF-1α hydroxylation, presumably through O₂ redistribution. Indeed, mitochondrial O₂ consumption in NRF2-inhibted cancer cells was low in comparison to that in the scRNA control.

Antibodies recognizing NRF2, VEGF, β-tubulin, and lamin were obtained from Santa Cruz Biotechnology. Antibodies against HIF-1α, hydroxyl (OH-Pro564) HIF-1α, and caspase-3 were purchased from Cell Signaling Technology, and PHD2 and HO-1 antibodies were from BD Biosciences and Abcam, respectively. The lentiviral system, including plasmids with predesigned human NRF2 shRNAs was purchased from Sigma-Aldrich. The reporter plasmid containing the human NQO1 ARE has been described previously (12). The luciferase reporter plasmid with the HIF-1α response element (HRE) contains 5 copies of the human VEGF HRE (26) and is a gift from Dr. You Mie Lee (Kyungpook National University, Daegu, South Korea). MG132 was from EMD chemicals. All other reagents including sulforaphane (SFN), CoCl₂, cycloheximide (CHX), and myxothiazol were purchased from Sigma-Aldrich.

Human colon cancer cell line HCT116 and HT29 were obtained from American Type Culture Collection. HCT116 was maintained in RPMI 1640 (Cambrex Bio Science) with 10% FBS (Hyclone) and penicillin/streptomycin (Hyclone). HT29 cells were maintained in Dulbecco's modified Eagle's Medium (Hyclone) with supplement of 10% FBS and penicillin/streptomycin. Cells were grown at 37°C in a humidified 5% CO₂ atmosphere.

Lentiviral particles with shRNA were produced in HEK 293T as described previously (27). Briefly, HEK 293T cells were transfected with 1.5 μg pLKO.1-NRF2 shRNA (shRNA #1, 5′-CCGGGCTCCTACTGTGATGTGAAATCTCGAGATTTCACATCACAGTA GGA-3′; shRNA #2, 5′-CCGGCCCTGTTGATTTAGACGGTATCTCGAGATACCGTCTAATCAACAGGGTTTTT-3′) and the Mission Lentiviral Packaging Mix (Sigma-Aldrich) using Lipofectamine 2000 (Invitrogen). The pLKO.1-scrambled RNA (scRNA) plasmid was used as a nonspecific control RNA (27). On the second day, the medium with transfection complex was removed and the complete medium was added into each well. Media containing lentiviral particles were harvested after 4 days and used for subsequent transduction.

HT29 and HCT116 cells in 6-well plates were transduced with lentiviral particles containing either nonspecific scRNA or NRF2 shRNA expression plasmid. Transduction was continued for 48 hours and followed by a 24-hour recovery in the complete medium. For the selection of cells with target plasmids, cells were grown in a medium containing 1 μg/mL puromycin (Sigma-Aldrich) for up to 4 weeks.

Total RNAs were isolated from cells using a Trizol reagent (Invitrogen). For the synthesis of cDNAs, a reverse transcriptase (RT) reaction was performed by incubating 200 ng of total RNA with a reaction mixture containing Moloney Murine Leukemia Virus RT (Invitrogen). PCR amplification for each gene was carried out with a thermal cycler (Bio-Rad). For quantitative real-time RT-PCR analysis, the Roche LightCycler was used with the Takara SYBR Premix ExTaq system. Primers were synthesized by Bioneer and the primer sequences for the human genes are: NRF2, 5′-ATAGCTGAGCCCAGTATC-3′ and 5′-CATGCACGTGAGTGCTCT-3′; NQO1, 5′-GATATTGTGGCTGAACAA-3′ and 5′-TGCTATATGTCAGTTGAG-3′; GCLC, 5′-AGACATTGATTGTCGCTG-3′ and 5′-TGGTCAGACTCATTAGCA-3′; HIF-1α, 5′-ACAGCAGCCAGACGATCATGCAG-3′ and 5′-AACTGGTCAGCTGTGGTAATCCACT-3′; VEGF, 5′-TTGTACAAGATCCGCAGACG-3′ and 5′-TTCTGTCGATGGTGATGGTG-3′; PHD1, 5′-ACGGGCTCGGGTACGTAAG-3′ and 5′-CCCAGTTCTGATTCAGGTAATAGATACA-3′; PHD2, 5′-GACCTGATACGCCACTGTAACG-3′ and 5′-CCCGGATAACAAGCAACCAT-3′; PHD3, 5′-AACTGAATCTGCCCTCACTGAAG-3′ and 5′-ATAATTCAGGAACCGTTACTAAAATGA-3′; nuclear respiratory factor-1 (NRF-1), 5′-CGCAGCCGCTCTGAGAACTTCA-3′ and 5′-TCGGGAGAAGAAGGCGAGTCTTCA-3′; peroxisome proliferator-activated receptor-γ coactivator-1α (PGC-1α), 5′-AGACAGCTTTCTGGGTGGACTCAA-3′ and 5′-TCAGCGCATCAAATGAGGGCAATC-3′; uncoupling protein-2 (UCP2), 5′-TCGGAGATACCAAAGCACCGTCAA-3′ and 5′-ATAGGTCACCAGCTCAGCGTT-3′; HO-1, 5′-GCTGCTGACCCATGACACCAAGG-3′ and 5′-AAGGACCCATCGGAGAAGCGGAG-3′; hypoxanthine-guanine phosphoribosyltransferase (HPRT), 5′-GGACTAATTATGGACAGGAC-3′ and 5′-TGCATTGTTTTGCCAGTGTC-3′. PCR products were resolved on 1.2% agarose gels and the images were captured by using a Visi Doc-It imaging system (UVP).

Proteins were separated on 6% to10% SDS-polyacrylamide gels and transferred to nitrocellulose membranes (Whatman). Membranes were blocked with 3% skim milk for 1 hour and incubated overnight with the primary antibodies. Then, membranes were incubated with the secondary antibody conjugated with horseradish peroxidase (Bio-Rad) for an hour. The detection was done with the Enhanced Chemiluminescence reagent (Amersham Biosciences) and LAS-4000 ImageReader (Fujifilm). Obtained images were quantified by Multi-gauge software (Fujifilm).

Cells were grown in 6-well plates for 24 hours and lysed with 5% metaphosphoric acid solution. For the measurement of total GSH content, optical densities were monitored for 4 minutes following an incubation of 30 μL cell lysates with 30 μL 5,′5-dithiobis (2-nitrobenzoic acid), glutathione reductase and β-NADPH. Protein concentration was determined by BCA protein assay kit (Pierce).

Cells were seeded in 24-well plates at a density of 2 × 10⁴ cells/well and grown for overnight. Next day, the transfection complex containing 0.5 μg of the reporter plasmid, 0.05 μg of pRLtk control plasmid (Promega) and the transfection reagent (Welgene Inc.) was added to each well. After 18 hours, the transfection complex was removed and cells were incubated in the complete medium for another 24 hours. Then, Renilla and Firefly luciferase activities were measured in cell lysates using the Dual Luciferase Assay System (Promega) with a luminometer (Turner Designs).

Cells were plated in 96-well plates at a density of 5 × 10³ cells per well. After the incubation of cells with doxorubicin for 24 hours, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) solution (2 mg/mL) was added and cells were further incubated for 4 hours. Following the removal of MTT solution, 100 μL of DMSO was added and absorbance was measured at 540 nm using a Versamax microplate reader.

HT29 (HT-SCi and HT-NRF2i) and HCT116 (HCT-SCi and HCT-NRF2i) cells were grown for overnight and harvested, washed twice with 1 × PBS. The suspension of 2.5 × 10⁶ cells in 0.2 mL serum-free medium was injected subcutaneously into the flank of 6-week-old BALBc (nu/nu) mice (Orient Bio Inc.), which were maintained in pathogen-free environment. The tumor growth was monitored twice a week by measuring 2 diameters of tumors with calipers. The tumor volume was calculated by the formula V = (a² × b)/2, where a and b are the width and the length of tumors in mm, respectively. Each group contained 4 to 5 animals. These experiments were performed according to the institutional guidelines for the Care and Use of Laboratory Animals as adopted by the United Sates National Institute of Health and the Yeungnam University Animal Care and Use Committee.

The tumor masses were fixed in 10% neutral buffered formalin and paraffin-embedded. Tissue sections with 3 to 4 μm thickness were prepared and stained with hematoxylin and eosin (H&E) for light microscope observation. The Masson's trichrome staining was performed for the detection of tumor cells with collagen fiber. The AZAN staining was done to visualize the basement membranes and erythrocytes, which represent blood vessels. Occupied regions by tumor cells (% overtrimmed tumor mass) and numbers of vessels (number/mm² of central regions of tumor mass) were counted using a digital image analyzer (DMI-300; DMI). The histomorphometrical analysis was conducted in 2 or 3 fields per tumor mass, and total 8 fields were used for the analysis of each group.

For the immunohistochemical analysis, tumor sections were deparaffinized and epitope retrieval was conducted. Briefly, tumor sections were incubated with 10 mmol/L citrate buffer (pH 6.0) at 95°C to 100°C for 20 minutes and were allowed to cool for another 20 minutes. Then tumor sections were treated with methanol and 0.3% H₂O₂ for 30 minutes to inactivate endogenous peroxidase, and the incubation with normal horse serum-containing blocking solution (Vector Lab. Inc.) was followed. The primary antibodies reacting with caspase-3 and VEGF were incubated overnight, and then the incubation with biotinylated secondary antibody and peroxidase substrate was followed. The numbers of positive cells, which occupy over 10% of immunoreactivity, were expressed as cell numbers among total 100 tumor cells in the central regions.

DNA synthesis of cells was measured using the cell proliferation ELISA kit (Roche Applied Sciences) according to manufacturer's instruction. This system is a colorimetric immunoassay based on the assessment of bromodeoxyuridine (BrdU) incorporation during DNA synthesis. Briefly, cells were plated in 96-well plates at a density of 5 × 10³ cells/well and grown for 48 hour. Then, cells were labeled with BrdU for 4 hours, and incorporated BrdU was quantified using a plate reader at wavelengths of 405 nm and 492 nm.

The 10 day-old chicken embryos were purchased from Baek-ja Farm and were incubated at 37°C with 55% relative humidity. A window of approximate length with 1.0 cm² was cut in egg shell using a small crafts grinding wheel (Dremel). Human colon cancer cells with either scRNA or NRF2 shRNA were mixed with Matrigel (BD Biosciences) and inoculated on the CAM (2 × 10⁶ cells/CAM). The CAMs were examined at 72-hour intervals using a SV6 stereomicroscope (Carl Zeiss) at ×50 magnifications. Digital images of CAM sections were captured using a 3-charge coupled color video camera system (Toshiba). The images were analyzed using Image-Pro software (Media Cybernetics) and the number of vessel branch points within a circular region was counted.

Hypoxic chambers (Thermo Scientific) were used for maintaining cells in either normoxic (21% O₂ and 5% CO₂), hypoxic (1% O₂, 5% CO₂, and N₂ balance), or anoxic (0.1% O₂, 5% CO₂, and N₂ balance) conditions at 37°C. Cells were exposed to hypoxia for 24 hours and immediately harvested, lysed, and frozen. In order to monitor HIF-1α stability, HT 29 cells were incubated in hypoxic chamber with MG132 for 24 hours, and washed to remove MG 132. Then, CHX was added and cells were further incubated in hypoxic or normoxic conditions for 10, 20, 40, or 60 minutes. For the effect of PHD inhibition on HIF-1α stability, ethyl-3,4-dihydroxybenzoate (EDHB, 150 μmol/L)-treated hypoxic cells were incubated for 10, 20, 40, and 60 minutes in the presence of CHX. All treatments were done within the chamber.

Human umbilical vein endothelial cells (HUVEC) were purchased from Lonza and cultured in EBM-2 (Lonza) supplemented with EGM-2 SingleQuots (Cambrex). To assess the effect of NRF2 on HUVECs tube formation, HT-SCi and HT-NRF2i cells in supplement-free EBM-2 medium were incubated in hypoxic condition for 24 hours, and supernatants (conditioned media) were collected. Then, HUVECs were plated in Matrigel-coated 48-well plates. Two volumes of conditioned media or fresh EBM-2 medium containing VEGF (10 ng/mL) were added into cells. After 18 hours, 4 or more random pictures were taken of each culture well using a digital camera system. Images were subsequently analyzed with an image analysis system (ImageInside Ver 3.32) for the quantitation of tube lengths.

BD Oxygen Biosensor System (BD Bioscience) was used for the measurement of cellular O₂ consumption. Briefly, cells were seeded in a 96-well BD Oxygen Biosensor System plate at a density of 5 × 10⁴ per well. After 48 hours of incubation, fluorescence intensities were read using a plate reader (Fluostar Optima; BMG Lab Technologies) with an excitation wavelength of 485 nm and an emission wavelength of 630 nm. Relative fluorescence units were normalized to the signal in air-saturated buffer. At the same time, to confirm that the measured fluorescence varies inversely with O₂ concentration in the medium, 100 mmol/L sodium sulfite was used as a positive control to eliminate all molecular oxygen. Protein concentration was determined for the final normalization.

Cells were plated in 96-well plates at a density of 2 × 10³ cells/well. The next day, 100 μL of 0.3% trichloroacetic acid was added to each well and incubation was followed for 30 minutes on ice. After removing protein precipitates by centrifugation, supernatants were obtained and 4 × volumes of 250 mmol/L Tris-acetate (pH 7.75) were added to neutralize pH of each sample. ATP contents were determined using an ATP assay kit (Enliten ATP assay kits; Promega) following an addition of 100 μL of substrate solution to 10 μL sample. Luminescence activities produced from ATP-dependent oxidation of luciferin were measured, and were converted to moles of ATP using a standard curve. Obtained ATP contents were then normalized by protein contents of each sample.

Statistical significance was determined by Student paired t test or 1-way ANOVA followed by Student–Newman–Keuls's comparison method (SigmaStat analysis software). For the analysis of histomorphometric results, multiple comparison tests for different groups were conducted. Variance homogeneity was examined using the Levene test. If the Levene test indicated no significant deviations from variance homogeneity, the obtain data were analyzed by 1-way ANOVA test followed by least-significant differences (LSD) multicomparison test. If significant deviations were observed at Levene test, a nonparametric comparison test, Kruskal–Wallis H test was conducted. When a significant difference is observed in the Kruskal–Wallis H test, the Mann–Whitney U test was conducted to determine the specific pairs of group comparison. Statistical analyses were conducted using SPSS for Windows (Release 14.0K; SPSS Inc.).

To investigate the role of NRF2 in tumor growth and angiogenesis, we established NRF2 knockdown colon cancer cell lines. For stable knockdown of NRF2, HCT116, and HT29 cells were transduced with lentiviral particles containing viral plasmids encoding nonspecific scRNA or NRF2-targeting shRNA, and stable cell lines were obtained following puromycin selection. In comparison to the scRNA control (HCT-SCi and HT-SCi), cells with stable shRNA expression (HCT-NRF2i and HT-NRF2i) showed repressed transcript levels of NRF2 and its target genes, such as the catalytic subunit of GCL (GCLC) and NQO1 (Fig. 1A). Reporter analysis with an ARE-luciferase plasmid verified the repression of NRF2: ARE-driven luciferase activity in HCT-NRF2i cells was only 15% of that in HCT-SCi control cells (Fig. 1B). In addition to reduced basal expression, sulforaphane (SFN)-inducible ARE luciferase activity was diminished in both NRF2-inhibited cell lines (Fig. 1B). Similar patterns were observed by NRF2 immunoblot analysis. Nuclear NRF2 levels were low in HCT-NRF2i cells in comparison to the SCi control, and SFN-inducible nuclear NRF2 accumulation was also repressed in both NRF2-inhibited cells (Fig. 1C). To assess NRF2 functional inhibition, cellular GSH content was determined. In results consistent with reduced GCLC expression, GSH levels in NRF2-knockdown cell lines were significantly diminished: total cellular GSH was reduced by 70% in HCT-NRF2i and 40% in HT-NRF2i (Fig. 1D). Previous reports have demonstrated that NRF2 enhances sensitivity to chemotherapeutic agents such as doxorubicin, a premise which we tested in our established cell lines. However, NRF2 knockdown in colon cancer cells did not yield significant changes in doxorubicin sensitivity. NRF2 inhibition partially sensitized HCT116 cells, while no effect was observed in HT29 cells (Fig. 1E).

To investigate the role of NRF2 in tumorigenesis, established colon cancer cells were implanted in nude mice and tumor growth was assessed. In the control group, HCT116 cells expressing scRNA (HCT-SCi) developed tumors 2 weeks after implantation and tumor size increased continuously up to 33 days (Fig. 2A). While, tumor growth of the HCT-NRF2i cells was greatly suppressed in comparison to the scRNA control. Although the degree of suppression was different, a similar pattern was observed in HT29 cells (Fig. 2B). When tumor cells were identified by the Masson's trichrome staining, we found that the percentages of tumor cell-occupied regions were significantly lower in the HCT-NRF2i and HT-NRF2i groups than in each control group (Fig. 2C). The AZAN staining, which visualizes erythrocytes within tumor masses, showed that NRF2-inhibited tumors developed fewer blood vessels than the scRNA control (Fig. 2C). Coincident with reduced tumor mass and tumor cell number, caspase-3–positive cell numbers significantly increased in NRF2-inhibited tumors (Fig. 2D). VEGF-immunoreactive cell numbers in NRF2 knockdown tumors were significantly reduced to 43% and 52% of the scRNA control tumors (Fig. 2E).

These results show that stable knockdown of NRF2 in colon cancer cells can suppress tumor growth and angiogenesis; however, it is not clear whether reduced vessel formation is a cause of tumor growth restriction, as small tumor masse develops less numbers of vessels compared to large mass. Therefore, we next assessed cell proliferation to ascertain that diminished angiogenesis contributed to tumor growth retardation. First, we monitored BrdU incorporation and found that NRF2 knockdown did not repress HT29 proliferation (Fig. 3A), whereas HCT-NRF2i cells showed a slight suppression in cell proliferation rate. Similarly, MTT analysis confirmed BrdU result (Fig. 3B). Our cell proliferation analysis appears to be concordant with the tumor growth pattern: NRF2 inhibition completely blocked HCT116-mediated tumor growth, whereas the repressive effect was smaller in HT29 cells. This implies that reduced vessel numbers and VEGF expression could be a cause of tumor growth restriction, at least in HT29 cells. Hence, to investigate the role of NRF2 in angiogenesis, we used HT29 cells for the subsequent experiments.

To verify the effect of NRF2 on angiogenesis, the chick CAM assay was performed following inoculation with matrigel-loaded colon cancer cells. When the branch points of formed blood vessels were counted as a marker of angiogenesis, NRF2-inhibited cancer cells developed fewer blood vessels: the numbers of vessel branch points in HCT-NRF2i and HT-NRF2i cells reduced to 73% and 58% of those in the control cells (Fig. 3C). As additional supportive evidence, we show that NRF2 knockdown HT29 cells displayed reduced endothelial tube formation under hypoxia. In this experiment, HT-SCi and HT-NRF2i cells were incubated in 1% O₂ for 24 hours with supplement-free EBM-2 medium, and conditioned media (supernatants) were collected. Tube formation of HUVECs was determined following the addition of conditioned media. Hypoxia-exposed SCi medium yielded enhanced tube formation, which is comparable to the results observed in the VEGF-treatment group (Fig. 3D). However, conditioned medium from HT-NRF2i cells did not increase tube formation. These results suggest that NRF2 knockdown inhibits angiogenesis; specifically, hypoxia-inducible angiogenesis can be blocked by NRF2-inhibition in HT29 cells.

Transcription factor HIF-1α plays a central role in the induction of angiogenesis-mediating genes, including VEGF. Therefore, we assessed the levels of HIF-1α following hypoxia to characterize the mechanism of antiangiogenic effect of NRF2 knockdown. HT29 cell lines were incubated under hypoxia (1% O₂) for 24 hours and HIF-1α protein levels were determined by immunoblot analysis. In contrast to normal and scRNA-expressing HT29 cells, which show substantial accumulation of HIF-1α protein, HIF-1α protein in NRF2-inhibited HT29 cells did not increase under hypoxia (Fig. 4A). The levels of hydroxyl Pro564-HIF-1α did not show notable changes between the SCi and NRF2i cells. Different HIF-1α protein levels in the scRNA control and HT-NRF2 cells were further evidenced by the analysis of time-dependent changes. HIF-1α accumulation was observed at 8 hours and reached maximal accumulation after 48 hours of incubation; however, the fold increase of HIF-1α protein in NRF2-inhibited cells was much lower throughout incubation (Fig. 4B). A similar pattern was observed in the analysis of HRE-reporter gene expression. Incubation of normal and HT-SCi cells under hypoxic condition for 24 hours led to increased HRE activity, whereas hypoxia-induced luciferase activity in HT-NRF2i cells was 48% of that in the scRNA control (Fig. 4C). Importantly, inhibited HIF-1α accumulation was confirmed in HCT116 cells (Fig. 4D). To verify that the impaired HIF-1α response is a NRF2-specific event, a different type of NRF2 shRNA was tested. For this purpose, cells were transduced with shRNA #1 (a construct used for stable cell establishment; 5′-CCGGGCTCCTACTGTGATGTGAAATCTCGAGATTTCACATCACAGTAGGA-3′) or shRNA #2 (5′-CCGGCCCTGTTGATTTAGACGGTATCTCGAGATACCGTCTAATCAACAGGGTTTTT-3′), and hypoxia-inducible accumulation of HIF-1α was monitored. HIF-1α inhibition was observed with both NRF2 shRNAs, implying that the effect of NRF2 shRNA on HIF-1α is NRF2-specific (Fig. 4E). Taken together, these results indicate that reduced tumor growth and angiogenesis observed in NRF2-inhibited xenografts is associated with diminished HIF-1α accumulation under hypoxia.

To determine the changes in HIF-1α and its target gene expression, transcript levels of HIF-1α, VEGF, PHD2, and HO-1 were analyzed in HT 29 cell lines. Stable inhibition of NRF2 in HT29 cells did not affect basal levels of HIF-1α mRNA and this indicates that the impaired HIF-1α accumulation was not due to repressed expression of HIF-1α (Fig. 4F and G). Of the measured HIF-1α target genes, HO-1 was the only gene for which basal expression was substantially repressed in NRF2-inhibited cells. Although HO-1 expression is known to be under regulatory control by NRF2 and HIF-1α, this result shows that basal HO-1 expression is largely influenced by NRF2. Transcript levels of HIF-1α target genes, including VEGF, PHD2, and HO-1 increased after 24 hour-hypoxia in both cell lines; however, inducible transcript levels were significantly lower in the HT-NRF2i cells than in the HT-SCi cells (Fig. 4F and G). Similarly, immunoblot analysis demonstrated that protein levels of hypoxia-inducible VEGF and PHD2 were lower in NRF2-inhibited cells (Fig. 4H). Next, NRF2 levels were assessed in hypoxic HT29 cells to determine whether induction of the NRF2 system involves the differential HIF-1α accumulation. Hypoxia exposure did not significantly alter nuclear NRF2 levels in both cell lines (Fig. 4I), although the level of nuclear NRF2 is lower in NRF2-inhibited cells. These results show that hypoxia cannot activate HIF-1α signaling in NRF2-knockdown colon cancer cells.

Next we asked the response of cells to CoCl₂ treatment as a hypoxia-mimicking chemical stimulus. Cells were treated with 50, 100, or 150 μmol/L CoCl₂ for 24 hours and HIF-1α accumulation was assessed. CoCl₂ treatment (100 and 150 μmol/L) increased HIF-1α protein in normal and scRNA expressing HT29 cells; however, HIF-1α accumulation was substantially diminished in NRF2-inhibited cells (Fig. 5A). Consistently, when higher concentrations of CoCl₂ (200 and 400 μmol/L) were applied, differential levels of HIF-1α accumulation were observed (Fig. 5B). This defect was mediated by protein destabilization: the treatment with proteasome inhibitor MG132 (10 μmol/L) could restore HIF-1α accumulation in CoCl₂-treaed HT-NRF2i cells (Fig. 5A). As a result of changed HIF-1α levels, HRE-driven luciferase activity was not inducible in NRF2-inhibited cells (Fig. 5C). The relationship between CoCl₂-inducible HIF-1α/VEGF and NRF2 was also confirmed in HCT116 cells (Fig. 5D and E). Of note, HO-1 activity was shown to stabilize HIF-1α protein, and CoCl₂ can increase HO-1 expression via NRF2 and HIF-1α signaling (28). Therefore, it may be possible that repressed HO-1 is linked to the HIF-1α dysfunction in CoCl₂-treated NRF2i cells. Our assessment of HO-1 protein showed that HO-1 induction by CoCl₂ was largely attenuated in NRF2 knockdown cells (Fig. 5F). Nuclear NRF2 levels did not increase in HT-SCi cells treated with 50 and 100 μmol/L CoCl₂, while 150 μmol/L CoCl₂ yielded a marginal increase (Fig. 5G). These results suggest that CoCl₂-mediated NRF2 activation and consequent induction of HO-1 may contribute to the differential response of HIF-1α to high concentrations of CoCl₂ (over 150 μmol/L).

To elucidate the mechanisms underlying the differential HIF-1α response, protein stability of HIF-1α was estimated. When HIF-1α protein accumulation was induced by hypoxia in the presence of MG132, and protein stability was observed in normoxic conditions with protein synthesis inhibitor CHX (50 μg/mL) treatment, HIF-1α in HT-NRF2i cells displayed a more rapid degradation pattern (Fig. 6A). The approximate half-life of HT-SCi HIF-1α is 18 minutes, whereas in HT-NRF2i cells, the half-life is 8 minutes. This discrepancy was much more prominent under hypoxia. HIF-1α decreased slowly in HT-SCi and about 80% remained at 40 minutes; however, HIF-1α in HT-NRF2i cells decreased rapidly and less than 20% remained at 40 minutes (Fig. 6B). These results clearly indicate that HIF-1α degradation is the primary cause of the impaired HIF-1α response to hypoxia.

To verify aforementioned results, first, cells were incubated with proteasome inhibitor MG132 (1 or 10 μmol/L) for 24 hours. MG132 increased HIF-1α in both scRNA control and NRF2-inhibited cells (Fig. 6C), and interestingly, we noted that HIF-1α protein levels were higher in NRF2-inhibited cells. As there was no difference in HIF-1α transcript levels (Fig. 4F and G), it can be hypothesized that enhanced HIF-1α protein synthesis might be occurring in these knockdown cells. Furthermore, when OH-Pro564 HIF-1α was monitored using a specific antibody, hydroxylated HIF-1α accumulation was observed in both cell lines and was particularly high in NRF2i cells (Fig. 6C). This implies that a short half-life of NRF2i HIF-1α in normoxic condition may be the result of higher PHD activity. Second, when cells were incubated in 1% O₂ in the presence of 10 μmol/L MG132, HIF-1α accumulation was completely recovered in comparison to the scRNA control (Fig. 6D). In addition, MG132 treatment led to a notable increase in OH-Pro564-HIF-1α in NRF2-inhibited HT29 cells but not in normal and scRNA controls, indicating that HIF-1α protein in NRF2 knockdown cells is continuously hydroxylated under 1% O₂ hypoxia.

Our collected results indicate that HIF-1α is degraded by hydroxylation under hypoxic condition. Therefore, we examined the direct involvement of PHD in the differential HIF-1α responses to hypoxia. First, we monitored PHD transcript levels and found that levels of PHD1, 2, and 3 were not notably different in these cells (Fig. 6E). This indicates that increased PHD expression does not associate with differential HIF-1α response. Second, to ascertain the involvement of PHD in HIF-1α stabilization, the effect of pharmacological PHD inhibitors on the HIF-1α response was examined. When HT-NRF2 cells were incubated in 1% O₂ in the presence of PHD inhibitor ethyl-3,4-dihydroxybenzoate (EDHB, 150 μmol/L) or deferoxamine (DFO, 150 μmol/L), hypoxia-inducible HIF-1α levels were largely restored (Fig. 6F). Consistent with this result, EDHB treatment prevented HIF-1α degradation and substantially increased the half-life of NRF2i HIF-1α under hypoxia (Fig. 6G). These results show that reduced HIF-1α accumulation in NRF2-inhibited cancer cells is linked to PHD activation even under hypoxic conditions.

It is notable that several reports showed that hypoxia elevates HO-1 expression and this in turn may stabilize HIF-1α through the generation of carbon monoxide (25, 28, 29). Our NRF2-inhibited colon cancer cells exhibit repressed HO-1 expression; therefore, there is a possibility that reduced HO-1 activity mediates the HIF-1α dysfunction. In our estimation of hypoxia-inducible HO-1 protein, although there was some increase in HO-1 mRNA levels (Fig. 4F and G), HT-SCi and HT-NRF2i cells did show notable elevations in HO-1 protein (Fig. 7A). Furthermore, treatment of hypoxic HT-SCi cells with HO-1 inhibitor Zn protoporphyrin (ZnPP, 20 μmol/L) did not diminish HIF-1α accumulation (Fig. 7B, left). In addition, hypoxia-inducible HIF-1α accumulation was not restored by the HO-1 inducer cobalt protoporphyrin (CoPP, 50 μmol/L) in NRF2-inhibited cells (Fig. 7B, right). These results show that HO-1 is unlikely to be linked to the HIF-1α defect in response to hypoxia shown in NRF2-inhibited colon cancer cells.

There is evidence that inhibited mitochondrial function can allow PHD to maintain activity in low O₂ environments (30). This view raises the possibility that NRF2-inhibited cells have reduced mitochondrial function mediating decreased O₂ consumption. As an initial experimental approach, we examined HIF-1α levels following treatment with a mitochondrial inhibitor. HT-SCi cells were incubated in 1% O₂ with or without myxothiazol (10 μmol/L) or rotenone (10 μmol/L) for 24 hours, and HIF-1α was monitored by immunoblot assay. Results show that treatment with mitochondrial inhibitors blocked hypoxia-mediated accumulation of HIF-1α (Fig. 7C). Myxothiazol-repressed HIF-1α accumulation could be restored by proteasome inhibition, and OH-Pro564-HIF-1α increase was observed following treatment with myxothiazol and MG132 (Fig. 7D). These results indicate that mitochondrial dysfunction in HT29 cells can cause continuous hydroxylation and subsequent degradation of HIF-1α in response to hypoxia. Next, in an attempt to link blunted HIF-1α accumulation to the O₂ redistribution hypothesis, we assessed mitochondrial function of NRF2-inhibted colon cancer cells. First, as a marker of mitochondrial function, cellular O₂ consumption was measured using the BD Oxygen Biosensor System. In this assay, fluorescence from oxygen-sensitive tris(1,7-diphenyl-1,10-phenanthroline) ruthenium chloride is reversibly quenched by O₂. When compared with the SCi control, the HT-NRF2i and HCT-NRF2i cells exhibited reduced O₂ consumption by up to 41% and 26% (Fig. 7E). Second, as another marker for mitochondrial function, ATP production was assessed. HT-NRF2i cells exhibited a 31% reduction in ATP production in comparison to the scRNA control cells (Fig. 7F). A similar result was obtained in HCT116 cells (data not shown).

To explain the reduced mitochondrial function, we analyzed several mitochondrial proteins in HT29 cell lines. We examined representative mitochondrial biogenesis factors and found that peroxisome proliferator-activated receptor-γ coactivator 1-α (PGC-1α) and nuclear respiratory factor-1 (NRF-1) showed similar expression levels in HT-SCi and HT-NRF2i cells. In addition, the level of uncoupling proteins (UCP2), which can reduce ATP production by uncoupling the respiratory chain, was not affected by stable NRF2 inhibition (Fig. 7G). Finally, to obtain confirmatory evidence of PHD activation in hypoxia-exposed NRF2 knockout cells, the HT-SCi and HT-NRF2i cells were incubated under anoxic conditions (0.1% O₂) for 24 hours and HIF-1α levels were monitored. Anoxia similarly increased HIF-1α protein levels in both SCi and NRF2i cells, showing that HIF-1α degradation cannot continue in the absence of oxygen (Fig. 7H). Taken together, these results suggest that mitochondrial function is reduced in NRF2-inhibited colon cancer cells, and this appears to be associated with HIF-1α destabilization.

Although the cytoprotective effects of NRF2 signaling have been confirmed by extensive studies performed during the past decade, there is increasing concern regarding the deleterious role of the NRF2 system in cancer cell biology. This question has been promoted by reports that NRF2 is often overactivated in several types of cancer cells. In a report from Padmanabhan and colleagues (31), somatic mutations were identified in the DGR domain of the KEAP1 protein from lung cancer cell lines and in the tumors of lung cancer patients. These mutations caused a glycine to cysteine substitution and led to the loss of KEAP1 repression of NRF2. Another independent study by Singh et al. (32) demonstrated that the KEAP1 protein showed a high frequency of mutations in the Kelch or intervening region domain, and proposed that these mutations are associated with constitutive NRF2 activation and drug resistance in small-cell lung cancer. After these early reports, KEAP1 mutations were found in breast and gallbladder cancers (19, 33). Besides KEAP1, NRF2 was also found to bear somatic mutations in tumors of the lung, esophagus, and skin (34): mutations were located in the DLG and ETGE motifs, which cause impaired recognition by KEAP1-Cul3 E3 ligase. Mutations in KEAP1 or NRF2 have been associated with poor prognosis in non–small-cell lung cancer patients (34). These findings led to the hypothesis that inhibition of NRF2 expression can reverse the phenotypic characteristics of cancer cells such as rapid proliferation, unresponsiveness to apoptosis, and drug resistance. In fact, independent researchers including our group have demonstrated that application of NRF2 interfering RNA could reduce expression levels of antioxidant proteins and the drug efflux system, leading to sensitization of cancer cells to cisplatin, carboplatin, doxorubicin, 5-fluorouracil, and etoposide (27, 34–37). In particular, when NRF2-inhibited lung cancer cells were implanted in nude mice, tumor growth was significantly attenuated in comparison to control cells (37). KEAP1-NRF2 mutations and their involvement in the growth advantages and drug resistance of cancer cells clearly reflect the importance of NRF2 signaling in cancer cell pathobiology.

The increase of HIF-1α protein during hypoxia is largely attributed to the regulation of protein stability (23, 38). HIF-1α stability is mediated by the O₂-dependent regulator PHD and pVHL-26S proteasome. Under aerobic conditions, PHDs (PHD1, PHD2, and PHD3) hydroxylate HIF-1α at specific proline residues (Pro401 and Pro564) in the presence of 2-oxoglutarate and ascorbate (39). Under hypoxic conditions (< 5% O₂), PHD activity is limited by the lack of substrate oxygen, leading to HIF-1α stabilization and accumulation. At the same time, hypoxia disrupts mitochondrial function and a consequent increase in cellular ROS can decrease the level of reduced Fe²⁺, which is a PHD cofactor. This second event also contributes to PHD inhibition in hypoxic environments (40, 41). In addition to O₂-dependent regulation, there is an alternative signal for HIF-1α stability control: hypoxia-induced SUMOylation can alter HIF-1α stability (42). Our study shows that stable knockdown of NRF2 represses tumor growth of xenografts with concomitant reductions in vessel numbers and VEGF expression. However, results from in vivo tumor xenografts do not directly reflect the effect of angiogenesis on tumor growth. Angiogenesis can be facilitated by increasing tumor mass; therefore, reduced VEGF expression and vessel numbers shown in NRF2-inhibited tumors can be the result of retarded tumor growth. In particular, there have been several reports showing that cancer cell growth is diminished by NRF2 knockdown (37). In HT29 cells, the assessment of DNA synthesis and MTT analysis showed that NRF2 inhibition did not show any reductions in these makers of cell proliferation. Further confirmatory assays with the CAM and endothelial tube formation revealed that cell-mediated angiogenesis was significantly repressed by NRF2 knockdown, whereas, in HCT116 cells, we observed that cell proliferation was slightly affected by NRF2 inhibition, although the CAM assay could confirm the inhibitory effect of NRF2 knockdown on angiogenesis. These support that reduced vessel formation in NRF2-inhibited tumors can be the consequence of repressed angiogenesis in HT29 cells. In case of HCT116 cells, repressed cell growth and angiogenesis appear to contribute together to the complete blockade of tumor growth of xenografts (Fig. 2A and B).

As an underlying mechanism of repressed angiogenesis, we demonstrate that hypoxia-induced HIF-1α regulation is disturbed in both NRF2 knockdown HT29 and HCT116 cells. As for the specificity of shRNA for NRF2, we demonstrated that 2 different sequences of NRF2 shRNA showed a similar result (Fig. 4E). An impaired HIF-1α response led to our initial hypothesis that the expression of HIF-1α might be repressed in NRF2-inhibited cells; however, there were no notable changes in HIF-1α mRNA levels in the SCi and NRF2i cells (Fig. 4F and G). Instead, we noted that a proteasome inhibition completely restores HIF-1α accumulation in hypoxic HT-NRF2i cells. Further, MG132 treatment accumulated OH-Pro536 HIF-1α under hypoxia, implying the involvement of PHD (Fig. 6D). In the analysis of protein stability, HIF-1α protein in NRF2-inhibited cells was found to be rapidly degraded under hypoxia compared with the SCi control (Fig. 6B). As direct evidence, pharmacological inhibition of PHD could restore hypoxia-inducible HIF-1α accumulation (Fig. 6F), and extended the half-life of hypoxic HIF-1α in NRF2-inhibited cells (Fig. 6G). These clearly support the participation of PHD in the HIF-1α dysfunction under hypoxia, and raise the possibility that PHD activity may be elevated in NRF2 knockdown cells. However, the levels of PHDs transcripts did not show notable changes between the SCi and NRF2i cells, although the half-life of HIF-1α in HT-NRF2i cells was measured to be shorter than that in the HT-SCi control (Fig. 6A and E). Further experimental results suggest that the O₂ redistribution effect, which is the result of reduced mitochondrial O₂ consumption, can enable PHD activation of HIF-1α hydroxylation in NRF2 knockdown cells (summarized in Fig. 7I).

The effect of intracellular O₂ redistribution on hypoxia-inducible HIF-1α regulation has been demonstrated in many studies (30, 43). Under inflammatory conditions, nitric oxide (NO) production inhibits mitochondrial respiration, resulting in decreased O₂ consumption and activation of PHD in hypoxic condition (30, 44). The O₂ redistribution effect was confirmed in our cell line systems: pharmacological inhibition of mitochondrial function abrogated HIF-1α accumulation under 1% O₂ (Fig. 7C). Accordantly, mitochondrial function monitored by the levels of O₂ consumption and ATP production was relatively low in NRF2 knockdown cells in comparison to the control cells (Fig. 7E and F). As clear evidence, HIF-1α accumulation was observed in anoxic (0.1% O₂) NRF2i cells (Fig. 7H). The underlying mechanisms of reduced mitochondrial O₂ consumption in NRF2 knockdown cancer cells remain unclear. In a brief analysis of the expression of mitochondrial biogenesis factors, no remarkable difference was observed in the levels of PGC-1α and NRF-1 (Fig. 7G). Further detailed mechanism studies will be required to understand this phenomenon; however, we propose 2 hypothetical solutions. First, under stable inhibition of NRF2 signaling, cells should operate the alternative adaptive response to survive under high levels of ROS. Because mitochondria are the major source of endogenous ROS, it might be an indispensible adaptation for cells to inhibit mitochondrial function to diminish ROS generation. Various components in mitochondrial function might be altered in a systemic manner during this adaptation. Second, the potential direct linkage between NRF2-target genes and mitochondrial function is still conceivable. In particular, there is considerable evidence that HO-1 plays a role in the control of mitochondrial function (45). Carbonic monoxide, which is a product of HO-enzymatic action, binds cytochrome c oxidase, leading to enhanced mitochondrial H₂O₂ production. In turn, H₂O₂ likely contributes to the activation of mitochondrial biogenesis. Because we observed that the basal level of HO-1 is substantially suppressed in NRF2-inhibited cell lines, low HO-1 activity may contribute to reduced mitochondrial function and consequent HIF-1α defect in NRF2-inhibited cells. However, hypoxia-inducible HIF-1α accumulation was not altered by pharmacological treatments with HO-1 inhibitor or inducer (Fig. 7B). This indicated that HO-1may not be the core component mediating the HIF-1α dysfunction. The conclusion that NRF2 target genes are not directly linked to the HIF-1α defect and angiogenesis can be further supported by several other observations. First, SFN, a well-characterized activator of NRF2 as well as HO-1, was known to exert antiangiogenic effect, which is conflicting to the proposed role of HO-1 in angiogenesis (46). Antiangiogenic effect of SFN has been explained by the apoptosis-inducing effect via NFκB and FOXO signaling, and the suppressive role in HIF-1α mRNA expression (47, 48). Second, when we investigated the HIF-1α response in the cells with KEAP1 knockdown, which exhibit elevated HO-1 expression, there were no differences in HIF-1α accumulation (data not shown). Our findings present the concept that cellular response systems to O₂ and ROS are tightly coregulated to counteract oxidative insults and promote cell survival, and mitochondria might be the core system coordinating these adaptive procedures.

Although our major focus is on the response of HIF-1α to hypoxia, it is notable that impaired HIF-1α function was also observed following CoCl₂ treatment, which is a hypoxia-mimicking chemical stimulus. The mechanisms underlying the hypoxia-like effect of cobalt are not yet known. However, several explanations have been suggested: depletion of cellular ascorbates, substitution of iron in the porphyrin ring of PHD, and generation of ROS (49). As these mechanisms account for the direct inhibitory effect of cobalt on PHD, the O₂ redistribution effect is unlikely to cause HIF-1α destabilization by CoCl₂. Therefore, in addition to O₂ redistribution, NRF2 knockdown cells might have developed more adaptive changes in HIF-1α signaling: one possibility is the role of HO-1 in HIF-1α stabilization. As a possible explanation, the role of HO-1 can be still speculated. Several reports have demonstrated that CoCl₂ increases HO-1 expression, leading to HIF-1α stabilization and angiogenesis (25, 28, 29). Together with the finding that HO-1 expression is regulated by HIF-1α and NRF2, we can hypothesize that HO-1 might be an important mediator of the interplay between NRF2 and HIF-1α in the CoCl₂ model. In fact, HT-NRF2i cells showed a defect in HO-1 induction with the loss of HIF-1α stabilization following CoCl₂ treatment. This may suggest that NRF2-dependent HO-1 induction takes part in differential accumulation of HIF-1α in the CoCl₂ treatment model. However, as we observed that HIF-1α half-life under normoxia is relatively short, and hypoxia accumulates OH-Pro564 HIF-1α in the NRF2i cells, it is still conceivable that differential PHD activity may participate in the HIF-1α defect upon CoCl₂.

We have not confirmed whether HCT116 and HT29 cells carry mutations in either KEAP1 or NRF2. However, given that NRF2 signaling is essential for the proper cellular response to an oxidizing environment, inhibition of NRF2 signaling in cancer cells could be advantageous to limit the chemosensitivity, cell growth, and apoptosis responses, regardless of NRF2 overactivation. For instance, murine embryonic fibroblasts (MEF) from nrf2-knockout mice were much more sensitive to cisplatin in comparison to wild-type MEFs, although wild-type MEFs are not cancerous cells (35). In addition, nrf2-deficient primary epithelial cells exhibited severe oxidative stress and DNA lesions with the impairment of cell-cycle progression and enhanced apoptotic response (50). Our study showing the association of NRF2 and HIF-1α signaling also supports the concept that NRF2 signaling is important for cancer cell survival and growth. In conclusion, our results suggest that NRF2 inhibition in colon cancer cells inhibits the activation of HIF-1α-VEGF signaling presumably through the adaptive reduction in mitochondrial O₂ consumption, resulting in diminished angiogenesis and tumor growth. This finding provides novel insight into the integrated cellular adaptive response to ROS and O₂, and supports the potential utility of NRF2 siRNA gene therapy to limit tumor growth and restore chemosensitivity.

No potential conflicts of interest were disclosed.

We would like to thank Sarala Manandhar, Chang-won Nam, Dong-ha Shin, and Bo-hyun Choi for their generous help with this study.

This work was supported by the National Research Foundation of Korea Grant (2010-0013857, 2010-0016808) funded by the Korean government.

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.