A genomewide transcriptomic approach identifies a novel gene expression signature for the detection of lymph node metastasis in patients with early stage gastric cancer

Abstract

Background: Although identification of lymph node (LN) metastasis is a well-recognized strategy for improving outcomes in patients with gastric cancer (GC), currently there is lack of availability of adequate molecular biomarkers that can identify such metastasis. Herein we have developed a robust gene-expression signature for detecting LN metastasis in early stage GC by using a transcriptome-wide biomarker discovery and subsequent validation in multiple clinical cohorts.

Methods: A total of 532 patients with pathological T1 and T2 GC from 4 different cohorts were analyzed. Two independent datasets (n = 96, and n = 188) were used to establish a gene signature for the identification of LN metastasis in GC patients. The diagnostic performance of our gene-expression signature was subsequently assessed in two independent clinical cohorts using qRT-PCR assays (n = 101, and n = 147), and subsequently compared against conventional tumor markers and image-based diagnostics.

Findings: We established a 15-gene signature by analyzing multiple high throughput datasets, which robustly distinguished LN status in both training (AUC = 0.765, 95% CI 0.667-0.863) and validation cohorts (AUC = 0.742, 95% CI 0.630-0.852). Notably, the 15-gene signature was significantly superior compared to the conventional tumor markers, CEA (P = .04) and CA19-9 (P = .005), as well as computed tomography-based imaging (P = .04).

Interpretation: We have established and validated a 15-gene signature for detecting LN metastasis in GC patients, which offers a robust diagnostic tool for potentially improving treatment outcomes in gastric cancer patients. FUND: NIH: CA72851, CA181572, CA14792, CA202797, CA187956; CPRIT: RP140784: Baylor Sammons Cancer Center polot grants (AG), VPRT: 9610337, CityU 21101115, 11102317, 11103718; JCYJ20170307091256048 (XW).

Keywords: Early stage; Gastric cancer; Gene signature; Lymph node metastasis; Prediction.

Fig. 1

Genome-wide discovery of a gene expression signature for the identification of lymph node metastasis status in early gastric cancers (GC). (A) Heatmap illustrating the expression levels of the genes expressed between patients with lymph node-positive (LNP) and lymph node-negative (LNN) gastric cancers. Of these, 84 genes were differentially expressed between LNP and LNN patients in the training dataset of 18 TCGA T1 patients. (B) ROC curve shows the diagnostic performance of the 15-gene signature for discriminating LNP from LNN TCGA T1 patients. (C) Waterfall plot shows the LN risk scores by LN status in the TCGA T1 and T2 cohort, and the ROC curve demonstrates the diagnostic performance in the expanded set of TCGA T1 and T2 patients. (D) Waterfall plot illustrates the risk scores by LN status and the its diagnostic potential in an independent set of ACRG T2 patients.

Fig. 2

Clinical validations of the 15-gene signature in identification of lymph node metastasis status in early GC. ROC curves show that our novel 15-gene signature had a higher diagnostic value for identification of LN metastasis over CEA and CA19_9 in (A) clinical cohort-1 (testing) and (B) clinical cohort-2 (validation), respectively. (C) ROC curves illustrate that the 15-gene signature had a higher performance compared to the clinical LN status determined by CT in clinical cohort-2. Comparison of AUC values were conducted by DeLong test (D) ROC curves illustrate that the combinatorial model integrating the 15-gene signature and clinical N stage further improved the predictive accuracy in clinical cohort 2.