Generation of anti-inflammatory adenosine byleukocytes is regulated by TGF-β



Levels of anti-inflammatory extracellular adenosine are controlled by the sequential action of the ectonucleotidases CD39 and CD73, whose expression in CD4+ T cells has been associated with natural regulatory T cells (nTregs). We here show that CD73 expression on activated murine CD4+ T cells is induced by TGF-β independently of Foxp3 expression, operates at the transcriptional level and translates into gain of functional capacity to generate adenosine. In the presence of AMP, CD73 induced by TGF-β generates adenosine able to suppress proliferation of activated CD4+ T cells in vitro. These effects are contextual and opposed by proinflammatory cytokines. CD73 is also upregulated by TGF-β in CD8+ T cells, DCs and macrophages, so providing an amplification mechanism for adenosine generation in tissue microenvironments. Together, these findings expose a novel anti-inflammatory role for TGF-β.


Despite negative selection in the thymus, self-reactive T cells are still disseminated to the periphery, where their activity is, at least in part, constrained by regulatory CD4+ T cells (Tregs). As yet, there is no consensus on mechanisms used by Tregs to control inflammation and immunity 1. Recently, CD4+Foxp3+ natural regulatory T cells (nTregs) have been shown to express high levels of both CD39 and CD73 able to deliver anti-inflammatory adenosine to the surrounding milieu 2–4.

CD73, an ecto-5′-nucleotidase, is a cell surface enzyme that converts extracellular adenosine 5′-monophosphate (AMP) to adenosine. Adenosine mediates a variety of effects through four G-protein-coupled purinergic receptors, A1, A2A, A2B and A3, expressed in different immune cells 5. Adenosine receptor A2A has been shown to be critical in the control of inflammation in several in vivo models 6, and extracellular adenosine acts on many immune cells to mediate anti-inflammatory effects 7 ranging from inhibition of Th1 and Th17 cells 8, to DCs 9, 10, NK cells 11, neutrophils 12, 13 and macrophages 14.

CD39 converts extracellular adenosine 5′-triphosphate (ATP) into AMP, thereby providing the substrate for CD73. In hypoxia or tissue injury, high intracellular concentrations of ATP (typically 104-fold higher than extracellular concentrations) are released from damaged cells causing a strong pro-inflammatory stimulus to T cells, DCs and neutrophils 15. By the sequential action of the cell surface nucleotidases CD39 and CD73, this pool of pro-inflammatory ATP can also be exploited to resolve inflammation by conversion to adenosine. Inflammation is, then, reduced both by removal of ATP and by the increase of adenosine concentrations 16.

Transforming growth factor-β (TGF-β) affects many cell types within the immune system and plays a crucial role in immune homeostasis 17, 18. Ablation of Tgfb1, the gene encoding the dominant isoform of TGF-β in the immune system, results in severe multifocal inflammatory disease 19. T cells are critical to this immunopathology as transgenic mice with T-cell specific ablation of signalling through TGF-β receptor-II show spontaneous activation of CD4+ T cells and also develop wasting multi-organ inflammatory disease 20, 21. TGF-β is known to inhibit both the proliferation and the effector function of T cells 18, 22. In addition, TGF-β can also induce expression of Foxp3 and confer regulatory capacity on T cells in the periphery 23, 24. We previously showed that TGF-β signalling in T cells is required for the generation of skin graft tolerance by therapeutic co-receptor-directed antibodies 23, through the conversion of naïve T cells to “induced” Foxp3+ regulatory T cells (iTregs) able to prevent graft rejection 25.

In contrast to the demonstration of CD39 and CD73 expression on nTregs, little is known regarding the relevance of these enzymes to the function of iTregs, nor how their expression is regulated in other immune cells. We here explore the effect of TGF-β on CD39 and CD73 expression and function. We show that TGF-β plays a key role in controlling expression of cell surface CD73 and generation of adenosine by iTregs and other immune cells, and that this effect can be antagonized by several pro-inflammatory cytokines. TGF-β-treated CD4+ T cells suppress proliferation of other CD4+ T cells in a CD73-dependent Foxp3-independent manner, provided substrate (AMP) is available. We propose that the generation of anti-inflammatory adenosine by both adaptive and innate immune cells provides an additional mechanism by which TGF-β controls inflammation.


CD39 and CD73 are widely expressed within lymphocyte populations

Coordinated expression of CD39 and CD73 resulting in accumulation of adenosine has been suggested as an inhibitory mechanism used by nTregs to modulate immune responses 4. Using C57BL/6J mice, we have confirmed high-level expression of these enzymes on CD4+ nTregs in the spleen (Supporting Information Fig. 1).

Natural Tregs constitute a minor population of splenic lymphocytes, and expression of CD39 and CD73, although enriched on Foxp3+ cells, was not unique to them. CD73 was detected on ∼80% of CD8+ T cells and ∼60% of total CD4+ T cells, whereas mouse B cells were CD73 (Supporting Information Fig. 1A and B). Aside from Foxp3+ cells, the majority of the other CD4+ T cells also expressed CD73, albeit at a lower level. Similarly, CD39 was enriched on Foxp3+ cells (∼80% expressing), but was also expressed on ∼15% of the CD4+Foxp3 T cells, and some CD8+ T cells. B cells were ∼95% positive for CD39 and comprised the majority of CD39+ cells in the spleen. Among CD4+ T cells in the spleen, Foxp3+ cells contributed around 50% of the CD39+ cells and 20% of the CD73+ cells (Supporting Information Fig. 1C). Expression of these molecules does, however, change following lymphocyte activation.

T-cell activation downregulates CD73 but upregulates CD39

We activated splenic T cells or B cells in vitro by culturing them for 3 days with anti-CD3/anti-CD28-coated beads or LPS, respectively. Surface expression of CD39 increased on both CD4+ and CD8+ T cells, with the majority of T cells expressing CD39 by day 3 of culture (Fig. 1A). The presence of CD39 on effector T cells conflicts with the idea that T-cell expression of CD39 is restricted to Tregs. In contrast, we observed that CD73 expression diminished on CD4+ T cells upon activation. This downregulation occurred mainly on Foxp3 cells while Foxp3+ cells maintained high levels of CD73 (Fig. 1B and C). CD39 and CD73 expression on B cells was unaffected by LPS activation (Fig. 1A).

Figure 1.

CD39 and CD73 expression on T cells is altered by activation and by TGF-β treatment. Naïve C57BL/6J splenic lymphocyte populations were autoMACS-sorted into CD4+ T cells, CD8+ T cells and B cells, and activated for 3 days: T cells were cultured with anti-CD3/anti-CD28-coated beads and B cells were activated with LPS. CD8+ T-cell cultures were supplemented with exogenous IL-2. (A) Expression of CD39 and CD73 on the surface of naïve cells is represented by the yellow filled histogram. Coloured lines show levels detected on the surface of cells activated in the presence of anti-TGF-β (1D11) (green), the isotype control antibody (13C4) (blue), or TGF-β (red). (B) CD73 surface staining of total CD4+, CD4+Foxp3 and CD4+Foxp3+ T cells, either unstimulated (D0), or after day 1 (D1) to day 3 (D3) of stimulation, is represented as the geometric mean fluorescence. 1D11: anti-TGF-β; 13C4: isotype control. (C) Surface expression of CD39 and CD73 and their co-expression with intracellular Foxp3 on naïve CD4+ T cells and on CD4+ T cells activated for 3 days in the presence of TGF-β or anti-TGF-β antibody is shown. In all panels (A–C) the data are representative of at least three independent experiments.

We next investigated whether the cytokine milieu in which T cells are activated affects expression of CD39 and CD73, and, in view of the known role of TGF-β in inducing Tregs and damping inflammation, whether TGF-β had an influential role.

TGF-β induces CD73 expression on the surface of T cells

TGF-β added to cultures of activated splenic T cells demonstrated a clear effect on CD39 and CD73 expression (Fig. 1A). It prevented activation-induced downregulation of CD73, and indeed enhanced its expression by CD4+ T cells. This enhancement of CD73 was seen in both Foxp3+ and Foxp3 cells (Fig. 1C). In contrast, a neutralizing anti-TGF-β antibody (1D11) did the opposite, and enhanced activation-induced downregulation of CD73, most likely by neutralizing calf serum-derived TGF-β (as 1D11 also reacts against bovine TGF-β 26) or TGF-β generated from the cultured cells (Fig. 1A).

The high expression of CD73 on naïve CD8+ T cells was only slightly reduced when TGF-β was neutralized, but increased with addition of TGF-β (Fig. 1A). Typically, at day 3, there was a ten-fold increase in fluorescence on TGF-β-treated CD8+ T cells when compared with cells treated with blocking anti-TGF-β antibody.

The upregulation of CD39 on CD4+ T-cell activation also seemed to be enhanced in the presence of TGF-β (Fig. 1A), and the expression of CD39 on CD8+ T cells during activation followed a similar pattern to that observed on CD4+ T cells. Expression of both CD39 and CD73 on LPS-activated B lymphocytes was unaffected by TGF-β (Fig. 1A).

De novo induction of CD73 by TGF-β is independent of Foxp3

To further study the effects of TGF-β on T-cell induction of CD73 we used TCR transgenic A1.RAG-1−/− mice (A1.RAG). All of the lymphocytes within these mice are CD4+ T cells, exclusively expressing a transgenic TCR specific for the male peptide Dby on a RAG−/− background. Because of their defined TCR affinity, selection within the thymus generates no Foxp3+nTregs 27. As described previously, in vitro culture of A1.RAG splenocytes activated in the presence of TGF-β induces Foxp3+ Tregs with suppressive capacity both in vitro and in vivo 23, 25. Naïve A1.RAG CD4+ splenocytes expressed negligible levels of CD73 and contained only very low numbers of CD39+ T cells compared to WT mice, making this an excellent model to study de novo induction of the ectonucleotidases on naïve CD4+ T cells (Fig. 2).

Figure 2.

TGF-β induces de novo expression of CD73 on A1.RAG CD4+ T cells. (A–C) AutoMACS-sorted splenic A1.RAG CD4+ T cells were cultured in vitro with anti-CD3/anti-CD28-coated beads and IL-2, in the presence or absence of TGF-β. Cells were assessed for intracellular expression of Foxp3 and surface expression of CD39 and CD73 at daily intervals over a four-day culture period. The percentages of CD4+ cells positive for either (A) Foxp3, (B) CD39 or (C) CD73 were determined by FACS and are presented as mean values+SEM from at least three independent biological replicates. For each time-point, unpaired Student's t-test was used to compare cells cultured in the presence or in the absence of TGF-β. (D) A representative staining of A1.RAG splenic naïve CD4+ T cells cultured as in (A) and treated with TGF-β or with isotype control antibody 13C4 for 4 days is shown. (E) The percentages of cells expressing Foxp3, CD39 and CD73 were also assessed on day 4 in response to titrated concentrations of TGF-β. (F) A1.RAG splenocytes were cultured for 7 weeks with BMDCs and Dby peptide in the presence of TGF-β, IL-6, anti-IL-4 (10 μg/mL) and anti-IFN-γ (10 μg/mL) antibodies to generate T cells expressing IL-17A. The cells were activated with PMA and ionomycin for 4 h, treated with brefeldin A for the last 2 h of activation, and permeabilized with saponin buffer prior to staining for IL-17A. Co-expression of surface CD73 and intracellular IL-17A was assessed by FACS. The staining is representative of three independent biological replicates.

A1.RAG splenic CD4+ T cells were activated in vitro with or without TGF-β. TGF-β-induced expression of Foxp3 as expected (Fig. 2A). The majority of the T cells upregulated CD39 independently of TGF-β (Fig. 2B), whereas TGF-β was necessary for substantial CD73 induction in CD4+ T cells (Fig. 2C and D). Both Foxp3 and CD73 induction were TGF-β dose-dependent (Fig. 2E).

We compared the expression patterns of CD73 and Foxp3. After activation in the presence of TGF-β for 4 days, only a proportion of the T cells expressed Foxp3 while many more expressed CD73 (Fig. 2E), and only a fraction of CD73+ T cells co-stained for Foxp3 (Fig. 2D). In addition to its anti-inflammatory effect, TGF-β in combination with IL-6 is pivotal for the induction of pro-inflammatory Th17 cells, which do not express Foxp3. A1.RAG T cells were stimulated for several weeks in the presence of IL-6 and TGF-β to generate Th17 cells and these IL-17+Foxp3 cells (as well as IL-17Foxp3 cells) were also found to be CD73+ (Fig. 2F), again showing that Foxp3 expression is not essential for the induction of CD73 expression. Furthermore, CD4+ T cells from Foxp3−/− mice activated in the presence of TGF-β induced CD73 at levels comparable to those of Foxp3+/+ T cells (Supporting Information Fig. 2). We conclude that it is exposure to TGF-β rather than Foxp3 expression or functional lineage of the T cells that determines CD73 induction on activation.

Pro-inflammatory cytokines antagonize the induction of CD73

A number of pro-inflammatory cytokines impaired the capacity of TGF-β to induce nucleotidase expression in naïve T cells. Of the cytokines tested, only TGF-β was capable of inducing CD73 on activated A1.RAG naïve CD4+ T cells (Fig. 3). Remarkably, this TGF-β-induced upregulation was prevented by Th1-related cytokines IL-12 and IFN-γ, the Th2 cytokine IL-4, and by IL-6, IL-10 and IL-21 (Fig. 3A and B), whereas IL-9 and IL-23 had no effect. Activation in the presence of certain inflammatory cytokines (e.g. IL-12) does not preclude later induction of CD73 by TGF-β once those cytokines are removed. In contrast, exposure to certain other cytokines (e.g. IL-4) seems to render the cells resistant to subsequent TGF-β-mediated CD73 induction (Supporting Information Fig. 3). IL-4, IL-6, IL-10, IL-12, IL-21 and IFN-γ also inhibited induction of Foxp3, albeit variably (Fig. 3C). The TGF-β treated cells that, despite inflammatory cytokine treatment, were capable of inducing Foxp3, show reduced intensity of CD73 expression, in particular those receiving IFN-γ (data not shown). In contrast to CD73, the level of CD39 upregulated on T-cell activation was not significantly altered by any of the cytokines used (Fig. 3D).

Figure 3.

CD73 induction by TGF-β is modulated by inflammatory cytokines. AutoMACS-sorted splenic A1.RAG CD4+ T cells were cultured in vitro with anti-CD3/anti-CD28-coated beads and IL-2 for 4 days, in the presence or absence of TGF-β, with or without other cytokines as indicated from day 0. 1D11: anti-TGF-β Ab; 13C4: isotype control. (A–D) Gating on CD4+ T cells, the percentages of positive cells for CD73 (A), Foxp3 (C), CD39 (D) and the fluorescence intensity of CD73 (geometric mean) (B) were determined by FACS at day 4 and presented as mean values+SEM of at least three independent biological replicates. Statistical significance was assessed using one-way ANOVA and each treatment was compared using Dunnett's multiple comparison post-test to the reference treatment “13C4”. Cells cultured with TGF-β plus other cytokines were compared using Dunnett's multiple comparison post-test to the reference treatment “TGF-β”. (E and F) CD73 mRNA expression normalized to HPRT is shown relative to values obtained for the reference sample “13C4” of cells activated in the absence of additional cytokines or neutralizing anti-TGF-β antibody (E); and is shown relative to values obtained for cells cultured with TGF-β alone (F). The mean values+SEM of three independent biological replicates are shown. The log transformed RQ values and one-way ANOVA were used to determine the statistical significance, and each treatment was compared to the reference treatment “13C4” in (E) or to the reference treatment “TGF-β” in (F) using Dunnett's multiple comparison post-test.

TGF-β upregulates CD73 mRNA expression in CD4+ T cells

The changes in surface CD73 on cytokine-influenced activated A1.RAG T cells were paralleled at the transcriptional level. Using qPCR we observed that levels of CD73 mRNA were reduced on activation (Fig. 3E). As seen with the surface staining, addition of TGF-β, but no other cytokine, caused a substantial increase in CD73 mRNA; more than 150-fold higher than that of the control activated population. Again, as for the surface staining, this induction of transcription was antagonized by certain pro-inflammatory cytokines, particularly IFN-γ (Fig. 3F).

TGF-β signalling is not required for constitutive expression of Foxp3 and CD73 on naïve T cells

Transgenic C57BL/6J mice expressing a T-cell specific dominant-negative form of TGF-β receptor type II (CD4.dnTGFβRII mice) develop multi-organ auto-immunity at around 3–4 months, illustrating the importance of TGF-β signalling to T cells for the maintenance of self-tolerance 20. We asked whether such mice expressed CD73 on their T cells if unable to receive a TGF-β signal (Supporting Information Fig. 4).

We observed only a minor difference between Foxp3 expression in naïve spleen CD4+ T cells from CD4.dnTGFβRII compared to control mice (Supporting Information Fig. 4A). CD73 expression in CD4+ and CD8+ T cells was similar between the two mouse strains and the expression on Tregs was also comparable (Supporting Information Fig. 4B and E). These data suggest that TGF-β signalling is not essential for constitutive expression of CD73 in naïve splenic T cells.

TGF-β induces CD73 expression on DCs and macrophages

T cells are not the only cells whose expression of CD73 could be modulated by TGF-β. The percentage of CD11c+CD73+ cells and also the level of CD73 expressed were strongly increased by inclusion of TGF-β in bone marrow-derived dendritic cell (BMDC) cultures (Fig. 4A and B), whereas CD39 is expressed constitutively in these cells. As seen with T cells, CD73 induction could be explained by changes in transcript abundance (Fig. 4C). Levels of CD73 were increased by TGF-β in both immature and LPS-matured BMDCs (data not shown). CD73 expression on elicited peritoneal macrophages was also increased by TGF-β, but not by IL-4, IL-21 or IFN-γ (Fig. 4D and E). Furthermore, CD73 induction by TGF-β was strongly inhibited by the addition of exogenous IL-4.

Figure 4.

TGF-β induces CD73 expression on DCs and macrophages. (A–C) Seven-day cultures of BMDCs were cultured for a further 4 days with the addition of 2 ng/mL of TGF-β or 10 μg/mL of neutralizing anti-TGFβ (1D11) or isotype control (13C4) antibody, and stained for FACS. (A) Representative stainings are shown indicating surface expression of CD39 and CD73 on CD11c+-gated cells. (B) The percentage of CD11c+ cells expressing CD73 and the geometric mean fluorescence intensity of CD73 expression on CD11c+ cells, as measured by FACS staining are shown as histograms. Mean values+SEM of three independent biological replicates are shown. Statistical significance was assessed using one-way ANOVA and each treatment was compared by Dunnett's multiple comparison post-test to the reference treatment “13C4”. (C) Expression of CD73 mRNA (normalized to HPRT) was determined using qPCR and is shown relative to the expression in the reference sample “13C4”. The log transformed RQ values and one-way ANOVA were used to determine the statistical significance, and each treatment was compared with the reference treatment “13C4” using Dunnett's multiple comparison post-test. (D and E) Biogel-elicited peritoneal macrophages were cultured for 4 days, prior to FACS staining to assess surface expression of CD73, with cytokines added as indicated from day 0 of the culture. (D) CD73 geometric mean on F4/80+ cells+SEM of FACS staining of three independent biological replicates are shown. Statistical significance was assessed using one-way ANOVA and each treatment was compared by Dunnett's multiple comparison post-test to cells that received no cytokine treatment or to samples grown in the presence of TGF-β. (E) Representative examples of CD39 and CD73 staining of F4/80+-gated cells cultured with or without TGF-β are shown.

TGF-β promotes the generation of adenosine by T cells and DCs in a CD73-dependent manner

De novo induction of CD73 on A1.RAG T cells mediated by TGF-β corresponded with a gain in capacity to generate adenosine from its AMP substrate (Fig. 5A). This could be blocked by the addition of the CD73 competitive antagonist, adenosine 5′-(αβ-methylene) diphosphate (APCP) 28. No other cytokine tested promoted adenosine generation, but the inflammatory cytokines IL-4, IL-12, IL-21 and IFN-γ suppressed the inductive effect of TGF-β as expected. In the same manner, TGF-β could enhance adenosine production by DCs (Fig. 5B).

Figure 5.

TGF-β-induced CD73 can generate de novo adenosine production by CD4+ T cells and DCs and mediate CD4+ T-cell suppression. Supernatant levels of adenosine generated from AMP were measured using the DiscoveRx PathHunter β-Arrestin assay with the ADORA2B receptor-transfected reporter cell line. (A) Levels of adenosine generated by autoMACS-sorted splenic A1.RAG CD4+ T cells cultured in vitro with anti-CD3/anti-CD28-coated beads and IL-2 for 4 days are shown. Cells were cultured in the presence or absence of TGF-β from day 0, plus additional cytokines as indicated. (B) Adenosine levels generated by 7-day BMDCs cultured for a further 4 days with TGF-β, anti-TGFβ antibody (1D11) or isotype control antibody (13C4) are shown. In (A and B), assays were made in triplicate and the values were compared with background levels detected using culture medium alone. Statistical significance was assessed using one-way ANOVA and each treatment was compared by Dunnett's multiple comparison post-test to the reference “medium” background. Unpaired Student's t-test was used to compare adenosine production by TGF-β-treated cells in the presence, or in the absence of the inhibitor APCP. In each case, data representative of three independent biological replicates with similar results are shown. (C and D) AutoMACS-sorted splenic CD4+ T cells from CD73−/− mice and female Marilyn Foxp3−/− mice were cultured in vitro with anti-CD3/anti-CD28-coated beads and IL-2 for 5 days, in the presence or absence of TGF-β from day 0. At day 5 of culture, cells were harvested and the “suppressor” cells were treated with mitomycin C to prevent their proliferation. Cells were then plated for a 72-h proliferation assay stimulated by IL-2 at a “suppressor” to responder cell ratio of 1 to 1. Proliferation was measured by tritiated thymidine incorporation in the last 18 h of the 72 h assay. AMP was added to the cultures at 1, 0.5, 0.1 and 0.05 mM as indicated and APCP added at 0.5 mM. Triplicate wells were set for each condition and the bars show the means of the three wells+SEM. The graphs shown are from one of the two independent biological replicates with similar results. Statistical significance was assessed using one-way ANOVA and each treatment was compared by Dunnett's multiple comparison post-test to the reference cells receiving no AMP in each condition. Unpaired Student's t-test was used to compare the suppression by TGF-β-treated Marilyn Foxp3−/− cells in the presence or in the absence of the inhibitor APCP.

These data suggest that any source of TGF-β in tissues can upregulate CD73 during activation of T cells, on DCs, macrophages and possibly other cells, so creating the potential for an adenosine-rich anti-inflammatory microenvironment, given sufficient levels of AMP are available and pro-inflammatory cytokines absent.

TGF-β-induced Treg are able to suppress lymphocyte proliferation in a CD73-dependent manner

To assess the functional relevance of TGF-β-induced CD73 on iTreg-mediated suppression, we adopted an in vitro T-cell proliferation assay where CD73 substrate AMP was provided (Fig. 5C and D). We avoided any background conversion of AMP to adenosine by using CD73−/− cells 29 as responder cells. The proliferation of pre-activated CD73−/− CD4+ T cells stimulated with IL-2 was not suppressed by TGF-β-treated cells in the absence of AMP. However, when AMP was provided, TGF-β-treated cells suppressed T-cell proliferation (Fig. 5D). This suppression was inhibited by the CD73 competitive inhibitor molecule APCP, so demonstrating CD73 involvement. In contrast, TGF-β treatment of CD73−/− CD4+ T cells did not endow those cells with the capacity to suppress (Fig. 5C). As shown before, Foxp3 expression is not required for CD73 induction by TGF-β. We used TGF-β-treated CD4+ T cells from Foxp3−/− TCR transgenic Marilyn mice to circumvent any confusing effects of Foxp3-mediated suppression. These results show that TGF-β induces CD73-dependent immune suppression independent of Foxp3 where substrate AMP concentrations were available.


Expression of CD39 and CD73, and a capacity to generate adenosine has been suggested as one mechanism within the arsenal of Treg suppressive actions. We here show that CD73 and its enzymatic activity can be upregulated by TGF-β in several immune cell types.

In vivo the sources of extracellular ATP for adenosine-based immune regulation are diverse. Intracellular concentrations of ATP can reach 3–10 mM, and extracellular release of ATP can occur by passive mechanisms, such as necrotic cell death and inflammation 15. ATP is also actively released by some immune cells, e.g. neutrophils 15 or monocytes 30. We propose a two-step model for ATP removal and adenosine regulation orchestrated by the local cytokine environment, with TGF-β performing a key role.

First, we propose that the widespread expression of CD39 across different lymphocyte populations limits ATP-mediated inflammation within healthy tissues. B cells and nTregs also express it constitutively. Recent publications have recognized that CD39 is also expressed on memory CD4+ T-cell populations. These cells can catabolise ATP, but fail to show regulatory capacity 31, 32. The broad expression of CD39 suggests that the ability to generate sufficient AMP substrate for CD73 is normally readily available in B cells and nTregs, and that CD4+ and CD8+ T cells can be induced to enhance CD39 expression and AMP generation upon activation. As TGF-β further increases the levels of CD39 induced by activation, more pro-inflammatory ATP is removed and further AMP substrate is made available for CD73. AMP was also shown to be released by activated neutrophils and additional sources might contribute more substrate for adenosine generation by CD73 33.

Second, CD73 is the enzyme that converts AMP into anti-inflammatory adenosine. CD73 is more limited in its expression within lymphocytes than CD39, and although found on CD8+ T cells and on subsets of CD4+ T cells, it is downregulated on activation. CD73 is induced rapidly on activated CD4+ T cells in the presence of TGF-β, unless antagonistic pro-inflammatory cytokines, such as IL-12, IFN-γ, IL-4 or IL-21, predominate. Inhibition of adenosine production by proinflammatory cytokines might ensure unopposed protective immunity against pathogens.

Generation of different functional types of T cells depends on the cytokine environment in which they are activated. Inflammatory Th1 and Th2 cytokines IL-12 and IL-4 did not induce CD73 expression, and polarized Th1 and Th2 clones resisted induction of CD73 by TGF-β (data not shown). Induction of cell surface expression of CD73 by TGF-β is affected by the presence of several proinflammatory cytokines. Control of CD73 expression by TGF-β occurred primarily at the level of transcription, and was negatively influenced by other cytokines, in particular IFN-γ. Although we did not investigate how IFN-γ interfered with TGF-β-induced transcription of CD73 mRNA, IFN-γ is known to inhibit TGF-β signalling 34, 35.

Constitutive expression of adenosine-generating machinery might be part of the suppressive mechanisms used by nTregs 2, 3. Although CD39 and CD73 are not exclusively expressed by nTregs, Tregs may play an important and coordinating role in perpetuating adenosine-mediated regulation, not only by generating adenosine themselves, but also by producing TGF-β; nTregs can express surface bound and secreted TGF-β thought to be necessary for nTregs to suppress in vivo 36, 37. Our data suggest that Treg-generated TGF-β might act on a variety of immune cells to induce CD73 expression and enhance adenosine generation in the tissues.

In previous studies, we observed that tolerance induced by co-receptor blockade required TGF-β signalling to T cells 25. TGF-β can induce several Treg-related molecules, including Foxp3. Our data suggest that TGF-β might act on T cells and also other cells in the immediate vicinity to affect their adenosine-generating capacity, offering one further mechanism by which TGF-β can dampen inflammation and confer suppressive function on naïve T cells 38.

CD73 expression and adenosine generation have been associated with immune suppression in several diseases where TGF-β plays an important role 39–43. These might now be explained, in part, by TGF-β effects on adenosine generation. So far, therapeutic exploitation of the ATP-adenosine pathway has focused on targeting adenosine receptors with agonists or antagonists 44, 45, or the use of exogenous CD73 16. Our results highlight a novel cytokine-dependent pathway for controlling adenosine metabolism that could now educate therapeutic manipulation in chronic inflammation, infection, transplantation and oncology.

Materials and methods


CBA/Ca, C57BL/6J, A1.RAG, CD4.dnTGFβRII, Marilyn. RAG1−/−46 and Marilyn.RAG1−/−Foxp3−/−mice were bred and maintained under specific pathogen-free conditions at the Sir William Dunn School of Pathology, University of Oxford, UK. C57BL/6 Foxp3−/− were generated by Alexander Rudensky's group 47 and were backcrossed twice to Marilyn. RAG−/− mice to generate Marilyn.RAG1−/−Foxp3−/−. CD73−/− mice were kindly provided by Linda Thompson and obtained from Paul Klenerman 29. All experimental procedures were conducted in accordance with the Home Office Animals (Scientific Procedures) Act of 1986 and received local ethical committee approval.

Antibodies and FACS

For FACS, cells were resuspended in PBS containing 1% v/v FCS and 0.01% w/v NaN3 with the appropriate monoclonal antibody and incubated at 4°C for 30 min. Anti-CD39 PE (24DMS1), anti-CD73 PE and biotin (eBioTY/11.8), anti-CD19 A647 (eBio1D3) and anti-Foxp3 (FJK-16s) antibodies were obtained from eBiosciences. Intracellular Foxp3 staining was performed using manufacturer's protocol. Anti-CD8 PercP (53-6.7), anti-CD4 APC (RM4-5), anti-CD4 FITC (H129.19), anti-IL-17A FITC (TC11-18H10.1) antibodies, and streptavidinPerCP and APC were obtained from BD Pharmingen. Cells were fixed in 2% w/v formaldehyde in PBS and four-color analysis performed using FACSCalibur (BD Biosciences) with the FlowJo software. Anti-TGF-β (1D11) 26 and its isotype control (13C4) obtained from Genzyme, anti-IL-4 (11B11) and anti-IFN-γ (XMG1.2), produced in-house, were used at 10 μg/mL.

Cell separation and cultures

Splenic CD4+ T cells (75–80% pure), CD8+ T cells (80–90% pure) and B cells (95% CD19+) were obtained using MACS isolation kits (Miltenyi Biotec) according to manufacturer's instructions, and selected using AutoMACS (Miltenyi Biotec). Anti-CD3/anti-CD28-coated beads (Invitrogen) were used at 1:1 ratio with T cells. LPS was used at 1 μg/mL to activate B cells.

Lymphocytes and BMDCs were cultured in RPMI-1640 medium (Lonza) containing 10% v/v FCS (Gibco), 50 μg/mL of penicillin–streptomycin, 1 mM sodium pyruvate solution, 0.1 M HEPES buffer solution (all from PAA Laboratories GmbH) and 50 μM 2-mercaptoethanol. DCs were produced as described in 48. GM-CSF was produced from X-63 transfectant, provided by Dr. D. Gray. Macrophages, elicited by injecting 0.5 mL of Bio-Gel P100 polyacrylamide beads (Bio-Rad Laboratories) intraperitoneally, were obtained by peritoneal lavage and cultured for 4 days in OptiMEM medium supplemented with 10% v/v FCS and 50 μg/mL of penicillin and streptomycin. For macrophage staining, 2.5 μM EDTA was added to prevent adhesion and Fc receptor binding was blocked with 2.4G2 at 10 μg/mL.


Adenosine, AMP sodium salt, APCP (CAS 3768-14-7), NBTI (CAS38048-32-7), EHNA hydrochloride (CAS 58337-38-5), PMA (used at 50 ng/mL), ionomycin (500 ng/mL), brefeldin A (10 μg/mL) and LPS (1 μg/mL) were bought from Sigma. Dby peptide REEALHQFRSGRKPI was used at 10 nM. Recombinant human TGF-β1 (used at 2 ng/mL) and recombinant mouse IL-6 (50 ng/mL), IL-9 (10 ng/mL), IL-12 (10 ng/mL), IL-21 (50 ng/mL) and IL-23 (20 ng/mL) were obtained from R&D Systems and Peprotech, recombinant mouse IL-2 (10 ng/mL), IL-4 (10 ng/mL) and IFN-γ (10 ng/mL) from BD Pharmingen and recombinant mouse IL-10 (15 ng/mL) from eBiosciences.

Quantitative PCR

Agilent-verified total RNA was reverse transcribed using random hexamer primers. Three biological replicates per sample were analysed as triplicate technical replicates using commercial qPCR assays Mm00446968_m1 and Mm00501910_m1 (Applied Biosystems). Expression of CD73 was normalized to HPRT and expressed relative to a given reference population using the comparative ΔΔCt method, essentially as described in ABI's User bulletin ♯2. Biological error was assessed using RQmax and RQmin derived using ±SD of the sample ΔCt values.

Measurements of adenosine production

Dead cells and APCs were removed from A1.RAG splenocyte cultures using Histopaque-1083 (Sigma). In all, 2×105 cells were incubated in 96-well plates in 200 μL of RPMI with 10% v/v FCS for 1 h in the presence of 5 mM AMP, 10 μM NBTI and 50 μM EHNA hydrochloride, with or without 1 mM of APCP. Supernatants were then tested with PathHunter assay ADORA2B according to the protocols provided by DiscoveRx. The assay can reliably measure adenosine concentrations above 1 μM, reaching signal saturation at around 100 μM. Concentrations of adenosine were calculated using a standard curve generated using adenosine diluted in culture medium.

Proliferation assay

“Responder” pre-activated CD73−/− CD4+ T cells were plated at 2.5×104 per well in 200 μL of RPMI containing 10% v/v FCS heat inactivated (56°C for 45 min) and stimulated with 10 μg/mL of IL-2. “Suppressor” cells were treated with 25 μg/mL of mitomycin C for 1 h at 37°C to prevent their proliferation and plated at 2.5×104 per well (co-cultures at 1:1 ratio of responders to “suppressors”). AMP was added to the cultures at 1, 0.5, 0.1 and 0.05 mM as indicated and APCP added at 0.5 mM. At 56 h, co-cultures were pulsed with 0.5 μCi/well of tritiated thymidine (3H-TdR) (Amersham International) and assessed after a further 18 h.

Statistical analysis

To determine statistical significance we carried out one-way ANOVA test with Dunnett's multiple comparison post-test or Student's t-test (unpaired, two-tailed) using the GraphPad Prism software, In the figures, p-values<0.05 are indicated by *, p<0.01 by ** and p<0.001 by ***, whereas non-significant p-values are labelled “ns”.


The authors thank the Pathology Support Building staff for their excellent contribution. This work was supported by UK Medical Research Council grants G7904009 (DH, KFN, SPC and HW) and G1000215 (DH, SPC and HW). F. S. R. received a PhD scholarship from Fundaçãopara a Ciência e Tecnologia, Portugal. They are also grateful to Linda Thompson, Paul Klenerman and Stuart Sims for the use of the CD73−/− mice, to Alexander Rudensky group that created the Foxp3−/− mice, and to Richard Flavell for the use of the CD4.dnTGFβRII mice.

Conflict of interst: The authors declare no financial or commercial conflict of interest.