Colonic nutritional and physicochemical parameters drive gut microbiota dysbiosis in obesity: what we learned from the Ob-M-ARCOL model development Authors: Ophélie Uriot, Charlotte Deschamps, Lucie Etienne-Mesmin, Morgane Brun, Mélanie Pouget, Sandrine Chalancon, Claude Durif, Cyril Chaudemanche, Monique Alric, Yves Boirie, Stéphanie Blanquet-Diot Published: 2026 ABSTRACT ABSTRACTBackgroundObesity has reached epidemic proportions worldwide with a substantial burden on both individual health and society. Alterations in gut microbiota are increasingly recognized as a key contributor in disease pathogenesis. However, mechanistic understanding at the ecosystem level remains limited. In line with EU and US regulations, in vitro gut systems can be useful to address these questions, but up to now, there is no validated in vitro model of the obese human colon.ResultsTo address this gap, we performed dietary surveys and conducted an extensive analysis of 250 articles to set-up and validate a unique in vitro colonic model, based on human in vivo data and reproducing obese-specific nutritional, physicochemical, and lumen vs mucus-associated microbial parameters. Then, in an original experimental set-up, we cross-compared data on gut microbiota structure and activities during fermentations performed with fecal microbiota from healthy (n = 4) or obese (n = 5) donors and operated under healthy or newly designed obese parameters. Interestingly, applying obese parameters on healthy fecal samples resulted in a significant reduction in microbial α-diversity and in taxa associated with health (e.g., Akkermanciaceae, Rikenellaceae, and Archaea), together with a tendency toward increased production of short-chain fatty acids and associated energy, in full agreement with in vivo data. Conversely, applying healthy parameters on obese fecal samples led to gut microbiota resilience.ConclusionsThese findings highlight the importance of nutritional and physicochemical environment in shaping colonic bacterial and archaeal populations in obesity. This innovative validated model represents a robust and useful platform for mechanistic investigations on gut microbiome in the absence of the host cells, as well as for preclinical evaluation of food and pharmaceutical strategies aiming to restore microbiota eubiosis in a personalized manner. Obesity has emerged as a major public health concern, affecting more than 890 million adults worldwide and its global prevalence has more than doubled since 1990. 1 Defined as an abnormal or excessive fat accumulation (body mass index (BMI) > 30 kg/m²), obesity is a complex and multifactorial disease resulting from an imbalance between energy intake (e.g., high-fat and high-sugar diets) and energy expenditure. It is associated with numerous comorbidities including type II diabetes, metabolic dysfunction-associated steatosis liver disease (MALSD, formerly known as non-alcoholic fatty liver disease), cardiovascular diseases, and cancer. 2 , 3 Growing evidence suggests that alterations of the gut microbiota may contribute to the onset or worsening of obesity with dysbiosis consistently observed in human fecal samples analysis, where characteristic dysbiosis signatures have been identified, including a loss of microbial markers such as Akkermansia, Faecalibacterium, Rikenellaceae, and methanogenic Archaea, alongside an increase in other populations like Prevotella, Fusobacterium, Lactobacillus, or Pseudomonadota. 3-7 These compositional changes are accompanied by functional alterations, including energy harvest from dietary carbohydrates through short-chain fatty acid (SCFA) production, and enhanced intestinal permeability to bacterial lipopolysaccharides (LPS) in response to high-fat diet consumption, exacerbating low-grade inflammation and insulin resistance. 8 , 9 Fecal microbiota transplantation studies in germ-free mice have provided decisive evidence for a causal role of the gut microbiota in obesity development. 10 However, the respective contributions of microbial composition versus host-derived nutritional and physicochemical conditions remain insufficiently disentangled. Although clinical studies remain the gold-standard approach, human in vitro gut models provide complementary experimental systems to investigate gut microbial mechanisms under controlled conditions that are difficult to assess in vivo. 11 , 12 Compared to rodent models, these systems can more accurately reproduce human-relevant diets, physicochemical digestive parameters and microbiota, and provide mechanistic insights into microbe-microbe interactions at the gut ecosystem level. 13 , 14 EFSA recently recognized dynamic in vitro gut models including SHIME (Simulator of the Human Intestinal Microbial Ecosystem), and M-ARCOL (Mucosal Artificial Colon) as suitable human-relevant systems to study gut microbiota dynamics, 15 in line with other broader international initiatives promoting alternatives to animal experimentation, including the European 3 R framework 16 and recent FDA recommendations supporting New Approach Methodologies (NAMs). 17 To date, most in vitro gut models have been developed to mimic the digestive conditions of healthy adults. 18-22 Only a limited number of studies have reported the use of dynamic in vitro gut models operated under obese conditions 23 , including the TNO gastroIntestinal Model (TIM-2), 24-27 the SHIME 28 , 29 and the Copenhagen MiniGut (CoMiniGut) 30 models. However, these approaches primarily relied on fecal inoculation from obese patients, without systematically adapting all the nutritional and physicochemical parameters to reflect the obese colonic environment, despite the well-established influence of these factors in shaping the gut microbiota. While ecological theory predicts that environmental conditions can shape community structure, experimental validation of this causal relationship within a complex human-derived gut ecosystem remains limited, especially under diseased conditions like obesity. Furthermore, robust in vitro-in vivo validation of these models doesn’t exist. Here, we developed and validated a newly sophisticated dynamic in vitro model of obese individuals' colonic environment that integrates multiple obesity-associated nutritional and physicochemical relevant parameters reported in the literature. This model called Ob-M-ARCOL was implemented using the dynamic M-ARCOL 31 system, which uniquely enables the simultaneous cultivation of luminal and mucus-associated microbiota under microbe-driven anaerobiosis. Ob-M-ARCOL constitutes an innovative bioengineered experimental framework designed to investigate causal relationships between colonic environmental factors under obesity and gut microbiota in the absence of host cell-related confounding factors. We therefore leveraged this model to cross-compare gut microbiota structure and metabolic activity during fermentations simulating either healthy or obese conditions, using stools collected from obese patients or healthy volunteers. We specifically tested the hypotheses whether (i) obesity-associated colonic nutritional and physicochemical parameters are sufficient to drive hallmark dysbiotic features in microbiota derived from healthy donors and (ii) conversely, whether healthy colonic conditions can restore microbial structure and metabolic activity in microbiota originating from obese individuals. Materials and methods Stool sampling and treatment Fresh stool samples were collected anaerobically from a cohort of 9 donors, respectively four healthy adults (healthy stools (HS): 2 women and 2 men, 36.3 ± 6.1 y old, BMI = 21.7 ± 2.0 kg/m2) and five obese adults (obese stools (OS): 4 women and 1 man, 41.8 ± 3.9 y old, BMI = 46.7 ± 2.1 kg/m2) (Figure 1a). All donors gave written, informed consent to take part in the study. The use of fecal microbiota of human origin was approved by the French Ethical research committee (Comité de protection des personnes Sud-Est I, CPP Sud-Est I, France) under registration number 2017-69 (OBESAR study, University Hospital CHU Gabriel Montpied, Clermont-Ferrand, EudraCT 2017-A01510-53). Exclusion criteria for both groups included antibiotic use or pre-/probiotic supplementation within the 2 months preceding enrollment, and bariatric surgery for obese participants. Stool samples were frozen at −80 °C within 6 h after defecation, without cryoprotectants. 31 For each donor, a 10% inoculum (w/v) was prepared under strict anaerobic conditions (COY Laboratory Products Inc, Grass Lake, MI, USA) by mixing feces in 30 mM sodium phosphate buffer (pH 6.5) supplemented with 1.9 mM cysteine. The resulting suspension was then filtered through a 500-µm sieve into a sterile bottle. Experimental design of the M-ARCOL model. a) Nine fecal samples were collected from 4 healthy (healthy stools (HS): H1-H4) and 5 obese (obese stools (OS): O1-O5) adult donors. b) Configuration of the M-ARCOL model under healthy (HP: healthy parameters) or obese (OP: obese parameters, new Ob-M-ARCOL set-up) colonic conditions. c) For each donor, two parallel bioreactors were run for 12 d, one under HP and one under OP. d) Luminal, mucosal and atmospheric samples were collected regularly to monitor microbiota composition and metabolic activities. BAI: primary bile acids, BAII: secondary bile acids, SCFA: short-chain fatty acids. Mucosal ARtificial COLon model M-ARCOL is a single-stage continuous fermentation system designed to replicate key physicochemical (pH, body temperature, retention time), nutritional (composition of ileal effluents used as a nutritive medium), and microbial (luminal and mucus-associated communities) characteristics of the human colon, based on in vivo data. 31 , 32 Each airtight glass bioreactor was filled with 200 mL of nutritive medium and initially flushed with O₂-free N₂ to ensure anaerobiosis before the addition of 100 mL of fecal inoculum. Anaerobic conditions were then maintained throughout the fermentation solely by the endogenous microbial metabolism and the airtight system design. 11 , 31 , 32 Bioreactors were kept at body temperature (37 °C) and colonic pH and redox potential were monitored continuously with automatic pH control using 2 M NaOH. The nutritive medium, containing various sources of carbohydrates, proteins, lipids, minerals, and vitamins, was continuously supplied into the bioreactor, while the fermentation medium was simultaneously withdrawn to maintain a defined retention time. Composition of the nutritive medium has been described previously 11 and was adapted for this study based on dietary surveys collected from obese individuals. For the obesity-associated medium, the main differences compared to the healthy medium are listed in Figure 1b and further detailed in the Results paragraph (see below set-up of the in vitro Ob-M-ARCOL model). The mucosal environment was simulated by recirculating a fraction of the fermentation medium through a separate external compartment containing mucin-alginate beads (~556 cm² surface area), prepared by dropwise gelation of 5% porcine mucin and 2% sodium alginate into 0.2 M CaCl₂, and maintained at body temperature. 31 , 33-35 These mucin-beads provide adhesion sites and mucin-derived substrates, allowing the preferential establishment of mucus-associated microbial communities following fecal inoculation while also limiting the washout effect inherent to continuous fermentation systems. 20 , 31-33 , 36 , 37 Experimental design Two bioreactors inoculated with the same stool sample were ran in parallel: one operated under healthy parameters (HP) and the other under obese-specific conditions (obese parameters -OP), in a new set-up called Ob-M-ARCOL, based on human in vivo data (Figure 1). As described previously, 33 , 36 , 37 the fermentation comprised a 24 h batch period allowing microbiota amplification, followed by 10 d of continuous fermentation (Figure 1c). Luminal samples collected daily (Figure 1c and d) were centrifuged at 18,000 g for 15 min at 4 °C. The pellets were stored at −80 °C for microbiota analysis, and supernatants were kept at −20 °C before SCFA quantification. 33 , 36 Anaerobiosis and gas composition were monitored daily, as well as excess gas volume. 33 , 36 As previously described, 31 , 33 , 36 , 37 mucin-alginate beads from the mucosal compartment were collected every 2 d, washed in sterile phosphate-buffered saline, refreshed under a continuous CO2 flow, and stored at −80 °C for mucosal microbiota analysis. DNA extraction Total luminal and mucosal genomic DNAs were extracted using the QIAamp Fast DNA Stool Mini Kit® (Qiagen, Hilden, Germany), using a protocol established in our earlier studies. 33 , 36 Briefly, samples were pretreated with citrate buffer (55 mM, 10 min, 37 °C), vortexed thoroughly, 38 then centrifuged (8000 g, 1 min, 20 °C). Pellets were mechanically disrupted in a bead beater (5 min, 20 beat/sec) with sterile glass beads (300 mg, ranging in diameter from 0.1 to 0.6 mm) and 1 mL of InhibitEX buffer from Qiagen kit. Then, DNA extraction was performed following the manufacturer's instructions. DNA integrity, purity, and concentration were assessed by agarose gel electrophoresis, NanoDrop™ 2000 (Thermo Scientific, Wilmington, DE, USA) spectrophotometry (A260/280 absorbance), and Qubit dsDNA Broad Range fluorometry (Invitrogen, Carlsbad, USA), respectively. 33 , 36 Purified DNA was stored at −20 °C before quantitative PCR analysis and 16S metabarcoding. Quantitative PCR Quantification of total bacteria, Bacillota, Bacteroidota, Gammaproteobacteria, Ruminococcaceae, Prevotellaceae, and Akkermansia muciniphila was carried out by qPCR as described in our previous studies. 31 , 33 , 36 Primers and corresponding annealing temperatures are listed in Supplemental Table 1. Amplifications were run in duplicate using the Takyon Low ROX SYBR 2X MasterMix blue dTTP (Eurogentec, Liège, Belgium) on a Biorad CFX96TM Real-Time System (Bio-Rad Laboratories, Hercules, USA) in a final volume of 10 μL, as previously described in Deschamps et al. 31 Thermal cycling included an initial denaturation at 95 °C for 5 min, followed by 40 cycles of 95 °C for 30 s, primer-specific annealing temperatures (Supplemental Table 1) for 30 s and 72 °C for 30 s. 33 , 36 A melting curve analysis was performed at the end to verify primer specificity. Quantification was based on standard curves generated from 10-fold serial dilutions of genomic DNA isolated from reference strains. Gut microbiota composition by 16S metabarcoding The bacterial V3-V4 region of 16S ribosomal RNA (rRNA) and archaeal 16S rRNA were amplified using primers listed in Supplemental Table 1. Amplicons were generated with a Fluidigm Access Array and sequenced on an Illumina MiSeq platform (Illumina, USA) at the Carver Biotechnology Center, University of Illinois (Urbana, USA). Bioinformatic analyzes were performed with R Studio using the rANOMALY package. 39 Raw reads were demultiplexed and quality-filtered with DADA2, 40 removing reads containing N bases, low-quality bases (Phred ≤ 2), or sequences shorter than 100 bp. Samples were retained for downstream analyzes only if they contained more than 1000 sequencing reads and exhibited a mean Phred quality score > 25. The average sequencing depth was 35,000 ± 10,000 reads per sample. PhiX sequences used as spike-in controls were also filtered out. Remaining sequences were dereplicated and chimeras were removed using DADA2. Amplicon sequence variants (ASVs) were inferred with the dada function from the DADA2 package producing an ASV count table per sample and representative ASV sequences in FASTA format. Taxonomy was then assigned with the idtaxa function from the DECIPHER 41 using the SILVA database release 138.2. 42 For unassigned or broadly classified ASVs (genus level or higher), BLAST alignments 43 were performed with ≥ 98% identity and coverage. A phylogenetic tree was constructed from ASV representative sequences using the phangorn package. 44 Gas analysis by gas chromatography In the M-ARCOL and Ob-M-ARCOL models, gas composition of the atmospheric phase (O2, N2, CO2, CH4, and H2) was analyzed as previously described. 11 , 33 , 36 Measurements were carried out using a 490 Micro‐Gas Chromatograph (Agilent Technologies, Santa Clara, CA, USA) equipped with a Molecular Sieve 5 A column and a PoraPlot U column, both coupled with thermal conductivity detectors (TCD), using argon as the carrier gas. Instrument calibration was performed using four reference gases: (i) ambient air, (ii) a CO2/H2/N2 mixture (5/5/90%), (iii) a CO2/H2 mixture (20/80%), and a CO2/CH4/H2/N2 mixture (19.9/19.9/20/40%). Gas composition was expressed as relative percentages. Short-chain fatty acids (SCFAs) analysis The main luminal SCFAs (acetate, propionate, and butyrate) were quantified following the procedures previously described. 11 , 33 , 36 Analyzes were carried out using High-Pressure Liquid Chromatography (Elite LaChrom, HITACHI, San Jose, CA, USA) coupled with a diode‐array detector. Samples were separated on a Concise Separation column (San Jose, CA, USA) maintained at 50 °C, with 0.04 M sulfuric acid as a mobile phase (0.6 mL/min). Detection was performed at 205 nm and chromatograms were analyzed using EZChrom Elite software. SCFA concentrations were determined from calibration curves generated with standard solutions (concentrations of 0, 10, 25 and 40 mM). 11 Results were expressed as relative percentages or absolute concentrations (mM). Total energy production (kJ/mol) was calculated using the following values for acetate (874.2 kJ/mol), butyrate (2183.5 kJ/mol) and propionate (1527.3 kJ/mol). 27 Statistical analysis Statistical analyzes were performed using GraphPad Prism 9.3.1 (GraphPad Software, San Diego, CA) and R (RStudio). Fermentation data were analyzed on the last 4 d (days 8–11) and daily measurements from this stabilized phase used as technical replicates, since RDA confirmed the absence of significant temporal variation during this period (p = 0.57). Normality was assessed using the Shapiro–Wilk test prior to applying the appropriate statistical tests. qPCR data were analyzed with GraphPad Prism using a mixed-effects model with Geisser–Greenhouse correction followed by Tukey's multiple comparisons test with individual variances computed for each comparison across the four tested conditions. SCFA, gas, and alpha-diversity for luminal and mucosal samples were analyzed in R using Kruskal–Wallis tests followed by Dunn's post-hoc tests with Holm correction for multiple comparisons. In addition to p-values, effect sizes were calculated for pairwise comparisons using Hedges’ g (bias-corrected Cohen's d) to quantify the magnitude of difference between treatments. Effect sizes are interpreted as small (|g|<0.5), medium (0.5≤|g|<0.8), or large |g|≥0.8). For stool alpha-diversity, group comparisons (HS vs OS) were performed using t-tests or Wilcoxon tests depending on data distribution. Beta-diversity was assessed using Bray-Curtis dissimilarities and visualized via principal coordinate analysis (PCoA). Statistical significance of beta-diversity patterns was assessed by PERMANOVA (ADONIS, 999 permutations), with p < 0.05 considered significant. Constrained ordinations (redundancy analysis, RDA) were performed using the vegan package 45 to investigate the effects of donors, colonic microenvironment, days, and donor or colonic parameters type (obese or healthy), with donor effects included as conditional variables when appropriate. Ellipses representing 95% confidence intervals were added to visualize group-specific effects and global as well as pairwise PERMANOVA tests were performed to evaluate the significance of differences between groups. Differential abundance analyzes were conducted using DESeq2, MetagenomeSeq, and Metacoder via the rANOMALY R-package. 39 Taxons detected by at least one method with p < 0.05 were reported. For taxons not detected by DESeq2, no adjusted value is provided. Results and discussion Set-up of an innovative in vitro colon model of obesity, the Ob-M-ARCOL, based on in vivo data In the context of the current worldwide “obesity epidemic” and the growing international shift toward human-relevant alternatives to animal experimentation, supported by recent European and FDA initiatives promoting NAMs, 15-17 we first developed a new in vitro model simulating the colonic ecosystem of obese individuals called Ob-M-ARCOL. This system was adapted from the one-stage M-ARCOL model, which, once inoculated with human fecal microbiota, reproduces the main microbial, nutritional, and physicochemical characteristics of the human colon. 31-33 An extensive literature review of more than 250 scientific articles was conducted, from which the twelve most relevant studies were selected to support the model development. 46-57 The literature screening was performed using a stepwise approach, including title and abstract screening followed by full-text evaluation. The search strategy was based on the following criteria: adult humans aged 18–60 y with BMI > 40 kg/m2; absence of bariatric surgery; no pharmacological treatment (including antibiotics); no probiotic or prebiotic supplementation; availability of quantitative physicochemical and/or nutritional data relevant to the colonic environment; direct comparison between healthy and obese groups under strictly similar experimental conditions; and whenever possible, prioritization of colonic over fecal data to ensure greater physiological relevance. For each selected article, quantitative data were manually extracted for key nutritional and physicochemical parameters relevant to colonic microbial ecology (e.g., substrate availability, pH, transit time). When multiple values were reported, parameter values were defined based on consistent trends observed across studies rather than single extreme values, with preference given to measurements obtained directly in the colon. Given the heterogeneity in study designs and analytical methods, a formal meta-analysis was not performed. Instead, parameter selection relied on a qualitative and comparative synthesis of the literature, focusing on reproducible and biologically consistent differences between obese and non-obese populations. In addition, dietary intake data collected from food surveys from obese patients meeting these criteria were obtained from the Clinical Nutrition Department of CHU Gabriel Montpied University Hospital (Clermont-Ferrand, France) and used to refine and contextualize the nutritional inputs implemented in the model. Figure 1b summarizes the main parameters adjusted to simulate obese condition. The first parameter adapted in Ob-M-ARCOL was the composition of the nutritive medium, which simulates ileal effluents continuously supplied to the bioreactor. Since no data on ileal effluent composition in obese individuals are available, the amount of lipids, carbohydrates, fibers, and salt was estimated using clinical dietary surveys, digestibility indices, and fecal concentrations reported in the literature. 46 , 47 Compared with the healthy condition, lipid content was increased 6-fold (using a 50:50 mixture of butter and palm oil), salt content was increased 1.2-fold, and dietary fiber content was reduced 4-fold and replaced with starch-based products to maintain the carbohydrate levels. We also adjusted bile acid composition by increasing the primary (BA I; cholic acid and chenodeoxycholic acid) to secondary (BA II; deoxycholic acid) bile acid ratio in the nutritive medium (60:40 in obese versus 15:85 in healthy situation), without changing total bile acid concentrations, these adjustments were made in accordance with data from obese stools. 48 , 49 This adjustment enabled the in vitro reproduction of bile acid malabsorption associated with compensatory overproduction in obesity. Incorporating obesity-associated bile acid alterations is particularly relevant, as these molecules strongly shape gut microbial community structure and activity. 50 , 51 Colonic pH was reduced by 0.3 units, based on observations in obese fecal samples, in which modest acidification has been reported and linked to increased SCFA levels. 52 , 53 Lastly, available data on colonic transit time were inconsistent across studies. 54-57 Given the lack of consensus and direct comparisons, we maintained the same retention time used under healthy conditions (i.e., 24 h). Once developed, the Ob-M-ARCOL system was uniquely validated by comparison with relevant human in vivo literature, prioritizing colon-derived data when available or stool-based data otherwise. Microbiota stabilization and shift from fecal to colonic profiles M-ARCOL and Ob-M-ARCOL were then used to run colonic fermentations under healthy or obese parameters, respectively, and inoculated with fecal material from either healthy donors or obese donors in independent runs (Figures 1 and 2). Daily samples were collected from luminal and mucosal compartments, as well as from the atmospheric phase, to monitor microbiota composition and metabolic activity. An adaptation period from in vivo (fecal microbiota) to in vitro (colonic-like profiles) preceded microbiota stabilization in the bioreactor. Individual gas (Supplementary Figure 1) and SCFA (Supplementary Figure 2) profiles indicated that stabilization was reached after 3–4 d under healthy parameters and after 5–6 d under obese parameters. Stabilization of gut microbiota composition required more time, particularly at lower taxonomic levels (Supplementary Figures 3 to 5). Redundancy analysis (RDA) confirmed the absence of a time effect (p = 0.57) at the ASV level from days 8 to 11 (Figure 3d). This stabilization phase reflects microbial adaptation to the in vitro environment, resulting in a significantly reduced richness and evenness compared with the initial stools (Figures 2e and 3a), as previously reported. 31–33,58 Notably, bacterial diversity was significantly higher in the mucosal compartment than in the luminal one (Shannon index of 3.4 and 3.1, respectively, p = 0.0012, data not shown) (Figure 3a) consistent with in vivo observations. 59 This mucosal environment provided additional ecological niches allowing the maintenance of low-abundance taxa originally present in the fecal samples. 35 , 60 , 61 A recent study by Duquesnoy and Chassaing in the MBRA batch model highlighted that integration of this mucosal niche selectively shapes microbial community structure, metabolic output, and pro-inflammatory potential in a donor-dependent manner, highlighting the relevance of this spatial context for intestinal microbiota research. 61 Stabilization of fecal microbiota in a colonic-like environment also led to a clear shift in microbial profiles (Figures 2c, 3c, 4, and 5), with enrichment of bacteria from the Bacteroidota and Pseudomonadota phyla (Supplementary Figure 3), consistent with patterns reported in human in vivo studies. 62 , 63 The mucosal compartment in the Ob-M-ARCOL model also displayed a distinct microbial composition (Figure 3c), characterized by a clear enrichment in Bacillota (Supplementary Figure 3), as reported in a clinical study 59 from obese patients. This enrichment includes species from the Clostridium, Butyrivibrio, or Roseburia genera, which can adhere to mucins or access substrates through cross-feeding with mucin-degrading microbes. 60 Although these taxa were also detected in the luminal compartment, their enrichment in the mucosal niche highlights preferential ecological adaptation rather than exclusive localization. Although luminal and mucosal compartments displayed distinct microbial compositions, functional analysis was not performed separately for each compartment in the present study. Nevertheless, based on in vivo evidence, mucosal-associated communities are thought to play a key role in host–microbiota interactions and quorum sensing, while luminal communities are primarily involved in substrate fermentation and digestion. 64 Finally, in agreement with in vivo literature, 65 the mucosal compartment was colonized by a lower bacterial load (mean: 7.41 ± 0.87 Log10 16S rRNA copies/g, Figure 5c) compared to the luminal compartment (mean: 9.98 ± 0.56 Log10 16S rRNA copies/g, Figure 4c). Characterization of initial fecal samples. Nine fecal samples from healthy volunteers (H1-H4) and obese donors (O1-O5) were analyzed for a) proportions of the three main short-chain fatty acids (acetate, propionate and butyrate), b-c) bacterial and archaeal taxonomic composition (16S rRNA gene metabarcoding), d) selected bacterial groups (qPCR) and e) microbial α-diversity (observed ASVs, Shannon and Simpson indexes, numbers in boxplots indicate median values). Group comparisons were performed using an unpaired t-test (Observed and Shannon indexes) or a Wilcoxon test (Simpson index), and no significant differences were detected. SCFA: short-chain fatty acids, HS: healthy stool, OS: obese stool. α- and β-diversity of microbial communities during M-ARCOL and Ob-M-ARCOL experiments. Diversity indexes were calculated from the ASV table; data points represent individual daily measurements from the stabilized phase (days 8–11, treated as technical replicates. a) α-diversity (Observed ASVs, Shannon numbers in boxplots indicate median values) showed no significant differences (Krustal–Walllis with Dunn–Holm correction). (b-g) β-diversity visualized using two-dimension redundancy analysis (RDA) plots. Panels (b) includes the donor effect; panels (c-g) exclude it. Ellipses represent 95% confidence interval for: donor (b), colonic microenvironment (c), fermentation day (d), donor type (e), parameter type (f), and donor × parameter combination (g). For panels (d-g), stool samples were excluded, and luminal and mucosal samples were analyzed separately. HS: healthy donors (green); OS: obese donors (blue); HP: healthy parameters (circle); OP: obese parameters (square); initial stool (diamond). Numbers inside shapes indicate individual donors (1–4: H1–H4; 5–9: O1–O5. Cross-compared effects of obese versus healthy stools and colonic parameters on bacterial and archaeal populations in the in vitro luminal environment. Two parallel M-ARCOL bioreactors per donor were run under HP or OP. Data correspond to the four endpoint days (stabilized period) of the luminal compartment. (a) Bacterial composition (16S rRNA gene metabarcoding). (b) Differential analysis (DeSeq2, Metacoder, and MetagenomeSeq). Only families identified as significantly enriched by at least one method are shown. (c) qPCR quantification of selected bacterial populations (mean 16S rRNA copies/g ± SEM, ANOVA, p < 0.05). (d) Abundance of methanogenic Archaea (16S rRNA gene metabarcoding). HS: healthy stools, OS: obese stools, HP: healthy parameters, OP: obese parameters. Cross-compared effects of obese versus healthy stools and colonic parameters on bacterial and archaeal populations in the in vitro mucosal environment. Two parallel M-ARCOL bioreactors per donor were run under HP or OP. Data correspond to the four endpoint days (stabilized period) of the mucosal compartment. (a) Bacterial composition (16S rRNA gene metabarcoding). (b) Differential analysis (DeSeq2, Metacoder, and MetagenomeSeq). Only families identified as significantly enriched by at least one method are shown. (c) qPCR quantification of selected bacterial populations (mean 16S rRNA copies/g ± SEM, ANOVA, p < 0.05). No significant difference was detected. (d) Abundance of methanogenic Archaea (16S rRNA gene metabarcoding). HS: healthy stools, OS: obese stools, HP: healthy parameters, OP: obese parameters. Obese colonic parameters are sufficient to induce obesity-related dysbiosis Carefully selected obesity parameters, diet, bile acid profiles, colonic pH and retention time, were implemented into the M-ARCOL model to investigate their impact on the gut microbiome from healthy donors (H1 to H4, Figure 2). Application of obese parameters to healthy microbiota induced a marked shift in microbial structure, characterized by lower abundances of Akkermansiaceae (p = 0.05), Rikenellaceae (p = 0.01), and Ruminococcaceae (p = 0.01) but increased levels of Prevotellaceae (ns, p = 0.12) and Veillonellaceae (p = 0.0005) (Figures 4 and 5, Table 1). Akkermansiaceae were almost completely lost from both the luminal (Figure 4a) and mucosal (Figure 5a) compartments of M-ARCOL under obese parameters (Ob-M-ARCOL), especially in donor H4, resulting in a significant decrease (DESeq2 log2FC = −21.65, p = 1.90 × 10−11) based on differential abundance analysis (Figure 5b). This drop was further confirmed by qPCR quantification of A. muciniphila, showing a significant 2-Log10 decrease following the shift from healthy to obese conditions in the luminal compartment (6.63 ± 0.88 vs 4.63 ± 0.90 Log10 16S rRNA copies, p = 0.021) and in the mucosal compartment (5.39 ± 1.03 vs 3.28 ± 0.06 Log10 16S rRNA copies, p = 0.19) (Figures 4c and 5c). Remarkably, most in vitro microbial changes closely mirrored alterations reported in stool samples from obese patients (Table 2), notably the loss of health-associated taxa such as Akkermansiaceae, Rikenellaceae, and Ruminococcaceae (see Table 2 for references). These shifts, and particularly those affecting Akkermansiaceae, are consistent with dietary patterns associated with low-fiber, high-fat, or high-salt intake, as demonstrated in mice models. 66-72 Conversely, some populations including Synergistota, Tannerellaceae, and Enterococcaceae displayed trends opposite to those reported in human in vivo datasets, with no clear explanation identified in the present study. Changes induced by obese conditions were associated with a trend toward reduced α-diversity in the luminal compartment (Shannon: 3.14 in HS‑HP vs 2.95 in HS‑OP; observed: 71.5 vs 64.0, respectively), although this effect did not reach statistical significance (Figure 3a). Such reductions in microbial diversity are consistently reported in obese adults 73-76 and align with the widely acknowledged definition of gut microbiota dysbiosis, defined as a combined alteration of microbial community composition and/or diversity together with changes in microbial metabolic activity, in accordance with recent ecological frameworks. 77 Despite interindividual variability, microbial profiles consistently clustered according to the applied physicochemical parameters rather than donor type, as shown by ß-diversity analyzes (Figure 3e-g). No significant donor type effect (p = 0.99) was detected, while healthy versus obese parameters significantly structured both luminal and mucosal communities (p = 0.001). Differences between both conditions HS-HP and HS-OP were also significant in both luminal (p = 0.001) and mucosal (p = 0.03) compartments. Table 1.Mean relative abundance of the 20 most abundant bacterial families across HS-HP, HS-OP, OS-HP, and OS-OP conditions, in both luminal and mucosal compartments.FamilyHS-HPHS-OPOS-HPOS-OPHS-HP vs HS-OPOS-HP vs OS-OPHS-HP vs OS-HPHS-OP vs OS-OP f__Acidaminococcaceae 0.44 ± 0.391.15 ± 1.800.35 ± 0.560.99 ± 1.20ns*nsns f__Akkermansiaceae 0.23 ± 0.650.00 ± 0.000.67 ± 1.050.05 ± 0.26***nsns f__Bacteroidaceae 59.55 ± 13.2944.41 ± 25.2670.08 ± 11.4556.13 ± 23.19ns**ns f__Bacteroidales_family 0.14 ± 0.160.10 ± 0.120.06 ± 0.100.15 ± 0.27nsnsnsns f__Barnesiellaceae 0.23 ± 0.450.11 ± 0.170.03 ± 0.070.03 ± 0.06nsnsnsns f__Christensenellaceae 0.19 ± 0.620.03 ± 0.140.08 ± 0.160.00 ± 0.00ns**nsns f__Clostridia_family 0.77 ± 1.230.15 ± 0.170.80 ± 0.680.51 ± 0.54nsnsnsns f__Desulfovibrionaceae 0.10 ± 0.120.07 ± 0.120.08 ± 0.180.18 ± 0.33nsnsnsns f__Enterobacteriaceae 0.46 ± 0.660.00 ± 0.000.15 ± 0.340.13 ± 0.28***ns** f__Gammaproteobacteria_family 0.17 ± 0.390.00 ± 0.000.02 ± 0.060.03 ± 0.11*nsnsns f__Lachnospiraceae 9.10 ± 5.4910.73 ± 9.2010.40 ± 5.5615.71 ± 7.58ns*ns* f__Oscillospiraceae 5.22 ± 3.542.35 ± 2.141.89 ± 0.891.32 ± 1.28******ns f__Oscillospirales_family 1.10 ± 2.331.67 ± 4.163.83 ± 6.953.03 ± 5.19nsnsnsns f__Prevotellaceae 7.93 ± 11.2730.15 ± 29.772.32 ± 4.3814.09 ± 19.14nsnsnsns f__Rikenellaceae 4.82 ± 6.611.10 ± 2.332.66 ± 3.471.45 ± 2.51**nsns f__Ruminococcaceae 5.20 ± 5.891.49 ± 2.063.34 ± 2.521.95 ± 1.71*nsnsns f__Sutterellaceae 3.59 ± 1.994.17 ± 3.361.91 ± 1.532.51 ± 2.03nsns*ns f__Synergistaceae 0.01 ± 0.020.01 ± 0.040.45 ± 1.050.21 ± 0.82nsnsnsns f__Tannerellaceae 0.18 ± 0.290.23 ± 0.380.26 ± 0.310.09 ± 0.14ns*nsns f__Veillonellaceae 0.38 ± 0.422.03 ± 2.010.37 ± 0.961.19 ± 1.39*******nsRelative abundances are expressed as percentages and reported as mean ± SD, averaged over stabilization days (days 8 to 11) for each bioreactor. Pairwise comparisons between experimental groups were performed using Ente Wilcoxon rank-sum tests with Benjamini-Hochberg correction for multiple testing. Significance is indicated as p < 0.05 (*), p < 0.01 (**), and p < 0.001 (***). HS: healthy stools, OS: obese stools, HP: healthy parameters, OP: obese parameters, ns: not significant. Mean relative abundance of the 20 most abundant bacterial families across HS-HP, HS-OP, OS-HP, and OS-OP conditions, in both luminal and mucosal compartments. Relative abundances are expressed as percentages and reported as mean ± SD, averaged over stabilization days (days 8 to 11) for each bioreactor. Pairwise comparisons between experimental groups were performed using Ente Wilcoxon rank-sum tests with Benjamini-Hochberg correction for multiple testing. Significance is indicated as p < 0.05 (*), p < 0.01 (**), and p < 0.001 (***). HS: healthy stools, OS: obese stools, HP: healthy parameters, OP: obese parameters, ns: not significant. Table 2.Shifts in gut microbiota composition and metabolic activities: in vitro/in vivo data comparisons. a: For each healthy donor (H1-H4), shift of ≥ 25 % in bacterial composition and activity between HS_OP and HS_HP (control) condition were considered meaningful. b: For each obese donor (O1-O5), shift of ≥ 25 % in bacterial composition and activity between OS_HP and OS_OP (control) condition were considered meaningful. : increase in relative abundance or concentration; : decrease in relative abundance or concentration; : no difference; -: non detected. c: In vivo data (colon-derived or fecal samples) were obtained from published human studies and were used for comparison with in vitro results generated in the M-ARCOL. Inclusive criteria for in vivo data were adult patients with BMI > 30 kg/m2, no drug treatment, no medical intervention such as bariatric surgery, no probiotic or prebiotic supplementation, and comparison with healthy controls reported within the same study. Ratio: proportion of articles in each category (less abundant/more abundant/no difference); Bolded ratio: consensus for in vivo data, ★★★: > 75% of studies in agreement; ★★: 50–75% in agreement; ★: < 50% in accordance or only one article available. Green symbol: in vitro data (this study) are consistent with in vivo literature; orange symbol: no clear conclusion, mainly due to lack of in vivo consensus; red symbol: in vitro and in vivo data are contradictory. HS: healthy stool, OS: obese stool, HP: healthy parameters, OP: obese parameters, SCFA: short-chain fatty acids. 108-135 Shifts in gut microbiota composition and metabolic activities: in vitro/in vivo data comparisons. a: For each healthy donor (H1-H4), shift of ≥ 25 % in bacterial composition and activity between HS_OP and HS_HP (control) condition were considered meaningful. b: For each obese donor (O1-O5), shift of ≥ 25 % in bacterial composition and activity between OS_HP and OS_OP (control) condition were considered meaningful. : increase in relative abundance or concentration; : decrease in relative abundance or concentration; : no difference; -: non detected. c: In vivo data (colon-derived or fecal samples) were obtained from published human studies and were used for comparison with in vitro results generated in the M-ARCOL. Inclusive criteria for in vivo data were adult patients with BMI > 30 kg/m2, no drug treatment, no medical intervention such as bariatric surgery, no probiotic or prebiotic supplementation, and comparison with healthy controls reported within the same study. Ratio: proportion of articles in each category (less abundant/more abundant/no difference); Bolded ratio: consensus for in vivo data, ★★★: > 75% of studies in agreement; ★★: 50–75% in agreement; ★: < 50% in accordance or only one article available. Green symbol: in vitro data (this study) are consistent with in vivo literature; orange symbol: no clear conclusion, mainly due to lack of in vivo consensus; red symbol: in vitro and in vivo data are contradictory. HS: healthy stool, OS: obese stool, HP: healthy parameters, OP: obese parameters, SCFA: short-chain fatty acids. 108-135 For the first time, under obese conditions, we also investigated a nonbacterial microbial fraction in the M-ARCOL and Ob-M-ARCOL systems, namely the archaeal community, with a particular focus on methanogenic Archaea. These microorganisms have been implicated in health and various diseases and are often reported to display an inverse association with obesity. 78-80 While a murine study suggested that interspecies H2 transfer between bacteria and Archaea may enhance energy harvest in obesity, 81 several human studies have instead reported a negative correlation between Archaea abundance and BMI. 53 , 73 , 82 , 83 In our study, the two CH4-producing healthy donors harbored detectable Archaea, upon implementation of obese parameters. These methanogenic Archaea were almost completely lost (Figures 4d and 5d, Supplementary Figure 1) and associated with a decrease in CH4 production, although it does not reach statistical significance (Hedge's g = 3.62, p = 0.10, Figure 6). Additional metabolic changes were observed under obese conditions, including an increase in CO2 relative abundance, which was significant in non-CH4 producers (Hedge's g = 3.48, p = 1.55 × 10−4). This shift may reflect an increased in carbohydrate fermentation and a tendency toward higher total gas production under obese conditions (Hedge's g = 1.17, p = 0.87, Figure 6). In donor H2 only, obese parameters also induced increased total daily SCFA concentrations (mean: 140.13 ± 3.82 vs 258.38 ± 7.05 mM (p = 0.021) for HS-HP and HS-OP, respectively, data not shown) and associated energy output (mean: 180.58 ± 4.87 vs 362.19 ± 13.69 kJ/mol (p = 0.021), data not shown, Figure 7 and Supplementary Figure 2). This increase is consistent with human studies reporting elevated SCFA levels in stools from obese individuals 52 , 53 , 84-86 and is in accordance, although based on a different experimental framework, with the work of Turnbaugh et al. 87 demonstrating an enhanced energy harvesting capacity of the obese microbiome in mice. The rise in total SCFA concentration was not accompanied by major changes in the relative proportions of the three major SCFA (Figure 7a), in accordance with several studies in human cohorts. 52 , 53 , 84-86 Cross-compared effects of obese versus healthy stools and colonic parameters on atmospheric gas production. Gas composition and production were measured daily in the atmospheric phase of the M-ARCOL model. Data points represent individual daily measurements from the stabilized phase (days 8–11, treated as technical replicates). (a) Gas composition for each donor under HP or OP. (b) Median percentages of CH4, CO2, H2, and O2 separated by CH4-producers and non-CH4 producers. (c) Daily gas volume produced in mL. Group differences were assessed by Kruskal–Wallis test with Dunn's test and Benjamini–Hochberg correction. Significant pairwise comparisons are shown: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***). Numbers in boxplots indicate median values. HS: healthy stools, OS: obese stools, HP: healthy parameters, OP: obese parameters. CH4: methane, CO2: carbon dioxide, H2: dihydrogen, N2: dinitrogen, O2: dioxygen. Cross-compared effects of obese versus healthy stools and colonic parameters on short-chain fatty acid production. Short-chain fatty acid composition and concentrations were measured daily in the luminal compartment. Data points represent individual daily measurements from the stabilized phase (days 8–11, treated as technical replicates). (a) Proportions of acetate, propionate, and butyrate for each donor under HP or OP. (b) Median daily concentrations in acetate, propionate, and butyrate (mM). (c) Daily total SCFA concentration (mM) and (d) corresponding energy production (kJ/mol). Group differences were assessed by Kruskal–Wallis test with Dunn's test and Benjamini–Hochberg correction. Significant pairwise comparisons are shown: p < 0.01 (**), p < 0.001 (***). Numbers in boxplots indicate median values. HS: healthy stools, OS: obese stools, HP: healthy parameters, OP: obese parameters. Lastly, comparison between healthy stools exposed to obese parameters (HS-OP) and obese stool exposed to obese parameters (OS-OP) revealed only minor differences in specific taxa, including Ruminococcaceae (DESeq2 log2FC = -1.39, p < 0.05) and Oscillospirales_family (DESeq2 log2FC = -4.70, p = 0.039) restricted to the luminal compartment (Figure 4b). When pooling luminal and mucosal environment, no significant differences were observed for most dominant taxa, with only Enterobacteriaceae (p = 0.017) and Lachnospiraceae (p = 0.017) showing significant differences between HS-OP and OS-OP (Table 1). β-diversity analyzes consistently demonstrated clustering driven by experimental parameters (healthy vs obese; luminal and mucosal, p = 0.001) rather than donor type (healthy vs obese donors; luminal and mucosal, p = 0.99), with HS-OP versus OS-OP comparisons showing significance only in the luminal compartment (p = 0.001; mucosal, p = 0.46), supporting previous observations (Figure 3e-g). Collectively, these results are consistent with the possibility that obesity-associated gut microbial dysbiosis can be adequately modeled in vitro from healthy donor stools solely by adjusting dietary and physicochemical conditions, suggesting that the use of obese donor stools is not strictly necessary for developing an in vitro obese colonic model. Healthy colonic parameters shift gut microbial composition toward eubiosis The next step of our study aimed to evaluate the impact of healthy versus obese parameters when fermenters were inoculated with stools from obese patients (Figure 1, donors O1 to O5). Compared to obese parameters applied on obese stool (OS-OP), the application of healthy parameters (OS-HP) resulted in a restoration of microbial taxa considered markers of gut health, including Akkermansiaceae (DESeq2 log2FC = 5.28, p = 0.048), Ruminococcaceae (DESeq2 log2FC = 1.39, p = 0.022) and Christensenellaceae (DESeq2 log2FC = 5.95, p = 0.0023) in both luminal and mucosal phases (Figures 4 and 5, Table 2 for in vivo references). These taxa had previously been lost when obese parameters were applied to healthy stools (HS-OP). Conversely, taxa typically enriched in obese stools 66 , 68 , 71 , 88-90 such as Lachnospiraceae (p = 0.02), Prevotellaceae (DESeq2 log2FC = -2.45, p < 0.05), and Veillonellaceae (DESeq2 log2FC = -2.85, p = 0.017) significantly decreased in healthy conditions (Figures 4 and 5, Tables 1 and 2, for in vivo references). As in the reciprocal experiments described above, Akkermansiaceae showed the most pronounced changes, with approximately 39-fold increase in the luminal phase (DESeq2 log2FC = 5.28, p = 0.048, Figure 4b), while Christensenellaceae displayed the strongest enrichment in the mucosal compartment, reaching an approximately 62-fold increase (DESeq2 log2FC = 5.95, p = 0.0023, Figure 5b). These findings were further supported by qPCR quantification of A. muciniphila, revealing a significant 1.6-Log10 increase under healthy conditions in luminal (5.87 ± 2.39 vs 4.24 ± 0.68 Log10 16S rRNA copies, p = 0.0009) and mucosal (4.53 ± 1.49 vs 2.88 ± 0.31 Log10 16S rRNA copies, p = 0.35) environments (Figures 4 and 5c). Healthy parameters also tended to restore bacterial richness (Observed ASV index) in both compartments (Figure 3a), likely reflecting the higher fiber content of the healthy nutritive medium, a known determinant of gut microbiota diversity. 91 Regarding methanogenic Archaea which was detected in four donors, restauration patterns were more heterogeneous. Archaea_family abundances increased under healthy conditions in donors O1 and O2 in the luminal phase (OS-HP = 115 ± 19.2 vs OS-OP = 30.2 ± 28.1), whereas Methanobacteriaceae decreased in donors O2, O4, and O5 (OS-HP = 6.83 ± 8.98 vs OS-OP = 27.8 ± 30.3) (Figure 4d). In donors O1 and O2, higher archaeal prevalence coincided with a marked decrease in Prevotellaceae (DESeq2 log2FC = −2.45, p < 0.05) (Figures 4a and d and 5a and d). Although a relationship between these populations has been previously suggested, 92 their putative synergistic role in obesity is not supported by our results. 81 Nevertheless, CH4 production was significantly higher (Hedge's g = 2.67, p = 1.52 × 10−4) under healthy compared with obese parameters, consistent with the observed increase in Archaea, again demonstrating microbiota reversibility (Figure 6). This effect represented the most notable metabolic change, as daily gas production was not modified (Hedge's g = 0.04, p > 0.05) while SCFA production (median values 113.84 vs 142.93 mM, Hedge's g = 0.65, p > 0.05), as well as associated energy production (median values 182.37 vs 160.98 kJ/mol, Hedge's g = 0.51, p > 0.05), tended to increase when healthy parameters were applied to obese stools (OS-HP) (Figures 6 and 7). β-diversity analysis further confirmed that obese stools under healthy parameters (OS-HP) clustered closely with healthy stools under healthy parameters (HS-HP). Community structuring was primarily driven by physicochemical parameters (p = 0.001 in both luminal and mucosal), whereas no significant donor type effect was detected in either compartment. OS-HP versus OS-OP differences were significant in both luminal (p = 0.001) and mucosal (p = 0.006) compartments (Figure 3e-g). Our results align with in vivo observations in obese patients undergoing dietary interventions such as calorie, fat, or carbohydrate restriction, which are associated with increases in Akkermansiaceae, Christensenellaceae, and Rikenellaceae (Alistipes), and decreases in Prevotella and Veillonellaceae. 93-98 Taken together, these results suggest that the M-ARCOL model captures key features of microbiota restoration processes associated with balanced dietary conditions. M-ARCOL retains individual features of the human gut microbiota Fecal microbial composition in human is highly individual-dependent, 62 , 99 and only part of the population harbors methanogenic Archaea capable of producing methane. In our study, RDA revealed that donor origin was the primary factor explaining microbiota dissimilarities (Figure 3b), demonstrating that in vivo interindividual variability was preserved both in the M-ARCOL and Ob-M-ARCOL models. Importantly, capturing this donor-dependent variability represents a major strength of our in vitro system, as it provides a more relevant representation of the heterogeneity observed in the human gut microbiota in vivo. Differences between donor stools in obesity-associated marker populations, including Akkermansiaceae, methanogenic Archaea, Rikenellaceae, and Christensenellaceae, were similarly maintained in vitro (Figure 2b-d, Supplementary Figures 4, 5, and 6). Donor-dependant variability in microbial composition was also reflected in functional outputs, including SCFA and gas production (Figure 2a, Supplementary Figures 1 and 2). In particular, the presence of methanogenic Archaea in stools and fermentation media strongly correlated with CH4 production in M-ARCOL (Figures 2c and d, 4d, 5d, and 6, and Supplementary Figure 1). These observations further support the relevance of using single-donor rather than pooled stools, as interindividual microbiota composition is itself a major driver of interindividual variability. 100 Ob-M-ARCOL: in vitro obese model for mechanistic insights and translational personalized interventions: opportunities and limits In line with recent EFSA and FDA guidelines and the European 3 R principles encouraging the development of in vitro approaches to reduce animal experimentation, 15-17 we developed and fully validated a cutting-edge in vitro model reproducing key microbial, nutritional, and physicochemical conditions of the human obese colon. Our findings demonstrate that this new Ob-M-ARCOL system accurately recapitulates both structure and metabolic features of the obese gut microbiota, capturing fine-scale disease-associated perturbations. To our knowledge, this is the first in vitro human gut model combining multiple obesity-associated environmental parameters with systematic validation against in vivo microbial and metabolic data (Table 3). Current obesity-associated in vitro models mainly rely on fecal inoculation from obese donors, 24-30 , 101 whereas Ob-M-ARCOL additionally integrates specific changes in available nutrients arriving at the colon, bile acid concentrations, pH, and transit time associated with obesity. Another major feature of the model is the simultaneous reproduction of luminal and mucus-associated microbiota, providing particular relevance for studying mucin-degrading populations such as Akkermansiaceae. For the first time, by cross-comparing healthy versus obese parameters and healthy versus obese stools, we demonstrate that inoculation with obese stools is not strictly required to reproduce obesity-associated dysbiosis in vitro, thereby potentially simplifying experimental and regulatory constraints. These findings further emphasize the dominant role of colonic physicochemical and nutritional parameters in shaping the gut microbial ecosystems. 102 Although the Ob-M-ARCOL model offers a controlled and reproductible system to perform mechanistic studies on microbial dynamics, it does not incorporate the host cell environment and therefore cannot capture direct host cell-microbiota interactions. Consequently, important pathophysiological aspects of obesity such as alterations in epithelial barrier integrity and immune or hormonal dysregulation 3 , 103 , 104 cannot be fully reproduced. To partially overcome these limitations, future studies will focus on coupling this model with host-based systems (intestinal epithelial, liver epithelial, immune cell lines, or intestinal organoids) to better approximate host‒microbiota crosstalk. In this study, we illustrated that instead of being considered only as a reductionist approach, Ob-M-ARCOL also showed the major advantage of allowing precise mechanistic investigations that are impossible to achieve in vivo due to host-related confounding factors. Particularly, our approach allows to distangle the respective contributions of gut microbiota composition (through inoculation with fecal samples from obese patients or healthy volunteers) and colonic environmental parameters (through the set-up of healthy or obese parameters in the M-ARCOL) in obese-related perturbations of the gut microbiome. Table 3.Comparison of Ob-M-ARCOL with the currently published dynamic model of the human obese colon including microbiota.NameMono/multi compart-mentedDigestive section simulatedType of fermen-tationType of microbiotaPhysicochemical parametersMedium composition (g/L)Anaero-biosisMucus-associated microbiotaEukaryota and/or Archaea monitoringValidation in vivo vs in vitro ReferencespH controlTemperature control (°C)Retention time (h)Redox monito-ringFibersSimple carbo-hydratesProteinsLipidsOligos/vitaminesBile acidsTIM-2MonoProximal colonSCFecal samples(obese) ✗ ✓ - - ✗ ✗ ✗ ✗ ✗ ✗ ✓ N2 flushing ✗ ✗ ✓ 24-27 SHIME-based modelMultiSmall intestine, proximal and distal colonSCFecal samples(obese) ✗ ✓ ✓ PC: 20 hDC: 32 h - ✗ ✗ ✗ ✗ ✗ ✗ ✓ N2 flushing ✗ ✓ Fungi ✗ 28 Triple SHIMEMultiStomach, small intestine, proximal and distal colonSCFecal samples(obese) ✗ ✓ ✓ PC: 20 hDC: 32 h - ✗ ✗ ✗ ✗ ✗ ✗ ✓ N2 flushing ✗ ✗ ✗ 29 CoMiniGutMonoColonBFecal samples(obese) ✗ ✓ - - ✗ ✗ ✗ ✗ ✗ ✗ ✓ N2 flushing ✗ ✗ ✗ 30 Ob-M-ARCOLMonoColonCFecal samples (healthy or obese) ✓ ✓ ✓ (24 h) ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ Microbiota only ✓ ✓ Archaea ✓ The present studyThe in vitro model reproduces based on in vivo data (✓), partially considers (✓) or does not simulate (✗) the corresponding parameter when simulating obese conditions.B: batch; C: continuous; DC: distal colon; PC: proximal colon; SC: semi-continuous. Comparison of Ob-M-ARCOL with the currently published dynamic model of the human obese colon including microbiota. The in vitro model reproduces based on in vivo data (✓), partially considers (✓) or does not simulate (✗) the corresponding parameter when simulating obese conditions. B: batch; C: continuous; DC: distal colon; PC: proximal colon; SC: semi-continuous. Beyond model development and mechanistic insights, Ob-M-ARCOL also opens important perspectives for translational microbiome research. The system could be used in preclinical studies, as an alternative to rodent experiments, to evaluate the impact of nutritional, microbial, or pharmaceutical interventions (e.g., ingredients with reduced obesogenic potential, plant extracts, prebiotics, probiotics, drugs, or fecal microbiota transplantation) on gut microbiota composition, diversity, and functions, with the objective of identifying innovative strategies to restore gut microbial eubiosis. Importantly, the preservation of donor-specific microbial signatures and activities allows the study of interindividual variability and provides an accurate capture of the in vivo situation in human. This will help in identifying responders and nonresponders to specific dietary or pharmacological interventions. 105 , 106 Such variability may be related to baseline microbial diversity which is known to influence responses to dietary restriction in obese patients. 107 Ultimately, this unique in vitro obese colon model represents a promising preclinical framework for the development of personalized nutritional or pharmaceutical strategies targeting the gut microbiota, which, in combination with host cells or gut-on-chip models, may contribute to elucidating the etiopathology of obesity and its associated comorbidities. Supplementary Material SuppFigS1.tif SuppFigS1.tif SuppFigS4.tif SuppFigS4.tif SuppFigS3.tif SuppFigS3.tif SuppFigS2.tif SuppFigS2.tif 2025 SuppFig caption and tableS1_R1.docx 2025 SuppFig caption and tableS1_R1.docx SuppFigS5.tif SuppFigS5.tif SuppFigS6.tif SuppFigS6.tif