Plasma exosome gene signature differentiates colon cancer from healthy controls

Liquid biopsies have become an integral part of cancer management as minimally invasive options to detect molecular and genetic changes. However, current options show poor sensitivity in peritoneal carcinomatosis (PC). Novel exosome-based liquid biopsies may provide critical information on these challenging tumors. In this initial feasibility analysis, researchers at Loma Linda University School of Medicine identified an exosome gene signature of 445 genes (ExoSig445) from colon cancer patients, including those with PC, that is distinct from healthy controls.

Plasma exosomes from 42 patients with metastatic and non-metastatic colon cancer and 10 healthy controls were isolated and verified. RNAseq analysis of exosomal RNA was performed and differentially expressed genes (DEGs) were identified by the DESeq2 algorithm. The ability of RNA transcripts to discriminate control and cancer cases was assessed by principal component analysis (PCA) and Bayesian compound covariate predictor classification. An exosomal gene signature was compared with tumor expression profiles of The Cancer Genome Atlas. Unsupervised PCA using exosomal genes with greatest expression variance showed stark separation between controls and patient samples. Using separate training and test sets, gene classifiers were constructed capable of discriminating control and patient samples with 100% accuracy. Using a stringent statistical threshold, 445 DEGs fully delineated control from cancer samples. Furthermore, 58 of these exosomal DEGs were found to be overexpressed in colon tumors.

Evaluation of sample clustering by ExoSig445

(a) Cancer patient and healthy control samples (columns) were hierarchically clustered by average linkage analysis using exosomal RNAs (rows) identified as overexpressed in colon cancer cases as follows. DEGs were determined by DESeq2 analysis via pairwise group comparisons (i.e., Cont vs. NM, Cont. vs. PC and Cont vs. VM). Genes having baseMean read values >10, and with Benjamini–Hochberg adjusted p-values <0.0001 for any one pairwise comparison were identified. After filtering of prevalent and overlapping tRNA transcripts, 445 distinct transcripts (termed ExoSig445) were identified as highly significantly overexpressed in the colon cancer cases relative to controls. Multiple gene expression patterns were identified that clustered colon cancer samples into 5–6 distinct exosomal subtypes. Healthy control samples clustered apart from colon cancer samples due to a relative reduction in transcript levels. Red color indicates above mean expression; green color denotes below mean expression. Color intensity reflects the magnitude of expression level. (b) Violin plot of ExoSig445 expression scores defined as the per-sample arithmetic mean of the 445 genes. Black vertical bars indicate the interquartile range; black circle denotes the mean. Horizontal dashed lines indicate the range of no overlap in expression scores between healthy controls and colon cancer cases. Cont control (n = 10), NM non-metastatic (n = 14), VM visceral metastasis (n = 16), PC peritoneal carcinomatosis (n = 12), DEGs differentially expressed genes

Plasma exosomal RNAs can robustly discriminate colon cancer patients, including patients with PC, from healthy controls. ExoSig445 can potentially be developed as a highly sensitive liquid biopsy test in colon cancer.