A Serum Protein Signature Associated with Outcome after Anti–PD-1 Therapy in Metastatic Melanoma

Abstract

A mass spectrometry analysis was performed using serum from patients receiving checkpoint inhibitors to define baseline protein signatures associated with outcome in metastatic melanoma. Pretreatment serum was obtained from a development set of 119 melanoma patients on a trial of nivolumab with or without a multipeptide vaccine and from patients receiving pembrolizumab, nivolumab, ipilimumab, or both nivolumab and ipilimumab. Spectra were obtained using matrix-assisted laser desorption/ionization time of flight mass spectrometry. These data combined with clinical data identified patients with better or worse outcomes. The test was applied to five independent patient cohorts treated with checkpoint inhibitors and its biology investigated using enrichment analyses. A signature consisting of 209 proteins or peptides was associated with progression-free and overall survival in a multivariate analysis. The test performance across validation cohorts was consistent with the development set results. A pooled analysis, stratified by set, demonstrated a significantly better overall survival for "sensitive" relative to "resistant" patients, HR = 0.15 (95% confidence interval: 0.06-0.40, < 0.001). The test was also associated with survival in a cohort of ipilimumab-treated patients. Test classification was found to be associated with acute phase reactant, complement, and wound healing pathways. We conclude that a pretreatment signature of proteins, defined by mass spectrometry analysis and machine learning, predicted survival in patients receiving PD-1 blocking antibodies. This signature of proteins was associated with acute phase reactants and elements of wound healing and the complement cascade. This signature merits further study to determine if it identifies patients who would benefit from PD-1 blockade. .

Mass Spectrometry-Based Multivariate Proteomic Tests for Prediction of Outcomes on Immune Checkpoint Blockade Therapy: The Modern Analytical Approach

“…Processes with p values for association with test classification <0.05 are identified as significant. The results were consistent with previously published data: Systemic responses of the innate immune system related to acute inflammation, acute phase, and complement activation showed significant associations with the tests studied, with higher levels of these processes in the poor prognosis test classification subgroups [43,51,53]. Components of complement system, acute phase reaction, and acute inflammation are now recognized as a major regulators of cancer development and progression [58][59][60].…”

“…In the case of immunotherapies, the most relevant tumor/TME biomarker is PD-L1 expression. No significant correlation has been observed between PD-L1 expression and classifications for the tests reviewed here: in the melanoma development set the p-values of association between test classifications and PDL 1 >5% vs. <5% was 0.693 for the BDX008 test [52] and 0.704 for the ICB test [43]. There was no indication of correlation of the PIR test classifications with PD-L1 expression (cut-off 1%) in the lung cancer development cohort [47].…”

“…These tests are based on Deep MALDI spectral acquisition from pretreatment serum samples. Sample preparation, spectral acquisition and spectral processing methods were similar for all three tests and are described in detail in Weber et al, [43]. Although certain elements varied depending on each test's design goals, several aspects of the machine learning used in test development were common to all three examples.…”

“…This requirement that samples be classified as good prognosis when assessed by different views of the whole dataset allows identification of a subgroup of patients most likely to have the best outcomes. In total, information from 209 spectral regions is used for the ICB classification [43].…”

“…Similar results are also observed for 3-year survival landmark, Table 2. Both tests have been validated in multiple independent cohorts of melanoma patients [43,46,53] treated with ICIs; the BDX008 test has also shown potential utility in NSCLC patients treated with nivolumab [50].…”