A_2A receptor signaling promotes peripheral tolerance by inducing T-cell anergy and the generation of adaptive regulatory T cells

Under normal physiologic conditions, the level of adenosine in the tissue microenvironment is relatively low.¹  At the height of inflammation, the destruction of host tissue by a vigorous immune response combined with damaged microcirculation and hypoxia leads to increases in extracellular adenosine. The presence of adenosine in the microenvironment prevents further tissue destruction by inhibiting inflammation.^2,3  The importance of the adenosine A_2A receptor (A_2AR) in mediating this negative feedback loop has been dramatically demonstrated in A_2AR null mice.⁴  These knockout mice are unable to control inflammation, which leads to extensive, often fatal, tissue destruction. For example, when immune-mediated liver inflammation was induced in mice by injecting Con A, the A_2AR null mice died because of massive immune–mediated liver destruction.

The inability of the thymus to eliminate all self-reactive T cells necessitates peripheral mechanisms of T cell tolerance to prevent tissue damage from autoreactive T cells. The role of clonal deletion and T-cell anergy in maintaining peripheral tolerance has been demonstrated in several in vivo models.^5,6  More recently, the roles of Foxp3⁺ and LAG-3⁺ regulatory T cells in peripheral tolerance has been better defined.^7,8  Likewise, cell-surface receptors such as CTLA-4 and PD-1 and cytokines such as interleukin (IL)-10 and transforming growth factor (TGF)-β are thought not only to inhibit T-cell activation but also to promote peripheral T-cell tolerance.^9,,–12  In this report, we identify extracellular adenosine as a novel inducer of peripheral T-cell tolerance. We propose that tissue-derived adenosine acts through the A_2AR to generate peripheral T-cell tolerance by inducing T-cell anergy and adaptive regulatory T cells.

Cyclosporine A (Calbiochem, Cambridge, MA) was dissolved in ethanol and used at 1 μM. CGS-21680 (Sigma, St Louis, MO) was dissolved in phosphate-buffered saline (PBS) and used at indicated concentrations. Anti-CD3 (2C11, BD PharMingen, San Diego, CA) was diluted in PBS and used at 1 μg/mL, as indicated. Soluble anti-CD28 (a kind gift from Dr J. Allison, University of California at Berkeley, Berkeley, CA) was used at a 1/1000 dilution. Hemagglutinin (HA) class II is an I-E^d specific peptide (SFERFEIFPKE), which was manufactured by the Johns Hopkins School of Medicine Oncology Department Peptide Synthesis Facility. Antiphospho-ERK (Santa Cruz Biotechnology, Santa Cruz, CA) was used at 1:400 in 5% nonfat dry milk (NFDM)/TBS-Tween (0.1%). Anti-p42/p44 MAPK (Cell Signaling Technology, Danvers, MA) was used at 1:500 in 5% BSA/TBS-Tween (0.1%). Anti–jun B (Santa Cruz) and anti-actin (Sigma) were both used at 1:1000 in 5% NFDM/TBS-Tween (0.1%) mAbs used for staining: biotinylated anticlonotypic 6.5 T-cell receptor (TCR) (generously provided by H. Von Boehmer); avidin-PE (Caltag, Burlingame, CA); FITC-conjugated anti-CD4 (RM4-4); Cy-chrome-conjugated anti-CD4 (RM4-5); PE-conjugated antiThy-1.1 (OX-7); FITC-conjugated anti-CD44 (1 M7); APC-conjugated anti-interferon (IFN)-γ (XMG1.2); PE-conjugated anti-IL-17 (TC11–18H10); biotinylated anti-Thy1.2 (53-2.1) (all purchased from BD PharMingen). Anti–IFN-γ and anti–IL-4 BD PharMingen) were used at 10 μg/mL. TGF-β (Sigma) was used at 5 ng/mL. IL-6 (Peprotech, Rocky Hill, NJ), was used at 20 ng/mL. FACSCalibur was used for flow cytometry event collection, and events were analyzed using FlowJo software (Ashland, OR). Sorting was done with FACSAria (BD Biosciences, San Jose, CA).

A.E7 is a clonotypic CD4⁺, Th1 T-cell clone specific for pigeon cytochrome c (PCC) and is maintained as previously described.¹³  A.E7s were rendered anergic with plate-bound anti-CD3 (1 μg/mL) or by incubation with APC, PCC (10 μM), and CGS (1 μM).

Splenocytes and lymphocytes from A_2AR wt or null 6.5⁺ mice (see “Transgenic mice”) were harvested and enriched for CD4⁺ T cells via negative selection with the CD4⁺ T cell isolation kit and LS MACs column (all from Miltenyi Biotech, Auburn, CA). The T cells were then cultured in the presence or absence of 10 μg/mL HA and 1 μM CGS. The T cells were rechallenged for 3 to 4 hours with 100 μg/mL HA, GolgiStop (BD PharMingen), and irradiated APCs.

Total cAMP of naive or previously activated cells were assayed with the Biotrak EIA system (Amersham Biosciences, Amersham, United Kingdom).

In vitro microarray was performed as previously described.¹⁴  In vivo microarray was performed as previously described.⁸ 

The C3HA expressing transgenic (recipient) mice express HA under the rat C3 promoter and has been previously described.¹⁵  The HA-specific TCR-transgenic mouse line 6.5 (donor mice) has been previously described.⁸  A_2AR^−/− mice were bred to this 6.5 mouse line. LAG-3 knockout mice on a C57/B6 were a generous gift of Dr Dario Vignali. The LAG-3 KO genotype was bred onto a B10.D2 6.5⁺ TCR background. All experiments involving the mice were performed in accordance with protocols approved by the Animal Care and Use Committee of the Johns Hopkins University School of Medicine.

5C.C7 TCR-transgenic Rag2^−/− mice on a B10.A background were purchased from Taconic (Petersburgh, NY). T cells from 5C.C7 mice are specific for pigeon cytochrome c.

Clonotypic CD4⁺ T cells were harvested from 6.5⁺ transgenic mice. The unfractionated population was resuspended to contain 1.2 × 10⁶ 6.5⁺ T cells in 200 μL HBSS for intravenous injection into C3HA mice. Recipient mice were given twice-daily intraperitoneal injections of vehicle (PBS alone) or CGS (2.5 mg/kg) on days 1 to 4 after the transfer. No CGS was administered after day 4.

The numbers of IFN-γ and IL-17 cells were determined by multiplying the number of infiltrating cells recovered from the entire lung by the percentage of donor (Thy1.1⁺) cells and the percentage of clonotypic cells that produce the cytokine.

For phospho-ERK and total ERK Western blots, splenocytes from 6.5⁺ mice were stimulated overnight with 10 μg/mL HA class II peptide then harvested and enriched for CD4⁺ T cells. CD4⁺ purified T cells were then stimulated in 500 μL with soluble anti-CD3 (10 μg/mL) and soluble anti-CD28 (1:20 dilution).

For the junB and actin Western blots, CD4⁺ purified, 6.5⁺ T cells were stimulated overnight in the presence or absence of 1 μM CGS. Nuclear extracts were then probed for junB and actin.

Nuclear extracts from CD4⁺ purified 6.5⁺ T cells that were prepared. The AP-1 probe was: 5′-CGC TTG ATG ACT CAG CCG GAA-3′. The NFκB probe was: 5′-AGT TGA GGG GAC TTT CCC AGG C-3′.

For in vitro studies, 5C.C7 splenocytes were stimulated overnight with 10 μM PCC in the presence or absence of 1 μM CGS. mRNA was isolated with Trizol and cDNA was generated. Reverse-transcribed polymerase chain reaction (RT-PCR) was performed as previously described.¹⁴ 

For in vivo studies, recipient C3HA mice were given 1.2 × 10⁶ 6.5⁺ donor T cells as described above. On day 3, donor T cells were sorted and cDNA generated as above. RT-PCR was performed as previously described.¹⁴ 

LAG-3 primers and probe sets used were: primer 5′-ACATCAACCAGACAGTGGCCA-3′/primer 5′-GCATCCCCTGGTGAAGGTC-3′/probe 5′-6FAM-CCCACTCCCATCCCGGCCC-TAMRA-3′.

FoxP3 primers and probe sets used were: primer 5′-GGC CCT TCT CCA GGA CAG A-3′/primer 5′-GCT GAT CAT GGC TGG GTT GT-3′/probe 5′-6FAM-ACT TCA TGC ATC AGC TCT CCA CTG TGG ATT AT-TAMRA-3′ IL-6 and TGF-β primers and probe sets were purchased from Applied Biosystems (Foster City, CA).

4 × 10⁶ 5C.C7 splenocytes were stimulated as previously described.¹⁶  Briefly, cells were stimulated in media containing TGF-β (5 ng/mL), IL-6 (20 ng/mL), and anti–IL-4 and anti–IFN-γ blocking antibodies (10 μg/mL).

Quantitative data were expressed as mean plus or minus standard deviation and compared using paired Student t tests. Values of P less than .05 were considered significant.

If cyclosporine A (CSA), which inhibits the translocation of NF-AT to the nucleus, is added during priming, the induction of anergy is inhibited (Figure S1A, available on the Blood website; see the Supplemental Materials link at the top of the online article). We established a microarray database of T cells stimulated under conditions that promote or inhibit the induction of anergy and devised a novel statistical algorithm (hypothesis-based analysis of microarray) to analyze these data.^14,17  The adenosine A_2AR was among the NF-AT-dependent genes that showed increased expression on TCR engagement and was inhibited by CSA (Figure S1B). For comparison, the adenosine A₁ receptor expression was not regulated by TCR signaling. The adenosine A_2B receptor displayed a similar, although less robust, expression profile to the A_2AR.

To confirm these microarray results, we assayed the CD4⁺ Th1 T-cell clone, A.E7, for their expression of functional A_2AR. We took advantage of the fact that the A_2AR is a G-protein coupled receptor and directly stimulates adenylyl cyclase. At a given dose of an A_2AR-specific agonist, differences in intracellular cAMP levels indicate the relative abundance of the A_2AR on the cell surface.¹⁸  We stimulated A.E7s with PBS alone (mock), plate-bound anti-CD3 or anti-CD3 + CSA overnight, rested the cells in fresh media for 24 hours, treated the cells with the A_2AR-specific agonist, CGS-21680 (CGS), and measured intracellular cAMP. Consistent with the microarray results, T cells treated with anti-CD3 alone produced more cAMP than mock-stimulated cells, indicating that the A_2AR is more highly expressed on activated T cells than resting ones (Figure 1A). Moreover, the presence of CSA during TCR stimulation prevented the cAMP increase, indicating that there is no up-regulation of the A_2AR when TCR-induced NF-AT activation is inhibited.

Next, we analyzed A_2AR expression in primary T cells from Wt and A_2AR^−/− mice. In the absence of T-cell stimulation, CGS treatment did not increase intracellular cAMP (“Wt Naive”). On activation, however, CGS treatment dramatically increased cAMP production in Wt T cells, again indicating that A_2AR protein expression has increased. Activated T cells from A_2AR^−/− mice did not produce cAMP in response to CGS treatment, which is expected and confirms the specificity of CGS for the A_2AR (Figure 1B). These results directly confirm our findings with the A.E7 T-cell clones and show that in primary T cells the A_2AR is expressed on TCR signaling in an NF-AT–dependent manner. Previous studies have clearly demonstrated the ability of the A_2AR to inhibit T-cell activation both in vitro and in vivo.^19,,,–23  In agreement with these studies, we found that A_2AR engagement during A.E7 stimulation markedly inhibited IL-2 production (Figure 1C). By comparison, the inhibition of IFN-γ by A_2AR stimulation was less robust. This finding was not secondary to CGS-induced cell death, as determined by trypan blue exclusion (data not shown). Furthermore, we failed to see any inhibition of T-cell proliferation by CGS in this dose-response range (Figure 1D). Thus, the A_2AR is induced by signal 1 alone in an NF-AT–dependent fashion and plays a role in inhibiting T-cell function, most specifically IL-2 expression.

In light of these initial findings, we wanted to determine the effect of A_2AR engagement on the induction of T-cell anergy. We examined whether A_2AR engagement during the initial antigen encounter could promote tolerance such that the T cells would be unresponsive on rechallenge in the absence of A_2AR stimulation. To this end, A.E7 T cells were stimulated for 4 days with irradiated APCs (signal 1 + 2) in the presence or absence of PCC (the cognate antigen for A.E7s) and in the presence or absence of CGS. The cells were then harvested, washed extensively, and rechallenged with irradiated APCs and PCC in the absence of CGS. Incubating A.E7s with CGS in the absence of peptide had no consequence on rechallenge with PCC (Figure 2A,C left bars). These results were consistent with the fact that A_2AR expression requires TCR engagement. However, if the T cells were activated by PCC in the presence of CGS, they failed to proliferate or produce IFN-γ on subsequent rechallenge (Figure 2B,C). Exogenous IL-2 can prevent and reverse clonal T-cell anergy.^24,25  Therefore, we explored the effect of exogenous IL-2 on A_2AR stimulation. The addition of exogenous IL-2 during the induction mitigated the effect of A_2AR engagement (Figure 2D,E). When IL-2 was present in the induction cultures, CGS had no effect on proliferation and IFN-γ production on rechallenge. Overall, our data indicate that A_2AR stimulation even in the presence of costimulation can induce a hyporesponsive state on rechallenge that resembles T-cell anergy.

Having determined that A_2AR engagement can promote anergy in A.E7 T-cell clones, we next wanted to examine whether A_2AR signaling can induce tolerance in primary T cells. We isolated CD4⁺ T cells from Wt or A_2AR^−/− 6.5⁺ mice and cultured them with irradiated APCs and HA peptide in the presence or absence of CGS. The T cells were harvested, extensively washed, and rechallenged with HA peptide in the absence of CGS. On rechallenge with HA, Wt T cells produced IFN-γ. In contrast, T cells that were initially incubated with HA peptide and CGS were hyporesponsive on rechallenge. Because no CGS was present during the rechallenge, the decreased cytokine production reflects the induction of anergy. This effect was A_2AR specific as T cells from A_2AR^−/− mice cultured initially in the presence or absence of CGS have equivalent numbers of IFN-γ⁺ T cells (Figure 3A). A_2AR signaling during an initial antigen encounter can induce a state of anergy in primary T cells such that they are hyporesponsive on rechallenge.

Next, we tested the hypothesis that A_2AR signaling might promote anergy by inhibiting TCR-induced AP-1 activation. A series of experiments was performed to determine the effect of A_2AR signaling on the Ras-MAP-kinase pathway. CD4⁺ 6.5⁺ T cells were stimulated with HA overnight to induce expression of the A_2AR. The cells were then harvested and washed before being stimulated with anti-CD3 and anti-CD28 plus or minus CGS. In the absence of CGS, ERK was rapidly phosphorylated. However, in the presence of CGS, this phosphorylation was prevented. Total ERK levels remain constant (Figure 3B). We then looked at the effect of A_2AR signaling on the expression of junB, a component of AP-1. In the absence of stimulation, CD4⁺ 6.5⁺ T cells had low levels of junB expression. On stimulation, junB levels drastically increased. But if the CD4⁺ 6.5⁺ T cells were stimulated in the presence of the A_2AR-agonist, junB levels remained similar to the unstimulated cells (Figure 3C). Finally, EMSA was performed to examine AP-1 activity in the nucleus. In the absence of stimulation, minimal AP-1 binding was observed. On stimulation, there was an increase in AP-1 in the vehicle-treated cells; however, in the presence of A_2AR signaling, AP-1 binding was inhibited (Figure 3D).

Using T-cell clones and TCR transgenic primary T cells, we have shown that A_2AR signaling promotes a state of hyporesponsiveness similar to T-cell anergy, even in the presence of costimulation. We then sought to determine whether endogenous, tissue-derived adenosine in vivo could promote T-cell tolerance. A series of experiments used a previously established model of T-cell tolerance induction and autoimmunity in vivo.^8,15,26,27  C3HA mice express HA as a self-antigen, with especially high HA expression in the lung and prostate. When HA-specific 6.5⁺ T cells are adoptively transferred into C3HA mice, the donor 6.5⁺ T cells are initially activated by HA, which is presented as a “self” antigen. But by day 4, they have been rendered anergic. If a sufficient number of 6.5⁺ T cells are transferred, the initial activation of the clonotypic cells dominates and the recipient mice die of autoimmune-induced pneumonitis. We wanted to determine whether the A_2AR is up-regulated on the HA-specific T cells in vivo. Clonotypic T cells were adoptively transferred into C3HA mice, isolated 4 days later and assessed by microarray analysis for A_2AR expression. Compared with the T cells transferred into nontransgenic mice or mice infected with vaccinia virus expressing HA, the A_2AR was markedly up-regulated on the in vivo tolerized clonotypic T cells (Figure 4A).

Having determined that the A_2AR is up-regulated on the adoptively transferred T cells in vivo, we investigated the role that endogenous adenosine might play in promoting tolerance. Wt and A2aR^−/− clonotypic T cells were adoptively transferred into C3HA mice. As seen in Figure 4B, the dose of Wt 6.5⁺ T cells was insufficient to induce lethal autoimmunity. The clonotypic T cells become tolerant and 100% of the mice survive. On the other hand, transfer of the A_2AR^−/− 6.5⁺ T cells led to 80% death because of autoimmune pneumonitis. Thus, in the absence of A_2AR signaling on the adoptively transferred T cells, a tolerizing interaction (A_2AR engagement) is removed leading to tissue destruction and death. Because only the adoptively transferred cells are A_2AR^−/−, our observations reflect the direct effect of endogenous tissue-derived adenosine on T cells and not an increase in the general inflammatory milieu (for example, the host APCs are Wt and thus are responsive to the tissue-released adenosine).

We then asked whether we could pharmacologically induce tolerance in vivo by administering CGS to hosts that receive A_2AR competent 6.5⁺ T cells. We transferred a lethal dose of T cells followed by a 4-day course of CGS (2.5 mg/kg, intraperitoneally, twice daily). CGS treatment resulted in long-term survival (30+ days) of 76% of the host mice, whereas only 24% of the vehicle-treated hosts survived past day 9 (Figure 5A). We wanted to rule out that the ability of CGS to induce tolerance was simply because of the deletion of the clonotypic cells. To test this, we transferred 6.5⁺ T cells into C3HA mice and harvested the donor 6.5⁺ T cells 2 or 3 days after transfer. CGS did not affect the frequency of donor 6.5⁺ T cells (Figure S2). The relatively equivalent numbers of clonotypic cells is consistent with our in vitro finding that CGS is a poor inhibitor of T-cell proliferation (Figure 1D). It also suggests that deletion is not the mechanism by which A_2AR signaling promotes tolerance in vivo. We next tested the clonotypic T cells from the treated and untreated mice for their function in response to rechallenge with HA antigen in vitro. Donor 6.5⁺ T cells harvested from vehicle- or CGS-treated mice on day 2 showed no differences in proliferation or IFN-γ and IL-2 production (Figure 5B-D gray bars). By day 3, however, donor 6.5⁺ T cells from CGS-treated hosts demonstrated greatly reduced proliferation and IFN-γ and IL-2 production compared with those from vehicle-treated hosts (Figure 5B-D open bars). A_2AR signaling, in vivo, promotes T-cell tolerance, resulting in reduced lung damage and increased survival.

Recently, it has been demonstrated that the lethality in the C3HA model is mediated, in part, by lung-infiltrating IL-17–producing cells (Harris and Drake, oral communication, June 2007). Thus, we tested whether A_2AR stimulation was able to affect the generation of Th17 effector cells in this model. Recipient C3HA mice were given 10 × 10⁶ 6.5⁺ T cells and treated with vehicle or CGS. The lungs were harvested, and intraparenchymal clonotypic T cells were stimulated with HA peptide in vitro. As seen in Figure 5E,F, there was a decrease in the number of not only IFN-γ producing cells but also IL-17–producing lung-infiltrating T cells in the mice treated with CGS. Thus, treatment with CGS resulted in a decrease in the tissue infiltration of Th1 and Th17 effector T cells.

The generation of IL-17–producing cells requires both IL-6 and TGF-β. In light of the anti-inflammatory properties of A_2AR stimulation, we hypothesized that A_2AR stimulation might inhibit the generation of IL-17–producing cells by altering the cytokines found in the inflammatory milieu. We tested this hypothesis using a previously defined method of generating Th17 cells in vitro from naive T cells from 5C.C7 TCR-transgenic Rag2^−/− mice.¹⁶  As seen in Figure 6A, CGS inhibited the expression of IL-6. In the same culture conditions, however, A_2AR activation led to a marked up-regulation of TGF-β (Figure 6B). These data suggest that CGS limits the generation of IL-17–producing cells in vivo, in part, by inhibiting IL-6 production. Indeed, CGS did not inhibit the induction of IL-17⁺ T cells in vitro if exogenous IL-6 was added to the culture conditions (Figure 6C).

Inducible Foxp3⁺ regulatory cells are generated when antigen recognition occurs in the presence of TFG-β.²⁸  Because TCR engagement in the presence of CGS led to an increase in TGF-β expression, we wanted to know if CGS could promote the expression of Foxp3. Indeed, CGS-induced inhibition of IL-6 and enhancement TGF-β was accompanied by an increase in Foxp3 expression (Figure 6D). This finding, in turn, prompted us to examine Foxp3 expression in the clonotypic cells derived from the CGS-treated mice. As seen in Figure 6E, there is increased expression of Foxp3 in the clonotypic T cells isolated 3 days after adoptive transfer from the CGS-treated mice compared with the untreated mice. Thus, under immune activating conditions, A_2AR activation not only inhibits the expression of proinflammatory cytokines but also enhances the transcription of inhibitory molecules such as TGF-β and Foxp3.

Recently, it has become clear that the CD4 homolog LAG-3 defines a subset of novel inducible regulatory T cells.^8,29,30  In the C3HA model, the induction of anergy correlates with increases in LAG-3 expression and the generation of these inducible regulatory T cells. In light of the ability of A_2AR stimulation to inhibit the generation of Th17 effector T cells in favor of Foxp3⁺ T cells, we inquired whether A_2AR engagement would also lead to an increase in LAG-3⁺ regulatory T cells. We stimulated CD4⁺ 6.5⁺ T cells in vitro with irradiated APCs and HA in the presence or absence of CGS and IL-2. The cells were harvested and the cDNA was assayed for LAG-3 expression by real-time PCR. T cells stimulated in the presence of CGS showed increased levels of LAG-3 transcript. The presence of exogenous IL-2, however, prevented the CGS-induced up-regulation of LAG-3 (Figure 7A). These results are consistent with our in vitro tolerance data suggesting that IL-2 can overcome A_2AR-induced inhibition (Figure 2D,E). We also analyzed LAG-3 expression from 6.5⁺ T cells that had been stimulated in vivo in C3HA host mice. Donor 6.5⁺ T cells were harvested from host mice 3 days after transfer and sorted. Wt T cells that were anergized in C3HA hosts displayed a dramatic increase in LAG-3 expression. In contrast, T cells from A_2AR^−/− 6.5⁺ mice did not have a robust increase in LAG-3 (Figure 7B). Because only the adoptively transferred cells are A_2AR^−/− (ie, host-derived APCs are Wt), these data suggest that endogenous adenosine plays a physiologic role in the up-regulation of LAG-3 in vivo. Consistent with this notion is the observation that pharmacologically inducing A_2AR signaling by treating hosts with CGS led to enhanced LAG-3 expression on donor T cells compared with donor T cells from vehicle-treated mice (Figure 7C).

Next we wanted to determine whether the increase in LAG-3 corresponded with increased regulatory activity in vivo. First, we transferred LAG-3^−/− 6.5⁺ T cells into C3HA hosts and treated them with PBS or CGS. Consistent with our data in Figure 5A, 75% of the vehicle-treated mice receiving Wt clonotypic T cells died, whereas 100% of the CGS-treated mice survived. On the other hand, eliminating LAG-3 completely abolished the ability of CGS to promote survival, as 100% of the mice that received LAG-3^−/− T cells died, regardless of whether they were treated with CGS or not (Figure 7D). These data suggest that the up-regulation of LAG-3 by A_2AR signaling plays a critical role in the ability of A_2AR signaling to promote tolerance and prevent tissue destruction.

Finally, we examined whether A_2AR signaling would lead to an increase in regulatory T cells that could mediate dominant-negative suppression on subsequent rechallenge with naive clonotypic cells. To this end, we designed an in vivo suppression assay. In the C3HA model, the mice that survive the initial transfer of autoreactive T cells develop LAG-3⁺ regulatory T cells that can suppress a second challenge of autoreactive 6.5⁺ T cells and prevent lethal tissue destruction.⁸  We injected the surviving mice of Figure 5A with a dose of T cells that would cause 100% death in naive mice. Indeed, by day 10, all of the naive mice receiving this high dose of clonotypic T cells died of lymphocyte-mediated tissue destruction of the lung (Figure 7E). Alternatively, consistent with previously published results, the 20% of the mice that survived the initial adoptive transfer of clonotypic T cells (Figure 5A) had developed Tregs, which suppressed the activity of the high dose of clonotypic cells and protected the mice from lung inflammation and death. Likewise, in accordance with our hypothesis that A_2AR signaling promotes the generation of Tregs, 100% of the CGS-treated hosts that survived the initial challenge of autoreactive T cells also survived the subsequent dose of T cells (Figure 7E). That is, 4 days of CGS treatment 1 month prior led to the generation of regulatory T cells that suppressed the activation of the newly transferred clonotypic T cells, preventing them from inducing tissue destruction and death.

A_2AR signaling on macrophages, T cells, and dendritic cells has been shown to directly inhibit effector function.^{19,,,–23,31,,–34}  Our data demonstrate that tissue-derived adenosine acts via the A_2AR to promote peripheral T-cell tolerance. Specifically, we demonstrate that A_2AR engagement can facilitate the induction of T-cell anergy, skew T-cell differentiation away from adaptive effector cells to adaptive regulatory cells, and promote the generation of LAG-3⁺ regulatory T cells.

T-cell anergy occurs when TCR engagement occurs in the absence of costimulation.³⁵  Anergic T cells produce less IL-2 and IFN-γ, and proliferate less on rechallenge to full stimulation (signal 1 + 2). In this report, we demonstrate that A_2AR stimulation can promote T-cell anergy even in the presence of costimulation. These observations are consistent with the ability of A_2AR stimulation to profoundly inhibit IL-2 production. However, not all inhibitors of IL-2 promote anergy. CSA, for example, potently abrogates IL-2 production but actually inhibits anergy induction by blocking NF-AT–inducible “anergic factors.” The Rao group has proposed that anergy is induced when TCR engagement results in NF-AT activation in the absence of AP-1 stimulation.³⁶  Along these lines, we are able to demonstrate that A_2AR stimulation can inhibit TCR-induced Erk phosphorylation and subsequent AP-1 binding (Figure 3B,D). These findings agree with a recent report demonstrating the ability of an A_2AR agonist to inhibit TCR-induced ZAP70 phosphorylation.²¹ 

Depending on the inflammatory microenvironment, TCR engagement can lead to the development of CD4⁺ adaptive effector cells, such as Th1, Th2, and Th17 cells.³⁷  Alternatively, in the setting of TGF-β without IL-6, TCR engagement can lead to the generation of “inducible” Foxp3⁺ regulatory T cells. Our data demonstrate the ability of tissue-derived adenosine, signaling through the A_2AR, to drive CD4⁺ T cells away from Th17 differentiation and toward Foxp3⁺ T-cell differentiation. Consistent with these observations, A_2AR stimulation resulted in increased TGF-β production with a concomitant decrease in the proinflammatory cytokine IL-6. Mechanistically, we propose that A_2AR stimulation enhances Foxp3 transcription by decreasing IL-6 expression and enhancing TGF-β expression. These observations are consistent with findings suggesting that A_2AR stimulation inhibits proinflammatory cytokine expression while sparing anti-inflammatory cytokine expression.²³  Interestingly, there have been several recent reports suggesting that adenosine acting, via the A_2AR, might mediate, in part, the suppressive function of regulatory T cells by engaging the A_2AR on the suppressed cells. It was found that CD39 and CD73 (both of which are ectoenzymes that promote the generation of extracellular adenosine) appear to be a more specific markers for Foxp3⁺ regulatory T cells than CD25.³⁸  The authors suggest that the generation of adenosine by regulatory T cells suppresses effector cells via the A_2AR. Likewise, in a colitis model, it was found that CD45RB^low or CD25⁺ T cells from A_2AR null mice fail to suppress CD45RB^high cells and prevent disease and effector cells from A_2AR null mice are not suppressed by suppressor cells derived from Wt mice.²³ 

In addition to the ability of A_2AR stimulation to promote anergy and the expression of Foxp3, we found that tissue-derived adenosine, engaging the A_2AR, promoted the generation of LAG-3⁺ T regulatory cells. LAG-3 is a CD4 homolog that is up-regulated on T-cell activation and remains elevated on a subset of T cells that possess regulatory cell function.⁸  Our in vitro and in vivo data demonstrate that A_2AR signaling promotes the generation of these regulatory T cells. T cells from LAG-3 null mice have diminished regulatory capacity, and the ectopic expression of LAG-3 confers the ability to suppress.^8,29  Recently, it has been shown that on TCR engagement, LAG-3 is cleaved from the cell surface of activated T cells by metalloproteinases.³⁹  This cleavage is necessary for robust T-cell proliferation and cytokine production, suggesting that LAG-3 also possesses cell intrinsic regulatory function. In our model, A_2AR agonists promote the generation of LAG-3 expressing T cells that have the capability of inducing tolerance and preventing autoimmunity. LAG-3 expression is required for suppressive function, as A_2AR agonists were unable to inhibit autoimmunity from the adoptive transfer of autoreactive LAG-3 null T cells. LAG-3 expression has been noted on tumor-infiltrating lymphocytes that have the capacity to suppress antitumor specific CD8⁺ T cells.³⁰  The fact that the tumor microenvironment contains relatively high levels of adenosine suggests that tumor-derived adenosine may promote the generation of these regulatory cells.^22,40 

For T cells, the consequences of TCR engagement are determined by the integration of multiple “accessory” signals.⁴¹  Just as IL-10 and TGF-β have been associated with inhibiting immune responses and promoting tolerance induction, we propose that tissue-derived adenosine acts as a “metabokine”⁴²  to perform a similar function. Because uric acid, HMG1-β, and low-molecular- weight species of extracellular matrix (ie, hyaluronic acid) might be considered examples of tissue-derived “danger” signals, we propose that tissue-derived adenosine, acting through the A_2AR, is one of many critical accessory signals that promote the induction of peripheral tolerance.^43,–45  As such, during “noninfectious” tissue injury, adenosine released into the microenvironment may contribute to the maintenance of peripheral tolerance to de novo released self-antigens.

The ability of A_2AR agonists to promote T-cell tolerance suggests that such agents could be used in the treatment ofT cell–mediated autoimmune disease and transplantation. In theory, A_2AR agonists might promote long-term, peripheral tolerance in the absence of continued immunosuppression. In support of such a hypothesis are studies showing that A_2AR agonists can prevent allograft rejection not only by acutely inhibiting T-cell activation but also by promoting the up-regulation of the inhibitory molecules PD-1 and CTLA-4.²¹  Alternatively, if tumor-derived adenosine inhibits antitumor responses by promoting the generation of anergic and suppressor T cells, A_2AR antagonists might prove to be important adjuvants to tumor immunotherapy. Indeed, a role for the A_2AR in protecting tumors from antitumor T cells has been described.⁴⁶ 

This work was supported by National Institutes of Health grants P50CA098252 and R01CA098109 and a grant from the Flight Attendants Medical Research Institute. C.G.D. is a Damon Runyon Clinical Investigator supported in part by the Damon Runyon Cancer Research Foundation. P.E.Z. and E.R.L. are doctoral candidates at the Johns Hopkins School of Medicine and this work is submitted in partial fulfillment of the requirement for the doctoral degree.

National Institutes of Health

Contribution: P.E.Z. performed the research and wrote the paper. C.T.H. performed research. E.R.L. performed research. J.K. assisted with microarray work. M.R.H. analyzed the data and assisted with research. J.L. provided the A_2AR null mice and helped design experiments. C.G.D. provided the C3HA and 6.5 mouse lines and analyzed the data. J.D.P. conceived of the project, designed the research and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Jonathan D. Powell, Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, CRB 443, 1650 Orleans Street, Baltimore, MD 21231; email: poweljo@jhmi.edu.