A Circulating Exosome RNA Signature Is a Potential Diagnostic Marker for Pancreatic Cancer, a Systematic Study

Abstract

Several exosome proteins, miRNAs and KRAS mutations have been investigated in the hope of carrying out the early detection of pancreatic cancer with high sensitivity and specificity, but they have proven to be insufficient. Exosome RNAs, however, have not been extensively evaluated in the diagnosis of pancreatic cancer. The purpose of this study was to investigate the potential of circulating exosome RNAs in pancreatic cancer detection. By retrieving RNA-seq data from publicly accessed databases, differential expression and random-effects meta-analyses were performed. The results showed that pancreatic cancer had a distinct circulating exosome RNA signature in healthy individuals, and that the top 10 candidate exosome RNAs could distinguish patients from healthy individuals with an area under the curve (AUC) of 1.0. Three (HIST2H2AA3, LUZP6 and HLA-DRA) of the 10 genes in exosomes had similar differential patterns to those in tumor tissues based on RNA-seq data. In the validation dataset, the levels of these three genes in exosomes displayed good performance in distinguishing cancer from both chronic pancreatitis (AUC = 0.815) and healthy controls (AUC = 0.8558), whereas a slight difference existed between chronic pancreatitis and healthy controls (AUC = 0.586). Of the three genes, the level of HIST2H2AA3 was positively associated with KRAS status. However, there was no significant difference in the levels of the three genes across the disease stages (stages I-IV). These findings indicate that circulating exosome RNAs have a potential early detection value in pancreatic cancer, and that a distinct exosome RNA signature exists in distinguishing pancreatic cancer from healthy individuals.

Keywords: KRAS mutation; RNA signature; early detection; exosome; pancreatic cancer.

Figure 1

Differential expression of circulating exosome RNAs in pancreatic cancer patients and healthy individuals. (A) Volcano plot of differential circulating exosome RNAs. The red dots represent significant genes with baseMean ≥20, the absolute value of log2FoldChange ≥5 and FDR <0.01, and the gray dots represent non-significant results. (B) PCA plot of circulating exosome RNAs showing the distance between the individuals. Light-coral dots represent healthy individuals and turquoise dots represent patients with pancreatic cancer. (C) Heatmap of significantly differential circulating exosome RNAs for patients with pancreatic cancer (light-coral bar) and healthy individuals (turquoise bar). The columns of the heatmap represent individuals, and the rows represent circulating exosome RNA genes. (D) Receiving operating characteristic (ROC) curve with the exosome RNAs of 10 candidate genes. AUC is the area under the curve.

Figure 2

Illustration of the top network of differentially expressed transcripts, related to “gene expression”, “RNA post-transcriptional modification” and “neurological diseases” in circulating exosome RNAs, compared between pancreatic cancer patients and healthy individuals. Red and green shading indicate the up- and downregulation of transcripts in circulating exosomes derived from patients with pancreatic cancer relative to healthy controls, respectively, with the color intensity corresponding to the degree of fold-change. Solid and dotted lines indicate direct and indirect relationships, respectively.

Figure 3

Boxplots of the candidate gene expression levels in pancreatic adenocarcinoma and normal pancreas tissues. The vertical (y) axis is the gene expression level of transcripts per million (TPM) in log2 (TPM +1), and the horizontal (x) axis is the status of tissues (left box = tumor (n = 179) and right box = normal (n = 171)). The solid line inside the box is the median and the box edges are the 25th and 75th quartiles (interquartile range, IQR). The whiskers are 1.5 × IQRs. The black dots are the expression levels of the gene in individuals. * p < 0.01.

Figure 4

Forest plot of the differential gene expression in primary pancreatic cancer. Random-effects meta-analysis results of the fold change in log2 for HIST2H2AA3 (A), LUZP6 (B) and HLA-DRA (C) in primary pancreatic ductal adenocarcinoma vs. normal pancreas tissues.

Figure 5

Violin plots of the three gene expression levels across the disease stages in pancreatic cancer. The vertical axis is the gene expression level of transcript per million (TPM) in log2 (TPM +1) for HIST2H2AA3 (A), LUZP6 (B) and HLA-DRA (C), respectively, and the horizontal axis is the disease stage. The violin shape is the frequency distribution of the gene expression levels; the inside boxplot represents the median (white dot), interquartile range (the box edge) and 95% confidence interval (the solid black line).

Figure 6

Violin plots of the three gene expression levels and KRAS mutation status in pancreatic cancer. The vertical axis is the gene expression level in log10 (FKPM) for HIST2H2AA3 (A), LUZP6 (B) and HLA-DRA (C), respectively, and the horizontal axis is the KRAS mutation status. The violin shape is the frequency distribution of gene expression levels; the inside boxplot represents the median (black line), interquartile range (the box edge) and 95% confidence interval (the solid black line).

Figure 7

ROC curves for the circulating exosome RNA signature of HIST2H2AA3, LUZP6 and HLA-DRA in pancreatic cancer patients and healthy controls; (A) pancreatic cancer vs. healthy controls; (B) pancreatic cancer vs. chronic pancreatitis; and (C) chronic pancreatitis vs. healthy controls, respectively. AUC is the area under the curve for the accuracy of the signature distinguishing the two groups.