ORIGINAL ARTICLE

Systemic oxygenation weakens the hypoxia and hypoxia induciblefactor 1α-dependent and extracellular adenosine-mediatedtumor protection

Stephen M. Hatfield & Jorgen Kjaergaard & Dmitriy Lukashev & Bryan Belikoff &Taylor H. Schreiber & Shalini Sethumadhavan & Robert Abbott & Phaethon Philbrook &

Molly Thayer & Dai Shujia & Scott Rodig & Jeffrey L. Kutok & Jin Ren & Akio Ohta &

Eckhard R. Podack & Barry Karger & Edwin K. Jackson & Michail Sitkovsky

Received: 29 April 2014 /Revised: 17 June 2014 /Accepted: 4 July 2014# Springer-Verlag Berlin Heidelberg 2014

AbstractIntratumoral hypoxia and hypoxia inducible factor-1α(HIF-1-α)-dependent CD39/CD73 ectoenzymes may governthe accumulation of tumor-protecting extracellular adenosineand signaling through A2A adenosine receptors (A2AR) intumor microenvironments (TME). Here, we explored the con-ceptually novel motivation to use supplemental oxygen as a

treatment to inhibit the hypoxia/HIF-1α-CD39/CD73-drivenaccumulation of extracellular adenosine in the TME in orderto weaken the tumor protection. We report that hyperoxicbreathing (60 % O2) decreased the TME hypoxia, as well aslevels of HIF-1α and downstream target proteins of HIF-1α inthe TME according to proteomic studies in mice. Importantly,oxygenation also downregulated the expression of adenosine-generating ectoenzymes and significantly lowered levels oftumor-protecting extracellular adenosine in the TME. Usingsupplemental oxygen as a tool in studies of the TME, we alsoidentified FHL-1 as a potentially useful marker for the conver-sion of hypoxic into normoxic TME. Hyperoxic breathingresulted in the upregulation of antigen-presenting MHC class Imolecules on tumor cells and in the better recognition andincreased susceptibility to killing by tumor-reactive cytotoxicT cells. Therapeutic breathing of 60 % oxygen resulted in thesignificant inhibition of growth of established B16.F10melanoma tumors and prolonged survival of mice. Takentogether, the data presented here provide proof-of principlefor the therapeutic potential of systemic oxygenation toconvert the hypoxic, adenosine-rich and tumor-protectingTME into a normoxic and extracellular adenosine-poorTME that, in turn, may facilitate tumor regression. Wepropose to explore the combination of supplemental oxy-gen with existing immunotherapies of cancer.

Key messages& Oxygenation decreases levels of tumor protecting

hypoxia.& Oxygenation decreases levels of tumor protecting extra-

cellular adenosine.& Oxygenation decreases expression of HIF-1alpha depen-

dent tumor-protecting proteins.

Electronic supplementary material The online version of this article(doi:10.1007/s00109-014-1189-3) contains supplementary material,which is available to authorized users.

S. M. Hatfield (*) : J. Kjaergaard :D. Lukashev : B. Belikoff :S. Sethumadhavan :R. Abbott : P. Philbrook :M. Thayer :A. Ohta :M. Sitkovsky (*)New England Inflammation and Tissue Protection Institute,Northeastern University, 360 Huntington Avenue, Boston,MA 02115, USAe-mail: s.hatfield@neu.edue-mail: m.sitkovsky@neu.edu

M. SitkovskyCancer Vaccine Center, Dana Farber Cancer Institute, HarvardInstitutes of Medicine, Boston, MA 02115, USA

T. H. Schreiber : E. R. PodackDepartment of Microbiology and Immunology, Miller School ofMedicine, University of Miami, Miami, FL 33136, USA

S. Rodig : J. L. KutokDepartment of Pathology, Brigham and Women’s Hospital, HarvardMedical School, Boston, MA 02115, USA

D. Shujia : B. KargerBarnett Institute and Department of Chemistry and ChemicalBiology, Northeastern University, Boston, MA 02115, USA

J. Ren : E. K. JacksonDepartment of Pharmacology and Chemical Biology, University ofPittsburgh School of Medicine, Pittsburgh, PA 15219, USA

J Mol MedDOI 10.1007/s00109-014-1189-3

& Oxygenation increases MHC class I expression and en-ables tumor regression.

Keywords Hypoxia . HIF-1α . Adenosine . CD73 . Cancer .

Immunology

Introduction

The intratumoral hypoxia and hypoxia inducible factor-1α(HIF-1α)-dependent pathways are tumor-protecting [1] andwere shown to upregulate the tandem activity of ectoenzymesCD39 (ecto-ATPase/ADPase) and CD73 (ecto-5′-nucleotidase)[2]. These enzymes represent the major source of extracellularadenosine in tissue microenvironments including the tumormicroenvironment (TME) [3].

The hypoxia-HIF-1α-CD39/CD73-mediated accumulationof adenosine results in signaling through the cAMP-elevatingA2A and A2B (A2AR/A2BR) adenosine receptors (Fig. 1).Extracellular adenosine and A2AR/A2BR were shown to becritical and nonredundant in the protection of inflamed normaltissues [4–11] and in the protection of tumors from the anti-tumor immune response [3–5].

There is a well-appreciated need in therapeutic treatmentsto (1) weaken tumor hypoxia and/or (2) inhibit HIF-1α [1,6–8], (3) inhibit CD39/CD73-mediated adenosine formation[9, 10], and (4) antagonize the A2AR/A2BR in the TME [3, 4,11–13]. In view of the importance of extracellular adenosinein tumor protection, one of the goals of this research was totest the hypothesis whether the early stages of hypoxia→HIF-1α signaling govern the downstream molecular events of thehypoxia-adenosinergic pathway (i.e., TME hypoxia → HIF-1α → CD39/CD73 → [Adenosine] → A2AR/A2BR →intracellular cAMP → PKA→ cAMP response element(CRE) → CREB pathway) (Fig. S1). Recently, the conceptof extracellular adenosine-mediated tumor protection has re-ceived additional strong support from the analysis of massivehuman data sets from more than 6,000 triple negative andchemotherapy-resistant breast cancers. Resistance to chemo-therapy and immunotherapy was shown to be due to anoverexpression of extracellular adenosine-generating CD73,resulting in stronger immunosuppression by adenosine-A2AR/A2BR signaling in the TME [14].

Here, we aimed to address the need in novel approaches todecrease the tumor-protecting effects of tumor hypoxia. Indeed,it could be therapeutically valuable to decrease the intensity oftumor hypoxia-driven and CD73-mediated accumulation ofextracellular adenosine in the TME by using already availableand safe drugs or treatments. We reasoned that by inhibiting theupstream events of hypoxia-HIF-1α-governed signaling path-ways, oxygenation may also inhibit the downstreamCD39/CD73-[extracellular adenosine]high-A2AR/A2BR-mediated tumor protection in the TME (Fig. 1). This, in turn,

may enable tumor rejection. The experimental observationsdescribed here validate these assumptions.

Methods

Animals Female C57BL/6 N (B6) mice, 8–12 weeks old,were purchased from Charles River Laboratories. These ani-mals were housed in a specific pathogen-free environmentaccording to the National Institute of Health guidelines. Allanimal experimentation conforms to protocols approved bythe Institutional Animal Care and Use Committee.

Tumors MCA205 fibrosarcoma is a 3-methylcholanthrene-induced tumor of B6 origin, and B16-F10.P1 is a poorlyimmunogenic subclone of the spontaneously arising B16/BL6 melanoma (SI, Ref. 10-12). B16-F10.P1 is the parentstrain of CL8-1 (transfected with MHC class I) [4]. To estab-lish tumors, B6mice were injected s.c. (1×105) or i.v. (3×105)with tumor cells resuspended in 200 μL of HBSS. Mice withsubcutaneous tumors were placed in either 21 or 60 % oxygenafter tumors became palpable (between days 7 and 10). Tumorsize was measured in two dimensions using vernier calipers.

Hyperoxic breathing Mice were placed in small animal careunits with well-controlled gas composition to mimic protocolsof supplemental oxygen delivery to humans [15]. Self-contained oxygen generators (Airsep) were used to ensuredesired levels of oxygen were maintained inside each unit.Hypercapnic acidosis was avoided by replacing traditionalmouse cage tops with aerated wire lids and by using Sodasorb(Grace &Co.) (SI, Ref. 1, 2). CO2 levels inside the chamber didnot exceed 0.4 %, while hypercapnia typically occurs at levelshigher than 2 %. Levels of CO2 and O2 in the chambers werecarefully monitored (see full details in supporting information).

Hypoxyprobe analysis Hypoxyprobe™-1 Plus Kit (Millipore)was used to detect hypoxia in tumors. Mice with establishedMCA205 tumors were placed in 21 or 60 % oxygen for 3 hthen injected with 80 mg/kg of Hypoxyprobe. After 1.5 h oflabeling, lungs were snap frozen in n-hexane and 4-μm cryo-sections were prepared from 10–20 different cutting surfaces,and immunohistochemistry was performed.

Immunohistochemistry Immunohistochemistry was per-formed at Harvard Medical School Pathology Department atBrigham and Women’s Hospital. Four-micrometer thickacetone-fixed, O.C.T-embedded tissue sections were pre-pared. The slides were soaked in -20 °C methanol-acetic acidfor 2 min then air-dried for 20min at room temperature. Slideswere pretreated with Peroxidase Block (DAKO). Primaryanti-Hypoxyprobe (Millipore) was applied at a concentrationof 1:100 at room temperature for 1 h. Rabbit anti-rat

J Mol Med

immunoglobulin antibody was applied at a concentration of1:750 in DAKO diluent for 1 h. Slides were stained with anti-rabbit Envision+kit (DAKO). Immunoperoxidase staining wasdeveloped using a DAB chromogen (DAKO) and counter-stained with hematoxylin. Tumor hypoxia was annotated usingSpectrum™ Plus and Aperio’s ScanScope® slide scanners bythe Pathology Department at Brigham and Women’s Hospital.See SI for additional details of fluorescent immunostaining.

Whole cell proteomics Mice bearing 7-day establishedMCA205 intradermal tumors were placed in either 21 or60 % oxygen. On day 21, tumors were excised and homoge-nized in protein extraction lysing buffer as recommended bymanufacturer (Thermo Fisher Scientific). After denaturing

and enzymatic digestion by trypsin, the 21 % O2 and 60 %O2 samples were labeled with differential stable-isotopiclabeling using dimethylation reaction. A two-dimensionalHPLC separation was performed by high/low pH reversephase HPLC and a Q-Exactive Orbitrap mass spectrometer(Thermo Fisher Scientific) to acquire the MS and MS/MSspectra of isotopically labeled peptide signals. Leading edgeanalysis using gene set enrichment analysis (GSEA) softwareshowed that five gene sets were highly correlated with theobserved HIF-1α regulated proteins. Only those protein setsthat fall into <0.25 false discovery rate (FDR) and weresignificant in the GSEA output are shown (SI, Ref 3-6). SeeSI for full details on proteomics and assays to prescreenhypoxic versus normoxic TMEs.

Fig. 1 Hyperoxia reduces intratumoral hypoxia and HIF-1α,downregulates HIF-1α-target genes, and increases levels of HIF-1αregulating proteins including four and a half LIM domains-1 (FHL-1). aImmunohistochemical staining of intratumoral hypoxia using the in vivohypoxia marker, Hypoxyprobe. Positive Hypoxyprobe staining is shownin brown in the light micrograph. Scale bars: 200 μm. Shown is theaverage intensity of staining OD/mm2 (mean±SEM,P<0.01, n=3). MicebearingMCA205 pulmonary tumors breathing 60% oxygen for 3 h wereinjected i.v. with Hypoxyprobe. b Expression of hypoxia-HIF-1α-depen-dent gene products in the TME as demonstrated by comparative wholecell proteomics analysis of known HIF-1α targets in the TME. The heatmap shows the clustered genes in the leading edge subsets with theexpression values represented as colors. The set-to-set graph (left) usescolor intensity to show the overlap between subsets: the darker the color

(on the left in green), the greater the overlap between the subsets. Leadingedge analysis using GSEA software showed that five gene sets werehighly correlated with the observed HIF-α regulated proteins. Only thoseprotein sets that fall into <0.25 false discovery rate (FDR) and weresignificant in the GSEA output are shown, as described previously (SI,Ref. 4-6). c The full names of hypoxia-regulated proteins identified bywhole cell proteome analysis. d Proteomics analysis demonstrate HIF-1αregulating proteins (FIH, VHL, and FHL-1) are increased in tumors frommice breathing 60 % oxygen. Tumor sections representing hypoxicregions (HIF-1αhigh) from mice breathing 21 % O2 were compared tonormoxic regions of the TME in mice breathing 60 % oxygen. e Immu-noblot assays from tumors of mice breathing 21 or 60 % oxygen. Shownare representative samples of HIF-1α (left panel) and FHL-1 (right panel)from five independent assays

J Mol Med

Western blot Mice with established tumors were placed ineither 21 or 60 % oxygen for 10 days, and tumors were snapfrozen in liquid nitrogen and then homogenized in lysis buffer.After fractionation using SDS/PAGE followed by semidrytransfer, samples were detected using mAb (1:500, SantaCruz) against mouse HIF-1α, FHL-1, or VEGF. Levels ofbeta-actin were detected using anti-mouse beta-actin mAb(1:5,000, Sigma-Aldrich). See SI for additional details.

In vivo extracellular adenosine measurements Mice withestablished MCA205 tumors were used to analyze levelsof extracellular adenosine. Microdialysis probes(Bioanalytical Systems) were placed in the center, edge,and in normal tissue next to tumors. Microdialysate(isotonic saline, 100 U of heparin, and 20 mM EHNAadenosine deaminase inhibitor) was perfused at 2mL/min for 2.5 h using an infusion pump (BraintreeScientific). Following the initial probe and an equilibra-tion period, mice were placed in 21 or 60 % oxygen for3 h and re-probed for levels of adenosine in the sametumor area. Adenosine levels were measured byreversed-phase liquid chromatography-tandem massspectrometry using a triple quadrupole mass spectrome-ter with 13C

10-adenosine as an internal standard (SI,Ref. 7-9). See SI for additional details.

RT-PCR Mice with established MCA205 tumors were placedin either 21 or 60 % oxygen for 72 h. RT-PCR was performedusing the RT2 SYBR green PCR master mix (Superarray) onan Applied Biosciences 7300 PCR platform. Primers werepurchased from SABiosciences (Qiagen, Valencia, CA).House-keeping genes were included as controls.

Cytotoxicity assay MCA205 cells were cultured in 21 % ox-ygen or 60 % oxygen for 4 days in vitro. Cells were collectedby trypsinization and labeled with 100 μCi sodium chromate(Amersham Biosciences) per 2×106 cells for 1 h at 37 °C. Invitro-activated T cells from the tumor-draining lymph nodesof MCA205 inoculated mice were used as effector cells andseeded into 96-well round-bottomed plates at the indicatedeffector/target (E/T) ratios. 51Cr-labeledMCA-205 cells at 1×104 cells per well were used as target cells. After 4 h incuba-tion at 37 °C, radioactivity released from lysed target cells wascounted on a γ-counter. See SI for calculations of percentspecific lysis.

Flow cytometry Solid or pulmonary tumors were homoge-nized and passed through a 70-μm strainer. Tumor cells andlymphocytes were recovered using 40 % Percoll separation,incubated with monoclonal antibodies (1:50, BD Pharmingenand eBioscience) in FACS buffer (PBS+0.5 % BSA) andacquired on a FACSCalibur. Analysis was performed usingCellQuest (BD Biosciences) and FlowJo (Tree Star) software.

Statistics Significance of differences in survival was calculat-ed using Log-rank (Mantel-Cox) Test and differences in tumorsize were evaluated by two-way ANOVA. Significance indifferences in immunohistochemistry, extracellular adenosinelevels, RNA, and flow cytometry was evaluated by theStudent’s T test.

Results

Our assays were designed to test whether breathing supple-mental oxygen (60 %O2) may inhibit the tumor hypoxia-HIF-1α-driven biochemical pathways, specifically the tumor-protecting CD39/CD73-dependent accumulation of extracel-lular adenosine ([Ado]high) in the TME. Sixty percent oxygenwas selected since it has been used clinically with humanpatients for other indications as well as in our previous studiesof lung inflammation in mice [15].

As expected, Fig. 1a demonstrates that continuous breath-ing of 60 % oxygen reduced intratumoral levels of hypoxia inmice. This was reflected in the significant reduction in immu-nostaining by the molecular hypoxia marker Hypoxyprobe,which identifies deeply hypoxic areas with oxygen tensionless than 1 % [16]. Figure S2 demonstrates that breathing60 % oxygen was also capable of decreasing the pathophys-iological hypoxia not only in tumors but also in inflamedtissue microenvironments. In parallel studies of mice withacute sepsis induced by cecal ligation and puncture (seesupplementary methods), breathing 60 % oxygen reducedhypoxia in surrounding tissues (Fig. S2).

Results of the studies shown in Fig. 1 confirmed thatbreathing 60 % oxygen increased the local intratumoral oxy-gen tension by decreasing the number of deeply hypoxic areas(0 to 1 % oxygen) in the TME. It is likely that 60 % oxygenalso oxygenated less hypoxic areas, but the use ofHypoxyprobe is limited by its inability to detect changes inhypoxia between 1 and 3 % oxygen tension [16].

In order to investigate the oxygenation-associated changesin the TME in more detail, we performed comparative wholecell proteomics analyses of the TME in mice breathing 21 %(ambient) or 60 % oxygen (Fig. 1b, c). The expectation wasthat breathing 60 % oxygen may lead to the degradation ofHIF-1α by increasing the oxygen tension in the TME above 1to 3 %, which is required for the stabilization of HIF-1α. Thiswas predicted to affect many proteins downstream of HIF-1α,including HIF-1α itself (Fig. 1e, left panel).

The results of the proteomics GSEA provide informationon the enriched functions/pathways in a differential proteinlist. Data in Fig. 1b–d and Fig. S3 revealed that breathing60 % oxygen reduced the protein expression of known HIF-1α targets and resulted in the downregulation of hypoxia- andHIF-1α-related gene sets. The list of HIF-1α-related proteins

J Mol Med

that were significantly downregulated are shown in Fig. 1c.These data confirm and extend to the protein level the previ-ous observations of other scientists [17–19] who have report-ed the oxygen-dependent manner of the expression of thesegene sets on the mRNA level. The proteins listed in Fig. 1c areknown to play important roles in many biological processesregulated by hypoxia (e.g., carbohydrate catabolic and meta-bolic processes, gluconeogenesis pathway, etc).

The observations and confirmations of expected changesin hypoxia-HIF-1α-dependent events due to oxygenation ofthe TME from an unbiased proteomics screen also providedan important internal control. Indeed, since the parallelanalyses of the TME confirmed that hyperoxic breathingdid reduce the intratumoral levels of HIF-1α (Fig. 1e, leftpanel), it provided more confidence in the interpretations ofsubsequent findings of the effects of hyperoxia on theexpression of other proteins, which were not yet identifiedas governed by hypoxia-HIF-1α signaling.

In order to further clarify differences in the proteome of thehypoxic versus normoxic TME and to identify proteins thatmay serve as molecular markers for normoxic TMEs, SDSPAGE separated proteins from different areas of tumors wereevaluated by HIF-1α immunoblotting prior to proteomics

analysis. In these experiments, tumor sections representinghypoxic regions (HIF-1αhigh) from mice breathing 21 % O2

were compared to normoxic regions of the TME in micebreathing 60 % oxygen (Fig. S4).

Interestingly, proteomics screen of the TME identified fourand a half LIM domains-1 (FHL-1), an inhibitor of HIF-1αsignaling (19), to be dramatically upregulated in the normoxicregions of tumors frommice breathing 60% oxygen (Fig. 1d).The strong increase in FHL-1 protein levels observed inproteomics studies was subsequently confirmed by westernblot analysis (Fig. 1e, right panel). In the proteomics analysis,FHL-1 was upregulated significantly higher than any othergene product in the TME of mice breathing 60 % oxygen(Fig. 1d), suggesting that FHL-1 could serve as a marker formore oxygenated TMEs. These observations have also iden-tified FHL-1 as a candidate molecule, which might mediatethe effects of 60 % oxygen by preventing association of HIF-1α with the critical co-activator p300/CBP [20].

Interestingly, the proteomics screen of the TME also dem-onstrated the 60% oxygen-mediated increase of other HIF-1αregulating proteins such as factor inhibiting HIF-1 (FIH-1) [1]as well as von Hippel-Lindau (VHL) [21] (Fig. 1d). Takentogether, these findings are in agreement with the biochemical

Fig. 2 Hyperoxic breathing inhibits both upstream and downstreamstages of the hypoxia-HIF-1α-CD39/CD73-A2 Adenosine receptortumor-protecting pathway. a Levels of extracellular adenosine in theTME following 60 % oxygen breathing. Microdialysis probes wereplaced in MCA205 intradermal tumors. After initial probe of the TME,mice were placed in 21 or 60 % oxygen for 3 h and new probes wereplaced in the same tumor area. Adenosine levels were measured byreversed-phase liquid chromatography-tandem mass spectrometry usinga triple quadrupole mass spectrometer with 13C

10-adenosine as an internalstandard (P=0.01) as in (SI, Ref. 7-9). b RT-qPCR analysis of CD73 and

CD39 in B16 s.c. tumors from mice breathing 21 or 60 % O2 for 72 h(P<0.05, n=3). c Expression of CD73 on the surface of TME-infiltratingT cells. Lymphocytes were isolated from the TME of mice bearingMCA205 pulmonary tumors breathing 21 and 60 % oxygen and theexpression of CD73 was measured by flow cytometry (CD4, P<0.02,n=4; CD8, P<0.02, n=4). d Breathing 60 % oxygen downregulatesimmunosuppressive A2AR, A2BR, and COX-2. Mice bearingestablished MCA205 pulmonary tumors were treated with 21 or 60 %oxygen for 72 h and total RNAwas extracted and subjected to RT-qPCRanalysis (P<0.05, n=3)

J Mol Med

assumptions that breathing 60 % oxygen reverses tumor hyp-oxia and HIF-1α regulated pathways in the TME.

It was important to test whether the chaotic tissue hypoxiaand vasculature in the TME would be affected by 60 %oxygen to such an extent that it would be accompanied by adecrease in the levels of tumor-protecting extracellular aden-osine. To this end, direct measurements of extracellular aden-osine in the TME were performed using equilibrium microdi-alysis probes placed intratumorally in a growing tumor.Following an initial microdialysis probe collection, mice wereplaced in 21 or 60% oxygen for 3 h and re-probed for levels ofadenosine in the same tumor area. Subsequent quantitativedetermination by high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS) revealed that tumorsfrom mice breathing 60 % oxygen had significantly reducedlevels of extracellular adenosine compared to those returned toambient oxygen (Fig. 2a). Importantly, tumor areas with ini-tially high levels of extracellular adenosine were significantlyreduced by supplemental oxygenation (Fig. S5).

It was also observed that intratumoral levels of extracellularadenosine were significantly higher than in normal adjacentdermal tissue, confirming the TME is rich in extracellularadenosine. Moreover, the high levels of adenosine are distrib-uted chaotically within different locations of the TME, whichmight be expected in a chaotically grown tumor (Fig. S5).

Further analysis of the TME demonstrated that breathing60 % oxygen was capable of reducing the expression of theadenosine-generating ectoenzymes CD39 and CD73(Fig. 2b). This provided a possible explanation for the de-crease in extracellular adenosine levels in the TME aftersupplemental oxygenation. The downregulation of CD73 byoxygenation in the TME was also observed during analysis oftumor-infiltrating CD8+ and CD4+ T cells (Fig. 2c).

Interestingly, breathing 60 % oxygen also resulted in thedownregulation of tumor-protecting [4] adenosine receptorsA2AR, A2BR, and even of key enzymes of other anti-inflammatory pathways, e.g., prostaglandin E synthase-2(COX-2) (Fig. 2d). In support of these in vivo TME studies,COX-2 has been shown to be regulated by hypoxia and A2AR-mediated cAMP induction in vitro [22, 23]. Since there is noA2A/A2B receptor reserve on T cells [24], the lower expres-sion of A2AR/A2BR translates directly into reduced produc-tion of intracellular cyclic AMP. These data indicate that oxy-genation decreases all stages of the hypoxia-driven andCD39/CD73-adenosine high-A2AR/A2BR-cAMP pathway.

The reduction in hypoxia and destabilization of HIF-1α inmice breathing 60 % oxygen also affected other HIF-1αdownstream targets, particularly those involved in the angio-genesis pathways. Figure 3a shows that hyperoxic breathingsignificantly reduced the levels of VEGF in the TME. Thiswas also confirmed by RT-qPCR analysis of tumor tissue(Fig. S6). We then hypothesized that the reduction in VEGFlevels would alter neovascularization of tumors in these mice.Indeed, immunohistochemical analyses of intratumoralPECAM-1 (CD31) demonstrated a slight but statistically sig-nificant reduction in the expression of the angiogenic markerin mice breathing 60 % oxygen (Fig. 3b, c). In parallel assays,no change was detected in the expression of the VEGF recep-tor on tumor cells (Fig. 3b).

The ability of supplemental oxygenation to favorablychange the TME by potentially weakening tumor escapemechanisms was confirmed in assays of several differenttumor models and anatomical locations. The anti-tumor ef-fects of oxygenation were reflected in the increased expres-sion of antigen-presenting MHC class I molecules on thesurface of tumors in mice breathing 60 % oxygen (Fig. 4a),

Fig. 3 Hyperoxia affects HIF-1αdownstream targets in angiogen-esis pathways. a Immunoblotanalysis of VEGF in the TME.Mice with MCA205 pulmonarytumors were exposed to 21 or60 % oxygen for 72 h; shown arerepresentative samples. b, c Im-munohistochemical staining ofintratumoral CD31 in micebreathing 21 % (left panel) and60 % oxygen (right panel) (×20magnification, P<0.01,n=3). Mice with establishedMCA205 pulmonary tumors wereplaced in either 21 or 60 % oxy-gen for 72 h. Also shown is theeffect of hyperoxic breathing onthe levels of the VEGF Receptor(CD309) on tumor cells

J Mol Med

suggesting that oxygenation may increase the susceptibility torecognition and lethal hit delivery by CTL. Strong support forthis interpretation is provided by observations that tumor cellsidentified to be more hypoxic (HPhi) in vivo had lower levelsMHC-I class I expression (Fig. 4b).

These data support the view that hypoxia within theTME may promote tumor evasion by making the tumorcells less recognizable. Supplemental oxygenation, there-fore, may prevent tumor evasion and facilitate the rec-ognition of tumors by T cells, weakening tumor protec-tion. Since antigen-presenting MHC class I molecules areessential for the priming of anti-tumor T cells, we hy-pothesized that tumor cells exposed to 60 % oxygenwould be more susceptible to recognition and lysis bytumor-reactive T cells. Figure 4c demonstrates that tumorcells with hyperoxia-induced increase in MHC class I

expression were significantly more vulnerable to lethalhit delivery by cytotoxic T lymphocytes.

To examine whether the improved tumor destruction dem-onstrated in in vitro assays would translate to in vivo tumorregression, we treated tumor-bearing mice with 60 % oxygen.Figure 4d shows that breathing 60 % oxygen results in signif-icant tumor regression in mice bearing established subcutane-ous B16 melanoma tumors (H2Kb transfected B16, CL8-1line). An increase in the MHC class I expression on the poorlyimmunogenic parent strain (Fig. S7) may also have contribut-ed to the observed tumor regression of B16 tumors followinghyperoxic breathing (Fig. 4e). Moreover, breathing 60 % ox-ygen resulted in the improved survival of B16 melanomatumor-bearing mice (Fig. 4f). These data confirm the originalassumption that oxygenation converts the TME from tumor-protecting to tumor-destructing.

Fig. 4 Hyperoxic breathing increases tumor cell recognition and killingby anti-tumor T cells, enhances tumor regression, and improves survivalof tumor-bearing mice. a Expression of antigen-presenting MHC class Imolecules on the surface of MCA205 fibrosarcoma and B16 melanomatumors (H2Kb transfected B16, CL8-1 line). Mean fluorescent intensityof MHC-I expression was measured by flow cytometry in mice breathing60 or 21 % oxygen (MCA205, P<0.04, n=3; CL8-1, P<0.04, n=4). bTumor cells exposed to higher levels of hypoxia had lower surfaceexpression of MHC class I (P<0.05, n=4). c Hyperoxia increases lethalhit delivery by cytotoxic T lymphocytes. MCA205 cells cultured in 21 or60 % oxygen for 4 days in vitro demonstrated an increase in the expres-sion of MHC-I as measured by flow cytometry (inset, P<0.01). The cellswere 51Cr-labelled and used in a cytotoxicity assay with culture-activatedT cells from MCA205 tumor-draining lymph nodes at different effector/

target (E/T) ratios. After 4 h, radioactivity released from lysed target cellswas counted on a γ-counter (*P<0.01). d Tumor regression of B16(H2Kb transfected B16, CL8-1 line) tumor-bearing mice following60 % oxygen breathing. After s.c. tumor establishment identified by thefirst appearance of palpable tumors, mice were placed in either 21 or 60%oxygen and tumor size was measured in two dimensions (*P<0.0.5,**P<0.01, n=5 mice). e Tumor regression of B16 melamona tumorsfollowing 60 % oxygen breathing. Mice bearing established B16 s.c.melanomas were placed in either 21 or 60 % oxygen and tumor sizewas measured in two dimensions (P<0.0001, n=10). f Prolonged surviv-al in mice with established B16 tumors breathing 60 % oxygen. Micebearing established B16 s.c. melanomas were placed in either 21 or 60 %oxygen. Tumor-bearing mice were euthanized when tumor reached 2 cmin one direction (P<0.05, n=10)

J Mol Med

Discussion

It has long been recognized that tumors are hypoxic, tumorhypoxia is a poor prognosis factor, and that established tumorsare protected by hypoxia and HIF-1α-mediated biochemicalpathways [1]. HIF-1 and HIF-2 as well as VHL may havedifferent roles in carcinogenesis [1, 21]. It was also demon-strated that hypoxia-driven and CD39/CD73-A2AR/A2BR-adenosinergic immunosuppression represents a major tumor-protecting mechanism (Fig. 5) [4, 12, 13, 25, 26].

However, while the potential benefits of anti-TME hypoxiatreatments and HIF-1α inhibitors are well-recognized [1, 6],there is still an acute medical need in such treatments in theimmunotherapy of cancer.

This study demonstrates the previously unappreciatedproperty of oxygen to predictably and therapeutically recon-dition the TME by reducing hypoxia and levels of extracellu-lar adenosine. Supplemental oxygenation converts the hypox-ic and extracellular adenosine-rich TME into a less hypoxicand extracellular adenosine-poor TME. This, in turn, mayresult in less favorable conditions for cancerous cells, whilesimultaneously creating less hostile conditions for the anti-tumor immunity. Thus, by altering the hypoxia-adenosinergiclandscape of the TME, hyperoxic breathing may create a moreimmunopermissive TME, subsequently increasing the vulner-ability of tumors to tumor-reactive immune cells (Fig. 5).

These data offer molecular justification to use supplementaloxygen in therapies of cancer in a way that has not been done,

and in combination with anti-tumor immunity. Indeed, ourmotivation to test 60 % oxygen is conceptually different fromother previous medical uses of oxygen. Here, we used oxygen-ation in order to weaken the immunosuppression in TME, whileprevious motivations to use supplemental oxygen were guidedby the hypothesis that high oxygen would enhance reactiveoxygen species (ROS) formation in the radiotherapy of cancer[27]; or that 70–100 % oxygen alone may be directly toxic totumors [28]. Therefore, the settings of tumor growth in preclin-ical studies were not designed to allow for the development ofthe anti-tumor immune response following the oxygen-mediated weakening of immunosuppression in the TME.These data provide mechanistic reasoning for breathing 60 %oxygen as a treatment to weaken all upstream and downstreamstages of the tumor-protecting hypoxia/HIF-1α-CD39/CD73-[Adenosine]High–A2AR/A2BR pathway (Fig. 5).

It was found that supplemental oxygen (1) decreases hypoxiain inflamed tissues or in tumors (Fig. 1, S2); (2) decreases levelsof extracellular adenosine and adenosine-generating enzymes inthe TME (Fig. 2); (3) reduces neovascularization (Fig. 3); and(4) increases the expression of tumor antigen-expressing mole-cules, thereby improving the recognition of tumors by tumor-reactive immune cells (Fig. 4). Importantly, the changes in theTME mediated by hyperoxia resulted in significantly delayedtumor growth in vivo and longer survival (Fig. 4).

It remains a likely possibility that the reduction in the levelsof extracellular adenosine following oxygenation could be notonly because of the decreased generation of adenosine byectoenzymes CD39/CD73 but also due to enhanced degrada-tion of extracellular adenosine by adenosine degrading en-zymes [29]. The experimental testing of the effects of oxygenon different mechanisms of adenosine metabolism and degra-dation may be complicated. Due to the critical necessity in theregulation of the levels of extracellular adenosine, there aremany different enzymes and pathways involved in adenosinemetabolism; which becomes increasingly complex when con-sidering the shuttling of adenosine through many adenosinetransporters and the very short half-life of extracellular aden-osine in vivo [30].

It is important to point out the possible differential roles ofsupplemental oxygenation on normal, nontransformed cellsduring tumorigenesis and on hypoxia and HIF-1α in the TMEduring the growth and rejection of established tumors. Ourdata suggest that the destruction of tumors due the lower levelsof immunosuppressive and tumor-protecting extracellularadenosine most likely takes place much earlier than thetumor-supporting effects of increased oxygenation of tumorsdue to, e.g., HIF-1α-dependent increase in EPO, VEGF andneovascularization [1, 7, 19].

As suggested by the data shown here, systemic oxygena-tion may indeed provide better oxygenation through existingblood vessels. While the increased oxygenation could poten-tially benefit some of the tumor cells, it simultaneously may

Fig. 5 Hyperoxic breathing may prevent tumor hypoxia-dependent ac-cumulation of tumor-protecting extracellular adenosine ([Ado]high) intumor microenvironments. Left panel: Tumor hypoxia/HIF-1α-drivenand CD39/CD73 ectoenzyme-generated extracellular [Adenosine]high

protects tumors by inhibiting T cells and NK cells through activation ofintracellular cAMP-elevating A2A or A2B adenosine receptors (A2AR/-A2BR). TME hypoxia may also protect tumors through the stabilizationof HIF-1α. At tissue oxygen tension above ~3 %, oxygen-sensing prolylhydroxylases (PHD) target HIF-1α for degradation, thereby preventing Tcell inhibition. Right panel: Hyperoxic breathing 60 % [Oxygen]high isexpected to weaken the first upstream stage (TME hypoxia) of hypoxia-HIF-1α signaling and thereby inhibit the formation of extracellular[Adenosine]high. This, in turn, may weaken tumor protection and enableT- and NK cells to reject tumors

J Mol Med

unleash the anti-tumor cytotoxic cells from hypoxia andadenosine-mediated inhibition. Thus, the overall outcomewould be the death of even the more oxygenated, “healthier”tumor cells. It is also possible that by reducing HIF-1α sig-naling, breathing 60 % oxygen may contribute to the normal-ization of tumor vasculature, further contributing to tumoroxygenation [31]. Future studies will further characterize theeffects of breathing 60 % oxygen on the timescale of neovas-cularization as well as on already established blood vessels.

Since supplemental oxygen is already a widely used andsafe medical treatment, these data may have accelerated clin-ical implications. We envision the combination of supplemen-tal oxygen with existing immunotherapies of cancer such ascancer vaccines, adoptive cell transfer, and/or blockade ofimmunological negative regulators (e.g., CTLA-4/PD-1).

However, as an important exclusion criteria, 60 % oxygenshould not be given to cancer patients during episodes ofsimultaneous acute inflammation. This includes episodes ofacute lung and liver inflammation, among others.Oxygenation may not only enable stronger anti-tumor activi-ties by de-inhibited tumor-reactive immune cells but alsoaccomplish stronger inflammatory damage of normal tissuesby de-inhibited myeloid cells or T cells activated by otherantigens. Earlier, we attracted attention to such possible ca-veats in using antagonists of A2AR, which act downstream ofthe hypoxia→HIF-1α stages [32].

Indeed, the delivery of 60 % oxygen should be consideredwith an understanding that while long-term treatment with 60%oxygen has been proven to be safe, our previous reports haveshown that 60 % oxygen exacerbates acute inflammatory lunginjury [15] by inhibiting the hypoxia-HIF-1α-[Adenosine]High–A2AR pathway. Thus, the weakening of this physiologicaltissue-protecting mechanism may increase inflammatory dam-age not only in cancerous tissues (Fig. 2) but also in normaltissues of vital organs (Fig. S2). This excessive collateral tissuedamage was originally shown to be due to overactive neutro-phils [15]. Similar effects on inflamed lungs were observed byothers who implicated macrophages [33] and pulmonary invari-ant natural killer T (iNKT) cells in the effects of hyperoxia [34].

It will be interesting in future studies to compare inparallel the effects of oxygenation on mice with increasedtumor burden with mice with acute lung inflammation.Studies in mice with high numbers of lung tumors maydirectly clarify whether tumor-reactive T cells de-inhibitedby supplemental oxygenation will not only inflict damageto cancerous tissue but also collateral damage to normaltissues. The expectation is that in contrast to myeloid cellsand due to their localized mechanism of lethal hit delivery[35], tumor-reactive T cells de-inhibited by 60 % oxygenwill still be selective for tumor antigens. Thus, the tumor-reactive T cells will likely cause much less collateral dam-age when de-inhibited by oxygen compared to myeloidcells during episodes of acute inflammation [15, 33, 34].

While considering the issues of patient compliance in using60 % oxygen-delivering masks, it will be appealing to explorethe re-purposing of hyperbaric chambers or hypoxic tents whichare used by athletes, as a more comfortable option for oxygen-ation of cancer patients. Future investigations will determinewhether other concentrations of oxygen, including hyperbaricoxygen, are capable of providing therapeutic benefit.

Taken together, these data highlight the potentially powerfulbenefits of supplemental oxygen as a co-adjuvant in combinationwith other existing cancer treatments in order to weaken thetumor hypoxia-HIF-1α-driven and CD39/CD73→A2AR/A2BR-mediated tumor protective pathway in the hypoxic TME.

Acknowledgments This work was supported by the funding fromNortheastern University and NIH grants (PI: M.S.) R01 CA-111985;R01GM-097320; R01CA-112561; R21AT002788.We thankDr. RichardMarsh, an expert in the field of energetics, kinematics, and kinetics forassistance with monitoring gas compositions in hyperoxia units.

Conflict of interest The authors declare they have no competing inter-ests as defined by Molecular Medicine or other interests that might beperceived to influence the results and discussion reported in this paper.

References

1. Semenza GL (2012) Hypoxia-inducible factors in physiology andmedicine. Cell 148:399–408

2. Synnestvedt K, Furuta GT, Comerford KM, Louis N, Karhausen J,Eltzschig HK, HansenKR, Thompson LF, Colgan SP (2002) Ecto-5'-nucleotidase (CD73) regulation by hypoxia-inducible factor-1 medi-ates permeability changes in intestinal epithelia. J Clin Invest 110:993–1002

3. Eltzschig HK, Sitkovsky MV, Robson SC (2013) Purinergic signal-ing during inflammation. N Engl J Med 368:1260

4. Ohta A, Gorelik E, Prasad SJ, Ronchese F, Lukashev D, Wong MK,Huang X, Caldwell S, Liu K, Smith P et al (2006) A2A adenosinereceptor protects tumors from antitumor Tcells. Proc Natl Acad Sci US A 103(35):13132–13137

5. Sitkovsky MV, Kjaergaard J, Lukashev D, Ohta A (2008) Hypoxia-adenosinergic immunosuppression: tumor protection by T regulatorycells and cancerous tissue hypoxia. Clin Cancer Res 14:5947–5952

6. Dewhirst MW, Cao Y, Moeller B (2008) Cycling hypoxia and freeradicals regulate angiogenesis and radiotherapy response. Nat RevCancer 8:425–437

7. ZhangH, Qian DZ, Tan YS, Lee K, Gao P, Ren YR, Rey S, HammersH, Chang D, Pili R et al (2008) Digoxin and other cardiac glycosidesinhibit HIF-1α synthesis and block tumor growth. Proc Natl Acad SciU S A 105:19579–19586

8. Lee K, Qian DZ, Rey S, Wei H, Liu JO, Semenza GL (2009)Anthracycline chemotherapy inhibits HIF-1 transcriptional activityand tumor-induced mobilization of circulating angiogenic cells. ProcNatl Acad Sci U S A 106:2353–2358

9. Feng L, Sun X, Csizmadia E, Han L, Bian S, Murakami T, Wang X,Robson SC, Wu Y (2011) Vascular CD39/ENTPD1 directly pro-motes tumor cell growth by scavenging extracellular adenosine tri-phosphate. Neoplasia 13:206–216

10. Stagg J, Divisekera U, McLaughlin N, Sharkey J, Pommey S,Denoyer D, Dwyer KM, Smyth MJ (2010) Anti-CD73 antibodytherapy inhibits breast tumor growth and metastasis. Proc NatlAcad Sci U S A 107:1547–1552

J Mol Med

11. Lukashev D, Ohta A, Sitkovsky M (2007) Hypoxia-dependent anti-inflammatory pathways in protection of cancerous tissues. CancerMetastasis Rev 26:273–279

12. Waickman AT, Alme A, Senaldi L, Zarek PE, Horton M,Powell JD (2012) Enhancement of tumor immunotherapy bydeletion of the A2A adenosine receptor. Cancer ImmunolImmunother 61:917–926

13. Jin D, Fan J,Wang L, Thompson LF, LiuA, Daniel BJ, Shin T, CurielTJ, Zhang B (2010) CD73 on tumor cells impairs antitumor T-cellresponses: a novel mechanism of tumor-induced immune suppres-sion. Cancer Res 70:2245–2255

14. Loi S, Pommey S, Haibe-Kains B, Beavis PA, Darcy PK, Smyth MJ,Stagg J (2013) CD73 promotes anthracycline resistance and poorprognosis in triple negative breast cancer. Proc Natl Acad Sci U S A110:11091–11096

15. Thiel M, Chouker A, Ohta A, Jackson E, Caldwell C, Smith P,Lukashev D, Bittmann I, SitkovskyMV (2005) Oxygenation inhibitsthe physiological tissue-protecting mechanism and thereby exacer-bates acute inflammatory lung injury. PLoS Biol 3:e174

16. Raleigh JA, Dewhirst MW, Thrall DE (1996) Measuring tumorhypoxia. Semin Radiat Oncol 6:37–45

17. Manalo DJ, Rowan A, Lavoie T, Natarajan L, Kelly BD, Ye SQ,Garcia JG, Semenza GL (2005) Transcriptional regulation of vascularendothelial cell responses to hypoxia by HIF-1. Blood 105:659–669

18. Elvidge GP, Glenny L, Appelhoff RJ, Ratcliffe PJ, Ragoussis J,Gleadle JM (2006) Concordant regulation of gene expression byhypoxia and 2-oxoglutarate-dependent dioxygenase inhibition: therole of HIF-1α, HIF-2α, and other pathways. J Biol Chem 281:15215–15226

19. Semenza GL (2014) Hypoxia-inducible factor 1 and cardiovasculardisease. Annu Rev Physiol 76:39–56

20. Hubbi ME, Gilkes DM, Baek JH, Semenza GL (2012) Four-and-a-half LIM domain proteins inhibit transactivation by hypoxia-inducible factor 1. J Biol Chem 287:6139–6149

21. Shen C, Kaelin WG Jr (2013) The VHL/HIF axis in clear cell renalcarcinoma. Semin Cancer Biol 23:18–25

22. Chida M, Voelkel NF (1996) Effects of acute and chronic hypoxia onrat lung cyclooxygenase. Am J Physiol 270:L872–878

23. Klein T, Shephard P, Kleinert H, Komhoff M (2007) Regulation ofcyclooxygenase-2 expression by cyclic AMP. Biochim Biophys Acta1773:1605–1618

24. Armstrong JM, Chen JF, Schwarzschild MA, Apasov S, Smith PT,Caldwell C, Chen P, Figler H, Sullivan G, Fink S et al (2001) Genedose effect reveals no Gs-coupled A2A adenosine receptor reserve inmurine T-lymphocytes: studies of cells from A2A-receptor-gene-deficient mice. Biochem J 354:123–130

25. Ohta A, Sitkovsky M (2001) Role of G-protein-coupled adenosinereceptors in downregulation of inflammation and protection fromtissue damage. Nature 414:916–920

26. Zhang B (2010) CD73: a novel target for cancer immunotherapy.Cancer Res 70:6407–6411

27. Busk M, Horsman MR (2013) Relevance of hypoxia in radiationoncology: pathophysiology, tumor biology and implications for treat-ment. Q J Nucl Med Mol Imaging 57:219–234

28. Margaretten NC, Witschi H (1988) Effects of hyperoxia on growthcharacteristics of metastatic murine tumors in the lung. Cancer Res48:2779–2783

29. Ren J, Mi Z, Stewart NA, Jackson EK (2009) Identification andquantification of 2',3'-cAMP release by the kidney. J PharmacolExp Ther 328:855–865

30. Longhi MS, Robson SC, Bernstein SH, Serra S, Deaglio S (2013)Biological functions of ecto-enzymes in regulating extracellularadenosine levels in neoplastic and inflammatory disease states. JMol Med (Berl) 91:165–172

31. Huang Y, Goel S, Duda DG, Fukumura D, Jain RK (2013) Vascularnormalization as an emerging strategy to enhance cancer immuno-therapy. Cancer Res 73:2943–2948

32. Ohta A, Sitkovsky M (2006) Caveats in promising therapeutictargeting of the anti-inflammatory A2 adenosine receptors: the notesof caution. Discovery, Nature Reviews Drug

33. Aggarwal NR, D'Alessio FR, Eto Y, Chau E, Avalos C, WaickmanAT, Garibaldi BT, Mock JR, Files DC, Sidhaye V et al (2013)Macrophage A2A adenosinergic receptor modulates oxygen-induced augmentation of murine lung injury. Am J Respir Cell MolBiol 48:635–646

34. Nowak-Machen M, Schmelzle M, Hanidziar D, Junger W, Exley M,Otterbein L, Wu Y, Csizmadia E, Doherty G, Sitkovsky M et al(2013) Pulmonary natural killer T cells play an essential role inmediating hyperoxic acute lung injury. Am J Respir Cell Mol Biol48:601–609

35. Sitkovsky MV, Paul WE (1988) Immunology. Global or directedexocytosis? Nature 332:306–307

J Mol Med