Gain of HIF-1α under Normoxia in Cancer Mediates Immune Adaptation through the AKT/ERK and VEGFA Axes

Immunotherapy has emerged as a promising approach for the clinical management of cancer. However, in many cases, it has been observed that the generation of a tumor-specific immune response does not translate into tumor regression in patients with cancer. A potential explanation for this is the overexpression by tumor cells of proteins that bestow them with enhanced survival, proliferation, and invasion capacity (1). In particular, hypoxia-inducible factor 1 (HIF1) is a key orchestrator of diverse biochemical pathways, from proliferation and survival to angiogenesis and invasion (2). Overexpression of HIF1 has been reported in virtually most of carcinomas (3), and HIF1 has been shown to drive cancer progression as well as resistance to chemotherapy and radiotherapy (2). Moreover, resistance of tumor cells to killing by NK cells or T cells has been reported to occur through HIF1 under hypoxia (4–9). Although HIF1 is a gateway to cancer progression, the manner in which it initially arises within tumor cells remains unknown. In fact, HIF1 is exquisitely sensitive to oxygen tension and is typically only present in stable form under hypoxia (10). Here, we found unexpectedly that stable expression of HIF1 in tumor cells occurs even under normal oxygen tension. We inferred that gain of HIF1 is a key element of cancer evolution that arises from selection pressure imposed by an antitumor immune response.

To explore this idea, we examined cancer evolution in the context of immune surveillance. Adaptation to immune defenses, in particular, those mounted by CD8⁺ cytotoxic T lymphocytes (CTL), has emerged as an early, indispensable, and host-intrinsic event in cancer progression (11). Thus, immune surveillance is an ideal selection pressure for the analysis of cancer evolution. We invented a system referred to as VICE, for Vaccination-Induced Cancer Evolution, in which variants of a parental tumor are derived through serial rounds of immune selection either in culture or in mice (12).

We used VICE to explore the role of HIF1 in cancer evolution under immune surveillance. Here, we show that the α subunit of HIF1 (HIF-1α) becomes markedly elevated during immune selection even under normoxia, and HIF-1α expression by tumor cells dictates the ability of cognate CTLs to control tumor growth. To our knowledge, gain of HIF-1α in tumor cells under normoxia in response to immune selection has not been previously reported. We found that the effects of HIF-1α on immune adaptation are transmitted through VEGFA-mediated activation of the AKT and ERK pathways, which induce the expression of a constellation of antiapoptotic molecules in tumor cells. Blockade of each of these pathways abrogated resistance of tumor cells to lysis by cognate CTLs, underscoring the importance of the HIF-1α/VEGFA axis in immune adaptation.

HPV-16 E7⁺ cells [TC-1, TC-1 P3, TC-1 P3 (A17), TC-1/no insert, and TC-1/HIF-1α] were used as a mouse tumor model. The production and maintenance of TC-1 (13) and TC-1 P3 A17 cells (14, 15) has been described previously. TC-1/HIF-1α cells were generated with the pMSCV/HIF-1α K532R vector (for TC-1/HIF-1α). For the production of human immune-resistant tumor cells, 10⁶ CaSki (CaSki P0) cells were pulsed with 10 μg/mL HLA-A2–restricted MART-1 M27 peptide (AAGIGILTV) for 2 hours and mixed with KKM MART-1–specific human CD8⁺ T cells (gift from Dr. Cassian Yee, MD Anderson Cancer Center, TX, USA) in a 6-well plate at 1:1 ratio for 4 hours. Surviving CaSki cells were expanded and termed CaSki P1 cells. The above procedure was twice repeated to create the CaSki P3 cells. Nonspecific antigen NY-ESO-1 (SLLMWITQC) peptide-pulsed CaSki cells pulsed with the irrelevant peptide NY-ESO-1 (SLLMWITQC) were incubated with MART-1–specific human CD8⁺ T cells to derive control CaSki cells (CaSki N1 and CaSki N3). For hypoxia culture, cells were maintained at 1% O₂ in a controlled atmosphere chamber (PLAS-LABS) with a gas mixture containing 1% O₂, 5% CO₂, and 94% N₂ at 37°C. Cells were exposed to hypoxia for 14 hours before analysis unless otherwise indicated.

Six- to 8-week-old female C57BL/6 mice were acquired from Daehan Biolink. Mice were maintained and handled under protocols approved by the Korea University Institutional Animal Care and Use Committee (KUIACUC-2013-210; Seoul, Korea).

Pharmacologic agents used to inhibit signaling axes were LY294002 (Calbiochem Corp, 25 μM) for the PI3K pathway, SB203580 (Calbiochem Corp, 10 μM) for p38 MAPK pathway, and PD98059 (Stressgen, 10 μM) for the ERK pathway.

To produce the pMSCV/human HIF-1α vector, DNA fragments encoding the human HIF-1α K532R mutant were PCR amplified from pCMV/HIF-1α K532R (gift from Dr. Lee Mi Ok, Sejong University, Seoul, Korea) with primers (5′-ATGGAGGGCGCCGGCGGCG-3′ and 5′-CAGAATTCTCAGTTAACTTGATCCAAA-3′). The amplified fragments were cloned into the BglII/EcoRI sites of the pMSCV retroviral vector (Clontech). Synthetic siRNA specific for HIF-1α was synthesized with a 2′-O-ACE-RNA phosphoramides backbone (Invitrogen). The sequences of mouse HIF-1α siRNA are 5′-CCAGATCTCGGCGAAGTAA-3′ (sense) and 5′-UUACUUCGCCGAGAUCUGG-3′ (antisense), human HIF-1α siRNA are 5′-CCUAUAUCCCAAUGGAUGAUG-3′ (sense) and 5′-CAUCAUCCAUUGGGAUAUAGG-3′ (antisense). A total of 10⁵ tumor cells were transfected in 6-well plates with 100 pmol of synthesized siRNAs with Lipofectamine 2000 (Invitrogen).

Total RNA was extracted with the RNeasy Mini Kit (Qiagen) and treated with DNase (Ambion). cDNA synthesis was performed with RT&Go Mastermix (MP Biomedicals), and real-time PCR was performed with Lightcycler FastStart DNA SYBR Green Master Mix (Roche) using mouse and human Vegfa-specific primers with the following sequences (Bioneer): mouse Vegfa, 5′-TGCACCCACGACAGAAGGA-3′ (forward) and 5′-GGCAGTAGCTTCGCTGGTAGAC-3′ (reverse); human Vegfa, 5′-CTGCTGTCTTGGGTGCATTGG-3′ (forward) and 5′-CACCGCCTCGGCTTGTCACAT-3′ (reverse). All data were normalized to expression level of β-actin mRNA.

Total RNA from TC-1 and CaSki cells was purified using TRIzol reagent (Invitrogen). Real-time PCR was conducted to detect Vegfa with TaqMan Universal SYBR Green Master Mix (Roche) using the mouse and human primer sets for Vegfa.

Samples probed with primary antibodies against phospho-AKT (Ser473), AKT, phospho-ERK (T202/Y204), ERK, p38 MAPK, BCL-w, BID, BIM, BAD, phospho-BAD (Ser 136), XIAP, Dual phospho-p38 MAPK (Cell Signaling Technology), HIF-1α, HIF-2α, Bcl-2, Bcl-xL, Bax, (Santa Cruz Biotechnology), BAK (BD Biosciences), or E7 (gift from Dr. Ju-Hong Jun, Seoul National University, Seoul, Korea). Antibodies were diluted 1:1,000, and bands were visualized by enhanced chemiluminescence (Elpis Biotech).

VEGFA concentration in supernatant of TC-1 P0, P3 (A17) and CaSki P0, P3, N3 cells collected 48 hours after incubation was measured with mouse and human VEGFA Quantikine ELISA Kit (R&D Systems). Supernatant of HIF-1α siRNA-treated cells was collected after 72 hours. Supernatant of cells treated with VEGFA-neutralizing Ab (100 ng/mL) was collected after 24 hours. VEGFA concentration was normalized to the number of cells in each sample and depicted as pg of VEGFA per 10⁶ cells.

TC-1 P0, P1, P2, P3, N1, N2, and N3 cells were fixed in 4% paraformaldehyde for 10 minutes. Cells were washed with PBS, treated with 0.2% Triton X-100, and blocked for 1 hour in 1% BSA. Cells were incubated overnight in the presence of primary mAb against HIF-1α in a humidified chamber at 4°C. Cells were stained with Alexa Flour 555-labeled goat anti-rabbit IgG. Nuclei were stained with SYTOX Green (Invitrogen). Expression of HIF-1α was examined by confocal laser scanning microscopy (Carl Zeiss).

TC-1 cells or CaSki cells were labeled with CFSE (10 μmol/L) in DMEM supplemented with 0.1% FBS. Cells were incubated for 10 minutes at 37°C and then pulsed with E7, NY-ESO-1, or MART1 peptide (10 μg/mL) for 1 hour, respectively. Tumor cells were then mixed with E7 or MART-1–specific CD8⁺ CTLs at a 1:1 ratio and incubated for 4 hours at 37°C. Cells were stained for active caspase-3 as an index of apoptosis and examined by flow cytometry. For CTL-mediated assay upon VEGFA blockade, VEGFA-neutralizing mAb (100 ng/mL) was added to cells 24 hours before the assay.

C57BL/6 mice were inoculated subcutaneously with 10⁵ TC-1 P0/no insert or TC-1 P0/HIF-1α cells. For immunization experiments as previously described (14), mice were administered with Vac-SEL days after tumor challenge. For adoptive transfer experiments, mice were administered with 2 × 10⁶ E7-specific CTLs or isotonic saline control by tail vein injection 7 days after tumor challenge. Tumor dimensions (length and width) were measured with a calipers every 3 to 4 days, and tumor volume was calculated as length (mm) × width² (mm²) × 0.52.

All data are representative of at least three independent experiments. Individual data points were compared using the Student t test. Statistical analysis was performed with SPSS version 17.0 software (SPSS Inc.). Values of P < 0.05 were considered statistically significant.

To investigate the molecular evolution of cancer under selection pressure from an antitumor immune response, we created a system termed VICE (Vaccination-Induced Cancer Evolution), in which the tumor is subjected to serial rounds of immune selection by antigen-specific CD8⁺ CTLs (12). We administered C57BL/6 mice with vaccinia virus encoding the E7 antigen of human papillomavirus type-16 linked to the sorting signal of Lamp-1 (Vac-SEL) to elicit CD8⁺ CTLs against E7 (16). We then inoculated the mice with parental TC-1 tumor cells (P0), which harbor E7 antigen. Outgrowth tumor was excised and inoculated into naïve mice that previously received Vac-SEL. This process was repeated for three cycles to derive the P3 line, which was impervious to lysis by CTLs in vitro and in vivo. As a control, we performed this procedure in mice implanted with TC-1 cells without vaccination to generate N1, N2, and N3 cells without immune selection. Notably, expression of E7 as well as MHC class I was nearly identical in P3 and P0 cells, indicating that the immune resistance of P3 was not due to antigen loss (14). Although the HIF family of proteins has been implicated in chemotherapy resistance, its role in resistance to CTLs has not been previously examined under normal oxygen tension (2). We therefore probed for expression of HIF-1α and HIF-2α protein in P0 and P3 cells. Although there was no difference in HIF2α expression between P0 and P3 cells, HIF-1α was upregulated nearly 20-fold in P3 relative to P0 cells under normoxia; this was confirmed in a clone from the P3 line (A17; Fig. 1A). Under hypoxia, HIF-1α expression was also elevated in P3 relative to P0 cells (Fig. 1B). To verify evolution toward HIF-1α expression under immune surveillance, we measured HIF-1α levels in tumor cells at different stages of selection (P0 to P3). We observed a consistent and gradual increase in HIF-1α from P0 to P3; in contrast, HIF-1α levels remained constant in tumor cells at parallel stages without immune selection (N0 to N3; Fig. 1C). We thus conclude that HIF-1α expression is induced in tumor cells by immune selection.

The overexpression of HIF-1α could be either due to elevated synthesis of basal HIF-1α by all tumor cells or to an increase the number of HIF-1α⁺ cells. To address this issue, we labeled P0 to P3 cells with fluorescence-labeled mAb against HIF-1α and determined the frequency of tumor cells with nuclear localization of HIF-1α by fluorescence microscopy. We observed a progressive increase in the frequency of nuclear HIF-1α⁺ tumor cells as immune selection progressed from the P0 to P3 stages (from 10%–80% of all tumor cells; Fig. 1D). Our results therefore demonstrate the molecular evolution of tumor cells toward HIF-1α under the immune selection pressure imposed by vaccination.

Because HIF-1α is upregulated in tumor cells during immune surveillance, we reasoned that this expression might be responsible for immune adaptation. To test this, we transfected P3 cells with siRNA against either HIF-1α (siHIF-1α) or GFP control (siGFP), mixed the cells with E7-specific CTLs, and measured the frequency of apoptotic tumor cells. We observed a higher frequency of apoptotic tumor cells in the siHIF-1α- compared with the siGFP-transfected group (Fig. 2A), suggesting a role for HIF-1α in immune resistance. To confirm this, we transduced P0 cells with a human HIF-1α K532R mutant (P0/HIF-1α) that is stable in normoxia and characterized their phenotype (Fig. 2B). There was no significant difference in surface MHC class I expression between P0/HIF-1α cells or their empty vector–transduced counterparts (P0/no insert; Fig. 2C). Furthermore, the activation and viability of E7-specific CTLs were similar when mixed with HIF-1α or vector-transduced cells (Fig. 2D and E). However, the frequency of apoptotic tumor cells was lower among P0/HIF-1α compared with P0/no insert cells when mixed with either E7-specific CTLs or with granzyme B (Fig. 2E and F).

To assess the role of HIF-1α in tumor immune adaptation in the setting of a host immune response, we inoculated C57BL/6 mice with P0/no insert or P0/HIF-1α, and then administered the mice with Vac-SEL 7 days later. Tumor growth was sharply accelerated in the vaccinated mice inoculated with P0/HIF-1α relative to P0/no insert cells (Fig. 2G). The presence of HIF-1α also rendered the tumor impervious to adoptive therapy with E7-specific CTLs (Fig. 2H). Notably, there was no significant difference in tumor growth between mice inoculated with P0/HIF-1α and P0/no insert cells without adoptive therapy, suggesting that HIF-1α facilitates cancer progression primarily by promoting immune adaptation.

We hypothesized that HIF-1α may cause immune adaptation by establishing an antiapoptotic state in tumor cells. To test this, we probed for the expression of a panel of pro- or antiapoptotic molecules in P3 cells transduced with siHIF-1α (relative to siGFP control) or in P0 cells transduced with HIF-1α (relative to empty vector control). We found that an array of antiapoptotic molecules, including Bcl-2, Bcl-xL, Bcl-W, XIAP, cIAP1, and cIAP2, were induced by HIF-1α (Fig. 3A). To elucidate the pathways by which HIF-1α controls expression of these molecules, we measured phosphorylation of AKT, ERK, and P38. We found that HIF-1α drives phosphorylation of AKT and ERK (Fig. 3B), and that pharmacologic inhibition of these pathways strongly reduces expression of antiapoptotic molecules (Fig. 3C) and renders tumor cells sensitive to lysis by CTLs (Fig. 3D). Therefore, HIF-1α-mediated immune adaptation occurs through the AKT and ERK signaling pathways, which trigger the activation of an antiapoptotic network within tumor cells.

We next sought to uncover the connection between HIF-1α expression and activation of AKT and ERK signaling. It has previously been shown that HIF-1α drives expression of its key target VEGFA, which is reported to induce the expression of antiapoptotic molecules in tumor cells (17, 18). We thus measured VEGFA expression and secretion in P0 versus P3 cells, and found both to be over 10-fold higher in P3 relative to P0 cells (Fig. 4A and B). We confirmed that HIF-1α controls expression of VEGFA by knocking down this transcription factor in P3 cells with siRNA (Fig. 4A and B), or by introducing it into P0 cells (Fig. 4C and D). To explore whether VEGFA triggers AKT and ERK signaling and the downstream antiapoptotic network, we incubated P3 or P0/HIF-1α cells with VEGFA blocking mAb and then measured AKT and ERK phosphorylation, as well as expression of antiapoptotic molecules. We observed reduced AKT and ERK phosphorylation among P3 or P0/HIF-1α cells incubated with α-VEGFA mAb relative to isotype control, together with lower expression of antiapoptotic molecules (Fig. 4E and F). Consistent with these data, P3 and P0/HIF-1α cells incubated with α-VEGFA mAb were more sensitive to killing by E7-specific CTLs compared with cells incubated with isotype control (Fig. 4G). We conclude that a core signaling axis linking HIF-1α, VEGFA, AKT/ERK, and an array of antiapoptotic molecules drives cancer immune adaptation.

Having established that the VEGFA-AKT/ERK axis induces HIF-1α-driven immune adaptation in a mouse model, we next sought to determine whether this is also true in human cancer. To do so, we pulsed parental CaSki human cervical cancer cells (P0) with a model peptide antigen, MART-1, (or an irrelevant antigen, NY-ESO-1 peptide) and then subjected these cells to immune selection with MART-1–specific human CTLs. After three rounds of selection, CaSki cells pulsed with MART-1 were termed P3, whereas those pulsed with irrelevant NY-ESO-1 peptide were termed N3. We mixed P0, P3, or N3 cells with MART-1–specific CTLs and determined the frequency of apoptotic tumor cells by flow-cytometric analysis of caspase-3 activation. We found that there were over two times less apoptotic P3 cells compared with either N3 or parental P0 cells, confirming, as expected, that immune selection of human cancer cells promotes immune adaptation (Fig. 5A). We also probed for HIF-1α expression in P0, P3, or N3 cells by Western blot analysis and observed over 10 times higher HIF-1α expression in P3 cells compared with either N3 or P0 cells (Fig. 5A). We also created two additional immune-resistant human tumor model systems by in vitro selection of parental MDA-MB-231 breast cancer cells or 526mel melanoma cells; we observed HIF-1α overexpression by immune-resistant cells compared with their parental counterparts in both of these systems (data not shown). Consistently, VEGFA mRNA level and protein secretion, AKT and ERK phosphorylation, and expression of the antiapoptotic factors cIAP-1 and cIAP-2, were upregulated in P3 cells relative to N3 and P0 cells as well (Fig. 5A and 5B and Supplementary Figs. S1 and S12). These results demonstrate that immune selection induces HIF-1α expression, VEGFA secretion, and activation of the AKT/ERK axis in human cancer cells, which prompts the expression of antiapoptotic factors. We found that silencing HIF-1α with siRNA reduced VEGFA secretion, AKT/ERK phosphorylation, and cIAP expression in CaSki P3 cells, showing that this sequence of signaling events is dependent on the presence of HIF-1α in immune-resistant human cancer cells (Fig. 5C and D). HIF-1α knockdown also rendered immune-resistant, MART-1-pulsed CaSki cells sensitive to lysis by human MART-1–specific CTLs (Supplementary Fig. S3). Likewise, HIF-1α knockdown in immune-resistant MDA-MB-231 or 526mel cells rendered them sensitive to these CTLs (data not shown). Furthermore, blockade of the VEGFA pathway with α-VEGFA mAb diminished AKT/ERK phosphorylation and cIAP expression, and restored sensitivity of P3 cells to MART-1–specific CTLs, illustrating that HIF-1α triggers the AKT/ERK axis, antiapoptotic signaling, and immune adaptation directly through VEGFA (Fig. 5E and F) in human cancer.

The immune adaption of tumor cells in response to host immune defenses has emerged as a major driving force for cancer onset and progression (19). This process may occur through host-intrinsic events, such as the induction of tolerance to tumor antigen in tumor-specific CTLs or the recruitment of regulatory T cells or myeloid-derived suppressor cells to the tumor microenvironment (20). Alternatively, it may be controlled by tumor-intrinsic events, such as loss of antigen or gain of resistance to apoptosis (21, 22).

Here, we demonstrate cancer immune adaption in response to a selection pressure imposed by tumor-specific CTLs can be triggered via the HIF-1α/VEGFA pathway.

We found that expression of HIF-1α is profoundly upregulated in tumor cells subjected to immune selection by tumor-specific CTLs compared with those without selection. Expression of HIF-1α in tumor cells mediated resistance to lysis by CTLs both in a culture system and in mice in a tumor-intrinsic manner; the activation, survival, and function of CTLs were not influenced by the HIF-1α expression status of tumor cells. Furthermore, we did not detect any effects of HIF-1α expression on either MHC class I expression or antigen processing and presentation in tumor cells, suggesting that HIF-1α conferred a direct survival capacity to tumor cells. We found that HIF-1α upregulated expression of a constellation of antiapoptotic molecules—Bcl-xL, Bcl-W, XIAP, cIAP1, and cIAP2—in an AKT- and ERK-dependent manner. Furthermore, the effects of HIF-1α on AKT/ERK activation and tumor survival were mediated through VEGFA signaling, as VEGFA blockade abolished the antiapoptotic properties of HIF-1α.

It is well established that HIF-1α stability is exquisitely sensitive to oxygen tension, such that HIF1 is typically only present in abundant amounts under hypoxia. In the context of the tumor microenvironment, the core of a solid tumor mass is characterized by hypoxia, and thus, HIF1 expression is frequently high in these areas. However, our data suggest that HIF1 expression is not only limited to the core of the tumor mass, but may also extend into the marginal areas of the tumor (which have normal oxygen tension). What would be the benefit for tumor cells of HIF1 expression in these areas? We believe that, because tumor cells in the periphery are the first to encounter infiltrating lymphocytes, these cells would be the most susceptible to immune attack. Therefore, gain of HIF-1α by these marginal tumor cells would confer a strong “fitness” advantage, potentially explaining the evolution of tumor cells toward HIF1 expression even under normoxia in response to an immune selection pressure. Independently, HIF1 expression under normoxia would induce VEGF secretion and promote angiogenesis, thereby contributing to tumor progression (3, 23). For instance, HIF-1α under normoxia has been implicated in patients with highly vascularized hemangioblastoma, in which tumor cells immediately adjacent to blood vessels were found to stain intensely for HIF-1α (3, 23, 24). Future work should aim to explore these theories.

Finally, it is intriguing to consider our results in light of the “adaptation model” of the immune system that has been recently proposed (25), in which the expression of certain receptors and corresponding ligands by immune cells or peripheral tissue cells tunes the magnitude of the immune response. In this context, HIF-1α may represent an adaptation-inducing factor that turns on the expression of ligands that directly modulate lymphocyte activation. This will be an interesting area for future investigation.

In conclusion, our study unravels a signaling pathway composed of VEGFA, AKT/ERK, and antiapoptotic molecules linking HIF-1α to cancer immune adaptation. Inhibitors of each component of this pathway should, in principle, be effective against cancer, particularly as adjuvants for immune-based therapies.

No potential conflicts of interest were disclosed.

Conception and design: Y.-H. Lee, H.C. Bae, K.-H. Song, S.-K. Ye, T.-C. Wu, T.W. Kim

Development of methodology: Y.-H. Lee, H.C. Bae, K.H. Noh, T.-C. Wu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y.-H. Lee, H.C. Bae, K.H. Noh, K.-M. Lee

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y.-H. Lee, H.C. Bae, K.H. Noh, K.-H. Song, C.-P. Mao, T.W. Kim

Writing, review, and/or revision of the manuscript: H.C. Bae, C.-P. Mao, T.-C. Wu, T.W. Kim

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

Study supervision: K.-M. Lee, T.W. Kim

This work was funded by the National Research Foundation of Korea (2012R1A2A2A01007527, 2013M3A9D3045881, and 2009-0086652), the Korea Healthcare Technology R&D Project (HI11C-0052-030013), the United States NIH Cervical Cancer Specialized Program of Research Excellence (SPORE; P50 CA098252), Head and Neck Cancer SPORE (P50 CA96784-06), and R01 grant (CA114425-01). Chih-Ping Mao is a recipient of the NIH F30 Research Fellowship (CA177221).

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.