The “Tricky Business” of Identifying Mechanisms of Resistance to Anti–PD-1 Free

In this issue of Clinical Cancer Research, Ascierto and colleagues present data from a melanoma patient who received treatment with nivolumab, during which six responding and four progressing lesions were monitored and subsequently resected at “warm autopsy” (1). This unusual case with extensive tumor sampling provided the authors with an ideal opportunity to investigate and compare the genomic, transcriptomic, and immune infiltrates of responding versus progressing lesions, without interpatient variability, in an effort to identify resistance mechanisms to anti–PD-1 immunotherapy.

Currently, the most effective single-agent immunotherapies for the treatment of advanced melanoma are the anti–PD-1 immune checkpoint inhibitors pembrolizumab and nivolumab. These drugs target the programmed cell death protein 1 receptor (PD-1) that is expressed on a variety of effector immune cells but most importantly on differentiated T cells. Activation of this receptor by its ligands, PD-L1 and PD-L2, suppresses the immune cell's cytotoxicity and proliferation and leads to their dysfunction (2). Preventing the PD-L1/PD-1 interaction with anti–PD-1 mAbs has achieved response rates of up to 44% in patients with advanced melanoma. Nevertheless, 43% of initially responding patients will progress in existing or new lesions within 3 years (3). Defining mechanisms of response and resistance to anti–PD-1 therapy not only lead to new strategies that overcome resistance but also enhance initial responses to checkpoint inhibition.

The clinical presentation of resistance to immunotherapies can take three major forms: (i) primary resistance where a metastasis continues to grow throughout treatment, (ii) acquired resistance in a previously responding tumor, and (iii) a new tumor develops and continues to grow during treatment. Some of the current known resistance mechanisms include alterations in genes involved in IFN-mediated signaling (JAK1/2) or antigen presentation (B2M, TAP, HLA), T-cell exclusion from the tumor via WNT/β-catenin signaling, immune escape via epithelial–mesenchymal transition, activation of alternate immunosuppressive pathways, or release of immunosuppressive cytokines/enzymes (4–6). With this in mind, Ascierto and colleagues analyzed multiple metastases from an individual patient showing heterogeneous clinical responses to nivolumab (1). The authors utilized whole-exome sequencing, RNA expression profiling, and quantitative pathology of immune cell infiltrates to evaluate potential mechanisms of response and resistance in six responding and four progressing tumors, all at the same time point on therapy (Fig. 1).

Genomic analysis did not reveal significant differences in the phylogenic evolution, mutation landscapes, or neoantigen burden between the responding and resistant tumors. All responding and resistant tumors were driven by a biallelic NF1 truncating mutation, shared many core mutations and neoantigens, and possessed a high burden of genomic changes and few private mutations. Notably, no mutations distinguished resistant from responding tumors. It is also important to highlight that although high mutation burden and tumor clonality are reportedly associated with improved response to immune checkpoint inhibitors (7), the presence of these features was not sufficient for response to PD-1 inhibition in this patient.

Epigenetic changes and tumor microenvironment factors are known to confer resistance to targeted BRAF and MEK inhibitors in melanomas, and Ascierto and colleagues (1) suggest that they may also influence response to immune checkpoint inhibitors (8). In particular, the laminin-3 subunit, LAMA3, was highly upregulated in the resistant lesions, in the absence of genetic alterations, and laminin-3 overexpression may alter the organization of the tumor microenvironment to influence immune cell migration. LAMA3 is altered by aberrant methylation in breast cancer cell lines, and this contributes to a more invasive phenotype (9). In addition, PD-1 expression is known to be altered by epigenetic processes, and BET inhibitors have direct actions on the PD-L1/PD-1 interactions (10). Therefore, epigenomic changes may hold some clues to this patient's resistance and provide a potential alternate combination strategy for the next wave of therapeutics.

Ascierto and colleagues performed differential expression analysis of the transcriptomic profiles of responding and progressing tumors that identified 13 differentially expressed candidate genes (1). These genes may relate to extracellular matrix formation and neutrophil infiltration in the progressing metastases. However, neutrophils were not quantified in this study, and the specific differential gene set identified by Ascierto and colleagues did not predict response in an independent, previously published dataset (5). Nevertheless, an invasive mesenchymal phenotype correlates with the upregulation of immune checkpoint molecules (11) and may be associated with innate PD-1 resistance (5). Ascierto and colleagues did not identify any differentially expressed immune-related signatures, which contrasts with several other intrapatient “response versus resistant” studies that reported IFN-related genes were upregulated in responding tumors (12). It is also important to note that the accuracy and content of differential RNA signatures will depend on the quality of tumor macrodissection and will reflect the percentage of tumor and immune cells in each sample. For example, genomic-based studies aiming to maximize tumor content in the DNA/RNA extractions will provide different results compared with studies that extract DNA/RNA from the entire biopsy and include the immune–tumor interface. Therefore, it is possible that RNA signatures are impacted by the methodology used to identify areas of tumor or tumor microenvironment for analysis.

Finally, Ascierto and colleagues quantified the densities and spatial distribution of immunoregulatory markers, including PD-1 and PD-L1, and various immune cell subsets in the responding and progressing metastases (1). It has been reported that PD-L1–positive tumors (>1% or >5%) experience higher response rates to anti–PD-1 therapies compared with PD-L1–negative tumors, although PD-L1–negative patients still respond (35% have some tumor reduction compared with 57% for patients with high tumor PD-L1 expression; ref. 4). Furthermore, other studies have found increased densities of activated and proliferating cytotoxic T cells in or immediately surrounding the tumor correlated with better responses to anti–PD-1 (5, 6). Surprisingly, Ascierto and colleagues found no significant difference in immune cell subsets, inhibitory receptor, or PD-L1 expression in responding and resistant metastases. Although, the authors did not perform co-staining of proliferative or activation markers, and it remains a possibility that the immune cells present within the progressing tumors were anergic and, therefore, functional, and unidentified immune cell differences may exist in the responding and resistant tumors. Nevertheless, the simplest explanation for the lack of phenotypic differences between the responding and resistant metastases is the timing of the biopsies (Fig. 1B). At the time of on-treatment biopsy, the initially responding metastases may have acquired or been reduced to the minority resistant subclone under the selective pressures of anti–PD-1 treatment. The subpopulations of tumor cells left over in the initially responding metastasis after anti–PD-1 therapy may have acquired the resistant phenotype of their primary resistant cousins, hence the lack of differentiation. Intrapatient studies that perform biopsies at multiple time points during anti–PD-1 have found the early on-treatment (4 weeks) biopsies display the greatest divergence in immune infiltrates between the responding and nonresponding patients. Therefore, studies that include extensive longitudinal biopsies, including biopsies procured early during treatment, are necessary to better understand the dynamics and timing of resistance in such patients.

In summary, as demonstrated by the current report, the identification of resistance mechanisms to immune checkpoint inhibitors is complex and difficult. The dynamic nature of response and resistance in treated patients is difficult to monitor and classify. Indeed, the development of unique, new response criteria to accommodate the kinetics of immune response underscores this. Combination strategies are overcoming some of the resistance seen in monotherapies; however, resistance remains a problem. More studies like the one performed by Ascierto and colleagues (1) on well-characterized tumors are needed, preferably including regular longitudinal biopsies with accurate response classification at each time point, to account and track the emergence of resistance in such patients.

No potential conflicts of interest were disclosed.

Writing, review, and/or revision of the manuscript: J.S. Wilmott, H. Rizos, R.A. Scolyer, G.V. Long