CD73's Potential as an Immunotherapy Target in Gastrointestinal Cancers.

CD73, a cell surface 5′nucleotidase that generates adenosine, has emerged as an attractive therapeutic target for reprogramming cancer cells and the tumor microenvironment to dampen antitumor immune cell evasion. Decades of studies have paved the way for these findings, starting with the discovery of adenosine signaling, particularly adenosine A2A receptor (A2AR) signaling, as a potent suppressor of tissue-devastating immune cell responses, and evolving with studies focusing on CD73 in breast cancer, melanoma, and non-small cell lung cancer. Gastrointestinal (GI) cancers are a major cause of cancer-related deaths. Evidence is mounting that shows promise for improving patient outcomes through incorporation of immunomodulatory strategies as single agents or in combination with current treatment options. Recently, several immune checkpoint inhibitors received FDA approval for use in GI cancers; however, clinical benefit is limited. Investigating molecular mechanisms promoting immunosuppression, such as CD73, in GI cancers can aid in current efforts to extend the efficacy of immunotherapy to more patients. In this review, we discuss current clinical and basic research studies on CD73 in GI cancers, including gastric, liver, pancreatic, and colorectal cancer, with special focus on the potential of CD73 as an immunotherapy target in these cancers. We also present a summary of current clinical studies targeting CD73 and/or A2AR and combination of these therapies with immune checkpoint inhibitors.


INTRODUCTION
Gastrointestinal (GI) cancers are some of the most common cancers worldwide and a major cause of cancer-related deaths (1)(2)(3)(4)(5). Immune checkpoint inhibitors (ICIs), including pembrolizumab (Keytruda) and nivolumab (Opdivo), antibodies against programmed death-1 (PD-1), recently gained Food and Drug Administration (FDA) approval for use in GI cancers ( Table 1) (6)(7)(8)(9)(10)(11). While their approval has been a significant step forward in advancing clinical care, currently, few patients benefit (12). Patients benefiting the most tend to have tumors harboring deficient DNA mismatch repair (dMMR) and high microsatellite instability (MSI-H) (8,13). dMMR and MSI-H occur together at a consistency of 90-95% (referred to as dMMR/MSI-H) (14,15). MMR deficiency leads to high mutational rates and subsequently high presence of neoantigens, making tumor cells more likely to be recognized and destroyed by antitumor immune cells (8,13,(16)(17)(18). Tumor-infiltrating lymphocytes are abundant in dMMR/MSI-H tumors and associate with favorable prognosis (19,20). For comparison, the somatic mutation frequency of dMMR/MSI-H tumors is 10-100-fold to that of proficient MMR tumors (21). In contrast, ICIs as single agents have not shown meaningful benefit for proficient MMR tumors (8), which are the vast majority of GI cancer cases. MSI-H tumors account for 6-22% of gastric, 1% of pancreatic, 3% of liver, and 14-16% of colorectal cancers (22)(23)(24)(25)(26)(27). Antibodies against PD-1/programmed death-ligand 1 (PD-L1) or cytotoxic T-lymphocyte-associated protein-4 (CTLA-4) are the most clinically advanced immunotherapy in cancer (12). The PD-1/PD-L1 axis promotes adaptive immune resistance by suppressing effector T cells and promoting the differentiation of regulatory T cells (Tregs). CTLA-4 also is a negative regulator of T cells; its engagement of B7-1 or B7-2 on antigen-presenting cells inhibits T cell activation (12). Preclinical and clinical efforts are pushing forward with combination ICI therapy as well as ushering in different approaches to harness the immune system to extend immunotherapy efficacy to more patients, including vaccines and viral therapy, adoptive cell transfer, and cytokine treatment (12,28). Challenges with improving efficacy include overcoming immunosuppression activity by the tumor microenvironment, unmasking pre-existing immune cell activity, and the ability to stimulate de novo immunogenicity (29). In recent years, antibodies and small molecular inhibitors against CD73 have made their way into clinical trials as an attractive target for restoring antitumor immunity (30)(31)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43)(44). This review provides a summary of current literature for CD73 in GI cancers and its potential as an immunotherapy target. We also discuss current clinical trials targeting CD73 and adenosine receptors in combination with ICI and conventional therapy and the clinical implications to GI tumors.
Whereas, A2AR activation on T regulatory (Treg) cells promotes Treg expansion and immunosuppressive activity [e.g., increasing forkhead box P3 (FoxP3) expression] (91). Mechanistically, these actions are linked together in a self-reinforcing loop. CD73 on Tregs generates extracellular adenosine and activates A2AR on effector T cells, suppressing effector T cell activity. Extracellular adenosine additionally activates A2AR on Tregs, promoting their expansion and activity (92). Human Tregs rarely express cell surface CD73 (93,94), unlike mouse Tregs (92,95). Instead, CD73 expression by surrounding cells or exosomes is considered to produce the extracellular adenosine. CD73 is expressed by populations of immune cells, stromal cells, epithelial and endothelial cells, cancer cells, and exosomes (96)(97)(98). Recently, CD39 co-expression with CD103 (integrin αE) was identified as a marker of antigen specific, tumor-reactive CD8+ T cells, having resident memory and a high capacity of recognizing and killing autologous tumor cells (99). These cells may be a strategy to improve adoptive cell therapy, which is limited by the ability to identify and expand tumor-reactive CD8+ T cells.
Here, using CD39+ CD103+ to enrich the cells prior to in vitro expansion may increase therapy success (99). Many ongoing studies are directed at capturing and/or reinvigorating T cellmediated antitumor responses. These studies will provide greatly to new approaches for extending and improving immunotherapy efficacy in cancer. A2AR activity also promotes peripheral T cell tolerance, skewing T cell differentiation from adaptive effector cells to adaptive FoxP3+ lymphocyte activation gene-3 (LAG-3)+ Tregs (100).
Frontiers in Immunology | www.frontiersin.org blockade of A2AR or respiratory hyperoxia restores NK cell maturation, proliferative capacity, and cytotoxic function, which improves control over tumor growth, delays tumor initiation and suppresses tumor metastasis (38,101,102). CD73 and/or A2AR blockade or supplemental oxygen in combination with therapies promoting NK cell activity may be relevant strategies to enhance antitumor immunity. Whole-body exposure to 60% oxygen reduces tumor growth by reversing hypoxiaextracellular adenosine-mediated immunosuppression. In these preclinical studies, extracellular adenosine levels and CD39, CD73, A2AR, and A2BR gene expression decreases and coincides with increased antitumor immunity (102,108). Hypoxiainducible factors (HIFs) are strongly linked to increasing CD73 (109), A2AR (110), and A2BR (111) gene expression and collaborates to increase extracellular adenosine/adenosine receptor signaling for dampening inflammation (46,47). Interestingly, recent studies show tumor cells can reprogram NK cells to gain immunosuppressive functions [e.g., increase IL-10 and transforming growth factor-β (TGF-β) production via signal transducer and activator of transcription 3 (STAT3) transcriptional activity, suppressing IFN-γ production] (112). The effects are not mediated through adenosine receptors, suggesting other mechanisms are involved and may not involve the production of extracellular adenosine (112).

Myeloid-Derived Suppressor Cells and Tumor-Associated Macrophages
CD39 and CD73 are upregulated on CD11b+ CD33+ peripheral blood and tumor-associated myeloid-derived suppressor cells (MDSCs) via TGF-β, which their ectonucleotidase activity inhibits T cell and NK cell activity (113). Granulocytic MDSCs expressing high CD39 and CD73 are described in colorectal cancer patients. These cells were found to exert robust immunosuppressive features (e.g., high PD-L1 expression) and activity that could be dampened by blocking CD39/CD73 (114). A2BR activation preferentially promotes the expansion and intratumoral accumulation of CD11b+ Gr1+ MDSCs (115). CD11b+ Gr1+ MDSCs express high CD73, which limits T cell proliferation. CD73 is also considered to facilitate MDSC expansion by generating extracellular adenosine to activate A2BR on myeloid progenitors (115). Accordingly, blocking A2BR reduces CD11b+ Gr1+ MDSCs immunosuppression and accumulation in tumors (87). Extracellular adenosine generated by cancer cells can recruit tumor-associated macrophages (TAMs), which their endonucleotiase activity, in collaboration with CD73 expression on other cells of the tumor microenvironment, further contributes (e.g., suppressing antitumor CD4+ T cell proliferation) to extracellular adenosine-mediated immunosuppression in tumors (116).

PRECLINICAL STUDIES TARGETING CD73 AND ADENOSINE RECEPTORS
CD73's potential as an immunotherapy target has advanced rapidly within the last decade (30)(31)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(67)(68)(69)(70). Current studies focus on combination strategies, including ICIs, adoptive transfer, chemotherapy, and targeted therapy. Preclinical studies show compelling evidence for both CD73 and A2AR blockade in enhancing anti-PD-1 and anti-CTLA-4 therapy. As single agents, both CD73 and A2AR blockade are effective in controlling tumor growth and metastasis. However, the combination of these therapies is far greater at reducing tumor growth, metastatic burden, and prolonging the life of mice. These effects depend on increased IFN-γ production and CD8+ T and NK cell activity (37,38,40,43,122,123). Notably, anti-PD-1 therapy is particularly synergized by inhibiting CD73, (37) and studies report A2AR combined with anti-PD-1 therapy is most effective with cancer cells expressing high CD73. The latter suggests CD73 expression may stratify patients likely to benefit from anti-PD-1 therapy-A2AR blockade combination (122,123). In melanoma, CD73 is a poor pretreatment biomarker for immunotherapy, however, its expression level in relapse tumors has predictive value (41). Therefore, CD73 as a biomarker may be tumor and sample (e.g., primary, metastasis, relapse) specific. Ciforadenant (formerly, CPI-444), an oral A2AR antagonist, recently completed a first-in-human study in patients with renal cell cancer (124). Preclinical studies have shown ciforadenant combined with anti-PD-L1 or anti-CTLA-4 therapy eliminates tumors in up to 90% of mice, restores antitumor immunity, and is effective in mice that failed prior anti-PD-L1 or anti-CTLA-4 therapy (68). Moreover, ciforadenant produced an antitumor memory response in which tumor growth was completely inhibited in mice with cleared tumors when later rechallenged (68). In clinical trials, cirforadenant combined with atezolizumab (anti-PD-L1 therapy) provide great disease control and survival benefit in patients, yet without high objective response rates (124). While reasons are unclear, the Fong and colleagues predict the response is due to persistent antitumor immunity that maintains durable control over tumor growth (124). Monotherapy ciforadenant also provided disease control in some individuals (124). Mechanistically, ciforadenant suppresses the expression of multiple checkpoint pathways on CD8+ effector T cells and CD4+ FoxP3+ Tregs and appears to have profound effects in restoring antitumor immunity at the draining lymph nodes by decreasing PD-1 and LAG-3 expression (69). Thus, a significant benefit of A2AR antagonism is its expansion of responsive cytotoxic T lymphocytes (69). A2AR and/or CD73 blockade also improves anti-CTLA-4 therapy efficacy in melanoma (43). Recently, anti-CD73 therapy combined with an agonist antibody to 4-1BB (4-1BB therapy) showed to restore antitumor immunity (125). 4-1BB is an activation-induced T cell costimulatory molecule that enhances cytotoxic T cell and NK cell activity (126,127). 4-1BB therapy has entered into clinical trials involving GI cancer patients (NCT03330561). Poor efficacy and toxicity have been a concern in the past with 4-1BB therapy (128). Further preclinical studies are warranted. Cancer vaccines educate the immune system to recognize cancer cells. Targeting A2AR in this setting also represents a promising strategy. Responses to melanoma and lymphoma tumor vaccines are increased in A2AR-deficient mice; these mice showed increased expansion of tumor-specific CD8+ T cells and increased survival compared to wild-type mice (129). The effectiveness of adoptive T cell transfer is also increased with genetic deletion or blockade of CD73 or A2AR. Tumorbearing mice benefit from improved tumor control and survival due to increased infiltration and activation of adoptive T cells (30,69,70).
Additionally, preclinical studies show chimeric antigen receptor T (CAR T) cell efficacy is greatly increased by A2AR antagonism (130). CAR activation increases A2AR expression and suppression of mouse and human CAR T cells, which can be reversed by A2AR antagonism or genetic targeting, increasing the therapy benefit of CAR T cells. Efficacy is increased further by combination therapy (A2AR blockade and anti-PD-1 therapy) (130). Increased CD73 expression is seen in patients progressing under adoptive T cell transfer therapy (41). Accordingly, future approaches targeting CD73 in combination with A2AR blockade, anti-PD-1 therapy, and/or adoptive T cell transfer may prove beneficial. Head-to-head comparison studies blocking CD39 and CD73 (44) or CD73 and A2AR (39) also show promise for significantly increasing antitumor immunity. Co-targeting CD73 with A2AR inhibits the compensatory response of A2AR blockade to increase CD73 (39). Whereas, co-targeting CD39 with CD73 is beneficial by targeting two different mechanisms (44). Blocking CD39 elevates ATP levels. High ATP levels promote DC and macrophage antitumor activity, which adds to the antitumor immunity benefits of blocking CD73 (44). Combining CD73 anti-antibodies or small molecule inhibitors with chemotherapy or targeted therapies [e.g., antibodies against epidermal growth factor receptor (EGFR)] also shows merit in preclinical studies (36,81). BRAF and MEK inhibitors combined with A2AR blockade show significant benefit in controlling melanoma tumor growth and metastasis in mice (42). A benefit of BRAF and MEK inhibitor combination is that it downregulates CD73 expression (42). Accordingly, this combination strategy provides the advantage of dampening CD73 expression without added drug/antibody therapy. Preclinical studies that focus on GI cancers will be essential in understanding the therapeutic potential of CD73 and/or adenosine receptor blockade in these tumors.

GASTRIC CANCER
Gastric cancer (GC) is the fifth and third most common cancer and cause of cancer deaths worldwide, respectively (1). Although incidence and death rates are declining (131), advancements in prevention and treatment remain a priority. Five year survival rates drop to 20-30% or less once the cancer moves beyond the lining of the stomach (132). The majority of GC cases are advanced stage (133). Treatment includes gastric resection, radiation, chemotherapy, and targeted therapy, including antibodies against (VEGF)/VEGF receptor 2 (VEGFR2), and HER2 (131). Recently, ICI therapy was approved for GC ( Table 1) (6). However, most patients do not benefit. Other immunotherapies being studied in GC include combination ICI therapy, adoptive cell transfer, vaccines [e.g., melanoma-associated antigen (MAGE) A3 peptides; Bacillus Calmette-Guerin (BCG)], and agonist antibodies for costimulatory receptors [e.g., OX40 (also known as tumor necrosis factor receptor superfamily, member 4), 4-1BB] (134).
Few studies have assessed CD73 expression in GC ( Table 2). CD73 expression is higher in GC vs. normal tissue and associates with poor tumor differentiation, increased depth of invasion, positive nodal status, presence of metastasis, advanced-stage disease, and poor overall survival ( Table 2) (75,135). Increased CD73 in GC may be due in part to hypoxia. Hypoxia-inducible factor-1α (HIF-1α) staining closely correlates with high CD73 expression in gastric tumors (75). In contrast, gene expression studies have shown high CD73 expression associates with favorable overall survival in GC (136). Notably, CD73 expression does not always correlate to protein expression, which may explain the differences between these studies (136). Additionally, significant heterogeneity for CD73 is seen in GC (75,135). For example, 30-50% of advanced stage, deeply invasive, and lymph node-positive tumors express low or no CD73 (75). Significant heterogeneity for CD73 expression is also described for melanoma (41,42). In melanoma, CD73 expression is influenced by sample type (e.g., primary, metastatic, or relapse tissue); therapy treatment; and presence of activating MAPK (e.g., BRAF) mutations, mitogenic and inflammatory signals [e.g., hepatocyte growth factor (HGF) and TNF-α], and necrosis (41,42). Increased CD73 expression with BRAF mutation is also seen in serous ovarian cancer; these patients have better clinical outcomes (137). BRAF mutations are found in 10-20% of colorectal cancer and frequently are MSI-H (138)(139)(140). CD73 expression is also impacted by NT5E promoter methylation, described for both melanoma and breast cancer (141,142). Suffice to say, multiple molecular and genetic factors can affect CD73 expression in human tumors.
Looking ahead, assessing CD73 expression to common molecular and/or genetic alterations of GC and The Cancer Genome Atlas (TCGA) may help to better understand CD73 in GC (23). Studies assessing the association of CD73 expression to immune checkpoints, such as PD-L1, may also  be helpful. Forty percent of GC cases are PD-L1 positive (143), and preclinical studies suggest high CD73 expression in PD-1/PD-L1 expressing tumors may identify patients that would benefit from combination anti-PD-1/PD-L1 therapy and CD73 and/or A2AR blockade (122,123). Few studies globally assess CD73 expression with other ecto-enzymes involved in ATP and adenosine synthesis and metabolism and its intracellular uptake (144), such as other E-NTPDases, ectonucleotide pyrophosphatases/phosphodiesterases (e.g., CD203a), nitcotinamide dinucleotide enzyme (e.g., CD38), prostatic acid phosphatase, alkaline phosphatase (45,145,146), adenosine deaminase, and equilibrative and concentrative nucleoside transporters (ENTs and CNTs, respectively). Reviewed recently by Boison and Yegutkin (144), this may present a major gap in developing effective adenosine-based therapies (144). Accordingly, a more global view of extracellular adenosine metabolism and signaling in GC may also prove significant. Considering CD73/extracellular adenosine's role in immune cell escape, studies of CD73's association to H. pylori-mediated tumorigenesis may provide additional insight. H. pylori infection is responsible for up to 60% of GC cases and arises in the background of inflammation (147,148). Immune cell evasion is important for H. pylori infection and supported by evidence of higher PD-L1 expression in H. pylori positive compared to negative gastric biopsies (149) and that H. pylori-induced PD-L1 expression on gastric epithelial cells converts naïve T cells to CD4+ FoxP3+ Tregs that inhibit T cell proliferation (150). CD73 expression by CD4+ CD25+ Tregs enhances H. pylori infection by increasing local extracellular adenosine, which suppresses IFN-γ production (151). Consistent with this, infected CD73-deficient mice experience worse gastritis and more severe inflammation (e.g., increased IL-2, TNF-α, and IFN-γ and impaired Treg function) (151). Taken together, these studies support that CD73/extracellular adenosine in collaboration with other immune checkpoints may downregulate immune cell responses necessary for recognizing and clearing transformed cells arising in chronically infected gastric tissues, thus supporting GC development. With H. pylori infection, CagA and VacA containing exosomes are released from gastric epithelial cells, stimulating pro-inflammatory responses and affecting the expression of tumor suppressor and oncogenic genes (152). Considering CD73 expression on exosomes promotes tumor immunosuppression (97,98), it would be interesting to see if CD73 is also expressed on H. pylori-mediated exosomes and if its presence or increased presence is a biomarker for the onset of GC.
Additional studies show CD73 promotes tumor cell proliferation, migration, invasion, and stemness in GC cells (135,153). Antitumor roles for extracellular adenosine are also reported, including AMP-kinase (AMPK)-mediated, caspaseindependent apoptosis, via intracellular uptake of extracellular adenosine through ENTs, and caspase-dependent apoptosis, mediated by A1R and A3R (154,155). ENTs passively transport nucleosides based on a concentration gradient (Figure 1) (156,157). A1R and A3R signaling both inhibit adenylyl cyclase activity and can activate multiple downstream signaling pathways, including phospholipase C, producing inositol 1, 4, 5-triphosphate (IP 3 ) and diacylglycerol (DAG), mitogenactivated protein kinase (MAPK), and phosphoinositide 3-kinase (PI3K) (Figure 1) (158)(159)(160). A3R agonist, CF102, is in clinical trials for antitumor benefit in liver cancer (NCT02128958). A3R is also reported to increase HIF-1α through a nontranscription-dependent, non-HIF-1α oxygen-dependent degradation mechanism in several cancer cell lines (161). Though the role of A3R-mediated upregulation of HIF-1α is unclear, these data suggest A3R may both suppress and promote tumor progression. A2AR expression is increased in human GC tissue and correlates with poor tumor differentiation, advanced stage, lymph node positivity, and worse patient outcomes (162). Studies show A2AR, via PI3K-AKT-mTOR signaling, promotes GC cell stemness, EMT, and tumor cell migration and invasion (162). Altogether, more work is necessary to understand the role of CD73/extracellular adenosine in GC. Targeting specific adenosine receptors (e.g., A2AR) may be promising, but represents an area in need of more research.

LIVER CANCER
Liver cancer is the fourth most common cause of cancer death and sixth in terms of incidence worldwide (2). Ninety percentage of liver cancers are hepatocellular carcinoma (HCC) (163). Chronic liver disease (e.g., cirrhosis and fibrosis) is a major risk factor and most commonly caused by hepatitis B or C infection or long-term alcohol abuse (2,164,165). The 5-year survival rate for HCC is 18% (2). Treatment includes tumor resection, liver transplant, and targeted therapy (e.g., multi-kinase inhibitor, sorafenib) (166). However, 70% of patients do not qualify for surgery, due to advance disease, and sorafenib therapy is limited in its benefit; patient survival is prolonged only by a few months (166). ICI therapy was recently approved as second-line therapy for HCC (Table 1) (7,9). Other promising immunotherapies are in development and are aimed at boosting existing or de novo immune responses, including vaccines and oncolytic viruses, and combination ICI therapy (167). Anticipation awaits the results of NCT03298451, a phase 3 clinical trial assessing anti-PD-L1 and anti-CTLA-4 combination therapy vs. monotherapy as better first-line options than sorafenib [HIMALAYA trial, (NCT03298451)].
In recent years, HCC has been a platform for the discovery of novel biology for CD73 in human tumors ( Table 2) (168,169). Studies by Snider and colleagues (168) identified an alternative splicing variant of NT5E, NT5E-2 expressed in liver cirrhosis and HCC. NT5E-2 produces a protein product, CD73-short (CD73s), and is a human-specific isoform that lacks enzyme activity and is unable to dimerize due to the loss of exon 7 with splicing (168). CD73s expression is limited to the cytoplasm and complexes with CD73 to promote proteasome-dependent degradation of CD73 (168). In HCC human tissues, CD73s expression is 6-8fold higher compared to normal liver, whereas CD73 expression is downregulated by more than 90% (168). Accordingly, these studies indicate CD73s may be the major source of "CD73" overexpression in HCC. In contrast, other studies (170,171) report CD73 is overexpressed in HCC and associates with poor tumor differentiation, microvascular invasion, and poor overall and recurrence-free survival (Figure 1) (170, 171). NT5E-2 expression was not assessed in these studies, which is a limitation. In line with CD73s expression, Sciarra et al. (172) especially noted significant cytoplasmic CD73 expression in tumors, particularly with invasive tumors (172) ( Table 2).
Many immunohistochemistry data for high CD73 expressing tumors, including gastric and pancreatic cancer, show significant cytoplasmic staining of CD73 (75,77). Current commercial antibodies are not marketed to distinguish between CD73 and CD73s. Thus, other human tumors with CD73 overexpression may overexpress CD73s. Notably, NT5E-2 is expressed at low levels in most normal human tissues and its expression increases with the onset of disease (123,159). Currently, NT5E-2 remains unstudied in other human tumors despite possible clinical implications. Similarly, recent studies by Alcedo et al. (169) report CD73 enzyme activity in HCC is significantly limited by aberrant glycosylation (169). The authors discovered that in HCC cells, unlike normal hepatocytes, CD73 carries abnormal N-linked glycosylation in its C-terminal catalytic domain, which greatly impairs the enzyme activity of CD73 (169). Aberrantly-glycosylated CD73 also showed to remain partially localized to the cytoplasm with golgi structural protein, GM130 (169). Importantly, these studies show that CD73 protein expression levels may not necessarily reflect its ability to generate extracellular adenosine. Studies by Snider et al. (168) and Alcedo et al. (169) are significant in that they demonstrate CD73 overexpression in human tumors can be misleading. Thus, CD73 immunohistochemistry may fall short in identifying patients likely to benefit the most from CD73 blockade therapy. As mentioned, commercial antibodies are unknown to be specific for recognizing CD73 vs. CD73s. Additionally, they are not primed for recognizing aberrant glycosylation. Instead, CD73 enzyme histochemistry is necessary, which is more challenging for clinical workups. These studies also raise questions as to how close preclinical studies of CD73/extracellular adenosine model human tumors. For instance, syngeneic and spontaneous mouse tumor models do not account for the biology of CD73s, which negatively regulates CD73 (168). A species-specific role of CD73 is also seen for arterial calcifications in humans and is not recapitulated in CD73-deficient mice (173,174). CD73 downregulation in human tumors has been described in endometrial cancer. Its loss associates with more aggressive disease and poor overall survival (175). In normal endometrium, CD73-generated adenosine protects epithelial integrity, which CD73 loss and subsequently the loss of cell-cell adhesions promotes tumor progression (175). In contrast, in normal breast tissue, CD73 is expressed in myoepithelial cells as opposed to differentiated cells (e.g., acinar and ductal epithelial cells) (176). Myoepithelial cells are stem cell-like and exhibit highly invasive behavior similar to tumor cells (177). Consistent with this, CD73 is upregulated in cancer cells of triple-negative breast cancer (TNBC), which are tumors characterized by a gene expression signature similar to basal/myoepithelial cells (36,177,178). Accordingly, studies that reconcile tissue-and cell-specific roles for CD73 in normal GI tissues may help better understand CD73 in GI cancers (179,180). Similar to endometrial cancer (175), CD73 is downregulated in cancer cells of bladder and prostate tumors and associates with poor prognosis (181,182). The role of CD73 in bladder and prostate epithelium is unknown. Notably, CD39 deficiency promotes both induced and spontaneous autochthonous tumors in the liver (183).
For adenosine receptors, studies show A2AR signaling via PI3K-AKT promotes HCC tumor growth and metastasis and is reversed by A2AR antagonist treatment (170). A2BR expression is increased in human HCC tissue and correlates with tumor progression and is likely due to hypoxia (184). HIF-1α increases A2BR expression in HCC cells and cancer cell proliferation (184). Recent studies by Lan and colleagues (185) show HIF-1α's induced expression of A2BR is essential in enriching breast cancer stem cells for the onset of recurrent disease (185). Studies linking A2BR to tumor progression include work in bladder (86), breast (186), colon (187), and prostate (188) cancer and involves A2BR activity on both immune and tumor cells. A2BR antagonist, ATL801, reduces metastases by more than 80% in mice, which is due to increased IFN-γ, IFNinducible chemokine CXCL10, a ligand for CXCR3, and tumorinfiltrating CXCR3+ T cells (86). Needless to say, interests in antagonizing A2BR in human tumors are rising. Studies by Vecchio et al. (188), in prostate cancer, describe a ligandindependent, constitutively active A2BR, which drives cancer cell proliferation (188). Importantly, these studies highlight an unappreciated view that adenosine receptors in tumors may not rely on CD73/extracellular adenosine. Aberrant ligandindependent G protein-coupled receptor constitutive activity is implicated in several cancers (188). In contrast, A3R expression is increased in human HCC and A3R promotes cancer cell apoptosis (189). A3R agonist, CF102, is being evaluated as second-line therapy for HCC (NCT02128958). Increased overall survival is reported with NCT02128958 and phase 3 studies are being planned (190). Taken together, adenosine receptors as opposed to CD73 may be better predictive targets for therapeutic benefit in HCC.

PANCREATIC CANCER
Pancreatic cancer is predicted to become the second leading cause of cancer-related deaths in the United States by 2030 (3,4). Ninety percentage of pancreatic tumors are pancreatic ductal adenocarcinoma (PDAC) while 3-5% are neuroendocrine tumors (PNETs) (191). Smoking, heavy alcohol consumption, obesity, H. pylori infection, and chronic pancreatitis are risk factors (192). Prognosis is incredibly poor, approximately 70% of patients will succumb to the disease in the first year (193). The 5-year survival rate is 9% (193). Standard of care for pancreatic cancer includes radiation therapy, chemotherapy, and targeted therapy (e.g., EGFR inhibitors) (192). The prevalence of therapy resistance to these treatments is a persistent problem. PDAC patients have not benefited from single agent or combination ICI therapy (194)(195)(196) despite increased expression of PD-L1 in tumors (197)(198)(199). Significant efforts are underway to improve immunotherapy efficacy, including studies investigating regulatory B cell inhibition (e.g., Bruton's Tyrosine Kinase (BTK) inhibitors), IDO inhibition, and vaccine therapy (200). Though a predominant target in B cell malignancies, BTK in PDAC is shown to induce B cell-and macrophage-mediated T cell suppression, which BTK inhibitors (i.e., ibrutinib) restore T celldependent antitumor immunity and improve responsiveness to chemotherapy in preclinical studies (201). BTK inhibitors also produce an unexpected anti-fibrotic effect (202). PDAC cancers are rich in stromal cells and fibro-inflammatory reactions, which support chemotherapy resistance (203). A phase 3 clinical trial of ibrutinib in combination with chemotherapy in PDAC was recently completed (April 2019; NCT02436668) (204). Results are not yet publicly available.
Studies of CD73 in human PDAC tissue have only recently emerged ( Table 2). CD73 is upregulated in PDAC compared to normal pancreatic tissue and correlates with increased tumor size, advanced stage, lymph node involvement, metastasis, and poor prognosis (77,80,172). While PDAC tumors are 100% positive for CD73 expression (172), interesting staining patterns for CD73 are seen. Well-and moderately-differentiated PDAC cells express mixed membrane and cytoplasmic CD73 staining. CD73 staining intensity is low to moderate in these tumors (172). In contrast, poorly-differentiated PDAC cells have aberrant CD73 staining, including very strong cytoplasmic CD73 expression (172). The increase of cytoplasmic CD73 expression in PDAC is unclear. We previously mentioned the discovery of CD73s in HCC (168). Studies assessing NT5E-2 (CD73s) expression may help to better understand CD73 in PDAC. CD73 expression in acinar cell carcinomas (ACC) is rare (172). ACC comprises 1-2% of pancreatic tumors and does not carry typical genomic alterations seen in PDAC, including KRAS and TP53 mutations (205), which is suggestive that CD73 expression in PDAC may be linked to KRAS and/or TP53 mutations. KRAS mutation occurs in nearly 100% of PDAC cases (206). In human colorectal cancer (CRC) and non-small cell lung cancer (NSCLC) tissue, CD73 staining is increased in KRAS mutant compared to wild-type tumor (207). KRAS alterations associate with increased CD73, CD39, A2AR, and A2BR gene expression in CRC and NSCLC cell lines, which correlates with anti-PD-1 resistance in KRAS mutant tumor models (207). Moreover, high CD73 expression and KRAS alterations associate with worse overall survival compared to patients with KRAS alterations and low CD73 expression tumors (207). EGFR alterations and high CD73 expression also associate with poor overall survival (TCGA pan-cancer) (207). EGFR alterations and KRAS mutations occur together in 67% of PDAC cases (208). Accordingly, EGFR alterations may also increase CD73 expression in PDAC. A positive association between CD73 expression and EGFR alterations is described in breast cancer (209).
CD73 expression (3+ staining) increases with aggressive disease in PDAC (172), which may be an indicator of an evolving or advancing immunosuppression phenotype. For instance, in PDAC, a decrease in CD8+ T cell infiltration into tumors is seen with the rise of infiltrating Tregs with disease progression (210). As mentioned, human Tregs rarely express cell surface CD73 (93,94), and it is considered that CD73-generated extracellular adenosine from other sources [e.g., cells (96) or exosomes (97,98)] activate adenosine receptors on immune cells for immunosuppression. Accordingly, the coinciding increase of CD73 in PDAC cells may be significant in promoting extracellular adenosine-mediated immunosuppression. Other cells and cell-derived products possibly contributing are CD4+ CD73+ T cells, B cells, and CD39+ CD73+ exosomes (211). CD73+ PDAC and NSCLC cell-derived exosomes activate A3R on intratumor and peripheral mast cells, which promotes remodeling of the tumor microenvironment through increasing the expression of angiogenic factors (212,213). Additionally, in PDAC models, tumor-infiltrating CD11b+ CD103-DCs promote tumor growth by inducing expansion of FoxP3 neg CD39+ CD73+ tumor-promoting Tregs (214).
Pancreatic neuroendocrine tumors and carcinomas (PanNET/PanNEC) account for 1-10% of pancreatic tumors (215,216). Thirty to fifty percent of PanNET/PanNEC express mild to moderate CD73 expression and associates with increased malignant potential, which is similar to gastrointestinal (GI)-NET/NECs (80,172,217). In GI-NET/NECs, CD73 expression positively correlates with PD-L1 expression (217), which possibly anti-PD-1/PD-L1 therapy with CD73 and/or A2AR blockade may benefit these patients. Increased expression of CD73 with PanNET also associates with cancer cell stemness (e.g., aldehyde dehydroxygenase expression) and aggressive behavior (80). Filippini et al. (218) recently reported a transplantable model of mouse pancreatic tumor organoids into immunocompetent mice that recapitulate human PDAC progression and that the system serves as a suitable model for immunophenotypic studies (218). The organoid-derived isographs induce the expression of many immunosuppressive/aggressive biomarkers with tumor development and evolution, including CD73 (218). Studies using such models may provide a significant understanding of CD73/extracellular adenosine signaling in immunosuppression and the immunoevolution of PDAC.
CD73 also shows to promote drug resistance and tumor growth in PDAC cells. For instance, high CD73 expression and low miR-30a-5p expression in PDAC cells result in chemotherapy (e.g., gemcitabine) resistance (77), and CD73 knockdown inactivates AKT and extracellular signal-regulated kinase (ERK) signaling and slows cancer cell growth (77). In contrast, studies show extracellular adenosine treatment in combination with AKT inhibitor, GSK690693, reduces PDAC growth and induces tumor cell apoptosis and senescence in patient-derived xenografts (PDX). Mechanistically, the intracellular uptake of extracellular adenosine via ENTs (Figure 1) appears important for this response, as dipyridamole (pan-ENT inhibitor) treatment remarkably recovers cell viability (219). The difference between these studies likely relates to the subcutaneous transplanting of tumors (77) vs. tumors transplanted to the tail of the pancreas (219). Indeed, for example, CD39 deficiency can promote the development of both induced and de novo tumors in the liver, which is in contrast to its role in antitumor immunity of subcutaneous transplanted tumors (183,220). It is considered that the surrounding microenvironment and interaction with these cells by the tumor likely produce different responses and outcomes. In the next several years, adopting in-depth and detailed characterization of CD73/extracellular adenosine in immunocompetent, autochthonous pancreatic cancer models, humanized models, and human organoids will be essential for better understanding the possible therapeutic benefit of targeting CD73 and adenosine receptors in pancreatic tumors.

COLORECTAL CANCER
Colorectal cancer (CRC) is the third most common cancer and the second cause of cancer-related deaths worldwide (5). CRC incidence rates are declining in the United States and are stable in most other Western countries, whereas rates are rising in Eastern Asia and Eastern Europe and likely reflect the adoption of a Western lifestyle (221,222). CRC risk factors include obesity, Western diet, lack of physical activity, excessive alcohol use, hereditary syndromes (e.g., Lynch syndrome), and smoking (223). Treatment includes surgery, combination chemotherapy, radiation therapy, and targeted therapy, including antibodies against VEGF/VEGFR or EGFR (224). Although advances in better screening and treatment have been made in the last decade, long-term survival remains poor for metastatic CRC patients. The 5-year survival rate is <15% (225). ICI therapy was recently approved for refractory dMMR/MSI-H metastatic CRC (Table 1) (8,10,11). Of CRC cases that are dMMR/MSI-H, only 4% are metastatic. Accordingly, several approaches, including IDO inhibitors, vaccine therapy, and combination ICI therapy are being studied to extend immunotherapy efficacy to more patients (226). A better understanding of CD73/adenosine receptor signaling in CRC may help in these efforts.
Early studies assessing CD73 in CRC were part of larger efforts examining enzymatic patterns of key enzymes involved with purine metabolism and salvage, including ADA, alkaline phosphatase, hypoxanthine-guanine phosphoribosyltransferase ( Table 2) (227,228). Studies by Camici et al. (227) reported no difference with CD73 enzyme activity between CRC and normal tissue (227). In contrast, Eroglu et al. (228) showed higher CD73 enzyme activity in tumors compared to normal tissue (228). No associations were found with high CD73 enzyme activity and poor clinical features. Instead, high CD73 enzyme activity was associated with well-differentiated tumors and low CD73 enzyme activity associated with poorlydifferentiated tumors (228). More recent studies show high CD73 expression correlates with poor tumor differentiation, lymph node involvement, advanced stage, and poor survival (78). In rectal cancer, CD73 expression in the different cell types carries different clinical prognosis (229). High CD73 expression in cancer cells and low CD73 expression in stromal cells associates with poor overall survival, whereas low CD73 expression in cancer cells and high CD73 expression in stromal cells is more favorable (229). Bladder cancer is similar. CD73 positive expression by epithelial cells predicts better progression-free survival and overall survival, whereas stromal cell CD73 positivity predicts poor outcome (230). Accordingly, these studies support that CD73 in tumors may suppress and promote tumor progression. Although unknown, targeting tumors with dual roles for CD73 may prove challenging for CD73 inhibitor therapy.
Tumor heterogeneity is likely one explanation for the reported differences of CD73 expression in CRC. CRC tumors carry significant inter-and intra-heterogeneity (231,232), so much so that in recent years an international consortium was formed to establish a robust molecular and genetic classification scheme for CRC. These global efforts led to the development of the consensus molecular subtypes (CMS) (233). Accordingly, future efforts assessing CD73 expression to the CMS groups (e.g., CMS1, CMS2, CMS3, CMS4) may provide a better understanding of CD73 in CRC and the possible molecular and genetic alterations that drive its downregulation and/or overexpression. Indeed, high CD73 expression in CRC may be associated with CMS2 tumors. In CRC, CD73 is a predictive biomarker of patient response to anti-EGFR therapy (234). In line with this, the CMS2 group predicts tumors that are more responsive to anti-EGFR and anti-HER2 therapy (235). Also consistent is that CD73 promotes CRC cell proliferation and tumor growth through β-catenin (WNT)/cyclin D1 signaling (236) and CMS2 tumors are characterized by WNT and MYC signaling (233). KRAS mutations/alterations are likely also linked to CD73 expression in CRC; discussed previously in the section on PDAC (207). Thus, investigating CD73 expression in KRAS mutant tumors may provide additional insight. A focus on metastatic samples may also be important. Liver metastasis occurs in 50% CRC patients (237). Recently, studies have shown high CD73 expression associates with significantly shorter time to recurrence and poor survival (238). In renal cancer patients, an adenosine high (AdenoSig hi ) expression signature was identified in pretreatment biopsies and associated with clinical response to A2AR antagonism (124). Similar efforts in identifying biomarker signatures may provide greatly to improving immunotherapy efficacy in CRC.
In preclinical studies, CD73 deletion increases CD8+ T cells and IFN-γ production to suppress the growth of MC-38 mouse colon cancer (32). The depletion of CD73 on CD4+ Foxp3+ Tregs also is significant in restoring antitumor immunity in this model (32). Similarly, CD39-deficient mice are resistant to MC-38 metastasis (239,240). Whereas, overexpression of CD39 increases MC-26 mouse colon cancer cell metastasis to the liver (220). CD39 deletion does not increase the development of primary MC-26 orthotopic transplant tumors in heterozygous CD39 mice or mice transgenic for human CD39 compared to wild-type mice (220). Recent studies show support for cotargeting CD39 and CD73 in combination with ICI therapy and/or chemotherapy (44). Tumor-bearing mice benefit from increased antitumor immunity in these studies, which is due to the recovery of DC, macrophage, and effector T cell antitumor activity (44). In line with these studies, inhibiting CD39 or CD73 on MDSCs from CRC patients is effective in dampening the immunosuppressive activity of these cells (114).
Adenosine receptors may also be possible therapeutic targets. High A2AR expression associates with larger tumor size, increased tumor invasion, and higher TNM (TNM Classification of Malignant Tumors) stage in CRC (241). High A2AR expression also predicts poor patient survival and is positively correlated with PD-L1 expression (241). Consistent with the possible benefit of combined A2AR antagonist and ICI therapy (43,68,122,123), studies with MC-38 cells, show A2AR antagonist, ciforadenant, combined with anti-PD-L1 or anti-CTLA-4 therapy eliminates 90% of tumors in mice by restoring antitumor immunity (68). Notably, MC-38 cells are normally highly sensitive to ICI therapy (37,123). Additionally, in many studies, MC-38 cells are grown subcutaneously. Accordingly, it is not known how close these preclinical studies model immunosuppression and immunotherapy efficacy for CRC. A2BR is also upregulated in CRC and likely is linked to tumor hypoxia and progression (187). In vitro studies show A2BR expression is upregulated in CRC cells by hypoxia and promotes cancer cell proliferation, which is dampened by A2BR antagonism (187). A2BR antagonism also dampens A2BRmediated CD73 expression by cancer-associated fibroblasts (CAFs) and CAF-associated immunosuppression activity (242). A3R is overexpressed in human CRC tissue and stimulates tumor growth via extracellular signal-regulated protein kinases 1 and 2 (ERK1/2) (243,244). In contrast, studies also report A3R activity inhibits tumor growth by modulating glycogen synthesis kinase-3β (GSK-3β) and NF-Kappaβ (NF-κβ). A3R agonist treatment inhibits CRC cell proliferation, limits liver metastasis, and increases the cytotoxicity of chemotherapy (e.g., 5-fluorouracil) (245)(246)(247). Interestingly, treatment of CRC cells with caffeine, a non-selective adenosine receptor antagonist, inhibits A3R-mediated stabilization of HIF-1α (248). It is unclear if HIF-1α mediated by A3R promotes tumor progression or antitumor activity. HIFs are described to have pro-and antitumor activity in CRC (249). Moreover, overexpression of HIF-1α does not increase CRC tumorigenesis and does not result in spontaneous tumor formation in mice (250). Taken together, while many adenosine pathway members show evidence for possible therapeutic targeting in CRC, detailed studies in human tumors and relevant preclinical models are greatly needed.

CLINICAL IMPLICATIONS
Inhibiting CD73 (and/or A2AR) restores antitumor immunity in many preclinical studies with combination approaches showing superior efficacy. Accordingly, several clinical trials inhibiting CD73 (e.g., antibodies against CD73 or small molecule inhibitors) in combination with ICI therapy, A2AR antagonism, targeted therapy, and/or chemotherapy are underway ( Table 3). Preliminary safety profiles report BMS-986179, an anti-CD73 humanized monocolonal antibody, and its combination with nivolumab (anti-PD-1 therapy) to be well-tolerated in patients (NCT02754141) (251). Recent studies in renal cell cancer (RCC) reported the feasibility and safety of A2AR antagonist, ciforadenant (124). Similar to preclinical studies, durable clinical benefit was associated with increased recruitment of CD8+ T cells (124). Additionally, combination therapy (ciforadenant and anti-PD-L1 therapy) showed benefit in patients who had progressed on anti-PD-1/PD-L1 therapy. Notably, patients in  Frontiers in Immunology | www.frontiersin.org these trials were heavily pretreated (≥3 prior treatments) (124). It will be interesting in the future to see if CD73 and/or A2AR therapy efficacy is increased further when used in earlier lines of therapy (124). Moreover, the authors discovered responding patients carry an AdenoSig hi signature (124). Assessing whether this signature can also be detected in pretreatment biopsies of other cancers and possibly primary tumors may be beneficial (124). Biomarkers or gene signatures will likely be key in identifying patients benefiting the most from CD73/adenosine receptor therapy. Clinical trials are underway for AB928, a dual A2AR/A2BR antagonist, and include a focus on GI cancers [e.g., esophageal cancer and CRC; NCT03720678 (Table 3)]. A favorable safety profile of AB928 combined with chemotherapy has been reported in patients (252). Future studies in GI cancers that focus on determining if adenosine-mediated resistance to immunotherapy therapy exists at diagnosis or evolves with therapy will also be of significant benefit. Encouraging early results for BMS-986179 combined with nivolumab report clinical benefit (partial response) in one or more patients with pancreatic and prostate cancer (NCT02754141) (251). Both are poorly immunogenic tumors. Preclinical studies show CD73/adenosine therapy (e.g., A2AR deletion) liberates CD8+ T cells for antitumor activity even against weakly immunogenic sarcomas (70). Therapy benefit in these studies is independent of the anatomical location of the tumor (70). Thus, therapeutic benefit across many tumors (immunogenic and non-immunogenic) is expected. Understanding factors preventing immune cells from recognizing and eliminating cancer cells will continue to be important in the advancement of immunotherapy strategies. Poor tumor immunogenicity can be a result of many features, including HLA class I molecule downregulation or loss (253); genetic, epigenetic, and chromosome alterations regulating presentation and processing of surface epitopes (254)(255)(256); expression and secretion of immunosuppressive factors (e.g., PD-1, TGF-β, adenosine) (257); and the inability of cancer cells to produce new surface epitopes that are different from what immune receptors have regularly experienced (258). Whether CD73 expression associates with dMMR/MSI-H in GI tumors and its blockade would further increase immunotherapy efficacy in these tumors is unknown. In NSCLC studies, tumor mutational burden and neoantigen burden does not associate with CD73 high or low expression (74). Taking advantage of strong associations of CD73 with molecular and genetic alterations (e.g., KRAS mutation and EGFR alterations) may benefit GI cancers. Combination studies of CD73 inhibitors with anti-EGFR therapy and/or tyrosine kinase inhibitors are in clinical trials for managing resistance ( Table 3). In CRC, high CD73 predicts patients benefiting from cetuximab (anti-EGFR therapy) (234). Benefits are the same for both wild-type and mutant KRAS tumors (234). It would be interesting to see the performance of combination cetuximab with CD73 inhibitors in preclinical CRC studies considering that inflammation is a mechanism of resistance to cetuximab (259). In melanoma, combination BRAF and MEK inhibitors with an A2AR antagonist induces significant tumor control in preclinical studies (41). MEK is a promising target for KRAS, NRAS, and BRAF mutant tumors and is being targeted in CRC (260). Recently, MEK inhibitor, cobimetinib, combined with anti-PD-L1 therapy (atezolizumab) failed to improve survival in microsatellite-stable metastatic CRC patients in a phase 3 clinical trial (261). Could the inclusion of A2AR antagonists be key to the success of these studies? AMG510, a selective inhibitor for KRAS (G12C) recently showed promising antitumor effects, including increasing ICI therapy sensitivity in preclinical models (262). Its combination with CD73/adenosine receptor blockade may be a promising future approach. AMG510 is in clinical trials (NCT03600883). Mentioned previously, hyperoxia induces antitumor immunity in preclinical studies, which involves the downregulation of many adenosine pathway genes (102,108). With drug toxicity being a concern with studies pushing past two targets, approaches like this that can simultaneously dampen multiple immune checkpoints may be better tolerated and provide greater benefit (263). Although a drawback of hyperoxia therapy is that it may/does cause tissue damage (263,264), it is interesting to consider whether this response also benefits in helping to recover antitumor immunity. Hyperoxia is in clinical trials for many conditions/diseases (ClinicalTrials.gov; hyperoxia, 87 studies).

CONCLUSIONS
Immunotherapy in GI cancers currently benefits only a few patients. Blocking adenosine signaling by inhibiting CD73 and/or A2AR/A2BR antagonism has the potential to improve antitumor immunity in these tumors. However, identifying which patients may benefit stands in the way. To aid in these efforts, a better understanding of CD73 in human GI cancers is greatly needed. This includes initiating studies that assess CD73 in addition to other ecto-enzymes involved in extracellular adenosine synthesis and metabolism as well as their association with key molecular and genetic features. A focus of CD73 expression in primary, pretreatment, and relapsed samples will also be of great value in addition to identifying predictive biomarkers or gene signatures relating to efficacy of CD73/adenosine receptor blockade. Mechanistically, studies assessing CD73/extracellular adenosine receptor activity in humanized and autochthonous tumor mouse models and patient-derived organoids will provide needed insight into the role of CD73/extracellular adenosine in these tumors. Moreover, studies in HCC have revealed CD73 overexpression in human tumors can be misleading. Future studies also incorporating this insight have the best chance of helping to better define CD73 in GI cancers.

AUTHOR'S NOTE
The figure was created using Servier Medical ART templates, which are licensed under a Creative Commons Attribution 3.0 Unported License; https://smart.servier.com.