BACKGROUND: Siewert's topographic classification (SW) has been used for treatment plan of esophagogastric junction adenocarcinomas (EGJA). But recently, molecular classifications categorized EGJA with similar characteristics and prognosis. It has been discussed whether SW3 tumors would have similar features as gastric cancer (GC), as well as whether SW2 would be considered either esophageal or gastric cancer, and how this would impact on therapeutic decisions. Thus, the aim of this study was to assess the differences in the gene expression of EGJA compared to SW classification and compared to gastric adenocarcinoma of the body (GCb) and antrum (GCa). METHODS: The study included a cohort of 45 patients with EGJA (SW1=15; SW2=17; SW3=13) and 42 with GC (GCb=20; GCa=22) treated at our institution. Gene expression analysis was conducted using a high-throughput nanofluidic qRT-PCR platform (Biomark HD System, Fluidigm) including 68 genes related to malignancy/metastasis, multiple growth signaling pathways, immune response and tumor suppressor genes. Survival of the EGJA patients was replicated in the external cohort of The Cancer Genome Atlas Esophageal Carcinoma (TCGA-ESCA). RESULTS: Among the 87 patients included, the median age was 65.2 years, and 58.6% were male. In an unsupervised analysis (no regard to their topographical localization), we detected a list of 10 top genes as potentially informative for grouping samples: CASP8, PECAM1, CD34, CDH5, IL6, IL1B, SOD2, H1F1A, CEACAM6 and CD44 (Figure 1). In a supervised analysis, we found 20 differentially expressed genes in specific pairwise comparisons (Figure 2). Among them, STAT3 gene presented similar expression between SW1/SW2/SW3, but different from GCa and GCb. In EGJA cases, we detected 8 genes that follow a gradient of transition in gene expression levels according to SW classification: CDH5 (SW1<SW2<SW3); GSTP1 and SOX7 (SW1>SW2>SW3 for both); HES1 (SW1=SW2>SW3); STAT3, IL18, IL12A and HMOX2 (SW2=SW3<SW1 for all). In survival analyses including only EGJA patients who underwent curative surgery, we found that high-AKT1 (p=0.002), high-ALDH3A1 (p=0.003), and low-CEACAM6 (p=0.009) gene expressions were related to worse overall survival. In external cohort of TCGA-ESCA, CEACAM6 was also associated with survival. CONCLUSIONS: EGJA may be molecularly defined based on their mRNA expression patterns, with genes differentially expressed in relation to GC. SW2 and SW3 tumors appear to have similar and lower expression of genes related to immune response than SW1; while SW1 and SW2 have greater expression of genes associated with cell proliferation and chemoresistance. Similar to TCGA, low-CEACAM6 expression had a negative prognostic impact on EGJA. Therefore, complexity and heterogeneity of EGJA tumors suggest that surgical and oncological treatment may be optimized according to molecular profile.

Figure 1. Heatmap of unsupervised analysis, showing the 10 top genes as potentially informative for grouping samples with no regard to their topographical localization.

Figure 2. Differentially expressed genes in specific pairwise comparisons between tumor locations (Kruskal-Wallis test with Bonferroni posthoc, -log10 p-values); SW: Siewert.

Tatiane Furuya

Appearances