Boosting immunotherapy of CD73-positive solid tumors by targeting purinergic signals with engineered natural killer cells

Extracellular adenosine (ADO) generated in response to tumor microenvironment (TME) hypoxia has been recognized as a critical immunosuppressive metabolite that impairs anticancer immune responses [1, 2]. ADO disables cytotoxic effector functions of both CD8⁺ T and natural killer (NK) cells predominantly via A_2A adenosine receptor signaling, enabling tumor immune evasion and escape [3, 4]. Furthermore, signal transduction through the A_2A adenosine receptor inhibits the Th1 CD4⁺ T-cell response, debilitating the cytokine environment necessary to support these effector cell types [5], and enhances proliferation of regulatory T cells (T_regs) and granulocytic MDSCs (myeloid-derived suppressor cells) [6]. Apart from its role in regulating global immune responses, ADO exerts powerful and specific immunomodulation of NK cell function [7]. Alongside inhibiting maturation of NK cells and limiting accumulation of cytotoxic CD56^dim subsets [4], ADO suppresses trafficking of NK cells to tumor sites by altering the chemokine milieu [8], and inhibits NK cell effector function against tumor targets [9]. Therefore, modulating ADO levels in the TME may ameliorate antitumor immunity and inhibit tumor growth. Recently, Hatfield et al. showed that downregulation of tumor-derived extracellular adenosine by oxygen supplementation improved the cytotoxic capacity of CD8⁺ T and NK cells [10]. Arabella et al. reported that co-inhibition of CD73 and adenosine A_2A receptor signaling enhanced anti-tumor immune responses by improving recruitment of NK cells to tumor sites [11]. It is becoming evident that targeting the adenosinergic pathway holds significant potential for cancer treatment and can enhance adoptive immunotherapy.

NK cells, specialized effectors of the innate immune system, can respond rapidly to cancer cells due to expression of germline-encoded activating receptors capable of directly binding to pathogen-derived or stress-induced self-antigens [12]. Allogeneic NK cells (such as clinically-studied NK-92 cells [13]) cause no graft versus host disease (GVHD) making their widespread, off-the-shelf use feasible [14]. Mature NK cells have a relatively limited life-span, permitting effective antitumor activity while reducing the probability of long-term adverse events, such as on-target/off-tumor effects [15]. Expression of chimeric antigen receptors (CARs) can increase the specificity and the cytotoxicity of NK cells against cancer targets [16] and rescue the downregulation of activating receptors induced by suppressive TME mechanisms such as hypoxia [17]. NK cells also have a better safety profile as they can avoid in vivo cytokine storm [18] and lack clonal expansion.

Here, we report pre-clinical data showing that combination immunotherapy with CD73 blockade and CAR-NK-92 cells shows improved anti-cancer effects against CD73⁺ solid tumor targets both in vitro and in vivo. NK cells, engineered using a genomically-stable and clinically-safe non-viral vector system to target NKG2D ligands via DAP10 and CD3ζ co-signaling domains based on the piggyBac transposon system, could mediate a powerful anti-tumor response in conjunction with purinergic blockade (Scheme 1b). Our results revealed that CD73 targeting enhances NK cell effector function via mechanisms that are ADCC- and adaptive immunity-independent. Taken together, this is the first report demonstrating the performance of non-viral piggyBac transposon system-engineered CAR-NK cells. We also demonstrate, for the first time, that targeting the CD73-adenosine axis could be used in combination with CAR-NK cell immunotherapy to redirect purinergic signaling away from adenosinergic immunosuppression as an immunometabolic treatment of solid tumors.

NK cells, innate lymphocytes, play a primordial role in tumor immunosurveillance and can detect and eliminate cancer cells that are deficient in MHC class I molecules [30]. Given their potent anticancer activity, therapeutic manipulation of NK cells provides an attractive strategy for cancer treatment. NKG2D, the best characterized among NK cells’ activating receptors, is directly involved in the recognition of tumor cells that express its ligands [16, 27, 30]. However, immunosuppressive cytokines and metabolites in the TME, such as TGF-β and ADO, alongside hypoxia, inhibit NK cell function by downregulating NKG2D alongside other activating receptors, leading tumors to escape surveillance by NK cell-mediated responses [31–34]. To mitigate loss of NKG2D, NK cells have been engineered with CARs that express NKG2D ligands fused to various co-signaling domains [35]. Endowing NK cells with endogenous transgenes encoding NKG2D enables them to restore the loss in NKG2D expression induced by the tumor microenvironment. It also enables them to maintain, or increase, NKG2D expression to promote sustained cytolytic ability. In order to enhance the clinical feasibility and improve the safety of adoptively-transferred, engineered NK cells, we developed non-viral, allogeneic NKG2D-CAR NK cells using PBAE447 and CAR-encoding piggyBac vectors, which bear the ability to induce increased and stable NKG2D expression in NK-92 cells via the genomic incorporation of CAR transgenes. We observed significantly more robust and sustained expression of NKG2D over time in transfected cells compared to non-transfected controls. In addition, preliminary data with primary human-derived NK cells showed an increase in NKG2D expression in transfected cells using the PBAE447/CAR-encoding piggyBac-based system (data not shown). This suggests that piggyBac-based genetic modification of NK cells could also be expanded to the autologous use of patient-derived cells. NKG2D.CAR-NK-92 cells also consistently showed higher degranulation activity, IFN-γ production and killing ability against CD73⁺ solid tumor targets (GBM43, GBM10, A549, and PC3). Besides, these NKG2D.CAR-NK-92 cells also showed stronger lytic activity against CD73^– liquid tumor cells (K562) (Additional file 1: Figure S10).

However, TME immunosuppression operates across a broad spectrum of mechanisms which collectively inhibit NK cell anti-tumor responses. Among these, the TGF-β pathway has been extensively characterized [36], while ADO, a purine metabolite, has more recently emerged as a potent immunosuppressant. ADO, which is generated and accumulates in the hypoxic TME, suppresses NK cells’ migration into tumor sites, while elevated concentrations of ADO can significantly suppress the proliferation of NK-92 cells. Activation with cytokines such as IL-2 is likely to be dampening ADO-induced immunosuppression, as evidenced by the high concentrations of ADO needed to achieve sustained inhibition of NK cell proliferation.

One of ADO’s key signaling mechanisms is via interaction with adenosine receptors, a group of four G-protein-coupled receptors, most notably the A_2A adenosine receptor, causing dysregulation of effector immune cell subsets (including tumor-infiltrating NK cells and CD8⁺ T cells), dampening their antitumor immune response [37]. This was evident by the effect of EHNA treatment in our study, which induced a significant decrease in the lytic activity of IL-2-primed NKG2D.CAR-NK-92 cells. EHNA can rescue ADO immunosuppression by inhibiting the enzyme adenosine deaminase (ADA), which converts ADO into inosine, leading to the extracellular accumulation of higher concentrations of ADO. In addition, ADO can directly regulate tumor proliferation, survival, adhesion, and migration by binding to A_2B adenosine receptors, which are expressed on cancer cells [38]. Based on the functions of ADO through the activation of distinct adenosine receptors, therapeutic strategies targeting adenosine receptors that inhibit ADO signal transduction are showing great potential. Among these, A_2A antagonists have been shown to be able to enhance anti-tumor immunity [39]. Moreover, some of them are undergoing clinical testing (NCT02655822↗, NCT02403193↗, and NCT02740985↗).

ADO accumulates upstream of its signaling interactions via its receptors, through an enzymatic cascade that concludes with CD73, an ectonucleotidase which converts AMP into ADO. This cascade makes CD73 a key modulator of the high levels of ADO in the TME of solid tumors. Overexpression of CD73 has been found in broad types of cancer cell lines and patients’ biopsies including breast cancer, colorectal cancer, ovarian cancer, gastric cancer, glioblastoma and gallbladder cancer [20]. Moreover, its expression is associated with poor clinical prognosis of many cancer patients [40]. In our study, all of the tested cancers cells, including patient-derived glioblastoma (GBM43 and GBM10), lung carcinoma (A549), and prostate cancer (PC3) cells were characterized by high CD73 expression. In addition to the immunoregulatory roles of CD73, the activity of CD73 was also revealed to be related to cancer proliferation, differentiation, invasion, and metastasis. CD73 induces the release of matrix metalloproteinases (MMPs) that facilitate breakdown of the extracellular matrix (ECM), thus enabling cancer cell invasion and migration [41, 42]. CD73 promotes cervical cancer cell proliferation and migration, via potentiating EGFR/Akt and VEGF/Akt pathways [43]. Moreover, CD73 regulates stemness and epithelial-mesenchymal transition (EMT) in ovarian cancer-initiating cells [44]. High CD73 expression is mainly driven by TME hypoxia, through a hypoxia-inducible factor 1α (HIF-1α) binding site in the CD73 gene promoter [45]. Hypoxia causes impairment of NK cell functions and downregulation of activating receptor NKG2D [17, 46]. Compounding this is TGF-β, which can increase CD73 expression on tumor-infiltrating cells [47]. Our data suggests that CD73 expression on CD73⁺ solid tumor cells (both GBM10 and A549) can be further increased due to TGF-β. Collectively, this creates a solid tumor niche that fuels cancer progression via mechanisms that lead to higher ADO concentrations and impaired NK cell functions. In this context, CD73 has risen to prominence as a biomarker and a meaningful target for cancer therapy.

In response to this, antibody-mediated CD73 blockade has been reported to show anti-cancer effects when used individually or in combination with other agents [11, 21, 22, 48]. So far, one human CD73 antibody (MEDI9447) has progressed to clinical trials (NCT02503774↗). However, most of the studies have been limited to evaluating the efficacy of CD73 blockade in murine tumor models, using mouse as the species reactivity of the anti-CD73 clones. Previously, Sebastian et al. showed that anti-human CD73 antibody could improve NK cell killing against CD73⁺ ovarian cancer cells in vitro by acting on mechanisms that may be ascribed to both ADCC and reduced adenosine levels [49]. In our study, we used an anti-human CD73 antibody (clone 7G2) which can bind specifically to CD73 and neutralize the biological activity of CD73. We found that treatment with anti-CD73 antibody could inhibit A549 tumor growth in vivo independently of adaptive immune cells or NK cells in the early stages of tumor growth. At the doses used in our study in vivo, once tumor burden increased, the effect of CD73 lessened. These results could indicate that while CD73 has the ability to act in an autocrine manner to promote tumor progression—thus supporting recent findings on the mechanistic action of CD73 on solid tumors [43, 44]—treatment regimen, dosage, and antibody isotype might play a role in controlling its therapeutic effect. Nonetheless, treatment with NKG2D.CAR-NK-92 cells in combination with CD73 blockade induced a sustained, consistent, and significantly heightened suppression of tumor growth. Because NK-92 cells used in this study do not express the FcγRIIIa receptor (CD16), they are unable to mediate ADCC [50], suggesting that mechanisms other than recruitment of NK cell activity via ADCC engagement enables altered anti-tumor function in response to blockade of CD73. This was supported by an enhanced intratumoral migration of engineered NK cells into CD73⁺ lung tumors in combination with CD73 blockade. Purinergic targeting via CD73 might thus be regulating tumor growth also via mechanisms that exclude immune cell recruitment, a finding that has important clinical implications.

Our results show that targeting CD73 activity can redirect purinergic metabolism and enhance immunotherapy of CD73⁺ solid tumors—including patient-derived primary and recurrent targets—with engineered NK cells by alleviating TME immunometabolic suppression induced by adenosinergic signaling, and enhance intratumoral migration of NK cells in vivo. We observed that hypoxia-induced CD73 ectonucleotidase activity is able to mediate both tumor control as well as innate immunity. By engineering NK cells non-virally using piggyBac vectors, we demonstrated the pre-clinical feasibility of a new cell engineering approach that can mediate stable genomic incorporation of CAR transgenes as a directly translatable immunotherapy of solid tumors.