Characterization of Gut Microbiota, Bile Acid Metabolism, and Cytokines in Intrahepatic Cholangiocarcinoma

Patients

Four groups (84 individuals) were included in this study. The demographic information and clinical characteristics of all patients are given in Supporting Table S1. Among the patients enrolled in the study, 28 patients with ICC (group I) and 28 patients with HCC (group H) were identified by pathological diagnosis. The other two groups included 16 patients with LC (group L) diagnosed using biopsy and 12 normal (healthy) individuals (group N). Healthy individuals with no prior cancer or liver disease diagnoses were selected among hospital researchers for group N. The histology of diseased and healthy peripheral liver tissues is shown in Supporting Fig. S1. The individuals enrolled in our study were treated at our hospital between 2016 and 2017. The exclusion criteria were as follows: (1) patients with hilar or distal cholangiocarcinoma, (2) patients with intrahepatic metastasis of extrahepatic cholangiocarcinoma, (3) patients with ICC mixed with HCC, (4) patients with other malignancies, (5) individuals with incomplete information, (6) individuals who abused alcohol, and (7) individuals who recently received treatment with antibiotic or probiotics. VI of ICC was also defined by the findings on final pathologic analysis. The study was approved by the ethics review committee of the Fifth Medical Center of PLA General Hospital. All participants provided informed consent for participation in this study.

Sample Collection

Blood plasma samples were collected for measurement of BAs and cytokines. Stool samples were collected to analyze BAs, and DNA was extracted for gut microbiota analysis.

DNA Extraction and 16S Ribosome RNA V4 Region Sequencing

DNA extraction was carried out according to the MoBio PowerSoil DNA Isolation Kit 12888-100 protocol (Qiagen, Hilden, Germany), and DNA was stored at −80°C in Tris-EDTA buffer solution (Sigma-Aldrich, St. Louis, MO) before use. To enable amplification of the V4 region of the 16S ribosomal RNA gene and add barcode sequences, unique fusion primers were designed based on the universal primer set, 515F (5′-GTGYCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACNVGGGTWTCTAAT-3′), along with barcode sequences. PCR mixtures contained 1 μL of each forward and reverse primer (10 μM), 1 μL of template DNA, 4 μL of deoxyribonucleotide triphosphates (2.5 mM), 5 μL of 10 × EasyPfu Buffer (TransGen Biotech Co., Beijing, China), 1 μL of EasyPfu DNA Polymerase (2.5 U/μL), and 1 μL of double distilled water in a 50-μL reaction volume. Thermal cycling consisted of an initial denaturation step at 95°C for 5 minutes, followed by 30 cycles of denaturation at 94°C for 30 seconds, annealing at 60°C for 30 seconds, and extension at 72°C for 40 seconds, with a final extension step at 72°C for 4 minutes. Amplicons from each sample were run on an agarose gel. The expected band size for 515f-806r was about 300-350 bp. Amplicons were quantified with the Qubit dsDNA HS Assay Kit (Q32854; Thermo Fisher, Waltham, MA/Invitrogen, Carlsbad, CA) following the manufacturer’s instructions. The amplicon library for high-throughput sequencing on the Illumina MiSeq platform was combined with an equal amount and subsequently quantified (KAPA Library Quantification Kit KK4824; Kapa Biosystems, Wilmington, MA) according to the manufacturer’s instructions.

Profiling of 16S Ribosomal RNA Gene Sequencing Data

Using the Quantitative Insights into Microbial Ecology 1.8.0 pipeline1, the raw sequences were processed to concatenate reads into tags according to the overlapping relationship; then, reads belonging to each sample were separated with barcodes, and low-quality reads were removed. The processed tags were clustered into the operational taxonomic units (OTUs) at the commonly used 97% similarity threshold. The OTUs were assigned to taxa by matching to the Greengenes database (Release 13.82). A phylogenetic tree of representative sequences was built. Alpha and beta diversity analyses were performed. Distances were calculated with R (3.3.1, flexmix package).

Bile Acid Detection15

The stool samples were weighted and transferred to a 1.5 mL-centrifuge tube followed by a supplement of 1 mL methanol. Then the tubes were shocked at room temperature for 1 hour and centrifuged at 13,200 rpm for 10 minutes. The supernatant was vacuum-dried and resuspended into 150 μL methanol. After a second centrifugation, the supernatant of stool extract was measured.

An aliquot of 200 μL of blood plasma was extracted with 600 μL of methanol containing the internal standards. The mixture was shocked at room temperature for 10 minutes and centrifugated at 13,200 rpm for 10 minutes. The 500 μL supernatant was vacuum-dried and resuspended into 100 μL methanol. After a second centrifugation, the supernatant of plasma extract was measured.

The stool and plasma extracts as well as the BA reference standards were analyzed with an UltiMate 3000 ultra-performance liquid chromatography (UPLC) coupled with a Q Exactive mass spectrometer (MS) (Thermo Fisher). The entire UPLC-MS system was controlled by Xcalibur 3.0 software. All chromatographic separations were performed with a HSS (High Strength Silica) T3 C18 column (1.7 μm, 100 mm × 2.1 mm internal dimensions; Waters, Milford, MA) and the injection volume was 1 μL. The mass spectrometer was operated in negative polarity with a 3-kv spray voltage. The heated capillary temperature was 320°C. The sheath gas and auxiliary gas were 30 psi and 10 psi, respectively. The parameters for the full mass scan were as follows: a resolution of 70,000, an auto gain control target under 1 × 10⁶, a maximum isolation time of 50 ms, and a mass-to-charge ratio range of 100-1,500. The calibration was customized for the analysis of Q Exactive to keep the mass tolerance of 5 ppm. The UPLC-MS raw data were analyzed using Xcalibur 3.0 software to obtain the calibration equations and the quantitative concentration of each BA in the samples. The BA standards were TUDCA, glycoursocholic acid (GUDCA), taurocholic acid (TCA), glycocholic acid (GCA), taurochenodeoxycholic acid, taurodeoxycholic acid, taurolithocholic acid, glycolithocholic acid, ursodeoxycholic acid, lithocholic acid, dehydrocholic acid, glycochenodeoxycholic acid (GCDCA), glycodeoxycholic acid (GDCA), cholic acid (CA), chenodeoxycholic acid (CDCA), and free deoxycholic acid (DCA).

Measurement of Cytokines and Histamine

The concentrations of blood plasma interleukin (IL)-4, tumor necrosis factor (TNF)-α, IL-6, IL-10, transforming growth factor (TGF)-β1, TGF-β2, IL-1β, monocyte chemoattractant protein (MCP)-1, and histamine were analyzed using the ProcartaPlex Multiplex Immunoassay system (eBioscience, Inc., San Diego, CA) and related enzyme-linked immunosorbent assay kits, according to the manufacturer’s instructions.

Murine Tumor Models

Male BALB/c nude mice (4 weeks old) were purchased from the Vital River Company (Beijing, China) and housed under specific pathogen-free conditions. Nude mice received TUDCA dissolved in 0.5% sodium carboxymethyl cellulose at a dose of 1,000 mg/kg once daily, and a control group received the same volume of vehicle by gavage 5 days per week for 8 weeks. The TUDCA/vehicle fed nude mice were divided into two groups at the second week, respectively, followed by hypodermic or intrahepatic injection of cholangiocarcinoma cell line QBC939 (2 × 10⁶ cell/mouse) into the mice of each group. The tumor size of each injected mice was measured at the end of the eighth week. All animal studies were approved by the Animal Welfare and Ethics Committee of the Fifth Medical Center of PLA General Hospital.

Statistical Analysis

Spearman’s correlation was calculated using R 3.3.1 software and SPSS 22.0 software. Correlation of gut microbiota with BAs was visualized through a heatmap.

Gut Microbiota Features of ICC, HCC, LC, and Healthy Individuals

The characteristics of stool bacteria among the four groups were analyzed using 16S sequencing. The phylum Firmicutes was the most abundant in all groups (Fig. 1A). α-Diversity metrics showed that group I had the highest biodiversity among all four groups (Fig. 1B). There were no differences in α-diversity among patients with HCC, patients with LC, and healthy individuals. The stool bacterial community profiles for patients with ICC were the least homogeneous among the four groups with regard to β-diversity (Fig. 1C). Additionally, the homogeneity of stool bacterial communities in patients with HCC was lower than that of the remaining two groups.

At the genus level, patients with ICC showed higher prevalence rates of Lactobacillus, Actinomyces, Peptostreptococcaceae, and Alloscardovia than the other groups by linear discriminant effect size (LEFSe) analysis (Fig. 2). The genera Ruminococcus and Leuconostocaceae were most abundant in group L, and the genera Lachnospira and Catenibacterium were most abundant in group N. However, compared with other groups, group H had no abundant genera.

Bile Acid Biomarkers for ICC Diagnosis

The gut microbiota can mediate the metabolism of primary BAs into secondary BAs and the deconjugation of BAs. In this study, we measured the levels of 16 BAs in blood plasma and stool samples from all participants using mass spectrometry. Among the four groups, there were no single BAs that could significantly distinguish ICC from HCC, LC, and healthy individuals. Therefore, we then calculated the plasma-stool ratios (PSRs) of BAs in each participant and found that the PSRs of two conjugated BAs were able to distinguish ICC from the other three groups. As shown in Fig. 3A, TUDCA-PSR and GUDCA-PSR were significantly higher in patients with ICC than in patients with HCC or LC and in healthy individuals. The diagnostic values of TUDCA-PSR and GUDCA-PSR were examined by receiver operating characteristic (ROC) analysis (Fig. 3B). The area under the curve (AUC) value of triglyceride (TG)-PSR, which indicated combining TUDCA-PSR and GUDCA-PSR between ICC and HCC, was 0.801 (95% confidence interval [CI]: 0.642-0.959). Additionally, the AUC value of TG-PSR between ICC and LC was 0.870 (95% CI: 0.730-1.000), and the AUC value of TG-PSR between ICC and healthy individuals was 0.906 (95% CI: 0.793-1.000). However, no PSRs of other BAs were significantly different among the four groups (Supporting Fig. S2A). Carbohydrate antigen 19-9 (CA 19-9), carcinogenic embryonic antigen (CEA), and cancer antigen 125 (CA 125) were used as positive controls in the comparisons of the four groups, and these three markers showed the highest concentrations in patients with ICC (Supporting Fig. S2B).

Genera Lactobacillus and Alloscardovia were Positively Correlated With TUDCA-PSR and Combined to Identify ICC

The correlation of the gut microbiota with TUDCA-PSR and GUDCA-PSR was analyzed. TUDCA-PSR was strongly positively correlated with the genera Lactobacillus and Alloscardovia (Fig. 4A). In contrast, GUDCA-PSR tended to be positively correlated with the genera Lactobacillus, Actinomyces and Alloscardovia, but no statistically significant differences were observed. The genus Catenibacterium was negatively correlated with GUDCA-PSR and TUDCA-PSR.

The abundances of Lactobacillus and Alloscardovia were further evaluated in each participant in the four groups. Lactobacillus was the most abundant. Notably, Alloscardovia was only observed in patients with ICC (Fig. 4B). Based on the logistic regression model, the genera Lactobacillus and Alloscardovia were combined to distinguish ICC from HCC, LC, and healthy controls by ROC curve analysis. As shown in Fig. 4C, the combination of the two genera could discriminate ICC from HCC with an AUC of 0.968. These two genera yielded AUCs of 0.965 and 0.987 to distinguish ICC from LC and healthy controls, respectively.

Gut Microbiota was Associated With VI in Patients With ICC

VI is associated with a poor prognosis in patients with cancer. In group I, the 3-year survival rates of patients with VI and without VI were 24.2% and 54.7%, respectively (Supporting Table S2). We further characterized the gut microbiota of patients with ICC with and without VI. Compared with patients with ICC without VI (group VI-N), patients with ICC with VI (group VI-Y) had similar α-diversity (Fig. 5A). In contrast, group VI-N showed higher β-diversity than group VI-Y (Fig. 5B). LEFSe analysis was used to determine the abundance of the gut microbiota in the two groups. The family Ruminococcaceae was more abundant in group VI-Y than in group VI-N. Compared with group VI-Y, group VI-N showed higher abundances of the family Eubacteriaceae and the genera Allobaculum, Pediococcus, Pseudoramibacter, and Peptostreptococcus (Fig. 5C). These results indicated that the gut microbiota was related to VI in patients with ICC.

Changes in Bile Acids and Inflammatory Cytokines in Patients With ICC With VI

Next, we compared stool and plasma BAs in groups VI-N and VI-Y. Six plasma BAs (i.e., CDCA, GCDCA, TCA, GDCA, TDCA, and TUDCA) had different concentrations in the two groups. The concentrations of plasma GCDCA, TCA, GDCA, TDCA, and TUDCA were higher in group VI-Y than in group VI-N. These five BAs were conjugated BAs. In contrast, the concentration of plasma CDCA (free BA) was higher in group VI-N (Fig. 6A).

BAs interact with their receptors (farnesoid X receptor and G protein-coupled BA receptor 1) to regulate inflammation.16, 17 We detected some cytokines, including IL-4, TNF-α, IL-6, IL-10, TGF-β1, TGF-β2, IL-1β and MCP-1, and histamine in blood plasma. As shown in Fig. 6B, patients with ICC with VI had higher levels of IL-4 and lower levels of IL-6 than patients with ICC without VI. However, the levels of TNF-α, IL-10, TGF-β1, TGF-β2, IL-1β, MCP-1, and histamine were similar between the two groups (Supporting Fig. S3A). Additionally, IL-4 levels were positively correlated with plasma TCA (r = 0.505, P = 0.023) by correlation analysis (Supporting Table S3).

Three conventional ICC markers (i.e., CA 19-9, CA 125, and CEA) were compared between the two groups, and only CA 19-9 was found to be increased in patients with ICC with VI (Supporting Fig. S3B). Moreover, the correlations of plasma CDCA, GCDCA, TCA, GDCA, TDCA, TUDCA, IL-4, IL-6, and CA 19-9 with various genera (Allobaculum, Pediococcus, Pseudoramibacter, and Peptostreptococcus) were analyzed. The genus Pseudoramibacter was negatively correlated with plasma TUDCA, whereas there were no correlations with the other genera (Fig. 6D).

TUDCA was Negatively Correlated With Survival in Patients With ICC but was Not Related to Tumor Size

These results indicated that plasma TUDCA may be related to ICC development. We analyzed the correlations of plasma TUDCA with survival, tumor number, tumor size, and lymph node metastasis again. Patients with ICC with lower plasma TUDCA levels had longer survival times (Fig. 7A). Moreover, plasma TUDCA was positively correlated with tumor number, but was not correlated with tumor size and lymph node metastasis (Supporting Table S4).

It had been reported that TUDCA inhibits the growth of Mz-ChA-1 cholangiocarcinoma cells18; therefore, we further determined the role of TUDCA in tumor growth using two murine cholangiocarcinoma models. As shown in Fig. 7B,C, mice fed TUDCA had tumor sizes similar to those in the vehicle-treated group in both mouse models, consistent with our clinical analysis.