Tumor-Infiltrating PD-1^hiCD8⁺-T-Cell Signature as an Effective Biomarker for Immune Checkpoint Inhibitor Therapy Response Across Multiple Cancers

Background: Stratification of patients who could benefit from immune checkpoint inhibitor (ICI) therapy is of much importance. PD-1^hiCD8⁺ T cells represent a newly identified and effective biomarker for ICI therapy response biomarker in lung cancer. Accurately quantifying these T cells using commonly available RNA sequencing (RNA-seq) data may extend their applications to more cancer types.

Method: We built a transcriptome signature of PD-1^hiCD8⁺ T cells from bulk RNA-seq and single-cell RNA-seq (scRNA-seq) data of tumor-infiltrating immune cells. The signature was validated by flow cytometry and in independent datasets. The clinical applications of the signature were explored in non-small-cell lung cancer, melanoma, gastric cancer, urothelial cancer, and a mouse model of breast cancer samples treated with ICI, and systematically evaluated across 21 cancer types in The Cancer Genome Atlas (TCGA). Its associations with other biomarkers were also determined.

Results: Signature scores could be used to identify the PD-1^hiCD8⁺ T subset and were correlated with the fraction of PD-1^hiCD8⁺ T cells in tumor tissue (Pearson correlation, R=0.76, p=0.0004). Furthermore, in the scRNA-seq dataset, we confirmed the capability of PD-1^hiCD8⁺ T cells to secrete CXCL13, as well as their interactions with other immune cells. In 581 clinical samples and 204 mouse models treated with ICIs, high signature scores were associated with increased survival, and the signature achieved area under the receiver operating characteristic curve scores of 0.755 (ranging from 0.61 to 0.91) in predicting therapy response. In TCGA pan-cancer datasets, our signature scores were consistently correlated with therapy response (R=0.78, p<0.0001) and partially explained the diverse response rates among different cancer types. Finally, our signature generally outperformed other mRNA-based predictors and showed improved predictive performance when used in combination with tumor mutational burden (TMB). The signature score is available in the R package "PD1highCD8Tscore" (https://github.com/Liulab/PD1highCD8Tscore).

Conclusion: Through estimating the fraction of the PD-1^hiCD8⁺ T cell, our signature could predict response to ICI therapy across multiple cancers and could serve as a complementary biomarker to TMB.