Development and Validation of an Immune-Related Signature for the Prediction of Recurrence Risk of Patients With Laryngeal Cancer

Abstract

Objective: Our purpose was to develop and verify an immune-related signature for predicting recurrence risk of patients with laryngeal cancer.

Methods: RNA-seq data of 51 recurrence and 81 non-recurrence laryngeal cancer samples were downloaded from TCGA database, as the training set. Microarray data of 34 recurrence and 75 non-recurrence cancer samples were obtained from GEO dataset, as the validation set. Single factor cox regression was utilized to screen prognosis-related immune genes. After LASSO regression analysis, an immune-related signature was constructed. Recurrence free survival (RFS) between high- and low- recurrence risk patients was presented, followed by ROC. We also evaluated the correlation between immune infiltration and the signature using the CIBERSORT algorithm. The genes in the signature were validated in laryngeal cancer tissues by western blot or RT-qPCR. After RCN1 knockdown, migration and invasion of laryngeal cancer cells were investigated.

Results: Totally, 43 prognosis-related immune genes were identified for laryngeal cancer. Among them, eight genes were used for constructing a prognostic signature. High risk group exhibited a higher recurrence risk than low risk group. The AUC for 1-year was separately 0.803 and 0.715 in the training and verification sets, suggesting its well efficacy for predicting the recurrence. Furthermore, this signature was closely related to distinct immune cell infiltration. RCN1, DNAJA2, LASP1 and IBSP were up-regulated in laryngeal cancer. RCN1 knockdown restrained migrated and invasive abilities of laryngeal cancer cells.

Conclusion: Our findings identify a reliable immune-related signature that can predict the recurrence risk of patients with laryngeal cancer.

Keywords: immune; laryngeal cancer; prognosis; recurrence; signature.

Figure 1

The LASSO coefficient profiles of immune-related genes. The optimal number of variables was determined using LASSO regression analysis. (A) 10-fold cross verification for selecting lambda in the LASSO model on the basis of the minimum criteria. (B) The optimal number of variables was determined using LASSO regression analysis.

Figure 2

Evaluation of the predictive efficacy of the signature for laryngeal cancer recurrence in the training set. (A) Heat maps depicting the expression patterns of these eight genes including RCN1, LAT2, DAPK2, DNAJA2, FUZ, LASP1, IBSP and HOOK2 between high- and low- risk score group. Red: up-regulation and blue: down-regulation. (B) Ranking of risk scores among all laryngeal cancer patients. (C) Distribution of recurrence time among all laryngeal cancer patients. Red dots express recurrence laryngeal cancer samples and blue dots represent non-recurrence samples. The dotted line indicates the optimal cutoff value of risk score. The left side of the line represents low-risk patients, while the right side represents high-risk patients. (D) Recurrence-free survival between high- and low- risk score groups. (E) ROC curve for predicting the patients’ 1-year recurrence. (F) ROC curves for predicting the patients’ 3 and 5-year recurrence. (G) ROC curves for prediction of 1, 3- and 5-year OS outcome.

Figure 3

Validation of the predictive efficacy of the signature for predicting recurrence of laryngeal cancer in the verification set. (A) Heat maps visualizing the differences expression patterns of these eight genes including RCN1, LAT2, DAPK2, DNAJA2, FUZ, LASP1, IBSP and HOOK2 between high- and low- risk score group. Red expresses up-regulation and blue indicates down-regulation. (B) Ranking of risk scores among all laryngeal cancer patients. (C) Distribution of recurrence time among all laryngeal cancer patients. Red dots express recurrence laryngeal cancer samples and blue dots represent non-recurrence samples. The dotted line indicates the optimal cutoff value of risk score. The left side of the line represents low-risk patients, while the right side represents high-risk patients. (D) Recurrence-free survival between high- and low- risk score groups. (E) ROC curve for predicting the patients’ 1-year recurrence. (F) ROC curves for predicting the patients’ 3 and 5-year recurrence.

Figure 4

Immune cell infiltration analysis. (A) The distribution of 22 kinds of immune cells between high- and low- risk groups via the CIBERSORT algorithm. (B) Heatmap showing the interaction between immune cells and the risk score among laryngeal cancer patients.

Figure 5

Violin plots showing the relationship between the signature and infiltration levels of 22 kinds of immune cells. (A) T cells CD4 memory resting; (B) NK cells; (C) macrophage M0; (D) T cells follicular helper; (E) T cells regulatory; (F) T cells; (G) plasma cells; (H) dendritic cells; (I) NK cells; (J) mast cells activated. Red indicates high recurrence risk group and blue indicates low recurrence risk group.

Figure 6

Evaluation of the clinical predictive efficacy of the signature and drug sensitivity. (A, B) Comparison of the predictive efficacy between (A) the signature and (B) grade in the training set in accordance with ROC curves at 1-, 3- and 5-year recurrence. (C, D) Comparison of the predictive efficacy between (C) the signature and (D) grade in the validation set based on ROC curves at 1-, 3- and 5-year recurrence. (E) Construction of a prognostic nomogram for predicting patients’ 1-, 3- and 5-year recurrence. (F) Heatmap showing the difference in drug sensitivity between high- and low-risk groups.

Figure 7

Validation of the genes in the prognostic signature. (A–I) Western blot for detection of the protein expression of (B) RCN1, (C) DNAJA2, (D) LASP1, (E) IBSP, (F) LAT2, (G) FUZ, (H) HOOK2, and (I) DAPK2 in laryngeal cancer and normal tissues. (J–Q) RT-qPCR for detecting the mRNA expression of (J) RCN1, (K) DNAJA2, (L) LASP1, (M) IBSP, (N) LAT2, (O) FUZ, (P) HOOK2, and (Q) DAPK2 in laryngeal cancer and normal tissues. Ns, not significant; *P < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Figure 8

RCNA1 knockdown lowers invasion of laryngeal cancer cells. (A–C) Western blot for detection of RCNA1 expression in TU686 and TU212 cells transfected with siRCNA1. (D–F) Transwell for detection of the invasion of TU686 and TU212 cells transfected with siRCNA1. *P < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Figure 9

RCNA1 knockdown restrains migration of laryngeal cancer cells. (A) Representative images of wound healing assay. (B, C) Quantification of wound distance of TU686 and TU212 cells transfected with siRCNA1. *P < 0.05; **p < 0.01.