PD-1 Inhibition—Trouble for Subsequent TIL Therapy in Patients with Melanoma?

In this issue of Clinical Cancer Research, Levi and colleagues investigate factors impacting the efficacy of adoptive cell transfer (ACT) in patients with anti–PD-1 treated metastatic melanoma (1). Immune checkpoint inhibition (ICI) targeting inhibitory receptors including PD-1, CTLA-4, and most recently LAG-3, have become the standard of care for the first-line treatment of patients with advanced melanoma. While ICI therapies have revolutionized the treatment of melanoma, currently only 40% of patients respond long-term and as such salvage therapies are needed for ICI-refractory patients.

Rosenberg's group at the NCI demonstrated over 30 years ago that adoptive transfer of in vitro expanded tumor-infiltrating lymphocytes (TIL) into lymphodepleted patients with advanced melanoma could lead to durable tumor responses, including complete responses. Prior to the wide availability of ICI most patients who received ACT were anti–PD-1 naïve. However, the ongoing clinical development of ACT—including industry-led registration trials—is predominantly pursued in patients with anti–PD-1 experienced melanoma. From previous analyses it has become evident that patients with anti–PD-1 experienced melanoma have lower response rates to ACT compared with anti–PD-1 naïve patients. The unique availability of samples and datasets from the “pre-ICI era” and the “ICI-era” at the NCI enabled the authors to explore the basis for this difference in response rate in the current study.

Neoantigens encoded by genomic tumor alterations are key targets of effective antitumor T-cell responses and are recognized by T cells mediating antitumor activity both in the context of ICI and adoptive TIL therapy. Melanoma is highly immunogenic, in part due to its high mutational burden and associated abundance of neoantigens. Investigating the role of neoantigens in mediating antitumor activity of ACT therapy is an important question in the field. The authors performed a retrospective analysis of patient data, samples, and ACT products. In their patient cohort, 150 patients with advanced melanoma received ACT (103 anti–PD-1 naïve and 47 anti–PD-1 experienced); 152 samples were available for mutational analysis, and 73 samples were assessed for TIL neoantigen reactivity.

In line with prior reports, the overall response rate was 55% in anti–PD-1 naïve and 26% in anti–PD-1 treated patients. In anti–PD-1 naïve patients, about half of the responses were complete, whereas only 8% of the responses were complete in anti–PD-1 experienced patients. To explore the underlying mechanism of these differing response rates, whole-exome sequencing was conducted to evaluate the tumor mutational burden (TMB). Tumors from anti–PD-1 naïve patients exhibited a higher median TMB compared with those from anti–PD-1 experienced patients. Higher TMB was found to be associated with response to ACT regardless of prior PD-1 therapy.

To understand the role of neoantigens in driving TIL-mediated tumor responses in the context of prior PD-1 inhibition, reactivity to neoantigens was assessed by coculturing TILs from prepared infusion products with autologous patient-derived antigen presenting cells that were either transfected with tandem minigenes expressing predicted neoantigens or pulsed with peptides encompassing neoantigens. Significantly higher recognition of neoantigens was found in the anti–PD-1 naïve group (78% of patients) compared with anti–PD-1 experienced patients (53%). This in vitro neoantigen discovery approach (in contrast to the more widely used in silico prediction) was pioneered by the group and is arguably the gold standard for the identification of relevant, immunogenic neoantigens. Limitations of this technology include the requirement of relatively large (usually surgical) tumor samples, the relatively small number of neoantigens that can be tested (ca.200), and the restriction of the discovery space to the spectrum of neoantigens recognized by endogenously primed T cells that are present in the tumor. To account for the differences in overall TMB across the study cohort, patients with similar TMBs were then compared and anti–PD-1 experienced patients were found to have fewer neoantigen-reactive TILs compared with anti–PD-1 naïve patients. Response to ACT therapy was associated with greater recognition of neoantigens in both groups. Taken together, these data suggest that neoantigen recognition by TILs is critical for the efficacy of ACT therapy, and that TILs from patients who received prior PD-1 therapy have a reduced ability to recognize these neoantigens.

In agreement with prior studies, the study highlights the importance of neoantigens as targets of effective immune responses driving efficacy of ACT therapy. Given the known association of TMB and response to anti–PD-1 therapy it is likely that the lower TMB seen in anti–PD-1 experienced patients is at least in part explained by the absence of long-term anti–PD-1 responders (with presumably higher TMB) in that group. However, the key additional observation that TIL neoantigen reactivity is also lower in these patients and that this holds up even when controlling for TMB raises the provocative question (and concern) that TIL neoantigen reactivity may be compromised by prior anti–PD-1 therapy and potentially other downstream salvage immunotherapies targeting neoantigens. The observations underscore a need to more deeply understand how T-cell populations evolve over the course of ICI.

While the study could not address this directly because no serial tumor samples were available, preclinical studies in mice have suggested that PD-1 blockade has the potential to impact the formation of memory T-cell responses (2), or lead to the development of dysfunctional T cells following suboptimal priming (3). This indicates that the function of TILs harvested following anti–PD-1 therapy could be compromised and thus these T cells may not expand as effectively in ACT preparation or target tumor neoantigens as efficiently (Fig. 1). The observations also bring to attention the implications of therapy sequencing. If the preparation of TILs for ACT is hampered by suboptimal expansion of functional tumor-specific T cells in this subset of patients, then it may not be ideal to use ACT subsequent to PD-1 blockade. Alternatively, ACT may be a more beneficial option as a combinatorial therapy alongside ICI.

The differences in neoantigen recognition can also be associated with the composition of TIL populations. Mechanistic studies in mouse models have shown that PD-1 blockade may preferentially target intra-tumoral stem-like T cells, leading to their expansion and differentiation, thus contributing to tumor control (4). Single cell sequencing analyses of TIL populations exhibiting the same markers in patients with melanoma showed correlations with response to ICI and overall survival suggesting that anti–PD-1 refractory patients to start may not have a TIL profile with a high enough frequency of these beneficial stem-like T cells. Indeed, an earlier study from the Rosenberg group thoroughly investigated the phenotypes of TILs associated with response to ACT and highlighted the importance of a “stem-like” T-cell population (5). Interestingly, these stem-like T cells had a similar transcriptional profile as T cells associated with response to ICI suggesting that a patient who has progressed following PD-1 therapy as a function of TIL composition likewise may not have optimal TILs that would similarly be beneficial for ACT.

Taken together, Levi and colleagues highlight gaps in our understanding regarding the impact of prior anti–PD-1 therapy on the activity of neoantigen-specific T cells. As these T cells are critical for tumor elimination, further studies should be aimed at elucidating the mechanisms inhibiting effective T-cell function in salvage immunotherapies. Through more in-depth studies, one could potentially identify methods to overcome these challenges and design improved next-generation ACT therapies.

E. Blass reports grants from the Cancer Immunology Training Program at Dana-Farber Cancer Institute (NIH T32 grant) during the conduct of the study. P.A. Ott reports grants and personal fees from Neon Therapeutics, Bristol-Myers Squibb, Merck, CytomX, Pfizer, Novartis, Celldex, Amgen, and Roche/Genentech; personal fees from Array; and grants from AstraZeneca/MedImmune, Armo BioSciences, Xencor, and Oncorus outside the submitted work.

E. Blass is supported by the Dana-Farber Cancer Institute Cancer Immunology Training Program (NIH T32CA207021). P.A. Ott is supported by the NCI (NCI-R01 CA229261), Team Science Award from the Melanoma Research Alliance, the Francis and Adele Kittredge Family Immuno-Oncology and Melanoma Research Fund, the Faircloth Family Research Fund, and the Susan and Bruce Hampton Research Fund.