BackgroundImmune checkpoint blockade (ICB) has significantly prolonged survival of non-small cell lung cancer (NSCLC) patients, although most patients develop mechanisms of resistance. Recently single-cell RNA-sequencing (scRNA-Seq) revealed a huge T-cell phenotypic and (dys)functional state variability. Accordingly, T-cell exhaustion is recognized as a functional adaptation, with a dynamic progression from a long-lived “pre-exhausted stem-like progenitor” to a “terminally exhausted” state. In this scenario it is crucial to understand the complex interplay between co-stimulatory and inhibitory molecules in CD8+ T-cell functionality.MethodsTo gain a baseline landscape of the composition, functional states, and transcriptomic signatures predictive of prognosis, we analyzed CD8+ T-cell subsets characterized by the presence/absence of PD1 and CD28 from periphery, adjacent non-tumor tissue and tumor site of a cohort of treatment-naïve NSCLC patients, by integrated multiparametric flow cytometry, targeted multi-omic scRNA-seq analyses, and computational pipelines.ResultsDespite the increased PD1 levels, an improved PD1+CD28+ T-cell polyfunctionality was observed with the transition from periphery to tumor site, associated with lack of TIGIT, TIM-3 and LAG-3, but not with Ag-experienced-marker CD11a. Differently from CD28+ T cells, the increased PD1 levels in the tumor were associated with reduced functionality in PD1+CD28− T cells. CD11ahigh, although expressed only in a small fraction of this subset, still sustained its functionality. Absence of TIGIT, TIM-3 and CTLA-4, alone or combined, was beneficial to CD28− T cells. Notably, we observed distinct TRM phenotypes in the different districts, with CD28+ T cells more capable of producing TGFβ in the periphery, potentially contributing to elevated CD103 levels. In contrast CD28− TRM mainly produced CXCL13 within the tumor.ScRNA-seq revealed 5 different clusters for each of the two subsets, with distinctive transcriptional profiles in the three districts. By interrogating the TCGA dataset of patients with lung adenocarcinoma (LUAD) and metastatic NSCLC treated with atezolizumab, we found signatures of heterogeneous TRM and "pre-exhausted" long-lived effector memory CD8+ T cells associated with improved response to ICB only in the presence of CD28.ConclusionsOur findings identify signatures able to stratify survival of LUAD patients and predict ICB response in advanced NSCLC. CD28 is advocated as a key determinant in the signatures identified, in both periphery and tumor site, thus likely providing feasible biomarkers of ICB response.Graphical