Targeting Hypoxia-Induced Carbonic Anhydrase IX Enhances Immune-Checkpoint Blockade Locally and Systemically

Immune-checkpoint blockade (ICB) with antibodies blocking CTLA-4 or PD-1 has now received FDA approval for multiple solid tumor types (1). Considerable response rates have been achieved in a subset of patients with metastatic melanoma (2, 3) and, to a lesser degree, patients with triple-negative breast cancer (TNBC; refs. 4, 5) when treated with ICB alone. Response rates in TNBC have seen considerable improvement when ICB was combined with chemotherapy (6). Despite this, many patients fail to respond or develop resistance, suggesting that a significant population of patients stand to benefit from enhancing response rates to ICB. Identifying determinants of response is critical to extending ICB benefit to a greater patient population. Although increased tumor mutational burden has been found to correlate with response to immunotherapy, it alone fails to explain the heterogeneity in responses (7–9). It is critical to understand the level of immune infiltration in the tumor and the environment in which the lymphocytes reside to predict which patients stand to achieve the greatest benefit from ICB (10, 11).

The tumor microenvironment (TME) has been identified as a critical hurdle to immune infiltration and activity (12). Physical constraints within the TME lead to decreases in perfusion and oxygen delivery, triggering the hypoxic response and stimulating a further increase in glycolytic activity (13). Elevated glycolytic activity of solid tumors results in increased nutrient competition between the tumor and immune cells (14), and increased extracellular accumulation of glycolytic metabolites, such as lactate, protons, and carbonic acids, leads to the acidification of the TME, further reducing normal immune function (13, 15–17). Indeed, the accumulation of lactate in the TME inhibits the efflux of lactate from cytotoxic T cells and blunts the production of cytotoxic effectors (15). Neutralization of acidity with bicarbonate within the TME or inhibiting lactate production in the tumor by reducing LDHA expression has been shown to increase immune activity and enhance the efficacy of ICB (18, 19). However, these approaches lack specificity for the tumor and are faced with challenges clinically (13, 20). Thus, there remains an unmet need to specifically restore pH homeostasis within the TME, while leaving the immune system unaffected.

Carbonic anhydrase IX (CAIX) is a hypoxia-induced, extracellular facing, cell-surface enzyme involved in pH regulation of hypoxic solid tumors (17). CAIX hydrates carbon dioxide to produce bicarbonate and protons. The protons accumulate in the extracellular space and contribute to the acidification of the TME, while the bicarbonate, via bicarbonate transporters, is returned inside the cell, facilitating the titration of intracellular pH. CAIX plays a role in tumor growth and survival, invasion and metastasis, and mobilization of immune-suppressive myeloid-derived suppressor cells, making it an attractive therapeutic target (21). Therapeutically targeting CAIX in hypoxic solid tumors with the specific small-molecule SLC-0111 inhibits tumor growth and metastases in preclinical models of breast and brain cancer (22–24). SLC-0111 has now entered clinical evaluation and has, thus far, been found to be well tolerated (25). Given its role in pH regulation within the hypoxic niche, targeting CAIX may restore pH homeostasis of the TME, relieving a critical hurdle to immune activity (23, 24, 26–28).

Herein, we investigated whether CAIX inhibition would increase the efficacy of ICB. We identified that CAIX was associated with increased tumor grade, risk of metastasis, and was independently predictive of worse overall survival in a melanoma patient cohort. We determined that CAIX inhibition reduced the capacity of melanoma and breast cancer cells to acidify the extracellular environment, leading to enhanced immune activity. We found that CAIX inhibition augmented anti–PD-1 and anti–CTLA-4 blockade to reduce melanoma tumor growth and breast cancer metastasis. Finally, we identified that CA9 expression was associated with decreased immune activity in the tumors of patients with a broad spectrum of solid malignancies, including metastatic melanoma and basal-like breast cancer.

The murine mammary adenocarcinoma 4T1 (CRL-2539) and murine skin melanoma B16F10 (CRL-6475) cell lines were purchased from the American Type Culture Collection and validated by STR analysis. 4T1 cells were maintained in DMEM (Gibco, #11995-065) plus 10% FBS (Gibco, #12483020) (D10) and 1× nonessential amino acids (Gibco, #11140-050). B16F10 cells were maintained in D10. Splenocytes were maintained in RPMI-1640 (Gibco, #11835-030) plus 10% FBS, 50 μmol/L β-mercaptoethanol (Sigma, #M3148), 1× penicillin and streptomycin (Gibco, #15140-122), and IL2 (50 U/mL; PeproTech, #200-02-10UG). Stable knockdown of CAIX was achieved by lentiviral transduction of short hairpin RNAs toward Car9 (23). Sequences contained within the constructs utilized are shNS: CTTACTCTCGCCCAAGCGAGAG; shCar9_1: V2LMM_48299—TAACTTCAGGTGGATCCTC; shCar9_2: V2LMM_57078—TTTCTTCCAAATGGGACAG. Cells were incubated with lentivirus for 24 hours prior to undergoing selection with puromycin for 72 hours. All cultures were maintained at 37°C under a 5% CO₂ atmosphere, and cell lines were cultured for a week prior to the initiation of an experiment. For studies involving hypoxia, cells were maintained in a 37°C incubator in a nitrogen-balanced atmosphere of 1% O₂ and 5% CO₂ and were routinely monitored for hypoxia-induced CAIX expression. Cell lines were routinely tested for Mycoplasma using the LookOut Mycoplasma PCR detection kit (Sigma-Aldrich; MP0035).

Antibodies to PD-1 (RMP1-14) and CTLA-4 (9H10) and isotype controls were purchased from Bio X Cell. Antibodies to 4-1BB/CD137 were purchased from R&D Systems (Supplementary Table S1). The ureido-sulfonamide CAIX inhibitor, SLC-0111 was previously described and provided by Welichem Biotech Inc. (23).

All studies involving mice were performed in accordance with and with the approval of the University of British Columbia Institutional Animal Care and Use Committee under approved animal study protocol A14-0058.

For studies done using the 4T1 model, 1 × 10⁶ tumor cells in 50 μL PBS were inoculated subcutaneously into the left fourth mammary fat pad of female Balb/c mice (7–9-week-old; Simonsen Laboratories). Tumor growth was tracked by digital caliper three times per week, and treatments were initiated once an average tumor volume for the cohort reached 100 mm³ using the modified ellipsoid formula (l × w² × π/6). Mice were then randomly distributed among groups in a manner to maintain equal size distributions across each treatment group. Treatment with SLC-0111 (50 mg/kg) or drug vehicle was initiated on day 1, and treatments were administered daily until study completion by oral gavage. The oral formulation (drug vehicle) of SLC-0111 consisted of 55.6% (w/w) phospholipon (Lipoid), 7.3% vitamin E TPGS (Antares Health Products Inc., #TG0101NF), 11.9% polyethylene glycol (PEG) 200 (Sigma, #P3015-500G), 16.3% PEG400 (Sigma, #202398-500G), and 8.9% propylene glycol (Sigma, #P4347-500ML). Antibodies to PD-1 (10 mg/kg), CTLA-4 (10 mg/kg), or isotype controls (10 mg/kg) were administered intraperitoneally (i.p.) on days 1, 3, 5, 7, 9, and 11.

For studies done using the B16F10 model, 5 × 10⁵ tumor cells in 100 μL PBS were inoculated subcutaneously onto the back on female C57Bl/6J mice (7–9 week-old; The Jackson Laboratory). Tumor growth was tracked as stated above. Due to the growth kinetics of the B16F10 tumor model and the rapid onset of signs of morbidity in the mice bearing these tumors, treatment was initiated when tumors were palpable in order to obtain reliable tumor growth measurements prior to the mice becoming moribund. When tumors became palpable, mice were randomized to treatment group by random number generation. Administration of SLC-0111(S) and vehicle (V) was performed as described above. Treatment with antibodies to PD-1 (P), CTLA-4 (C), and isotype (I) controls was performed on days 1, 3, 5, 7, 9 and days 1 and 6 for 4-1BB/CD137 (BB) (1 mg/kg; Supplementary Table S1). Tumor growth was tracked until permitted size limits (1,200 mm³) were reached ahead of animals becoming moribund.

To assess the growth of CAIX-depleted cell lines, cell lines were inoculated subcutaneously on the back of female C57Bl/6J mice (5 × 10⁵ B16F10) or in the left fourth mammary fat pad of female NOD/SCID (BC Cancer Research Centre in house breeding of Jackson Laboratory mice) mice (1 × 10⁶ 4T1), and tumors were allowed to grow until the endpoints were reached (mentioned above for subcutaneous B16F10 on the back and 500 mm³ for orthtopic 4T1 in the mammary fat pad). Tumor growth was measured as mentioned above.

For flow cytometry experiments profiling immune populations, B16F10 tumor-bearing mice were treated with immune-checkpoint or isotype control antibodies on days 1, 3, 5, and 7, or SLC-0111 or vehicle daily and tumors were harvested 24 hours following the final dose of immune-checkpoint antibody. This included a reduction of antibody treatments to four in an effort to maximize recovery of tumor material from treated groups. Spleens were removed from mice and mashed through a 0.45-μm-mesh filter with the plunger of a 3-mL syringe. Tumors were removed and minced with a razor blade prior to tissue disaggregation using the mouse tumor dissociation kit, according to the manufacturer's instructions (Miltenyi Biotec, #130-096-730). Peripheral blood was collected by cardiac bleed into a potassium EDTA-coated tube. Following disaggregation, all collected tissues were subjected to erythrocyte lysis with ammonium chloride buffer (Stem Cell Technologies, #07800). Samples were stained in HBSS containing 2% FBS for 30 minutes at 4°C.

For detection of transcription factors and intracellular cytokines, cells were fixed with FoxP3/transcription factor staining buffer (Thermo Fisher; #00-5523-00). Viability was tracked using the LIVE/DEAD fixable yellow viability stain (Thermo Fisher; #L34959). Antibodies and dilutions utilized are listed in Supplementary Table S1. Gating strategy used for all main cell subsets can be found in Supplementary Fig. S1. All samples were analyzed using the BD LSRFortessa (BD Biosciences), and data were analyzed using FlowJo v10 (FlowJo LLC).

B16F10 cells were seeded in 6-well dishes (BD Falcon) at 5,000/cm² and incubated at 21% O₂ overnight in D10 media. The following morning, media were replaced with fresh D10 growth medium or medium containing 100 μmol/L SLC-0111. Cells were then incubated in 21% or 1% O₂ for 72 hours. Following the 72-hour incubation, media were collected, and pH measured immediately with an Accumet pH meter (Fisher Scientific) and Accumet pH electrode (Fisher Scientific; #13-620-299A) as previously described (23).

B16F10 cells were seeded into 96-well plates (BD Falcon; 5,000 cells/well) in D10 media. The following day, fresh D10 media were added and the cells were cultured in 1% O₂ for 72 hours in the presence or absence of 100 μmol/L SLC-0111. Intracellular pH (pHi) measurements were carried out using the Fluorometric Intracellular pH Assay Kit (Sigma; cat no: MAK150) according to the manufacturer's instructions. Briefly, the growth media were removed and cells were loaded with 50 μL of BCFL-AM dye loading solution for 30 minutes at 1% O₂. After loading, the cells were treated with 100 μmol/L SLC-0111 and incubated for additional 15 minutes. Ratiometric measurements were carried out using a SpectraMax i3x microplate reader (Molecular Devices). A nigericin-based calibration (Sigma; N7143) standard curve was prepared as previously described (29) with buffers of pH ranging from pH 5.5–8.0 in 0.5 pH unit increments. A sigmoidal 4PL nonlinear regression model was used to fit the calibration curve and interpolate the experimental pHi values using GraphPad Prism 7.

Splenocytes (5 × 10⁵) from naïve C57Bl/6J mice were obtained as described above, and T cells were stimulated with plate bound anti-CD3/anti-CD28 (5 μg/mL and 1 μg/mL, respectively; Supplementary Table S1) for 48 hours and then expanded for an additional 48 hours with fresh medium provided every 24 hours. B16F10 cells were cultured for 72 hours in 21% or 1% O₂ at 5,000/cm². Cells were then harvested and seeded onto 96-well plates at the same density overnight in D10 medium containing the nuclear tracking dye, Nuclight Rapid Red (1:2000; Essen Biosciences, #4717), in 21% or 1% O₂. The following morning, T cells and indicated treatments were added in culture medium containing the membrane permeability dye, Sytox green (250 nmol/L; Thermo Fisher, #S7020), to track dead cells. Cells were treated with SLC-0111 (100 μmol/L) dissolved in culture medium. Cultures were incubated in 21% or 1% O₂ and imaged longitudinally (four images per well per time point) using the IncuCyte live-cell imaging system (Essen Bioscience). Cytotoxicity indices were calculated by measuring the number of dead cancer cells as a percentage of total cancer cells (Sytox green/mm²/Nuclight Rapid Red/mm²).

Cell lines were seeded at 5 × 10³ cells/cm² and incubated in 21% or 1% O₂ for 72 hours. Following this incubation period, cells were lysed in RIPA buffer in 21% or 1% O₂ and quantified by BCA assay. Lysate (30 μg) was separated on a 4% to 12% Bis-Tris gel, transferred to PVDF, and incubated overnight with antibody to CAIX and CAXII (Supplementary Table S1). Blots were developed using the Super Signal West femto maximum sensitivity ECL reagent (Thermo Fisher Scientific; #34096).

Two hours prior to tumor excision, mice were injected intraperitoneally (i.p.) with a saline solution containing bromodeoxyuridine (BrdUrd; 1,500 mg/kg; Sigma; B5002) and pimonidazole (60 mg/kg; Hypoxyprobe; HP2-100Kit). Seven minutes prior to tumor excision, mice were injected intravenously (i.v.) with DiOC₇ as previously described (23). Formalin-fixed paraffin-embedded sections were deparaffinized in xylene, gradually rehydrated with incubations in ethanol baths containing decreasing ethanol concentrations, and incubated in PBS. For B16F10 tumor sections, melanin was bleached by 40-minute incubation in 65°C in 10% H₂O₂ (30). Antigen retrieval was performed by incubation in 0.01 mol/L citrate, pH 4.0, by microwaving on high for 10 minutes. Tissue sections were incubated with primary antibody overnight (GLUT1 1 μg/mL; CAIX 2 μg/mL, MCT4—2 μg/mL; pimonidazole—1:1,500, CD3—1:100; Supplementary Table S1) and incubated with species-specific ImmPRESS HRP secondary Abs (MP-7405 and MP-7500, Vector Laboratories) for 30 minutes the following morning, according to the manufacturer's instructions. Staining for and detection of pimonidazole (hypoxia), BrdUrd (proliferation), CD31 (blood vessels), and DiOC₇ (perfusion) in the 4T1 model are previously described (23). Detection was carried out using the DAB Peroxidase (HRP) Substrate Kit (SK-4100; Vector Laboratories). Immunofluorescent staining of tumor tissue required 4 μg/mL CAIX and 300 μg/mL anti-pimonidazole (Hypoxyprobe) antibody concentrations. Tissues were treated with the TrueView autofluorescence quenching kit (SP-8400; Vector Laboratories) per the manufacturer's instructions following incubation with Hoechst 33342 (5 μg/mL). Histochemical analysis of lung and tumor sections was done by the Centre for Translational and Applied Genomics (CTAG) at the BC Cancer Centre.

Sample procurement and TMA construction are previously described (31). The TMA interrogated in this study now differs by 10 patients due to 6 being removed because of misdiagnosis and 4 additional patients being lost to follow-up. The TMA contains samples from the following melanoma subtypes: SS: superficial spreading, S: spindle-like, AL: acral lentiginous, N: nodular, DES: desmoplastic, U: unspecified. Note that all metastatic cases within the TMA had a primary diagnosis of unspecified, so this classification defines metastatic melanoma cases. Sections were stained for CAIX using the M75 monoclonal antibody from Bioscience on the Ventana system as previously described (ref. 23; Supplementary Table S1). Scoring of CAIX expression either 0 (no staining) or 1 (any staining) was done by a board-certified pathologist (D. Gao) and confirmed independently by S.C. Chafe.

Orthotopic breast tumors and plasma from mice treated with 25, 50, or 100 mg/kg SLC-0111 were recovered 24 hours following the final dose and flash frozen in liquid nitrogen and stored at −70°C until processing. Tumors were homogenized by combining two parts blank plasma with one part tumor using the T10 basic Ultra-Turrax tissue homogenizer (IKA Works Inc.). Twenty-five microliters of plasma or 50 μL of tumor homogenate were separated by reversed phase ultrahigh-performance liquid chromatography on a Kinetix C18 column (2.6 μm, 75 × 3 mm; Phenomenex) at a flow rate of 0.4 mL/min and analyzed by LC-MS/MS on an API 4000 LC-MS/MS system (Sciex). For mass spectrometric detection, electrospray ionization and multiple reaction monitoring in positive ionization mode was used. SLC-0111 concentrations were determined in each sample by back calculation from standard curves generated with SLC-0111 solutions of known concentration (25–30,000 ng/mL).

Bright field immunohistochemical images were captured on a Leica DM2500 microscope attached to a CCD camera. Images were processed using Photoshop CS5. Whole lung and tumor hematoxylin and eosin (H&E)–stained sections were scanned using the Panoramic Midi (3D Histec) using a 20× objective and images analyzed using Imagescope (3D Histec). IHC image quantitation of CAIX expression on whole tumor sections was done using thresholding in ImageJ. Fluorescent images were captured on an LSM Airyscan 800 Zeiss confocal microscope using a 63× oil immersion objective using Zen Blue software.

To assess the impact of CAIX inhibition on metabolic activity of melanoma cells, B16F10 cells (5,000/cm²) were treated with SLC-0111 and incubated in 1% or 21% O₂ for 72 hours, and glycolytic and respiratory function were measured using a glycolysis stress test and cell mito stress test assays on the Seahorse Extracellular Flux Analyzer (Agilent). Cells (5,000) were seeded per well in an XFe96 assay plate. Cells were exposed to 1 μmol/L oligomycin and 0.5 μmol/L FCCP (carbonyl cyanide-p-trifluoromethoxyphenylhydrazone) where indicated. Data were collected with the Wave software (Agilent), were normalized to cell number using CellTiter-Glo (Promega, #G7570), and plotted in GraphPad Prism v7.0. Rot/AA: cocktail of rotenone and antimycin A.

Raw count data from the TCGA project were retrieved using the TCGA biolinks package (v2.6.12) in Bioconductor. HTSeq counts generated from transcriptomes were downloaded from the GDC harmonized database for primary tumor samples. These data were combined with clinical data separately for each cancer type. Breast cancer data were generated for each molecular subtype based on PAM50 classification (32, 33).

Expression count values were normalized by the DESEQ2 variance stabilization method and adjusted to z-score for each gene (34). Z-scores were capped at 6 and −6 for increased and decreased expression, respectively. This arbitrary threshold was chosen to reduce the contribution of outliers during clustering. Hierarchical clustering was performed using Euclidean distance and “Ward.D2” clustering method in R. Heat maps were generated using the heat map function based on a gene set representing Th1-, cytotoxicity-, and HLA-related genes and tumor-infiltrating lymphocyte (TIL) abundance at the tumor site (9, 35). Correlation analysis assessing the relationship of expression between two genes was generated using Spearman correlation.

To show clinical correlation between these clusters and the clinical outcome, patient clinical data were downloaded and plotted for each patient across each hierarchical cluster. Analysis was performed after dividing the patients into two or three trees. Survival analysis was performed using “TCGAanalyze_survival” R function, and survival curves were plotted for each group of patients.

The transcriptome fastq files for the TCGA provisional data sets for breast invasive carcinoma (TCGA-BRCA), skin cutaneous melanoma (TCGA-SKCM), pancreatic adenocarcinoma (TCGA-PAAD), lung adenocarcinoma (TCGA-LUAD), lung squamous cell carcinoma (TCGA-LUSC), bladder urothelial carcinoma (TCGA-BLAD), and sarcoma (TCGA-SARC) were aligned to hg18 using STAR aligner and bam files were generated with recommended settings from the STAR-Fusion user guide (https://github.com/STAR-Fusion/STAR-Fusion/wiki). Cufflink was used to call the FPKM values and expression counts. The matrix of count values was sent to DESEQ2 for differential analysis. This matrix plus a conditions file annotating the condition and treatment of each sample of interest was then input into the Bioconductor DESEQ2 package (v1.10.1) in R, and differential analysis was conducted as instructed by the user guide (https://bioconductor.org/packages/devel/bioc/html/DESeq2.html). After filtering for nonempty samples and keeping only annotated samples of interest, the three key DESeq2 commands were executed as suggested in the user guide. A table of significantly differentially expressed genes between the two conditions of interest was generated, which specified both fold change and significance value for each gene. The results were written to a text output for the subsequent gene set enrichment and pathway analysis with the Bioconductor ReactomePA (v1.14.4) package in R.

Statistical analyses were performed using GraphPad Prism v7.0. Tests for normality were calculated using the D'Agostino and Pearson test. Nonparametric comparisons for two groups were calculated by Mann–Whitney U test, and analyses were two sided. Comparisons for more than two groups were calculated by ANOVA followed by Tukey multiple comparisons test for data with normal distribution or the Kruskal–Wallis test and Dunn multiple comparison test for data with nonnormal distribution. Survival analyses by the Kaplan–Meier method were compared by log-rank test. Clinicopathologic associations were compared using the Fisher exact test. Multivariate survival analysis by Cox regression was performed using R version 3.5.2 with statistical packages “survival” and “survminer.” Comparisons of patient gene-expression profiles were calculated using Student t test with Bonferroni correction. P < 0.05 was deemed the threshold for significance in all statistical tests performed.

To determine whether CAIX expression at the protein level played a role in malignant melanoma, we stained a tissue microarray (TMA) comprised of 449 patient tumor samples for CAIX expression (31). Membranous staining for CAIX was identified in 9% of cases interrogated (Fig. 1A). The clinicopathologic and follow-up data linked to the TMA identified that CAIX expression was associated with increased grade (P = 0.0003) and risk of metastasis (P = 0.0003; Supplementary Table S2–S3). CAIX expression was detected across all melanoma subtypes contained within the TMA, although expression was predominantly associated with cases that were metastatic and did not contain information on subtype classification of the primary lesion (Fig. 1B). Multivariate analysis, including known prognostic factors and those associated with CAIX expression, categorized CAIX as an independent marker of worse patient survival with a hazard ratio of 2.03 (Fig. 1C; Supplementary Table S4). These data suggest that CAIX may be an important therapeutic target for patients with metastatic melanoma.

To understand the role that CAIX plays in melanoma tumor growth, we interrogated CAIX expression in B16F10 melanoma tumors by IHC (Fig. 1D). B16F10 tumors contained regions of hypoxia-induced CAIX expression, as seen by colocalization with the glucose transporter GLUT1 (Fig. 1D, middle) and upregulation in pimonidazole-positive niches (Supplementary Fig. S2A). Upregulation of both GLUT1 and CAIX was also accompanied by upregulation of the lactate transporter MCT4 (Fig. 1D, right), suggesting that these tumors are glycolytic and contain an acidic extracellular milieu within the hypoxic TME.

We stably depleted CAIX expression in B16F10 cells with two independent short hairpin RNAs (Supplementary Fig. S2B) and assessed their ability to form tumors (Fig. 1E). Depletion of CAIX with shCar9_2 resulted in a 90% reduction in tumor growth relative to control shRNA (shNS) and a less effective shRNA (shCar9_1; Fig. 1E). Analysis of the tumors revealed significant CAIX expression in the tumors of shNS and shCar9_1 mice and the effective reduction of CAIX expression for the shCar9_2 tumors (Supplementary Fig. S2C), suggesting that CAIX expression in the tumor was required for growth. We also demonstrated that the closely related carbonic anhydrase isoform CAXII was not compensating for loss of CAIX expression in this model (Supplementary Fig. S2B–S2C). Because shCar9_2 tumors were regressing at the point when control tumors reached endpoint, we assessed whether this coincided with increased T-cell infiltration and identified increased CD3⁺ infiltration upon CAIX depletion (Fig. 1F–G).

Because acidic pH and lactate accumulation has been shown to reduce immune activity (15, 18, 19, 36, 37) and CAIX contributes to acidification of the TME (26, 28), we determined whether CAIX inhibition with the small-molecule inhibitor SLC-0111 (22–24) impacted the ability of B16F10 cells to acidify the extracellular milieu (Fig. 2A). Growth of B16F10 cells in hypoxic conditions resulted in a substantive increase in glycolytic flux, leading to a significant reduction in extracellular pH (pHe) relative to normoxic conditions. Treatment with SLC-0111 reduced the extent to which the pHe was acidified in hypoxic conditions but was ineffective when CAIX was absent in normoxic conditions (Fig. 2A). These observations are consistent with our previous findings in murine basal-like breast cancer cells (23). To determine whether CAIX inhibition altered the metabolic activity of melanoma cells grown in hypoxic conditions, we measured the extracellular flux of protons and oxygen. SLC-0111 treatment reduced the ability of melanoma cells to rely on glycolysis for energy production, known as their glycolytic reserve, when energy production by oxidative phosphorylation was inhibited (Fig. 2B and C). When cellular respiration was measured, SLC-0111–treated cells had a reduced basal respiration rate compared with controls (Fig. 2D and E). The decreased metabolic activity of CAIX-inhibited melanoma cells was exacerbated when the cells respired at maximal rates, known as their spare respiratory capacity, revealing a metabolic defect (Fig. 2F).

To explain the impact of CAIX inhibition on metabolic activity, we assessed whether CAIX inhibition altered intracellular pH (pHi; Fig. 2G). Enzymatic function is dependent on narrow pH ranges for optimal activity, and glycolytic enzymes are impacted by changes in intracellular pH (38). Melanoma cells treated with SLC-0111 had a more acidic pHi than controls (Fig. 2G). To determine whether these conditions would lead to enhanced T-cell activity, we cocultured activated T cells at multiple ratios with B16F10 cells (Fig. 2H and I) in the presence and absence of SLC-0111. In the presence of SLC-0111, a dose-dependent increase in T cell–mediated killing of the cancer cells was seen (Fig. 2H and I).

To assess whether inhibiting CAIX activity in the TME would enhance the efficacy of ICB, we treated B16F10 tumors with a combination consisting of SLC-0111 and antibodies to PD-1 and CTLA-4 (Fig. 3A). In agreement with previous data, treatment with single-agent anti–PD-1 or anti–CTLA-4 failed to provide significant benefit (Supplementary Fig. S3A; refs. 39, 40). Single-agent treatment with SLC-0111 was as effective as both antibodies in delaying growth relative to controls (Fig. 3B; Supplementary Fig. S3A). Combining SLC-0111 with either immune-checkpoint antibody offered no additional benefit over either single agent (Supplementary Fig. S3A).

As also demonstrated previously by others, combining anti–PD-1 and anti–CTLA-4 delayed tumor growth and progression compared with controls and either antibody alone (Fig. 3B and C; Supplementary Fig. S3A; ref. 41). Because ICB has been shown to result in vascular normalization in certain preclinical models (42), we assessed CAIX expression after treatment and found no change from control across the indicated treatment groups (Fig. 3D; Supplementary Fig. S3B). The addition of SLC-0111 to the combination of anti–PD-1 and anti–CTLA-4 further delayed tumor growth and progression (Fig. 3B and C; Supplementary Fig. S3A and S3C), and this triple combination increased the number of complete responders to 30% (3/10), resulting in the extension of survival of these mice (Fig. 3C, E, and F).

To determine whether CAIX inhibition could be combined with additional interventions, we assessed combination treatment of melanoma tumors with an antibody to CD137/4-1BB, which stimulates proliferation and survival of T cells (Supplementary Fig. S3D–S3G; refs. 43, 44). Anti–4-1BB treatment alone resulted in a complete response rate of 10% and combination with anti–PD-1 failed to result in any complete responses (Supplementary Fig. S3F). The addition of SLC-0111 to anti–PD-1 and anti–4-1BB increased the complete response rate to 20% (Supplementary Fig. S3F).

To determine whether infiltrating T cells into the hypoxic TME expressed CAIX, we stained tumor sections for CAIX and CD3 simultaneously (Supplementary Fig. S3H). CAIX was upregulated on the tumor cells in the hypoxic TME, but CD3⁺ T cells localized to these areas did not produce any detectable CAIX, suggesting that targeting CAIX with SLC-0111 would not adversely impact the responding immune cells when treated in combination with ICB.

To extend our melanoma observations to other hypoxic solid tumors, we assessed CAIX inhibition in combination with ICB in the 4T1 model of basal-like breast cancer. The 4T1 model is a glycolytic and hypoxic tumor model that expresses a substantive amount of CAIX in areas that are both positive and negative for the hypoxia marker pimonidazole (Fig. 4A). To this end, we previously demonstrated that shRNA-mediated depletion of CAIX in this model results in significant inhibition of tumor growth (23). We demonstrated here that this was dependent upon the presence of a functional immune system, as growth of these cell lines in immunocompromised mice results in only a minor growth delay of the tumors formed from CAIX-depleted cell lines relative to the nonsilenced control tumors (Supplementary Fig. S4A). These data provide additional evidence that CAIX expression in the TME negatively impacted immune function.

We treated mice bearing orthotopic 4T1 tumors with SLC-0111, anti–PD-1, and anti–CTLA-4 (Fig. 4B). We first demonstrated that SLC-0111 preferentially accumulated in hypoxic breast tumors over the plasma in a dose-dependent manner (Supplementary Fig. S4B), and this correlated directly with decreases in tumor volume (Supplementary Fig. S4C). In agreement with previous findings, single-agent anti–PD-1 and anti–CTLA-4 treatment had very little impact on tumor growth (Supplementary Fig. S4D–S4E; ref. 45). Combining CAIX inhibition with anti–PD-1, anti–CTLA-4, or both inhibitors simultaneously only marginally provided additional therapeutic capacity in controlling tumor growth (Supplementary Fig. S4D–S4E), despite equivalent CAIX expression present across all treatment cohorts (Fig. 4C; Supplementary Fig. S4F). Because we failed to achieve tumor growth delays in mice with these treatments, we assessed the therapeutic efficacy histologically and evaluated necrosis (Fig. 4D; Supplementary Fig. S4G). The combination of anti–PD-1 and anti–CTLA-4 increased central necrosis within the tumors relative to controls (Fig. 4D; Supplementary Fig. S4G), and the addition of SLC-0111 to this combination increased intratumoral necrosis even further, suggesting that this combination was effectively activating the immune system to target the tumor (Fig. 4D). Because the 4T1 model represents a basal-type/triple-negative tumor with a high metastatic propensity, we examined metastatic burden in the mice (Fig. 4E). As shown in Fig. 4F and Table 1, we observed up to a 90% reduction in gross lung metastatic burden across the treatment cohort relative to controls. CAIX inhibition in combination with anti–PD-1 or anti–CTLA-4 was more effective in reducing lung metastasis than either single-agent treatment (Table 1). Triple combination treatment had the greatest effect in reducing the lung metastatic burden (Fig. 4F; Table 1) and resulted in extension of median survival over control mice (Fig. 4G). Histologic examination of lung tissues from mice across the treatment cohort confirmed the findings of the macroscopic evaluation of the lungs (Supplementary Fig. S4H).

To evaluate changes in immune cell composition in mice treated with SLC-0111, anti–PD-1, and anti–CTLA-4, we collected tumors, blood, and spleens (Fig. 5A; Supplementary Fig. S4A). Analysis of the leukocyte composition within the tumors revealed no change in the frequency of T cells and B cells (Fig. 5B). Because overall T-cell frequency did not change with the combination of SLC-0111 with anti–PD-1 and anti–CTLA-4, we assessed the phenotype of the CD4⁺ T-helper cells present within the TME (Fig. 5C). The CD4 profile of control-treated tumors was split between T-bet⁺ (Th1) and RORγt⁺ (Th17) cells, whereas the presence of regulatory T cells (Treg) and GATA-3⁺ Th2 cells was also detected. Treatment with SLC-0111 reduced the presence of Tregs and Th17 cells and increased the frequency of Th1 cells. Combining anti–PD-1 with anti–CTLA-4 further reduced the presence of Tregs within these tumors, but had a similar Th1 frequency to SLC-0111 treatment alone. Treatment with all three inhibitors maintained the SLC-0111-induced Th1 cells, while maintaining low frequencies of Tregs and eliminating Th2 cells, suggesting that the enhanced therapeutic efficacy achieved in the combination of all three inhibitors may be due to the removal of immune-suppressive cell populations within the TME.

To assess whether the increased presence of T-bet⁺ Th1 cells led to an increase in CD8⁺ TILs skewed in that direction, we evaluated T-bet expression on the CD8⁺ TILs (Fig. 5D and E). Similar to the observations in the CD4⁺ TILs, T-bet expression was increased in SLC-0111-treated tumors relative to control treatments. Anti–PD-1 and anti–CTLA-4 treatment resulted in a similar increase in the presence of T-bet–expressing CD8⁺ TILs, whereas treatment with all three led to an enhanced frequency of T-bet⁺ CD8⁺ TILs within the TME. We also assessed whether this was accompanied by an increase in EOMES expression on the CD8⁺ TILs (Fig. 5F and G). Similar to the T-bet observations, SLC-0111 treatment alone or in combination with anti–PD-1 and anti–CTLA-4 increased the frequency of EOMES-expressing CD8⁺ TILs to a similar degree relative to control treatments, suggesting the skewing of T cells toward a cytotoxic phenotype.

We next assessed the expression of PD-1 (Fig. 5H–J), LAG-3 (Supplementary Fig. S1B), and TIM-3 (Supplementary Fig. S1C) on antigen-experienced CD8⁺ TILs. Treatment with controls or SLC-0111 was ineffective at reducing PD-1, LAG-3, or TIM-3 expression. PD-1 was decreased with triple treatment, but this was not seen for LAG-3 or TIM-3 (Fig. 5H–J; Supplementary Fig. S1B–S1C). To assess whether this reduction impacted the release of cytolytic molecules from the CD8⁺ TILs, we assessed the TIL production of granzyme B in these tumors (Fig. 5K and L). The heterogeneity in the production of granzyme B in the TME masked any changes between groups, so we assessed granzyme B–producing T cells in the blood and identified that the addition of SLC-0111 increased the frequency of granzyme B–producing T cells in circulation (Fig. 5M). Together, these data suggest that CAIX inhibition in the TME enhanced antitumor Th1 responses.

Profiling the CD4 abundance within the TME showed a reduction in the frequency of Tregs in all treatments relative to the controls (Supplementary Fig. S1D–S1E). We further investigated the relationship of this reduction with respect to the global ratio of CD8:CD4 T cells, and this was unchanged (Supplementary Fig. S1F). We next assessed whether the CD8:Treg ratio changed (Fig. 5N). The CD8:Treg ratio was greatest in the triple combination but was not significant. We also assessed the frequency of an effector CD4⁺ population associated with response to anti–CTLA-4 therapy (46). The triple combination resulted in increased circulating ICOS⁺CD4⁺ T cells (Fig. 5O). This suggests that the enhanced efficacy achieved by the triple combination was due to an increased Th1 response in a less suppressive TME.

We next assessed CA9 expression in patient tumors available in TCGA data sets for association with a T cell–inflamed gene signature consisting of immune activity markers (Fig. 6). Using the skin cutaneous melanoma (TCGA-SKCM) data set (n = 479), we performed hierarchical clustering according to high or low intartumoral expression of CD3E, CD8A, and CD4 (Fig. 6A). Higher CA9 expression was significantly associated with lower expression of CD3E, CD8A, and CD4 (Fig. 6B). To evaluate whether this was attributable to the metabolic phenotype of the tumors, we assessed whether this was also true for expression of lactate dehydrogenase (LDHA), the monocarboxylate transporter MCT-1 (SLC16A1), and glucose transporter GLUT-4 (SLC2A4), gene products involved in glucose uptake, utilization, and lactate extrusion (Supplementary Fig. S5A). We identified that the reduction in expression of CD3E, CD8A, and CD4 in the TME was associated with worse outcome (Fig. 6C).

We then used an immune gene set previously demonstrated to predict colorectal cancer metastasis and to identify T-cell–inflamed TMEs to address the immune activity of T cells within tumors and their association with CA9 expression (Fig. 6D; refs. 9, 34). Higher CA9 expression was associated with decreased expression of genes associated with an effective immune response (Fig. 6E). This observation was also true for expression of LDHA and SLC2A4 (Supplementary Fig. S5B). Reduced expression of genes associated with an effective antitumor immune response was predictive of worse survival in these patients (Fig. 6F). Because CAIX is associated with the hypoxic TME and the glycolytic phenotypes associated within these regions of solid tumors, we assessed how the expression of CA9, SLC2A1, SLC2A4, LDHA, SLC16A1, SLC16A3, and PDK1 correlated with the expression of genes in the immune gene set (Fig. 6G). Our analysis identified that increased expression of CA9, SLC2A1, SLC2A4, LDHA, and SLC16A1 were the most significantly associated with decreased expression of the gene set. We next performed hierarchical clustering of patients according to the expression of CA9, SLC2A1, SLC2A4, LDHA, SLC16A1, SLC16A3, and PDK1 and assessed survival (Supplementary Fig. S5C–S5D). Although the expression of these genes clustered together, a glycolytic phenotype of the tumors alone did not offer further predictive value in determining patient survival in our analyses (Supplementary Fig. S5C–S5D).

We next assessed whether these findings were restricted to melanoma, and assessed additional TCGA solid tumor data sets, in particular, basal-like breast cancer where we have previously demonstrated CAIX to be an independent biomarker for worse overall survival (23). Stratifying the TCGA-BRCA data set according to the PAM50 classifier, we confirmed that CA9 expression had the highest expression in the basal-like subtype, followed by the HER2, Luminal A, and Luminal B subtypes (Supplementary Fig. S5E) and these findings agree with our previous observations (23). We then evaluated the basal-like subset identically to our melanoma analyses (Fig. 6H–N). Increased CA9 expression was associated with reduced expression of CD3E, CD8A, and CD4 in the tumors of these patients (Fig. 6H and I). This was also true for SLC2A1, suggesting that glucose utilization in the basal-like subtype may negatively affect immune infiltration (Supplementary Fig. S5F). Reduced expression of these genes within the TME was associated with worse overall survival (Fig. 6J).

We next evaluated the expression of the immune gene set against the basal-like stratified patients (Fig. 6K) and identified that higher CA9 expression was associated with decreased expression of genes associated with immune activation (Fig. 6L). Decreased immune activity was associated with worse overall survival in this cohort (Fig. 6M). We explored whether the reduction in the immune gene set was associated with a glycolytic/acidic gene-expression signature similar to melanoma, we identified that again the expression of SLC2A1, LDHA, and CA9 was negatively associated with markers of immune activation (Fig. 6N), suggesting that the acidic TME was associated with decreased immune activity in breast cancer. The expression of SLC2A4 and SLC16A1 was not predictive in this analysis, suggesting cancer-specific metabolic alterations need to be considered (Supplementary Fig. S5G). However, together, the data demonstrated that increased glucose utilization and conversion to lactate by the tumor and the concomitant acidification of the TME associates with immune dysfunction in metastatic melanoma and basal-like breast cancer patients, and further demonstrates that CA9 expression could be a promising target to improve antitumor immune responses in cancers with similar characteristics (Supplementary Fig. S6A–S6J).

The suppressive TME is recognized as a critical hurdle to antitumor immunity (12, 47). Aberrant vasculature and accompanying deficiencies in perfusion and oxygen diffusion create a metabolically challenging environment where nutrients are scarce and pH is critically low. Interventions to restore tissue oxygenation and perfusion have been shown to be beneficial to immune function (42, 48–50). Here, we demonstrated that therapeutically targeting CAIX altered metabolism of the tumor cells, decreasing glycolytic output and respiratory capacity. We demonstrated that CAIX inhibition reduced the capacity of melanoma and breast cancer cells to acidify the extracellular environment, leading to enhanced immune activity. CAIX inhibition in models of melanoma and metastatic breast cancer improved the efficacy of anti–PD-1 and anti–CTLA-4. We have also demonstrated that although CAIX is negatively associated with expression of Th1, cytotoxic, and HLA-related genes in patients with metastatic melanoma and basal-like breast cancer, its inhibition increased the Th1 response in the preclinical models evaluated.

CAIX is associated with poor prognosis in many solid tumor types, including breast cancer (51). We have previously shown that CAIX is an independent marker of worse overall survival and is associated with the occurrence of distant metastasis in TNBC (23). We established that CAIX was an independent marker of worse overall survival in melanoma and was associated with increased metastasis. In support of these findings, increased CAIX expression has been detected previously in melanoma patients with distant metastasis compared with those with lymph node metastasis (52). Thus, CAIX is a critical therapeutic target in both TNBC and metastatic melanoma. The identification of biomarkers or signatures of response/resistance to ICB is critical to stratifying patients into those most likely to derive maximal benefit from therapy. We identified that CA9 expression was associated with decreased expression of CD3E, CD8A, and CD4 genes and 31 genes of a T cell–inflamed signature in tumors from patients with metastatic melanoma, basal-like breast cancer, pancreatic ductal adenocarcinoma, bladder cancer, lung adenocarcinoma, lung squamous cell carcinoma, and sarcoma. The broader implications that CA9 expression is negatively correlated with the expression of T-cell activation markers across a range of solid tumor types, many of which are currently the subject of interventions with ICB, strengthen our rationale for combining CAIX inhibition with ICB to improve immune activity clinically. We demonstrated that SLC-0111 treatment in combination with anti–PD-1 and anti–CTLA-4 leads to an increased Th1 response, in part, by reducing PD-1 expression on antigen-experienced T cells within the TME. Chronic exposure to antigens in models of chronic viral infection or cancer leads to epigenetic changes in T cells that ultimately lead to an unrecoverable state of exhaustion (53, 54). However, a window of opportunity exists whereby T cells can be reinvigorated by therapeutic intervention with anti–PD-1 (55, 56).

Tumor hypoxia is an appreciated impediment to immune function (21). The hypoxic niche contains numerous nodes of immune suppression, including enhanced expression of PD-L1 on tumor cells (21, 57). A study investigating markers of resistance to anti–PD-1 therapy identified hypoxic gene signatures among those expressed in tumors of patients who failed to respond to therapy (58). Interestingly, mining their data set revealed increased CA9 expression in those patients who failed to respond. These findings were further supported in a study that also identified increased CAIX expression, at both the RNA and protein levels, in patients failing to respond to anti–PD-1 therapy (59). Combined, these studies provide evidence that a clinical cohort stands to potentially benefit from CAIX inhibition in combination with PD-1 blockade. CAIX is a well-known downstream effector of the hypoxia-inducible factor (HIF)-1, and its expression in relationship to exogenous markers of hypoxia, as shown here, is well documented (23). HIF1, although known for its stabilization under conditions of low oxygen, is also induced by other means, e.g., oncogenic activation (60), which could lead to a hypoxia-independent activation of the HIF program and subsequent expression of downstream HIF effectors. Thus, there are circumstances where CAIX expression may be induced in a hypoxia-independent manner (51). In support of this, there is evidence available that CAIX expression is induced in acidic pH environments independent of hypoxia, which suggests that CAIX expression and exogenous markers of hypoxia may not always directly correlate (61). Nevertheless, elimination of the CAIX-positive fraction of the tumor between the perfused vessels and the pimonidazole area through CAIX inhibition, in combination with a reinvigorated immune response, is an effective approach for restoring immune function in the TME. Our findings further support the importance of neutralizing the acidic pH within the TME by exploiting the tumor-specific expression of CAIX, while offering a clinically viable therapeutic option using the small-molecule SLC-0111 (25).

P.C. McDonald has ownership interest (including stock, patents, etc.) in SignalChem Lifesciences Corporation. S.P. Shah has ownership interest (including stock, patents, etc.) in Contextual Genomics Inc. and is a consultant/advisory board member for the same. S. Dedhar has ownership interest (including stock, patents, etc.) in SignalChem Lifesciences Corp and is a consultant/advisory board member for Welichem Biotech Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S.C. Chafe, P.C. McDonald, S. Dedhar

Development of methodology: S.C. Chafe, P.C. McDonald, S. Saberi, O. Nemirovsky, G. Venkateswaran, A.H. Kyle, Y. Zhou, S.P. Shah, S. Dedhar

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.C. Chafe, P.C. McDonald, S. Saberi, O. Nemirovsky, G. Venkateswaran, S. Burugu, A. Delaidelli, A.H. Kyle, J.H.E. Baker, J.A. Gillespie, Y. Zhou

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.C. Chafe, P.C. McDonald, S. Saberi, O. Nemirovsky, G. Venkateswaran, D. Gao, A. Delaidelli, J.A. Gillespie, A.I. Minchinton, S.P. Shah, S. Dedhar

Writing, review, and/or revision of the manuscript: S.C. Chafe, S. Saberi, S. Dedhar

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Saberi, Y. Zhou, S. Dedhar

Study supervision: S.C. Chafe, A. Bashashati, S. Dedhar

This work was supported by research grants from the Canadian Institutes of Health Research (FDN-143318) and the Canadian Cancer Society Research Institute (CCSRI #703191) to S. Dedhar.

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.