Despite decades of BA studies, diverse biological roles for BAs have been identified just recently due to developments in understanding the human microbiota[20]. Fecal microbiota transplantation (FMT), as the reconstruction technique of gut microbiota, represents a paradigm shift by restoring microbial BA-metabolizing functions. Despite mechanistic links between dysbiosis and BAM in IBD, clinical evidence for the efficacy of FMT remains sparse. Prior studies lack multi-omics integration and longitudinal cohort validation. Thus, in this study we aimed to evaluate the efficacy of FMT in IBD patients with BAM or BAD identified by serum C4, and explore the value of the random forest model in predicting the occurrence of BAM in IBD patients based on baseline microbiota.
This study recruited healthy Chinese volunteers as a control group (healthy controls matched with patients, age, and sex) and donors of FMT and was derived and part of two registered trials (ClinicalTrials.gov: NCT01790061 and NCT01793831) which conducted at the Second Affiliated Hospital of Nanjing Medical University. We selected the study cohort (37 UC and 69 CD) based on the availability of paired clinical samples and associated data before and after FMT treatment in patients with IBD who underwent microbiota transplantation between 2013 and 2020. Demographics and clinical characteristics were thoroughly recorded before FMT. The daily score of abdominal pain, frequency of defecation, and clinical efficacy of treatments were recorded and evaluated at baseline, within 1 week, and 1 and 3 months after the initial FMT. In UC patients, clinical response was defined as a decrease of partial Mayo score ≥ 3 and ≥ 30% from baseline, in addition to a decrease in the rectal bleeding sub-score of ≥ 1 or final rectal bleeding sub-score of ≤ 1[21]. Clinical remission was defined as a partial Mayo Score ≤ 1. In CD patients, clinical response was defined as the decrease of Harvey Bradshaw index (HBI) score > 3 and clinical remission was defined as HBI score ≤ 4. Loss to follow-up was excluded.
This study recruited 24 healthy Chinese volunteers as healthy controls. Healthy Chinese volunteers were those who have regular bowel habits every day with the Bristol Stool Scale ranging from 3 to 4; newly appeared gastrointestinal symptoms (e.g., abdominal pain, diarrhea, nausea, vomiting) and newly appeared illness or general symptoms (various organ systems) for at least 6 months were excluded. The inclusion criteria for patients with IBD are the following: patients who were diagnosed as IBD based on a combination of typical clinical symptoms, endoscopy, and histological criteria for at least 6 months; patients with mild, moderate, and severe active UC (Mayo score from 3 to12) or CD (Harvey Bradshaw index ≥ 5); patients exhibited a poor response (defined by recognized international guidelines or expert consensus) and patients who failed to achieve satisfactory efficacy from conventional medications such as 5-aminosalicylic acid, steroids, cyclosporine, azathioprine, anti-TNF antibody, and traditional Chinese medicine; patients who were available of complete and paired clinical samples and associated clinical records before and after FMT treatment in patients with IBD who underwent microbiota transplantation. Excluded criteria are the following: patients missed blood samples; had a follow-up period < 3 months; under the age of 10 years; suffered from other intestinal diseases, e.g., Clostridioides difficile infection, or from other severe diseases, such as malignant neoplasm, serious liver or kidney diseases, or cardiopulmonary failure, refused to complete the follow-up, and underwent FMT.
Adolescents (preferably aged 6-24 years old) who passed questionnaire screening, face-to-face screening, and laboratory screening step-by-step are potential donors in the clinical practice according to our previous study. Before April 2014, fecal microbiota was prepared by manual methods. Beginning in April 2014, the method for preparation of washed microbiota is improved and based on the automatic microbiota purification system (GenFMTer, FMT Medical, Nanjing, China) followed with centrifugation plus suspension for three times in a Good Manufacture Practice (GMP) level laboratory room, which is a development on methodology for FMT[22,23,24].
All experimental protocols were approved by the animal ethics committee of Nanjing Medical University. The male C57BL/6 mice weighted 20-22 g at 6-8 weeks from Animal Center of Nanjing Medical University were used (each group n = 8). Antibiotics pretreatment included the following: Ampicillin: 0.5 g/L, vancomycin: 0.25 g/L, metronidazole: 0.5 g/L, neomycin: 0.5 g/L and gentamicin: 0.5 g/L along with 1 packet of artificial sweetener in per 250 ml of drinking water as previously described[25], beginning 1 week prior to 2.5% DSS (Dextran Sulfate Sodium, MW: 36,000-50,000 Da, MP Biomedicals, USA) administration. The supplementation was CA (400 mg/kg) and CDCA (100 mg/kg) at the dose of 0.2 mL per mouse[26]. (1) Acute modeling group (Fig. 6a): All mice treated with antibiotics for 1 week before 2.5% DSS administration. In injury phase, Ctrl and CA/CDCA groups drunk normal water and were gavaged with PBS. CA or CDCA once daily for 1 week. DSS and CA/CDCA + DSS groups drunk 2.5% DSS water and were gavaged with PBS, CA, or CDCA once daily for 1 week. In recovery phase, all mice had water for 2 days. (2) Treatment group (Fig. S7a): All mice treated with antibiotics for 1 week before 2.5% DSS administration. In injury phase, Ctrl group drunk normal water and were gavaged with PBS. CA/CDCA + DSS groups drunk 2.5% DSS water and were gavaged with PBS, CA, or CDCA once daily for 1 week. FMT treatment group drunk 2.5% DSS water and were gavaged with PBS, CA, or CDCA together with fecal microbiota for 1 week. In recovery phase, all mice had water for 2 days. (3) Chronic modeling group (Fig. S9a): Ctrl, CA, and CDCA groups treated with normal water for the whole process and were gavaged with PBS, CA, or CDCA for 1 month. Anti + Ctrl/CA/CDCA groups treated with antibiotics for 1-week before returning to normal water and were gavaged with PBS, CA, or CDCA for 1 month. Serum, liver, spleen, and colon were taken to evaluate the inflammation and for Elisa (IL-1b, IL-10, and TNF-a) and qPCR (Muc2, Occludin, ZO-1, IL-1b, IL-10, and TNF-a) examination.
Serum samples at baseline before FMT and 1 week after FMT and IBD patients were collected and stored at - 80℃ until untargeted metabolomics analysis based on liquid chromatography mass spectrometry (LC-MS/MS, Waters 2D UPLC system: Waters, Milton, MA; Q-Exactive mass spectrometer: Thermo Fisher Scientific, Wilmington, MA, USA, at a resolution of 70,000) [27]. High-resolution mass spectrometry (HRMS) was combined with the high-quality secondary spectrum information database (XCMS, mzCloud, Metlin, KEGG, HMDB, LIPIDMAPS) to match and identify the molecular characteristic peaks. Quality control (QC) samples were used (combined with BGILibrary) and a Coefficient of Variance (CV) less than 30% in QC samples were retained. Differential metabolites screening thresholds (OPLS-DA model): |log2FC|≥ 1 & OPLS-DA_VIP ≥ 1 & P-value ≤ 0.05, combined with the RSD of QC samples was less than 15%. Due to the lack of SeHCAT retention in China, the gold standard for quantifying BAM [28] and uncertain value of C4 in different studies, we used the highest upper limit (relative abundance based on LC-MS/MS) of C4 in the serum of healthy controls to define the pathological increase of C4 in IBD patients. Higher than the highest upper limit of healthy controls was defined as BAM, otherwise defined as non-BAM.
Stool samples at baseline before FMT and 1 week after FMT from healthy donors and IBD patients were collected and stored at - 80℃. And then stool samples were sequenced for the V3-V4 region of 16S rRNA genes. Microbial DNA extraction and sequencing were carried out using an Illumina MiSeq platform (Illumina, Inc., San Diego, CA, USA). The mothur (version 1.33.3, http://www.mothur.org/), UPARSE 7.1 (http://drive5.com/uparse), and R (version 4.0.2, https://www.r-project.org/) software applications were used for processing the raw 16S rRNA gene sequences, Operational taxonomic units (OTUs) clustering, and analysis [29]. Specifically, UPARSE pipeline was employed to cluster OTUs having a sequence similarity of ≥ 97%. Taxonomic differences were identified through linear discriminant analysis of effect size (LEfSe) and abundance levels of genera were considered in the LEfSe and Wilcoxon signed-rank test. In addition, principal coordinates analysis (PCoA) based on the distance matrix of Bray-Curtis dissimilarity was performed. Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) (http://picrust.github.io/picrust/tutorials/genome_prediction.html) program based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) database was used to predict the functional alteration of microbiota [30, 31]. The interaction between differential microbiota and specific KEGG pathway was conducted by the Mantel test [32]. To examine the interactions among different species and metabolites, we constructed co-occurrence patterns based on 16S rRNA and LC-MS/MS data using the Spearman correlation coefficient. The co-occurrence patterns between significant correlated taxa or microbiota were illustrated by Cytoscape version 3.10.1 (https://cytoscape.org/).
All analyses were carried out using IBM SPSS Statistics 23.0.0 (SPSS Inc., Chicago, IL, USA), R version 4.0.0 (R Development Core Team, Vienna, Austria), and Python 3.7.7. Continuous data were expressed with mean ± standard deviation (SD). Categorical data were described as numbers (percentages). We compared groups using unpaired Student's t test and one-way ANOVA for normal continuous variables, a Mann-Whitney U test for skewed continuous variables, and a chi-square test or Fisher's exact test for categorical variables. Differences were considered significant when P < 0.05. The Spearman correlation was used for correlation analysis. The differences between serum metabolites and fecal microbiota detected before and after FMT were analyzed by Wilcoxon matched pairs signed rank. Machine learning using the "caret" package was performed to examine the complex relationships among metabolites and thus predict patient outcomes as accurately as possible. Training and testing datasets were derived and used in the feature selection in the training dataset, and then random forest (RF) algorithms were used to develop a classification model. The sklearn package of python was used to process the samples according to ratio of the training set: testing set to be 0.8:0.2. The model parameters of the training set were adjusted by fivefold CV, and the optimal parameters were selected and evaluated by the test set. The performance of the markers was analyzed by calculating the area under the receiver-operating characteristic curve (AUC).