Assessment of the Isolated and Combined Impact of β-Glucan and Lacticaseibacillus rhamnosus on Cystic Fibrosis Gut Microbiota Using a SHIME® System

The selection of the type of dietary fibre with prebiotic potential and the probiotic strain was based on findings from a previous study on children with CF, conducted by our research group [8]. A non-starch polysaccharide, β-glucan (Oat; Medium Viscosity) (Main Chain Glycosidic Linkage: β-1,4 and β-1,3) obtained from Megazyme by Neogen^® (SKU No. P-BGOM) (Michigan, USA), and a probiotic strain from the Lactobacillaceae family, Lacticaseibacillus rhamnosus GG (ATCC 53103TM), were selected. The probiotic was cultured in MRS broth (Scharlab, Barcelona, Spain), and the final concentration was adjusted to 10¹⁰ UFC/mL, using sterile peptone water to preserve it until use (Scharlab, Barcelona, Spain). With these two items, three supplementation treatments were designed as follows: the prebiotic (250 mg/day beta-glucan); probiotic (1.5 mL/day L. rhamnosus (concentration of 10⁸ UFC/mL at the colonic compartment)); and synbiotic (250 mg/day beta-glucan and 1.5 mL/day L. rhamnosus) (Table 1). These doses were selected based on our previous studies [17,18]. The effects of the three treatments were assessed by simulating supplementation with a dynamic colonic fermentation model on the basal faecal microbiota of a child with cystic fibrosis.

A stool sample was collected to obtain the faecal inoculum from one donor (n = 1) to simulate colonic fermentation. The sample was obtained from a paediatric patient with cystic fibrosis recruited in the Valencian Community (Spain). The procedure was approved by the Ethics Committee of Universitat Politècnica de València (P03_25-07-2022). The subject was a 12-year-old male with the F508del/G85V genotype, who was not on CFTR modulators therapy and had pancreatic insufficiency (faecal elastase < 15 µg) supplemented with pancreatic enzyme replacement therapy (PERT). The subject presented adequate fat (3.6 g/24 h faecal sample) and protein (0.7 nitrogen g/24 h faecal sample) absorption and anthropometric nutritional status indicators, suggesting adequate weight and growth. The inclusion criterion was a confirmed diagnosis of cystic fibrosis while the exclusion criterion was having taken antibiotics in the last three months. The child’s legal guardians signed the informed consent, and the study was conducted under the regulations of the Declaration of Helsinki.

The decision to work with one faecal donor was based on both scientific and practical considerations. First, our primary objective was to conduct an in-depth, proof-of-concept study using the SHIME^® platform to investigate how the cystic fibrosis gut microbiota responds under controlled conditions. Employing a single, well-characterised donor allowed us to minimise inter-individual variability, which is particularly high in CF patients due to differences in genetics, clinical history, and treatment regimens. By controlling this variability, we were able to focus on the mechanistic effects of the intervention without the confounding influence of donor heterogeneity.

The dynamic gastrointestinal digestion and colonic simulation system SHIME^® was employed in this study. The system consisted of six double-jacketed bioreactors, simulating the stomach and small intestine (n = 3) and colon (n = 3). The gastric and intestinal conditions were simulated adjusting pancreatin source (pancreatic enzyme supplements (Kreon^® 10000 LU) (Viatris, Canonsburg, PA, USA), intestinal pH (pH = 6), and bile salts concentration (1 mM) to approach the most to the altered conditions found in CF [19,20]. A programmable pump system regulated the flow of digestive contents and the addition of 0.1 M NaOH or 0.1 M HCl to maintain the pH (5.6–5.9). To simulate the in vivo colonic environment, the three colonic vessels were inoculated with faecal material from the CF donor. A nutritional medium (PDNM002B, Prodigest^®) (Ghent, Belgium) was administered in three daily feeding cycles to support the microbial viability.

The experimental design consisted of four consecutive stages. The first was a stabilisation stage lasting 18 days, aimed at allowing the colonic microbiota to adapt to the in vitro environment. This was followed by a control stage of 15 days, during which no intervention was applied, to confirm the microbial community’s stability. The third was a treatment stage of 15 days, during which the selected formulations—prebiotic, probiotic, and synbiotic—were administered to evaluate their effects over time. Finally, a post-treatment stage of 10 days was carried out, during which no supplementation was provided, to observe whether any effect persisted after the intervention.

The digestion process was controlled and parameterised through dedicated software, which regulated volumetric flow rates, the timing of culture medium and pancreatic juice addition, the anaerobic environment, pH, and agitation conditions in the bioreactors. Aliquots were collected from each of the three colonic compartments at specific time points: at the end of the control stage (day 15), along the treatment stage (days 2 [T2], 5 [T5], 10 [T10], and 15 [T15], and mid-term and at the end of the post-treatment stage (days 5 [PT5] and 10 [PT10], respectively) (Figure 1). All samples were immediately stored at −80 °C for subsequent analytical determinations.

To analyse faecal microbiota composition, DNA was extracted from faecal samples using the Stool DNA Isolation Kit (Norgen Biotek Corp^® in Thorold, ON, Canada), following the manufacturer’s instructions. The DNA concentration extracted was determined using a Qubit fluorometer (Invitrogen Co., in Carlsbad, CA, USA). Microbiota profiling was performed by amplification of the V3-V4 regions of the 16S SSU rRNA gene using specific primers and Illumina technology. Primer selection followed the guidelines provided [21], and the sequences were in accordance with standard IUPAC nucleotide nomenclature. Primers 515F–806R [22,23] were used following the Illumina amplification libraries protocol. DNA amplicon libraries were generated using a limited cycle PCR (KAPA HiFi HotStart ReadyMix, KK2602) (Roche Diagnostics, Basel, Switzerland), then lllumina sequencing adaptors and dual-index barcodes from Nextera XT index kit v2, FC-131-2001 (Illumina, San Diego, CA, USA) are added to the amplicon in a second PCR. Libraries are then normalised and pooled prior to sequencing. The sample containing indexed amplicons was loaded onto the Nextseq2000 P1 reagent cartridge (reference 20100981, Illumina) and onto the instrument, along with the flow cell. Automated cluster generation and paired-end sequencing with dual indexes reads was performed (2×300bp run) on a MiSeq-Illumina platform at the FISABIO sequencing service in Valencia, Spain. FISABIO sequencing service conducted filtering and quality assessment of raw sequences using the fastp programme [24], which involved removing low-quality nucleotides from the 3′ end (within 10-nucleotide windows with an average quality score under 20) and discarding sequences shorter than 50 base pairs in length. Illumina forward and reverse sequences were joined using the FLASH programme v1 [25]. After that, joined data were processed using the DADA2 package (1.32.0) on R software (R version 4.4.1, released on 14 June 2024). Reads containing undetermined nucleotides and those matching the phiX genome were removed. A denoising algorithm was employed to infer exact ASVs, and chimaeras were removed with default parameters. Taxonomy was assigned to ASVs up to the species level using the SILVA database species train set file (version 138.1).

Short-chain fatty acid (SCFA) analyses were performed following the protocol described by [26], a SUPELCOWAX™ 10 GC capillary column (30 m × 0.25 mm × 0.25 μm) (Merck, Rahway, NJ, USA) on an Agilent GC7890B-5977B GC-FID gas chromatograph (Santa Clara, CA, USA) equipped with flame ionisation detector FID (Agilent Techonologies, Santa Clara, CA, USA). The oven temperature was initially set at 90 °C for 1 min, then increased to 190 °C at a rate of 5 °C/min, and finally held at 250 °C for 30 min. The column flow rate was 1 mL/min, with helium as the carrier gas. A liquid–liquid extraction was performed as follows: a 2 mL aliquot of samples was transferred into a 15 mL Falcon tube, followed by the addition of 0.5 mL of 9.2 M H₂SO₄, a small amount of NaCl, 0.4 mL of internal standard solution (52.9 mM 2-Methylhexanoic acid), and 2 mL of diethylether. The mixture was vortexed for 1 min and centrifuged for 3 min at 3000 rpm, and 550 µL of the organic phase was transferred to a chromatography vial for further analysis. The following SCFAs were quantified: linear (acetic acid (AA), butyric acid (BA), propionic acid (PA), valeric acid (VA)) and branched (BCFAs) (isovaleric acid (IVA), isobutyric acid (IBA)). Results were expressed in terms of millimolar (mM) concentration.

Ammonia and lactate concentration were determined using commercial kits: the Enzytec™ Liquid Ammonia kit (R-BioPharm, Darmstadt, Germany) and the Lactate Assay kit (Sigma-Aldrich, St. Louis, MO, USA), respectively. Both analyses were conducted according to the manufacturers’ instructions. Lactate results were expressed in micromoles per litre (μM) and ammonia results in millimoles per litre (mM).

The microbiota composition and metabolite production were measured in triplicate and reported as mean and standard deviation. Statistical analysis was performed with Statgraphics Centurion XVIII^® software (version 19.6.03; Warrenton, VA, USA). A two-way ANOVA followed by Dunnett’s post hoc test was applied to assess statistically significant differences (p < 0.05) in metabolic activity and relative microbial abundance at the phylum and genus levels across the different experimental stages: control vs. treatment (after 2, 5, 10 and 15 days of supplementation) and post-treatment (after 5 and 10 days). Relative abundance at genus taxonomic level data were transformed and expressed as increases and decreases with respect to the control values in each treatment. Two-way ANOVA analyses were also performed on a selection of key variables to compare the overall effect of the synbiotic supplementation treatment against the prebiotic and the probiotic. All graphs were generated using GraphPad Prism^® software (version 8.4.3, Boston, MA, USA). The alpha diversity (Shannon and Chao indexes) was obtained using R software. The beta diversity (Bray–Curtis scale) was obtained using MicrobiomeAnalyst (web platform; accessed September 2025). Differentially abundant genera were identified with LEfSe in MicrobiomeAnalyst (web platform; accessed November 2025). Prior to biomarker discovery, the feature table was kept at genus level and expressed as relative abundances (TSS-scaled percentages). To avoid artefacts from “count” filters on proportional data, minimum count was set to 0; we applied a prevalence filter of 10% (features present in ≥10% of samples) and excluded very rare features by imposing a minimum relative abundance of 0.1%. No variance filter was applied for LEfSe. For the SHIME design, the class variable was Treatment (prebiotic, probiotic, synbiotic); Bioreactor was specified in the Subject (paired) field to account for within-reactor dependency. When stage-specific effects were assessed, analyses were repeated within each Stage (Control, Treatment, Post) or Stage was used as Subclass. LEfSe was run with Kruskal–Wallis α = 0.05, pairwise Wilcoxon α = 0.05, all-against-all strategy, and an LDA effect size threshold of log10 (LDA) ≥ 2.5. Reported biomarkers therefore reflect features that are statistically significant across groups, consistent within groups, and exhibit a large effect size.

Changes in microbiota composition accounted by the three different treatments (supplementation with prebiotic, probiotic, or synbiotic) were assessed at different simulation time points and compared to the baseline values corresponding to the end of the control period and just before supplementation. Of note, the baseline microbiota, despite coming from the same faecal inoculum, resulted in varying proportions in the three different colon bioreactors (i.e., one per treatment) at the end of the control period.

In terms of microbial alpha diversity, only the supplementation with the prebiotic (β-glucan) achieved a significant increase as referred to the situation in the control period (Figure 2a). Concretely, supplying the prebiotic increased the evenness (how equal the community is numerically), with statistically different values (p < 0.05) of Shannon index after 10 days of treatment and (T10) (4.31) and at 5 days of post-treatment stage (4.36), compared to the control (4.21). In terms of richness (number of different phylotypes present in a community), the supplementation with the prebiotic led to no significant differences over the experiment time points. In terms of beta diversity (Figure 2b), the microbial communities of the different treatments showed statistically significant differences (F-value: 48.019; R-squared: 0.61548; p-value: 0.001), with prebiotic supplementation reflecting the greatest change in diversity, while treatment with the synbiotic generated a more consistent and prolonged effect over time. The variations in the microbiota are detailed in greater depth below, comparing relative abundances at the phylum and genus levels.

Although supplementation with the probiotic and the synbiotic did not significantly improve microbiota diversity, all three treatments induced significant changes in the relative abundance of phyla, as compared to the composition in the control period (Figure 3). At the control stage, the three main phyla were found in the following ranges (three different bioreactors): Bacillota (61.20–70.15%), Bacteroidota (15.69–22.20%), and Proteobacteria (7.67–18.21%), while Actinobacteriota was <1% and Verrucomicrobiota was not detected. As expected, the proportion among the three most abundant phyla was altered compared to healthy children. Previous studies have defined those children with CF present with different physiological alterations and specific treatments leading to CFRGD, which is characterised by an unbalanced ratio between Bacillota/Bacteroidota and also increased Proteobacteria [5,27,28].

Of note, a balanced Bacillota/Bacteroidota ratio has been associated as a marker of gut health [29,30], while keeping Proteobacteria abundance low is also considered as a target of supplementation and dietary strategies to improve gut microbiota [31]. As for the experimental results, the prebiotic and the probiotic supplementation treatments were successful in significantly reducing the abundance of Bacillota while increasing Bacteroidota abundances, helping to balance the ratio (Table S1). Concretely, prebiotic treatment allowed the reduction of Bacillota from 68.6% at the control stage to 64.1% at the end of the treatment (T15) (FDR p-value = 0.0043) and to 57.9% at the end of post-treatment (PT10) (FDR p-value < 0.0001) (i.e., −10%). Similarly, the probiotic treatment also reduced Bacillota abundance significantly, from initial 72.9% to 62.0% at the end of the treatment period (FDR p-value < 0.0001) and to 65.9% at the end of post-treatment (PT10) (FDR p-value < 0.0001). In parallel, both treatments enabled the growth of Bacteroidota, from 4.1% (control stage) to 19.4% (PT10) with the prebiotic and from 11.8% (C) to 18.3% (PT10) with the probiotic. However, the synbiotic supplementation treatment showed an inverse effect, leading to a significant statistical increase in Bacillota and a reduction in Bacteroidota, which was not statistically significant. As for the changes in Proteobacteria, fluctuations in relative abundance were registered over the experimental time points with all three treatments. Considering the overall change, the prebiotic and synbiotic treatments accounted for a discrete though significant decrease (i.e., ≈4%) at the end of post-treatment (PT10) (FDR p-value < 0.0001).

Going into the details of the genus taxonomic level (Figure S1), a more specific interpretation of the results was enabled. While the bioreactor-specific environment led to slight differences in the microbiota composition found at phylum level, in case of the genera, wider ranges were found, and at the end of the control stage the most abundant genera were Faecalibacterium (17.20–46.23%), Bacteroides (4.15–23.58%), Klebsiella (10.09–18.64%), Lachnospira (0.12–19.13%), and Clostridium sensu stricto 1 (4.72–6.51%). In fact, statistically significant differences were found between the three bioreactors at this baseline period (Table S2).

Changes in genera relative abundance accounted by each supplementation treatment were expressed as increases and decreases to each corresponding control value (Figure 4). More specifically, a significant increase in the Bacteroides genus was observed with all treatments, with the prebiotic and probiotic contributing the most to the increase (p < 0.0001) at the end of treatment (+5.09%) and (+6.56%), respectively. In contrast, in the post-treatment period the greatest increase (p < 0.0001) (+15.36%) was observed with the prebiotic, and a slight but significant decrease (p < 0.0001) (−1.63%) with the synbiotic. Bacteroides predominantly produce propionate in the colon and have anti-inflammatory properties for the host [32].

As for Clostridium sensu stricto 1, there was a significant increase (p < 0.0001) in relative abundance after the probiotic (+6.74%) and synbiotic (+9.43%). Of note, Clostridium ss 1 comprises a large group of commensals but also pathogenic Clostridium species, which in fact has been identified as one of the unusually increased genera in the gut microbiota of cystic fibrosis children [33]. A possible reason for its increased presence relates to the lung–gut axis, as some species of Clostridium are among the most resistant pathogens in the respiratory system in this population [2]. Therefore, the increase in Clostridium ss 1 could be considered as a negative finding.

In contrast, Faecalibacterium was significantly reduced during treatment with the probiotic supplement (p < 0.0001) (−12.11%) and after discontinuation (p < 0.0001) (−10.18%), unlike the synbiotic treatment, in which the relative abundance increased significantly at T15 (p < 0.0001) (+5.47%) and PT10 (p < 0.0001) (+9.58%). In the context of cystic fibrosis, it is common to observe a reduced abundance of Faecalibacterium [5,27,34]. This genus is a commensal taxon that is considered a marker of intestinal health [35] and is associated with the biosynthesis of SCFAs [28], especially butyric acid [36]. The species F. prausnitzii is known for its ability to suppress pro-inflammatory pathways by inhibiting the NF-κB factor and stimulating anti-inflammatory cytokines such as IL-10 and TGF-β [37]. At the same time, butyrate activates G protein-coupled receptors (GPR109A) and increases the expression of tight junction proteins, improving the integrity of the epithelial barrier and reducing mucosal inflammation [38]. Therefore, its increase with synbiotic treatment, as opposed to the effect observed with prebiotic or probiotic treatments, is a remarkable result. So, its increase with the synbiotic treatment, as opposed to the effect with the prebiotic or probiotic treatments, is a noted result.

Another relevant finding was the decrease in Veillonella, both during treatment and post-treatment with the probiotic (p < 0.0001) (decreases of 5.04 and 5.76%) and the synbiotic (p < 0.0001) (decreases of 7.50 and 8.75%). Veillonella is a characteristic genus in cystic fibrosis, which dominates the pulmonary microbiota in this population [39,40]. Its commonly increased presence in the colon is linked again with the lung–gut axis, and the mechanisms involved in sputum swallowing [41,42]. The prevalence of some species of Veillonella has been associated with increased virulence of other pathogens [43]. However, its role in cystic fibrosis disease is not yet fully understood.

On the other hand, the prebiotic treatment drove the increase (p < 0.0001) of Agathobacter during the treatment period with an increase of 5.88%, the highest among all the genera. Likewise, the probiotic and the synbiotic induced an increase in this genus in the two stages, but the differences were not statistically significant. Some species within Agathobacter are considered as anti-inflammatory microorganisms and are associated with the maintenance of intestinal health [44].

In the case of Klebsiella, the prebiotic showed a significant reduction (p < 0.0001) only at the post-treatment stage (−4.57%), while the probiotic showed an increase (p < 0.0001) (+3.32%) in the treatment. In turn, the synbiotic showed a significant decrease at all stages, in the treatment (p < 0.0001) (−2.62%) and post-treatment (p < 0.0001) (−5.33%). Klebsiella spp., especially the pathogenic species K. pneumoniae, and Megasphaera spp., are opportunistic pathogens associated with lung infections—though with low evidence, and thus its clinical impact on the disease of cystic fibrosis is still limited [45,46]—and are also found to be specific to the microbiota of children with cystic fibrosis [17]. In this study, the genus Megasphaera was significantly reduced (p < 0.0001) with the synbiotic along treatment (−1.88%) and post-treatment stages (−1.83%), while the probiotic only achieved a significant reduction (p < 0.0001) in the post-treatment stage (−2.20%). A previous SHIME^®-based study on the effect of L. plantarum supplementation also showed a reduction in Klebsiella and Megasphaera [17]. The opposite occurred with prebiotic supplementation, where a significant increase was evident 10 days after cessation of treatment (p = 0.0161) (+2.06%), possibly related to the capability of some Megasphaera species to ferment soluble fibres (FOS) [47].

Another result that deserves attention is the reduction of the genus Stenotrophomonas, considered as a biomarker of respiratory infections in cystic fibrosis [48], especially the pathogenic species S. maltophilia, which has been associated with more recurrent pulmonary exacerbations [49], and its presence in the gut could be related to the gut–lung axis [30]. In this study, the prebiotic β-glucan was able to induce a marked decrease in Stenotrophomonas abundance at the treatment stage (p = 0.0002) and post-treatment (p = 0.0003) with a reduction of 2.88% in both cases. While reductions were also observed with the synbiotic, the change was only statistically significant (p = 0.0346) at 15 days of treatment, with a decrease of 0.99%.

Moving onto the metabolic activity (Figure 5), the following markers were assessed. Lactate and linear-chain SCFAs (acetic, butyric, and propionic acids) as potentially beneficial metabolites, and ammonia and branched-chain SCFAs (isobutyric and isovaleric acids) as potentially harmful metabolites [50]. The concentrations at the different experimental time points were determined and compared to the values of the control stage (baseline values) for all three supplementation treatments (prebiotic, probiotic, and synbiotic).

Focusing first on lactate, the concentration (Figure 5a) was significantly reduced (p < 0.05) with L. rhamnosus at 15 days of treatment (24.33 µM) and after 10 days of stopping supplementation (24.15 µM) compared to the control stage (32.07 µM), while no significant changes were observed with the other treatments. This result differs from previous studies, where supplementation with the probiotic L. rhamnosus significantly increased lactate [8,17]. However, this result is not contradictory as lactate is an intermediate metabolite in the degradation of non-digestible carbohydrates, which in turn is used as substrate by some bacteria to produce SCFAs [51].

Ammonia production (Figure 5b) decreased significantly (p < 0.05) with the prebiotic and the synbiotic at the T10 stage and at the post-treatment stage all three treatments continued to reflect a decrease in ammonia concentration. Despite this reduction being statistically significant, it may not be biologically relevant, as other authors have suggested that reductions > 20–30% could be considered as beneficial [52]. The slight reduction could be expected due to the absence of proteins in the culture medium paralleled with the presence of fermentable carbohydrates in the treatments with the prebiotic and synbiotic (beta-glucan) that could have led to the preference for this type of substrate [53].

Similarly to production of lc-SCFAs (Figure 5c), the concentration of acetic acid was only affected by the probiotic treatment, which caused a significant decrease in this metabolite in the post-treatment stage (3.15 mM to 2.71 mM). In contrast, butyric acid production changed significantly with the prebiotic and synbiotic, which significantly increased the control values from 5.35 to 6.24 mM to 6.61–9.62 mM at 15 days of treatment and the increase was maintained to 6.48–7.86 mM at 10 days post-treatment. This finding suggests that the synbiotic combination in this CF context could stimulate cross-fermentation pathways, certain colon bacteria degrading β-glucan into oligosaccharides and monosaccharides, which are then fermented mainly into lactate and/or acetate [54]. In turn, supplementation with the probiotic L. rhamnosus contributes to increased lactate production [55]. These metabolites (lactate/acetate) can act as a substrate that other bacterial strains can use for butyrate production via the butyryl-CoA pathway [56]. Regarding propionic acid concentration, a slight but significant decrease was observed with the prebiotic at both stages: treatment (T15) (3.76 mM) and post-treatment (PT10) (3.73 mM) compared to the control value (4.10 mM). This metabolic shift in all the treatments is considered as a relevant finding of the study, as lc-SCFAs serve several metabolic functions associated with colonic health [57].

Finally, bc-SCFA production was influenced by two treatments: the prebiotic significantly increased the concentration of the control from 2.09 to 2.43 mM 10 days after cessation of treatment and the synbiotic showed significant increases at the T15 (2.60 mM) and PT10 (2.52 mM) versus C (2.30 mM) stage. bc-SCFAs result from the degradation of amino acids in the colon and their physiological effects on health are negative, although more evidence is needed to understand their role in cystic fibrosis [58].

Overall, although decreases in certain lc-SCFAs and significant increases in bc-SCFAs were observed in our study, the levels achieved may not be biologically relevant.

In order to provide a practical scope to this study, key study outcomes were selected and compared between the supplementation treatments: synbiotic vs. prebiotic and synbiotic vs. probiotic, to have an objective criterium to answer to the question posed in the introduction (Table 2). According to the FDR-corrected p-values resulting from the two-way ANOVA analysis, the prebiotic treatment seemed the most suitable for reducing Bacillota and increasing Bacteroidota. However, it was the synbiotic treatment that exhibited the most positive outcomes, including the highest reduction in Proteobacteria, the highest increase in lc-SCFAs, a not so high increase in bc-SCFAs, an increase in Faecalibacterium, and the highest decrease in the CFRGD characteristic genera Veillonella and also (along with the prebiotic treatment) in Megasphaera. In contrast, the probiotic treatment seemed to be the least beneficial to improve CFRGD, as it accounted for an increase in Proteobacteria, a reduction in lc-SCFAs, and a significant reduction in Faecalibacterium (compared to the other two treatments).

Furthermore, to avoid the assessment of individual outcomes and strengthen the conclusions regarding microbial and metabolic shifts mentioned above, LEfSe analysis was performed. This analysis identified 41 discriminatory genera (Figure 6). The highest effect sizes were observed for Faecalibacterium (enriched under the synbiotic condition), followed by Bacteroides (enriched with the probiotic treatment) and Lachnospira (enriched with the prebiotic). Several butyrate-producing taxa (e.g., Faecalibacterium, Eubacterium eligens, Agathobacter) were more abundant in synbiotic samples, whereas other taxa (e.g., Clostridium sensu stricto 1, Klebsiella, Lachnospiraceae) were associated with the alternative treatments. These results highlight biomarkers that are not only statistically significant but also biologically meaningful due to their large effect sizes. For transparency, we also report group medians and log2 fold-changes for significant features to complement LDA scores and facilitate biological interpretation (Table S3).

Before leading to the conclusions, a major limitation of the study has to be acknowledged. While the use of the SHIME^® system provided valuable mechanistic information, the use of a single donor limits the direct extrapolation of results to patients with cystic fibrosis, as this approach restricts biological representativeness. However, previous research shows that supplementation with biotics (especially probiotics, the most commonly used in in vivo studies) does not produce uniform effects, but rather responses that depend on each individual [59,60,61]. For this reason, in this study a pooled faecal inoculum was not used, with the aim of providing an initial and specific characterisation of the effect of in vitro treatments on an individual baseline microbiota that could inform exploratory clinical studies and guide future clinical studies. Furthermore, when working with dynamic models such as SHIME^®, the number of available bioreactors limits the experimental design. In this experiment, we included several treatments, so it was not technically possible to dedicate independent units to more than one donor at a time [17,18]. On the other hand, some aspects of the simulation were approached in terms of the most translationally relevant and representative conditions in CF, including the simulation of the altered intestinal conditions and the use of supplement-feasible doses of the pre- and probiotic compounds. Finally, our findings should be interpreted as hypothesis-generating and not predictive of clinical efficacy, as only integrated multi-donor and clinical trials can determine the extent to which these microbiota-mediated improvements translate into clinical outcomes.