Fecal samples were collected from three healthy donors—acknowledging that this sample size may still be considered limited for capturing interindividual variability—aged between 25 and 45 years who consumed a mixed Western diet, had no history of chronic diseases, and had no antibiotic use during the 6 months preceding the experiment. Fecal suspensions were prepared, mixed with a cryoprotectant (31), aliquoted, flash-frozen, and stored at −80 °C until needed. Effects of the test products on the gut microbiota were evaluated relative to a fiber-free control containing only the fecal microbiota of a given donor and a background nutritional medium. The collection and use of the fecal samples were performed in accordance with the protocol approved by the Ethics Committee of the University Hospital Ghent (reference number: ONZ-2022-0267; approved on 29 July 2022). Informed consent was obtained from all subjects involved in the study.

Fiber mixtures evaluated in this study contained chicory inulin, wheat dextrin, and cellulose in varying proportions: (a) Fiber mixture 1: chicory inulin 43.8%, wheat dextrin 18.8%, cellulose 37.5%; (b) Fiber mixture 2: chicory inulin 35.0%, wheat dextrin 45.0%, cellulose 20.0%; and (c) Fiber mixture 3: chicory inulin 83.3%, wheat dextrin 16.7%. The mixtures are used as fiber sources in tube feeds and oral nutritional supplements (Fresubin® formula by Fresenius Kabi), which are designed as food for special medical purposes for patients with or at risk of malnutrition. Additionally, the prebiotic tapioca dextrin was evaluated as a single fiber source. All test products were provided by Fresenius Kabi, Bad Homburg, Germany.

A 40 g/L stock solution of each test product was prepared and exposed to conditions that simulate oral, gastric, and small intestinal passage (32). Following this, the solutions were placed inside 0.5 kDa dialysis membranes, which were then sealed. The solutions were dialyzed in 3.75 g/L NaHCO₃, pH 7.0, for 24 h to remove monosaccharides and disaccharides from the predigested solutions, retaining the non-digestible fibers. Additionally, a predigestion blank medium was generated in parallel to be added to the fiber-free controls in the colonic simulations. This was achieved by running each step of the predigestion (including dialysis) in the absence of the test products.

Each reactor was first filled with 56 mL of nutritional medium (PD01; ProDigest, Gent, Belgium). Then, 7 mL (10% [v/v]) of predigested/dialyzed test product or predigestion-conditioned blank medium was added to each reactor, followed by 7 mL (10% [v/v]) of freshly prepared fecal inoculum from one of three healthy human donors. The total reactor volume was 70 mL, with a starting test product concentration of 4 g/L (assuming no absorption). The reactors were incubated at 37 °C with continuous shaking (90 rpm) in an anaerobic atmosphere for 48 h. Each condition was tested in duplicate.

At 0, 6, 24, and 48 h, changes in pH, gas pressure, SCFAs, branched chain fatty acids (BCFA), and lactate were measured; ammonium levels were determined at 0, 24, and 48 h. Changes in pH were measured with a Senseline F410 pH meter (ProSense, Oosterhout, The Netherlands), and gas production was monitored using a handheld pressure indicator (CPH6200; Wika, Echt, The Netherlands). The methods of De Weirdt et al. were used to measure acetate, propionate, and butyrate, and BCFAs (isobutyrate, isovalerate, and isocaproate) (33). Lactate levels were evaluated using an enzymatic assay kit from R-Biopharm (Darmstadt, Germany) according to the manufacturer’s instructions. Ammonium levels were assessed using the method of Tzollas et al. (34).

Total DNA was isolated as described by Duysburgh et al. (35). 16S-targeted Illumina sequencing utilized primers spanning two hypervariable regions (V3–V4) of the 16S ribosomal RNA (rRNA) gene (341F, 5′-CCTACGGGNGGCWGCAG-3′; 785R, 5′-GACTACHVGGGTATCTAAKCC-3′) (36, 37). Sequencing of 2 × 250 bp (paired sequencing) yielded 424-bp amplicons (LGC Genomics GmbH, Berlin, Germany). Amplicons of this size are taxonomically more useful than smaller fragments. Read assembly and cleanup were performed according to the MiSeq SOP (Schloss lab) (37, 38). Briefly, reads were assembled into contigs, and alignment-based quality filtering was performed (alignment to the mothur-reconstructed SILVA SEED alignment, version 138.0). Chimeras were then removed (vsearch version 2.13.3), and a naïve Bayesian classifier (39) and SILVA NR version 138_1 were used to assign taxonomy. Contigs were clustered into Operational Taxonomic Units (OTUs) at 97% sequence similarity using mothur (version 1.44.3) (OTU-based clustering was used instead of amplicon sequence variant (ASV)-based approaches (e.g., deficiency of adenosine deaminase 2 [DADA2]) as part of prior validated workflows). Sequences that were classified as Archaea, Eukaryota, mitochondria, or chloroplasts, or those that could not be classified, were removed. The representative sequence was defined as the most abundant sequence within an OTU. Reads with a maximum abundance ≤5 across samples were removed as they were considered to be bacteria with low biological impact or artefacts.

The abundances of the two dominant colonic bacterial phyla, Firmicutes and Bacteroidetes, and two specific groups of interest (due to their links to health-promoting effects), Bifidobacterium and Akkermansia muciniphila, were monitored by quantitative PCR (qPCR). qPCR for Firmicutes and Bacteroidetes was performed as previously described by Guo et al. (40). The method reported by Collado et al. was used to quantify A. muciniphila (41), and the method reported by Rinttilä et al. was used to quantify Bifidobacterium spp. (42).

To determine the total number of bacterial cells, samples were analyzed by flow cytometry using a BD Accuri C6 Plus Flow Cytometer (BD Biosciences, Franklin Lakes, NJ, USA) set to the high flow rate. Bacterial cells were separated from medium debris and signal noise by applying a threshold level of 700 on the SYTO channel. Parent and daughter gates were set to determine the populations. Using these cell counts, metagenomics data on relative abundance were converted to absolute abundance by multiplying each sample’s relative abundance by the total cell count, as previously reported by Vandeputte et al. (43).

Co-culture experiments were performed using Caco-2 (HTB-37; American Type Culture Collection) and phorbol-12-myristate-13-acetate (PMA) differentiated THP1-Blue™ cells (InvivoGen, San Diego, CA, USA) as previously described (44, 45). First, Caco-2 cells were seeded in 24-well semi-permeable inserts and cultured for 14 days. THP1-Blue™ cells were seeded in 24-well plates and treated with PMA to induce differentiation. Subsequently, Caco-2-bearing inserts were positioned onto the PMA-differentiated THP1-Blue™ cells for further experimentation, following previously established protocols (44). Briefly, sterile filtered (0.22 μm) colonic suspensions were added to the co-cultures. After 24 h of incubation, transepithelial electrical resistance (TEER) was measured in Caco-2 cells. Next, cells were stimulated with 500 ng/mL ultrapure lipopolysaccharide (LPS; Escherichia coli K12) (InvivoGen, San Diego, CA, USA) for 6 h. Supernatants were then collected and processed to assess nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) activity and levels of select chemokines and cytokines. NF-κB activity was determined using the QUANTI-Blue reagent (InvivoGen, San Diego, CA, USA) to measure SEAP levels in basolateral supernatants according to the manufacturer’s instructions. Levels of human interleukin-1β (IL-1β), IL-6, IL-8, IL-10, tumor necrosis factor (TNF)-α, and C-X-C motif chemokine ligand 10 (CXCL10) in the co-culture basolateral supernatants were measured using Luminex® multiplex (ProcartaPlex immunoassays, ThermoFisher Scientific) according to the manufacturer’s instructions. Samples from the colonic incubations were obtained as biological duplicates and used in the cell co-culture assay as technical duplicates, resulting in quadruplicate values for each test condition.

Total SCFA level was calculated by adding together the levels of acetate, propionate, and butyrate for each condition at each timepoint. For analysis of microbial metabolic activity, paired two-sided t-tests were used to determine significant differences between the test product and the fiber-free control.

Per donor and for each bacterial group, fold changes (FC; ratio of abundance at 6 or 24 h [T1] vs. abundance at start of incubation [T0]) were calculated for test products and fiber-free control. Treatments with fold changes that differed from fiber-free control across the donors were identified. A bacterial enrichment was considered significant if it exceeded the enrichment observed in the fiber-free control.

α-Diversity was evaluated using four methods: (a) observed taxa (measure for species richness), (b) Chao1 (measure for species richness), (c) Shannon (measure for species richness and evenness), and (d) Simpson (measure for species richness and evenness, giving more weight to common or dominant species). Paired two-sided t-tests were used to determine significant differences between test products and the fiber-free control. β-Diversity was assessed using Discriminant Analysis of Principal Components (DAPC), which joins two methods to assess effects on population structure. To accomplish this, sequence data were transformed using principal component analysis, and clusters were then identified using discriminant analysis (46), aiming to maximize among-group variation and minimize within-group variation.

Linear discriminant analysis effect size (LEfSe) (47) and treeclimbR (48) analyses were performed to identify the taxa most likely to explain differences between treatments. All features shown in the LEfSe plots met a criterion of p ≤ 0.05 for the Kruskal–Wallis and Wilcoxon tests; the 50 most significant features were included in the plot. No restrictions were put forward with respect to minimal linear discriminant analysis (LDA) scores, but, in general, LDA scores ≥ 2.0 were considered biologically relevant. The higher the LDA score, the higher the difference in abundance between the two test conditions. The results of the treeclimbR analysis are shown using volcano plots. The cut-off for statistical significance was set at p < 0.05 (or -log₁₀[0.05] = 1.3 on the y-axis). Bacterial enrichments with a –log (p-value) > 1.3 were considered statistically significant. Taxa were classified into four different categories: (a) not significant and not biologically relevant (−2 < log₂FC < +2, and −log₁₀[p-value] < 1.3), (b) biologically relevant, but not statistically significant (log₂FC < −2 or log₂FC > +2, and −log₁₀[p-value] < 1.3), (c) statistically significant, but not biologically relevant (−2 < log₂FC < +2, and −log₁₀[p-value] > 1.3), and (d) biologically and statistically significant (log₂FC < −2 or log₂FC > +2, and −log₁₀[p-value] > 1.3).

To assess differences in TEER and immune markers between the test products and fiber-free control for individual donors, a two-way analysis of variance (ANOVA) with Dunnett’s multiple comparisons test against the fiber-free control was used. Cytokine data are shown normalized to the LPS control. To assess differences in TEER and immune markers between the test products and the fiber-free control, paired two-tailed t-tests were performed on the average of the three donors, using the average of the technical replicates for each donor as input values.

A p-value of < 0.05 was considered statistically significant. All statistical analyses were performed using GraphPad Prism version 9.5.0 for Windows (GraphPad Software, San Diego, CA, USA).

General markers of microbial fermentation, pH decrease, and gas production were affected by the fiber mixtures and tapioca dextrin. pH decreased significantly starting from 6 h following test product supplementation, and gas production significantly increased relative to the fiber-free control in all test conditions (Figures 1A,B).

Measures of saccharolytic fermentation (i.e., SCFA levels) demonstrated activity with the test products, as significant increases in total SCFA levels were observed for all test products when compared to the fiber-free control over the complete incubation period (FC = 1.7 with p = 0.0011, FC = 1.9 with p = 0.0109, FC = 2.2 with p = 0.0236, and FC = 1.8 with p = 0.0175 for fiber mixtures 1, 2, 3, and tapioca dextrin, respectively). Levels of acetate and propionate were significantly increased relative to the fiber-free control with fiber mixture 1 or fiber mixture 2, and numerically increased with fiber mixture 3, with increases observed as early as 6 h after supplementation (Figures 2A,B). Tapioca dextrin induced a numeric increase in acetate and a significant increase in propionate production relative to the fiber-free control. Levels of butyrate were numerically increased with all test products relative to the fiber-free control in the time interval of 6–24 h after supplementation (Figure 2C). A numeric increase in lactate production relative to the fiber-free control was also observed for all test products during the 0–6 h incubation period, which decreased at the later timepoints (Figure 2D). Measures of proteolytic fermentation, BCFA, and ammonium were numerically and significantly decreased, respectively, with all test products relative to the fiber-free control (Figures 2E,F). Data for individual donors are shown in Supplementary Figure S1. Donor to donor variations in SCFA production were observed; Donor A produced lower levels of butyrate relative to the other two donors.

Bacterial species richness (measured using the Observed and Chao1 indices) was decreased with the fiber mixtures and increased with tapioca dextrin relative to the fiber-free control (Figure 3A), though not reaching statistical significance. Bacterial species evenness (measured by the Shannon and Simpson indices) was reduced compared with the fiber-free control with all test products, reaching significance for fiber mixture 3 (Shannon: p = 0.0426) and tapioca dextrin (Shannon indices: p = 0.0295; Simpson indices: p = 0.0270). Furthermore, treatment with the fiber mixtures and tapioca dextrin significantly impacted community composition (β-diversity), as indicated by their segregation from the fiber-free control in Figure 3B. Fiber mixtures 1, 2, and 3 were found to share some similarities in community composition that were slightly distinct from tapioca dextrin, even though differences from the fiber-free control were more pronounced.

TreeclimbR analysis demonstrated a biologically and statistically significant increase in the abundance of the genera Parabacteroides (Bacteroidetes), Fusicatenibacter (Firmicutes), and unclassified Bacteroidales (Bacteroidetes) with all three fiber mixtures relative to the fiber-free control (Supplementary Figures S2A–C). The Parabacteroides increase was confirmed in LEfSe analysis for all three fiber mixtures, and the Fusicatenibacter was confirmed for fiber mixes 1 and 2 (Figures 4A–C). Biologically significant increases that were common to all three fiber mixes included Holdemanella (Firmicutes) and Streptococcus (Firmicutes). In addition to the common changes, fiber mixture 1 induced a biologically significant increase in Blautia (Firmicutes) (Supplementary Figure S2A). Fiber mixture 2 also stimulated a biologically and statistically significant increase in Marvinbryantia (Firmicutes) and a biologically significant increase in Blautia, Erysipelotrichaceae (family) (Firmicutes), and Prevotellaceae (family) (Bacteroidetes) (Figures 4B; Supplementary Figure S2B). Biologically significant increases unique to fiber mixture 3 were Dialister (Firmicutes), Collinsella (Actinobacteria), and Bifidobacterium (Actinobacteria) (Supplementary Figure S2C). Tapioca dextrin induced biologically and statistically significant increases vs. the fiber-free control for Parabacteroides and Fusicatenibacter and biologically significant increases in Bacteroidales, Bifidobacterium, Streptococcus, and Prevotellaceae (family); the increase in Parabacteroides was also observed with LEfSe analysis (Figures 4D; Supplementary Figure S2D).

At both 6 and 24 h, the FC in Bifidobacterium abundance from baseline (0 h) was significantly greater with all three fiber mixtures and tapioca dextrin relative to the fiber-free control across donors (Figures 5A; Supplementary Figure S3A). When looking at individual donors, the FC in Bifidobacterium abundance with all test products relative to the fiber-free control was high for Donors A and B, while the effect was less pronounced for Donor C at both timepoints (Figures 5B; Supplementary Figure S3B). There were no significant differences for any of the test products relative to the fiber-free control for FC in Akkermansia abundance from baseline to 6 h or 24 h (Figures 5C; Supplementary Figure S3C). There was a high level of variability in FC for Akkermansia abundance among donors, with Donor A having the greatest effect, followed by Donor C, and very little change in any condition for Donor B (Figures 5D; Supplementary Figure S3D). However, improved Akkermansia growth was observed at 24 h for all test products for Donors A and C. Across donors, FC was significantly higher from baseline in Firmicutes and Bacteroidetes with fiber mixtures 1 and 2 and with tapioca dextrin compared with the fiber-free control at 6 h and with all test products at 24 h (Supplementary Figures S4A–C). When examining individual donors, interindividual differences in response were minimal (Supplementary Figures S4B–D).

For all treatments, donor-dependent effects on the intestinal epithelial barrier integrity were observed. Treatment with all test product fermentations significantly increased TEER in Donor C compared to their respective fiber-free control. Therefore, TEER values indicated that fiber mixtures and tapioca dextrin fermentations provided significant protection against inflammation-induced barrier disruption relative to fiber-free fermentations in Donor C (Figure 6A). No significant protection was observed with any of the fiber mixtures for Donors A and B; there was a significant decrease in protection with tapioca dextrin for Donor B. Following stimulation with LPS, anti-inflammatory IL-10 was greatly induced by test product fermentations, which reached significance for Donors A and C with fiber mixture 1, across donors with fiber mixtures 2 and 3, and for Donor C with tapioca dextrin (Figure 6B). NF-κB activity was increased with fiber mixture 1 (Donor B and C, across donors), fiber mixture 2 (Donor A and C, across donors), and fiber mixture 3 (across donors) but not with tapioca dextrin (Supplementary Figure S5A). The test product fermentations had little effect on IL-6, IL-1β, TNF-α, CXCL10, or IL-8 levels (Supplementary Figures S5B–F).