PRECISE: A prognostic thyroid follicular cell-derived gene signature for papillary thyroid carcinoma

Purpose: Papillary thyroid carcinoma (PTC) exhibits heterogeneous behavior. There is a need for effective biomarkers and improved risk stratification methods in PTC. Gene signatures derived from bulk RNA sequencing capture composite transcriptional profiles that are confounded by tumor microenvironment cells. We developed a thyroid follicular cell-derived gene signature, P rognostic R NA E xpression C ell-specific I ntegrated S ignaturE (PRECISE), using single-cell and nucleus RNA sequencing techniques and evaluated its prognostic utility across independent cohorts. Experimental Procedures: PRECISE was developed using a discovery cohort of PTC patients (MDACC n=109) with a median follow-up of 14 years. Within the cohort, 11 PTC tumors and 4 normal thyroid samples were successfully sequenced using single-nucleus RNA sequencing. Genes that were significantly downregulated in malignant cells compared to normal thyroid follicular cells were integrated with previously identified differentially expressed genes from single-cell RNA sequencing. Prognostic significance was assessed using bulk RNA sequencing in the discovery cohort and validated in 2 additional cohorts (VUMC n=65; TCGA n=370). A rank-based single-sample method was used for score calculation. Associations between PRECISE score and progression-free (PFS) and disease-specific survival (DSS) were evaluated using multivariate Cox proportional hazards models, and predictive models were compared using Harrell's C-statistic and ANOVA. Results: PRECISE comprised of 41 genes downregulated in PTC tumor cells and upregulated in normal follicular cells, capturing dysregulation of thyroid function and metabolism pathways. Higher PRECISE score was significantly associated with shorter PFS across all 3 cohorts (MDACC HR=1.64, P =0.002; VUMC HR=2.54, P <0.001; TCGA HR=1.63, P =0.012) and remained significant after TNM stage adjustment in two cohorts with ≥5 years follow-up (MDACC aHR=1.42, P =0.038; VUMC aHR=2.12, P =0.024). PRECISE score was also associated with DSS in these cohorts (MDACC HR=4.16, P <0.001; VUMC HR=2.23, P =0.01), independent of TNM stage in the MDACC cohort (aHR=2.83, P =0.015). Incorporating PRECISE significantly improved predictive performance for PFS (MDACC P =0.031; VUMC P =0.024) and DSS (MDACC P =0.008) beyond stage-based models. Conclusions: PRECISE is a thyroid epithelial gene-derived signature with independent prognostic value in PTC.