Overcoming Acquired PD-1/PD-L1 Resistance with CD38 Blockade

Therapies targeting the immune checkpoint molecule programmed death 1 (PD-1) and its ligand, PD-L1, have revolutionized cancer treatment in the past decade. Although one third of patients with cancer have benefited from PD-1/PD-L1 blockade, a large subset of patients remains refractory to these monotherapies (innate resistance). Alarmingly, disease progression is also seen in patients who initially demonstrated a response to PD-1/PD-L1 therapy (acquired resistance; ref. 1). Understanding mechanisms underlying such innate and acquired resistance to PD-1/PD-L1 blockade and developing novel therapeutic approaches to overcome resistance is critical to significantly improve clinical outcomes.

In this issue, Chen and colleagues highlight another molecular pathway that regulates extracellular adenosine levels and may restrict the long-term benefit of anti–PD-1/PD-L1 (2). This study determined that tumors treated with PD-1/PD-L1–blocking antibodies developed resistance through the upregulation of CD38, which was induced by all-trans retinoic acid (ATRA) and type I IFN in the tumor microenvironment (Fig. 1). In vitro and in vivo studies demonstrate that CD38 inhibited CD8⁺ T-cell function via adenosine receptor signaling, and that CD38 or adenosine receptor blockade were effective strategies to overcome the resistance in the development of combination immunotherapeutic strategies (2).

An innate anti–PD-1 resistance gene signature called IPRES was identified by transcriptomic analysis of melanoma biopsies isolated from patients prior to anti–PD-1 treatment. IPRES includes a group of 26 transcriptomic patterns associated with regulation of mesenchymal transition, cell adhesion, extracellular matrix remodeling, angiogenesis, and wound healing. It was suggested that inhibition of biological processes relevant to the IPRES signature might enhance the efficacy of PD-1/PD-L1 blockade (3). In contrast, mechanisms of acquired resistance to immunotherapy have been discovered in patients through tumor sequencing and genetic screens, and most evidence so far largely points to tumor alterations that converge on the antigen presentation and IFNγ signaling pathways (4). CD8⁺ T cell–induced IFNγ initiates an antitumor response through cognate IFNγ receptor 1 and 2, and downstream signaling molecules JAK1/2 and IFN regulatory factor 1 (IRF1). Exposure to IFNγ is known to induce PD-L1 expression, and therefore mutations in any of these IFNγ-regulating genes limit the level of PD-L1 expression, making these tumors resistant to anti–PD-1/PD-L1 treatment (4). Recently, the blockade of TNF signaling was shown to enhance anti–PD-1 treatment efficacy through attenuation of PD-1–mediated TIM3 expression and associated activation-induced CD8⁺ T-cell death (5).

A complete analysis of the downstream events following PD-1/PD-L1 inhibition and resistance will (i) help in the identification of novel biomarkers that could predict response to PD-1/PD-L1 therapy and (ii) offer superior combination strategies that will augment efficacy of PD-1/PD-L1 blockade. To this end, Chen and colleagues used mouse models of lung cancer and melanoma to investigate gene signatures that were differentially regulated upon PD-1/PD-L1 blockade. CD38 emerged as a promising candidate whose expression was upregulated manifold in progressively anti–PD-L1-resistant tumors. CD38 is an ectoenzyme expressed by tumor cells and activated immune cells such as T cells, B cells, natural killer (NK) cells, and myeloid-derived suppressor cells (MDSC). CD38 catalyzes metabolically important NAD to ADP-ribose (ADPR) and cyclic ADPR, and this is further dephosphorylated to adenosine in the presence of tumor-localized CD203a and CD73 (6). CD38 expression in the tumor microenvironment was not exclusive to anti–PD-1-resistant tumor cells but was also expressed on regulatory T cells and MDSC that had infiltrated the tumor microenvironment.

To directly establish the link between PD-L1 and CD38, the authors generated PD-L1 knockout (KO) tumors and sorted these cells on the basis of CD38 expression. Interestingly, PD-L1^KO CD38⁻ tumors failed to grow in wild-type mice compared with PD-L1^KO CD38^hi tumors. Consistently, CD38 knockdown tumors grew significantly slower than CD38 wild-type tumors in wild-type mice. This protection was abolished in the absence of CD8⁺ T cells, suggesting that CD38-expressing tumor cells impair CD8⁺ T-cell function. Mechanistically, the authors showed that CD38 promoted adenosine generation presumably through the CD38–CD203a–CD73 axis (ref. 7; Fig. 1). This subsequently led to the activation of the immunosuppressive adenosine receptors A2AR and A2BR in T cells and stunted cytotoxic T-cell functions (Figure 1).

The immunosuppressive effects of adenosine on the tumor microenvironment are well documented (7). This study provides a major role for CD38-mediated adenosine in restricting anti–PD-1/PD-L1 efficacy. Similarly, blocking adenosine generation or signaling via CD73 or A2AR, respectively, increased the sensitivity of tumors to anti–PD-1 therapies (7). Conversely, adenosine also increases PD-1 levels on CD8⁺ T cells (7). This further suggests that a vicious cycle exists between PD-1/PD-L1 and adenosine in the tumor microenvironment, justifying the need to cotarget PD-1/PD-L1 and adenosine pathways.

Given this striking relationship between PD-L1 and CD38, and CD38 and adenosine, Chen and colleagues next tested whether combining anti–PD-L1 with anti-CD38 antibodies or with adenosine receptor antagonists might improve therapeutic outcome in mouse cancer models (2). Coinhibition of PD-L1 and CD38 or PD-L1 with the combination of A2A/A2B adenosine receptor antagonists showed superior antitumor efficacy over monotherapy alone. This was also accompanied by marked increase in the percentage of CD8⁺ effector T cells, and significant reduction in the percentage of CD4⁺ regulatory T cells and CD45⁺ MDSC. The authors used the combination of both A2A and A2B adenosine receptor inhibitors in their experiments; therefore, it is difficult to evaluate the relevance of A2A versus A2B receptor expression. Because A2A receptor expression was relatively greater than A2B receptor expression on CD8⁺ T cells and A2A is a higher-affinity receptor than A2B, it is likely that the majority of antitumor effects observed were associated with A2AR on CD8 T cells.

The anti-CD38 antibody daratumumab is an effective treatment for patients with relapsed or refractory multiple myeloma; however, a proportion of patients develop therapy resistance. A detailed investigation into these myeloma cells indicated that CD38 expression was significantly reduced 14 weeks after daratumumab infusion. Exogenous addition of ATRA restored CD38 expression on these bone marrow–derived myeloma cells, thus making these cells again sensitive to daratumumab therapy (8). An exciting hypothesis would be to test whether reduced CD38 expression following daratumumab affected PD-1/PD-L1 expression in the tumor microenvironment, thus sensitizing myeloma cells to anti–PD-1/PD-L1 treatment.

Besides CD38, the CD39–CD73 axis can also generate adenosine, albeit through a different catalytic pathway (6, 7). How these nonoverlapping and adenosine-generating pathways are regulated in patients displaying adaptive anti–PD-1/PD-L1 resistance would also need assessment. Importantly, CD38 is also a receptor for CD31 or PECAM1, a marker of endothelial cells. High CD31 expression is related to tumor cell migration and may correlate with poor survival in patients with cancer (9). It is hence likely that the antitumor activity observed with antagonistic CD38 antibody might occur through additional nonadenosine pathways such as reduced angiogenesis.

The authors also probed whether these findings could be translated to humans by analyzing tumor specimens from patients with lung cancer and melanoma. CD38 expression was associated with an active immune infiltrate and also coincided with the expression of FOXP3, CTLA4, TIM3, PD-1, PD-L2, LAG3, BTLA, CCL2, HVEM, and IDO. Furthermore, CD38 expression was significantly increased in patients on treatment receiving nivolumab (anti–PD-1 therapy) compared with pretreatment groups. This provides further rationale to trial anti-CD38 antibody in patients developing acquired resistance to anti–PD-1 or anti–PD-L1 therapy. Although the anti-CD38 antibody daratumumab is clinically in use for multiple myeloma, the value of blocking CD38 is yet to be tested in patients with solid tumors. Interestingly, intracellular CD38 is also critical for the antitumor effects of NK cells (10). NK cell–tumor cell interaction activates intracellular CD38 in NK cells by protein kinase A, which results in the production of ADPR that targets transient receptor potential melastatin 2 (TRPM2), Ca²⁺ permeable channels, on cytolytic granules. This TRPM2-mediated Ca²⁺ is important for NK-cell degranulation and their antimetastatic function (10). This contrasts with Chen and colleagues (2), who show that CD38 on tumor cells produces immunosuppressive adenosine. Therefore, in-depth research is required to delineate the role of CD38 on tumor versus immune cells, and whether CD38 blocking will be truly therapeutic in metastatic malignancies may need further investigation.

Together, the findings of Chen and colleagues strongly suggest that targeting adenosine may offer therapeutic benefit for patients displaying progressively anti–PD-1/PD-L1–resistant tumors. Experimental evidence supports that other adenosinergic molecules such as CD39, CD73, and adenosine receptors may be critical in anti–PD-1/PD-L1 resistance (7). A comparative analysis of these molecules across human tumor types is warranted, and further mechanism-of-action studies must be performed to determine how redundant these molecules might be with respect to anti–PD-1/PD-L1 resistance.

M.J. Smyth reports receiving commercial research grants from Bristol-Myers Squibb, Tizona Therapeutics, and Aduro Biotech and is a consultant/advisory board member for Tizona Therapeutics. No potential conflicts of interest were disclosed by the other authors.