Identification and validation of a novel senescence-related biomarker for thyroid cancer to predict the prognosis and immunotherapy

Abstract

Introduction: Cellular senescence is a hallmark of tumors and has potential for cancer therapy. Cellular senescence of tumor cells plays a role in tumor progression, and patient prognosis is related to the tumor microenvironment (TME). This study aimed to explore the predictive value of senescence-related genes in thyroid cancer (THCA) and their relationship with the TME.

Methods: Senescence-related genes were identified from the Molecular Signatures Database and used to conduct consensus clustering across TCGA-THCA. Differentially expressed genes (DEGs) were identified between the clusters used to perform multivariate Cox regression and least absolute shrinkage and selection operator regression (LASSO) analyses to construct a senescence-related signature. TCGA dataset was randomly divided into training and test datasets to verify the prognostic ability of the signature. Subsequently, the immune cell infiltration pattern, immunotherapy response, and drug sensitivity of the two subtypes were analyzed. Finally, the expression of signature genes was detected across TCGA-THCA and GSE33630 datasets, and further validated by RT-qPCR.

Results: Three senescence clusters were identified based on the expression of 432 senescence-related genes. Then, 23 prognostic DEGs were identified in TCGA dataset. The signature, composed of six genes, showed a significant relationship with survival, immune cell infiltration, clinical characteristics, immune checkpoints, immunotherapy response, and drug sensitivity. Low-risk THCA shows a better prognosis and higher immunotherapy response than high-risk THCA. A nomogram with perfect stability constructed using signature and clinical characteristics can predict the survival of each patient. The validation part demonstrated that ADAMTSL4, DOCK6, FAM111B, and SEMA6B were expressed at higher levels in the tumor tissue, whereas lower expression of MRPS10 and PSMB7 was observed.

Discussion: In conclusion, the senescence-related signature is a promising biomarker for predicting the outcome of THCA and has the potential to guide immunotherapy.

Keywords: cellular senescence; immunotherapy; prognosis; signature; thyroid cancer; tumor microenvironment.

Figure 1

Identification of the three senescence clusters. (A) Consensus CDF in consistent clustering (k = 2–9). (B) Relative change in area under the CDF curve from k 2–9. (C) Tracking plot of the THCA samples (K = 2–9). (D) Consensus heatmap defining the three clusters (k = 3). (E) K-M survival analysis showing significant prognosis between the three clusters. (F) Boxplot presenting the significant expression difference of immune checkpoints between the three clusters. ns, no significance. * indicated P<0.05; ** indicated P<0.01; *** indicated P<0.001.

Figure 2

Immune cell infiltration in the three clusters. Violin plot showing CD8 T cells, cytotoxic lymphocytes, endothelial cells, fibroblasts, monocytic lineage, myeloid dendritic cells, neutrophils, T cells, and NK cells.

Figure 3

Identification and validation of the senescence-related signature. (A) Univariate Cox regression analysis identifying 23 prognostic DEGs. (B) Coefficients of the LASSO analysis. (C) The senescence-related signature obtained six prognostic genes with a minimum lambda value. (D–F) K-M survival analysis showing a significant survival difference between low- and high-risk THCA across the TCGA-all, TCGA-training, and TCGA-test subsets. (G–I) ROC analysis showing the stable prediction ability of the senescence-related signature across TCGA-all, TCGA-training, and TCGA-test subsets.

Figure 4

Stability of the senescence-related signature and construction of a nomogram. (A–C) Survival curve of the THCA patients across TCGA-all, TCGA-training, and TCGA-test subsets. (D–F) Survival status of the THCA patients across TCGA-all, TCGA-training, and TCGA-test subsets. (G–I) Heatmaps showing the expression of signature genes in THCA patients across the TCGA-all, TCGA-training, and TCGA-test subsets. (J–L) PCA showing the perfect separation of low- and high-risk THCA across the TCGA-all, TCGA-training, and TCGA-test subsets. (M) The nomogram constructed with the senescence-related signature, age, gender, T stage, M stage, N stage, and clinical stage. (N) The calibration curve used to estimate the prediction accuracy of the nomogram. (O) Multi-index ROC curve of the senescence-related signature and other factors.

Figure 5

Correlation analysis showing that the senescence-related signature is associated with age, gender, T stage, N stage, M stage, and clinical stage.

Figure 6

K-M survival analysis presenting the significance of prognosis between low- and high-risk THCA in subgroups of age > 65, female, male, N0, N1, T1–2, T3–4, clinical stage I–II, and clinical stage III–IV.

Figure 7

Immune cell infiltration pattern in low- and high-risk THCA. (A) Correlation analysis of risk score and diverse immune cells using the XCELL, TIMER, QUANTISEQ, MCPCOUNTER, EPIC, CIBERSORT-ABS, and CIBERSORT algorithms. (B) Boxplot showing the expression difference of immune checkpoints in low- and high-risk THCA. (C) Violin plot showing infiltration of CD8 T cells, cytotoxic lymphocytes, endothelial cells, fibroblasts, monocytic lineage, myeloid dendritic cells, neutrophils, NK cells, and T cells in low- and high-risk THCA. ns, no significance. * indicated P<0.05; ** indicated P<0.01; *** indicated P<0.001.

Figure 8

Immunotherapy response of low- and high-risk THCA. (A) Difference of the TIDE, Exclusion, Dysfunction, CD274, IFNG, Responder, Merck18, and MDSC score between low- and high-risk THCA. (B) IPS score of the low- and high-risk THCA in the of CTLA4- PD1-, CTLA4- PD1+, CTLA+ PD1-, and CTLA+ PD1+ subgroups. * indicated P<0.05; ** indicated P<0.01; *** indicated P<0.001.

Figure 9

Functional enrichment analysis of low- and high-risk THCA. (A) GO enrichment results across TCGA-THCA including BP, CC, and MF analysis. (B) KEGG enrichment results showing the top related pathways across TCGA-THCA. (C) GSEA identifying the top five gene sets in low- and high-risk THCA.

Figure 10

Drug sensitivity in low- and high-risk THCA, including AKT inhibitor VIII, GSK1070916, rapamycin, 5-fluorouracil, bleomycin, crizotinib, doxorubicin, erlotinib, and gemcitabine.

Figure 11

Expression of the signature gene. (A) Gene expression differences across TCGA dataset. (B) Gene expression differences across GSE33630. (C) RT-qPCR verifying the gene transcription in tumor and normal cells. * indicated P<0.05; ** indicated P<0.01; *** indicated P<0.001.