Extracellular adenosine 5′-triphosphate (ATP) acts on many immune cells to promote inflammation. Conversely, the ATP metabolite adenosine is mainly an anti-inflammatory molecule. The ecto-enzymes CD39 and CD73 can dephosphorylate extracellular ATP to adenosine, thereby controlling this important pathway of immune modulation. Despite their established roles in the immune system, little is known of how CD39 and CD73 are themselves regulated. Recent data have shown that CD73 expression and adenosine generation are up-regulated by transforming growth factor-β, depending on the cytokine content of the local microenvironment. We review here these recent findings and discuss their implications in disease.
The differential response to allografts depending on their location gave rise to the concept of ‘intrinsic immune privilege’. Organs where even minor episodes of inflammation could have long-term consequences for the survival or reproduction of the organism, such as eye, brain, testis or the placenta, were seen to curtail damaging immune responses, and were considered immune sanctuaries. This operational ‘unresponsiveness’ was shown not to be a failure to see antigen but, rather, a consequence of active mechanisms. It has long been considered that an understanding of these protective processes might provide targets for therapeutic intervention. Evidence emerging from our own studies, where allografts were tolerated following immunotherapy (e.g. co-receptor blockade with antibodies), have led our group to propose the concept of ‘acquired immune privilege’, whereby tolerated tissues harbour many overlapping privileged microenviroments that are generated de novo. Acquired immunological privilege also proved to be dependent on active mechanisms, and regulatory T cells (Treg) were shown to play an important role in their co-ordination [2–5]. The literature abounds with potential tissue protective mechanisms , including well-established anti-inflammatory cytokines such as transforming growth factor (TGF)-β or interleukin (IL)-10  and less conventional immune mediators such as carbon monoxide, biliverdin and free iron produced by haem-oxygenase [8–10], or arginine and tryptophan catabolism by arginase  and indoleamine 2,3 dioxygenase [12–14]. Our recent work has suggested that many other amino acid catabolic enzymes, in addition to those depleting tryptophan or arginine, may also contribute to ‘acquired immune privilege’. Murine Treg appear to have the capacity to induce the expression of these catabolic enzymes in dendritic cells (DCs) and, in the environment of low essential amino acid concentrations, T cell effector responses may be curtailed. Nutrient sensing processes act through the GCN2 and mammalian target of rapamycin (mTOR) signalling pathways to suppress proliferation and, in synergy with TGF-β, to increase the conversion of naive T cells to ‘induced’ forkhead box protein 3 (FoxP3)-expressing Treg (iTreg), highlighting a new role for inhibition of mTOR in immune privilege and infectious tolerance . Recently, much attention has been given to the adenosine 5′-triphosphate (ATP)–adenosine pathway as a mechanism of immune regulation in the microenvironment. Consequently, the generation of anti-inflammatory adenosine will provide the focus for the remainder of this paper.
Extracellular adenosine is anti-inflammatory while extracellular ATP is a ‘danger signal’
In the extracellular space, adenosine can act as a signalling mediator for a variety of physiological effects in many organs. The effects of extracellular adenosine in the heart have been exploited therapeutically for the treatment of supraventricular tachycardia, but it has been shown to be important in diverse body systems: vascular, neurological, visual, renal, endocrine and pulmonary systems, and also in cancer and inflammation [17,18].
Adenosine acts via four cell surface G-protein-coupled P1 purinergic receptors (GPCR): A1, A2A, A2B and A3. Adenosine receptors work through inhibition or stimulation of adenylyl-cyclase to decrease or increase intracellular cyclic adenosine monophosphate (cAMP) levels: A1 and A3 receptors provoke an inhibitory regulative G-protein (Gi)-mediated decrease in cAMP levels, whereas stimulative regulative G-protein (Gs) receptors of the A2 family increase cAMP concentrations. Other signalling pathways have been associated with A1 and A3 receptors . Adenosine might also enter cells by passive and active transporter channels and exert its effects inside the cells independently of receptor engagement .
In the immune system, extracellular adenosine was shown to act on diverse immune cells to mediate mainly anti-inflammatory effects . The adenosine receptor A2A was shown to be critical to the control of inflammation in several in-vivo models , and a large body of evidence exists on the effects of adenosine on lymphocyte functions. Acting via the A2A receptor, extracellular adenosine can prevent activation and proliferation of CD4+ T cells  and has been proposed to inhibit T helper type 1 (Th1)  and Th17 responses in vivo, while also promoting the induction of Treg. Adenosine also mediates anti-inflammatory effects on dendritic cells, neutrophils, macrophages and other immune cells [25,26]. Further evidence of the importance of adenosine in the immune system arises from patients with inherited deficiency of adenosine deaminase (ADA). ADA is mainly a cytosolic enzyme, but it also operates at the cell surface (ecto-ADA). Ecto-ADA irreversibly deaminates extracellular adenosine to inosine, and in humans (but not mice) is associated with CD26 on the cell surface . Patients with ADA deficiency have increased extracellular levels of adenosine and 2′-deoxyadenosine, and present with severe lymphopenia and immunodeficiency that can lead to death early in life, highlighting the anti-inflammatory role of adenosine (among other pathological phenotypes where adenosine is thought to play an important role, e.g. within renal, neural, skeletal and pulmonary systems) .
ATP exists in high concentrations (3–10 mM)  inside cells where it participates in energy metabolism. In contrast, ATP concentrations are very low (104-fold lower than inside the cells) in the extracellular space. When cell damage occurs, for instance during tissue injury or inflammation, ATP is released to the extracellular space. Furthermore, ATP can be released physiologically by different cell types to mediate autocrine and paracrine effects. Extracellular ATP is sensed by two subfamilies of P2 purinergic receptors, P2X and P2Y [25,29], and mediates several physiological functions on both adaptive and innate cells of the immune system to evoke mainly proinflammatory effects, although in certain circumstances functional inhibition may be seen . ATP released by activated lymphocytes seems to operate in an autocrine manner, to prolong activation and IL-2 secretion . Interestingly, FoxP3+ Treg were shown to release lower amounts of proinflammatory ATP than effector T cells. Treg express the ATP receptor P2X7 and ATP can interfere with their suppressive capacity and phenotypic stability . ATP may also be important during T cell differentiation, as the inhibition of the P2X receptor during activation of naive CD4+ T cells could suppress effector T cell differentiation and promote iTreg induction . ATP can also promote Th17 polarization in the lamina propria . The general effects of adenosine and ATP in immune cells have been reviewed by Bours  and Kumar , and the autocrine effects were reviewed recently by Junger .
CD39 and CD73 convert ATP to adenosine
CD39 and CD73 are two cell surface ectoenzymes that dephosphorylate ATP to produce adenosine, thus controlling adenosine and ATP levels in the extracellular space. CD39, also known as ecto-apyrase, degrades extracellular ATP to AMP, thereby providing the substrate for CD73 conversion of AMP to adenosine. CD73, also known as ecto-5′-nucleotidase, is a glycosyl phosphatidylinositol (GPI)-linked, membrane-bound glycoprotein with enzymatic capacity to catalyze the dephosphorylation of extracellular nucleoside monophosphates into nucleosides, with 5′-AMP as the preferred substrate generating adenosine. CD39 and CD73 actions reduce inflammation by removing ATP and also by generating anti-inflammatory adenosine . Figure 1 illustrates some of the pathways involved in the ATP/adenosine production, degradation and signalling.
In the mouse and human immune systems CD39 expression has been described in several cell types, such as neutrophils, monocytes and B cells, and is present in the surface of subsets of T cells [36,37]. CD39 expression and activity in human lymphocytes was described by Pulte et al. . CD73 is expressed in diverse mouse cells . In both human and mouse immune systems, CD73 is expressed in subsets of T cells, myeloid cells, bone marrow stromal cells, thymic epithelial cells and human B cells .
Expression of CD39/CD73 and adenosine generation is a feature of Tregs
Recently, the conversion of ATP into adenosine by CD39 and CD73 has been suggested as one of the multiple mechanisms employed by natural regulatory T cells (nTreg) in mediating suppression. Analysing the transcriptome of murine FoxP3+ cells, Gavin et al. noticed an increased expression of CD39 and CD73 mRNA . At the cell membrane protein level, murine FoxP3 nTreg were shown to express high levels of CD39 and CD73 [41–43], and these ecto-enzymes were claimed to play a role in the suppression by these cells, as Treg lacking one of the ectoenzymes appeared to show impaired suppressive capacity in vivo. Further characterization of CD39 expression on T cells was obtained by Zhou et al.: CD39+ cells could be separated into two groups, one comprising Treg expressing FoxP3 and also CD73, while the other exhibited a memory phenotype (FoxP3-CD39+CD44+CD62L-CD25-CD73dim/-), and showed no regulatory capacity, and even rejected allografts faster than FoxP3-CD39- T cells . Also, CD73 expression on CD4+ T cells was not exclusive to Treg, with CD73 expressed on uncommitted primed precursor Th cells . Our own studies have confirmed enhanced expression of CD39 and CD73 on FoxP3+ T cells and shown that neither is a specific marker for FoxP3+ T cells, as other T cell subsets and B cells express them .
In humans, the expression of CD39 is not homogeneous within the FoxP3+ population: Mandapathil and colleagues found that ∼80% of human FoxP3+ Tregs were CD39+ and others have proposed that CD39 expression is restricted to Treg with a memory CD45RO+CCR6+ phenotype . Fletcher et al.  suggested that human Treg, defined as CD4+CD25high, could be divided into two groups based on the expression of CD39. Both CD39+ and CD39- could suppress Th1 responses, but only the CD39+ prevented IL-17 production by Th17. CD39+ FoxP3- cells were found to produce interferon (IFN)-γ and IL-17 and to be proinflammatory despite a high ATP hydrolytic capacity , and CD39 was seen to be up-regulated in the inflamed joint of juvenile idiopathic arthritis patients compared to its expression in peripheral blood.
CD73 was found to be enriched intracellularly in human FoxP3+ Treg compared to effector T cells, but surface expression of CD73 was present only in 1–7% of human Treg[46,49]. Interestingly, CD26, the ‘anchor’ surface protein for ADA, was found to be expressed modestly in nTreg, but expressed highly on effector cells . Using an in-vitro mixed lymphocyte assay, it was further shown that inhibition of CD73 or A2A receptor signalling reduced the suppressive capacity of human Treg. Conversely, the inhibition of ADA on effector cells enhanced suppression by Treg. Together, these data suggest that effector cells are incapable of producing adenosine from ATP and, in addition, are effective at removing the anti-inflammatory adenosine via CD26-bound ADA. CD39+CD73+CD26- Treg have the opposite ability, which is to be able to increase adenosine generation. Treg expression of CD39/CD73 to remove ATP and generate adenosine might be important not only for suppression of damaging functions of other cells, but also for autocrine protection of Treg against ATP effects in inflammatory environments, as ATP is able to interfere with their suppressive capacity .
CD73 expression is induced by TGF-β and modulated by proinflammatory cytokines
Despite several reports characterizing CD39 and CD73 expression on naive lymphocytes, and an early report showing that CD73 could be induced specifically by CD38 cross-linking in a variety of human lymphocytes and cell lines , the control of ecto-nucleotidase expression upon activation is incompletely understood. Experiments in our laboratory examined the expression of CD39 and CD73 on murine lymphocytes and other immune cells, and how these are modulated by the tissue microenvironment in which the cells are activated, in particular the cytokine milieu .
We observed that CD39 is up-regulated in all murine CD4+ T cells upon activation. This is consistent with the previously described observations that CD39 is expressed highly on populations with memory phenotypes [44,48]. In contrast, we found that naive T cells that express CD73 down-regulate it upon activation. This might be necessary to prevent autocrine/paracrine suppressive adenosine generation during T cell activation, so permitting progression of the lymphocyte response.
We also found that CD73 expression could be modulated by the cytokine milieu present on activation and, in particular, by TGF-β: TGF-β (but not other cytokines) prevented activation-induced down-regulation of CD73 on CD4+ T cells and strongly up-regulated CD73 expression (both at mRNA and cell surface protein expression levels). TGF-β treatment provided the cells with the capacity to produce adenosine from AMP, and to suppress in-vitro proliferation of responding T cells in a CD73-dependent manner.
TGF-β can induce FoxP3 expression and a regulatory phenotype [51,52], and these iTreg express CD73 . However, FoxP3 was not essential for CD73 induction by TGF-β, as FoxP3−/− T cells normally up-regulated CD73 after TGF-β treatment. TGF-β could also increase the expression of CD73 on other immune cells such as CD8+ T cells, macrophages and dendritic cells, providing a possible amplification mechanism for adenosine production. This TGF-β-induced expression of CD73 on CD4+ T cells was counteracted by inflammatory cytokines such as IL-4, IL-12, IL-21 and IFN-γ. This suggests that inflammatory and anti-inflammatory cytokine microenvironments are important determinants of CD73 expression and, therefore, of adenosine levels in the tissues.
Our findings are somewhat reinforced by the recently published data showing that in-vitro TGF-β-induced regulatory T cells (iTregs) from non-obese diabetic (NOD) mice (but not ex-vivo nTreg) had an impaired capacity to suppress T cell proliferation in vitro when compared to similarly produced iTregs from control wild-type mice . Microarray and quantitative polymerase chain reaction (qPCR) comparison between NOD iTregs and wild-type iTregs revealed that CD73, among other genes, was down-regulated in NOD iTregs. Whether ATP release and this incapacity to up-regulate CD73 in iTregs contributes to the development of diabetes in NOD mice is not yet established.
A role for adenosine in tumour immune evasion?
Adenosine generation could offer a route by which certain tumours evade host immunity. CD73 has been detected in a number of malignant cell types, and associated with poorer outcomes in breast cancer patients . CD73 expression and accumulation of adenosine have been connected to cancer privilege , and mice with a genetic deletion of the A2A adenosine receptor appear to reject melanoma and lymphoma malignancies more efficiently than the wild type, in a mechanism mediated by T cells . More recently, the expression of CD73 and adenosine production has been implicated directly in tumour immune evasion: in animal models, shRNA knock-down of CD73 in tumour cells was necessary for efficient anti-tumour T cell immunotherapy , and anti-CD73 antibody therapy reduced tumour growth and metastasis in processes dependent on adaptive immunity . Using CD73-deficient mice, a role for adenosine generation has been proposed in perpetuating tumour growth in melanoma models , and the expression of CD39/CD73 and adenosine receptors has been suggested as a mechanism of autocrine suppression of apoptosis in chronic lymphocytic leukaemia . Ablation of CD39 expression in both the vasculature or leucocytes can inhibit the growth of melanoma hepatic metastasis , and CD73 expression on tumour cells and non-tumour cells was found to contribute to tumour progression, although in distinct processes: non-haematopoietic expression of CD73 was shown to be important in controlling T cell homing to the tumour, while CD73 on leucocytes limited systemic anti-tumour T cell proliferation and effector function .
Exosomes are 30–100-nm membrane vesicles containing a multitude of cellular molecules, such as cytosolic and membrane proteins. These vesicles are released by most cell types, including immune cells and tumour cells, and are capable of immune regulation [63,64]. Exosomes produced from various cancer cells and mesothelioma effusions were shown to express CD39/CD73 and convert ATP to adenosine, and to suppress T cell activation .
In human cancers, higher serum levels of TGF-β have been associated with disease progression and poorer prognosis in gastric, breast, colon and lung cancer . Many tumours secrete TGF-β, apparently promoting later-stage tumour growth and angiogenesis, while blockade of TGF-β signalling on T cells has been shown to lead to eradication of tumours . It may well be that CD73 expression and adenosine production in tumours may, at least in part, be explained by TGF-β production by the cancerous cells or cancer-associated cells. Whatever the cell source of adenosine generation in tumours, this might be supplemented by that generated through infiltration of Treg, and conversion of naive T cells to iTreg in the tumour microenvironment.
The co-ordinated action of CD39/CD73 results in the rapid conversion of extracellular ATP to adenosine, a mechanism with potentially profound immunological consequences: adenosine mediates mainly anti-inflammatory effects, while ATP in the extracellular space acts as a strong proinflammatory stimulus, acting, among others, on T cells, dendritic cells and neutrophils. Several reports have characterized the expression of CD39 and CD73 in murine and human leucocyte populations. However, ecto-nucleotidase-based regulation in the resting and activated states of immune and non-immune cells currently remains incompletely understood. Our results with murine T cells highlight a co-ordinating role for TGF-β and the cytokine microenvironment in ecto-nucleotidase expression. Whether CD73 expression in human T cells is also subject to similar cytokine control is unknown, and merits a fuller analysis. Human Treg were shown to express low levels of cell-surface CD73, but possess high levels of CD73 intracellularly and therefore have the potential to up-regulate surface ectonucleotidase activity in the appropriate circumstances.
We propose that generation of anti-inflammatory adenosine is an additional mechanism by which TGF-β contributes to the creation of privileged microenvironments within tissues. Several of the already known effects of TGF-β might be mediated by CD73 and adenosine. The importance of TGF-β in immune regulation is already well established. With our new understanding, it will be important to establish whether adenosine has any role in the induction of Treg and infectious tolerance.
With regard to our novel finding that FoxP3 induction is controlled by mTOR , one can hypothesize that extracellular adenosine might feed into this central fate pathway. There are two ways in which extracellular adenosine could lead to mTOR inhibition, both acting by increasing the levels of AMP within the cell. First, at higher concentrations (≥1 mM) of adenosine there would be an uptake directly by the equilibrative adenosine transporters . Adenosine, once within the cell, is converted immediately to AMP by adenosine kinase which then activates AMP kinase to phosphorylate and inhibit mTORC1 both directly and indirectly via TSC1/TSC2 and Rheb. Secondly, the A2A/B adenosine receptors, which work by activating or inhibiting adenylate cyclase to convert ATP to cAMP, tend to work at micromolar concentrations. This pathway could also generate increased levels of AMP within the cell, because the level of cyclic AMP is dependent on the rate of generation by adenylate cyclase from ATP and the breakdown by specific phosphodiesterases that convert the cAMP into AMP. This AMP could then also be sensed by the mTOR inhibition via AMP kinase, as above. AMP kinase is also stimulated by the drug metformin, used in type II diabetics, and the actions of adenosine on mTOR might be mimicked by metformin treatment.
One should also bear in mind that the effects of adenosine in the immune system might be regulated at stages other than its production by CD39/CD73. It may well be that distribution of adenosine receptors is a key determinant of outcome, as the sensitivity of the cells to ATP and adenosine depends on the expression and distribution of receptors among the immune system cells [17,18,68,69]. The influences of cell activation and cytokine microenvironment on the distribution of adenosine receptors (P1) and ATP receptors (P2), and consequently sensitivity to adenosine and ATP, remain largely unexplored. A further step in regulation might occur at the degradation and removal of adenosine from the microenvironment. Extracellular adenosine might be degraded to inosine by ecto-ADA, and this enzyme has already been shown capable of in-vivo regulation of adenosine levels . In humans, ecto-ADA exists on the surface of cells linked to CD26 and the regulation of CD26 expression might determine the capacity of the cells to bind ADA released to the extracellular space. Human activated CD4+ T cells seem to up-regulate CD26 (and ADA) for adenosine removal to allow immune responses to occur. Conversely, in order to achieve high levels of adenosine and immune suppression, the expression of ecto-ADA and adenosine degradation might need to be tightly controlled. In short, the local cytokine environment may not only impact adenosine production, but also tailor the expression of ecto-ADA for adenosine removal. In light of our data on TGF-β promotion of adenosine generation, it will be interesting to establish how this and other cytokines impact the degradation pathways of adenosine, and the expression of adenosine receptors.