Differential circulating miRNA profiles identified miR-423-5p, miR-93-5p, and miR-4532 as potential biomarkers for cholangiocarcinoma diagnosis

Chemicals and reagents

EDTA blood tubes were purchased from Becton Dickinson. The miRNAeasy® Serum/Plasma Kit (Qiagen, Gaithersburg, MD, USA) was used for the extraction of plasma cir-miRNA. The measurement of miRNA concentration was done by using a microvolume spectrophotometer (NanoDrop 2000; Thermo Fisher Scientific, Greenville, DE, USA) and Qubit™ miRNA Assay Kit (Invitrogen, Waltham, MA, USA). The miRNA profile was analyzed by nCounter^® SPRINT Profiler using Human V3 miRNA panel (NanoString Technologies, Seattle, WA, USA). The quantitation of miR-423-5p, miR-93-5p, and miR-4532 were investigated by ID3EAL miRNA qPCR Assay (MiRXES Pte Ltd, Singapore).

Study cohort

This study cohort was collected during year 2017–2019 and approved by the Human Ethic Committee of Udonthani Cancer Hospital, Udon Thani, Ministry of Public Health, Thailand (UCH-CT 11/2563). All subjects received written informed consent and gave their approval before the study.

Since chronic O. viverrini (OV) infection is the etiology of CCA, detection of OV infection and treatment with praziquantel are essential to prevent CCA development. However, detecting OV infection has been challenging due to the limitations of the gold standard fecal examination sensitivity. According to a previous study, most patients infected with OV were asymptomatic (Prasopdee et al., 2023). When OV patients are not treated, chronic liver fluke infection and CCA can develop. Therefore, differential diagnosis of OV is a crucial step in preventing the development of CCA. In this study, therefore, subjects with OV infection and CCA were recruited to identify potential differential diagnostic biomarkers for OV/CCA as well as diagnostic biomarkers for CCA.

Briefly, recruitment process of the subjects was done based on health statuses including healthy control (HC), O. viverrini infected subject, and cholangiocarcinoma subject. The healthy control group, which included the subjects with a normal physical examination, no liver enlargement, no jaundice, and who tested negative for O. viverrini egg on fecal examination; the group with O. viverrini-infected subjects, which included the subjects who tested positive for O. viverrini egg on fecal examination with normal liver and bile duct by ultrasonography; and the group with cholangiocarcinoma subjects, which included subjects with hepatomegaly and/or jaundice on physical examination, abnormal liver and bile duct ultrasonography, and confirmed cholangiocarcinoma by tissue histopathology. All the subjects in the OV group were asymptomatic. This study was divided into two phases including discovery phase by nCounter^® SPRINT Profiler and validate phase by RT-qPCR.

Heterogeneity is commonly observed in various cancers. Individual samples can exhibit variability due to genetic, environmental, and lifestyle differences among subjects. To address this heterogeneity, plasma samples were consolidated by pooling to minimize variability, thereby providing a more representative sample for this study (Schisterman et al., 2010; Kendziorski et al., 2005). For the discovery phase, 10 µl of each of the 10 plasma samples of each group were pooled individually for three pools (30 plasma samples/group and 90 plasma samples in total) and further subjected for nCounter^® SPRINT Profiler. For the validated phase, another 17 subjects of HC, 16 subjects of OV, and 35 subjects of CCA were recruited for RT-qPCR analysis. The demographics and clinical statuses of the subjects, including sex, age, alcohol consumption, smoking, eating habits of raw fish and fermented foods, and history of OV infection, were summarized in Table 1 for both the discovery and validation phases.

Extraction of circulating miRNA

The cir-miRNA was isolated from 200 μl of plasma samples using miRNeasy Serum/Plasma Kit according to the manufacturer’s protocol. Briefly, 1 ml QIAzol lysis reagent was added to 200 μl thawed plasma, mixed and incubated at room temperature for 5 min. A total of 3.5 μl Spike in control (MIRXES, Singapore) was added into the sample lysis buffer. A total of 200 μl chloroform was added, followed by shaking, incubation and centrifugation. The upper aqueous phase was transferred and 900 μl ethanol was added into a new tube and then transferred to the RNeasy MinElute column. The column was washed with RWT, RPE, and 80% Ethanol. following centrifugation to dry the membrane, RNAs were eluted with 14 μl RNase-free water. The isolated plasma circulating RNA (cfRNA) concentration was evaluated using the NanoDrop 2000 spectrophotometer.

Investigation of cir-miRNA profiles

The 100 ng of isolated plasma RNA were processed following the manufacturer’s manual. The cir-miRNA profile was analyzed by nCounter^® SPRINT Profiler using Human V3 miRNA panel. The data of raw count was analyzed by the nCounter^® SPRINT Profiler for image capture (190 fields of view). Analysis of cir-miRNA data was performed using nSolver Analysis (version 4.0) software. Subtraction of each cir-miRNA count data was done by using the geometric mean of the negative controls. For data integrity, the counting copy number of cir-miRNA with less than 25 were excluded. Profiling data were normalized by the geometric mean of the positive controls and the geometric mean of the top 100 most highly expressed microRNAs. The cir-miRNA that log2 fold-change (log2FC) >1.5 and P < 0.05 were used for differential cir-miRNA expression by using a heat map. The predicted targets of differential cir-miRNA were done by miRSystem (http://miRsystem.cgm.ntu.edu.tw) (Lu et al., 2012). For reactome analysis (https://reactome.org/PathwayBrowser/#TOOL=AT) (Fabregat et al., 2018), targeted mRNAs were then converted to UniProtKB ID by using gene cards of organism-specific databases (https://www.uniprot.org/uploadlists/) (The UniProt Consortium, 2023) and subjected for reactome analysis.

Validation of miR-423-5p, miR-93-5p, and miR-4532 by RT-qPCR

Plasma circulating RNAs from 200 μl plasma of 68 samples which include healthy volunteer (N = 17), OV-infected patiences (N = 16), and CCA patients (N = 35) were isolated by using miRNeasy Serum/Plasma kit (QIAGEN, Hilden, Germany) according to the manufacturer’s instructions except for the following modification: a spike in control (MIRXES, Singapore) was added into the sample lysis buffer and measured by NanoDrop 2000 spectrophotometer and Qubit 4 Fluorometer for total RNA and cir-miRNA, respectively. The 100 ng of isolated RNAs were reversed to cDNAs by using ID3EAL cDNA Synthesis System and ID3EAL RT primers (MIRXES, Singapore). The mixture was incubated at 42 °C for 30 min followed by 95 °C for 5 min. We examined three cir-miRNAs, including miR-423-5p, miR-93-5p, and miR-4532. The cir-miRNA expression was performed by RT-qPCR using ID3EAL miRNA qPCR assay (MIRXES, Singapore) on CFX96 Touch™ Real-time PCR detection system (QIAGEN, Germantown, MD, USA). The cDNA products were conducted according to the manufacturer’s protocol. Briefly, 5 µl of 1:10 diluted cDNA was combined with 10 µl ID3EAL miRNA qPCR Master Mix and 2 µl ID3EAL miRNA qPCR assay. The total volume of the mixture product was 20 µl. The incubation of the mixture product was carried out at 95 °C for 10 min, 40 °C for 5 min followed by 40 cycles of 95 °C for 10 s and 60 °C for 30 s. The Cq values were used for analysis. The relative gene expression values for cir-miRNAs were normalized by using spike-in control and compared by using ΔCt. The spike in control is synthetic small RNAs (~22 nt) with sequence distinct from endogenous miRNAs.

Statistical analysis

The statistical analysis was conducted utilizing IBM SPSS Statistics for Windows, Version 26.0 (IBM Corp., Armonk, NY, USA). For each selected cir-miRNA dataset (miR-423-5p, miR-93-5p, and miR-4532), the mean of 2^−ΔCt, along with the corresponding standard deviations (SD), was documented. Prior to conducting a one-way analysis of variance (ANOVA), values identified as extreme outliers by boxplot, defined as those above the third quartile (Q3) plus three times the interquartile range (IQR) or below the first quartile (Q1) minus three times the IQR, were excluded from the dataset. To find differences among the three health status groups, an initial assessment of homogeneity of variance via the Levene test was performed. Subsequently, if the obtained P-value exceeded 0.05, an ANOVA F-test followed by a Tukey Honestly Significant Difference (HSD) analysis was employed. Conversely, if Levene’s test yielded a P-value equal to or less than 0.05, indicating that the variances between groups are not equal, an ANOVA Welch test, which is unaffected by unequal variances, was conducted, followed by a Dunnett’s C post hoc analysis. Graphpad was used for scatter plot and area under curve (AUC) of the receiver operating characteristic (ROC) curve (GraphPad Prism version 10.0.0 for MacOS, GraphPad Software, Boston, Massachusetts, USA, www.graphpad.com). Maximum likelihood ratio was used for selecting cut-off. The diagnostic parameters including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were calculated by MedCalc (https://www.medcalc.org/calc/diagnostic_test.php; Version 22.017; accessed January 13, 2024).

Circulating miRNA profiles

At least 428 circulating miRNAs have been identified from plasma using the nCounter^® SPRINT Profiler (File S1). Using nSolver Analysis program to analyze the significant up- and down-regulated cir-miRNAs of three differential profiles including OV vs. HC, CCA vs. HC, and CCA vs. OV were listed along with the heat map (Fig. 1). Prediction of CCA vs. HC down-regulated cir-miRNA target, the chromosome structure and DNA repair were shown (File S2) while CCA vs. HC up-regulated cir-miRNA targeted to cell signaling pathway including Rho GTPase, TGFR family, Interleukin 4, 10, and 13, Autophagy, and DNA repair (File S3). For CCA vs. OV up-regulated cir-miRNA targeted to the immune system: Calcineurin activates NFAT and CLEC7A (Dectin-1) induced NFAT activation, and CAM-PDE 1 activation (File S4).

Investigation of potential circulating miRNA expression by RT-qPCR

To differentiate CCA from HC and OV, the top three cir-miRNAs with the highest fold changes in CCA compared to HC (P < 0.05) were selected for further validation as biomarkers for cholangiocarcinoma which include miR-4532 (15.26), miR-93-5p (4.18), and miR-423-5p (3.88) (Fig. 1). Moreover, miR-93-5p was specifically up-regulated in CCA compared to OV (2.19-fold, P = 0.0492), which is beneficial for distinguishing CCA from patients infected with OV.

Based on the 2^−ΔCt levels of miR-423-5p, a Welch’s ANOVA indicated differences among at least two groups (Welch F [2, 41.495] = 12.469, P < 0.001). Post hoc multiple comparisons using Dunnett C tests revealed significant differences between the CCA group and both the HC group (P < 0.05, 95% CI [0.033499–0.099281]) and the OV group (P < 0.05, 95% CI [0.034357–0.100437]). However, no significant difference was observed between the HC group and the OV group (P > 0.05, 95% CI [−0.007820 to 0.009833]) (Fig. 2).

The one-way ANOVA indicated differences in the 2^−ΔCt levels of miR-93-5p among at least two groups (F [2, 63] = 3.899, P = 0.025). Subsequent Tukey HSD tests revealed a significant difference between the CCA group and the OV group (P = 0.019, 95% CI [0.000159–0.002121]). However, no significant differences were observed between the HC group and the OV group (P = 0.275, 95% CI [−0.000397–0.001846]) or between the HC group and the CCA group (P = 0.556, 95% CI [−0.001377 to 0.000545]) (Fig. 2).

Furthermore, the 2^−ΔCt levels of miR-4532 exhibited a similar trend, indicating differences among at least two groups (Welch F [2, 36.106] = 18.969, P < 0.001). The Dunnett C test identified significant differences between the CCA group and both the HC group (P < 0.05, 95% CI [0.001294–0.002996]) and the OV group (P < 0.05, 95% CI [0.001090–0.002890]). Nevertheless, no significant difference was noted between the HC group and the OV group (P > 0.05, 95% CI [−0.000546 to 0.000236]) (Fig. 2).

Diagnostic potential of circulating miRNA

The AUC of the ROC curve and diagnostic parameters including sensitivity, specificity, PPV, NPV, and accuracy of miR-423-5p, miR-93-5p, and miR-4532 were calculated by using Graphpad and MedCalc, respectively, and were summarized in Table 2. The result showed all cir-miRNA could be used for differential CCA from HC and OV groups. The highest AUC of the ROC curve was found for differential CCA from HC group (0.8983) by using miR-4532. The high specificity with low sensitivity were observed in miR-423-5p (cut-off <25.24: 57.14% sensitivity and 96.97% specificity) and miR-93-5p (cut-off <29.32: 20.00% sensitivity and 96.97% specificity). The 77.14% sensitivity and 78.79% specificity were obtained from miR-4532 when using <30.42 as a cut-off. In combination of these three potential cir-miRNAs, the 85.71% sensitivity, 76.92% specificity, 76.92% PPV, 85.71% NPV, and 81.08% accuracy were obtained.