Serum miR-626 and miR-5100 are Promising Prognosis Predictors for Oral Squamous Cell Carcinoma

Abstract

Although serum microRNAs (miRNAs) are currently being considered as promising noninvasive biomarkers for cancers, their role in the prognosis of oral squamous cell carcinoma (OSCC) has not been elucidated. Here we aimed to identify serum miRNA biomarkers that could be used as prognosis predictors of OSCC. Methods: A cohort of 260 serum miRNA samples was assessed in a three-step approach that included a screening stage, a training stage, and a testing stage. The correlation between prognosis of OSCC and the miRNAs expression was comprehensively analyzed. Results: A two-miRNA signature involving miR-626 and miR-5100 has been developed. Patients defined to be high-risk group by the two-miRNA signature had significantly shortened median survival time compared with the low-risk group. In multivariate analysis, this two-miRNA signature was independently predictive of survival, and achieved a superior predictive value compared with that of traditional clinicopathologic factors such as pathology grade as well as tumor and node metastasis (TNM) stage. An integrated prognostic model combining the TNM stage and miRNA signature displayed the highest prognostic performance (AUC value: 0.787, specificity: 0.884, sensitivity: 0.573) compared to the TNM stage-alone (AUC value: 0.630, specificity: 0.526, sensitivity: 0.733) or miRNA signature-alone model (AUC value: 0.771, specificity: 0.768, sensitivity: 0.773). In addition, we found that OSCC tumor cells not only expressed a high level of these two miRNAs, but also secreted certain miRNAs into the extracellular environment, suggesting these miRNAs may originate from tumor cells. Conclusion: In our study, we established a two-miRNA signature that was strongly and independently associated with prognosis in OSCC, and may serve as a promising prognosis predictor.

Keywords: miRNA; oral squamous cell carcinoma; prognosis; serum.

Figure 1

Schematic of study design.

Figure 2

Identification of prognosis-related serum miRNAs in the screening cohort. (A), Unsupervised hierarchical cluster analysis identified two distinct groups in the screening cohort (OSCC patients, n= 25); rows stand for miRNAs, and columns stand for patients. Clinicopathological characteristics of each patient were indicated. Differences in survival rates were found between the two groups. Tree analysis revealed that segregation of these two subgroups was accompanied by the generation of three major miRNA clusters. (B), Serum levels of nine miRNAs (red color) were statistically higher in patients who died. (unpaired t test, P< 0.01). DOI: Depth of invasion.

Figure 3

A two-miRNA signature developed in the training cohort. (A), A comparison of indicated serum miRNA levels in OSCC patients and healthy controls (unpaired t test, ns: not significant, ***P< 0.001). (B), A comparison of indicated serum miRNA levels in OSCC patients who died, OSCC patients who survived and healthy controls. (one-way ANOVA, *P< 0.05, ***P< 0.001) (OSCC patients, n= 40; healthy controls, n= 15). (C, D), Kaplan-Meier analysis of disease free survival (upper) and overall survival (below) indicated that patients having high expression of miR-626 (C) or miR-5100 (D) displayed significantly shortened MST. (E), Kaplan-Meier analysis of disease free survival (upper) and overall survival (below) showed that patients having high expression of both risk miRNAs displayed significantly decreased MST than those carrying zero or one risk miRNA. (disease free survival, P= 0.000846; overall survival, P= 0.004097). (F), Patients having higher risk-scores of the two-miRNA signature showed a shortened disease free survival (left) and overall survival (right). (disease free survival, P= 0.000007; overall survival, P= 0.00005). (G), The two-miRNA signature yielded better AUC value of 0.920 with 0.810 specificity and 0.842 sensitivity to predict survival of OSCC patient than one-single-miRNA. AUC: area under the receiver operating characteristic; CI: confidence interval; HR: hazard ratio; MST: median survival time; ns, no significant difference; OSCC, oral squamous cell carcinoma.

Figure 4

The two-miRNA signature was validated in independent cohorts. (A), Risk-scores distribution. The Y axis shows the index of risk scores. (B), Prognosis in patients. The dotted line refers to the median miRNA signature cutoff (1.509). (C), Color-gram of miRNA expression profiles; rows stand for miRNAs, and columns stand for patients. (D, E), Kaplan-Meier analysis indicated that high-risk patients in both the testing cohort (D) and all-combined cohorts (E) showed significantly shortened disease free survival (left) and overall survival (right). (P< 0.001). (F, G), Comparison of each single miRNA and the miRNA signature for DFS prediction in both the testing cohort (F) and all-combined cohorts (G). (H), Comparison of each single miRNA and the miRNA signature for recurrence prediction in all-combined cohorts. (I), Comparison of the miRNA signature alone, TNM staging system alone and an integrated model combining the TNM staging system and miRNA signature for DFS prediction. AUC: area under the receiver operating characteristic; CI: confidence interval; DFS: disease free survival; HR: hazard ratio; MST: median survival time; OSCC, oral squamous cell carcinoma; ROC: receiver operating characteristic.

Figure 5

Identification of the source of these two-risk miRNAs. (A), Comparison of serum miR-626 (left) and miR-5100 (right) in paired pre-operative and post-operative samples (n= 40). (paired t test, **P< 0.01, ***P< 0.001). (B), Correlations between the reduction of serum miR-626 (left) and miR-5100 (right) expression and tumor volume. (Spearman's rank correlation test, n= 40, P< 0.001). (C), Morphology of primary oral mucosal epithelium cells (left) (indicated by arrow) and OSCC cells (right) (indicated by arrow) growing with irradiated 3T3 J2 fibroblasts (indicated by arrow head). Scale bar, 100 μm. (D), Comparison of the miR-626 (left) and miR-5100 (right) expression levels in tumor-derived CRC lines, normal mucosa-derived CRC lines and their culture media. (paired t test, n= 8, **P< 0.01, ***P< 0.001). (E), Correlation between miR-626 (left) (P= 0.002) and miR-5100 (right) (P= 0.001) levels in tumor-derived CRC lines and their matched culture media. (Spearman's rank correlation test, n= 8). CRC: conditionally reprogrammed cell; OSCC: oral squamous cell carcinoma.