In this trial of programmed cell death-1 (PD-1) blockade with toripalimab in previously treated Chinese patients with melanoma, unique histologic and molecular features may explain why the objective response rate is lower than those defined in Western populations. This work suggests future avenues for investigating mechanisms of melanoma formation and resistance to PD-1 blockade.
See related article by Tang et al., p. 4250
In this issue of Clinical Cancer Research, Tang and colleagues (1) report molecularly annotated efficacy data of programmed cell death-1 (PD-1) immune checkpoint inhibition in melanomas arising in patients from China. Since the initial approval of pembrolizumab and nivolumab in 2014–15, the clinical development of PD-1 blockade has revolutionized the treatment of advanced melanoma. These have uniformly demonstrated long-term objective response rates of approximately 45%, with >40% surviving greater than 5 years. Several biomarkers of response to PD-1 blockade in melanoma have been identified, including tumor and stromal expression of PD-1 ligand (PD-L1) and increasing tumor mutational burden (TMB; ref. 2). Many patients with melanomas that lack PD-L1 and high TMB can nonetheless respond to PD-1 blockade; this has led to the orthogonal development of expression-based markers to identify tumors that are “T-cell inflamed.” For example, tumor samples expressing higher levels of antigen presentation genes and an 18-gene signature of IFNγ signaling are associated with efficacy of PD-1 blockade independent of TMB and PD-L1 (3, 4).
The efficacy of PD-1 blockade cannot be generalized to melanoma worldwide. The pembrolizumab and nivolumab cohorts were largely accrued from North America, Europe, and Australia, and where Caucasians with sun-exposed cutaneous melanomas dominate accrual. In these trials, “rare melanomas” arising from the acral lentiginous surfaces of the palms, soles, and nailbeds or the mucosal surfaces of the gastrointestinal or genitourinary tracts represent approximately 10% of the overall cohort. The activity of PD-1 blockade in “rare melanomas” was described in smaller subgroup analyses of prospective trials and retrospective case series. In one series of acral and mucosal melanoma, for example, the PD-1 response rate was estimated to be in the range of 23%–32% (5). It has long been recognized that these subsets are molecularly distinct (6), with a lower rate of BRAF V600 alterations, TMB, and higher rate of copy-number alterations (CNA) all potentially contributing to poorer outcomes. Many clinical trials enrolling patients with melanoma refractory to PD-1 blockade specifically exclude them from enrollment. The lack of rigorous prospective data for PD-1 blockade in noncutaneous melanomas belies the fact that, globally, these rare melanomas are not all that rare. In a recent large series from China, for example, acral (43%) and mucosal melanomas (23%) were more common than cutaneous melanomas (22% ref. 7).
In this trial, 127 Chinese patients with advanced melanoma received toripalimab, a novel IgG-4 mAb against PD-1, after progression on prior systemic therapy. Patient demographics were distinct from those enrolled in prior registrational trials in several key ways. This was a relatively heavily pretreated cohort, with nearly half of patients receiving three or more prior lines of therapy. The prior therapies vary from Western populations, with a much higher rate of prior chemotherapy and only 9% receiving prior ipilimumab. Forty percent of patients had acral lentiginous primaries, 17% had mucosal primaries, 20% had an unknown primary, and only 23% had nonacral cutaneous melanomas. Of the 110 samples available for IHC analysis of PD-L1, only 26 (23%) were positive, defined as ≥1% staining on tumor cells; this contrasts with the nivolumab registration trial, CheckMate-067, which had a positivity rate of 58% using a similar definition.
With these demographics, it is perhaps unsurprising that the overall response rate (ORR) for the entire cohort was 17%. Looking more closely, however, we see some clear similarities to prior Caucasian cohorts. Among patients with nonacral cutaneous melanomas or unknown primary melanomas (which are almost always regressed cutaneous melanomas), response rates were 31% and 23%, respectively. Noting limitations of cross-trial comparisons, this appears similar to the established activity of nivolumab (27%) and pembrolizumab (25%) in the post-ipilimumab setting (8, 9). Among 50 patients with acral melanoma, the ORR was 14%. Interestingly, among patients with mucosal melanoma, zero patients responded. This may reflect the heavily pretreated population but also possible molecular and immunologic differences between Asian and Caucasian populations.
As part of this work, a subset of patients underwent whole-exome sequencing (N = 98), whole-transcriptome sequencing (N = 46), or both (N = 39). These have been shared in a public repository to enable collaborative work (https://bigd.big.ac.cn/gsa-human/s/IlFfvCn6). For the first time, the field can begin to investigate genomic similarities and differences between melanomas treated in China versus North America, Europe, and Australia, and how these molecular findings might associate with differences in patient outcomes.
These molecular correlative studies are critical as we consider the emergence of novel immunotherapy agents from non-Western populations. To compare the efficacy of agents like toripalimab against nivolumab and pembrolizumab, it is important to account for variables known to influence progression-free survival and overall survival with PD-1 blockade, such as IFNγ-associated gene expression and TMB. In addition, understanding genomic differences between Chinese and Western populations will better illuminate fundamental mechanisms of melanoma progression and cancer evolution in general.
Several molecular distinctions can be seen from the initial comparison of genomic data between this cohort (Fig. 1A) and a recent publication by Liu and colleagues (ref. 4; Fig. 1B). Overall, as expected, the mutational burden appears much lower in the Chinese cohort. Similarly, mutation rates in select genes involved in the MAPK signaling pathway vary by population. While BRAF V600 alterations are similar in both groups, NRAS Q61 and NF1 alterations appear less common in the Chinese dataset. Markers of preexisting T-cell inflammation, including rates of expression of antigen presentation and a modified version of the 18-gene T-cell–inflamed gene expression profile, appear less frequent in the subset of Chinese patients with available data versus their Western counterparts.
This allows us to generate informed hypotheses on the lower efficacy of PD-1 blockade in the Chinese versus Western population. The combination of lower rates of T-cell inflammation, antigen presentation machinery, and TMB may mean melanomas arising in Chinese patients are intrinsically less responsive to PD-1 blockade as monotherapy. Further investigation of combination PD-1 strategies is warranted. Other host-specific factors like HLA type and microbiome composition that vary by ethnic and geographic background may also influence outcomes like severity of immune-related adverse events and response to PD-1 blockade.
Of note, not all relevant genetic information could be gathered in Fig. 1. CNA data, for example, were not available for inclusion, and the expression data are normalized only within a cohort, not across the two cohorts. The bioinformatics pipelines also vary by cohort, which likely affects detection of nucleotide variants and CNAs. More work is needed to directly compare these cohorts and may lead to interesting future lines of inquiry. For example, if coding variants in the MAPK pathway are indeed much less frequent in melanomas diagnosed in China, protein-based comparisons of MAPK activation would be required. This could elucidate alternate mutational drivers or epigenetic mechanisms of melanoma formation.
Melanoma therapeutics have been at the forefront of cancer drug development and immuno-oncology particularly. The authors are applauded for placing their data into a public repository and it is critical to continue to increase the diversity of available clinically annotated genomic datasets from around the world. Having the ability to interrogate tumor samples that reflect greater ethnic diversity of disease will accelerate the pace of improving treatment outcomes, regardless of which continent patients call home.
Disclosure of Potential Conflicts of Interest
A.N. Shoushtari reports personal fees and non-financial support from Bristol-Myers Squibb and Immunocore, personal fees from Castle Biosciences, and non-financial support from Xcovery outside the submitted work. J.J. Luke reports consultant fees from TTC Oncology, 7 Hills, Alphamab Oncology, Pyxis, Spring Bank, Tempest, AbbVie, Akrevia, Algios, Array, Astellas, Bayer, Bristol-Myers Squibb, Cstone, Eisai, EMD Serono, Ideaya, Incyte, Janssen, Merck, Mersana, Novartis, PTx, RefleXion, Regeneron, Rubius, Silicon, Tesaro, and Vividion, equity in Actym, Alphamab Oncology, Arch Oncology, Kanaph, Mavu (now part of AbbVie), Onc.AI, Pyxis, and Tempest, research support (all to institution for clinical trials unless noted) from AbbVie, Agios (IIT), Array (IIT), Astellas, Bristol-Myers Squibb, CheckMate (SRA), Compugen, Corvus, EMD Serono, Evelo (SRA), Five Prime, FLX Bio, Genentech, Immatics, Immunocore, Incyte, Leap, MedImmune, Macrogenics, Necktar, Novartis, Palleon (SRA), Merck, Spring Bank, Tesaro, Tizona, and Xencor, travel expenses from Akrevia, Bayer, Bristol-Myers Squibb, EMD Serono, Incyte, Janssen, Merck, Mersana, Novartis, Pyxis, and RefleXion, and the following patents (both provisional): Serial #15/612,657 (Cancer Immunotherapy), PCT/US18/36052 (Microbiome Biomarkers for Anti-PD-1/PD-L1 Responsiveness: Diagnostic, Prognostic and Therapeutic Uses Thereof). No potential conflicts of interest were disclosed by the other author.
A.N. Shoushtari acknowledges NCI Core Grant P30 CA008748.