Myeloid Maturity: ATRA to Enhance Anti–PD-1?

In this issue of Clinical Cancer Research, Tobin and colleagues report the results of their phase Ib/II study of pembrolizumab combined with all-trans retinoic acid (ATRA) for treatment-naïve patients with advanced melanoma (1). Though some patients with advanced melanoma enjoy a transformative benefit from anti-PD-1 therapy, most patients do not respond (2), and combinations with additional immune checkpoint inhibitors (ICI) leave most patients experiencing primary or acquired resistance (3, 4). The discrepancy in outcomes among patients responding to anti-PD-1 therapies versus not highlights the imperative to target resistance mechanisms. In this study, the authors propose ATRA as a possible therapy to enhance anti-PD-1 through targeting of immunosuppressive myeloid-derived suppressor cells (MDSC). Across multiple discrete pathways, MDSCs are implicated in impairing effective antitumor immunity (5). This pilot study of 24 patients shows that pembrolizumab plus ATRA was safe with a low rate of grade 3–4 adverse events, and that the combination generated a surprisingly-high response rate, albeit in a small patient population. Changes in the peripheral blood compartment showed an overall decrease in circulating MDSCs and increase in HLA-DR+ myeloid cells after treatment with ATRA.

MDSCs are a plastic and diverse group of predominantly immature myeloid cells. MDSC myelopoiesis is thought to be triggered by pathologic stimuli to provide innate host defense; however, under conditions of chronic inflammatory stimuli, as with advanced cancers, MDSC production may be sustained as a protective mechanism against inflammation-mediated tissue damage (6). For example, MDSCs levels correlate with tumor burden and response to therapy among patients with advanced cancer (7). Unlike mature myeloid cells, MDSCs have low phagocytic potential, and instead exert broad suppression of effector T-cell functions. Specific mechanisms of MDSC immunosuppression include the expression of immune checkpoints, scavenging of metabolites key to T-cell function, production of reactive oxygen species (ROS) toxic to T cells, and secretion of soluble immunosuppressive cytokines (5). The diversity of MDSC populations and their functions has generated multiple therapeutic strategies to overcome their tumorigenic effects. For example, MDSC recruitment and differentiation into M2-polarized macrophages can be targeted through inhibition of macrophage-specific colony-stimulating factors such as CSF1R (8). In addition, the growth factor TGFβ, secreted by MDSCs, is a known mechanism of resistance to anti-PD-1 therapy. TGFβ can also be directly targeted to potentially overcome this MDSC-mediated resistance mechanism (9, 10). The therapeutic strategy with ATRA, however, is to alter the differentiation state of MDSCs, converting them to a less immunosuppressive and more mature myeloid cell.

ATRA is in the retinoid family and binds the nuclear retinoic acid receptors (RAR) and retinoid X receptors that modulate transcription initiation and functions of multiple different genes (11). ATRA is perhaps best known for its fundamental role in treating acute promyelocytic leukemia (APL). This disease is the result of a chromosomal translocation creating the PML-RARα fusion protein that in turns halts differentiation of myeloid differentiation in the promyelocytic stage. ATRA is able to induce degradation of this oncoprotein thus allowing normal maturation of APL cells. In turn, the vast majority of patients with APL achieve a complete remission of their leukemia (12). While it is frequently observed that ATRA reduces MDSCs in patients with cancer by inducing maturation, the additional mechanisms by which ATRA exerts its effect on MDSCs is less clear. Recent studies show that ATRA induces accumulation of glutathione in MDSCs, which in turn scavenges ROS in MDSCs and induces their differentiation (11). This differentiation then shifts MDSCs along the myeloid lineage to mature dendritic cells, macrophages, and granulocytes (Fig. 1; refs. 13, 14). As the authors highlight, ATRA significantly reduces peripheral blood MDSCs when combined with other immunotherapy partners such as anti-CTLA-4 antibodies and dendritic cell vaccines. While as a single agent, ATRA is unlikely to induce similar antitumor efficacy as seen with APL, the prior studies suggest that it may synergize with more potent ICI partners such as pembrolizumab.

In this single-arm, single-institution study, Tobin and colleagues show that ATRA plus pembrolizumab was well tolerated and generated encouraging responses with concurrent decreases in circulating MDSCs. Of the 24 treated patients, 5 experienced high-grade adverse events, mostly attributed to pembrolizumab. One dose-limiting toxicity of meningitis/stroke was attributed to ATRA, and the starting dose of pembrolizumab plus ATRA was maintained as the recommended phase two dose. A total of 71% (n = 17) of patients achieved a RECIST response with 50% (n = 12) achieving a complete response. Peripheral blood MDSCs were significantly decreased among responders as compared with nonresponders. Altogether, this pilot study demonstrated that ATRA plus pembrolizumab generates the expected decrease in MDSCs with higher-than-expected responses that one would expect with pembrolizumab alone. The tolerability of the regimen suggests that if the efficacy signals were maintained in a larger patient sample, this approach would compare favorably to current ICI combinations often used for treatment-naïve patients with advanced melanoma. The authors appropriately highlight the study's limitations—namely, the lack of a pembrolizumab control arm as well as tumor biopsies to compare against clinical responses and the observed peripheral blood correlates. It will be of great interest to see the results from the KEYMAKER-U02 treatment-naïve melanoma study (NCT04305041) evaluating pembrolizumab with and without ATRA as well as triplet combinations of pembrolizumab, ATRA and other agents such as anti-TIGIT and lenvatinib.

The small sample size with a majority of patients having mostly M1a-b and normal lactate dehydrogenase may enrich for responders to pembrolizumab monotherapy. Similarly, the changes in the MDSCs could conceivably reflect the known antitumor benefit of pembrolizumab alone. A larger study with access to tumor biopsies may help to better correlate the added benefit of ATRA. Historical examples of pairing agents without known single-agent activity along with anti-PD-1 therapy offer a word of caution. In the ECHO-301 study of anti-PD-1 plus the IDO-inhibitor epacadostat, promising results from the early phase study failed to translate in the subsequent randomized phase III study (15). Similar phase III failures were also observed in studies adding both bempegaldesleukin and talimogene laherparepvec to anti-PD-1 therapies (16, 17). While multiple hypotheses have been offered to explain the failure to translate promising early results, the failure to interrogate intratumoral dynamics closed the window to understanding many of these reasons. This cautionary example offers a strong rationale to integrate pretreatment and on-treatment biopsies into subsequent development of ATRA plus pembrolizumab.

Characterizing intratumoral correlates of immune response to ATRA plus pembrolizumab could also allow optimization of patient selection. On the basis of the broad mechanisms by which MDSCs inhibit antitumor immunity, patients set to benefit most from MDSC depletion might be those with non–T-cell–inflamed or “cold” tumors. Should ATRA fundamentally overcome the negative immune effects of MDSCs in the tumor microenvironment, this would permit those patients to then benefit from anti-PD-1 therapy. Conversely, those patients with a baseline T cell–inflamed tumor may experience a lesser benefit to the addition of ATRA. This paradigm of rigorous interrogation of baseline tumors could then pair ATRA plus pembrolizumab based on tumor phenotype, as we have suggested a paradigm of immunotherapy development (18). Similarly, patient selection by baseline tumor phenotype offers a logical extension of testing ATRA plus pembrolizumab in additional tumor histologies such as lung cancers where MDSCs mediate resistance to anti-PD-1 therapies (19). Tobin and colleagues have completed an important pilot study to justify the study of ATRA with pembrolizumab, and we look forward to further exploration of this regimen.

D.J. Olson reports personal fees from MJH Life Sciences, Novartis, GLG Consulting, and Iovance and grants from American Cancer Society outside the submitted work. J.J. Luke reports participation on data and safety monitoring boards of AbbVie, Agenus, Immutep, and Evaxion; membership on scientific advisory boards for (no stock): 7 Hills, Affivant, BioCytics, Bright Peak, Exo, Fstar, Inzen, RefleXion, and Xilio, and (stock): Actym, Alphamab Oncology, Arch Oncology, Duke Street Bio, Kanaph, Mavu, NeoTx, Onc.AI, OncoNano, physIQ, Pyxis, Saros, STipe, and Tempest; consultancy with compensation from AbbVie, Agenus, Alnylam, Atomwise, Bayer, Bristol-Myers Squibb, Castle, Checkmate, Codiak, Crown, Cugene, Curadev, Day One, Eisai, EMD Serono, Endeavor, Flame, G1 Therapeutics, Genentech, Gilead, Glenmark, HotSpot, Kadmon, KSQ, Janssen, Ikena, Inzen, Immatics, Immunocore, Incyte, Instil, IO Biotech, Macrogenics, Merck, Mersana, Nektar, Novartis, Partner, Pfizer, Pioneering Medicines, PsiOxus, Regeneron, Replimmune, Ribon, Roivant, Servier, STINGthera, Synlogic, and Synthekine; and research support (all to institution for clinical trials unless noted) from AbbVie, Astellas, AstraZeneca, Bristol-Myers Squibb, Corvus, Day One, EMD Serono, Fstar, Genmab, Ikena, Immatics, Incyte, Kadmon, KAHR, Macrogenics, Merck, Moderna, Nektar, Next Cure, Numab, Palleon, Pfizer, Replimmune, Rubius, Servier, Scholar Rock, Synlogic, Takeda, Trishula, Tizona, and Xencor. In addition, J.J. Luke holds patents (both provisional) 15/612,657 (cancer immunotherapy) and PCT/US18/36052 (microbiome biomarkers for anti–PD-1/PD-L1 responsiveness: diagnostic, prognostic, and therapeutic uses thereof). No other disclosures were reported.

J.J. Luke acknowledges NIH UM1CA186690-06, P50CA254865-01A1, P30CA047904-32, and R01DE031729-01A1.