Comprehensive molecular pathology analysis of small bowel adenocarcinoma reveals novel targets with potential for clinical utility

Abstract

Small bowel accounts for only 0.5% of cancer cases in the US but incidence rates have been rising at 2.4% per year over the past decade. One-third of these are adenocarcinomas but little is known about their molecular pathology and no molecular markers are available for clinical use. Using a retrospective 28 patient matched normal-tumor cohort, next-generation sequencing, gene expression arrays and CpG methylation arrays were used for molecular profiling. Next-generation sequencing identified novel mutations in IDH1, CDH1, KIT, FGFR2, FLT3, NPM1, PTEN, MET, AKT1, RET, NOTCH1 and ERBB4. Array data revealed 17% of CpGs and 5% of RNA transcripts assayed to be differentially methylated and expressed respectively (p < 0.01). Merging gene expression and DNA methylation data revealed CHN2 as consistently hypermethylated and downregulated in this disease (Spearman -0.71, p < 0.001). Mutations in TP53 which were found in more than half of the cohort (15/28) and Kazald1 hypomethylation were both were indicative of poor survival (p = 0.03, HR = 3.2 and p = 0.01, HR = 4.9 respectively). By integrating high-throughput mutational, gene expression and DNA methylation data, this study reveals for the first time the distinct molecular profile of small bowel adenocarcinoma and highlights potential clinically exploitable markers.

Keywords: CHN2; Kazald1; Pathology Section; p53; small intestine cancer.

Figure 1

A. Gene expression of the tumor top 25% most variant probes overlain with associated clinical and mutational information using WGCNA. Clusters containing Crohn's and coeliac disease cases are outlined in red. B. Limma volcano plot of differential expression (non-variance filtered) with the top 30 probes annotated. C. Random forest classification on 75% of samples utilizing kmeans to offer data reduction of probe numbers for utility in validation machines of lower complexity. Top ranked variable importance in each kmeans group used in final n = 20 signature. D. WGCNA analysis revealed module ‘5′ had a strong correlation with PIK3CA mutation by Kendall correlation. Depicted here is the most strongly correlated network within the module, filtered across a bonferroni corrected all-possible-correlations threshold of p < 0.001 and visualized by NetBioV.

Figure 2

A. ‘Gene’ region methylation expression of the top 25% most variant probes overlain with associated clinical and mutational information. B. Volcano plot of differential expression (non-variance filtered) with the top 30 probes annotated, ranked by ‘combined rank’ in RnBeads. C. random forest classification on 75% of samples ‘gene’ regions utilizing kmeans to offer data reduction of numbers for utility in validation machines of lower complexity. Top ranked variable importance in each kmeans group used in final n = 20 signature. D. WGCNA analysis revealed module ‘3’ had a strong correlation with KRAS mutation by Kendall correlation. Depicted here is the most strongly correlated network within the module, filtered across a bonferroni corrected all-possible-correlations threshold of p < 0.001 and visualized by NetBioV.

Figure 3. Correlation between DNA methylation and gene expression

A–E. Inverse correlation for the top 5 genes from array data, n = 45 (primary axis – methylation, secondary axis – expression). F. Inverse correlation between pyrosequencing and immunohistochemistry for CHN2, n = 56.

Figure 4

A. Kaplan–Meier curve showing survival advantage for wild-type p53 cases, n = 27. B. Kaplan–Meier curve showing survival disadvantage for cases with Kazald1 hypomethylation, n = 26.

Figure 5. Immunohistochemistry

A. p53 wild type tumor showing no staining (extreme negative). B. p53 mutant tumor showing extreme positive staining. The C > T substitution can be seen as confirmed by Sanger sequencing C. No CHN2 protein expression in tumor. A low T and high C peak show a highly methylated CpG D. CHN2 protein expression in normal small bowel. A high T and low C peak show an unmethylated CpG E. Tumor with no Her2 expression. F. Tumor with a Her2 c.2329G > T (Substitution) mutation expressing the protein.