Inulin changes gut bacteria chemicals linked to brain function in children with obesity

This study was designed as a randomized, double-blind, placebo-controlled trial carried out from August 2017 to July 2020 at King Chulalongkorn Memorial Hospital (KCMH) in Thailand. A comprehensive description of the protocol can be found in a previous publication¹⁸. The study involving human participants, materials, and data adhered to the principles outlined in the Declaration of Helsinki. The protocol was approved by the Institutional Review Board of the Faculty of Medicine, Chulalongkorn University (IRB no. 240/60). Informed consent was obtained from all subjects and/or their legal guardian(s). The trial was registered on clinicaltrials.gov under the identifier NCT03968003 (Registered 30/05/2019). The inclusion and exclusion criteria were identical to those used in our previous study¹⁸. Briefly, children aged 7 to 15 years with obesity, defined as a body mass index (BMI) exceeding the median plus two standard deviations (SDs) based on the World Health Organization growth reference¹, were eligible for inclusion. Exclusion criteria comprised syndromic or monogenic obesity, endocrine causes of obesity (e.g., hypothyroidism or growth hormone deficiency), use of medications known to affect appetite or body weight, allergy to inulin, and participation in other concurrent weight reduction programs.

The complete study design is described elsewhere¹⁸. In summary, 165 participants were randomly divided into three groups. Participants in the inulin group consumed 13 g of extracted inulin powder from Thai Jerusalem artichoke daily for 6 months, approximately 30 min before dinner. The powder was mixed with 150 ml of warm water and stirred until dissolved. The placebo group consumed 11 g of isocaloric maltodextrin (Oligocarb; Ma-Jusmin Company Limited, Bangkok, Thailand) daily, using the same preparation method. The inulin extraction process was performed according to our patented method (Inventors: Chonnikant Visuthranukul and Supakarn Chamni, Chulalongkorn University and National Science and Technology Development Agency, Thailand, Thai Patent No. 15858, issued January 23, 2020). The third group was given dietary fiber guidance that included portion size illustrations to age-appropriate intake¹⁹. All participants were provided with the same general advice regarding diet, exercise, and behavior modification, along with monthly follow-ups for six months.

The full details are available in a previous publication¹⁸. In summary, a dietitian assessed dietary intake using 3-day dietary records, with daily energy and nutrient intake calculated via the Institute of Nutrition, Mahidol University Calculation-Nutrients (INMUCALs) Version 3²⁰. Physical activity was evaluated through questionnaires. Trained personnel collected anthropometric and body composition data, following methods from the prior study¹⁸.

Blood was drawn at the KCMH after participants fasted overnight. Citrate vacutainers underwent centrifugation for 15 min at 1,500×g at room temperature. Following this, 120 µL aliquots were taken for plasma and preserved at -80 °C until use. The AbsoluteIDQ^® p180 kit, supplied by Biocrates Life Science AG (Innsbruck, Austria) and Helena Thai Laboratories co., Ltd (Bangkok, Thailand), included all essential reagents, internal and calibration standards, quality controls, test mix, and a 96-well filter plate. This kit employs an automated assay that utilizes phenylisothiocyanate (PITC) derivatization to analyze target analytes in biological fluids, with internal standards used for precise quantification. Amino acids and biogenic amines were detected using liquid chromatography-mass spectrometry (LC-MS). For plasma sample preparation, as per the manufacturer's protocol, 10 µL of the sample was applied to the 96-well plate and dried under nitrogen. A 5% PITC solution (50 µL) was then added to derivatize the amino acids and biogenic amines, followed by another drying step. Metabolites were extracted using 300 µL of 5 mM ammonium acetate in methanol, transferred to the lower 96-well plate, and further diluted with MS solvent A (comprised 0.2% formic acid in water) for analysis. Quantification was based on 7 internal standards and a calibration curve with human plasma-based quality controls in 3 concentration levels (low, medium, high)²¹. The list of all amino acids and biogenic amines measured using the AbsoluteIDQ^® p180 Assay is provided in Supplementary Table S1. Our primary hypothesis for this study specifically focuses on the GBA, and therefore, our discussion and in-depth analyses have focused on compounds directly involved in this pathway, which were Glu, glutamine (Gln), histidine (His), histamine, Trp, Tyr, dopamine, putrescine, spermidine, spermine, and 5-HT.

The LC-MS/MS configuration included an ACQUITY UPLC system (Waters, Milford, MA, USA) connected to a QTRAP5500 mass spectrometer (AB Sciex, Redwood City, CA, USA) operating in electrospray ionization (ESI) mode. This setup was used to analyze amino acids and biogenic amines in positive mode via LC-MS. A volume of 2 µL from the sample extract was applied to an ACQUITY BEH C18 column (2.1 × 7.5 mm, 1.7 μm), kept at a temperature of 50 °C. The method featured a 7.3-minute solvent gradient, where solvent A and solvent B contained 0.2% formic acid in acetonitrile.

The full details are described in a previous publication²². In brief, after a 12-hour fast, venous blood samples were collected to assess inflammatory cytokines, interleukin-1β (IL-1β), IL-6, and tumor necrosis factor-α (TNF-α) at baseline and month 6. Plasma inflammatory cytokines were measured with Bio-Plex Pro™ Human Cytokine Assays (Bio-Rad, USA) using the Bio-Plex Suspension Array System, based on Luminex xMAP technology.

Baseline fresh fecal samples were collected for gut microbiota and SCFAs analysis, as described in a prior publication^5,23. Participants received sterile collection kits and were instructed to gather samples at home. Each sample was placed in a 50 ml sterile container, sealed in a zip-lock bag, and stored in a home freezer (around -20 °C) before being delivered to the lab within 24 h. Upon receipt, samples were stored at -80 °C until analysis. Illumina paired-end reads of 16S rRNA were then analyzed as previously outlined. For SCFA analysis, fecal samples were 10-fold diluted with phosphate-buffered saline (pH 8.0) and homogenized using a stomacher blender (Stomacher^® 80 Biomaster; Seward, Worthing, UK) for 5 min, adapted from Kisuse et al.²⁴. The resulting 1 mL fecal slurry was then centrifuged at 13,000×g for 5 min, with the supernatant stored at -80 °C until analysis. Lactic acid and SCFAs (acetic, butyric, and propionic acid) were quantified from these prepared samples via high-performance liquid chromatography (HPLC) (Water 1525, USA). HPLC sample preparation followed previously established, modified methodologies^25,26, with all analytical parameters assessed using Agilent Technologies 7890 A equipment (Santa Clara, USA).

Participants provided 12-hour fasting serum samples, which were immediately placed on a test tube rack to clot at room temperature for 10-20 min. Within 30 min of collection, the blood underwent centrifugation at 1,820×g for 20 min. The resulting serum samples were then preserved at -80 °C until subsequent analysis. Levels of glucagon-like peptide 1 (GLP-1) and peptide YY (PYY) were determined using Human GLP-1 and Human Peptide YY Enzyme-Linked Immunosorbent Assay (ELISA) Kits (MyBioSource, Inc., San Diego, CA, USA), respectively. Both hormones were quantified using the Sandwich ELISA technique. An experienced technician performed the analysis and calculated results using computer-based curve-fitting software. The GLP-1 assay had a detection range of 15 to 3,000 ng/L with a sensitivity of 7.29 ng/L, while the PYY assay ranged from 3 to 900 pg/mL with a sensitivity of 1.56 pg/mL.

A principal component analysis (PCA) was applied to amino acid and biogenic amine profiles at both baseline and after the intervention to visualize changes in the metabolite profiles across the placebo, inulin, and dietary fiber advice groups over the six-month period. To determine the statistical significance of the PCA, we applied the PCAtest function²⁷, through random permutations to generate null distributions for key PCA statistics. These included testing whether the observed correlational structure among variables was random or not, determining the statistical significance of each principal component, and evaluating the contribution of each variable to the significant components. Bootstrap replicates were used to estimate sampling variance around the mean statistics. S-plots derived from the orthogonal partial least squares discriminant analysis (OPLS-DA) model identified the amino acid and biogenic amine profiles responsible for the largest separation between these metabolite profiles measured before (month 0) and after the intervention (month 6)²⁸. Significant metabolites were identified in the three study groups over the six-month period using a VIP score greater than or equal to 1, P < 0.05, along with correlation less than -0.5 or greater than 0.5²⁸.

The across-group differences between amino acid and biogenic amine profiles at baseline and the changes from month 0 to month 6 were assessed using Kruskal-Wallis (for baseline profiles) and one-way analysis of variance (ANOVA) (for month 0 to month 6 changes). To identify specific pairs of groups with significant differences in the changes of Putrescine from month 0 to month 6, a series of pairwise independent sample t-tests were performed.

All statistical analyses were performed using R statistical software²⁹. PCA and OPLS-DA were conducted using FactoMineR and ropls packages, respectively. Data visualizations, including PCA trajectory plots and S-plots, were generated using ggplot2³⁰.

Associations between GBA-related amino acid and biogenic amines, gut microbiota genus abundances, and clinical parameters were measured using Spearman's rank correlation coefficient. P-values were first calculated based on the assumption that the parameters are normally distributed and uncorrelated, as implemented in the spearmanr function of the SciPy package³¹, and subsequently corrected for multiple testing with Benjamini-Hochberg procedure at 5% false discovery rate. Only significant associations were visualized in the circus plot, using pycirclize package³², and the network plot, using CytoScape³³. It should be noted that not all measurements are normally distributed (tested using SciPy's normaltest function with P-value cutoff of 0.5), but this assumption was used only to estimate the significance for Spearman's rank correlation. The heatmaps of correlation coefficients (Supplementary Figs. S2 and S4) are also provided for full transparency.

An overview of demographic data and baseline characteristics across all groups is provided in Supplementary Table S2, which has been reported previously¹⁸. No significant differences were observed in clinical data, nutrient intake, physical activity, or biochemical parameters. Additionally, baseline GBA-related plasma amino acids and biogenic amines showed no significant differences between the three groups (all P > 0.05) (Table 1).

The PCA plots illustrate the changes in amino acids and biogenic amines between the placebo, inulin, and dietary fiber advice groups over the 6-month intervention (Fig. 1). A permutation-based test was used to determine the significance of PCA model, using 1000 permutations and 1,000 bootstrap iterations through the PCAtest function. The test indicated that the first three principal components were statistically significant, accounting for 61.3% of the total variation. PC1 accounted for 33.0% of the total variation (95% CI: 31.9-34.8), PC2 explained 23.1% (95% CI: 21.3-26.2), and PC3 contributed 5.3% (95% CI: 4.8-6.2). As per the permutation test, 33 metabolites were found to significantly load PC1 and PC2, while five were found to significantly load PC3 (Supplementary Table S3). For a given month, the three study groups were clustered together, with profiles measured at month 6 exhibiting a pronounced shift from that measured at month 0. Additionally, the deviation along PC2 (indicated by error bars) at month 6 was also larger than that at month 0. The loading plots for individual amino acids and biogenic amines for the first two principal axes (PC1 and PC2) are shown in Supplementary Fig. S1.

S-plots derived from OPLS-DA were used to identify key amino acids and biogenic amines within each study group over the 6-month period. Significant changes were observed, particularly an enhancement of GBA-related amino acid and biogenic amines in the inulin group. Specifically, Tyr showed a significant change (VIP score = 1.244, correlation P-value: 9.58E-09). Similarly, putrescine (VIP score = 1.396, correlation P-value: 3.98E-11) and spermine (VIP score = 1.236, correlation P-value: 1.25E-08) also demonstrated significant changes within the inulin group during this period. No significant changes were noted in the placebo and dietary fiber advice groups (Fig. 2). The complete results for all compounds from these within-group analyses are presented in Supplementary Table S4.

Notably, inulin supplementation significantly upregulated putrescine levels over time compared to the placebo group (one-way ANOVA P < 0.01, t-test P = 0.021) (Fig. 3). Putrescine levels over the 6-month period across the three groups are presented as mean (MeanZScore), standard deviation (SD), and standard error (SE). The mean values are plotted over time with ± SE ribbons, where SE was calculated as SE = SD/(√N), indicating the deviation around the mean estimates. A pairwise t-test was used to compare putrescine levels between groups. The Bonferroni correction was applied to the P-values to correct for multiple comparisons (time and intervention). All compounds from the between-group comparisons are presented in Supplementary Table S5.

Among amino acids, Glu positively correlated with Streptococcus, Gln with Bifidobacterium, and His with Bifidobacterium, Collinsella, Blautia and Streptococcus, but negatively with Parabacteroides, Fusobacterium, and Prevotella. Trp was linked to Bifidobacterium and Blautia, but inversely to Prevotella. Tyr was positively associated with Lactobacillus but negatively associated with Alistipes. For biogenic amines, dopamine positively correlated with Bifidobacterium but negatively with Prevotella. Putrescine showed positively related to Bacteroides and Fusobacterium, and spermine to Bifidobacterium and Blautia (all P < 0.05 after Benjamini-Hochberg correction with FDR = 0.05) (Fig. 4 and Supplementary Fig. S2).

Glu positively correlated with butyrate, while His had a positive relationship with GLP-1. Trp and Tyr negatively related to screen time. His, Gln, and Trp demonstrated negative correlations with the inflammatory cytokines, IL-1β, IL-6, and TNF-α. For biogenic amines, dopamine also negatively correlated with screen time and these cytokines but had a positive association with butyrate. Spermine was positively related to butyrate but negatively related to screen time and TNF-α. (all P < 0.05 after Benjamini-Hochberg correction with FDR = 0.05) (Fig. 4 and Supplementary Fig. S3).

Following intervention, we observed different associations in the changes in metabolites and gut microbiota both across groups (Supplementary Fig. S4) and with the baseline (Supplementary Fig. S2). Within the inulin supplementation group, there were positive correlations between changes in putrescine with changes in gut microbiota such as Bacteroides and Olsenella. Changes in Trp, His, and spermine were positively correlated with changes in Chloroplast, and Peptostreptococcus, but negatively correlated with changes in Prevotella and Oscillibacter for example. However, some of these associations are either lost or reverted in the placebo or dietary advice groups.

There were consistent associations between putrescine and gut microbiota that were observed both at baseline and after treatment. For example, as high baseline putrescine was associated with high baseline Bacteroides and Fusobacterium (Fig. 4), increase in putrescine was also associated with increase in Bacteroides following inulin supplementation (Fig. 5 and Supplementary Fig. S4).

The present study is the largest and first randomized, double-blinded, placebo-controlled trial to investigate the effects of prebiotic supplementation on GBA-related amino acids and bioactive molecules in children with obesity. By demonstrating that inulin significantly enhanced key markers of gut-brain communication, including putrescine, spermine, and Tyr. These findings highlight the potential of inulin as a targeted intervention for managing pediatric obesity by improving gut-brain signaling. Furthermore, the use of advanced technique, specifically LC-MS/MS, ensured the depth and robustness of the analyses. The limitations of the study include its 6-month duration, which may not capture the long-term effects of inulin supplementation on gut-brain communication and clinical outcomes. Additionally, while the study identified changes in GBA-related compounds, detailed mechanistic pathways were not directly explored. Addressing these aspects in future research could provide a more comprehensive understanding.