Improving combination therapies: targeting A_2B-adenosine receptor to modulate metabolic tumor microenvironment and immunosuppression

Abstract

Background

We investigated the role of A_2B-adenosine receptor in regulating immunosuppressive metabolic stress in the tumor microenvironment. Novel A_2B-adenosine receptor antagonist PBF-1129 was tested for antitumor activity in mice and evaluated for safety and immunologic efficacy in a phase I clinical trial of patients with non-small cell lung cancer.

Methods

The antitumor efficacy of A_2B-adenosine receptor antagonists and their impact on the metabolic and immune tumor microenvironment were evaluated in lung, melanoma, colon, breast, and epidermal growth factor receptor–inducible transgenic cancer models. Employing electron paramagnetic resonance, we assessed changes in tumor microenvironment metabolic parameters, including pO₂, pH, and inorganic phosphate, during tumor growth and evaluated the immunologic effects of PBF-1129, including its pharmacokinetics, safety, and toxicity, in patients with non-small cell lung cancer.

Results

Levels of metabolic stress correlated with tumor growth, metastasis, and immunosuppression. Tumor interstitial inorganic phosphate emerged as a correlative and cumulative measure of tumor microenvironment stress and immunosuppression. A_2B-adenosine receptor inhibition alleviated metabolic stress, downregulated expression of adenosine-generating ectonucleotidases, increased expression of adenosine deaminase, decreased tumor growth and metastasis, increased interferon γ production, and enhanced the efficacy of antitumor therapies following combination regimens in animal models (anti–programmed cell death 1 protein vs anti–programmed cell death 1 protein plus PBF-1129 treatment hazard ratio = 11.74 [95% confidence interval = 3.35 to 41.13], n = 10, P < .001, 2-sided F test). In patients with non-small cell lung cancer, PBF-1129 was well tolerated, with no dose-limiting toxicities; demonstrated pharmacologic efficacy; modulated the adenosine generation system; and improved antitumor immunity.

Conclusions

Data identify A_2B-adenosine receptor as a valuable therapeutic target to modify metabolic and immune tumor microenvironment to reduce immunosuppression, enhance the efficacy of immunotherapies, and support clinical application of PBF-1129 in combination therapies.

The solid tumor microenvironment comprises a complex milieu of hypoxia, acidic extracellular pH, altered redox potential, low nutrient levels, increased interstitial fluid pressure, and elevated adenosine triphosphate levels. Cancer cell demise is accompanied by the release of adenosine triphosphate (1), metabolized to adenosine by adenosine triphosphate–hydrolyzing ectoenzymes, including CD39, CD73, and CD203, expressed on tumor and immune cells (2-7).

Adenosinergic signaling is a powerful immuno-metabolic checkpoint in solid tumors (8-12). Adenosine-activated pathways promote tissue adaptation in hypoxic environments. High expression of adenosine-generating enzymes is associated with poor clinical outcomes in immunotherapy-resistant cancers (13-16). Adenosine binds G-protein–coupled adenosine receptors A₁ and A_2A, A_2B, and A₃ to modulate cellular and physiologic processes (4,15), including pleiotropic effects on the tumor microenvironment–modulating activity of T cells, dendritic cells, natural killer cells, macrophages and neutrophils, and accumulation of myeloid-derived suppressor cells, type II macrophages, and regulatory T cells (Treg) within the tumor (17-29). Tumor microenvironment metabolic derangement alters their function and induces T-cell irresponsiveness (30-32).

We developed a set of paramagnetic probes for noninvasive concurrent assessment of tumor microenvironment parameters in vivo by the L-band electron paramagnetic resonance spectroscopy (33,34). The triple functionality monophosphonated triarylmethyl (pTAM) probe enables multifunctional sensitivity and correlation of extracellular pH, pO₂ and inorganic phosphate (P_i) to characterize metabolic tumor microenvironment (35,36).

Preclinical studies provided the basis for clinically targeting adenosinergic signaling in cancer (13,15,18,37,38). Ongoing phase I/II clinical trials in solid malignancies evaluated the safety and efficacy of drugs targeting the adenosine system as a single treatment or in combination with various therapies, including 4 anti-CD73 antibodies, 1 anti-CD38 antibody, and 5 different A_2A antagonists (13,15,18,37). Adenosine receptors A₁ and A_2A are an especially attractive target because of their low affinity, which allows their activation only at pathologically high adenosine levels observed with tumor hypoxia and ischemia.

Here, we show that adenosine receptor A_2B signaling aggravates the metabolically stressful tumor microenvironment, whereas adenosine receptor A_2B inhibition moderated the tumor microenvironment metabolic stress in animal studies. PBF-1129, a novel oral adenosine receptor A_2B antagonist, demonstrated efficacy alone or in combination treatments in decreasing tumor growth and metastasis and reversing adverse immunologic effects. Reduced expression of CD39 and CD73 ectoenzymes in antigen-presenting cells following PBF-1129 treatment in patients with non-small cell lung cancer (NSCLC) in a phase I clinical trial (NCT03274479) suggested its pharmacologic efficacy.

Methods

Information about the cell lines and mouse models used for this study as well as detailed experimental procedures are provided in the Supplementary Materials (available online).

Adenosine receptor expression in human lung tumor, normal tissue, and cancer cells

For the analysis of adenosine receptor expression in human lung tumors and normal tissues, we used previously generated data from our shotgun proteomics profiling analysis of NSCLC samples (39). Tissue specimens included cancer tissue from patients with pathologic stage I lung cancer and no previous cancer history. Normal lung tissues were obtained from patients undergoing lung resection for suspicion of lung cancer but with no confirmed diagnosis of lung or other cancer. Demographic information, details on the sample preparation, methodology, and data analysis have been described earlier (39). Spectral counting (total number of spectra identified for a protein) was used as a quantitative metric to examine protein expression profiles for each tissue type.

Lung adenocarcinoma gene expression from The Cancer Genome Atlas from 517 tumors were downloaded from the genomic data commons using the cgdsr package in R (R Foundation for Statistical Computing, Vienna, Austria). A Spearman correlation was used to relate CD73 and A2B expression values. Kaplan-Meier curves for A2B and CD73 were created using the survival and survminer packages in R, with stratification of high and low expression at quantiles 0.55 and 0.45, respectively.

In human and mouse cancer cell lines, expression of adenosine receptor A_2B and adenosine system genes was evaluated by reverse transcriptase–polymerase chain reaction with sets of specific primers or by immunoblot with the specific antibodies (Supplementary Table 1 and 2, available online).

PBF-1129 treatment of human patients and analysis of peripheral blood mononuclear cells

NCT03274479 is an open-label, phase I dose escalation study of PBF-1129 in patients with metastatic NSCLC whose disease has progressed on standard-of-care therapies. A 3 + 3 dose escalation design was employed and 4 dose levels are being evaluated. Eligible patients must have metastatic NSCLC, have no curable options, and have experienced disease progression on standard-of-care immunotherapy and chemotherapy. Patients with select driver alterations are eligible, provided they have been treated with all standard targeted therapies. Exclusion criteria include active or symptomatic brain metastases, autoimmune disorders requiring treatment, and concurrent anticancer therapies. Patients receive oral PBF-1129 daily on 28-day cycles and continue treatment until disease progression, death, unacceptable toxicities, or withdrawal of consent. The primary endpoint is safety and tolerability and to determine recommended phase II dose. As part of protocol correlatives, blood draws for peripheral blood mononuclear cells are conducted at cycle 1 day 1 and cycle 2 day 1. Written informed consent from participating patients was obtained, and the study was approved by the institutional review board at The Ohio State University Comprehensive Cancer Center (Columbus, OH). Accrual to this study is ongoing. Pharmacokinetic studies have confirmed full adenosine receptor A_2B target engagement at the 80-mg dose (Supplementary Tables 3-6, available online). For 7 patients treated as part of this clinical trial, blood was drawn before and after the treatment with 40-, 80-, or 160-mg dose of PBF-1129. Peripheral blood mononuclear cells were isolated by gradient centrifugation using Ficoll-Paque media (Fisher Scientific, Waltham, MA), and cells were analyzed by flow cytometry for the expression of cell surface markers (1 patient received a 40-mg dose and did not have evaluable measures). The demographics for all 7 patients include 5 White men and 2 White women, with a median age of 60 years (range, 51-69 years); all patients had received standard-of-care immunotherapy and chemotherapy. For single ascending dose and multiple ascending doses, pharmacokinetic studies were conducted in Spain (CUNFI-1617, EudraCT2016-003135-38 and CUNFI-1709, EudraCT-2017-001901-34); White male healthy volunteers 30 years of age or younger (n = 56) were recruited. Written informed consent was obtained, and the studies were approved by the Central Unit for Clinical Trials at Clínica Universidad de Navarra (Navarra, Spain).

Statistical analysis

Data were analyzed using GraphPad Prism 9 (GraphPad Software Inc, San Diego, CA) and SAS (SAS Institute, Cary, NC) software and presented as mean (SD). Comparisons between treatment and control groups were performed using the Kruskal-Wallis test. Comparisons between 2 groups for animal studies were performed using the Mann-Whitney test. Tumor electron paramagnetic resonance treatment effects were estimated using a linear regression model to study the association of pH, pO2, or P_i with the treatment groups while controlling the tumor volume as a cofounder. Significant differences between the groups were assessed using an F test. Electron paramagnetic resonance spectroscopy data for lymph nodes were approximated by plateau followed by 1 phase decay model and compared by nonlinear least square fit. For tumor growth data, a linear mixed-effect model was used to assess the association of tumor volume with the treatment groups across all time points, where a subject-specific random intercept was used to account for the association among repeated measures from the same mouse. The fixed effects in these regression models would include the treatment group, assessment days, and interactions between treatment group and assessment days. The Bonferroni method was used to adjust for multiple comparisons. Details and complete results of the analysis of electron paramagnetic resonance spectroscopy and tumor growth data are presented in the Supplementary Statistical Analysis in the Supplementary Materials (available online). Survival curves were compared using the Mantel-Haenszel log-rank test. Comparisons between human pre– and post–PBF-1129 treatment groups were performed using a Wilcoxon matched-pairs signed rank test. All tests were 2-sided, and P < .05 was considered statistically significant.

Results

Impact of adenosine receptor A_2B signaling on metabolic tumor microenvironment parameters and function of tumor-infiltrating immune cells

Electron paramagnetic resonance spectroscopy with a pTAM paramagnetic probe enables access to tumor microenvironment parameters, including pH, pO₂, and P_i in the interstitial milieu, as well as monitoring of their alterations across a range of tumor sizes and time course of treatments (35,36,40,41). Following the approved study protocols (see Supplementary Methods, available online), multiple subcutaneous and orthotopic tumors exhibited the exacerbation of tumor acidosis and hypoxia as well as the buildup of inorganic phosphate during tumor growth (Figure 1, A and B; Supplementary Figure 1, available online).

Figure 1.

Effect of adenosine receptor A_2B signaling on metabolic tumor microenvironment parameters. A) Alterations of metabolic parameters during the time course of subcutaneous Lewis lung carcinoma tumor growth in mice. B) Treatment with adenosine receptor A_2B antagonist PSB-603 ameliorates metabolic stress in the tumor microenvironment during Lewis lung carcinoma tumor growth. Each dot in (A) and (B) represents 1 measurement for an individual mouse, with 5 mice per group for 3 to 4 L-band electron paramagnetic resonance spectroscopy measurements per mouse. The relationship between tumor electron paramagnetic resonance spectroscopy–based measurements and pH, pO₂, P_i, and tumor size were modeled using linear regression (formulas and Pearson correlation P values are shown), with an interaction between treatment and tumor volume assessed by F test (details are presented in the Supplementary Statistical Analysis, available online). C) Differences between the slopes of linear regressions approximating alterations of the tumor microenvironment’s metabolic parameters in untreated and PSB-603–treated mice. Y-axes show parameter change per 1 mm³ of tumor volume. Tumor-bearing mice were untreated or treated with adenosine receptor A_2B antagonist PBS-603 (10 mg/kg), and assessment of tumor microenvironment parameters was performed during tumor growth using the mono-phosphonated triarylmethyl electron paramagnetic resonance spectroscopy probe. The difference between the groups was assessed using an F test; pairwise comparisons, *P < .05. P_i = inorganic phosphate.

To assess the effect of adenosine receptor A_2B–specific antagonist PSB-603, a linear regression model was used to test the association of pH, pO₂, or P_i with the treatment groups. While controlling the tumor volume as a cofounder; significant statistical differences between the groups were assessed using an F test. For Lewis lung carcinoma tumors, blocking adenosine receptor A_2B signaling affected the relationship between tumor microenvironment parameters and tumor volume by significantly increasing acidosis (P = .04) and decreasing P_i buildup (P = .01) (Figure 1, A-C, Supplementary Statistical Analysis, available online).

In subcutaneous B16F10 melanoma and orthotopic 4T1 breast tumor models, adenosine receptor A_2B modulation profoundly affected tumor microenvironment metabolic parameters, confirming a link between the metabolic tumor microenvironment, tumor aggressiveness, and metastases. In the B16F10 model, however, PSB-603 decreased metabolic stress by ameliorating hypoxia and P_i accumulation and reversing acidosis, and the adenosine receptor A_2B agonist BAY60-6583 exacerbated the metabolic stress and increased tumor growth (Figure 2, A and B). Moderation of metabolic tumor microenvironment stress correlated with improved responses of tumor-infiltrating T cells (mean [SD] numbers of interferon [IFN]-γ spots/10⁶ cells of tumor cell suspension for control vs PBS-603–treated mice: 16.69 [1.21] vs 55.88 [20.13], respectively; n = 5, P = .02) (Figure 2, C).

Figure 2.

Correlation of metabolic tumor microenvironment with tumor characteristics and functionality of infiltrating immune cells. A-C) Tumor microenvironment parameter changes (A), tumor growth curves (B), and the number of IFN-γ–producing cells within tumor single-cell suspension (C) in a subcutaneous B16F10 melanoma tumor model. In (B), a linear mixed-effects model with repeated measures was used to assess the association of tumor volume with the treatment groups across all time points. The Bonferroni method was used to adjust for multiple comparisons (pairwise comparison P values; details are presented in the Supplementary Statistical Analysis, available online). D, E) The changes in tumor microenvironment parameters between post- and pretreatment measurements by L-band electron paramagnetic resonance spectroscopy (D) and the number of lung metastases (E) in an orthotopic 4T1 breast cancer model; tumor-bearing mice were treated for 15 days, but the difference in tumor size among the groups was statistically insignificant. Tumor-bearing mice were untreated or treated with adenosine receptor A_2B antagonist PBS-603 (10 mg/kg) or agonist BAY60-6583 (2 mg/kg); a monophosphonated triarylmethyl probe was used for the electron paramagnetic resonance spectroscopic analysis, with 8 to 10 mice per group for tumor growth or metastasis studies and 5 mice per group for 3 to 4 electron paramagnetic resonance spectroscopic measurements per mouse. (A, D) The difference between treatment groups while controlling the tumor volume was assessed by F test (pairwise comparison P values; details are presented in the Supplementary Statistical Analysis, available online). In (C) and (E), mean (SD) or mean with individual values, respectively, are shown; P values are from Mann-Whitney tests; *P < .05; **P < .01. IFN = interferon; P_i = inorganic phosphate.

Similar effects were observed in orthotopic 4T1 breast tumors. Electron paramagnetic resonance spectra taken before and after 15-day treatment with adenosine receptor A_2B antagonist or agonist exhibited amelioration or aggravation of metabolic stress, correlating with the decreased or increased number of tumor lung metastases, respectively (mean numbers of lung metastases for control vs PBS-603–treated mice: 15.50 vs 4, difference = ‒11.50, 95% confidence interval [CI] = 0 to 17.00, P = .02; for control vs BAY60-6583–treated mice: 15.50 vs 27.00, difference = 11.50, 95% CI = 0 to 37.00, n = 9-11, P = .02) (Figure 2, D and E).

Effect of adenosine receptor A_2B inhibition on tumor-infiltrating myeloid-derived suppressor cells and regulation of CREB-dependent adenosine-generating autocrine loop

Analysis of tumor and tumor-infiltrating leucocytes (TILs) in multiple models demonstrated that accumulation of CD11b⁺Gr1⁺ myeloid-derived suppressor cells, and the activity of the adenosine generation system in tumors was decreased by adenosine receptor A_2B antagonist (PSB-603, Figure 3, A-C). Adenosine receptor A_2B blockade led to reduced expression of a key adenosine-producing ectonucleotidase CD73 in CD45⁺ TILs and CD45^‒ cells in Lewis lung carcinoma tumors (mean [SD] proportions of CD45⁺CD73⁺ TILs and CD45^-CD73⁺ cells for control vs PBS-603–treated mice: 17.80% [2.03%] vs 7.84% [2.17%] and 15.27% [1.37%] vs 9.05% [1.38%]; P = .005 and P = .01, respectively; n = 10) (Figure 3, C). Restraining immunosuppressive factors by PSB-603 improved the TILs’ antitumor function (Figure 3, D).

Figure 3.

Impact of adenosine receptor A_2B inhibition on tumor accumulation of cells with a myeloid-derived suppressor cell phenotype, immune responses, adenosine-generating autocrine loop, and the efficacy of oncogene-targeted therapy. A-C) Blockade of adenosine receptor A_2B decreases tumor accumulation of CD11b⁺Gr1⁺ cells (A), decreases expression of CD73 ectonucleotidase in CD11b⁺Gr1⁺ cells (B) and in CD45⁺ and CD45^‒ cells in tumors (C) and increases the number of IFN-γ–producing tumor-infiltrating lymphocytes (D) in multiple tumor models; mean (SD), 8 to 10 mice per group in (A, C, D). E) Adenosine receptor A_2B signaling upregulates expression of CD73 in tumor cells through the protein kinase A / cAMP response element binding protein (PKA/CREB) pathway. Representative immunoblots demonstrate the expression of CD73 and pCREB in Lewis lung carcinoma and HCC4006 cells untreated or treated with pan-specific adenosine receptor agonist 5ʹ-N-ethylcarboxamidoadenosine, adenosine receptor A_2B antagonists PSB-603 and PBF-1129, adenosine receptor A_2B agonist BAY60-6583, adenosine receptors A₁ and A_2A antagonist ZM241385 or PKA inhibitor H89 for 36 hours (concentrations of reagents are indicated in the Supplementary Materials, available online). F, G) Adenosine receptor A_2B inhibition improves tumor and immune responses in combination with oncogene-targeted treatment. Transgenic mutant EGFRL858R mice with lung tumors were treated with erlotinib (25 mg/kg) alone or in combination with PSB-603 (10 mg/kg) from days 15 to 25 upon tumor induction with doxycycline. Lung tumor (white opacities) growth was evaluated by MRI and quantified for volume; representative MRI images at 60 days are shown (inserts). Immune cell subsets were evaluated in lung single-cell suspension of treated mice by fluorescence activated cell sorting at day 60; parental cell populations are indicated; mean (SD), 7 mice per group. P values are from Mann-Whitney tests; *P < .05; **P < .01. IFN = interferon; LLC = Lewis lung carcinoma; MRI = magnetic resonance imaging; NECA = 5ʹ-N-ethylcarboxamidoadenosine; TIL = tumor-infiltrating leucocyte.

Mouse and human adenosine receptor A_2B–expressing cancer cell (Lewis lung carcinoma, HCC4006) treatments with pan-specific analog 5ʹ-N-ethylcarboxamidoadenosine or adenosine receptor A_2B agonist BAY-606583 induced CD73 protein expression, which corresponded with the enhanced phosphorylation of cyclic adenosine 3′, 5′- monophosphate (cAMP) response element-binding protein (CREB) protein (Figure 3, E). This effect was abrogated by adenosine receptor A_2B antagonists PSB-603 and PBF-1129, adenosine receptor A_2A antagonist ZM241385, or protein kinase A inhibitor H89 (Figure 3, E). In mouse splenocytes, 5ʹ-N-ethylcarboxamidoadenosine enhanced the frequency of CD73-expressing cells, whereas inhibition of protein kinase A by H89 reversed this effect (Supplementary Figure 2, available online). Collectively, these data implicate the cAMP/protein kinase A/CREB pathway in regulating adenosine receptor A_2B–mediated expression of CD73 in cancer and immune cells.

Impact of adenosine receptor A_2B blockade on oncogene-targeted and antiangiogenic therapies in oncogenic mouse models

We evaluated the antitumor efficacy of adenosine receptor A_2B blockade in combination with oncogene-targeted therapy using tetracycline-inducible transgenic mice that express L858R mutant human epidermal growth factor receptor (EGFR) in lung epithelial cells (42). Mice were treated with EGFR inhibitor erlotinib alone or in combination with adenosine receptor A_2B antagonist PSB-603 for 10 days (days 15-25). Although both groups were responsive to erlotinib treatment, mice that received the combination therapy had delayed tumor recurrence and decreased lung tumor burden. Mean (SD) lung tumor volumes in the erlotinib group vs the erlotinib plus PSB-603 group at day 50 were 105.51 (25.93) vs 38.22 (12.37) mm³ and at day 60 were 132.15 (30.95) vs 44.01 (19.27) mm³, respectively (n = 7, P = .02 at day 50 and P = .003 at day 60) (Figure 3, F). Adenosine receptor A_2B antagonist administration greatly enhanced antitumor immune response. At day 60, the proportions of IFN-γ–producing TILs and CD11b⁺CD11c⁺ dendritic cells in diseased lungs were remarkably increased, while CD11b⁺Gr1⁺ myeloid-derived suppressor cell proportion decreased (Figure 3, G).

Lewis lung carcinoma mouse tumor is highly vascularized; a blocking anti–vascular endothelial growth factor antibody shows an antitumor effect in this model, as does the adenosine receptor A_2B antagonist PSB-603 (Supplementary Figure 3, A, available online). The combination treatment was significantly superior to any single treatment, however. Adenosine receptor A_2B inhibition improved antitumor immunity, augmented tumor infiltration by IFN-γ–producing lymphocytes, and skewed response toward Th1-type away from Th2 (Supplementary Figure 3, B, available online).

Modulation of the metabolically stressful tumor and draining lymph node microenvironment by clinically relevant adenosine receptor A_2B antagonist PBF-1129

We tested adenosine receptor A_2B antagonist PBF-1129, recently US Food and Drug Administration approved for clinical trials, for its ability to modulate metabolic and immunologic tumor microenvironment. The inhibition constants (K_i) against adenosine receptor A_2B were 24 nM and 35 nM in 2 different assays and more than 500 nM against the other adenosine receptors (Table 1). PBF-1129 inhibited the cAMP production induced by 10 µM 5ʹ-N-ethylcarboxamidoadenosine with K_B = 28 nM, demonstrating potent antagonistic activity and good selectivity against the other adenosine receptors (Table 1). Additional characteristics and pharmacology data for this compound are in Supplementary Tables 3 to 6 (available online).

Table 1.

Evaluation of inhibition constant for PBF-1129 and antagonistic activity of PBF-1129 in a cellular assay of cyclic adenosine monophosphate accumulation K_B (nM) for different adenosine receptors

Adenosine receptors	Ki, nM	Ki, nM	IC(90), nM	KB, nM
	Study 1	Study 2
A1	1060	649.5	NDa	1609
A2A	510	4866	26930	4270
A2B	35	23.9	596	28.1
A3	>100 000	>100 000	ND	ND

^aThe compound activity is too low to determine. IC = inhibitory concentration; K_i = inhibition constant; ND = not determined. K_B =equilibrium dissociation constant for a competitive antagonist.

In colon carcinoma CT26, PBF-1129 moderated metabolic tumor microenvironment by reversing acidosis, improving oxygenation, and decreasing P_i levels; adenosine receptor A_2B agonist BAY-606583 had the opposite effects, decreasing tissue oxygenation and enhancing P_i accumulation (Figure 4, A; Supplementary Figure 1, available online). They correlated with attenuated or accelerated tumor growth, respectively (Figure 4, B). A linear mixed-effects model with repeated measures demonstrated a statistically significant difference between control and PBF-1129– or BAY60-6583–treated groups, with P = .01 for each (n = 8-10; details of the analysis are in the Supplementary Statistical Analysis, available online).

Figure 4.

Effect of administration of clinically relevant adenosine receptor A_2B antagonist PBF-1129 on metabolic stress in the tumor and draining lymph node microenvironments, tumor growth, and antitumor immunity. A, B) Tumor microenvironment metabolic parameter changes (A) and growth (B) of CT26 tumor in mice untreated or treated with adenosine receptor A_2B antagonist PBF-1129 (100 mg/kg) or with adenosine receptor A_2B agonist BAY60-6583 (2 mg/kg); 0.2 × 10⁶ CT26 cells were inoculated subcutaneously in the flank. A) A monophosphonated triarylmethyl electron paramagnetic resonance spectroscopy probe was used for the analysis; y-axes show parameters change per 1 mm³ of tumor volume; the difference between treatment groups while controlling the tumor volume was assessed by F test (pairwise comparison P values; details are presented in the Supplementary Statistical Analysis, available online). B) A linear mixed-effects model with repeated measures was used to assess the association of tumor volume with the treatment groups across all time points. The Bonferroni method was used to adjust for multiple comparisons (pairwise comparison P values; details are presented in the Supplementary Statistical Analysis, available online), with 5 mice per group for 3 to 4 electron paramagnetic resonance spectroscopic measurements per each mouse; 8 to 10 mice were used for tumor growth experiments. C-F) Implantation of a lithium octa-n-butoxy-naphthalocyanine electron paramagnetic resonance spectroscopic probe into mouse lymph node (C), alterations of oxygen pressure (D), and numbers of IFN-γ–producing cells in tumor-draining lymph nodes (E) and Lewis lung carcinoma tumor volume (F) of untreated mice or mice treated with PBF-1129 (100 mg/kg). D) Four mice per group for electron paramagnetic resonance spectroscopy studies, with 6 electron paramagnetic resonance spectroscopic measurements per mouse; each dot represents 1 measurement for an individual mouse; data were approximated by plateau followed by 1 phase decay model and compared by nonlinear least square fit. IFN-γ–producing cells from draining lymph nodes were evaluated in parallel groups of mice that did not receive electron paramagnetic resonance spectroscopic probe implants, with 5 mice per group (E) and 8 to 10 mice per group for tumor volume evaluation measured at day 25 after inoculation (F); mean (SD), P values are from Mann-Whitney tests. G) Tumor microenvironment metabolic parameter changes in Lewis lung carcinoma subcutaneous tumor in mice untreated, treated with anti–PD-1 antibody (8 mg/kg), PBF-1129 (100 mg/kg), or a combination, with 5 mice per group for 3 to 4 electron paramagnetic resonance spectroscopic measurements per mouse. The difference between treatment groups while controlling the tumor volume, was assessed by F test (pairwise comparison; details are presented in the Supplementary Statistical Analysis, available online); *P < .01; **P < .005. EPR = electron paramagnetic resonance; IFN = interferon; LLC = Lewis lung carcinoma; PD-1 = programmed cell death 1 protein; P_i = inorganic phosphate.

Electron paramagnetic resonance spectroscopy monitoring in Lewis lung carcinoma tumors revealed that adenosine receptor A_2B inhibition affected the metabolic microenvironment and immune response in tumor-draining lymph nodes. A lithium octa-n-butoxy-naphthalocyanine electron paramagnetic resonance spectroscopy probe allows pO₂ evaluation in the interstitial milieu (43, 44). This microcrystal probe was implanted into mouse inguinal lymph node, and the baseline pO₂ levels were determined (Figure 4, C and D; the arrow indicates the probe deposit). Six days later, subcutaneous Lewis lung carcinoma tumors were established on the same flank. Mice were treated or not with PBF-1129. In lymph nodes of intact mice, oxygen levels were low compared with the tumor or normal tissues; during tumor progression, pO₂ was significantly decreased in the untreated mice (Figure 4, D). Lewis lung carcinoma tumor cells appear in draining lymph nodes between days 10 and 14 after inoculation; thus, the observed pO₂ decline could be the result of the high oxygen consumption by proliferating lymphocytes or tumor cells. Pharmacologic inhibition of adenosine receptor A_2B retained oxygen levels closer to normal (Figure 4, D). This effect correlated with the improved draining T-cell function, as assessed by IFN-γ production and decreased tumor volume (Figure 4, E and F; mean [SD] numbers of IFN-γ–producing cells in control vs PBF-1129–treated groups were 4.91 [0.71] and 12.9 [2.40], respectively; n = 5, P = .005). Thus, clinically relevant adenosine receptor A_2B blockade could help T- and other immune cell proliferation by improving lymph node oxygenation and supporting their metabolic requirements in the secondary immune organs.

Effect of PBF-1129 on metabolic stress, immunosuppression, and efficacy of anti–PD-1 immunotherapy

We investigated the effect of adenosine receptor A_2B receptor antagonist PBF-1129 on anti–programmed cell death 1 protein (PD-1) therapy in relation to tumor microenvironment metabolic parameters, tumor growth, and antitumor immunity. Anti–PD-1 treatment alone aggravates the metabolic stress in Lewis lung carcinoma tumors by increasing acidosis and hypoxia, whereas adenosine receptor A_2B inhibition with PBF-1129 makes the microenvironment less stressful during combination treatment (Figure 4, G). A linear regression model to evaluate the association of the tumor microenvironment parameters with the treatment groups while controlling the tumor volume as a cofounder confirmed that pH, pO₂, and P_i were statistically significant different between treatment groups (see Supplementary Statistical Analysis, available online).

Lewis lung carcinoma tumor–bearing mice receiving PBF-1129 demonstrated reduced tumor growth and increased survival (Figure 5, A and B), whereas anti–PD-1 treatment showed marginal efficacy. Linear mixed-effect statistical analysis to account for the association among repeated measures showed that the anti–PD-1 and PBF-1129 combination was highly efficient at reducing tumor growth (Figure 5, A; details are in the Supplementary Statistical Analysis, available online). The combination treatment provided survival benefit superior to any single treatment (mean survival probability for control, anti–PD-1, PBF-1129, and combination groups were 23, 25.5, 42.5, and 31.5 days, respectively; control vs PBF-1129, hazard ratio = 10.04, 95% CI = 2.88 to 34.96, P < .001; anti–PD-1 vs combination, hazard ratio = 11.74, 95% CI = 3.35 to 41.13, P < .001; PBF-1129 vs combination hazard ratio = 3.78, 95% CI = 1.27 to 11.17, P < .005, 2-sided F test).

Figure 5.

Effect of treatment by adenosine receptor A_2B antagonist PBF-1129 in combination with anti–PD-1 immunotherapy on tumor growth, survival, immunosuppression, activity of the adenosine-generating system, and antitumor immune responses. Animals were treated, as in Figure 4, G, 10 mice per group. A, B) Decreased tumor growth (A) and enhanced survival benefit (B) of PBF-1129 treatment alone or in combination with anti–PD-1 antibody in Lewis lung carcinoma tumor-bearing mice. A) Differences between treatment groups while controlling the tumor volume were assessed by F test (pairwise comparison P values; details are presented in the Supplementary Statistical Analysis, available online). B) PBF-1129 vs control, PBF-1129-positive anti–PD-1 vs PBF-1129, and PBF-1129-positive anti–PD-1 vs control P < .001, P < .005, and P < .001, respectively; log-rank test. C) Decreased tumor accumulation of myeloid-derived suppressor cells and enhanced infiltration by CD8 and CD11c⁺ dendritic cells moderated the activity of the adenosine-generating enzymatic system and increased expression of adenosine deaminase in tumors of mice that received PBF-1129 or combination treatment with an anti–PD-1 antibody; proportions of immune cell types within parental cell subsets and parental subsets are shown. D, E) Reduced functional activity of myeloid-derived suppressor cells as evaluated by the ability to inhibit T-cell proliferation and IFN-γ production (D) and enhanced cytotoxicity of tumor-infiltrating T cells assessed by the expression of markers of cytotoxicity (E) in mice treated with PBF-1129 or combination with an anti–PD-1 antibody. Ratio of 2 types of myeloid-derived suppressor cells from different groups of experimental mice to donor T cells was 1:5 (D); mean (SD), n = 8, P values are from Mann-Whitney tests; *P < .01; **P < .008. DC = dendritic cell; IFN = interferon; M = myeloid; MDSC = myeloid-derived suppressor cell; PD-1 = programmed cell death 1 protein; PMN = polymorphonuclear.

Adenosine receptor A_2B blockade resulted in positive changes in the distribution and functionality of TILs (Figure 5, C-E). PBF-1129 and the combination regimen decreased immunosuppression by reducing tumor accumulation of CD11b⁺Gr1⁺ cells, including CD11b⁺Gr1⁺Ly6C⁺ myeloid and CD11b⁺Gr1⁺Ly6G⁺ polymorphonuclear myeloid-derived suppressor cells with T-cell inhibitory activity. Expression of adenosine-generating ecto-5′-nucleotidase CD73 was significantly reduced in infiltrating myeloid-derived suppressor cells, CD45⁺ cells, and CD45^-EpCAM⁺ tumor cells, whereas T cells had significantly decreased surface ectonucleotidase CD39 (Figure 5, C). In the PBF-1129 and combination treatment groups, CD45^-EpCAM⁺ tumor cells had upregulated expression of adenosine deaminase compared with control (mean [SD] = 16.88 [3.97] and 17.5 [4.09] vs 6.55 [1.97], respectively; n = 8, P = .01 and P = .008, respectively) (Figure 5, C). Thus, adenosine receptor A_2B blockade diminished the extracellular adenosine-generating pathway activity but enhanced adenosine-degrading pathway activity in tumor tissue.

Tumor CD11b⁺Gr1⁺ cells effectively suppressed T-cell activity in a dose-dependent manner (Supplementary Figure 4, available online). Myeloid-derived suppressor cells from groups of mice untreated or treated with PBF-1129, anti–PD-1, or combination demonstrated different activity in inhibiting donor T-cell function, as evaluated by T-cell proliferation or IFN-γ production. Myeloid-derived suppressor cells from combination-treated mice were less immunosuppressive than cells from control mice with polymorphonuclear- or myeloid-type myeloid-derived suppressor cells following similar patterns (Figure 5, D; flow cytometry gating strategy shown in Supplementary Figure 5, available online). Chemokine receptor CCR1 expression in myeloid-type myeloid-derived suppressor cells was lower in the combination treatment group than in control (Supplementary Table 7, available online), suggesting that the combination therapy moderates myeloid-derived suppressor cells’ functional activity. The metabolic and immune effects of the combination treatment correlated with the improved tumor infiltration, CD8⁺ cell activation, and enhanced T-cell cytotoxicity, as assessed by granzyme-B upregulation and IFN-γ expression (Figure 5, C and E).

In plasma, interleukin 22 levels increased significantly in the PBF-1129 and combination groups, whereas interleukin 9 levels were elevated in anti–PD-1 or combination treatments (Supplementary Table 8, available online). Plasma lactic acid concentrations were unchanged between groups (not shown). These data imply moderate systemic but robust immune effects of adenosine receptor A_2B–blocking therapy alone or in combination with anti–PD-1 antibody.

Adenosine receptor A_2B expression in human lung tumors as a negative prognostic indicator of patient survival

RNA expression analysis of multiple human and mouse lung tumor cell lines revealed that A_2B and A_2A receptors and CD73 ectonucleotidase are predominant in most of the tested cells (Figure 6, A;Supplementary Figure 6, available online). Immunoblot analysis confirmed RNA expression data (Figure 6, A insert).

Figure 6.

Expression of adenosine receptor A_2B in lung cancer cell lines and tissues and correlation with CD73 expression and survival of patients with lung cancer. A) RNA was isolated from cultured human and mouse lung tumor cells, and levels of messenger RNA transcripts of adenosine system genes were determined by quantitative RT-PCR. Mean (SD), n = 5. Note the expression of adenosine receptor A_2B and CD73 ectonucleotidase in most of the cell lines; (Insert) adenosine receptor A_2B protein expression in human lung tumor cells, as assessed by immunoblot. B) Box and whisker plot (10th-90th percentile) demonstrating expression levels of adenosine receptors in normal human lung and lung cancers. Expression of the receptors was evaluated using shotgun proteomics analysis. Number of patients: 19 with NL, 22 with LUAD, 19 with LUSC, and 7 with LCC; P values are from Mann-Whitney tests; *P < .05; **P < .001. C-E) Correlation analysis of the expression of A2BAR and CD73 genes (C) and survival plots in relation to expressions of A2BAR (D) and CD73 genes (E) for 517 patients with LUAD using publicly available RNA-seq expression and survival data from The Cancer Genome Atlas database. C) The Spearman rank-based correlation P value is shown. In (D) and (E), stratification of high and low expression was made at quantiles 0.55 and 0.45, respectively; log-rank test. LCC = lung large cell carcinoma; LLC = Lewis lung carcinoma; LUAD = lung adenocarcinoma; LUSC = lung squamous carcinoma; NL = normal lung tissue; RT-PCR = reverse transcriptase–polymerase chain reaction.

Proteomic analysis of human lung tumors vs normal lung tissue showed that only adenosine receptor A_2B was significantly upregulated in lung adenocarcinoma and large cell carcinoma tumor tissues compared with normal tissue (Figure 6, B).

RNA-seq and survival data from The Cancer Genome Atlas correlated with A2BAR and CD73 expression in lung adenocarcinoma tumors, confirming that adenosine receptor A_2B signaling can upregulate CD73 gene expression (Figure 6, C). Higher expression of A2BAR and CD73 was associated with poor survival (Figure 6, D and E), demonstrating that the adverse effect was possibly mediated through their regulation of immune and metabolic tumor microenvironment.

Pharmacologic safety and immunologic impact of PBF-1129 treatment in patients with NSCLC

PBF-1129 represents a novel, potent, orally administered, selective antagonist of the adenosine A_2B receptor developed at Palobiofarma, SL (Barcelona, Spain) as a first-in-class drug for the treatment of lung cancer. Pharmacokinetic, safety, and toxicity data for the PBF-1129 compound are in Supplementary Data and Supplementary Tables 3 to 6 (available online).

We are currently conducting a phase I study of the adenosine receptor A_2B antagonist PBF-1129 in patients with NSCLC who have progressed on standard-of-care therapies (NCT03274479). Under a written informed consent (see Supplementary Methods, available online), patients received 40, 80, or 160 mg daily of oral PBF-1129. Pharmacokinetics studies confirmed full adenosine receptor A₂ target engagement at the 80-mg dose (Supplementary Tables 3-6, available online). We collected peripheral blood before and after the first 28-day treatment cycle and performed peripheral blood mononuclear cell phenotypic analyses from 7 patients with NSCLC (1 patient received a 40-mg dose and did not have evaluable measures). Wilcoxon matched-pairs signed rank tests showed a significant post-treatment decrease of HLA-DR⁺ cells expressing CD39 (mean [SD] = 23.98 [10.99] vs 16.47 [10.85], P = .03), and CD73 (mean [SD] = 17.46 [8.03] vs 9.49 [5.31], P = .01), HLA-DR^-CD11b⁺CD33⁺ myeloid-derived suppressor cells (mean [SD] = 3.12 [1.91] vs 1.86 [1.13], P = .02), and CD4⁺CD25⁺CD127⁺ Treg cells (mean [SD] = 2.70 [0.48] vs 2.19 [0.73], P = .05) (Figure 7, A and B; n = 7). Evaluation of plasma cytokines relevant to the antitumor immune responses did not show differences between pre-and post-treatment levels. Although accrual to this dose-escalation study is ongoing, data confirm the pharmacologic efficacy of PBF-1129 and indicate its ability to modulate adenosine system and immune response parameters in patients with NSCLC.

Figure 7.

Immunophenotyping of peripheral blood mononuclear cells from patients with non-small cell lung cancer before and after treatment with the adenosine receptor A_2B antagonist PBF-1129. A) Blood was drawn before and after the first treatment cycle, peripheral blood mononuclear cells were analyzed by flow cytometry, and cell proportions within parental subsets were evaluated. B) Representative dot plots demonstrate pre- and post-treatment proportions of cell subsets within peripheral blood mononuclear cell lymphocyte fractions. PBF-1129 daily oral doses (40, 80, or 160 mg) are indicated, and P values from Wilcoxon matched-pairs signed rank tests are shown; n = 7; P < .05.

Discussion

We establish the link among metabolic tumor microenvironment parameters, tumor characteristics, and tumor-infiltrating immune cell functionality. Data reveal that manipulation of adenosine receptor A_2B signaling has profound effects on tumor interstitial pH, pO₂, and P_i levels and shapes metabolic and immune tumor microenvironment, thus presenting a novel therapeutic approach for tumor microenvironment modulation. Inhibition of adenosine receptor A_2B moderates tumor acidosis and hypoxia, making the tumor microenvironment less metabolically stressful and more immunologically permissive. Interstitial P_i has emerged as a highly correlative factor and a cumulative measure of tumor microenvironment stress and adenosine-mediated immunosuppression.

Altered metabolism and reliance on aerobic glycolysis led to tumor microenvironment acidification. Adaptation to acidosis supports malignancy through increased invasion and metastasis, drug resistance, and immunosuppression (45). Our electron paramagnetic resonance spectroscopy data and the adenosine receptor A_2B signaling targets point to the profound involvement of adenosine receptor A_2B in regulating tumor metabolism. This effect is likely through activities of mitogen-activated protein kinases, including extracellular signal-regulated kinase (ERK), c-Jun N-terminal kinase (JNK) and phosphatidylinositol 3-kinase /protein kinase B (AKT) PI3K/AKT, implicated in regulation of glycolytic and metabolic enzymes (46, 47). Adenosine receptor A_2B activates the Rap1-MEK-ERK pathway through adenylate cyclase stimulation and cAMP accumulation (48). Protein kinase A activation by cAMP and adenosine receptor A_2B–activated protein kinase C can mediate JNK activation (47, 49, 50). cAMP cellular accumulation could lead to PI3K activation via Epac1. Subsequent PI3K-mediated AKT/mTOR axis activation links adenosine receptor A_2B to HIF-1α transcription factor, a key regulator of glycolytic enzymes (47).

Adenosine was shown to improve the oxygen Supply and demand ratio, had vasodilatory effect, and stimulated vascular endothelial growth factor production (25, 48, 51, 52). We hypothesized that the hypoxia decrease upon adenosine receptor A_2B blockade was the result of the reduced tumor oxygen consumption by decreased tumor cell proliferation. The hypoxia levels correlated with the tumor growth in animals treated with adenosine receptor A_2B antagonist or agonist. Amelioration of the metabolic stress and hypoxia significantly benefits proliferating T- and other immune cells (53, 54).

Our studies with an oxygen-sensitive electron paramagnetic resonance spectroscopy probe implanted into lymph nodes generated unique data directly characterizing metabolic alterations in tumor-draining lymph nodes. Increased metabolic demand by activated T cells and tumor metastasis results in highly hypoxic microenvironment. This finding suggests that metabolic factors could play an immunoregulatory role in lymph nodes at the level of T-cell priming and that adenosine receptor A_2B blockade could ameliorate hypoxia in lymph nodes. Electronic paramagnetic resonance spectroscopy studies demonstrated that immunotherapeutics could aggravate the metabolic tumor microenvironment. As acidosis and hypoxia are harmful for T-cell antitumor activity (55, 56), including adenosine receptor A_2B antagonist in the combination regimen could ameliorate this adverse effect.

Targeting adenosine receptor A_2B proved to be an effective approach to interfering with the adenosine-generating system and disrupting an adenosine-mediated immunosuppressive loop by downregulating adenosine-generating enzymes and upregulating adenosine-degrading adenosine deaminase. Our data implicate the role of the cAMP/PKA/CREB pathway in adenosine receptor A_2B–mediated CD73 upregulation. Adenosine receptor A_2B can enhance CD39 and CD73 expression by stabilizing the expression of HIF-1α, a key factor driving their expression in tumor, stromal, and infiltrating Treg cells as well as myeloid-derived suppressor cells (18). Interference with adenosine receptor A_2B signaling breaks this pathway. Studies have reported a profound effect of adenosine receptor A_2B inhibition on myeloid and lymphoid immune compartments in cancer and improved antitumor immunity (24, 25, 57, 58). Our investigation demonstrates that in addition to blocking immunosuppressive pathways, these effects are mediated through metabolic tumor microenvironment stress amelioration and adenosine downregulation.

Human lung tumor tissue analysis and The Cancer Genome Atlas data showed adenosine receptor A_2B upregulation in lung tumors compared with lung tissue and a negative correlation with overall survival in patients with adenocarcinoma. This finding corroborates other studies showing that adenosine receptor A_2B is the predominant receptor expressed more strongly in tumors than in normal tissue (46, 59, 60).

Several A_2A but only few A_2B-adenosine receptor antagonists have been studied in clinical trials (4, 23-25). We are conducting a clinical trial to evaluate the safety profile and immunologic efficacy of the adenosine receptor A_2B antagonist PBF-1129 in patients with NSCLC. PBF-1129 demonstrated pharmacologic efficacy, as measured by CD39 and CD73 decrease in antigen-presenting cells. Collectively, our data with PBF-1129 and other adenosine receptor A_2B antagonists identify the A_2B-adenosine receptor as a valuable therapeutic target to reduce tumor-supporting immunosuppression and modify tumor microenvironment metabolic conditions to enhance the efficacy of cancer immune checkpoint therapies.

Data availability

The data supporting the presented findings are available in the article and in its online supplementary material. The study data will be shared on request to the corresponding author. Information on the clinical trial can be found at https://clinicaltrials.gov/ct2/show/NCT03274479. Deidentified data from the clinical trial can be requested from the study principal investigator (Dwight Owen, MD, MS, dwight.owen@osumc.edu) under a data use agreement from The Ohio State University.

Author contributions

Jason V. Evans, PhD (Data curation), Valery V. Khramtsov, PhD (Conceptualization; Methodology; Supervision; Writing—original draft), Sergey V. Novitskiy, PhD (Data curation; Formal analysis; Methodology), David P. Carbone, MD, PhD (Conceptualization; Methodology; Supervision), Jay L. Zweier, MD (Methodology; Supervision), Daniel J. Spakowicz, PhD (Formal analysis; Writing—review & editing), Dwight H. Owen, MD, PhD (Conceptualization; Data curation; Supervision; Writing—original draft; Writing—review & editing), Thanigaivelan Kanagasabai, PhD (Data curation; Writing—review & editing), Alla Ivanova, PhD (Writing—review & editing), Julio Castro, MD, PhD (Methodology; Supervision; Writing—original draft), Takefumi Kikuchi, PhD (Data curation), Kyungho P. Park, BS (Data curation), Igor Feoktistov, MD, PhD (Data curation), Sergey V. Ryzhov, PhD (Data curation), Aaditya Pallerla, BS (Formal analysis), Anil Shanker, PhD (Conceptualization; Methodology; Supervision; Writing—review & editing), Lai Wei, PhD (Formal analysis), Joseph Amann, PhD (Data curation), Alexandre Samouilov, PhD (Data curation; Formal analysis), Rebecca Hoyd, PhD (Data curation; Formal analysis), Roman V. Uzhachenko, PhD (Data curation), Benoit Driesschaert, PhD (Data curation), Andrey A. Bobko, PhD (Data curation; Formal analysis), Irina Ilgisonis, MD, PhD (Data curation), Longzhu Piao, PhD (Data curation), Anneliese Antonucci, MS (Data curation), Rajeswara Rao Arasada, PhD (Data curation), Shuxiao Guan, PhD (Data curation), Elena E. Tchekneva, MD (Conceptualization; Data curation; Writing—original draft; Writing—review & editing), Mounika Uttam L. Goruganthu, MS (Data curation), Shankar Suman, PhD (Data curation), Ruohan Wu, PhD (Data curation), Mikhail M. Dikov, PhD (Conceptualization; Formal analysis; Methodology; Supervision; Writing—original draft; Writing—review & editing).

Funding

The work was supported by National Institutes of Health grants R01 CA248741 and R01 CA175370 (M.M.D. and D.P.C.) and R01 CA138923 (M.M.D.); Dallapezze Fund (M.M.D. and D.P.C.); OSU Pelotonia Awards (D.H.O. and M.M.D.) and (D.H.O. and D.J.S.); Meharry Clinical and Translational Research Center Pilot grant U54 MD007593, U54 CA163069, and SC1 CA182843 (An.S.); R01 CA 149013 and CA192064 (V.V.K.). Owen is a Paul Calabresi Scholar supported by the OSU K12 Training Grant for Clinical Investigators (K12 CA133250).

Conflicts of interest

D.H.O. received a research grant for conducting clinical trial NCT03274479, and M.M.D. received a research grant for conducting preclinical studies with PBF-1129, respectively, from Palobiofarma S.L. Other authors declare no conflict of interest.