Targeting both HIF-1 and HIF-2 in human colon cancer cells improves tumor response to sunitinib treatment

Sunitinib is a multitargeted receptor tyrosine kinase inhibitor that exhibits inhibitory activity against multiple targets including c-KIT, vascular endothelial growth factor receptors (VEGFR; VEGFR1, VEGFR2, and VEGFR3), platelet-derived growth factor receptors (α and β), FLT3, CSF-1R, and RET (1–7). Inhibition of these receptor tyrosine kinases blocks transduction of signals important for tumor growth, survival, and angiogenesis. Phase II and III clinical trials with sunitinib have shown clinical benefit, with significantly improved time to progression in patients with imatinib-resistant gastrointestinal stromal tumors and advanced renal cell cancer (6–9). However, this clinical response to sunitinib is heterogeneous. In phase II clinical trials in patients with other solid tumors, including colorectal cancer, breast cancer, non-small cell lung cancer, hepatocellular carcinoma, and neuroendocrine tumors, sunitinib has shown detectable but low objective response rates (10–15). To improve on sunitinib efficacy, combination therapies with cytotoxic chemotherapy, radiation therapy, and targeted agents are currently under active clinical investigation.

It is likely that resistance to sunitinib treatment develops due to the existence of compensatory pathways in both tumor blood vessels and tumor epithelial cells. At the tumor blood vessel level, preclinical and clinical studies have shown that treatment with antiangiogenic or antivascular agents leads to the compensatory induction of proangiogenic factors including VEGF, placental growth factor, interleukin-8, and basic fibroblast growth factor (16–19). These mechanisms may be responsible for tumor blood vessel resistance to sunitinib inhibition of VEGFRs and platelet-derived growth factor receptors. At the tumor epithelial cell level, sunitinib inhibition of receptor tyrosine kinase phosphorylation decreases survival signaling through AKT and extracellular signal-regulated kinase and decreases glucose uptake (20–23). However, inhibition of tumor cell growth is not long-lasting as shown by clinical trials using fluorodeoxyglucose positron emission tomography imaging. In these studies, sunitinib treatment incompletely inhibited fluorodeoxyglucose uptake or induced rebound fluorodeoxyglucose uptake during the initial post-treatment rest period (22, 23).

Overexpression of the hypoxia-inducible transcription factors, HIF-1α or HIF-2α has been strongly associated with cancer progression (24–29). HIF-1 is composed of the HIF-1α and HIF-1β subunits, and HIF-2 is composed of the HIF-2α and HIF-1β subunits. Whereas HIF-1β is constitutively expressed, HIF-1α and HIF-2α protein stability and synthesis are regulated by intratumoral hypoxia, growth factors, and genetic alterations (30–38). Induction of HIF-1 and HIF-2 transactivate gene programs important for inducing tumor angiogenesis, cancer cell growth and survival, and energy metabolism (39–42). As such, tumor induction of HIF-1 and HIF-2 may be a mechanism of escape from the effects of sunitinib therapy.

We hypothesized that inhibition of HIF-1 and HIF-2 in cancer cells would improve response to sunitinib treatment. To test this hypothesis, we generated isogenic human colon cancer cell lines with specific genetic disruption of HIF-1α, HIF-2α, or both HIF-1α and HIF-2α (HIF-αs). The effects of sunitinib treatment, both in vitro and in vivo, were compared in parental and knockout cell lines.

All cancer cell lines were acquired from the American Type Culture Collection. HCT116, LOVO, and HCT15 are human colon cancer cell lines. Isogenic cell lines with somatic disruption of HIF-1α, HIF-2α and VEGF genes were derived from HCT116 (HCT116^HIF-1α-/-, HCT116^HIF-2α-/-, HCT116^{HIF-1α-/-HIF-2α-/-}, and HCT116^VEGF-/-), LOVO (LOVO^HIF-1α-/-), and HCT15 (HCT15^HIF-1α-/-) cells as described previously (43). Cells were grown in McCoy’s 5A medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin (Invitrogen).

Cells were trypsinized, counted, and plated. For cell proliferation studies, cells were harvested daily for 7 days and stained with trypan blue, and viable (nonblue) cells were counted on a hemocytometer. Student’s paired t test was used to determine statistical significance between groups. For analysis of colony formation, cells were allowed to grow undisturbed at 37°C in 5% CO₂ for 10 days and then stained with crystal violet.

Six-week-old female athymic nu/nu mice (Harlan) were implanted with 20 × 10⁶ cells by s.c. injection into the flanks as described previously (43). Tumor sizes in two dimensions were measured with calipers, and volumes were calculated with the formula: (L × W²) × 0.5, where L is length and W is width. Student’s paired t test was used to determine statistical significance between groups. Mice were housed in barrier environments, with food and water provided ad libitum as approved by the University of Michigan Animal Care and Use Committee. Xenografts were harvested for subsequent analyses when they reached ∼0.5 cm³.

For sunitinib treatment, implanted tumors were allowed to grow for 2 to 4 days, and the mice were treated daily by oral gavage with sunitinib malate (Pfizer) at 40 mg/kg/d for 21 days.

Harvested xenografts were fixed in Tissue-Tek OTC compound (Sakura Finetek) and stored at −80°C. Frozen sections, 10 μm in thickness, were prepared with a Leica Microsystems cryostat. For tumor angiogenesis, microvessel density was determined by staining sections with a monoclonal antibody against the endothelial cell-specific marker CD31 (BD Pharmingen) followed by a biotinylated secondary antibody (Jackson ImmunoResearch Laboratories) and rhodamine-streptavidin (Vector Labs). At least three sections from each of three tumors were examined under fluorescence microscopy. Results are illustrated and quantified in Fig. 3 using Image-Pro Plus software (Media Cybernetics).

Mice bearing tumors were injected i.v. with Hoechst 33342 (40 mg/kg) 2 min before sacrifice. Tumors were fixed in Tissue-Tek OTC compound and stored at −80°C. Frozen sections, 10 μm in thickness, were prepared with a Leica Microsystems cryostat and examined under fluorescence microscopy. At least three sections from each of three tumors were examined under fluorescence microscopy. Results are illustrated and quantified in Fig. 3 using Image-Pro Plus software (Media Cybernetics).

HCT116 and HCT116^{HIF-1α-/-HIF-2α-/-} xenografts were harvested at ∼0.4 cm³ and total RNA was extracted. Gene expression analyses on the samples were done at the University of Michigan Comprehensive Cancer Center Affymetrix Core Facility. Commercial high-density oligonucleotide arrays were used (GeneChip Human Genome U133A; Affymetrix) following protocols and methods developed by the supplier.

Total RNA from xenografts were extracted, treated with DNase I, and reverse-transcribed. Real-time PCRs were done in triplicate on reverse transcription-derived cDNA, and relative values were calculated. Student’s paired t test was used to determine statistical significance between groups.

Whole-cell protein extracts were prepared from cell lines, separated by electrophoresis, transferred to nitrocellulose membranes, and probed with antibodies. Antibodies were obtained from Cell Signaling Technologies (peroxidase-conjugated anti-rabbit antibody and peroxidase-conjugated anti-mouse antibody), BD Transduction Laboratories (anti-human HIF-1α), Novus Biologicals (anti-human HIF-2α), and Sigma (anti-α-tubulin). Antibody dilutions were as recommended by the manufacturer.

To determine the effects of specifically inhibiting HIF-1 on tumor response to sunitinib, we generated a series of human colon cancer cell lines that do not express HIF-1α: HCT116^HIF-1α-/-, HCT15^HIF-1α-/-, and LOVO^HIF-1α-/-. Exons 3 and 4 of the HIF-1α gene were targeted for disruption by homologous recombination as described previously (43, 44). Complete loss of HIF-1α protein was confirmed by Western blot analysis (Fig. 1). To determine the effects of inhibiting HIF-2 on tumor response to sunitinib, we disrupted the HIF-2α gene in HCT116 to generate HCT116^HIF-2α-/- cells. Exons 5 and 6 of the HIF-2α gene loci were targeted for disruption by homologous recombination as described previously (43, 44). Complete loss of HIF-2α protein was confirmed by Western blot analysis (Fig. 1). The residual lower bands in the HCT116^HIF-2α-/- and HCT116^{HIF-1α-/-HIF-2α-/-} cell lysates likely represent nonspecific cross-reactivity of the HIF-2α polyclonal antisera to HIF-1β. Specific loss of HIF-2α function is confirmed by the down-regulation of expression of known HIF-2-regulated genes.⁴

Disruption of HIF-1α led to growth delays in the three human colon cancer xenografts tested. HCT116^HIF-1α-/- xenografts had a ∼4.5-day growth delay compared with HCT116 xenografts, HCT15^HIF-1α-/- xenografts had a ∼10-day growth delay compared with HCT15 xenografts, and LOVO^HIF-1α-/- xenografts had a ∼5-day growth delay compared with LOVO xenografts (Fig. 2A-C; Table 1). Disruption of HIF-2α-/- led to ∼1.5-day growth delay in HCT116^HIF-2α-/- xenografts compared with HCT116 xenografts (Fig. 2D; Table 1).

Sunitinib treatment alone led to growth delays in the three parental xenografts, with additional delay in their respective HIF-1α-/- or HIF-2α-/- xenografts. Sunitinib therapy delayed growth in HCT116 xenografts by ∼4 days (Fig. 2A; Table 1). In HIF-1α or HIF-2α disrupted xenografts, sunitinib delayed growth to ∼7.5 and ∼18 days, respectively (Fig. 2A and D; Table 1). This additive inhibition was also observed in the HCT15 and LOVO series of cells (Fig. 2B and C; Table 1).

We speculated that the additive inhibitory effect on xenografts growth was due to improved inhibition of tumor angiogenesis and/or tumor epithelial cells growth. To evaluate tumor blood vessels, tumor angiogenesis and microvessel perfusion were assessed. In HCT116 xenografts, sunitinib treatment significantly decreased both tumor angiogenesis and microvessel perfusion, respectively (Fig. 3). When compared with sunitinib treatment alone, the combination of sunitinib treatment and disruption of HIF-1α did not lead to further decreases in tumor angiogenesis and microvessel perfusion (Fig. 3). These findings were replicated in the HCT15 and LOVO series of cells (Fig. 3). Similarly, when sunitinib treatment was combined with disruption of HIF-2α, there was only a slight tumor blood vessel response when compared with sunitinib treatment alone (Fig. 3). Thus, although the combination of sunitinib treatment plus disruption of HIF-1α or HIF-2α additively inhibits tumor xenograft growth, the combination does not additively inhibit tumor angiogenesis and microvessel perfusion (Figs. 2 and 3). This suggests that the underlying mechanism of additive inhibition might also involve alterations in tumor epithelial cell growth and is examined below in Fig. 4.

HIF-1α and HIF-2α have been shown to share partial structural and functional similarities and bind to HIF-1β to form the HIF-1 and HIF-2 heterodimers, respectively (41, 42, 45–48). HIF-1 and HIF-2 bind to the same hypoxia response element promoter motif and transactivate both unique and overlapping sets of genes, important for tumor angiogenesis and tumor cell growth and survival (42, 49, 50). For these reasons, we hypothesized that combined inhibition of both HIF-1 and HIF-2 would further improve response to sunitinib.

To determine the effects of inhibiting both HIF-1 and HIF-2 on tumor response to sunitinib, we disrupted both HIF-αs genes in HCT116 cells to generate the HCT116^{HIF-1α-/-HIF-2α-/-} cells. Complete loss of both HIF-1α and HIF-2α proteins was confirmed by Western blot analysis (Fig. 1). Disruption of both HIF-αs led to >30-day xenograft growth delay (Fig. 4A; Table 1). Sunitinib treatment of HCT116^{HIF-1α-/-HIF-2α-/-} xenografts further delayed their growth, with sustained complete remission in 50% of mice (Fig. 4A; Table 1). We next examined the effects on tumor angiogenesis and microvessel perfusion. Disruption of both HIF-αs led to a significant decrease in both tumor angiogenesis and microvessel perfusion (Fig. 3). Combining sunitinib treatment with disruption of HIF-αs led to an almost complete inhibition of tumor angiogenesis and microvessel perfusion (Fig. 3).

Because both HIF-1 and HIF-2 are well known to induce VEGF production, we sought to determine if more efficient inhibition of VEGF was responsible for the enhanced antiangiogenic response to sunitinib in the HIF-αs disrupted xenografts. Indeed, VEGF expression was partially decreased in HCT116^HIF-1α-/- and HCT116^HIF-2α-/- xenografts and further decreased in HCT116^{HIF-1α-/-HIF-2α-/-} xenografts when compared with HCT116 xenografts (Fig. 4C).

To more directly test whether superior inhibition of VEGF enhanced tumor response to sunitinib, we disrupted the VEGF gene in HCT116 cells by homologous recombination. VEGF disruption delayed tumor growth by >30 days in HCT116^VEGF-/- compared with their parental counterparts (Fig. 4B). However, addition of sunitinib did not further delay VEGF-/- xenograft growth (Fig. 4B). Tumor vessel analyses show that VEGF disruption significantly decreased both tumor angiogenesis and microvessel perfusion (Fig. 3). Notably, the decrease in tumor angiogenesis and microvessel perfusion in VEGF-/- xenografts was less pronounced than that observed in HIF-1α-/-HIF-2α-/- xenografts (Fig. 3). Sunitinib treatment of HCT116^VEGF-/- xenografts also resulted in a lesser decrease in tumor angiogenesis and microvessel perfusion compared with sunitinib treatment of HCT116^{HIF-1α-/-HIF-2α-/-} xenografts (Fig. 3). These data suggest that the enhanced antiangiogenic effects from combining HIF-αs disruption and sunitinib treatment are not due to the inhibition of VEGF alone.

We next sought to determine whether additional angiogenic factors are regulated by HIF-1 and HIF-2. To systematically identify putative target genes, HCT116 and HCT116^{HIF-1α-/-HIF-2α-/-} xenografts were subjected to global gene expression analyses using an Affymetrix U133A GeneChip. Genes with at least a 4-fold change in expression were identified. As target genes are regulated by the binding of HIF-1 and/or HIF-2 to hypoxia response elements, the genes were further filtered by the existence of the hypoxia response element sequence [(A/G)CGTG] in their promoters.

Disruption of HIF-1α and HIF-2α led to decreased expression of the gene encoding the proangiogenic factor angiopoietin-like protein 4 (ANGPTL4) and increased expression of the gene encoding the antiangiogenic factor thrombospondin 1 (TSP1). To validate these microarray results, ANGPTL4 and TSP1 mRNA expression was measured in tumor xenografts by real-time reverse transcription-PCR. Compared with parental HCT116 xenografts, ANGPTL4 expression was decreased by ∼8-fold in HCT116^{HIF-1α-/-HIF-2α-/-} xenografts (Fig. 4C). Conversely, compared with parental HCT116 xenografts, TSP1 expression was increased by ∼8-fold in HCT116^{HIF-1α-/-HIF-2α-/-} xenografts (Fig. 4C). These data show that HIF-1 and HIF-2 inhibition not only down-regulate VEGF but also down-regulate another proangiogenic factor, ANGPTL4, and also up-regulate the antiangiogenic factor, TSP1. As such, antiangiogenic response is superior when sunitinib is combined with HIF-1 and HIF-2 inhibition compared with VEGF inhibition, because the combination disarms multiple angiogenic pathways.

Disruption of HIF-1α or HIF-2α enhanced tumor response to sunitinib but did not significantly alter tumor blood vessel response (Figs. 2 and 3). This suggested that disruption of HIF-1α or HIF-2α might directly affect tumor epithelial cell growth.

To determine the effects on epithelial cell growth, cell proliferation and clonogenic survival assay were done. In HCT116 cells, disruption of either HIF-1α or HIF-2α led to decreased cell growth, and disruption of both HIF-αs further decreased growth (Fig. 4D and E). These findings in HIF-1α-/- cells were replicated in HCT15 and LOVO cell lines (Supplementary Fig. S2).⁵

These data illustrated that disruption of HIF-1α and HIF-2α in colon cancer cells had direct antitumor effect on cell growth, whereas sunitinib treatment did not. As such, targeting HIF-1 and HIF-2 would complement the antiangiogenic effects of sunitinib by directly inhibiting tumor epithelial cells.

Altogether, our data show that the antitumor effect of sunitinib treatment on colon cancer xenografts is predominantly mediated by inhibition of tumor angiogenesis. In combination with sunitinib, disruption of both HIF-1α and HIF-2α additively inhibited tumorigenesis through two mechanisms. The first mechanism is by modulation of additional angiogenic pathways that complement those affected by sunitinib. The second mechanism is through the direct effects of HIF-1α and HIF-2α on tumor epithelial cell growth. We have shown previously that HIF-1α promotes cell proliferation in the nonhypoxic compartment of tumor xenografts (43). Based on our in vitro data, HIF-2α also promotes cell growth under nonhypoxic conditions. As the nonhypoxic compartment is where tumor cells continue to proliferate because of ongoing delivery of oxygen and nutrients, this compartment might allow for tumor escape despite antiangiogenic therapy. Therefore, targeting tumor epithelial cells directly through inhibition of HIF-1α and/or HIF-2α would complement the antiangiogenic effects of sunitinib.

Our preclinical findings suggest an approach in which sunitinib may be combined with inhibitors of HIF-1 and/or HIF-2 for cancer treatment. Combining sunitinib with HIF-1 and/or HIF-2 inhibition led to additive inhibitory effects in xenografts derived from multiple cell lines. These preclinical findings suggest that this combination in patients would also lead to additive clinical benefit, with improved clinical endpoints including progression-free survival, time to progression, and likely overall survival.

To date, high-throughput small-compound screens and mechanistic studies have identified several classes of anticancer agents that disrupt HIF-1 function, including inhibition of its transcriptional activity and HIF-1α protein synthesis or stability (39, 51–56). Additional screening is ongoing to identify inhibitors of HIF-1α and HIF-1β interactions (57). Based on partial structural and functional similarities between HIF-1 and HIF-2, it is possible that many of the already identified HIF-1 inhibitors would inhibit HIF-2. Our data suggest that it may be advantageous to systematically identify compounds that are effective at inhibiting both HIF-1 and HIF-2 for cancer treatment.

The improved antiangiogenic effect of HIF-αs disruption plus sunitinib treatment is not mediated by an enhanced inhibition of the VEGF/VEGFR axis but likely by the inhibition of another proangiogenic factor, ANGPTL4, and induction of the antiangiogenic factor, TSP1. Our data are consistent with previous preclinical studies showing compensatory induction of other proangiogenic factors in addition to VEGF on initiation of antiangiogenic therapy (17, 18). These data also suggest that combining sunitinib and agents primarily targeting the VEGF pathway might not improve clinical response or benefit. A potential advantage of combining sunitinib with agents targeting HIF-1 and HIF-2 is their effects on complementary angiogenic factors.

In conclusion, our study shows that effective antitumor treatment may be achieved by maximal inhibition of both tumor blood vessel and tumor epithelial cell compartments. However, important questions remain, and further investigations are ongoing in our laboratory. First, the mechanism by which HIF-1 and HIF-2 coordinate and induce tumor angiogenesis through the regulation of proangiogenic factor(s) and antiangiogenic factor(s) remains to be determined. Second, it is not entirely known how HIF-1 and HIF-2 interact and regulate cancer cell growth and survival. It is likely that the mechanisms by which HIF-1 and HIF-2 regulate cell growth and survival are unique and overlapping, and target genes controlling nutrients metabolism, growth, and survival pathways need to be identified. Further mechanistic studies in these areas will likely yield additional targets for drug development and novel combinations for cancer treatment.

No potential conflicts of interest were disclosed.

We thank Drs. Jacques Nor, Lawrence Gardner, William Ensminger, Mark Zalupski, and Douglas Blayney for critical review of the article; Joe Washburn, James McDonald, and Kerry Innes (University of Michigan Comprehensive Cancer Center Microarray Core Facility) for assistance with microarray analyses; and Linda Hartwig and Pamela Varga for administrative assistance.