UCSF

BACKGROUND: The treatment of esophageal squamous cell carcinoma (ESCC) patients varies considerably depending upon whether lymph node metastasis (LNM) is present. Patients with ESCC can particularly benefit from neoadjuvant therapy if LNM is accurately diagnosed before surgery. Long noncoding RNA (lncRNA) has been confirmed to be closely related to the development of metastases in ESCC, but much remains unknown regarding the relationship between LNM and lncRNA. The purpose of our study was to investigate relationship between LNM and lncRNA, and create a diagnostic model for predicting LNM in ESCC before surgery. METHODS: We used quantitative real-time polymerase chain reaction (qRT-PCR) to detect the expression of LINC02381. We also verified the in vitro effect of LINC02381 on the growth and metastasis of ESCC in the KYSE510 and KYSE180 cell lines. We used the Kaplan-Meier (KM) method and the log-rank test to confirm the differences of overall survival (OS) and disease-free survival (DFS) in LINC02381 expression. We used univariate and multivariate logistic regression analyses to screen for clinical characteristics and assessed their clinical diagnostic efficacy using receiver operating characteristic (ROC) curves. The model was validated with the area under the curve (AUC) and calibration curves and visualized through a nomogram. RESULTS: qRT-PCR suggested a significant elevation of LINC02381 expression in ESCC tissues compared with normal esophageal epithelial tissues (P<0.001) and in ESCC tissues with LNM (P<0.001). Analysis of OS and DFS indicated that the high expression of LINC02381 and lymph node positivity were associated with poor prognosis. Combined analysis showed that patients with both a high expression of LINC02381 and lymph node positivity had the worst prognosis. High expression of LINC02381 was associated with poor differentiation, tumor-node-metastasis (TNM) staging, and LNM in ESCC. Presence of LNM was also closely associated with tumor differentiation and primary tumor staging. Univariate and multivariate logistic regression analyses identified that primary tumor staging, tumor differentiation, and LINC02381 expression were independent influencing factors. In the ROC curve analysis of the risk model, the AUC for LINC02381 expression was 0.822 and increased to 0.913 when primary tumor staging and tumor differentiation were added. We further conducted calibration curve analysis to display the calibration of our final model. A nomogram was used to display the predictive variables. The in vitro experiments demonstrated that the knockdown of LINC02381 could inhibit the growth and metastasis of ESCC. CONCLUSIONS: LINC02381 may serve as a biomarker for predicting LNM. Our risk model can assist in predicting LNM in clinical practice, inform the decision to implement neoadjuvant therapy before surgery, and therefore improve prognosis.