Delineation of immunotherapeutic predictive versus prognostic transcriptional programs to identify SLC22A5-centric carnitine metabolism-driven resistance to anti-PD-L1 treatment in advanced non–small-cell lung cancer.

Abstract

2623 Background: Prognostic factors indicate the natural course of a disease regardless of treatment, whereas predictive factors determine the likelihood of response to specific therapies. Distinguishing between predictive and prognostic factors is essential for separating treatment-specific outcomes from the inherent progression of cancer, thereby guiding clinical decision-making. We aim to dissect the predictive and prognostic transcriptional programs underlying the efficacy of anti-PD-L1 versus chemotherapy in advanced non-small cell lung cancer (NSCLC) to uncover mechanisms specific to immunotherapy resistance. Methods: Clinical and baseline tumor transcriptomic data were collected from two randomized controlled trials comparing atezolizumab with docetaxel: OAK (n=697, discovery cohort) and POPLAR (n=192, validation cohort). Transcriptional program scores for each biological process and metabolic pathway from the Reactome database were calculated using gene set variation analysis for each patient. Cox regression and P-value for interaction tests were conducted to differentiate predictive versus prognostic effects of transcriptional programs. Tumor microenvironment and cell-cell communication underlying immunotherapy resistance were explored using bulk and single-cell transcriptomic data. Results: Transcriptional programs in the OAK discovery cohort were divided into four categories associated with different predictive effects specific to atezolizumab or docetaxel. Carnitine metabolism was the most prominent process contributing to atezolizumab-specific resistance, while porphyrin metabolism drove docetaxel-specific resistance. SLC22A5, the only high-affinity carnitine transporter, was upregulated in atezolizumab-resistant patients. The predictive effect of SLC22A5-centric carnitine metabolism for resistance to atezolizumab rather than docetaxel was confirmed in the POPLAR validation cohort. Integrative analyses of bulk and single-cell transcriptomes revealed that cancer cell-specific SLC22A5 expression induced M2 macrophage polarization and decreased CD8+ T cell infiltration via carnitine uptake, thus forming an immunosuppressive microenvironment. Conclusions: Our study elucidates the distinction between predictive and prognostic factors in advanced NSCLC from a metabolic perspective. Cancer cells uptake of carnitine via SLC22A5 mediates resistance to anti-PD-L1 treatment. Combining inhibition of SLC22A5-centric carnitine metabolism with anti-PD-L1 agents might be a promising strategy to reverse immune escape in advanced NSCLC. Keywords: Predictive, Prognostic, Non-small cell lung cancer, Carnitine metabolism, Resistance.