Integrative single-cell and bulk transcriptome analyses reveal a carnitine-related prognostic signature and fibroblast plasticity in HNSCC

This study constructed a six-gene risk model based on carnitine-related genes for predicting outcomes in head and neck squamous cell carcinoma, with some associations to immune cell infiltration.

Summary and findings

This study identified a six-gene carnitine-related gene (CRG) risk model for head and neck squamous cell carcinoma (HNSCC) using transcriptome analyses. The model achieved area under the curve (AUC) values of 0.65, 0.66, and 0.65 for 1-, 3-, and 5-year predictions in the training set, and 0.70, 0.64, and 0.60 in external validation. The study also reported decreased CD8+ T cells and regulatory T cells in the high-risk group.

Abstract

<title>Abstract</title> <p>Background Dysregulated carnitine metabolism contributes to tumor metabolic reprogramming, but the prognostic value of carnitine-related genes (CRGs) in head and neck squamous cell carcinoma (HNSCC) remains unclear. Methods Differentially expressed genes (DEGs) were identified in The Cancer Genome Atlas (TCGA)-HNSCC cohort and intersected with CRGs to obtain candidate genes. Prognostic genes were screened by univariate and multivariate Cox regression to construct a risk model, evaluated by time-dependent receiver operating characteristic (ROC) curves and Kaplan-Meier (KM) analysis, and externally validated in GSE41613. CIBERSORT was used to compare immune cell infiltration between the high-risk group (HRG) and low-risk group (LRG), and single-cell RNA sequencing (scRNA-seq) data (GSE181919) were analyzed to identify the key cell type associated with prognostic gene expression. Results Intersection of 4114 DEGs with 91 CRGs yielded 26 candidate genes. Six prognostic genes (ACADL, SERPINA5, PRKAA2, ACACB, SLC25A20, GPR12) were identified; the resulting risk model achieved 1-, 3-, and 5-year area under the curve (AUC) values of 0.65, 0.66, and 0.65 in the training set and 0.70, 0.64, and 0.60 in external validation, remaining an independent prognostic factor (P < 0.001). CD8 + T cells and regulatory T cells were significantly decreased in the HRG (P < 0.01). Single-cell analysis identified fibroblasts as the dominant signaling hub; five prognostic genes were predominantly expressed in fibroblasts and progressively downregulated along pseudotime. Conclusions A six-gene CRG-based risk model was constructed and externally validated for HNSCC prognosis, showing associations with immune infiltration and fibroblast-mediated microenvironment remodeling, offering a potential reference for prognostic stratification.</p>

Limitations

Not reported in abstract.

Elsewhere in the L-Carnitine (Levocarnitine) corpus