The miRNAome of Opisthorchis viverrini induced intrahepatic cholangiocarcinoma

Intrahepatic cholangiocarcinoma (ICC) is an aggressive cancer, arising in the biliary ducts that extend into the liver. The highest incidence of ICC occurs in Southeast Asia, particularly in the Mekong River Basin countries of Thailand, Laos, Cambodia, and Vietnam, where it is strongly associated with chronic infection by the food-borne liver fluke Opisthorchis viverrini (OV), one of only three eukaryote pathogens considered Group one carcinogens. Intrahepatic cholangiocarcinoma is usually diagnosed at an advanced stage, with a poor prognosis and survival often less than 24 months. Hence, biomarkers that enable the early detection of ICC would be desirable and have a potentially important impact on the public health in the resource-poor regions where this cancer is most prevalent. As microRNAs (miRNAs) remain well preserved after formalin fixation, there is much interest in developing them as biomarkers that can be investigated using tumor biopsy samples preserved in formalin fixed paraffin embedded (FFPE) tumor blocks. Recently, we reported the first comprehensive profiling of tissue-based miRNA expression using FFPE from the three most common subtypes of OV-induced ICC tumors: moderately differentiated ICC, papillary ICC, and well-differentiated ICC. We observed that each subtype of OV-induced ICC exhibited a distinct miRNA profile, which suggested the involvement of specific sets of miRNAs in the progression of this cancer. In addition, non-tumor tissue adjacent to ICC tumor tissue on the same FFPE block shared a similar miRNA dysregulation profile with the tumor tissue than with normal (non-tumor) liver tissue (individuals without ICC or OV infection). Herein, we provide a detailed description of the microarray analysis procedures used to derive these findings.

Intrahepatic cholangiocarcinoma (ICC) is an aggressive cancer, arising in the biliary ducts that extend into the liver. The highest incidence of ICC occurs in Southeast Asia, particularly in the Mekong River Basin countries of Thailand, Laos, Cambodia, and Vietnam, where it is strongly associated with chronic infection by the food-borne liver fluke Opisthorchis viverrini (OV), one of only three eukaryote pathogens considered Group one carcinogens. Intrahepatic cholangiocarcinoma is usually diagnosed at an advanced stage, with a poor prognosis and survival often less than 24 months. Hence, biomarkers that enable the early detection of ICC would be desirable and have a potentially important impact on the public health in the resource-poor regions where this cancer is most prevalent. As microRNAs (miRNAs) remain well preserved after formalin fixation, there is much interest in developing them as biomarkers that can be investigated using tumor biopsy samples preserved in formalin fixed paraffin embedded (FFPE) tumor blocks. Recently, we reported the first comprehensive profiling of tissue-based miRNA expression using FFPE from the three most common subtypes of OV-induced ICC tumors: moderately differentiated ICC, papillary ICC, and well-differentiated ICC. We observed that each subtype of OV-induced ICC exhibited a distinct miRNA profile, which suggested the involvement of specific sets of miRNAs in the progression of this cancer. In addition, nontumor tissue adjacent to ICC tumor tissue on the same FFPE block shared a similar miRNA dysregulation profile with the tumor tissue than with normal (non-tumor) liver tissue (individuals without ICC or OV infection). Herein, we provide a detailed description of the microarray analysis procedures used to derive these findings.
©  (Table 2) derived from liver biopsies of individuals suspected of severe steatosis or steatohepatitis prior to gastric bypass surgery were included as normal non-tumor tissue (N-NT) to assess baseline liver histology of individuals with no ICC and do not reside in an OV endemic region. Details of the specimens, including histological confirmation and preparation of FFPE samples, can be found in [1].
The Institutional Review Boards from both KKU and GWU determined that the samples did not meet the definition of human subjects research, i.e., a living individual about whom an investigator conducting research obtains: a) data through intervention or interaction with the individual or b) private identifiable information. This determination was made since the samples were limited to pre-existing, de-identified specimen analysis labeled with a random code.

RNA isolation
Total RNA was extracted from each of the FFPE blocks using the miRNeasy FFPE kit (Qiagen) following the manufacturer's protocol [2] and also further detailed in [1]. RNA quality and integrity were examined by spectrophotometry (Nano Drop 2000, Thermo Scientific) and by using an Agilent 2100 Bioanalyzer (RNA 6000 Nano and Small RNA Kits, Agilent). The 260/280 ratios obtained were approximately 2.0, indicating that the RNA was pure. The purified RNA exhibited 260/230 ratios greater than 1.9, or only slightly less than the range of 2.0-2.2 expected, indicating no significant contaminants.
Analysis by the Agilent Bioanalyzer produced RNA integrity numbers (or RIN scores) of two to three, with 28S and 18S peaks largely absent. This value was below the quality criteria (RIN greater than or equal to eight) and indicated degraded RNA. RNA obtained from FFPE tissue samples often has slight modifications and degradation due to the process of formalin fixation and duration of storage is expected [3,4]. However, due to the stable nature of miRNAs in the FFPE matrix, as shown by us [1,2] as well as others [5,6], the extracted RNA was determined to be suitable for subsequent analysis of miRNA profiles by microarray.

Microarray analysis
Purified RNA from all 46 cases samples including 16 CTT and both types of controls (15, D-NT and 13, N-NT) (Tables 1, 2) were profiled on the Agilent human miRNA microarray (miRBase Release 16.0). Hybridization and further details are available in the Gene Expression Omnibus (GEO) database and also found in [1].

Data normalization and analysis
The feature intensity of each sample was transferred to digital data and then log 2 transformed using Agilent Feature Extraction (V.10.7). All the raw data files in text (.txt) format were analyzed (including statistical analysis) with Agilent GeneSpring software (GX 12.6) (http://genespring-support.com/files/gs_12_6/GeneSpring-manual.pdf) [7].
To determine the presence of aberrantly expressed miRNAs, a new "project" and new "experiment" were created within GeneSpring (GX 12.6), with miRNA selected for analysis type and the data import wizard used for the workflow type. After uploading the raw intensity files into GeneSpring, 31181_v18_0 technology was selected ( Table 3). The threshold raw signals were set to 1.0 and 90 percentile and "shift normalization" was performed to standardize the statistical parameters across all samples (Table 3). No baseline transformation was performed.
Four different methods were used to analyze this sample set as shown in Fig. 1 Table 3. Dysregulated miRNAs were reported as associated with either ICC itself or with ICC stratified by histological subtype as reported in [1]. In all analyses below, subsequent "interpretations" were created in GeneSpring (GX 12.6) with the conditions selected, including "nonaveraged" selected over replicates, and the measurements flagged as "default." In most analyses, unless otherwise noted, the probe sets were filtered by expression value (i.e. 30-336133.0) with at least 50% of the values greater than 30 in any one condition within range.

and in
Analysis One: CTT versus N-NT versus D-NT by one-way ANOVA 1) In this first experimental design (Table 3), the FFPE samples were grouped accordingly: 13 N-NT, 15 D-NT, 2 necrotic samples, and 16 CCT samples (without consideration of histological subtype).
2) For "Quality Control", the correlation coefficient (CC) and 3D Principal Component Analysis (PCA) scores were used to determine associations among the samples. The CC of sample Y62-N1 (D-NT from a papillary ICC block) was below an acceptable CC (b0.7) and the sample removed from subsequent analyses. Principle Components Analysis demonstrated that N-NT samples clustered together, and were distinct from the other types of samples (CTT and D-NT). Further details regarding principal component analysis can be found below: • Algorithm: Principal Components Analysis • Parameters: ○

4)
To analyze the differences between CTT versus D-NT versus N-NT, a one-way ANOVA was conducted (with no post hoc pairwise testing) with asymptotic p-value computation without correction completed. Fold changes (≥ 2.0) were analyzed under pairs of conditions with two minimum numbers of pairs. Fewer differences were observed between tumor tissue (CTT) and distal non-tumor tissue (D-NT) than between D-NT and N-NT, confirming the PCA and Hierarchical Clustering results.
Analysis Two: CTT versus D-NT by paired Student's t-test 1) CTT samples were again grouped by histological subtype and compared to adjacent non-tumor tissue (D-NT) (see Table 1): i.e., moderately differentiated CTT versus D-NT from the same block, well-differentiated CTT versus D-NT from the same block, and papillary CTT versus D-NT from the same block (Table 3, Box 2 Fig. 1).
2) To filter probe sets by expression value, in each case the "entity" and "interpretation" were selected and filtered by raw data values. The lower cut-off value was set to 30 and at least 50% of the values in any one condition within range for welldifferentiated CTT versus its D-NT and papillary CTT versus its D-NT. In the case of moderately differentiated CTT versus necrosis versus N-NT, at least 50% of the values in one condition were greater than 20.    3) In the "analysis workflow" in GeneSpring (GX 12.6), paired Student's t-test (p b 0.05) was performed to analyze differences among CTT versus D-NT samples of well-differentiated ICC and papillary ICC by asymptotic p-value computation without correction for multiple testing. Further a fold change cut-off of 2.0 between conditions (CTT against D-NT) was performed. Only fold changes ≥ 4.0 were analyzed for moderately differentiated CTT versus moderately differentiated necrosis versus moderately differentiated D-NT but statistical significance was not calculated from this limited sample set consisting of only two moderately differentiated CTT samples and a single D-NT sample. 4) A Venn diagram was plotted to determine the shared dysregulated miRNAs among the groups of analyzed after filtering, statistical analysis, and fold change selection (data not shown).

2) When conducting unpaired Student's t-test (with Benjamini
Hochberg False Discovery Rate correction undertaken due to multiple testing), each comparison was given its own p-value cut, when testing the three histological subtypes of ICC versus normal liver tissue (N-NT) by asymptotic p-value computation: p b 0.01 was set for well-differentiated ICC versus N-NT and p b 0.005 for papillary and moderately differentiated ICC versus N-NT (p-values after Benjamini Hochberg False discovery rate correction). A fold change cut-off of 2.0 between conditions (ICC versus N-NT) was also performed. 3) A Venn diagram was plotted to show the shared dysregulated miRNAs after filtering, statistical analysis and fold change selection (Center Box, Fig. 1).
Analysis Four: Histological subtype analysis of ICC by one-way ANOVA 1) Here the subtypes of ICC (CTT, necrotic, and D-NT samples were grouped and referred to as "tumor" or "ICC" as a whole) were analyzed against each other (Table 3, Box 4 Fig. 1).
2) Fold change selection was performed by selecting a cut-off of 2.0 in two pairs of conditions. Further a one-way ANOVA (no pairwise poc hoc testing) with Benjamini Hochberg False Discovery Rate correction was performed among the pairs of conditions by asymptotic p-value computation and p b 0.05 (p-values after Benjamini Hochberg False discovery rate correction).

3) Significantly dysregulated miRNAs, as determined in Analysis
Four, were plotted by Venn diagram and compared to the dysregulated miRNAs determined by Analysis Three for observation of overlap (Center Box, Fig. 1). Fig. 1. Flow diagram of analyses and associated miRNAs found to be dysregulated. A final qPCR verification step was also completed to confirm the magnitude and expression levels determined by microarray but not described herein [1]. A Venn diagram [8] highlights overlapping miRNAs identified in analysis three and analysis four, yielding seven overlapping miRNAs. Abbreviations are as follows: Intrahepatic cholangiocarcinoma tumor tissue (CTT); distal non-tumor tissue (D-NT); and normal non-tumor tissue (N-NT).