Robust circulating microRNA signature for the diagnosis and early detection of pancreatobiliary cancer

Abstract

Background: A new circulating biomarker superior to carbohydrate antigen 19-9 (CA19-9) is needed for diagnosing pancreatobiliary cancer (PBca). The aim of this study was to identify serum microRNA (miRNA) signatures comprising reproducible and disease-related miRNAs.

Methods: This multicenter study involved patients with treatment-naïve PBca and healthy participants. The optimized serum processing conditions were evaluated using t-distributed stochastic neighbor embedding (t-SNE) visualization. Serum miRNA candidates for disease association were selected using weighted gene coexpression network analysis (WGCNA). A miRNA signature combining multiple serum miRNAs was tested in exploratory, validation, and independent validation sets. The synthesis and secretion of diagnostic miRNAs were evaluated using human pancreatic cancer cells.

Results: In total, 284 (150 healthy and 134 PBca) of 827 serum samples were processed within 2 h of blood collection before freezing, distributed in the same area as that in the t-SNE map, and assigned to an exploratory set. The 193 optimized samples were assigned to either the validation (50 healthy, 47 PBca) or independent validation (50 healthy, 46 PBca) set. Index-1, a combination of five serum miRNAs (hsa-miR-1343-5p, hsa-miR-4632-5p, hsa-miR-4665-5p, hsa-miR-665, and hsa-miR-6803-5p) with disease association in WGCNA, showed a sensitivity and specificity of > 80% and an AUC outperforming that of CA19-9 in the exploratory, validation, and independent validation sets. The AUC of Index-1 was superior to that of CA19-9 (0.856 vs. 0.649, p = 0.038) for detecting T1 tumors. miR-665, a component of Index-1, was expressed in human pancreatic cancer cells, and its transfection inhibited cell growth.

Conclusions: The serum miRNA signature Index-1 is useful for detecting PBca and could facilitate the early diagnosis of PBca. These findings can help improve clinical PBca detection by providing an optimized biomarker that overcomes the limitations of the current standard.

Keywords: Biomarker; Circulating microRNA; Pancreatobiliary cancer.

Fig. 1

Clinical consort diagram. PBca pancreatic or biliary tract cancer, HC healthy control

Fig. 2

Optimal conditions for measuring serum miRNA levels. The kinetics of 147 miRNA signals in whole blood (a) and serum (b) are dependent on the time of serum preparation. The x-axis shows the time to centrifugation from blood collection (a) or time to freezing from the start of centrifugation after 30 min of coagulation (b). The y-axis shows the signal of each miRNA. c t-SNE plot of the 827 participants shown in Fig. 1, using a serum expression dataset of 25 PBca-related miRNAs. The 827 participants were grouped according to their sampling period. PBca, pancreatic or biliary tract cancer; HC, healthy control

Fig. 3

Selection of diagnostic miRNA candidates based on robustness and relationship with PBca. a Selection diagram of diagnostic miRNA candidates and miRNA signature for diagnostic index. b Examples of robustness evaluation of miRNAs. Left: example of robust miRNA, hsa-miR-665. Right: example of unrobust miRNA, hsa-miR-125a-3p. Evaluation was performed using two tests [black circle: HC in the exploratory set (set1, n = 150), red circle: HC in prior obtained data (set2, n = 150)]. x-axis: mean of three internal control miRNAs; y-axis: miRNA signals; rsq: regression squared of regression curve; slope: slope of regression curve. c WGCNA results: module–trait relationship between eigengene modules and clinical factors. Pearson correlation coefficients (r) and p-values are shown in each column. d Ridge regression coefficient for detecting PBCA in the exploratory set. x-axis: absolute values of ridge regression coefficients. Red dashed lines indicate the threshold for the selection of coefficients. ROC curves using Index-1, Index-2, or CA19-9 for discriminating patients with PBca from healthy participants in the exploratory (e), validation (f), and independent validation sets (g). AUC values are shown in each figure. Int-con, mean of internal control miRNAs; rsq, regression squared; slope, slope of regression line; PBca, pancreatic or biliary tract cancer; HC, healthy control; logCEA, CEA (U/mL) in logarithm with base 2; logCA19-9, CA19-9 (U/mL) in logarithm with base 2; cT, cN, cM: Union for UICC clinical TNM classification; PLT, platelet count

Fig. 4

Subgroup analysis for diagnostic abilities of Index-1. The AUCs of Index-1 and CA19-9 in subgroups are shown horizontally. Circles represent the average AUC, and whiskers represent the lower and upper 95% confidence intervals. The red sign represents Index-1, whereas the black sign represents CA19-9. PBca, pancreatic or biliary tract cancer; HC, healthy control; AUC, area under the receiver operating characteristic curve; CI, confidence interval; CA19-9, carbohydrate antigen 19–9; ACC, Aichi Cancer Center; ASA, Asahikawa Medical University Hospital; HCC, Hyogo Cancer Center; KCC, Kanagawa Cancer Center; KYU, Kyusyu University Hospital; NCC, National Cancer Center East Hospital; ONO, JA Onomichi General Hospital; SHO, Showa University Hospital

Fig. 5

Index-1, CA19-9, and miRNAs of Index-1 in groups of cT. Healthy control (HC, black) and pancreatic/biliary tract cancer (PBCa, red) distributions are shown according to the UICC for International Cancer Control clinical T classification. Circle spots represent each case, boxes represent the interquartile range (IQR), and whiskers represent 1.5 × IQR from the hinge. The blue dotted lines in Index-1 and logCA19-9 indicate the thresholds. The p-value was calculated using Students’ t-test, either linearly (Index-1, CA19-9) or using log base of 2 (miRNAs), as shown in each graph. PBca, pancreatic or biliary tract cancer; HC, healthy control; logCA19-9: CA19-9 (U/mL) in logarithm with base 10. cT1 ~ cT4: UICC clinical T classification

Fig. 6

miRNA expression in pancreatic cancer lines and miR-665 effects on cell proliferation. a miRNA expression profile of eight cell lysate and supernatant samples. The upper color bar shows the origin tissue of each sample (red, cell lysate; blue, supernatant). Logarithmic miRNA expression is shown via color gradation. b Cell proliferation activity (ATP fluorescence) 4 days after induction with the miRNA mimic sequence. Data are presented as mean and SD, n = 3.**: p < 0.01, obtained via Student’s t-test. sup, supernatant; cell, cell lysate