Identification and validation of an immune-related prognostic signature and key gene in papillary thyroid carcinoma

Abstract

Background: Papillary thyroid carcinoma (PTC) is the most common pathological type of thyroid cancer. The effect of traditional anti-tumor therapy is not ideal for the patients with recurrence, metastasis and radioiodine resistance. The abnormal expression of immune-related genes (IRGs) has critical roles in the etiology of PTC. However, the effect of IRGs on PTC prognosis remains unclear.

Methods: Based on The Cancer Genome Atlas (TCGA) and ImmPort databases, we integrated IRG expression profiles and progression-free intervals (PFIs) of PTC patients. First, we identified the differentially expressed IRGs and transcription factors (TFs) in PTC. Subsequently, an IRG model that can predict the PFI was constructed by using univariate Cox regression, least absolute shrinkage and selection operator (LASSO) regression and multivariate Cox regression analyses of the differentially expressed IRGs in the TCGA. Additionally, a protein-protein interaction (PPI) network showed the interactions between the differentially expressed genes (DEGs), and the top 30 genes with the highest degree were extracted from the network. Then, the key IRG was identified by the intersection analysis of the PPI network and univariate Cox regression, which was verified the differential expression of by western blotting and immunohistochemistry (IHC). ssGSEA was performed to understand the correlation between the key IRG expression level and immune activity.

Results: A total of 355 differentially expressed IRGs and 43 differentially expressed TFs were identified in PTC patients. Then, eight IRGs were finally utilized to construct an IRG model. The respective areas under the curve (AUCs) of the IRG model reached 0.948, 0.820, and 0.831 at 1, 3 and 5 years in the training set. In addition, lactotransferrin (LTF) was determined as the key IRG related to prognosis. The expression level of LTF in tumor tissues was significantly lower than that in normal tissues. And the results of ssGSEA showed the expression level of LTF is closely related to immune activity.

Conclusions: These findings show that the prognostic model and key IRG may become promising molecular markers for the prognosis of PTC patients.

Keywords: Immune-related genes; PPI; Papillary thyroid carcinoma; Prognosis; Tumor infiltrating immune cells.

Fig. 1

Flow chart of the analysis process in our study

Fig. 2

Differentially expressed IRGs and TFs between PTC and normal thyroid samples and functional enrichment analysis. a, b Volcano plots of the differentially expressed IRGs and TFs. c, d GO enrichment analysis and e–f KEGG enrichment analysis of the differentially expressed IRGs and TFs

Fig. 3

Construction of the prognostic risk model and analysis. a, b LASSO regression analysis of the PFI-associated IRGs. c The hazard ratios and p-values from the multivariate Cox regression are shown in the forest plot. d Nomogram showing the PFIs at 1, 3 and 5 years of patients in the TCGA database. e, f Calibration curves of the nomogram to predict the PFIs at 1 and 3 years. g, h Two-dimensional PCA plot and three-dimensional PCA plot showing distribution in the high-risk group and low-risk group

Fig. 4

Prognostic risk model in PTC patients from the training set and the whole set. a, b Patients ranked by risk score, corresponding survival status and heatmap of the training set and the whole set. c, d Kaplan–Meier survival curve of PFIs of PTC patients in the training set and the whole set according to the median cutoff value. e, f ROC curves at 1, 3, and 5 years in the training set and the whole set

Fig. 5

Analysis of the prognostic risk model. a, b Univariate Cox regression analysis and multivariate Cox regression analysis of clinical parameters and the risk score in the whole set. c–h Correlation of the risk score with age, tumor burden, T stage, N stage, M stage and TNM stage of PTC

Fig. 6

PPI network and univariate Cox regression analysis. a PPI network of the DEGs. b Bar plot showing the top 30 genes ordered by the number of nodes. c Forest plot showing the prognosis-related IRGs screened by the univariate Cox regression analysis. d Venn diagram displaying the common genes shared by the top 30 nodes in the PPI network and prognosis-related IRGs. e Interaction network between TFs and prognosis-related IRGs. Triangles: TFs; circles: IRGs; red circles: IRGs that positively correlated with PFIs; green circles: IRGs that negatively correlated with PFIs; green line and red line indicate a negative correlation and positive correlation, respectively

Fig. 7

The differential expression of LTF and its association with survival and potential functional mechanism in PTC patients. a LTF expression in multiple tumor and normal tissues based on the TIMER database. b Differentiated expression of LTF in tumor and normal tissues from the TCGA database. c Paired analysis of LTF expression between tumor and normal tissues from the same patient in the TCGA database. d KM survival curve of PFIs in patients in the low LTF and high LTF expression groups in the TCGA database. e Diagnostic efficacy of the ROC curve of LTF. f Methylation level of LTF according to UALCAN. g LTF protein expression levels in the normal thyroid follicular epithelial cell line Nthy-ori 3–1 and PTC cell lines. h Examples of IHC staining of LTF in PTC tissues and adjacent noncancerous tissues. i Comparison of LTF protein expression in 30 pairs of matched paraffin section samples by IHC

Fig. 8

GSEA and ssGSEA scores between the different LTF expression groups. a Enriched gene sets in the C7 collection in the high LTF expression group. b Enriched gene sets in the HALLMARK collection in the high LTF expression group. c Boxplots showing the scores of 16 immune cells in the different LTF expression groups. d Boxplots showing the scores of 13 immune-related functions in the different LTF expression groups. The p-values were uniformly replaced with the following symbols: *p < 0.05; **p < 0.01; ***p < 0.001