An immune-related lncRNA pairs signature to identify the prognosis and predict the immune landscape of laryngeal squamous cell carcinoma

Abstract

Background: Laryngeal squamous cell carcinoma (LSCC) is the most common squamous cell carcinoma. Though significant effort has been focused on molecular pathogenesis, development, and recurrence of LSCC, little is known about its relationship with the immune-related long non-coding RNA (lncRNA) pairs.

Methods: After obtaining the transcriptome profiling data sets and the corresponding clinical characteristics of LSCC patients and normal samples from The Cancer Genome Atlas (TCGA) database, a series of bioinformatic analysis was conducted to select the differently expressed immune-related lncRNAs and build a signature of immune-related lncRNA pairs. Then, the effectiveness of the signature was validated.

Results: A total of 111 LSCC patients and 12 normal samples' transcriptome profiling data sets were retrieved from TCGA. 301 differently expressed immune-related lncRNAs were identified and 35,225 lncRNA pairs were established. After univariate Cox analysis, LASSO regression and multivariate Cox analysis, 7 lncRNA pairs were eventually selected to construct a signature. The riskscore was computed using the following formula: Riskscore = 0.95 × (AL133330.1|AC132872.3) + (-1.23) × (LINC01094|LINC02154) + 0.65 × (LINC02575|AC122685.1) + (-1.15) × (MIR9-3HG|LINC01748) + 1.45 × (AC092687.3|SNHG12) + (-0.87) × (AC090204.1|AL158166.1) + 0.64 × (LINC01063|Z82243.1). Patients were classified into the high-risk group (> 1.366) and the low-risk group (< 1.366) according to the cutoff value (1.366), which is based on the 5-year riskscore ROC curve. The survival analysis showed that the low-risk group had a better prognosis (P < 0.001). The riskscore was better than other clinical characteristics in prognostic prediction and the area under the curves (AUCs) for the 1-, 3-, and 5-year survivals were 0.796, 0.946, and 0.895, respectively. Combining age, gender, grade, stage, and riskscore, a nomograph was developed to predict survival probability in LSCC patients. Then, the riskscore was confirmed to be related with the content of tumor-infiltration immune cells and the model could serve as a potential predictor for chemosensitivity.

Conclusion: We successfully established a more stable signature of 7 immune-related lncRNA pairs, which has demonstrated a better prognostic ability for LSCC patients and may assist clinicians to precisely prescribe chemo drugs.

Keywords: Laryngeal squamous cell carcinoma (LSCC); Long non-coding RNA (lncRNA); Prognostic signature; The Cancer Genome Atlas (TCGA); Tumor-infiltrating immune cell.

Fig. 1

Flow chart of the study

Fig. 2

Identification of differently expressed immune-related lncRNAs and construction of a lncRNA pairs signature. A The heatmap and (B) volcano plot of the differently expressed immune-related lncRNAs. C The generalized cross-validation curve of paired likelihood deviance. D The generalized cross-validation curve of coefficients. E A forest map showed 7 differently expressed immune-related lncRNA pairs identified by Cox proportional hazard regression in the stepwise method

Fig. 3

The cut-off value and survival curve between high- and low-risk groups. A The 1-, 3-, and 5-year ROC of the optimal model. B The maximum inflection point is the cut-off point obtained by the AIC in 5-year ROC curve. C The survival curve between high- and low-risk groups. D Riskscores and survival outcome (E) of each case

Fig. 4

Riskscore was an independent factor in predicting prognosis. (A)Univariate and (B) multivariate Cox regression analysis results were showed by forest maps. (C-E) Riskscore was better than other clinical characteristics in prognostic prediction in 1,3,5- year.

Fig. 5

Correlations between the riskscore and other clinical characteristics. A strip chart (A) and the scatter diagrams showed the correlations between riskscore and age (B), gender (C), grade (D), stage (E), T(F), N(G), M(H)

Fig. 6

A nomogram was developed to predict survival probability. A A nomogram was developed in which age, gender, grade, stage and riskscore were integrated. B The calibration curve of 3-year survival. C The calibration curve of 5-year survival

Fig. 7

Exploration the relationship between tumor-infiltration immune cells and riskscore. (A) tumor-infiltration cells, such as B cell plasma, B cell, cytotoxicity score, T cell CD4 + effector memory, T cell CD8 + , T cell follicular helper, were negatively associated with the riskscore. The scatter diagrams showed that B cell plasma (B), B cell (C), cytotoxicity score (D), T cell CD4 + effector memory (E), T cell CD8 + (F, G), T cell follicular helper (H, I) were higher in low-risk group

Fig. 8

Immune related genes were differently expressed between high-and low risk groups and riskscore was a potential predictor for chemotherapeutics.VCAN (A) was highly expressed in high-risk group while TNFRSF4 (B) and TNFRSF18 (C) were highly expressed in low-risk group. Bexarotene (D) and bicalutamide (E) had lower IC50 value in high-risk group