ATP and cancer immunosurveillance

Abstract

While intracellular adenosine triphosphate (ATP) occupies a key position in the bioenergetic metabolism of all the cellular compartments that form the tumor microenvironment (TME), extracellular ATP operates as a potent signal transducer. The net effects of purinergic signaling on the biology of the TME depend not only on the specific receptors and cell types involved, but also on the activation status of cis‐ and trans‐regulatory circuitries. As an additional layer of complexity, extracellular ATP is rapidly catabolized by ectonucleotidases, culminating in the accumulation of metabolites that mediate distinct biological effects. Here, we discuss the molecular and cellular mechanisms through which ATP and its degradation products influence cancer immunosurveillance, with a focus on therapeutically targetable circuitries.

This review summarizes our current view on the dual roles of signaling via adenosine triphosphate and its degradation products in the tumor microenvironment, affecting intracellular metabolism as well as crosstalk between malignant and immune cells.

Introduction

According to a widely accepted model, malignant transformation is initiated by relatively common genetic or epigenetic alterations that incapacitate tumor‐suppressing mechanisms (for the most part, mechanisms that ensure the preservation of cellular homeostasis) as they activate oncogenic drivers (generally, processes that enable accrued anabolism in support of hyperproliferation) (Hanahan & Weinberg, 2011; Timp & Feinberg, 2013). The vast majority of newly formed malignant cells, however, appears to be controlled by the host immune system prior to forming symptomatic tumors (Vesely & Schreiber, 2013; Lopez‐Otin & Kroemer, 2021). While in most cases such control involves the definitive “eradication” of malignant cell precursors, in some instances newly formed cancer cells can resist immune attacks and generate a rudimentary tumor microenvironment that enables some degree of proliferation, a dynamic battle between emerging tumors and their host commonly referred to as “equilibrium” (Vesely & Schreiber, 2013). In this context, malignant cells can acquire additional genetic and epigenetic alterations that either (i) impair their ability to initiate anticancer immune responses, such as the loss or downregulation of genes coding for endogenous danger signals, (ii) limit their visibility to immune effector cells, such as the loss of MHC class I‐coding genes or beta‐2‐microglobulin (B2M), (iii) increase their resistance to immune effector molecules, such as the loss of caspase 8 (CASP8), or (iv) establish a state of local immunosuppression, such as the upregulation of CD274 (best known as PD‐L1) (Galluzzi et al, 2018). In this context, the equilibrium between cancer cells and the host immune system ceases to exist in favor of an “escape” phase culminating in uncontrolled tumor growth and metastatic dissemination (Dunn et al, 2002; Dersh et al, 2021).

Importantly, the TME of malignancies that escaped immunosurveillance (Rao et al, 2019) undergoes a considerable reconfiguration, generally involving the accumulation of immunosuppressive myeloid and lymphoid cells including M2‐like tumor‐associated macrophages (TAMs) and CD4⁺CD25⁺FOXP3⁺ regulatory T (T_REG) cells at the expense of M1‐like TAMs, type I conventional dendritic cells (cDC1s), T_H1 CD4⁺ T cells, CD8⁺ cytotoxic T lymphocytes (CTLs), and natural killer (NK) cells, all of which promote tumor‐targeting immunity (Talmadge & Gabrilovich, 2013; Mantovani et al, 2017; Lee & Radford, 2019; Sprooten et al, 2019; Togashi et al, 2019). Beyond such a general trend, however, the precise immune contexture of each neoplasm exhibits considerable heterogeneity (De Sousa et al, 2013; Vitale et al, 2021) and has a major impact on disease course and response to therapy (Fridman et al, 2017). Indeed, it has now become clear that the efficacy of most anticancer agents commonly employed in the clinic, encompassing cytotoxic chemotherapeutics, radiation therapy (RT), and targeted anticancer agents, relies at least partially on the (re)activation of immunosurveillance (Galluzzi et al, 2020; Rodriguez‐Ruiz et al, 2020; Petroni et al, 2021). In line with this notion, considerable efforts are being dedicated to the identification of clinically relevant approaches to alter the TME in favor of treatment efficacy, especially for tumors that exhibit rather scarce infiltration by immune effector cells, such as luminal breast cancer and pancreatic carcinoma (Kroemer et al, 2015; Ho et al, 2020).

All cellular components of the TME including malignant and immune cells engage in a dynamic competition for nutrients, oxygen, and trophic signals (all of which are generally scarce as a consequence of relatively poor vascularization) (Martinez‐Outschoorn et al, 2017; O'Sullivan et al, 2019; Garner & de Visser, 2020). Moreover, the availability of nutrients, oxygen, and trophic signals is not equal across all tumor regions and is not constant over time (e.g., before and after therapy), hence constituting a major driver of intratumoral heterogeneity (ITH) (De Sousa et al, 2013; Vitale et al, 2021). Indeed, such restrictions de facto operate as Darwinian pressures, fostering the selection of cells with an accrued capacity to harness alternative carbon sources (e.g., lactate, extracellular amino acids) for catabolic and anabolic reactions in support of proliferation and tolerance to hypoxia (Chang et al, 2015; Xiao et al, 2019).

Adenosine triphosphate (ATP) occupies a key position in the overall configuration of the TME. On one hand, intracellular ATP is crucial for each cellular TME component to survive and mediate its functions (including proliferation, for malignant and non‐terminally differentiated immune cells) (Leone & Powell, 2020; Bergers & Fendt, 2021). On the other hand, the pool of ATP that accesses the TME upon active secretion by living or dying cells into the extracellular space constitutes a major signal transducer (Di Virgilio et al, 2018). The net effect of ATP signaling on the TME, however, depends on multiple factors, including the presence of extracellular ATP‐degrading enzymes as well as the expression pattern of purinergic receptors. Here, we will critically discuss the molecular and cellular mechanisms through which extracellular ATP and its degradation products influence the crosstalk between malignant and immune cells and present recent advances on the purinergic system as a potential target for the development of novel anticancer interventions.

Extracellular ATP homeostasis in the TME

Since ATP cannot be synthesized in the extracellular milieu, the microenvironmental levels of ATP are entirely controlled by the balance between its secretion/release and degradation (Fig 1).

Figure 1. Extracellular ATP homeostasis in the tumor microenvironment.

The concentration of extracellular ATP in the tumor microenvironment is determined by the balance between ATP release and degradation. A variety of cells release ATP either as part of their physiological state or as they respond to stress and potentially die, including cancer cells, dendritic cells (DCs), tumor‐infiltrating neutrophils (TINs), tumor‐associated macrophages (TAMs), and platelets. Extracellular ATP is catabolized by the sequential activity of ectonucleoside triphosphate diphosphohydrolase 1 (ENTPD1, best known as CD39), which converts ATP into ADP and AMP, and 5'‐nucleotidase ecto (NT5E, best known as CD73), which converts AMP into adenosine (ADO). CD39 and CD73 are expressed by multiple cell type that populate the tumor microenvironment, including some malignant cells, cancer‐associated fibroblasts (CAFs), exhausted cytotoxic T lymphocytes (CTLs), regulatory T (T_REG) cells, an immunosuppressive subset of natural killer (NK) cells, M2‐like TAMs, and myeloid‐derived suppressor cells (MDSCs). Hypoxia is a major driver of ATP degradation in the tumor microenvironment.

Adenosine triphosphate secretion is an active, regulated process that can occur via multiple mechanisms and involve different cellular sources. Molecular mechanisms for ATP secretion encompass: (i) exocytosis of ATP‐containing vesicles, a process that may (but does not necessarily) involve cell death (Imura et al, 2013; Martins et al, 2014) and mechanistically relies on vesicular loading by solute carrier family 17 member 9 (SLC17A9) (Imura et al, 2013; Cao et al, 2014) and the SNAP receptor (SNARE)‐ and Rho‐associated, coiled‐coil containing protein kinase 1 (ROCK1)‐dependent fusion of exocytosis‐competent ATP‐rich vesicles with the plasma membrane (Imura et al, 2013; Martins et al, 2014); (ii) liberation of cytosolic ATP molecules via gap junction protein alpha 1 (GJA1, best known as CX43) hemichannels at gap junctions (Stout et al, 2002; Eltzschig et al, 2006; Kang et al, 2008); and (iii) gradient‐driven efflux via oligomeric pannexin 1 (PANX1) channels (also known as “pannexons”) (Dahl, 2015), which are sensitive to activation by mechanical forces (Bao et al, 2004), by the pro‐inflammatory CASP1 (Narahari et al, 2021), and by apoptotic caspases such as CASP3 and CASP7 (Chekeni et al, 2010; Medina et al, 2020). That said, while both vesicular ATP secretion and PANX1‐dependent release have been documented in living and dying malignant cells (Martins et al, 2014; Martin et al, 2017), CX43 hemichannels appear to be mostly operational in non‐malignant cells of the TME, including astrocytes (Stout et al, 2002), as well as (at least potentially) neutrophils (Dosch et al, 2018) and macrophages (Dosch et al, 2019). Indeed, while both neutrophils and macrophages have been shown to release ATP via CX43 hemichannels in non‐oncological settings, whether such function is preserved in TAMs and tumor‐infiltrating neutrophils (TINs) remains to be elucidated. Along similar lines, platelets are known as major sources of extracellular ATP upon degranulation (Yeaman, 2014), but their contribution to extracellular ATP availability in the TME has just begun to emerge (Schumacher et al, 2013; Gaertner & Massberg, 2019). Additional cellular compartments that may secrete ATP in the TME encompass (at least in some settings) endothelial cells (Sáez et al, 2018; Yang et al, 2020d), fibroblasts (Pinheiro et al, 2013; Murata et al, 2014), dendritic cells (DCs) (Tappe et al, 2018; Martinek et al, 2019), and activated CTLs (Tokunaga et al, 2010). Importantly, while some cells spontaneously secrete at least some ATP in their physiological status, for the most part, ATP is actively released in the context of adaptive responses to microenvironmental perturbations, including mechanical cues (Bao et al, 2004), inflammatory signals (Beckel et al, 2018), hypoxia (Lim To et al, 2015), and exposure to a variety of cancer therapeutics (Michaud et al, 2011; Tatsuno et al, 2019; Rodriguez‐Ruiz et al, 2020). In most such instances, abundant ATP secretion by stressed cells (which is key for extracellular ATP to mediate immunostimulatory effects, see below) involves functional autophagic responses (Michaud et al, 2011), potentially linked to the ability of autophagy to preserve intracellular ATP pools during stress (Rybstein et al, 2018; Anderson & Macleod, 2019). Consistent with this notion, genetic and pharmacological interventions aimed at blocking or boosting autophagic responses in cancer cells have been consistently associated with reduced and increased ATP secretion, respectively, in response to immunogenic chemotherapy (Michaud et al, 2011; Pietrocola et al, 2016; Chen et al, 2019; Kepp & Kroemer, 2020; Wang et al, 2020). Obviously, all dying cells abruptly release their cytosolic ATP pool when they undergo plasma membrane permeabilization (PMP) as the final step of cellular demise. However, while PMP itself has now been shown to be an active (rather than an osmosis‐driven) process even in the context of post‐apoptotic, secondary necrosis (Kayagaki et al, 2021), the consequent spillage of cytosolic content into the extracellular milieu remains a largely unregulated phenomenon.

Extracellular ATP is rapidly catalyzed by the sequential activity of two ectonucleotidases, that is, ectonucleoside triphosphate diphosphohydrolase 1 (ENTPD1, best known as CD39), which converts ATP into ADP and AMP, and 5'‐nucleotidase ecto (NT5E, best known as CD73), which converts AMP into adenosine as the rate‐limiting step of this enzymatic cascade (Allard et al, 2020; Moesta et al, 2020). CD39 is mostly expressed by T_REG cells (Borsellino et al, 2007), M2‐like TAMs (d'Almeida et al, 2016), and myeloid‐derived suppressor cells (MDSCs, an immature population of myeloid cells with potent immunosuppressive activity) (Li et al, 2017), as well as by specific cancer cell types, such as adult T‐cell leukemia/lymphoma cells (Nagate et al, 2021), type 16 human papillomavirus (HPV‐16)‐associated cervical carcinoma cells (de Lourdes Mora‐García et al, 2019), and ovarian carcinoma cells (Häusler et al, 2011). Moreover, CD8⁺ CTLs undergoing terminal exhaustion as a consequence of chronic antigen stimulation generally exhibit a CD39⁺ phenotype (Canale et al, 2018). Conversely, CD73 is expressed by a wide variety of malignant cells as well as by cancer‐associated fibroblasts (CAFs) (Yu et al, 2020), T_REG cells (Stagg et al, 2011), and a regulatory subset of NK cells (Neo et al, 2020). Interestingly, the whole‐body deletion of purinergic receptor P2X 7 (P2RX7), which codes for one of the main receptors of extracellular ATP, has a major impact on extracellular ATP levels in the TME of experimental P2RX7‐competent melanomas (De Marchi et al, 2019), at least in part as a consequence of altered tumor infiltration by CD39⁺ and CD73⁺ T_REG cells and decreased ATP release by TAMs (De Marchi et al, 2019). Such an effect, however, cannot be recapitulated by the pharmacological P2RX7 antagonist A740003 as a result of its mixed activity on immune cells (it fails to alter tumor infiltration by T_REG cells, decreases the abundance of intratumoral CD39⁺ and CD73⁺ effector T (T_EFF) cells, and inhibits ATP secretion by TAMs) and malignant cells (it favors ATP secretion by malignant cells) (De Marchi et al, 2019).

Of note, extracellular ATP degradation does not necessarily require the expression of CD39 and CD73 on the same cell (in cis), but can also occur efficiently when these ectonucleotidases are expressed by different cellular compartments that are in proximity to each other within the TME (Schuler et al, 2014). CD73 is abundant in T_REG cell‐derived exosomes (Smyth et al, 2013), which are highly mobile and hence further promote the overall catalytic efficiency of ATP degradation within the TME. Moreover, CD38 (also known as cyclic ADP‐ribose hydrolase) and ectonucleotide pyrophosphatase/phosphodiesterase 1 (ENPP1), which are expressed by some cancer cells and exhausted T cells, can compensate for limited CD39 activity as they catalyze the conversion of extracellular NAD⁺ into ADP ribose and AMP (Morandi et al, 2015). Finally, the expression of both CD39 and CD73 can be upregulated by hypoxia, which is relatively common in the TME of solid neoplasms, via a transcriptional mechanism that involves hypoxia‐inducible factor 1 subunit alpha (HIF1A, best known as HIF‐1α) (Giatromanolaki et al, 2020; Synnestvedt et al, 2002).

In summary, the levels of extracellular ATP in the TME are dynamically determined by the mutually opposed inputs of secretion/release vs. degradation. As the factors governing these aspects of the ATP biology exhibit a considerable degree of ITH, regional and temporal fluctuations in extracellular ATP levels are likely to play a major role in the outcome of purinergic signaling in the TME, as discussed further below.

Immunostimulation by extracellular ATP

Extracellular ATP mediates two main functions: (i) It operates as a chemotactic cue for myeloid cells, upon binding to the purinergic receptor P2Y2 (P2RY2), a metabotropic receptor (Elliott et al, 2009; Chekeni et al, 2010), and (ii) it promotes activation of the inflammasome and hence CASP1‐dependent secretion of interleukin 1 beta (IL1B) and IL18 upon binding to P2RX7, an ionotropic receptor (Perregaux et al, 2000). Importantly, both these effects are required for the optimal activation of tumor‐specific immune responses by (and hence the complete efficacy of) immunogenic chemotherapeutics such as anthracyclines and oxaliplatin, as demonstrated in P2ry2 ^−/−, P2rx7 ^−/−, Casp1 ^{−/− ‐} and Il18 ^−/− mice, as well as mice lacking a core component of the inflammasome (Nlrp3 ^−/− mice), the main IL1B receptor (Il1r1 ^−/− mice) or treated with a purinergic receptor antagonist (suramin) or an IL1B‐blocking antibody (Ghiringhelli et al, 2009; Aymeric et al, 2010; Ma et al, 2013).

In this setting, DC precursors newly recruited to the TME via ATP released by cancer cells succumbing to immunogenic cell death (ICD) not only mature upon ATP‐driven inflammasome activation and migrate to tumor‐draining lymph nodes or tertiary lymphoid structures to prime adaptive anticancer immunity, but also recruit a population of IL17‐producing γδ T cells that is critical for tumor infiltration by primed CTLs (Ma et al, 2011). In accordance with this notion, optimal anticancer immune responses (and consequent superior therapeutic efficacy) driven by immunogenic chemotherapeutics are compromised in Il17a ^−/− and Il17ra ^−/− mice (Ma et al, 2011). Intriguingly, it has recently shown that the chemotactic activity of ATP on DCs also involves P2RX7 and PANX1 (Saez et al, 2017), suggesting the existence of a feed‐forward loop whereby intracellular ATP stores may contribute to DC migratory capacity (Saez et al, 2017). Moreover, elevated levels of extracellular ATP appear to induce pyroptosis in P2RX7⁺ M2‐like TAMs, hence supporting T cell‐mediated antitumor immunity upon the depletion of immunosuppressive cells from the TME (Bidula et al, 2019).

Importantly, the net immunomodulatory effect of extracellular ATP depends on the activation of additional signaling pathways. Indeed, the PANX1‐dependent co‐release of ATP and a wide panel of metabolites including ADP, AMP, GMP, creatine, spermidine, and glycerol‐3‐phosphate (G3P) by dying cells reportedly promote the removal of cell corpses while preventing the initiation of inflammatory reactions (Medina et al, 2020; Narahari et al, 2021). Moreover, extracellular ATP can have direct tumorigenic functions. Specifically, the cancer cell‐driven release of ATP from platelets initiates a P2RY2‐dependent signaling cascade that promotes tumor extravasation and metastatic dissemination upon the opening of endothelial barriers (Schumacher et al, 2013; Chen et al, 2019; Wang et al, 2020). The autophagy‐dependent secretion of ATP by melanoma cells has been shown to promote invasiveness and resistance to the BRAF inhibitor vemurafenib, a process that requires P2RX7 expression in the malignant cell compartment (Martin et al, 2017). Similar findings have been obtained with human triple‐negative breast cancer (TNBC) MDA‐MB‐231 cells upon the ATP‐dependent activation of the transcription factor SRY‐box transcription factor 9 (SOX9) (Yang et al, 2020a). Finally, NME/NM23 nucleoside diphosphate kinase 1 (NME1, best known as NDPK‐A) and NME2 (best known as NDPK‐B) expression on extracellular vesicles from MDA‐MB‐231 cells reportedly support the formation of pulmonary metastatic niches as a consequence of extracellular ATP generation in situ and consequent activation of purinergic receptor P2Y1 (P2RY1) (Duan et al, 2021).

Consistent with the multipronged effects of extracellular ATP on the TME, a large body of clinical literature suggests that genetic or epigenetic defects affecting ATP signaling influence disease outcome in cancer patients in a context‐dependent manner (Table 1). For instance, while loss‐of‐function polymorphisms in P2RX7 (rs3751143; rs208294) have been associated with advanced stage or poor disease outcome in cohorts of patients with breast carcinoma (Ghiringhelli et al, 2009), chronic lymphocytic leukemia (CLL) (Thunberg et al, 2002; Wiley et al, 2002; Zhang et al, 2003), and papillary thyroid carcinoma (PTC) (Dardano et al, 2009), no impact on clinicopathological variables could be attributed to rs3751143 in other cohorts of subjects with CLL (Starczynski et al, 2003; Nückel et al, 2004), multiple myeloma (Paneesha et al, 2006), and PTC (Dardano et al, 2009), while increased expression levels of P2RX7 have been linked to disease progression in an independent cohort of CLL patients (Adinolfi et al, 2002). Likewise, elevated P2RY2 levels have been associated with gastric malignant transformation (Aquea et al, 2014). A variety of immunohistochemical and transcriptional signatures of proficient autophagic responses have been linked to worsened disease outcome in cohorts of breast (Yamazaki et al, 2020), gastric (Kim et al, 2019; Wang et al, 2021), pancreatic (Ko et al, 2013; Cui et al, 2019), and head and neck (Jiang et al, 2012) cancer patients, while the contrary held true (or there was no impact on clinicopathological variables) in independent series of patients with breast (Ladoire et al, 2015; Tang et al, 2015; Ladoire et al, 2016), ovarian (Chen et al, 2020c), hepatocellular (Lee et al, 2013; Qin et al, 2018), gastric (Wang et al, 2021), colorectal (Li et al, 2020), and salivary gland (Li et al, 2019a) carcinoma.

Table 1.

Pathophysiological relevance of extracellular ATP signaling in human cancer.

Cancer	No. patients	Variable	Technology	Impact	References
Breast cancer	225	P2RX7 rs3751143	SNP analysis	Metastatic dissemination Decreased OS	Ghiringhelli et al (2009)
	1,067 1,992	BECN1	Gene expression profiling	Improved disease outcome	Tang et al (2015)
	152	MAP1LC3B	IHC	Improved MFS Improved OS	Ladoire et al (2016)
	152 1,646	MAP1LC3B	IHC	Improved PFS	Ladoire et al (2015)
CLL	36	P2RX7 rs3751143	SNP analysis	Disease stage	Wiley et al (2002)
	144	P2RX7 rs3751143	SNP analysis	Marginally decreased OS	Zhang et al (2003)
	170	P2RX7 rs3751143	SNP analysis	Decreased OS	Thunberg et al (2002)
	111	P2RX7 rs3751143	SNP analysis	No correlation with clinical outcome	Nückel et al (2004)
	121	P2RX7 rs3751143	SNP analysis	No correlation with clinical outcome	Starczynski et al (2003)
	21	P2RX7	Immunoblotting	Disease progression	Adinolfi et al (2002)
Colorectal cancer	2,297	MAP1LC3B	Gene expression profiling	Increased OS	Li et al (2020)
Gastric cancer	14	P2RY2	Gene expression profiling	Disease	Aquea et al (2014)
	354	MAP1LC3C	Gene expression profiling	Improved OS	Wang et al (2021)
	354	ATG4D	Gene expression profiling	Decreased OS	Wang et al (2021)
	402	MAP1LC3B and SQSTM1	IHC, immunoblotting a,nd RT–PCR	Decreased OS	Kim et al (2019)
Head and neck cancer	79	MAP1LC3B	IHC	Disease stage	Jiang et al (2012)
Hepatocellular carcinoma	190	MAP1LC3B	IHC	Improved OS	Lee et al (2013)
	1,086	BECN1	Gene expression profiling	Improved OS	Qin et al (2018)
Multiple myeloma	136	P2RX7 rs3751143	SNP analysis	No correlation with clinical outcome	Paneesha et al (2006)
Ovarian cancer	1,497	BECN1	Gene expression profiling	Improved OS and PFS	Chen et al (2020c)
	1,497	MAP1LC3B	Gene expression profiling	No correlation with clinical outcome	Chen et al (2020c)
Pancreatic cancer	73	BECN1	IHC	Disease progression	Ko et al (2013)
	86	BECN1 and MAP1LC3B	IHC and RT–PCR	Metastatic dissemination Disease stage Decreased OS	Cui et al (2019)
Papillary thyroid cancer	121	P2RX7 rs3751143	SNP analysis	Disease type	Dardano et al (2009)
	121	P2RX7 rs208294	SNP analysis	No effect on disease stage	Dardano et al (2009)
Salivary gland carcinoma	48	BECN1 and MAP1LC3B	IHC	Low disease stage	Li et al (2019a)

CLL, chronic lymphocytic leukemia; IHC, immunohistochemistry; MFS, metastasis‐free survival; N/A, not available; OS, overall survival; PFS, progression‐free survival; SNP, single nucleotide polymorphism.

In summary, while an abundant preclinical literature mechanistically implicates ATP secretion by stressed and dying cells in the initiation of anticancer immune responses (Fig 2), additional, hitherto poorly characterized factors and mechanisms appear to influence the net effect of ATP signaling in the TME.

Figure 2. Immunostimulation by extracellular ATP.

Extracellular ATP mediates chemotactic effects on myeloid cells, upon binding to purinergic receptor P2Y2 (P2RY2), and promotes the inflammasome‐dependent secretion of interleukin 1 beta (IL1B) and IL18 upon binding to purinergic receptor P2X 7 (P2RX7). These effects are critical for mature dendritic cells (DCs) to prime cytotoxic T lymphocytes (CTLs) against tumor‐derived antigens and hence initiate adaptive anticancer immunity. Alongside, extracellular ATP triggers pyroptosis in tumor‐associated macrophages (TAMs), hence depleting the tumor microenvironment of generally immunosuppressive cells along with the emission of pyroptosis‐dependent immunostimulatory signals.

Immunosuppression by extracellular ATP metabolites

Extracellular ATP degradation by the sequential activity of CD39 and CD73 mediates immunosuppressive effects not only as a consequence of limited ATP‐dependent immunostimulation (see above), but also due to the generation of adenosine, which per se promotes tumor progression via immunological and non‐immunological mechanisms (Allard et al, 2020). Consistent with this notion, transgene‐driven overexpression of CD39 by malignant cells has been associated with impaired anticancer immunity driven by immunogenic chemotherapy (and hence poor disease outcome) in syngeneic immunocompetent mouse models of fibrosarcoma (Michaud et al, 2011; Pietrocola et al, 2016). Similarly, experimental CD73 overexpression in human cervical carcinoma as well as human and mouse breast carcinoma cells has been shown to promote invasiveness and metastatic potential (Zhou et al, 2007; Gao et al, 2017), at least in part as a consequence of autocrine–paracrine adenosine signaling via adenosine A2b receptor (ADORA2B) and ADORA2B‐driven neovascularization (Stagg et al, 2010; Mittal et al, 2016; Ludwig et al, 2020). Together with ADORA2A, ADORA2B is indeed the main receptor for extracellular adenosine signaling in the TME (Allard et al, 2020). However, while ADORA2B is expressed by various malignant cell types including breast cancer (Stagg et al, 2010; Lan et al, 2018), cervical cancer (Torres‐Pineda et al, 2020), and melanoma (Mittal et al, 2016) cells, as well as by endothelial cells (Ludwig et al, 2020), DCs (Chen et al, 2020b), and M2‐like TAMs (Cohen et al, 2015), ADORA2A expression is mostly restricted to myeloid cells (Nakamura et al, 2020), NK cells (Young et al, 2018), CTLs (Kjaergaard et al, 2018; Shi et al, 2019), and lymphatic endothelial cell precursors (Allard et al, 2019).

From a mechanistic standpoint, ADORA2A‐ and ADORAB‐driven immunosuppression results from the activation of intracellular cyclic AMP (cAMP) signaling, the ultimate functional outcome of which depends on the specific immune cell type expressing these receptors. Thus, cAMP signaling directly inhibits TCR activation and cytokine production in CTLs (Lappas et al, 2005), promotes forkhead box P3 (FOXP3) synthesis and the upregulation of the co‐inhibitory receptor cytotoxic T lymphocyte‐associated protein 4 (CTLA4) in T_REG cells (Klein & Bopp, 2016), impairs NK cell effector and secretory functions (Raskovalova et al, 2005), inhibits NF‐κB‐dependent inflammatory responses (Minguet et al, 2005) and drives the secretion of immunosuppressive cytokines such as IL10 from TAMs and MDSCs (Németh et al, 2005; Cekic et al, 2014), and perturbs the maturation of DCs (Kayhan et al, 2019). Of note, both ADORA2A and ADORA2B can be upregulated by hypoxia (Sitkovsky et al, 2014; Lan et al, 2018), in thus far resembling CD39 and CD73. Thus, particularly hypoxic regions of the TME are expected to exhibit robust adenosinergic signaling and hence (i) potent immunosuppression, (ii) accrued neovascularization, (iii) increased vascular permeability, and (iv) enhanced cancer cell motility, de facto representing ideal niches for metastatic dissemination. This is further aggravated by immunometabolic circuitries driven by hypoxia. Specifically, hypoxic tumor regions are rich in glycolytic lactate, which mediates multipronged immunosuppressive effects involving CTLs (Brand et al, 2016), NK cells (Husain et al, 2013), T_REG cells (Watson et al, 2021), and MDSCs (Yang et al, 2020c).

Consistent with the potent immunosuppressive effects of adenosine signaling in the TME, an abundant clinical literature links elevated expression levels of adenosine‐generating enzymes or adenosine receptors to poor disease progression in cohorts of patients with various tumors (Table 2). Thus, high levels of CD39 have been associated with advanced grade or poor disease outcome in multiple cohorts of patients with CLL (Pulte et al, 2011), renal cell carcinoma (Wu et al, 2020a), endometrial tumors (Aliagas et al, 2014), pancreatic carcinoma (Künzli et al, 2007), and non‐small cell lung cancer (NSCLC; Li et al, 2017). Similar clinical findings have been correlated with various single nucleotide polymorphisms affecting Entpd1 (i.e., rs10748643, rs11188513, rs2226163) in patients with colorectal carcinoma (Tokunaga et al, 2019; Gallerano et al, 2020), but the functional impact of these variants on CD39 functions remains unclear. Moreover, abundant tumor‐infiltrating or circulating levels CD4⁺ or CD8⁺ cells expressing CD39 have been linked to disease progression or resistance to therapy in various cohorts of individuals with CLL (Perry et al, 2012), colorectal carcinoma (Gallerano et al, 2020), head and neck squamous cell carcinoma (Gallerano et al, 2020), pancreatic cancer (Gallerano et al, 2020), NSCLC (Koh et al, 2020), and renal cell carcinoma (Qi et al, 2020). Along these lines, intratumoral or circulating biomarkers of CD73 proficiency (including expression levels and enzymatic activity) have been correlated with poor disease outcome in patients with diffuse large B‐cell lymphoma (Wang et al, 2019), ovarian cancer (Turcotte et al, 2015), pancreatic carcinoma (Chen et al, 2020a; Tahkola et al, 2021), NSCLC (Li et al, 2017), renal cell carcinoma (Tripathi et al, 2020), breast cancer (Loi et al, 2013; Buisseret et al, 2018), glioma (Xu et al, 2013), colorectal carcinoma (Messaoudi et al, 2020), and metastatic melanoma (Turiello et al, 2020). The Nt5e polymorphism rs2229523 (of hitherto unclear functional significance) correlated with limited overall survival in patients with colorectal carcinoma (Tokunaga et al, 2019), as did tumor infiltration by CD8⁺CD73⁺ cells in prostate cancer patients (Leclerc et al, 2016). Finally, ADORA2A expression or signaling was linked to poor disease outcome in individuals with diffuse large B‐cell lymphoma (Wang et al, 2019), renal carcinoma (Kamai et al, 2021), and a variety of tumors from The Cancer Genome Atlas (Sidders et al, 2020), as was the ADORA2B polymorphism rs2015353 in patients with colorectal cancer (Tokunaga et al, 2019), although the functional implications of rs2015353 on ADORA2B activity remain to be elucidated.

Table 2.

Pathophysiological relevance of adenosinergic signaling in human cancer.

Cancer	No. patients	Variable	Technology	Impact	References
Breast cancer	122	CD73	IHC	Decreased OS	Buisseret et al (2018)
	>6,000	NT5E	Gene expression profiling	Reduced pCR	Loi et al (2013)
CLL	34	Circulating CD4+CD39+ and CD8+CD39+ cells	Flow cytometry	Disease stage	Pulte et al (2011)
	62	Circulating CD4+CD39+ T cells	Flow cytometry	Disease stage	Perry et al (2012)
Colorectal cancer	107	ADORA2B rs2015353	SNP analysis	Decreased OS	Tokunaga et al (2019)
	107 215	ENTPD1 rs11188513	SNP analysis	Decreased OS	Tokunaga et al (2019)
	215	ENTPD1 rs2226163	SNP analysis	Decreased OS	Tokunaga et al (2019)
	107	NT5E rs2229523	SNP analysis	Decreased OS	Tokunaga et al (2019)
	129	ENTPD1, NT5E, and ADORA2B	SNP analysis	No correlation	Tokunaga et al (2019)
	60	Circulating CD8+CD39+ T cells	Flow cytometry	Disease progression	Gallerano et al (2020)
	60	ENTPD1 rs10748643	SNP analysis	Disease progression	Gallerano et al (2020)
	215	CD73	IHC	Disease progression Decreased OS	Messaoudi et al (2020)
	193	Circulating CD73	ELISA	Decreased OS	Messaoudi et al (2020)
DLBCL	91	ADORA2A on TILs	IHC	Decreased OS	Wang et al (2019)
	91	CD73	IHC	Decreased OS	Wang et al (2019)
Endometrial tumors	29	CD39	IHC	Tumor grade	Aliagas et al (2014)
Glioma	500	NT5E	Gene expression profiling	Limited DFS	Xu et al (2013)
HNSCC	19	Circulating CD8+CD39+ T cells	Flow cytometry	Disease progression	Gallerano et al (2020)
Melanoma	546	Circulating CD73	AMPase activity	Disease progression Decreased OS	Turiello et al (2020)
NSCLC	132	Circulating CD8+CD39+ T cells	Flow cytometry	Decreased OS	Koh et al (2020)
	24	Circulating CD8+CD39+ MDSCs	Flow cytometry	Tumor infiltration by MDSCs	Li et al (2017)
Ovarian cancer	208	CD73	IHC	Decreased OS	Turcotte et al (2015)
Pancreatic cancer	28	ENTPD1	RT–PCR	Decreased OS	Künzli et al (2007)
	3	Circulating CD8+CD39+ T cells	Flow cytometry	Disease progression	Gallerano et al (2020)
	110	CD73	IHC	Decreased OS	Tahkola et al (2021)
	168	NT5E	Gene expression profiling	Disease progression Decreased OS	Chen et al (2020a)
Prostate cancer	285	Circulating CD8+CD73+ T cells	IHC	Disease progression	Leclerc et al (2016)
Renal cell carcinoma	60	ADORA2A expression	IHC	Metastatic dissemination Decreased OS	Kamai et al (2021)
	138	CD73	IHC	Decreased OS	Tripathi et al (2020)
	243	CD8+CD39+ T cells	IHC	Disease progression	Qi et al (2020)
	367	CD39	IHC and RT–PCR	Disease stage Disease progression	Wu et al (2020a)
Various tumor types	N/A	ADORA2A‐regulated gene expression	Gene expression profiling	Decreased OS	Sidders et al (2020)

CLL, chronic lymphocytic leukemia; DBLCL, diffuse large B‐cell lymphoma; DFS, disease‐free survival; ELISA, enzyme‐linked immunosorbent assay; IHC, immunohistochemistry; MDSC, myeloid‐derived suppressor cell; N/A, not available; OS, overall survival; pCR, pathological complete response; SNP, single nucleotide polymorphism; TIL, tumor‐infiltrating lymphocyte.

Of note, the discovery that the TME contains unusually high levels of extracellular ATP has prompted innovative therapeutic approaches that specifically harness this biochemical feature. For instance, a monoclonal antibody specific for the immunostimulatory receptor TNF receptor superfamily member 9 (TNFRSF9, best known as CD137) has been engineered to drive CD137 signaling only in the presence of nearly millimolar ATP levels, thus inducing a potent anticancer immune response in the absence of adverse effects due to extratumoral activity (Kamata‐Sakurai et al, 2020). This agent is currently being tested in combination with the immune checkpoint inhibitor (ICI) atezolizumab (a PD‐L1‐blocking antibody) for safety and preliminary efficacy in a phase I clinical trial enrolling patients with solid tumors (JapicCTI‐205153). Along similar lines, it has recently been shown that a novel positive allosteric modulator of P2RX7 (which only acts in the presence of high extracellular ATP levels) potentiates the therapeutic effects of an ICI specific for programmed cell death 1 (PDCD1, best known as PD‐1) by stimulating DCs to release IL18 in support of the effector functions of tumor‐infiltrating NK cells and CTLs (Douguet et al, 2021). Incidentally, these therapeutic applications provide an independent and convincing demonstration (mutatis mutandis, almost an “ex adiuvantibus” proof) of the accuracy of early measurements of extracellular ATP concentration in the TME (Pellegatti et al, 2008).

Taken together, these observations point to CD39/CD73‐dependent adenosine generation and consequent adenosinergic signaling via ADORA2A and ADORA2B as a key immunosuppressive mechanism supporting the progression and metastatic dissemination multiple tumors (Fig 3). Of note, while additional adenosine receptors including ADORA1 and ADORA3 (which inhibit cAMP signaling) are expressed by malignant cells and some tumor‐infiltrating immune cells (Stagg & Smyth, 2010), their role in immunosurveillance remains poorly investigated.

Figure 3. Immunosuppression by extracellular ATP metabolites.

Extracellular adenosine (ADO) not only favors metastatic dissemination by binding adenosine A2b receptor (ADORA2B) on malignant cells, hence promoting invasiveness, and endothelial cells (ECs), thus promoting neovascularization, but also modulates the functions of multiple immune cells upon interaction with ADORA2B or adenosine A2a receptor (ADORA2A). Specifically, adenosine signaling inhibits dendritic cell (DC) maturation and interferes with the effector functions of cytotoxic T lymphocytes (CTLs), B cells, and natural killer (NK) cells, while favoring immunosuppression by M2‐like tumor‐associated macrophages (TAMs), myeloid‐derived suppressor cells (MDSCs), and regulatory T (T_REG) cells.

Targeting purinergic signaling for cancer therapy

Consistent with the key role of extracellular ATP and its degradation products in the control of immunosurveillance, a variety of pharmacological and genetic approaches designed to boost ATP‐driven immunostimulation or inhibit adenosine‐dependent immunosuppression (alone or combined with other treatments) have been shown to mediate prominent antineoplastic effects in immunocompetent mouse models of cancer.

Boosting ATP secretion with autophagy‐activating maneuvers, including weekly cycles of nutrient deprivation and administration of so‐called caloric restriction mimetics (CRMs, i.e., molecules that induce autophagy and cause other biochemical correlates of nutrient deprivation in the absence of sizeable weight loss), has been linked to improved therapeutic responses to immunogenic chemotherapies in immunocompetent mouse models of fibrosarcoma, correlating with decreased infiltration by T_REG cells, via a mechanism that depends on expression of the essential autophagy gene autophagy‐related 5 (ATG5) in cancer cells and intact immune responses (Pietrocola et al, 2016; Castoldi et al, 2020; Wu et al, 2020b). Along similar lines, short‐term starvation reportedly boosts the responsiveness of mouse breast cancer cells growing in immunocompetent syngeneic hosts to radiation therapy (RT) (Saleh et al, 2013; Simone et al, 2016). However, proficient autophagic responses in mouse breast cancer cells limit the efficacy of RT in vivo as a consequence of an improved disposal of permeabilized mitochondria that would otherwise release mitochondrial DNA in the cytosol and trigger cyclic GMP‐AMP synthase (CGAS) signaling coupled to type I interferon (IFN) secretion (Medler et al, 2019; Yamazaki et al, 2020). This suggests that whole‐body autophagy activation by short‐term fasting may support the efficacy of RT by mechanisms unrelated to ATP secretion in the TME, potentially linked to improved autophagic responses in immune cells, most of which rely on autophagy for optimal functions (Clarke & Simon, 2019). That said, even though cyclic/short‐term nutrient deprivation could be safely implemented in at least some cancer patients (Krstic et al, 2020), clinicians remain cautious on implementing clinical trials involving such a nutritional measure. Similarly, while many CRMs with robust anticancer activity in preclinical models are currently available as over‐the‐counter medications (e.g., aspirin) (Castoldi et al, 2020) or vitamin supplementations (e.g., nicotinamide) (Buqué et al, 2020), clinical development remains at bay, at least in some cases reflecting potential dosing issues.

Blocking CD39 or CD73 has also been associated with robust antineoplastic effects in immunocompetent mouse models of cancer. For instance, whole‐body Entpd1 deletion as well as reconstitution of radiosensitive hematopoietic cells with Entpd1 ^−/− precursors have been shown to inhibit in vivo growth and metastatic dissemination of mouse melanoma and colorectal carcinoma cells (Sun et al, 2010; Perrot et al, 2019). Similar findings have been obtained in Entpd1 ^−/− mice xenografted with mouse melanoma or fibrosarcoma cells and treated with immunogenic chemotherapy or ICIs (Perrot et al, 2019), as well as with (i) monoclonal antibodies targeting human CD39 in human ENTPD1 knock‐in mice bearing mouse fibrosarcoma cells (Perrot et al, 2019), (ii) monoclonal antibodies targeting mouse CD39 in wild‐type mice used as hosts for mouse melanoma, fibrosarcoma, or colorectal carcinoma cells (Li et al, 2019c; Yan et al, 2020), (iii) antisense oligonucleotides targeting Entpd1 in wild‐type mice bearing syngeneic breast cancer cells (Kashyap et al, 2019), and (iv) pharmacological CD39 inhibitors in wild‐type mice xenografted with mouse colorectal carcinoma cells (Michaud et al, 2011). Of note, in the latter models, the anticancer effects of CD39 blockage could be abrogated by P2rx7 or Nlrp3 whole‐body deletion, as well as by co‐administration of monoclonal antibodies targeting IL18 (Li et al, 2019c; Yan et al, 2020), formally linking therapeutic activity to accrued extracellular ATP signaling (rather than to mere adenosine depletion). Mice lacking Nt5e have been shown to be poorly permissive to the growth of syngeneic glioblastoma, lymphoma, melanoma, ovarian cancer, colorectal carcinoma, and breast cancer cells (Stagg et al, 2011; Wang et al, 2011; Yegutkin et al, 2011; Young et al, 2016; Yan et al, 2019). Moreover, blockage of CD73 with monoclonal antibodies or pharmacological agents mediates standalone therapeutic effects or enhances the efficacy of other treatments (encompassing chemotherapeutics, RT, and ICIs) in immunocompetent mouse models of breast cancer (Loi et al, 2013; Allard et al, 2014; Young et al, 2016; Wennerberg et al, 2020), colorectal carcinoma (Allard et al, 2013; Hay et al, 2016; Tsukui et al, 2020), melanoma (Iannone et al, 2014; Young et al, 2016), ovarian cancer (Häusler et al, 2014; Li et al, 2019b), head and neck squamous cell carcinoma (Deng et al, 2018), and prostate cancer (Allard et al, 2013). In line with such an abundant preclinical literature, several monoclonal antibodies targeting CD39 or CD73 are currently being tested for their safety and anticancer efficacy, either as standalone therapeutics or combined with ICIs, in clinical trials (Table 3). Preliminary findings from such studies point to an acceptable safety profile and promising clinical activity encompassing disease stabilization and (at least in some patients) partial or complete responses (Mobasher et al, 2019; Bendell et al, 2020).

Table 3.

Ongoing clinical trials targeting purinergic signaling for cancer therapy. ^a

Agent	Target	Indications	Phase	Status	Notes	Ref.
AZD4635 (AstraZeneca)	ADORA2A	CRPC	II	Recruiting	In combination with oleclumab (anti‐CD73), durvalumab (anti‐PD‐L1)	NCT04089553
		NSCLC	I/II	Active	In combination with oleclumab (anti‐CD73)	NCT03381274
		Solid tumors	I	Active	Single agent and in combination with durvalumab (anti‐PD‐L1), oleclumab (anti‐CD73), docetaxel, abiraterone acetate, enzalutamide	NCT02740985
		Solid tumors	I	Recruiting	Single agent	NCT03980821
Ciforadenant CPI‐444/V81444 (Corvus Pharmaceuticals)	ADORA2A	NSCLC	I/II	Recruiting	Single agent and in combination with atezolizumab (anti‐PD‐L1)	NCT03337698
		CPRC Renal cell carcinoma	I	Recruiting	Single agent and in combination with atezolizumab (anti‐PD‐L1)	NCT02655822
EOS100850 (iTeos Therapeutics)	ADORA2A	Solid tumors	I	Recruiting	Single agent	NCT03873883
Etrumadenant AB928 (Arcus Biosciences)	ADORA2A ADORA2B	CRC GEC	I	Active	In combination with mFOLFOX	NCT03720678
		NSCLC	I/I	Recruiting	In combination with carboplatin, pemetrexed and pembrolizumab	NCT03846310
		Ovarian cancer TNBC	I	Recruiting	In combination with IPI‐549 (PI3γ inhibitor), doxorubicin, paclitaxel	NCT03719326
		Solid tumors	I	Active	In combination with zimberelimab (AB122, anti‐PD1)	NCT03629756
Taminadenant NIR178/PBF‐509 (Pablobio/Novartis)	ADORA2A	NSCLC	I	Active	Single agent and in combination with spartalizumab (anti‐PD‐1)	NCT02403193
		TNBC	I	Recruiting	In combination with spartalizumab (anti‐PD‐1) and LAG525 (anti‐LAG‐3)	NCT03742349
		Solid tumors	I	Recruiting	Single agent and in combination with NZV930 (anti‐CD73), spartalizumab (anti‐PD1)	NCT03549000
		Solid tumors	I	Recruiting	Single agent and in combination with spartalizumab (anti‐PD‐1); NZV930 (anti‐CD73), KAZ954	NCT04237649
		Solid tumors	II	Recruiting	Single agent and in combination with spartalizumab (anti‐PD‐1)	NCT03207867
SRF617 (Surface Oncology)	CD39	Solid tumors	I	Recruiting	Single agent and in combination with paclitaxel, gemcitabine, pembrolizumab (anti‐PD‐L1)	NCT04336098
TTX‐030 (AbbVie)	CD39	Lymphoma Solid tumors	I/I	Recruiting	Pembrolizumab (anti‐PD‐L1), docetaxel, gemcitabine	NCT03884556
		Solid tumors	I	Recruiting	In combination with budigalimab (anti‐PD‐1), mFOLFOX	NCT04306900
BMS986179 (Bristol Myers Squibb)	CD73	Solid tumors	I/II	Active	Single agent and in combination with nivolumab (anti‐PD‐1)	NCT02754141
CPI‐006 (Corvus Pharmaceuticals)	CD73	NHL Solid tumors	I/I	Recruiting	Single agent and in combination with ciforadenant (ADORA2A antagonist), pembrolizumab (anti‐PD‐L1)	NCT03454451
LY3475070 (Eli‐Lilly)	CD73	Solid tumors	I	Recruiting	Single agent or in combination with pembrolizumab (anti‐PD‐L1)	NCT04148937
NZV930 (Surface Oncology)	CD73	Solid tumors	I	Recruiting	Single agent or in combination with spartalizumab (anti‐PD‐1), NIR178 (ADORA2A antagonist)	NCT03549000
Oleclumab MEDI9447 (AstraZeneca)	CD73	Bladder cancer	I	Recruiting	In combination with durvalumab (anti‐PD‐L1)	NCT03773666
		NSCLC	II	Recruiting	In combination with durvalumab (anti‐PD‐L1)	NCT03334617
		Ovarian cancer	II	Recruiting	Single agent or in combination with durvalumab (anti‐PD‐L1)	NCT03267589
		Pancreatic cancer	I/II	Recruiting	In combination with gemcitabine, paclitaxel, durvalumab (anti‐PD‐L1), FOLFOX	NCT03611556
		TNBC	I/II	Recruiting	In combination with durvalumab (anti‐PD‐L1), paclitaxel	NCT03742102
TJ004309 TJD5 (Tracon Pharmaceuticals)	CD73	Solid tumors	I	Recruiting	In combination with atezolizumab (anti‐PD‐L1)	NCT03835949

CRC, colorectal cancer; CRPC, castration‐resistant prostate cancer; GEC, gastroesophageal cancer; NHL, non‐Hodgkin lymphoma; NSCLC, non‐small cell lung cancer; TNBC, triple‐negative breast cancer.

^aRestricted to active and recruiting studies, as per www.clinicaltrials.gov on February 15, 2021.

Genetic and pharmacological strategies for ADORA2A and ADORA2B inhibition have also been shown to mediate anticancer effects in preclinical cancer models. For instance, (whole‐body or myeloid cell‐specific) Adora2a or Adora2b deletion has been demonstrated to inhibit tumor growth in immunocompetent mice bearing syngeneic melanoma (Ohta et al, 2006; Cekic et al, 2014; Chen et al, 2020b), lymphoma (Waickman et al, 2012; Nakamura et al, 2020), breast carcinoma (Beavis et al, 2013), or lung cancer cells (Chen et al, 2020b). Consistent with this notion, pharmacological ADORA2A (e.g., CPI‐444, SCH58261, AZD4635) or ADORA2B (e.g., PSB‐1115) inhibitors have been attributed robust antineoplastic properties in syngeneic mouse models of melanoma (Willingham et al, 2018), fibrosarcoma (Beavis et al, 2017), breast cancer (Beavis et al, 2013; Mittal et al, 2014; Beavis et al, 2015; Beavis et al, 2017), lymphoma (Nakamura et al, 2020), colorectal carcinoma (Beavis et al, 2015; Willingham et al, 2018), multiple myeloma (Yang et al, 2020b), and renal cell carcinoma (Willingham et al, 2018), especially when combined with ICIs or other immunotherapies. Importantly, at least in some of these models, co‐inhibition of ADORA2A and CD73 mediated superior tumor control as compared to inhibiting ADORA2A or CD73 alone (Young et al, 2016), suggesting a non‐redundant role for these two factors in the establishment of local immunosuppression. Consistent with these preclinical observations, a number of ADORA2A or dual ADORA2A/ADORA2B inhibitors are currently being investigated for safety and activity in clinical trials (Table 3). Preliminary findings from these studies indicate that many ADORA2A/ADORA2B inhibitors exhibit an acceptable safety profile and at least some degree of clinical activity (Chiappori et al, 2018; Fong et al, 2020; Lim et al, 2020).

Additional approaches that may be harnessed to target adenosine signaling in the TME encompass the use of CD38 or ENPP1 blockers and strategies that revert tumor hypoxia (e.g., respiratory hyperoxygenation) (Hatfield et al, 2014; Hatfield & Sitkovsky, 2020). However, CD38‐specific agents (including the FDA‐approved monoclonal antibody daratumumab) are currently used with the aim of eradicating CD38‐expressing myeloma cells (Facon et al, 2019), and ENPP1 blockers are still in preclinical development (Carozza et al, 2020). Similarly, hyperoxygenation has been employed in the past to improve radiosensitivity (because the ability of radiation therapy to cause DNA damage in cancer cell depends on local oxygen tension) but is no longer used for this purpose.

In summary, although a variety of strategies have been successfully used to target purinergic signaling in preclinical tumor models, CD73, ADORA2A, and (less so) CD39 blockers are the only drugs currently in clinical development as anticancer agents, for the most part in combination with standard of care therapeutics or ICIs (Table 3).

Concluding remarks

In summary, extracellular ATP and its degradation products play a key role in the regulation of the tumor immune contexture, hence have a major influence on the propensity of human neoplasms to respond to therapy. While various agents aimed at boosting extracellular ATP concentrations and/or limiting adenosine signaling are already in clinical development, multiple questions to be addressed and avenues to be explored remain. First, it will be important to determine the contribution of P2RX7 and/or P2RY2 signaling in cancer cells to tumor growth and metastatic dissemination in specific settings. Accumulating evidence indicates that some cancer cells can harness extracellular ATP in support of disease progression and resistance to therapy (Martin et al, 2017), suggesting that CD39 and/or CD73 inhibition may not mediate optimal anticancer effects in some settings, and calling for the identification of cancer cell‐targeted P2RX7 or P2RY2 inhibitors. Second, it will be crucial to develop combinatorial approaches based on the dual blockage of ATP degradation and adenosine signaling, potentially in the context of ICI‐based immunotherapy. Indeed, it appears that the efficacy of ADORA2A and/or ADORA2B blockers can be significantly boosted by CD73 inhibition (Young et al, 2016), pointing to a functional non‐redundancy that may be harnessed for therapeutic purposes. Also, while inhibiting ATP degradation or adenosine signaling mediates anticancer effects per se (at least in various models), blocking additional immunosuppressive pathways such as those controlled by co‐inhibitory receptors appears to provide improved disease control in most cases (Beavis et al, 2017; Leone et al, 2018; Goswami et al, 2020). Finally, it will be interesting to elucidate whether and how purinergic signaling can be targeted in patients with innate or acquired resistance to ICI‐based immunotherapy, reflecting the fact that exhausted T cells generally express high levels of CD39 (Canale et al, 2018). Irrespective of these and other incognita, purinergic signaling stands out as a particularly promising target for the development of novel anticancer agents.

Author contributions

LG conceived the article. OK and LG wrote the first version of the manuscript with critical input from LB, TY, FDV, MJS, and GK. OK, TY, and LB prepared display items under supervision from LG. All authors approved the final version of the article.

Conflict of interest

OK reports research funding from Daiichi Sankyo and a co‐founder role of Samsara therapeutics. FDV is a member of the Scientific Advisory Board of Biosceptre, Ltd. GK reports research funding from Bayer Healthcare, Eleor, Genentech, Glaxo Smyth Kline, Institut Mérieux, Lytix, PharmaMar, Sotio, and Vasculox (completed), funding from Samsara, consulting/advisory honoraria from The Longevity Labs and Lytix, membership of the Executive Board of Bristol Myers Squibb Foundation France, co‐founder role of everImmune, Samsara therapeutics and Therafast‐Bio. LG reports research funding from Lytix and Phosplatin (completed), consulting/advisory honoraria from Boehringer Ingelheim, AstraZeneca, OmniSEQ, The Longevity Labs, Inzen, Onxeo, and the Luke Heller TECPR2 Foundation. All other authors declare that they have no conflict to interest.

The EMBO Journal (2021) 40: e108130.