The somatic mutation landscape of premalignant colorectal adenoma

Abstract

Objective: There are few studies which characterised the molecular alterations in premalignant colorectal adenomas. Our major goal was to establish colorectal adenoma genome atlas and identify molecular markers of progression from colorectal adenoma to adenocarcinoma.

Design: Whole-exome sequencing and targeted sequencing were carried out in 149 adenoma samples and paired blood from patients with conventional adenoma or sessile serrated adenoma to characterise the somatic mutation landscape for premalignant colorectal lesions. The identified somatic mutations were compared with those in colorectal cancer (CRC) samples from The Cancer Genome Atlas. A supervised random forest model was employed to identify gene panels differentiating adenoma from CRC.

Results: Similar somatic mutation frequencies, but distinctive driver mutations, were observed in sessile serrated adenomas and conventional adenomas. The final model included 20 genes and was able to separate the somatic mutation profile of colorectal adenoma and adenocarcinoma with an area under the curve of 0.941.

Conclusion: The findings of this project hold potential to better identify patients with adenoma who may be candidates for targeted surveillance programmes and preventive interventions to reduce the incidence of CRC.

Keywords: colorectal adenoma; conventional adenoma; sessile serrated adenoma; somatic mutation.

Figure 1

Frequency of somatic mutations in patients with colorectal cancer or adenoma. (A) Number of mutations per million base pairs: red, non-silent mutations; green, silent mutations. (B) Microsatellite instability status (MSI): red, MSI high; orange, MSI low; green, microsatellite stable; grey, data not available. MLH1 expression (MLH1): red, upregulation; blue, downregulation; green, normal; grey, data not available. Pathological subtypes (Subtype): yellow, colorectal cancer; blue, sessile serrated adenoma; green, conventional adenoma.

Figure 2

Frequently mutated genes with driver patterns demonstrated by WES and TS in adenomas. (A) CNAD WES. (B) SSA WES. (C) CNAD TS. (D) CNAD WES and TS. Driver gene q value: false-discovery rate of driver gene probability. Genes in panels (A) and (B), p<0.05; genes in panels (C) and (D), q<0.1. CNAD, conventional adenoma; OG, oncogene; SSA, sessile serrated adenoma; TS, targeted sequencing; TSG, tumour suppressor gene; UTR, untranslated region; WES, whole-exome sequencing.

Figure 3

Prevalence and composition of non-silent mutations in genes displaying a significant trend with disease progression. Left: non-advanced adenoma. Middle: advanced adenoma. Right: colorectal cancer. The diameters of the pie charts indicate relative prevalence of non-silent mutations. Colour: mutation type. All genes had trend test q<0.1.

Figure 4

Prediction model for malignant progression from conventional adenoma to non-hypermutated colorectal cancer. ROC, receiver operating characteristic.