Tumor-derived serum proteotypes to predict response to immune therapy in patients with limited disease small-cell lung cancer.

8103 Background: The STIMULI clinical trial (NCT02046733) tested the combination of ipilimumab and nivolumab (IPI/NIVO) after standard chemoradiotherapy in patients with limited disease small cell lung cancer (LD-SCLC). The study showed no improvement in the primary endpoint of progression-free survival (PFS) in unselected patients. Accessing tumor samples is challenging in LD-SCLC after chemoradiotherapy; hence, we analyzed serum and whole blood RNA samples as biomarkers to stratify patients for clinical benefit. Methods: We used a semi-quantitative large-scale proteomic assay (SomaScan) to analyze more than 7000 proteins. We also profiled whole-blood RNA collected before and after chemoradiotherapy and performed Bulk RNA barcoding and sequencing (BRBseq). We analyzed 131 serum samples, of which one failed quality control. 119 whole blood Paxgene RNA before chemoradiotherapy and 124 after were also analyzed. We performed unsupervised hierarchical clustering on protein and RNA samples and defined differentially expressed proteins/transcripts and pathway enrichment in specific clusters. We also performed RNA deconvolution with CibersortX. In addition, we performed organ-specific aging prediction from serum proteomics using the orange Python pipeline. Results: We identified five patient proteotype clusters. Unexpectedly, the proteotypes corresponded to three types of clinical outcomes. Clusters 1, 3, and 4 were linked with a lack of benefit from immune therapy. Patients in cluster 2 had improved outcomes from immune therapy (HR: 0.436 CI: 0.1976 to 0.9078, p: 0.0272) and likely benefited most. In contrast, patients in cluster 5 had a decreased survival (HR: 5.53, CI: 1.261 to 24.28, p: 0.0234). These patients might represent a hyper-progression phenotype on immune therapy. Protein-based pathway analysis showed that cluster 1 patients had upregulated TGF𝛽, and IL-6 signaling, whereas patients in cluster 5 activated the complement pathway. In addition, preliminary analysis of organ aging showed that the patient clusters could also be linked with differential immune aging. Together, our prototypes suggest differential proteomic states of patients' underlying responses. In contrast with the protein data, RNA analysis did not lead to outcome predictors. We compared the protein signatures to single-cell RNAseq data of SCLC and found that these proteins might originate from the tumors. Conclusions: Proteotypes based on large-scale proteomic analysis could predict both the benefit and lack thereof for IPI/NIVO therapy in limited disease small cell lung cancer. The clusters also provide clues into differential pathway activation and could serve as potential new biomarkers and targets for precision adjuvant immune therapy of SCLC and beyond.