Ethical statements
This study was reviewed and approved by the local ethical committee CPP Ile de France IV 2018/69, ID-RCB: 2018-A02155-50. It was conducted between November 2018 and June 2019 at the Human Nutrition Research Center of Rhône-Alpes (CRNH-RA) and carried out in accordance with the Second Declaration of Helsinki and French Jardé’s law. It was reported and registered on http://www.clinicaltrials.gov NCT03773900 (12/12/2018). All participants received and signed informed consent before the initiation of any study-related procedure.
Study participants
Sixteen subjects with CM risk profile were recruited. Inclusion criteria included age: 30–65 years old, BMI: 25–35 kg/m2, waist circumference > 80 cm for women and > 94 cm for men, daily fiber consumption < 25 g/day, exhaled H2 (max–min > 20 ppm) in response to lactulose test, stable weight and moderate physical activity, no known gastrointestinal disease, no previously bariatric surgery, no use of antibiotics or other drugs interfering with microbiota composition in the 3 months prior to the beginning of the study.
Study design
This was a dietary monocentric, randomized, double blind and cross-over study. Each subject received two 3-week dietary interventions in a random order according to a randomization list per block of permutations established with SAS software 9.4 by the statisticians. The two dietary interventions were, 1) 4.5 g/d of chitin-glucan (CG) and 2) 4.5 g of maltodextrin as control (Table S3). A 4 to 6-week washout period separated the two interventions. Before and after each intervention, fecal samples were collected for gut microbiota composition and its derived-metabolites ‘analysis. Fasting and postprandial metabolic parameters were evaluated during metabolic assessment days (Figure S3).
CG and CTL had the same galenic form and visual aspect but mainly differed by their fiber content (Table S3). Both were presented as a 3 g sachet powder to be diluted in water. Each sachet included 1.5 g of CG or 1.5 g control, thus subjects consumed 3 sachets per day (20–30 min before the 3 meals of the day) to reach the target dose of 4.5 g/d. The set of sachets to be consumed for each period was given to subjects at the beginning of each period during which, they were asked to keep the same diet and usual physical activity. A French National Authority for Health (HAS) questionnaire assessed with NUTRILOG software (version 3.10b, released in February 2017) was used to evaluate subjects’ dietary fiber intake at inclusion. To evaluate dietary intake and ensure a good compliance during the intervention periods, subjects were ordered to fill in a 3-day dietary record. Data from this 3-day dietary record were also processed using NUTRILOG software which allowed to estimate energy, macronutrient, and dietary fiber intakes. Study’s products consumption compliance was evaluated by counting full and empty sachets they brought to the center at the end of every intervention period.
During the 3 days before metabolic assessment day, subjects were ordered to collect stool samples using a specific kit. Stool samples were kept in the subject’s freezer until they brought them to the center using a furnished cooler bag. On metabolic assessment days, subjects arrived at CRNH-RA after an 8-h overnight fast following the ingestion of a standardized low dietary fiber evening meal (one serving of fish, rice, cheese, rusk or soft bread, dried biscuits and butter). Body weight and fat mass percentage were measured using standardized methodologies using a calibrated weighing scale and a Bodystat Quadscan 4000 (BQ4000; Bodystat Ltd. Douglas, UK) respectively. RMR was measured by indirect calorimetry using a QUARK calorimeter (Cosmed, Rome, Italy). Subjects were served a dietary fiber-enriched breakfast at T0 (rye bread, hazelnut carob spread, fruit juice) and a standardized challenge test meal: the FlexMeal challenge at T240. The FlexMeal is a revised version of the “PhenFlex drink”43. It consists on a standardized challenge test meal composed by a soup and a dairy dessert presenting 923 kcal, 32.5% of carbohydrates, 58.9% of lipids and 8.7% of proteins. Fasting and 4 h postprandial blood (T0, T30, T60, T90, T120, T180, T240, T255, T270, T300, T330, T360, T420) were collected using an antecubital vein catheter. Fasting and postprandial exhaled gases were also collected (T0, T60, T120, T180, T240, T300, T360, T420, T480, T540, T600).
Biochemical blood analyses
Collected blood was centrifuged immediately for 10 min at 4500 rpm. Plasma was stored at − 20 °C until the assays were conducted.
Glycemia was measured by spectrophotometry according to Architect Abbott Hexokinase method; TC, HDL-C and TG by spectrometry using Architect Module Chimie Abbott method; insulin by radio immunoassay according to RIA CisBio IBA method, NEFA by automatized spectrometry with PENTRA 400 HORIBA ABX. LDL-C was calculated using Friedwald formula and HOMA as plasma glucose (mmol/L) x plasma insulin (mUI/L) / 22.5.
Exhaled gases analyses
Exhaled gases have been collected using EASYSAMPLER Breath Test Kit (Quintron, Milwaukee, WI, USA). The concentrations of H2, CH4 and CO2 were then measured by gas chromatography using a QUINTRON BREATH TRACKER analyzer (Quintron, Milwaukee, WI, USA).
Microbiota analysis
Fecal microbiome sequencing analysis
Bacterial DNA was extracted from fecal samples using the QIAamp DNA Stool Mini Kit (QIAGEN, Hilden, Germany), following the ‘Protocol Q’ that was described by Costea and colleagues44 with a slight modification: a reduction in time of bead beating step. Cells are mechanically lysed by running the FASTPREP Instrument for 2 min at max speed (beating 1 min and resting 5 min).
Library construction and the Illumina sequencing protocol has been previously described in detail45. In brief, the V5-V6 regions of the 16S rRNA gene were targeted for PCR amplification using primers 784 F [5′-RGGATTAGATACCC-3′] and 1064 R [5′- CGACRRCCATGCANCACCT-3′]. 16S rRNA gene amplicons were sequenced by the MiSeq platform (300 bp paired-end length) at the University of Minnesota Genomics Center. All samples of this study were sequenced in the same run. Paired-end reads were merged, demultiplexed and conducted quality control implementation (length = 281 bp, mean sequence quality score ≥ 30) using QIIME2 pipeline with DADA246,47, which we refer to herein as amplicon sequences variants (ASV). The average of the sequence in all samples (n = 60) was 19,884. An even depth of 8,290 sequences per sample was used to conduct microbiome diversity. We assigned the sequences to taxonomic categories including kingdom, phylum, class, order, family and genus levels using a pre-trained Naive Bayes classifier based on Silva 132 99% OTUs database48. For significant ASV, in order to have a higher resolution of the ASV identification, online 16S rRNA databases on both NCBI blast and EzBioCloud platforms were also used.
SCFA analysis
The methodology has been described before15. In brief, the native fecal samples were homogenized and subsamples of 400–500 mg were diluted 1:4 in ultrapure water and analyzed using a capillary gas chromatograph (HP6890 Series; Hewlett Packard Corp., Paolo Alto, California, USA). Fecal dry mass was assessed by drying 300–500 mg of native sample overnight at 103 °C.
Markers of intestinal permeability
All biomarkers included were validated as markers of intestinal permeability before (https://pubmed.ncbi.nlm.nih.gov/34009040/). Zonulin and calprotectin were measured using enzyme-linked immunosorbent assay kits (K5600; K6927; Immundiagnostik AG, Bensheim, Germany) following the manufacturer’s protocol. The fecal samples were diluted to the working concentration in sample buffer using stool sample tubes (K6998SAS; Immundiagnostik AG, Bensheim, Germany). LBP (REFs: DY870-05 and DY008; Bio-Techne GmbH, Wiesbaden, Germany) was measured in plasma.
Bile acids analysis
Bile acids were analyzed using method validated method adapted from Guillemot-Legris et al.49. Briefly, lyophilized feces (5 mg) were homogenized in ice-cold distilled water prior to protein precipitation using acetone containing seven deuterated bile acids used as internal standards. Samples were next centrifuged, and the supernatant was evaporated to dryness under nitrogen steam. The resulting residue was resuspended in methanol and injected in the HPLC–MS system consisting of an LTQ-Orbitrap XL mass spectrometer (Thermo Fisher Scientific) coupled to an Accela HPLC system (Thermo Fisher Scientific). Analyte separation was performed on an Ascentis Express C-18 column (2.7 µm, 100 × 4.6 mm) (Sigma-Aldrich) using a gradient between acetonitrile and water, both containing formic acid. Mass spectrometry analysis was performed using an electrospray ionization source in the negative mode. Calibration curves were prepared in the same conditions.
LCFA analysis
To determine the LCFA profile in feces, we used 40 mg of previously lyophilized feces during 48 h (Labconco, freeze dryer 4.5). Forty microliter of C19 were added as extraction standard for homogenization with methanol:chloroform (1:2 V/V) by brief sonication in ice (Labsonic U, B. Braun). Homogenates were then filtered with Whatman filters n°1 (10 μm of porosity). Filters were rinsed with 2 ml of chloroform and 1 ml of methanol. Homogenates were purified with KCl 0.88% and KCl 0.88%: methanol (1:1 V/V). After centrifugation (1500 g, 5 min), the chloroform phase was collected in new tubes and evaporated under nitrogen flux until samples were completely dry. The esterified fatty acids were then subjected to alkaline hydrolysis (saponification) and free fatty acids were methylated and quantified by gas chromatography with flame ionization detector as previously described50.
Statistics
The primary outcome of this study was the measure of SCFA production. Due to the lack of available published data on this specific fiber, a standard power calculation to obtain the number of subjects needed for the study was not possible. The choice of the primary outcome relayed on data from two studies. The first one, carried out in an in vitro model simulating the intestine showed that a dose of 4.5 g of CG was able to increase the production of SCFA13. The second one showed that a supplementation of 3 g/day for 2 months of BG, a component of CG and a dietary fiber which structure is closely related CG, resulted in a significant increase of SCFA in the feces of 26 subjects51.
All statistical analyses and graphs were performed with R software v3.6.0 and SAS software 9.4 TS Level 1M652.
To find out whether CG had a different effect from the effect of control on clinical variables and to assess if the difference was statistically significant, we performed a linear mixed model for repeated measures, with Toeplitz or autoregressive structure (AR) as covariance structure. In order to account for variability between subjects and to adjust for any non-specific differences, subjects were included as random effects. To quantify the effect of CG compared to the effect of control, we estimated the difference in terms of change ETD = After Fiber − Before Fiber − (After Control − Before Control) to which the p value of the interaction treatment*time is associated.
For gut microbiota analysis, relative abundances performed in Qiime2 were calculated in R for each taxa and ASV. To avoid analyzing spurious sequences, ASV with an average relative abundance below 0.1% in all samples were removed. The same cut-off was applied for analysis of bacterial genera. To assess if a given microbial feature was differently impacted by the two treatments, we first computed the delta (after–before) in abundance for both treatments. Then, a Wilcoxon signed-rank test has been performed to assess if the difference between the two deltas was statistically significant; if so, we evaluated the impact of each treatment by comparing the abundances before and after its consumption with a Wilcoxon signed-rank test.
Beta-diversity indices were compared by PERMANOVA using the Adonis function in the vegan package on R Software and data were visualized using non-metric multidimensional scaling.
For subgroup analysis, we computed Δ exhaled H2 = exhaled H2 peak – baseline and defined two subgroups: “High-H2” = subjects with Δ exhaled H2 > Δ exhaled H2 median and “Low-H2” = subjects with Δ exhaled H2 < Δ exhaled H2 median. The interaction treatment*time*subgroup was tested for all variables and the linear mixed model for repeated measures was only performed when the interaction was significant (p < 0.05).
For correlation analyses, we computed the deltas (after–before) following CG supplementation for each metabolic parameter and bacterial feature which were differently impacted by the two intervention. Spearman’s correlations were performed.
All statistical analyses were performed with adjustment using false discovery rate (FDR < 0.05) for multiple tests according to the Benjamini–Hochberg procedure and none of the adjusted p values were significant. Nevertheless, non-adjusted p values < 0.05 were considered significant due to the explorative character of these approaches.
