It is important to note that adenosine pathway related molecules, such as CD39 and CD73, are also expressed by tumor infiltrating lymphocyte (TIL) populations (37, 38). In human colon and lung tumor samples, analysis of TIL populations found that only CD8 T cells that express CD39 represent the stimulated form of T cells whereas those lacking CD39 play a bystander role (39). An in vitro study found similar results that T cells rapidly upregulate CD39 expression upon activation but that CD73 expression remains unchanged (37). In contrast to these findings, another group found that CD39 positive CD8 T cells represent the exhausted phenotype with decreased expression of both TNF and IL-2 through analysis of both human and mouse samples from breast and melanoma tumors (40). It is important to note however that heterogeneity within samples was high with some TIL samples expressing high levels of CD39 and some quite low (39, 40). An additional study utilizing a melanoma mouse model found that both CD73 and CD39 are upregulated on T cells during immune response, but that high CD73 limits effector function through a decrease in mitochondrial capacity (41). This suggests a potential mechanism that T cells control levels of extracellular ATP and AMP to self regulate their activity.

In addition to T cells, CD39 and CD73 function has also been explored on NK cells. A recent report studying both breast and sarcoma tumors found that CD73 positive NK cells within the tumor express higher levels of immune checkpoint molecules such as LAG-3 compared to CD73 negative NK cells found in the peripheral blood. These NK cells have suppressive functions on CD4 T cells (42). An additional study utilizing a mouse model of melanoma found that mice with tumors show modest increase in expression of both CD39 and CD73 on NK cells as compared to tumor naïve mice (38). Furthermore, another study found that CD39 expression on NK cells was not essential for cytotoxic function (43). It is important to note that immune cell populations such as B cells (44), T regulatory cells (45), and MDSC’s (46) have also been shown to express high levels of CD39 and CD73 resulting in the production of adenosine. Interplay between the diverse cells within the tumor microenvironment resulting in adenosine production displays the complexity of this pathway.

Recently, multiple groups have focused on targeting the adenosine pathway in EGFR mutant NSCLC leading to swift discoveries in the recent years (26, 67). Standard of care for EGFR mutant tumors are tyrosine kinase inhibitor (TKI)s (68). EGFR mutant tumors are not very responsive to ICB therapy possibly due to their low tumor mutational burden (TMB) as compared to tumors carrying another common oncogenic driver of NSCLC, KRAS, which are more responsive to ICB (69, 70). Somatic mutations can serve as neoantigens and high TMB status is associated with increased sensiviity to ICB therapies on NSCLC (71). Response rates to ICB therapies are between 7-16% depending on EGFR mutation type, compared to 22% in EGFR wild type tumors (72). Acquired resistance to TKIs, such as Osimertinib is common and occurs on average around 18 months after treatment initiation (73). Multiple studies reported that CD73 upregulation was one of the mechanisms of resistance to TKIs (74, 75). We recently analyzed genes related to immune cell function in EGFR mutant versus wild type LUAD and found that NT5E and ADORA1, which encodes for one of the adenosine receptors, were two of the top upregulated immune-suppresive genes in this subtype (26). EGFR mutant NSCLC has a suppressed tumor microenvironment with few NK cells and CD8 T cells, but increased dendritic cell populations, and other immune cell populations remain similar to those seen in EGFR wild type tumors (26). In an EGFR mutant mouse model, we showed that treatment with a CD73 antibody alone resulted in decreased lung tumor growth displaying proof of concept as a treatment option (26). Another group observed similar correlations with EGFR pathway activation and adenosine signaling. Interestingly this group showed Oleclumab is effective in treating EGFR mutant tumors in combination with PD-L1 antibody in a human peripheral blood mononuclear cell (PBMC) transplanted mouse model (76).

Another recent report focused on the connection between MET amplification, CD73 and suppression of STING pathway in EGFR mutant tumors (67). Authors found that MET amplification induces CD73 expression and restrains the function of STING resulting in reduced T cell activation while also causing resistance to 3^rd generation EGFR TKIs. MET amplification was found to be specifically associated with CD73 expression in EGFR mutant LUAD. Genetic knockout of MET in lung cancer models resulted in decreased CD73 expression therefore displaying a connection between these two pathways (67). This study links CD73 intrinsic function to extrinsic function through immune cell regulation.

Although adenosine pathway expression has been studied the most in the EFGR mutant subset of LUAD, other common mutations such as KRAS, BRAF, and MET are just beginning to be explored. A recent study utilizing the TCGA data set found that CD73 expression is significantly elevated in KRAS mutant LUAD (25, 77). Similar induction of CD73 expression was also seen with ALK gene alterations. As proof of concept, they found that pharmacological inhibition of ALK led to a decrease in CD73 mRNA expression (25). This data suggests a link between the adenosine pathway and oncogenic drivers of LUAD. A new study in pancreatic ductal adenocarcinoma (PDAC), where almost 80% of cases are KRAS mutant, demonstrated that activating mutation in KRAS alone is sufficient to elevate levels of CD73 suggesting that CD73 induction with oncogenes extends beyond lung cancer (78). Results from these studies thus far demonstrate oncogenic drivers of NSCLC possibly drive increased CD73 expression and that underling mechanism behind this warrants further investigation.

CD73 expression is seen in a subset of SCLC patient CDX models and cell lines and its expression correlated with the non-neuroendocrine or YAP1 positive subtype (62, 63). A recent study focused on the metabolomics of primary SCLC tumors. This study demonstrated that AMP and adenosine were the two significantly upregulated metabolites in primary patient samples in the MYC high subtype versus MYC low subtype (79). This study also examined the link between chemoresistance and purine biosynthesis and found that chemo-resistant SCLC cell lines have increased levels of AMP and ATP compared to parental cell lines. These pre-cursors could result in adenosine accumulation within the tumor microenvironment (79). A similar finding was seen in vivo with chemotherapy relapsed tumors displaying increased purine biosynthesis, including AMP (79). Therefore these interesting findings suggest that purine biosynthesis maybe advantageous for SCLC relapse after treatment.

Cellular plasticity and evolution of the tumor microenvironment with disease progression has been reported in NSCLCs. Transdifferentiation of EGFR mutant NSCLC to SCLC is one of the reported resistance mechanisms to EGFR TKIs. This occurs in approximately 3-10 percent of EGFR mutated NSCLCs (80). Studies have found that SCLC transformed tumors in these patients no longer respond to TKIs but they do respond to platinum etoposide therapy at least initially, similar to SCLCs (80–82). A recent report found that loss of extracellular-signal regulated kinase (ERK) signaling dependency was the main regulator in this transformation and this resulted in expression of neuroendocrine transcription factors which was normally repressed by ERK signaling (83). SCLC transdifferentiation was also reported in therapy resistant KRAS mutant LUAd (84). It would be interesting to study whether these transformed tumors retain expression of CD73 and can be targeted by CD73 targeting molecules.

In the field of SCLC multiple groups have found that SCLC primarily switches from neuroendocrine to a more non-neuroendocrine like state. This was shown to be driven through NOTCH driven Myc expression (85, 86), while other groups show that this is driven by EMT gene signature including expression of YAP1 (87). Neuroendocrine and non-neuroendocrine tumor types have shown to have different immune microenvironments with neuroendocrine tumors having far less immune cell infilrates (48, 88, 89). A new report examined 146 SCLC patient samples through IHC and found that 2.3% of tumors expressing YAP1 dominantly. Additionally, the authors found that 17.6 percent of tumors expressed 2 of the SCLC lineage markers and 2.8% were positive for 3 different lineage markers. These areas of different lineage marker staining tended to cluster away from one another suggesting that different cell populations were in various parts of the tumor (90). A recent study highlights that the SCLC intertumoral composition is constantly evolving and that there maybe continuous subtype switching (91). Multiple studies have suggested that one of the target genes of the YAP/TAZ axis is CD73, suggesting the idea that induction of CD73 expression may provide a selective advantage to YAP1 expressing SCLC cells under treatment (92–94). This connection must be explored further, but it is possible that the YAP1 positive cells are contributing to adenosine production, with high expression of CD73 seen in YAP1 positive CDX and SCLC cell line models (63, 95).

Recent studies have shown that CD73 contributes to EMT (epithelial mesenchymal transition) and is a novel target beyond the current key genes that are implicated in EMT such as E-cadherin, vimentin, and N-cadherin (109). A recent study utilizing a breast cancer model with genetic loss or pharmacological inhibition of CD73 resulted in far fewer and less invasive organoids in vitro and decreased lung metastasis when injected into an immune competent mouse model. It was reported that when CD73 is knocked down that there is increased E-cadherin expression and that cells appeared to revert from EMT (12). A similar finding was also shown in a preclinical model of ovarian cancer (110). Another recent study in breast cancer found that SNAIL, a key regulator of EMT, bound at promotors of NT5E through chromatin immunoprecipitation (ChIP) sequencing analysis suggesting that there could be interplay between these two genes and the EMT program (111). An additional recent report in NSCLC line A549 showed that CD73 overexpression leads to increased invasion and metastasis both in vitro and in vivo (25). Further studies employing the A549 model demonstrated that CD73 promotes cell proliferation through binding to EGFR which leads to activation of AKT/mTOR pathway (104). Across numerous cancer types genetic knockdown or knockout of CD73 in human cancer cell lines grown in immune deficient mice demonstrated that loss of this molecule reduces tumor growth and metastatic potential even when a functional immune system is not present (12, 110, 112, 113).

Outside of the scope of the EMT program, studies have demonstrated that CD73 plays a role in other biological processes as well. A few different studies have linked CD73 to control of cell cycle progression (112, 114, 115). A specific study in pancreatic ductal adenocarcinoma (PDAC) found that when CD73 is knocked down, this results in arrest of cells at G1 phase through AKT/ERK/Cyclin D signaling (114). Other studies have focused on how CD73 expression leads to treatment resistance. A recent in vitro study in LUAD demonstrated that knock down of CD73 altered cell cycle progression and sensitized cells to cisplatin chemotherapy treatment (115). Similar results were seen in a model of glioblastoma when they knock down CD73 in their model this resulted in increased temozolomide sensitivity, and that resistance is caused by CD73-A_2BAR signaling (116). An additional in vitro study in breast cancer model found that CD73 deficient MDA-MB-231 cells to be significantly more sensitive to Olaparib likely due to decreased PARP activity. They also found that loss of CD73 suppressed mitochondrial respiration and led to increased genomic instability (117).

Studies thus far have demonstrated an interesting link between expression of CD73 and sensitivity to commonly used therapeutic agents. Of note, in the tumors resistant to numerous therapies CD73 expression is increased. These include radiation (118), chemotherapy (119), TKIs (75), as well as monoclonal antibodies (120). This has been shown to occur through different mechanisms across cancer types. A study in melanoma found that there was increased CD73 expression in the tumors resistant to both adoptive T cell transfer and ICB therapy. They found that both in mouse models and human tissues that MAPK signaling resulted in class switching from an epethial to mesenchymal state and that there was clonal selection for a more “invasive” cell population (121). In hepatocellular carcinoma (HCC) CD73 was shown to contribute to resistance to Lenvatinib, a VEGF inhibitor, through AKT overactivity which resulted in increased SOX9 expression and stemness of HCC cells (13). Within the tumor micronevironment, a pre-clinical study in triple negative breast cancer (TNBC) found that after chemotherapy that there was an increase in CD73, CD47, and PD-L1 positive tumor cells (122). Such examples of therapy resistance show that CD73 is tied to multiple mechanisms of therapy resistance.

CD73 expression correlated with adenosine production in most studies, however it is technically difficult to separate the ectonucleotidase activity from other functions of CD73 in genetic inactivation or antibody blockade studies. One group recently thoroughly studied the non-enyzmatic role of CD73 by blocking the nucleotidase function. They reported a direct physical interaction between CD73 and Src in the ER resulting in Src activation (100). They found that CD73 ectonucleotidase activity did not contribute to Gemcitabine resistance but that resistance was rather caused by AKT pathway activation by CD73. Furthermore they demonstrated that when they mutated the zinc finger binding domain of CD73, which severely inhibits ectonucleotidase activity, that there was still resistance to Gemcitabine suggesting again that it was not the enzymatic activity leading to resistance (100). An additional study in HCC showed that CD73 localizes within the ER of cancer cells and is more abundant in tumors than normal tissue. They propose that there is both a short and and full length of CD73 proteins with the long form only having ectonuclotidase activity and the shorter version found solely localized in the ER (123). To our knowledge these are the only two studies to yet examine proximity of CD73 in the cell and therefore further understanding of the differences between mebrane bound and intracellular CD73 are needed.

NK cells are a key component of the innate immune system and display cytotoxic reponse. Recent advances in the field of lung cancer have shown that NK cells play an essential role in controlling tumor growth and response to ICB therapy (134–137). NK cells can target cancer cells independent of antigen presentation on major histocompatibility complex (MHC) making them a valuable therapeutic target as MHC-I expression is lost or reduced in tumors as one of the mechanism of immune evasion (138). Loss of MHC-I is seen across lung cancer subtypes as well (139). However, loss of MHC-I can sensitize tumors to NK cell killing as MHC-I is an immune inhibitory molecule for NK cells (140). Recent advances in the field of lung cancer have shown that NK cells play an essential role in controlling tumor growth and response to ICB therapy (134–137). Recent studies have demonstrated that around 10-20 percent of circulating lymphocytes in the lungs are NK cells, which is higher than NK cell levels found in peripheral blood (141). However, there are numerous reasons why these cells are not able to infiltrate tumors, and reasons include different immune suppressive metabolites and inhibitory molecules. Additionally, tumors have been shown to shed the NK cell activating ligands to escape NK cell tumor killing and antibodies are currently being tested that can prevent this (142).

It was reported that adenosine impedes NK cell function (143–145). A₂AR inhibition in a co-culture system resulted in increased NK cell proliferation and signaling response as compared to untreated cultures (146). As a mechanism of tumor resistance researchers found that upon binding of the 4-1BBL domain that tumor cells can hijack NK cells and induce CD73 expression in the NK cells in a breast cancer model. These adenosine producing NK cells additionally had higher levels of immune inhibitory checkpoints such as LAG-3 (42). A similar finding was also found in in vitro co culture methods when cancer cell lines with high CD73 were found to induce CD73 expression in an established NK cell line, NK-92 cells. Cancer cell lines with low CD73 expression did not induce CD73 expression in NK cells (147). A recent study used a chimeric antigen receptor (CAR)-NK cells to target CD73 in a preclinical lung cancer model. Authors observed that both in vitro and in vivo that their CD73 directed CAR inhibited lung cancer growth while not attacking normal tissue (126). These findings demonstrate that adenosine as a cancer target must be explored further.