Bile acid profiling as an effective biomarker for staging in pediatric inflammatory bowel disease

ABSTRACT

Rapid and accurate clinical staging of pediatric patients with inflammatory bowel disease (IBD) is crucial to determine the appropriate therapeutic approach. This study aimed to identify effective, convenient biomarkers for staging IBD in pediatric patients. We recruited cohorts of pediatric patients with varying severities of IBD to compare the features of the intestinal microbiota and metabolites between the active and remitting disease stages. Metabolites with potential for staging were targeted for further assessment in both patients and colitis model mice. The performance of these markers was determined using machine learning and was validated in a separate patient cohort. Pediatric patients with IBD exhibited distinct gut microbiota structures at different stages of disease activity. The enterotypes of patients with remitting and active disease were Bacteroides-dominant and Escherichia-Shigella-dominant, respectively. The bile secretion pathway showed the most significant differences between the two stages. Fecal and serum bile acid (BA) levels were strongly related to disease activity in both children and mice. The ratio of primary BAs to secondary BAs in serum was developed as a novel comprehensive index, showing excellent diagnostic performance in stratifying IBD activity (0.84 area under the receiver operating characteristic curve in the primary cohort; 77% accuracy in the validation cohort). In conclusion, we report profound insights into the interactions between the gut microbiota and metabolites in pediatric IBD. Serum BAs have potential as biomarkers for classifying disease activity, and may facilitate the personalization of treatment for IBD.

KEYWORDS:

• Pediatric IBD
• intestinal microbiology
• bile acid
• serum biomarkers
• diagnosis and staging

Introduction

Inflammatory bowel disease (IBD) is a chronic, nonspecific inflammatory disease of the gastrointestinal tract that involves interactions between genetics, environmental factors, immunity, and the microbiome.^1,2 Crohn’s disease (CD) and ulcerative colitis (UC), the two main subtypes of IBD, differ in the affected region and presence of lesions.³ Reportedly, approximately 25% of IBD patients are diagnosed before age 20, and the incidence of IBD in children has risen worldwide over the past decade,⁴ presenting a heavy economic and healthcare burden. Compared to adults, pediatric patients with IBD are often associated with longer disease duration, more extensive disease, and systemic complications, such as growth retardation, malnutrition, and osteoporosis.^3,5 However, the scarce research on IBD in pediatric populations has limited the development of effective treatment strategies and management approaches.

Typical clinical symptoms of IBD include abdominal pain, diarrhea, and hematochezia.¹ Currently, the primary treatment paradigm for IBD involves a “step-up” or “top-down” strategy,⁶ both of which require careful, periodic monitoring of disease activity to evaluate the treatment response.^7,8 Multiple approaches have been developed to evaluate disease activity in pediatric IBD patients. Although endoscopy combined with biopsy is considered the most effective way to establish an IBD diagnosis and manage the disease, it is expensive, invasive, poorly tolerated by pediatric patients, and carries the risk of complications.⁹ Another approach is to use scores from the Pediatric Crohn’s Disease Activity Index (PCDAI) for CD and Pediatric Ulcerative Colitis Activity Index (PUCAI) for UC.¹⁰ However, these disease activity indices (DAIs) rely on subjective assessment of symptoms by patients and caregivers, and have limited utility in asymptomatic patients. Many laboratory-based IBD staging indicators, such as fecal calprotectin, C-reactive protein, cytokines, and perinuclear anti-neutrophil cytoplasmic antibody, have gained widespread usage in recent years,^11,12 but have limited accuracy.¹³ Further research to develop reliable, convenient diagnostic biomarkers specifically tailored to assess disease activity in pediatric IBD is crucial.

Multiomics has emerged as a reliable way to find new biomarkers.^14,15 While previous multiomics studies have compared individuals with IBD and healthy controls, and patients with the CD and UC subtypes,^16–18 insufficient attention has been given to the variance in different activity stages, particularly in children. In this study, we recruited three cohorts of pediatric patients with varying degrees of IBD severity. We conducted 16S rRNA gene sequencing and untargeted liquid chromatography-tandem mass spectrometry (LC-MS/MS) profiling to identify differences in specific microbial species and metabolites between these groups. Next, we targeted promising differential metabolites in serum and stool samples, validating our findings in an independent patient cohort from an alternate healthcare facility and in an animal model of colitis. Our primary objectives were to clarify changes in the microbiome and metabolome during disease progression and remission, and to identify promising classes of biomarkers to assess disease severity in pediatric IBD patients.

Results

We present our research approach in Figure 1, depicting the three cohorts we collected. Cohort A comprised 37 patients with active IBD, 33 patients with remitting IBD, and 11 healthy controls from Ruijin Hospital, Shanghai, China. Cohort A underwent fecal 16S rRNA testing and fecal metabolomics analysis. Cohort B, also from Ruijin Hospital, comprised 59 patients with active IBD, 80 patients with remitting IBD, 61 patients with enteritis other than IBD, and 38 controls without apparent gastrointestinal inflammation, whose serum bile acid (BA) profiles were used to construct a diagnostic model. Cohort C comprised 17 patients with active IBD and 31 patients with remitting IBD from Shanghai Children’s Hospital, Shanghai, China, whose data were used to validate the diagnostic potential of serum BAs. For patient classification, all IBD patients, regardless of type and stage, were categorized as IBD-Any, while active IBD patients were denoted as IBD-Act, remitting IBD patients as IBD-Rem, and healthy controls as HC.

Figure 1. Schematic diagram of the techniques used to study the three patient cohorts and model mice. In cohort A, stool samples were collected for multiomics testing. The number of collected stool samples varied among patients, resulting in different patient sample sizes for each omics dataset. In cohorts B and C, serum samples were collected for bile acid testing. Cohort B was used to establish the diagnostic model, and cohort C was used to verify it. Additionally, a colitis mouse model was employed to investigate serum bile acid changes and establish cross-species homogeneity.

Intestinal microbiota characteristics in pediatric IBD patients

First, we compared the microbiota composition between IBD-Any and HC, which revealed lower alpha diversity in the gut microbiota of the IBD patients (Supplemental Figure S1(a)). However, when we performed Principal Coordinates Analysis (PCoA) to assess beta diversity, we did not find any significant differences between the two groups (Supplemental Figure S1(b)). To investigate the possibility that the lack of distinction between the groups was a result of heterogeneity within IBD-Any, we further categorized the IBD patients by disease stage (remitting or active). Patients in remission had a similar microbiome structure to that of controls, while patients in the active stage showed significant discrepancies (Figure 2(a)). When we compared the microbiota between patients with different disease subtypes (CD and UC) and localizations, there were no noticeable differences (Supplemental Figure S1(c,d)). However, our analysis did reveal significant correlations between the principal component 1(PC1) axis of the microbiota and two common disease severity indicators: fecal calprotectin and the DAI (Figure 2(b)).

Figure 2. Alterations of gut microbial structures in IBD. (a) PCoA of IBD-Act, IBD-Rem, and HC based on Bray-Curtis dissimilarity metrics, with statistical significance and variance of dissimilarity data assessed using ANOSIM. (b) Correlations between PC1 of microbiota and DAI (PCDAI/PUCAI: n = 70; fecal calprotectin: n = 58) in all IBD patients, as analyzed by Pearson correlation. (c) Average relative proportions of the main phyla and genera in the three groups. The significantly increased phylum in active-phase patients is mainly Proteobacteria, while the significantly decreased phylum is mainly Bacteroidota. (d) LEfSe used to identify essential differences in bacterial abundance (from phylum to genus levels) between IBD-Act and IBD-Rem (LDA threshold > 4.0). Differentially abundant microbial species are highlighted using red and blue boxes in panels D, with red indicating an increase in active IBD and blue indicating a decrease in active IBD. (e) Heatmap of Spearman’s correlation between IBD severity-related microbial taxa and DAI; *p < .05, **p < .01, ***p < .001).

To examine differences between patients with active and remitting IBD, we conducted multilevel species discriminant analysis using a threshold of linear discriminant analysis (LDA) > 4 (Figure 2(c,d)). At the phylum level, there was a significant increase in Proteobacteria, a noticeable decrease in Bacteroidetes, and a reduction in the Clostridia class of the Firmicutes phylum in patients with active IBD compared with remitting IBD. At the genus level, patients with active IBD exhibited greater proportions of Escherichia-Shigella and Haemophilus, while patients with remitting IBD showed dominance by Bacteroides, Phascolarctobacterium, and Lachnoclostridium. Furthermore, we found significant correlations between the abundance of each of these bacteria and the DAI (Figure 2(e)). These findings indicate the significant role of microbiota in modulating disease activity in pediatric patients with IBD.

Fecal metabolomic profiling corresponds to the severity of IBD

Recognizing the crucial role of metabolites in facilitating communication between bacteria and hosts, we expanded our analysis by conducting untargeted metabolomics studies on the fecal samples from patients with active or remitting IBD and controls. The PCoA plot depicted clear group distinctions in the compositions of fecal metabolites, especially between patients at different activity stages (Figure 3(a)). Moreover, the PC1 of metabolites was significantly correlated with the levels of fecal calprotectin and the DAI, indicating an association between the metabolome and disease progression (Supplemental Figure S2(a)).

Figure 3. Changes in fecal metabolites and their correlations with the microbiota during IBD progression. (a) PCoA of IBD-Act, IBD-Rem, and HC based on Bray-Curtis dissimilarity metrics, with statistical significance and variance of dissimilarity data assessed using ANOSIM. (b)Volcano plot showing the differential metabolites between IBD-Act and IBD-Rem. Blue and red dots represent depleted and enriched metabolites, respectively (VIP >1 and FDR < 0.5, corrected by BH). (c) T-statistics for each compound from an independent sample t-test between IBD-Act and IBD-Rem. Wilcoxon signed-rank tests were applied on the individual differential abundance trends of metabolites. p values were corrected for multiple testing using a BKY two-stage linear step-up procedure. Only metabolite classes containing ≥ 10 putative members in the HMDB subclass dataset are displayed. Boxplot ‘boxes’ indicate the first, second, and third quartiles of the data. (d) Bubble diagram illustrating the KEGG enrichment analysis. The size of each bubble represents the number of metabolites enriched in the pathway, and the color gradient indicates the significance of enrichment. Pathways with an adjusted p value < 0.05 were annotated with a box. Enrichment analysis was conducted using Fisher’s exact test, and the resulting p values were corrected for multiple testing using the BH method. (e) Integrated network of IBD severity-related microbial species and fecal metabolites depicted using the compound Spring Embedder layout method. Metabolites are represented by different shapes based on their HMDB classification. The size of the edge indicates the degree of Spearman’s correlation. Direct correlations are highlighted as red edges, while inverse correlations are portrayed as blue edges, with a cutoff of p < .05 and r > 0.6 or r<–0.6.

Of the 2562 metabolites detected, patients with active IBD exhibited significant increases in 141 metabolites and significant decreases in 440 metabolites compared to those in remission (Variable Importance in Projection [VIP] scores > 1; false discovery rate [FDR] <0.05; Figure 3(b)). Next, we classified the differential metabolites by annotation in the Human Metabolome Database (HMDB), and focused on metabolite classes containing at least 10 putative members in the HMDB subclass dataset. Using rank-based enrichment analysis, we identified molecular categories that were significantly increased or decreased between the active and remitting phases of IBD (Figure 3(c)). Compared to IBD-Rem, IBD-Act exhibited significant increases in glycerophospholipids and steroidal glycosides, with a notable depletion of amino acids, carbonyl compounds, prenol esters, vitamin D, BAs, and fatty acid-related molecules. We proceeded to categorize the distinctive metabolites on the basis of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment to examine their potential function. This analysis revealed that bile secretion, choline metabolism in cancer, vitamin digestion and absorption, and sphingolipid signaling were the primary affected pathways between the two stages of IBD (FDR threshold < 0.05; Figure 3(d)).

The most significantly altered pathway was bile secretion, which is associated with a series of BAs, including deoxycholic acid (DCA), taurocholic acid (TCA), and their derivatives (Supplemental Figure S2(b)). Additionally, phospholipid metabolism displayed a disturbance, as evidenced by increased levels of lysophosphatidylcholine and a concurrent decrease in its hydrolysis product diacylglycerol (Supplemental Figure S2(c)). This finding is consistent with several previous studies on targeted sphingolipid metabolism among pediatric patients with IBD. ^19,20

Given the intimate relationship between the microbiota and metabolome in the intestine, we conducted a comprehensive association study. The application of Inter-omic Procrustes analysis on fecal samples revealed a significant similarity (Supplemental Figure S2(d): Monte Carlo p = .001). Furthermore, we observed a strong correlation between the PC1s of the microbiota and metabolome (Supplemental Figure S2(e)). Subsequently, we used a threshold of r > 0.6 and p < .05 (Figure 3(e)) to identify Escherichia-Shigella, Bacteroides, Parabacteroides, and Lachnoclostridium as core microbial species that are significantly correlated with several lipids, including triacylglycerols, BAs, and sphingolipids. These results suggested that microbiota-mediated lipid metabolism may play a critical role in IBD pathogenesis. Overall, these findings provide novel insights into the underlying mechanisms linking microbial dysbiosis and metabolic disorders in IBD.

Targeted metabolomics sheds light on BA abnormalities in pediatric IBD

Because the levels of a number of BAs exhibited significant differences between active and remitting IBD, we performed targeted metabolomic analysis to elucidate the relationship between BA metabolism and disease activity in pediatric IBD. Altogether, we identified a total of 43 distinct BAs in the stool samples obtained from Cohort A. Consistent with the untargeted metabolic findings, we observed a significant divergence in BA composition between individuals at different stages of disease activity (Figure 4(a)). We identified key differential metabolites by setting a threshold of VIP > 1 and p < .05. Remarkably, pediatric IBD patients showed an increased concentration of primary BAs, particularly during the active stage, together with a reduced concentration of secondary BAs (Supplemental Figure S3(a,b)). Additional correlation analysis employing BA classification revealed that primary conjugated BAs, including TCA, glycocholic acid (GCA), taurochenodeoxycholic acid (TCDCA), and taurohyocholic acid (THCA), were positively associated with disease activity. Conversely, secondary unconjugated BAs, including ursodeoxycholic acid (UDCA), DCA, lithocholic acid (LCA), and isolithocholic acid (isoLCA), were negatively correlated with disease activity (Figure 4(b)). Considering the bioprocessing-driven conversion of primary to secondary BAs by intestinal bacteria, we constructed a correlation network of representative microbial categories and fecal BAs found in pediatric IBD patients (Figure 4(c)). By employing the Group Attributes Layout of Cytoscape to categorize the BAs, we uncovered a significant negative correlation between primary and secondary BA types. Our analysis further revealed that that Escherichia-Shigella exhibited a positive correlation with primary BAs, but a negative correlation with secondary BAs. In contrast, Bacteroides, Phascolarctobacterium, Parabacteroides, and Lachnoclostridium showed the opposite correlation with primary and secondary BAs (Supplemental Figure S3(c)). These findings highlight the need for further comprehensive investigations into the dynamic interactions between patients and microbes. Likewise, the potential diagnostic and therapeutic implications of targeting BAs and their associated bacteria should be considered.

Figure 4. Changes in fecal bile acid composition in pediatric IBD. (a) PCoA of IBD-Act, IBD-Rem, and HC based on Bray-Curtis dissimilarity metrics, with statistical significance and variance of dissimilarity data assessed using ANOSIM. (b) Heatmap of Spearman’s correlation between differential fecal bile acids and clinical indicators; *p < .05, **p < .01, ***p < .001. primary and secondary bile acids are depicted by orange and green bars, respectively. (c) Integrated network of IBD severity-related microbial species and fecal bile acids using the group attributes layout method of cytoscape. The size of the edge indicates the degree of Spearman’s correlation. Direct correlations are highlighted as red edges, and inverse correlations as blue edges, with a cutoff of p < .05 and r > 0.4 or r<–0.4. Bile acid categories that could not be accurately classified are not listed.

IBD clinical activity correlates with serum BA profiles

Although previous studies have demonstrated the diagnostic potential of fecal BAs, it can be challenging to implement such findings given the difficulty obtaining stool samples from pediatric patients. As an alternative, considering the unique metabolic process of enterohepatic circulation of BAs,^21,22 in which approximately 95% of intestinal BAs are recirculated to the liver via the portal vein, we sought to establish a novel, convenient set of diagnostic BA markers using serum.

First, we measured and analyzed the levels of 15 common BAs found in serum in four groups of children from Cohort B (IBD-Act, IBD-Rem IBD, Enteritis, and Control) and identified the BA categories with significant between-group differences using a multiple comparison test (Figure 5(a)). Compared with the Control, IBD-Act exhibited particularly obvious reductions in six secondary BAs: GLCA, GUDCA, UDCA, TDCA, GDCA, and DCA. Compared with IBD-Rem, IBD-Act showed increased levels of primary BAs, such as GCA and TCA, and decreased levels of secondary BAs, such as DCA, UDCA, and GUDCA (Supplemental Figure S4(a)). To account for the synthesis and transformation of BAs, we also developed the following comprehensive indices: ratio of primary BAs to secondary BAs (PBA/SBA), ratio of DCA to cholic acid (CA) plus DCA (DCA/(CA+DCA)), and ratio of conjugated BA to unconjugated BA. Remarkably, there were significant differences in these indices between IBD-Act and IBD-Rem patients (Figure 5(b)), and each index was significantly correlated with the fecal concentration of calprotectin in IBD-Any patients (Supplemental Figure S4(b–d).

Figure 5. The composition of serum bile acids in pediatric IBD patients and DSS colitis mice. (a) Graph illustrating the absolute concentration of serum bile acids among the four groups of cohort B; *p < .05, **p < .01, ***p < .001 between IBD-Act and IBD-Rem; #P < .05, ##P < .01, ###P < .001 between IBD-Act and control. (b) Comparisons of three comprehensive BA indices between the four groups. For panels a and B, hypothesis testing was conducted using the Kruskal-Wallis test, followed by multiple comparisons and corrections using Dunn’s test. (C) Schematic diagram of the murine colitis model induced by DSS. (d) Representative measurements of serum bile acids in the three mouse groups. Hypothesis testing was performed using one-way ANOVA, while multiple comparisons and corrections were carried out using the Bonferroni method. (e) Pearson correlation analysis of the PBA/SBA ratio and disease indicators in all DSS mice; *p < .05, **p < .01, ***p < .001.

To mitigate the impact of potentially confounding factors related to individual patient differences on our results, we conducted similar microbial and metabolic investigations using the DSS mouse model. By manipulating the concentration of DSS, we obtained model mice with both mild and severe colitis (Figure 5(c)). The validity of the colitis modeling was confirmed by histopathological scoring of colon tissue sections, changes in body weight, and DAI scoring, which exhibited significant variation among the mild, severe, and control groups (Supplemental Figure S5(a,b)). We observed notable differences in the serum BA profile between DSS mice and control mice (Supplemental Figure S5(c)). These changes were characterized by increased levels of CA, and were accompanied by marked changes in the PBA/SBA ratio (Figure 5(d)). Furthermore, the PBA/SBA ratio was significantly correlated with disease activity indicators, including the histopathological score, DAI, and end-weight (Figure 5(e) and Supplemental Figure S5(d)). The qPCR analyses revealed that, as the disease progressed, Escherichia coli abundance increased while Bacteroides vulgatus abundance decreased (Supplemental Figure S5(e)). The results of our prior study of microbiota changes in DSS mice using 16S rRNA sequencing²³ were consistent with the present findings (Supplemental Figure S5(f)). In summary, our findings thus far revealed cross-species homogeneity of serum BA changes in IBD.

Evaluation of the diagnostic performance of specific microbial taxa and metabolites

Next, we aimed to apply our findings to the diagnosis and follow-up of patients with IBD, particularly in the staging of disease activity. The diagnostic efficacy of microbial genera that exhibited significant variance between the active and remitting periods was assessed (Supplemental Figure S6(a)), with each displaying an area under the ROC curve (AUC) surpassing 0.7. However, utilization of a single genus of bacteria as a clinical biomarker may be impractical. Therefore, to explore suitable serum biomarkers, we opted to utilize serum BAs to train an RF classifier. Using the 10-fold cross-validation method, we assessed the predictive accuracy and discriminatory performance of the classifier in 139 patients from Cohort B (Supplemental Figure S6(b)). After assessing the performance of various serum BA categories using ROC curve analyses, we identified GUDCA and UDCA as having the highest AUC values. These categories were combined with 13 other serum BAs for a combined index with a diagnostic efficiency of 0.88 (Figure 6(a)). Regarding ease of use and clinical practicality, PBA/SBA may prove to be the most suitable for widespread clinical application, exhibiting an AUC of 0.84 (Figure 6(a)). Notably, among 84 patients who underwent traditional stool-based fecal calprotectin detection (AUC = 0.85), PBA/SBA performed even better (AUC = 0.90). Furthermore, when combined, the diagnostic efficiency of the two markers exceeded 0.9 (Figure 6(b)).

Figure 6. Construction and validation of diagnostic models. (a) ROC curves illustrating the potential of serum BAs and comprehensive indices as biomarkers for staging pediatric IBD. (b) Comparison of the diagnostic sensitivity of fecal calprotectin, serum PBA/SBA, and the combination. (c) Confusion matrix of the diagnostic performance of PBA/SBA in cohort C. Clinical diagnosis was based on endoscopic and mobility scores. (d) PBA/SBA analysis at different disease phases in the same patient using a paired t-test. CI, confidence interval.

We validated our diagnostic model in a separate and distinct patient cohort (Cohort C; n = 48) to ensure that the test performance would not benefit from potential overfitting of the RF classifier to its training data. Once again, there was a noteworthy link between PBA/SBA and disease activity, as demonstrated by the DAI and calprotectin levels (Supplemental Figure S6(c,d)). The ability of PBA/SBA to distinguish patients with active versus remitting disease was assessed using a confusion matrix. Indeed, 77% of patients were classified accurately (Figure 6(c)), with misclassifications primarily comprising patients in remission as having some degree of active disease. Notably, many patients in remission exhibited PBA/SBA values concentrated near the critical value, which may have contributed to a degree of error. In contrast, the accuracy of PBA/SBA in predicting active disease was 94%.

We also conducted a separate analysis of IBD patients with a change in DAI ≥ 10 or a change in clinical stage from both cohorts B and C. Specifically, we used a paired t-test to compare serum BA levels within the same patient. Our findings revealed that 20 of 23 eligible patients (86.96%) exhibited higher PBA/SBA levels during active disease (Figure 6(d)). This approach accounted for potential inter-individual differences within the group, further supporting the potential of PBA/SBA as a valuable tool for the evaluation of treatment response in pediatric IBD.

Discussion

The dynamic nature of IBD, which often results in considerable variability in disease activity, sometimes within the same patient, poses challenges in devising appropriate treatment strategies and personalized care plans. Therefore, the need to identify precise, easily accessible biomarkers is paramount in the management of IBD. Here, we implemented a comprehensive multiomics analysis utilizing multiple cohorts of pediatric IBD patients to discover microbial and metabolic indicators corresponding to distinct disease activities. We then formulated and evaluated the diagnostic efficacies of different serum BA measurements, substantiating their potential as diagnostic tools.

The strong association between IBD and gut dysbiosis is well established.^24,25 An important conclusion of our study was the existence of distinctive microbial composition signatures at the active and remitting stages of pediatric IBD. The intestinal bacterial composition of patients in remission exhibited obvious similarities to that of healthy controls, with a convergence of dominant species, hereby termed “Bacteroides-dominant enterotype”. Furthermore, related enriched species such as Bacteroides ²⁶ and Eubacterium ²⁷ often serve as producers of metabolites, including short-chain fatty acids, indole derivatives, and secondary BAs, which have widely recognized anti-inflammatory effects. Certain bacterial species linked to clinical remission have demonstrated effectiveness as standalone microbiome-modulating therapies.²⁸ By contrast, our patients in the active phase of disease exhibited a bacterial composition that we termed “Escherichia-Shigella-dominant enterotype”. This enterotype was associated with enrichment of Escherichia-Shigella, Haemophilus, and other bacteria that are closely linked to the production of multiple virulence factors, such as adhesins and aggregation substances.^29,30 Indeed, previous studies documented differences in gut microbial composition between adult IBD patients and healthy individuals; namely, a decrease in the phylum Firmicutes and an increase in the phylum Proteobacteria in IBD patients.³¹ Our study enriches these results in new cohorts of pediatric patients. While gut microbiome dysbiosis in children with IBD is a topic of ongoing discussions, we must acknowledge that the link between gut dysbiosis and IBD is likely to be intricate and ever-changing, rather than a simple cause-and-effect relationship.³² Inflammatory conditions, such as the destruction of the intestinal mucus layer, elevated pH levels, increased oxygen content, and abnormal activation of immune cells, induce changes in the intestinal microenvironment.³³ These changes further promote the proliferation of anaerobic microorganisms, upsetting the balance between different species and ultimately disrupting the ecology of the host microbiota. Translating these insights into practical diagnoses and treatments can be challenging.

As central regulators of metabolic disorders, metabolites derived from the gut microbiota hold unquestionable significance. We focused on exploring clinical and translational research in the direction of BAs. BAs exert their effects in various host tissues, primarily through two receptors: farnesoid X receptor (FXR) and G-protein-coupled BA receptor 1 (GPBAR1; also known as TGR5). The activation of TGR5 promotes the proliferation and reconstruction of intestinal epithelial cells and can inhibit the inflammatory response.³⁴ Similarly, FXR plays a major role in regulating inflammation, immunity, and liver regeneration, in addition to being a regulator of BA synthesis.³⁵ Knockout and activation experiments in mice showed that Fxr and Tgr5 were closely related to the occurrence and remission of colitis.^36,37 The secondary BAs LCA and DCA are the most important ligands of FXR and TGR5.^38,39 Additionally, certain BA metabolites, such as LCA, 3-oxoLCA, and isoalloLCA, can directly regulate the balance between T helper 17 and regulatory T cells, thereby influencing the host’s immune response.⁴⁰ Integrating the microbial transformation of BAs (primarily bile salt hydrolase [BSH] and 7-alpha hydroxylase-associated microorganisms) with their importance in pattern recognition by the immune system, the significance of BAs as a highly distinct metabolite class becomes evident. (Supplemental Figure S7).

Alteration of the fecal BA profile in adults with IBD has been extensively investigated.⁴¹ In this study, considering the uniquely effective enterohepatic circulation of BAs, we utilized serum BA levels to indirectly reflect the composition of fecal BAs and thus serve as an ideal indicator for disease staging. Moreover, serum BA measurements are a reliable, established biomarker for various diseases, including intrahepatic cholestasis,⁴² colon cancer,⁴³ and Alzheimer’s disease.⁴⁴ Notably, there is reportedly an association between serum secondary BA concentrations and the likelihood of achieving remission with anti-cytokine therapy in adult IBD, emphasizing the potential value of serum BAs as biomarkers.⁴⁵ Nevertheless, to our knowledge, our study is the first to integrate fecal and serum BA metabolic profiles for stratification of pediatric IBD patients, and includes the diagnostic performance assessment of serum BA profiles in two cohorts. In addition to detecting known BAs (e.g. DCA, LCA, GCA), we found that hyocholic acid and its derivatives also showed significant differences in IBD. Considering their reported roles in other disease states, such as diabetes⁴⁶ and sepsis,⁴⁷ we suggest that the therapeutic or diagnostic potential of these uncommon bile acids in the context of IBD warrants further investigation.

There are some limitations to this study. First, although we performed complete metabolomics and 16S rRNA detection, the number of samples for untargeted metabolic profiling was comparatively small (n = 32), which limited the construction of the interaction network. Larger cohort studies are needed to verify our results. Second, our microbial detection was limited to 16S rRNA sequencing, which is not as accurate as metagenomic sequencing for inferring the functions of specific microbial categories. To ensure the accuracy of our results, we limited the description of microbial characteristics to the genus level. Furthermore, although we verified our results in a DSS mouse model, the roles of specific BA classes in the pathogenesis of IBD require mechanistic exploration.

In summary, we employed a systematic multiomics approach utilizing fecal and blood samples to investigate the interplay between the gut microbiota and metabolites in pediatric IBD progression. We successfully established a new serum BA biomarker, with the PBA/SBA ratio as its core indicator. Our findings provide a foundation for personalized treatment of IBD, and insights that may lead to elucidation of the underlying mechanisms of this disease.

Materials and methods

Study design

For this descriptive, observational study, we enrolled three cohorts between 2021 and 2023, all of which comprised individuals aged ≤18 years. Participants with severe enteric infections, or liver or kidney disease, were excluded. And none of the participants involved in our present study had a history of cancer. Baseline information of patients on the day of admission from each patient were collected, including age, gender, BMI, white blood cell count, fecal calprotectin, interleukin-6.

IBD was defined in accordance with recommendations from European Crohn’s and Colitis Organization and European Society of Gastrointestinal Endoscopy,¹³ and the phenotyping was based on endoscopy and radiological findings as described in the Paris Modification of the Montreal Classification for IBD.⁴⁸ Additionally, as the reference standard, endoscopic scores and clinical activity scores were calculated to determine the staging of the disease. Patients in remission were identified as having a colonoscopy score ≤ 3 or a PCDAI/PUCAI ≤10, while active IBD patients were identified as having a colonoscopy score > 3 or a PCDAI/PUCAI >10. If there was a conflict between the two measures, endoscopic examination results were given priority. The study was approved by the Ethics Committee of Ruijin Hospital, with the ethics number C0168CRD4020, all patients (or their guardians) signed the consent form allowing the use of their bodily fluids and other specimens for scientific research upon admission. More details on the basic information of patients can be found in the online supplemental material 1.

Sample collection and processing

Stool samples were collected from patients immediately following defecation, frozen at − 20°C, and transferred to a − 80°C freezer for long-term storage. Each specimen was divided into three parts (if applicable): one for intestinal microbial composition analysis via 16S rRNA sequencing, one for gut metabolomic profiling via untargeted LC-MS/MS analysis, and one for BA metabolomics via targeted LC-MS/MS. Fecal microbiota and metabolite analyses were performed by Majorbio (Shanghai, China) and are described in further detail in online supplemental material 2.

Similarly, serum samples were collected and frozen at − 20°C. The detection of serum BAs was performed using the AB SCIEX Triple Quad™ 4500MD.LC-MS samples were prepared from 50 μl of serum by protein precipitation using methanol, with the addition of a bile acid internal standard solution. The samples were centrifuged at 14,000 rpm for 10 minutes at 4°C and then diluted at a ratio of 1:2. A 20 μl aliquot was taken for LC-MS/MS analysis. Phenomenex columns were used for chromatographic separation, with a mobile phase consisting of water (with 0.1% EBBA1) as solvent A and methanol as solvent B. The column temperature was set at 50°C, and a 9-minute gradient elution was performed. Mass spectrometry scans were conducted using an electrospray ionization (ESI) source and multiple reaction monitoring (MRM) mode. The ion source utilized was Turbo V™.

Dextran sodium sulfate (DSS)-induced colitis mouse model

Male mice aged 6 weeks were purchased from Charles River (Beijing, China) and housed at a constant temperature (20°C −22°C) with a 12-hour light-dark circle under specific pathogen-free conditions. The mice were acclimatized for 1 week before modeling. To induce acute colitis, the mice were administered DSS (molecular weight: 36000–50,000 Da; MP Biomedicals, California, USA) in drinking water for 5 days, followed by regular drinking water for 2 days. The mild colitis group received 1.8% (w/v) DSS and the severe colitis group received 2.5% (w/v) DSS. During the modeling, the mice were weighed and scored daily. Disease activity scores were based on weight loss, stool consistency, and bleeding, as described previously.⁴⁹ Samples, including serum, and colonic tissue and contents, were collected on day 9. Histological analysis and real-time quantitative polymerase chain reaction are described in online supplemental material 3. All research reported in submitted papers has been conducted in an ethical and responsible manner. The animal experiments were approved by the Ethics Committee of Ruijin Hospital.

Statistical analysis

Statistical analyses were conducted using SPSS software version 16.0. A p value of < 0.05 was considered to indicate statistical significance. Data were generated using GraphPad Prism software version 8.0 and R software version 4.0.5 for graphical representation. Statistical comparisons between two groups were performed using either Student’s t-test or the Mann-Whitney test, depending on the normality of the data distribution. Correlations between species, metabolites, and clinical factors were assessed using Pearson correlation analysis. Rank-based enrichment analysis was performed using the Wilcoxon signed-rank test. Specific statistical methods and the correction of p-value can be found in online supplemental material 4.

To identify the most effective biomarkers, the random forest (RF) algorithm from the R randomForest package was employed. Feature selection was performed using the Recursive Feature Elimination with Cross-Validation algorithm, employing 10-fold cross-validation and 500 decision trees. The diagnostic performance of biomarkers was evaluated using the receiver operating characteristic (ROC) curve from the R plotROC package.

Acknowledgments

We thank Michelle Kahmeyer-Gabbe, PhD, from Liwen Bianji (Edanz) (www.liwenbianji.cn) for editing the English text of a draft of this manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The sequence dataset generated from this study has been deposited in the NCBI database under BioProject number PRJNA998101 in https://www.ncbi.nlm.nih.gov/bioproject/. The data that support the findings of this study are available from the corresponding author upon reasonable request.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19490976.2024.2323231

Additional information

Funding

This work was supported by the National Natural Science Foundation of China (82172324, 81971993, 82002200, and 82202581).