Gene Signature in Sessile Serrated Polyps Identifies Colon Cancer Subtype

Abstract

Sessile serrated colon adenoma/polyps (SSA/P) are found during routine screening colonoscopy and may account for 20% to 30% of colon cancers. However, differentiating SSA/Ps from hyperplastic polyps (HP) with little risk of cancer is challenging and complementary molecular markers are needed. In addition, the molecular mechanisms of colon cancer development from SSA/Ps are poorly understood. RNA sequencing (RNA-Seq) was performed on 21 SSA/Ps, 10 HPs, 10 adenomas, 21 uninvolved colon, and 20 control colon specimens. Differential expression and leave-one-out cross-validation methods were used to define a unique gene signature of SSA/Ps. Our SSA/P gene signature was evaluated in colon cancer RNA-Seq data from The Cancer Genome Atlas (TCGA) to identify a subtype of colon cancers that may develop from SSA/Ps. A total of 1,422 differentially expressed genes were found in SSA/Ps relative to controls. Serrated polyposis syndrome (n = 12) and sporadic SSA/Ps (n = 9) exhibited almost complete (96%) gene overlap. A 51-gene panel in SSA/P showed similar expression in a subset of TCGA colon cancers with high microsatellite instability. A smaller 7-gene panel showed high sensitivity and specificity in identifying BRAF-mutant, CpG island methylator phenotype high, and MLH1-silenced colon cancers. We describe a unique gene signature in SSA/Ps that identifies a subset of colon cancers likely to develop through the serrated pathway. These gene panels may be utilized for improved differentiation of SSA/Ps from HPs and provide insights into novel molecular pathways altered in colon cancer arising from the serrated pathway. Cancer Prev Res; 9(6); 456-65. ©2016 AACR.

Figure 1

Differentially expressed annotated protein coding and non-coding RNAs in SSA/Ps and traditional hyperplastic polyps (HPs) identified by RNA sequencing. Panel A – Differentially expressed genes with a ≥ 2-fold change and FDR < 0.05 in SSA/Ps (n=12 for syndromic and n=9 for sporadic) compared to control right colon (n=10) and HPs (n=10) compared to control left colon (n=10). Panel B – Relative abundance of protein coding and non-coding RNAs differentially expressed in SSA/Ps. Non-coding RNAs included antisense non-coding RNAs, long intergenic non-coding RNAs (lincRNAs), pseudogenes and other miscellaneous RNAs including immunoglobulin and intronic RNAs.

Figure 2

Differentially expressed genes in syndromic and sporadic SSA/Ps and HPs by RNA sequencing. Panel A – Genes with ≥ 4-fold change and FDR < 0.05 in syndromic SSA/Ps (n=12), sporadic SSA/Ps (n=9) and HPs (n=10). Syndromic and sporadic SSA/Ps were compared to control right colon and HPs were compared to control left colon. Panel B – Relative expression of 27 protein-coding genes in syndromic SSA/Ps, sporadic SSA/Ps, HPs and control left and right colon. Log₂ ratios comparing each individual sample to the mean of all samples were used for hierarchical clustering. Two right-sided HPs are labeled in red and five left-sided SSA/Ps are labeled in green. Panel C – Mean fold change expression of the same 27 protein coding genes described in Panel B in normal colon, adenomas, HPs and sporadic and syndromic SSA/Ps.

Figure 2

Differentially expressed genes in syndromic and sporadic SSA/Ps and HPs by RNA sequencing. Panel A – Genes with ≥ 4-fold change and FDR < 0.05 in syndromic SSA/Ps (n=12), sporadic SSA/Ps (n=9) and HPs (n=10). Syndromic and sporadic SSA/Ps were compared to control right colon and HPs were compared to control left colon. Panel B – Relative expression of 27 protein-coding genes in syndromic SSA/Ps, sporadic SSA/Ps, HPs and control left and right colon. Log₂ ratios comparing each individual sample to the mean of all samples were used for hierarchical clustering. Two right-sided HPs are labeled in red and five left-sided SSA/Ps are labeled in green. Panel C – Mean fold change expression of the same 27 protein coding genes described in Panel B in normal colon, adenomas, HPs and sporadic and syndromic SSA/Ps.

Figure 3

Evaluation of a 28-gene signature to distinguish SSA/Ps from HPs. The 28-gene panel was developed using a leave-one-out cross validation approach on 31 independent serrated polyps (21 SSA/Ps and 10 HPs) samples. Panel A – Principal component analysis of the 28 gene log2 ratios for each individual serrated polyp compared to the mean of all serrated polyps. Principal component 1 (PC1) accounted for 28% of the variation in the data and separated most SSA/Ps (red) from HPs (blue). Twenty-eight of thirty-one serrated polyps (~90%) clustered correctly similar to the nominal error rate found in the cross validation results. Panel B – Relative expression (log of normalized reads (RPKM) of the same 28 genes described in Panel A in SSA/Ps and HPs. Six genes (circles) were overexpressed in SSA/Ps compared to HPs (range 2.8 to 3.7 fold) and 22 genes (squares) were underexpressed in SSA/Ps compared to HPs (range −2.2 to −6.7).

Figure 4

Evaluation of a 51 SSA/P gene signature in colon cancer RNA sequencing datasets from The Cancer Genome Atlas (TCGA). Panel A – Log2 ratios comparing individual colon cancers (n=72) and SSA/Ps (n=21) to the mean of 14 uninvolved and 10 control colon samples (n=24) were used for hierarchical clustering. “Tissue” color bar shows colon adenocarcinomas (orange) and SSA/Ps (yellow). “MSI status” color bar shows microsatellite stable (MSS) cancers (dark blue), MSI-H cancers (red) and MSI-L cancers (light orange). SSA/Ps and colon cancers not evaluated for MSI (light blue). Panel B – Percentage of TCGA colon cancers showing overexpression of FSCN1, ZIC2, ZIC5, CRYBA2, MUC6, TRNP1 and/or SEMG1 described in Table 1. 195 colon cancers with RNA expression and MLH1 methylation data in the TCGA database were evaluated using the cBioPortal for Cancer Genomics.