Roseburia intestinalis-derived inosine improves intestinal motility by activating TGF-β1/p-Smad3/Transgelin signaling axis

Abstract

Background

Functional constipation (FC) is a prevalent gastrointestinal disorder marked by impaired intestinal motility, affecting millions worldwide and significantly diminishing quality of life. Current therapeutic strategies provide only transient symptom relief and fail to restore colonic motor function. This study aims to explore the causal role of gut microbiota in FC, focusing on the mechanisms by which Roseburia intestinalis (RI) and its metabolite inosine improve intestinal motility.

Methods

To identify protective gut taxa, we first performed bidirectional Mendelian randomization (MR) integrated with 16S rRNA sequencing on fecal samples from 30 FC patients and 30 healthy controls. Loperamide-induced constipation mouse models evaluated RI gavage effects on motility indicators, including fecal output (frequency and water content), gastrointestinal transit time, intestinal propulsion rate, colonic bead expulsion time, and ex vivo organ bath contractions. To elucidate the underlying mechanism, we conducted RNA sequencing on colonic tissues. Furthermore, comparative metabolomics was employed to identify the key bacterial metabolites responsible for RI’s effects. Finally, the proposed mechanism of action for both RI and its key metabolite was validated through a combination of histology, immunohistochemistry, immunofluorescence, quantitative PCR, and Western blot analyses.

Results

MR and sequencing identified RI as a protective factor (OR = 0.8675, P < 0.05), with reduced abundance in FC patients negatively correlating with symptom severity and quality-of-life scores. In constipation mouse models, RI gavage promoted intestinal motility, enhanced fecal water content and gastrointestinal transit, and ameliorated mucosal injury while improving mucus secretion. RNA sequencing revealed upregulation of Tagln (transgelin) and enrichment of the TGF-β pathway after RI intervention. Metabolomics analysis pinpointed inosine as a key RI-derived small-molecule metabolite, enriched in RI supernatant and depleted in FC patient feces. Inosine gavage improved motility and histology, increased TGF-β1, p-Smad3, and transgelin, and these effects were abolished by the TGF-β receptor inhibitor SB431542.

Conclusions

These findings established that RI and its metabolite inosine promote intestinal motility via TGF-β1/p-Smad3 signaling and identified transgelin as the terminal effector of this cascade, offering mechanistic insight into microbiota–host interactions and pointing to therapeutic targets for gastrointestinal motility disorders.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-07366-6.

Background

Functional constipation (FC) is a prevalent and burdensome gastrointestinal disorder that affects more than 10% of the global population [1], with prevalence rising to 18.9% in older adults [2]. This condition arises from disrupted intestinal motility, where coordinated contractions of colonic smooth muscle play a central role in propelling contents through the gut [3]. Our previous research has found that within this contractile apparatus, transgelin—an actin-binding cytoskeletal protein—helps stabilize the contractile network and sustain tonic force [4]. Reductions in transgelin expression have been linked to impaired smooth-muscle performance in gastrointestinal disease contexts [4]. Current therapeutic strategies, such as dietary fiber supplementation, osmotic laxatives, and stimulant laxatives, offer symptomatic relief to some patients but are often associated with limited efficacy, adverse side effects, and a failure to address the underlying pathophysiological mechanism [5]. These limitations underscore the need for mechanism-based treatments that restore gut motility.

Emerging evidence highlights the gut microbiota as a key regulator of intestinal motility [6]. Although findings are not fully consistent, many studies link constipation to dysbiosis characterized by decreased beneficial acid-producing taxa—such as Bifidobacterium, Lactobacillus, and Roseburia—and increased pro-inflammatory or potentially pathogenic bacteria [7, 8]. Microbial communities can influence the enteric nervous system (ENS) and smooth-muscle activity through diverse mediators, such as serotonin and short-chain fatty acids [9]. Despite these associations, most research on gut microbiota in constipation just relies on observational data from 16S rRNA sequencing, which identifies correlations but fails to establish causality. Mendelian randomization (MR) based on genome-wide association studies (GWAS) offers a complementary strategy by using genetic variants as instrumental variables to infer causal links between specific microbes and disease outcomes, thereby reducing confounding and reverse causation [10]. Integrating MR with microbial profiling combines genetic inference with real-world community signals, strengthening biological plausibility and translational relevance. However, such integrated analyses in constipation remain scarce.

Building on existing work and our preliminary analyses, Roseburia has been identified as a bacterial group associated with fecal consistency and positively correlated with looser stools [11]. Reduced abundance of Roseburia intestinalis (RI) has been observed in fecal samples from patients with FC, especially slow-transit constipation (STC) [12, 13], suggesting a potential protective role against constipation. RI has also shown protective effects against various digestive and systemic diseases, including experimental colitis, colorectal cancer (CRC), and depression [14–16]. Despite these links and the recognized importance of RI in gut homeostasis, a direct causal role for RI in regulating intestinal motility has not been demonstrated. Moreover, the specific effector molecules and mechanisms by which RI might enhance motility remain undefined.

Prior studies have largely been associative, leaving two critical gaps: (i) genetic-level causal evidence linking RI to constipation risk, and (ii) a defined mechanistic chain connecting a specific RI-derived metabolite to colonic motility. Here, we identified microbial taxa with potential therapeutic relevance for constipation by integrating bidirectional MR with 16 S rRNA sequencing, and prioritizing RI as a protective factor. We then combined clinical indices with in vivo validation in constipation mice model, and deployed an integrated multi-omics strategy, including RNA sequencing and metabolomics, to map downstream host pathways and isolate key RI-derived metabolites. Our findings showed that RI and its metabolite inosine enhance colonic motility by activating the TGF-β1/p-Smad3/Transgelin axis. These findings not only provided new insights into the intestinal motility mechanism but may also identified potential targets for the treatment of constipation, offered a practical path for microbiome-guided interventions.

Methods

Detailed methods related to the instrumental variables selection, sensitivity analysis of MR, the details of questionnaires, the details of animal experiments’ design, 16 S rRNA gene sequencing, DNA/RNA extraction and qPCR, bulk RNA sequencing, LC-MS/MS analyses and western blot are provided in Supplementary methods. The antibodies used in this study are listed in Table S7.

Data source

The characteristics of the associated GWAS data sources are outlined in Table S1. The genetic instrumental variables (IVs) of each bacterial taxon were obtained from the largest meta-GWAS of human gut microbiota, which comprised 18,340 individuals from 24 cohorts, 14,363 of whom (>78%) were of European ancestry (Table S2) [17]. The microbiome GWAS was adjusted by age, sex, technical covariates and genetic principal components. After removing 15 unknown bacterial taxa, the GWAS data we obtained finally covered a total of 196 taxa (sorted by taxonomy): 9 phyla, 16 classes, 20 orders, 32 families and 119 genera (Table S3) [17]. The constipation GWAS summary statistics were derived via the FinnGen Consortium R9 release (https://r9.risteys.finngen.fi/), encompassing 341,255 controls and 36,022 constipation cases.

Bidirectional univariate MR (UVMR) analysis

The design of the MR study is illustrated in Fig. 1A. Two-sample UVMR analysis was conducted to estimate the total causal effects of: (i) GM on constipation and (ii) constipation on GM. Multiple analytical methods were employed, including MR Egger, inverse variance weighting (IVW), simple mode, weighted mode, and weighted median. IVW served as the primary method, with Wald ratio tests applied to features with a single IV [18]. All causal estimates are reported as odds ratios (ORs) with 95% confidence intervals (CIs). Results were considered statistically significant when both MR-Egger and IVW analyses were consistent, and the IVW P-value was < 0.05. The Benjamini–Hochberg method was used to correct the P-value and avoid false-positive rates in multiple hypothesis testing.

Fig. 1.

Mendelian randomization analyses of gut microbiota and constipation. (A) The flowchart of the hypothesis of MR analyses. (B) All results of MR analysis and sensitivity analysis between gut microbiota and constipation. (C) Forest plots summarizing the MR results of gut microbiota with a causal relationship to constipation using the IVW method (P < 0.05). IVW, inverse-variance weighted; MR, Mendelian randomization; OR, odds ratio

Patients

This cross-sectional study was performed at the General Hospital of Tianjin Medical University in Tianjin, China, between August 2022 and June 2024. We enrolled participants aged 18 to 75 years who met the diagnostic criteria for FC as defined by Rome IV criteria [19]. To be eligible, individuals had to exhibit at least two of the following symptoms: (a) straining during more than 25% of defecations; (b) lumpy or hard stools in over 25% of defecations; (c) a sensation of incomplete evacuation for more than a quarter of defecations; (d) a feeling of anorectal obstruction or blockage in more than a quarter of defecations; or (e) a frequency of fewer than three spontaneous bowel movements per week. The study explicitly excluded individuals with irritable bowel syndrome (IBS) and opioid-induced constipation. Further exclusion criteria were: (a) use of probiotics or antibiotics within the past three months; (b) any surgical procedure within the last six months; (c) and the presence of any suspected organic lesions or alarm symptoms such as fever, bloody stools, or melena. Participants with physical disabilities, or severe cardiovascular, gastrointestinal, hepatic, renal, or endocrine diseases were also ineligible. Ultimately, a total of 60 participants successfully completed the questionnaire and provided samples. For a subset of this cohort, colon samples were procured via colonoscopy from five individuals in each group.

Questionnaire survey

The Patient assessment of constipation symptom (PAC-SYM) and Bristol stool scale (BSS) were selected to evaluate the patients’ constipation symptoms [20, 21]. The Patient assessment of quality of life (PAC-QOL) was used to assess the impact of constipation on quality of life over the previous two weeks [22]. Details of the questionnaires are provided in the Supplementary methods.

Bacterial strains and culture conditions

As previously described [23], Roseburia intestinalis (DSMZ-14610; Deutsche Sammlung von Mikroorganismen und Zellkulturen, Braunschweig, Germany) was obtained via the American Type Culture Collection (ATCC) and cultured in modified reinforced clostridial medium (RCM; MZMD039B, Mingzhou Biotechnology Co., Ltd, China) under strict anaerobic conditions at 37 °C. Cultures were centrifuged at 4,000 × g for 20 min at 4 °C to separate pellets and supernatants. The clarified supernatant was passed through a 0.22-µm membrane to yield sterilized supernatant. For size-based partitioning, ultrafiltration spin columns were used with a 3-kDa molecular weight cut-off (MWCO) to operationally enrich low-molecular-weight metabolites in the < 3 kDa permeate and proteins/large peptides in the ≥ 3 kDa retentate.

Animal experiments

Male C57BL/6J mice (age: 6 weeks, weight: 18–23 g) were obtained from the Beijing Animal Research Center. The mice were maintained under specific pathogen-free (SPF) conditions in a temperature-controlled barrier facility with a 12-hour light/dark cycle and provided ad libitum access to food and water for a 1-week adaptation. The design of animal experiment 1–4 are provided in the Supplementary methods (Figs. 3A, 5I and 6A, and Fig. 8A).

Fig. 3.

RI improves intestinal motility in mice model of constipation. (A) Experimental design for animal experiment 1 in SPF mice. (B) Representative images of fluorescence in situ hybridization results for colon of mice. Probes: EUB338 CY3 hybridized bacterial cells, RI FITC hybridized bacterial cells. The arrow indicates the colonization of RI. (C) Defecation frequency (pellets/h), (D) Fecal water content (%), (E, F) Intestinal propulsion rate (%), (G) whole gut transit time(min), and (H) Bead expulsion time (sec) between NC, LOP and RI gavage group (n = 6 per group). (I) The effect of PBS, loperamide and RI stimulation on the contraction of ex vivo mouse intestinal segments. (J) Max force (g) of mouse colon before and after PBS stimulation (n = 3 per group). (K) Max force (g) of mouse colon before and after loperamide stimulation (n = 3 per group). (L) Max force (g) of mouse colon before and after RI stimulation (n = 3 per group). Representative images of H&E(M) and AB-PAS staining(N) for the colonic tissue in NC, LOP and RI group. Data presented as the mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, no significance by One-way ANOVA with Dunnett’s multiple-comparisons test and Bartlett’s test of equal variances in C, D, F-H. Two-tailed paired t test in J, K; i.g, intragastrical; NC, normal control; LOP, loperamide; RI, Roseburia Intestinalis

Fig. 5.

RI activates the TGF-β1/p-Smad3/transgelin axis in vivo. Relative mRNA expression of (A) Tagln, (B) TGF-β1, (C) TGF-β2, and (D)TGF-β3(n = 6 per group). (E) Representative images of immunohistochemical staining of TGF-β1, p-Smad3, and transgelin for the colonic tissue in NC, LOP and RI group (n = 5 per group). The mean option density (MOD) of (F)TGF-β1, (G) p-Smad3, and (H) transgelin. (I) Experimental design for animal experiment 2 in SPF mice. (J) Representative images of Immunofluorescence co-localization of TGF-β1 and transgelin. Dapi, blue. (K) Representative images of Immsssunofluorescence co-localization of p-Smad3 and transgelin. Dapi, blue. The Relative fluorescent intensity of transgelin (%) (L), TGF-β1 (%) (M), and p-Smad3 (%) (N) (n = 5 per group). Data presented as the mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, no significance by One-way ANOVA with Dunnett’s multiple-comparisons test and Bartlett’s test of equal variances; i.g, intragastrical; i.p, intraperitoneal injection; NC, normal control; LOP, loperamide; RI, Roseburia Intestinalis

Fig. 6.

Small-molecule metabolites are the key mediators through which RI enhances intestinal motility. (A) Experimental design for animal experiment 3. (B) Defecation frequency (pellets/h), (C)Fecal water content (%), (D) Whole gut transit time(min), (E) Bead expulsion time (sec), and (F, G) Intestinal propulsion rate (%) of group in animal experiment 3 (n = 6 per group). (H) The effect of RCM and RI supernatant, RI supernatant (< 3 kDa), and RI supernatant (≥ 3 kDa) stimulation on the contraction of ex vivo mouse intestinal segments. Max force (g) of mouse colon before and after (I) RCM, (J) RI supernatant, (K) RI supernatant (< 3 kDa), and (L) RI supernatant (≥ 3 kDa) stimulation (n = 3 per group). Data presented as the mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, no significance by One-way ANOVA with Dunnett’s multiple-comparisons test and Bartlett’s test of equal variances in B-E, G. Two-tailed paired t test in I-L; i.g, intragastrical; LOP, loperamide; RCM, reinforced clostridial medium; RI-S, Roseburia Intestinalis supernatant

Fig. 8.

Inosine recapitulates RI’s pro-motility effects via the TGF-β1/p-Smad3/Transgelin axis. (A) Experimental design for animal experiment 4. (B)Fecal water content (%), (C) whole gut transit time(min), (D) Bead expulsion time (sec), and (E) Intestinal propulsion rate (%) of group in animal experiment 4 (n = 6 per group). Representative images of H&E(F) and AB-PAS staining(G) for the colonic tissue in LOP, INO-L, INO-H, and INO-H + SB431542 group. (H) Representative images of Immunofluorescence co-localization of TGF-β1 and transgelin. Dapi, blue. (I) Representative images of Immunofluorescence co-localization of p-Smad3 and transgelin. Dapi, blue. The Relative fluorescent intensity of transgelin (%) (J), TGF-β1 (%) (K), and p-Smad3 (%) (L) (n = 5 per group). (M) Representative Western blots of TGF-β1, p-Smad3, and transgelin in colonic tissue from Control, LOP, LOP + RI, and LOP + Inosine groups, each tested with (+) or without (–) the TGF-β receptor inhibitor SB-431,542. GAPDH and total-Smad3 served as the loading control. The protein levels of (N)p-Smad3, (O)TGF-β1, and (P)transgelin (n = 3 per group). Data presented as the mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001, ns, no significance by One-way ANOVA with Dunnett’s multiple-comparisons test and Bartlett’s test of equal variances; i.g, intragastrical; i.p, intraperitoneal injection; LOP, loperamide; INO, inosine; RI, Roseburia Intestinalis

Microbial FISH analysis

5-µm paraffin transversal sections were prepared. Oligonucleotide probes targeting the 16S rRNA were obtained from ProbeBase (http://www.microbia-ecology.net/probebase/ ). The “universal bacterial” probe, EUB338 (FITC-labeled), with the sequence 5’-GCTGCCTCCCGTAGGAGT-3’, was used for general bacterial detection. For RI detection, probes were designed by retrieving the 16 S rRNA sequences of closely related bacteria or strains from the SILVA database (https://www.arb-silva.de/search/). The sequences were compared and analyzed to identify conserved regions, and probes were designed based on the reverse complementary principle with a melting temperature (Tm) range of 55–65 °C. Specificity was confirmed by aligning the probe sequences with the NCBI database. The RI-specific probe sequence 5’- GCTTACCCGCTGGCTACT-3’ was synthesized and labeled with Cy3. Fluorescent microscopy (Olympus, Japan) was used to analyze the samples.

Histology and immunohistochemistry

Colon tissues were collected immediately after euthanasia, fixed in 4% (w/v) paraformaldehyde in buffered saline, paraffin-embedded, and sectioned at 5 μm. For routine histology, sections were deparaffinized, rehydrated, and stained with hematoxylin and eosin (H&E) according to standard procedures, then examined under a light microscope to assess epithelial and mucosal architecture as well as inflammatory changes. Alcian blue–periodic acid–Schiff (AB-PAS) was used for mucin visualization. Briefly, deparaffinized and rehydrated sections were sequentially processed with alcian blue and periodic acid–Schiff reagents, counterstained with hematoxylin, and evaluated for goblet-cell mucin content and distribution. For immunohistochemistry (IHC), paraffin sections were deparaffinized and rehydrated, subjected to heat-induced (microwave) antigen retrieval, and incubated overnight at 4 °C with primary antibodies against TGF-β1, p-Smad3, or transgelin. After washing, sections were incubated with appropriate secondary antibodies followed by streptavidin–horseradish peroxidase, developed with diaminobenzidine, counterstained with hematoxylin, and examined by light microscopy. Quantification of immunohistochemistry staining was carried out using ImageJ software (USA). After background correction and color separation using the Color Deconvolution function, the mean optical density (MOD) was measured from at least five randomly selected, non-overlapping high-power fields (HPFs) per sample. The MOD values were used as a semi-quantitative indicator of protein expression intensity.

Immunofluorescence

Colon tissue was fixed in 4% (w/v) paraformaldehyde in buffered saline, paraffin-embedded, and sectioned at 5 μm. Sections were incubated in blocking buffer and then with primary antibodies against transgelin, TGF-β1, and p-Smad3 overnight at 4 °C. The next day, sections were incubated with Alexa Fluor 488- or 594-conjugated secondary antibodies for immunofluorescence detection, and nuclei were counterstained with DAPI for 3 min. Images were acquired using a laser-scanning confocal microscope (Olympus, Japan).

Fecal assesement

To evaluate fecal parameters, mice were individually housed in metabolic cages one hour after gavage. During a one-hour period from 10:00 AM to 11:00 AM, both food and water were withheld. All fecal pellets excreted by each mouse were gathered, counted, and their wet weight was measured immediately. Subsequently, the pellets were dehydrated in an oven at 100 °C for a minimum of one hour to determine their dry weight. The fecal water content percentage was then calculated as: Fecal water content (%) = [(wet weight − dry weight) / wet weight] × 100%.

Measurement of gastrointestinal transit time (GITT) and intestinal propulsion rate (IPR)

We measured GITT and IPR following a previously described protocol [24]. For GITT, mice were fasted for 16 h and then given an oral gavage of 0.1 mL of 1.5% methylcellulose solution (Solarbio, China) containing 5% Evans blue (Solarbio, China). The time elapsed between the gavage and the excretion of the first blue-stained fecal pellet was recorded as the GITT. To measure IPR, animals were euthanized 30 min after receiving the Evans blue solution. The distance the blue marker traveled from the pylorus to the cecum was measured. The IPR was calculated as the percentage of the total small intestine length covered by the dye: IPR (%) = (distance traveled by Evans blue / total length of the small intestine) × 100%.

Measurement of colonic motility

In vivo colonic motility was evaluated using a bead expulsion test, adapted from the Previously reported method [25]. After a 12-hour fasting period, a single 3 mm glass bead was gently inserted 2 cm into the distal colon of each mouse. The time required for each animal to expel the bead was recorded as a measure of colonic transit.

Ex vivo organ bath

To assess colonic contractility ex vivo, we employed a modified organ bath protocol based on previously reported protocols [26]. Colon segments from euthanized mice were mounted longitudinally in an organ bath containing Krebs-Ringer solution (Giboco, USA), maintained at 37 °C and continuously oxygenated. The tissue was attached between a force transducer and a fixed hook. After a 15-minute equilibration period under a preload of approximately 0.5 g, spontaneous contractions were recorded using LabChart software. Subsequently, the colon segments were respectively exposed to 200 µL of LOP (1 mg/mL), RI (1 × 10⁹ CFU), RCM edium (2% v/v), RI supernatant (2% v/v), and inosine (30 mg/mL), with a 5-minute stabilization period following each addition. Contraction curves were observed and recorded. For experiments involving live RI stimulation, the colon tissue was incised longitudinally prior to mounting.

Statistical methods

Data are presented as mean ± SEM from at least three independent experiments. Statistical significance was set at P < 0.05. For comparisons between two groups with normally distributed data, a two-tailed unpaired or paired t-test was used, while the nonparametric Mann–Whitney U test was applied for non-normally distributed data. One-way ANOVA was utilized to compare differences among multiple groups. The correlation was assessed using Spearman correlation analysis. Data processing and visualization were performed using R software (version 4.3) and GraphPad Prism 9.0. MR analysis was conducted utilizing the R packages “TwoSampleMR”. Additionally, “MR_PRESSO” was utilized for multiplicity tests [27].

Results

Bidirectional and UVMR analyses of GM and constipation

Initially, 2,028 SNPs linked to GM were detected as potential IVs through large-scale GWAS following the removal of palindromic SNPs (Table S3). As mentioned in Table S4, 32 SNPs were correlated with constipation in the FinnGen database. Significantly, all IVs displayed F-statistics > 10, indicating the absence of weak instrumental bias in the present research. After conducting heterogeneity and pleiotropy tests utilizing Cochrane’s Q test and MR-Egger test, a total of 11 causal correlations were identified from GM characteristics (7 genera, 1 family, 1 order, and 2 class) to constipation traits via the IVW method (Fig. 1B, C and Table S5). Notably, the genetically predicted genus Roseburia (95%CI: 0.766–0.982), genus Haemophilus (95%CI: 0.876–0.998), genus Eubacterium rectale group (95%CI: 0.800–0.997), genus Eubacterium brachy group (95%CI: 0.899–0.991), genus Coprococcus2 (95%CI: 0.747–0.994), order Methanobacteriales (95%CI: 0.903–0.999), family Methanobacteriaceae (95%CI: 0.903–0.999), class Methanobacteria (95%CI: 0.903–0.999), and class Betaproteobacteria (95%CI: 0.817–0.990) were linked to a decreased risk of constipation (Fig. 1C). Conversely, genus Victivallis (95%CI: 1.003–1.101) and genus Holdemanella (95%CI: 1.002–1.121) were related to a heightened constipation risk (Fig. 1C). Following this, a reverse analysis was performed, revealing no evidence of a causal impact of constipation on the aforementioned GM (Table S6).

Reduced RI abundance in FC patients and its negative correlation with disease severity

We performed 16S rRNA gene sequencing on fecal samples from patients with FC and healthy individuals to characterize gut-microbiota alterations. PCoA demonstrated a clear segregation between patients with FC and controls, indicating a marked shift in overall community structure (Fig. 2A). Hierarchical clustering of the 30 most abundant genera underscored this dysbiosis: FC samples were enriched for opportunistic or facultative taxa—including Limosilactobacillus, Lawsonibacter, Dysosmobacter, and several Bacillus/Bacillaceae lineages, whereas healthy samples retained a predominance of fiber-fermenting, butyrate-producing genera such as Faecalibacterium, Roseburia, and Blautia (Fig. 2B). Consistently, LEfSe analysis (LDA > 3.0) confirmed over-representation of these taxa in FC (Fig. 2C, Supplementary Fig. 1A). Specifically, facultative, lactate-producing Bacillaceae and other opportunistic taxa dominated the FC microbiota, whereas key butyrate-producing Oscillospirales/Lachnospirales lineages were markedly depleted (Fig. 2C). Cross-referencing the differential GM with MR results pinpointed Roseburia as the sole overlapping genus (Fig. 2D). Notably, Roseburia abundance was significantly reduced in FC patients and displayed a protective effect of constipation, suggesting a protective role for RI in maintaining colonic motility (Figs. 1C and 2C).

Fig. 2.

Reduced RI abundance in FC patients and its negative correlation with disease severity. (A) Principal-coordinates analysis of unweighted UniFrac distances for constipation and healthy control fecal microbiota. (B) Heat-map with hierarchical clustering of the 30 most abundant genera in constipation and control samples. (C) LEfSe bar plot (LDA > 3) depicting differential bacterial taxa between constipation and control groups. (D) Venn diagram showing the overlap between 16 S-identified genera and Mendelian-randomization hits. (E) Representative images of fluorescence in situ hybridization results for colon of constipation patients and control. Probes: EUB338 CY3 hybridized bacterial cells, RI FITC hybridized bacterial cells. The arrow indicates the colonization of RI. (F) Relative abundance of RI in fecal samples measured by qPCR (n = 30 per group). Data presented as the mean ± SEM. ****P < 0.0001 by unpaired t test comparison of variances. Spearman correlation between fecal RI relative abundance and PAC-SYM score (G), PAC-QOL score (H) and BSS score (I) in functional constipation patients (n = 30). RI, Roseburia Intestinalis; PAC-SYM, the Patient assessment of constipation symptom; PAC-QOL, the Patient assessment of quality of life; BSS, Bristol stool scale

To further investigate changes in RI abundance and its clinical relevance in FC patients, we enrolled 30 patients with confirmed FC and 30 age- and sex-matched healthy controls. Clinical characteristics of all participants are detailed in Table 1. Colonic mucosal colonization of RI was evaluated by microbial FISH, revealing predominant localization of RI to the mucosal layer and a markedly lower detection rate in FC patients versus controls (Fig. 2E). qPCR of fecal samples confirmed a significant reduction in RI levels in FC patients, corroborating our prior 16S rRNA sequencing results (Fig. 2F). Constipation severity and its impact on quality of life were assessed using the PAC-SYM and PAC-QOL questionnaires, respectively; higher scores indicate worse symptoms and poorer quality of life (Fig. 2G, H). Stool form and transit rate were measured by the BSS score, where higher scores reflect looser stools and faster transit (Fig. 2I). Compared with healthy controls, FC patients exhibited significantly higher PAC-SYM (P < 0.0001) and PAC-QOL (P < 0.0001) scores, and significantly lower BSS scores (P < 0.0001) (Table 1). Spearman correlation analysis demonstrated that fecal RI abundance was negatively correlated with PAC-SYM (r = -0.545, P = 0.0018) and PAC-QOL (r = -0.4636, P = 0.0099) scores, and positively correlated with BSS score (r = 0.3880, P = 0.0341). These findings suggest that RI reduction is associated with more severe constipation and that RI may play a protective role in preserving gut motility.

Table 1.

Clinical characteristics of the research subjects

Variables	Control (n = 30)	Functional Constipation (n = 30)	Test statistic	P
Age	47.23 ± 2.207	49.93 ± 2.019	t = 0.9026	P = 0.3705
Sex[n(%)]			χ2 = 0.0667	P = 0.7961
Female	15(50)	14(46.7)
Male	15(50)	16(53.3)
PAC-SYM score	2(1,3)	18(15.75,21.25)	U = 0	P<0.0001
PAC-QOL score	13.7 ± 0.824	58.27 ± 2.627	t = 16.19	P<0.0001
BSS score	4(4,5)	2(1,2)	U = 12.5	P<0.0001

Continuous variables were first assessed for normality using the Shapiro–Wilk test. Age and PAC-QOL scores, which were normally distributed, are presented as Mean ± SEM and compared by independent-samples t-test. PAC-SYM and BSS scores, which deviated from normality, are reported as Median (P25, P75) and compared by Mann–Whitney U test. Categorical variables (sex) were compared using the χ² test

RI improves intestinal motility in mice model of constipation

To determine whether RI improves intestinal motility, we established a constipation model in SPF mice. Following a 7-day adaption fed, mice received daily gavage with loperamide for 14 days (LOP group), and with RI gavage after 6 h (RI group) (Fig. 3A). Successful colonization of RI within the colonic mucosa and submucosa was verified by microbial FISH (Fig. 3B). Functionally, RI-treated mice exhibited significantly increased defecation frequency, fecal water content, and small-intestinal transit compared with the LOP group (Fig. 3C-F). Furthermore, whole-gut transit time assessed by evans blue was markedly shortened by RI treatment (Fig. 3G), and bead expulsion time—an indicator of colonic transit—was also significantly reduced (Fig. 3H). To investigate contractile physiology directly, we performed ex vivo organ-bath recordings. Loperamide markedly suppressed spontaneous contractions and maximal contractile force of colonic segments (Fig. 3I, K). In contrast, acute exposure to PBS and live RI did not influence augment contractility ex vivo (Fig. 3I, J, L), suggesting that RI likely promotes motility in vivo via metabolites or secreted factors rather than by direct contact with the muscle layer.

Given that constipation can alter colonic histology and mucus secretion [28], we assessed tissue morphology with H&E staining and goblet-cell mucin by AB-PAS staining. Relative to NC group, LOP mice displayed mucosal thinning and crypt shortening, both of which were rescued by RI intervention (Fig. 3M). AB-PAS staining further showed reduced mucus in the LOP group, whereas NC and RI groups exhibited comparable mucin content (Fig. 3N). Collectively, these data demonstrate that RI successfully colonizes the murine colon and significantly improves intestinal motility while ameliorating constipation-associated mucosal injury.

RNA sequencing analysis reveals potential mechanisms of RI-induced improvements intestinal transit

To define global transcriptional changes associated with RI-induced improvements in intestinal transit, we performed RNA-seq on colonic tissues from loperamide-treated mice with or without RI gavage (n = 3 per group). Principal component analysis (PCA) demonstrated clear separation between groups, indicating robust, treatment-driven transcriptomic remodeling (Fig. 4A). Differential expression analysis (Padjust < 0.05; |log₂FC| ≥ 1) identified 1,406 DEGs—387 upregulated and 1,019 downregulated in the RI group relative to LOP. Notably, Tagln (transgelin)—a gene critical for contractile cytoskeletal organization—was markedly upregulated in RI-treated colons (Fig. 4B). Hierarchical clustering coupled with pathway term mapping highlighted coherent gene modules that distinguished RI from LOP (Fig. 4C). Specifically, RI-shifted genes were enriched for calcium signaling, tight junction, cell adhesion molecules, and Wnt signaling (Fig. 4C). Consistent with these patterns, GO analysis revealed enrichment for epithelial/epidermal programs—including epithelium/epidermis development, keratinization, cornification, and epithelial cell differentiation—together with Wnt-mediated cell–cell signaling (Fig. 4D), suggesting that RI may promote epithelial barrier differentiation/remodeling to protect barrier integrity and support motility. KEGG analysis further highlighted TGF-β signaling, Wnt signaling, and calcium signaling as central axes, alongside cytokine–receptor interaction, TRP channels, and several neuro-endocrine regulatory pathways (e.g., serotonergic synapse, adrenergic signaling) (Fig. 4E), collectively aligning with pathways linked to epithelial barrier regulation and contractile/cytoskeletal function. Building on prior evidence that Tagln is a newly identified, direct downstream target of canonical TGF-β/Smad3 signaling [29], and together with our RNA-seq findings, we hypothesize that RI may improve intestinal transit by regulating the TGF-β pathway to upregulate transgelin, thereby facilitating intestinal contraction and enhancing gut motility.

Fig. 4.

RNA sequencing analysis reveals potential mechanisms of RI-induced improvements intestinal transit. (A) The PCoA of differential proteins based on the proteomics. Blue: RI group; Orange: LOP group. (B) The volcano plot about differentially expressed genes (DEGs) of RI and LOP group. Red: upregulated; Blue: downregulated. (C) Hierarchical clustering of differential proteins based on proteomics. Red: upregulated; Blue: downregulated. (D) GO analysis reveals the diversity of molecular biological processes, cellular components, and molecular functions. (E) KEGG analysis shows the significantly enriched pathways. LOP, loperamide; RI, Roseburia Intestinalis

RI activates the TGF-β1/p-Smad3/Transgelin axis in vivo

To validate the mechanism of RI action, we analyzed colonic tissues from the three groups in animal experiment 1. qPCR revealed that Tagln and TGF-β1 were markedly increased in the RI group compared with the LOP group, whereas TGF-β2 and TGF-β3 showed no apparent differences (Fig. 5A–D). Using immunohistochemistry on serial adjacent sections from the same specimens across all three groups, we observed increased expression of TGF-β1, p-Smad3, and transgelin in RI-treated colons (Fig. 5E–H). To further confirm pathway dependence, animal experiment 2 included an additional group that received intraperitoneal SB431542 (a TGF-β receptor inhibitor) on top of RI gavage (Fig. 5I). Under SB431542, the RI-induced upregulation of transgelin was abolished, indicating that RI regulates transgelin via the TGF-β1/p-Smad3 pathway in vivo (Fig. 5J-M).

Small-molecule metabolites are the key mediators through which RI enhances intestinal motility

To identify the active components of RI responsible for its pro-motility effects, we compared constipation mice gavaged with RI supernatant versus sterilized culture medium (RCM) (Fig. 6A) and, using a 3-kDa MWCO ultrafiltration, separated the supernatant into a < 3-kDa fraction (low-molecular-weight metabolites) and a ≥ 3-kDa fraction (proteins/large peptides) for parallel testing (Fig. 6A). RCM itself had no effect on intestinal motility, whereas mice receiving RI supernatant or the < 3-kDa fraction showed significantly increased defecation frequency and fecal water content (Fig. 6B, C) and shortened whole-intestinal transit time (Fig. 6D). Consistently, bead expulsion time was markedly reduced (Fig. 6E), and the small-intestinal propulsion rate was significantly enhanced (Fig. 6F, G) in both the RI supernatant and < 3-kDa groups. In contrast, the ≥ 3-kDa fraction failed to improve motility in any assay. Ex vivo organ-bath recordings reinforced these findings: both RI supernatant and the < 3-kDa fraction increased spontaneous contractions and maximal contractile force of isolated intestinal segments, whereas RCM and the ≥ 3-kDa fraction were inert (Fig. 6H-L). Together, these results indicate that RI improves intestinal motility through non-proteinaceous, small-molecule metabolites rather than protein components.

Comparative metabolomics identifies inosine as a candidate effector metabolite in RI supernatant

Building on prior results, we hypothesized that metabolites in the RI supernatant mediate its pro-motility effects. To pinpoint candidates, we performed comparative metabolomics of RI supernatant versus RCM medium. Metabolite profiling revealed significant differences between the RI supernatant and RCM medium, highlighting a distinct metabolic signature for the RI supernatant (Supplementary Fig. 2A, B). Figure 7A shows the 30 top-ranked metabolites based on variable importance in projection (VIP) scores through OPLS-DA analysis with thresholds FDR < 0.05. Beyond the expected enrichment of butyrate, we found multiple purine-related metabolites were differentially abundant: adenosine and adenine were decreased, whereas inosine and its downstream product hypoxanthine were increased in the RI supernatant. What’s more, KEGG pathway analysis indicated enrichment of the purine metabolism pathway (Fig. 7B). Within this pathway, adenosine is deaminated to inosine by adenosine deaminase (ADA), and inosine can be converted to hypoxanthine by purine nucleoside phosphorylase (PNP)—a pattern consistent with our observed directions of change and nominating inosine as a candidate RI-derived effector metabolite.

Fig. 7.

Comparative metabolomics identifies inosine as a candidate effector metabolite in RI supernatant. (A) Matchstick analysis of the top 30 metabolites ranked by VIP in the RI supernatant and RCM groups (Padjust < 0.05). Red: downregulated metabolites; Blue: upregulated metabolites. (B) KEGG analysis shows the significantly enriched pathways. Matchstick analysis of the metabolites ranked by VIP in the (C) feces and (D) peripheral blood samples of constipation patients and control (P < 0.05, |log₂FC| ≥ 0.5, VIP > 1) Red: downregulated metabolites; Blue: upregulated metabolites. (E) Spearman correlations between selected differential expressed genes and metabolites. *P < 0.05; **P < 0.01; ***P < 0.001, ****P < 0.0001. RCM, reinforced clostridial medium; RI, Roseburia Intestinalis

We next performed untargeted metabolomics on fecal and peripheral blood samples from constipation patients (thresholds: P < 0.05, |log₂FC| ≥ 0.5, VIP > 1). We finally identified 40 and 32 differential metabolites in feces and blood, respectively (Fig. 7C, D). In feces, inosine was significantly reduced in constipation patients, accompanied by increased adenine and shifts in additional purine derivatives (e.g., 2-aminopurine), a pattern that converges with the RI supernatant findings on purine metabolism (Fig. 7C). In blood, while inosine and adenosine were not significantly altered, hypoxanthine was decreased, and allopurinol—a xanthine oxidase inhibitor and hypoxanthine analog were also reduced (Fig. 7D). Collectively, these data indicate a purine-metabolism imbalance in constipation and, together with the supernatant profile, support inosine as a key RI-derived small-molecule metabolite potentially underlying the pro-motility effect.

Integrating these metabolomic findings with transcriptomic KEGG enrichment, we selected DEGs mapped to the TGF-β signaling pathway together with the top 30 VIP-ranked differential metabolites for Spearman correlation analysis (Fig. 7E). Inosine, hypoxanthine, and adenosine each correlated with multiple pathway genes; specifically, inosine showed a negative correlation with Bmp8a and a positive correlation with Dcn. Notably, inosine and hypoxanthine displayed significant positive correlations with Tagln. These associations support the hypothesis that inosine may serve as an RI-derived effector engaging the TGF-β axis to promote transgelin expression and enhance intestinal motility, which motivated our subsequent experiments to test the role of inosine in motility regulation.

Inosine recapitulates RI’s pro-motility effects via the TGF-β1/p-Smad3/Transgelin axis

To determine whether inosine is a key metabolite mediating RI-induced effects, we evaluated its dose-dependent impact and mechanism in loperamide-treated mice (Fig. 8A). Guided by prior studies, we tested a low dose (100 mg/kg) and a high dose (300 mg/kg) to assess dose responsiveness [30–32]. Compared with the LOP group, mice gavaged with high-dose inosine showed increased defecation frequency, fecal water content, and intestinal propulsion rate, along with shortened whole-gut transit time and bead expulsion time (Supplementary Fig. 3A, B; Fig. 8B–E). In the low-dose inosine group, no appreciable effects were observed on fecal water content or intestinal propulsion rate versus LOP (Fig. 8B, E). Across endpoints, the high-dose group outperformed the low-dose group, indicating a dose-dependent trend. Ex vivo organ-bath recordings further showed that inosine enhanced spontaneous contractions and increased maximal contractile force of isolated intestinal segments (Supplementary Fig. 3C, D). In contrast, co-administration of the TGF-β receptor inhibitor SB431542 abolished the inosine-associated improvements (Supplementary Fig. 3A, B; Fig. 8B–E).

Histologically, mice gavaged with low-dose and high-dose inosine both improved mucosal thickness, crypt depth, and goblet-cell number/mucus secretion, with more pronounced changes in the high-dose group (Fig. 8F, G). Mechanistically, qPCR showed marked increases in Tagln and TGF-β1 following inosine treatment (Supplementary Fig. 3E, F). Immunofluorescence further demonstrated higher TGF-β1, p-Smad3, and transgelin in the high-dose inosine group than in the constipation group, with levels exceeding those in the low-dose group (Fig. 8H–L). Notably, intraperitoneal SB431542 eliminated both the histologic rescue and activation of the TGF-β1/p-Smad3/transgelin axis, along with motility benefits disappear (Supplementary Fig. 3E, F; Fig. 8F–L). Finally, Western blotting further confirmed that RI and inosine elevated colonic TGF-β1, p-Smad3, and transgelin protein levels, whereas SB431542 blocked these effects (Fig. 8M–P). Collectively, these findings support inosine as a RI-derived metabolite that improves intestinal motility through activation of the TGF-β1/p-Smad3/Transgelin pathway (Supplementary Fig. 4).

To further explore the potential receptor mediating inosine’s effects, we conducted animal experiment 5 (Supplementary Fig. 5A), in which loperamide-treated mice were randomized to receive high-dose inosine alone or in combination with the selective A2A receptor antagonist ZM241385. Immunofluorescence analysis showed that co-administration of ZM241385 did not abolish the inosine-induced increases in TGF-β1, p-Smad3, and transgelin expression (Supplementary Fig. 5B–F). These findings suggest that the pro-motility action of inosine in this model might not be mediated through the A2A receptor.

Discussion

The pathophysiology of FC is complex and multifactorial, involving dietary habits, colonic motor function, and psychological factors, with effective therapeutic options remaining limited [33]. Numerous evidence has implicated gut microbiota dysbiosis in FC, but most were associative and did not establish causality or delineate mechanisms. Our study addressed this gap and established a causal, protective role for RI in maintaining intestinal motility. We also showed that RI, via its purine metabolite inosine, alleviated constipation by activating a novel pro-motility TGF-β1/p-Smad3/Transgelin cascade. These findings not only uncover a new therapeutic axis for FC but also, for the first time, propose inosine as a potential drug for treating constipation, providing further insights into microbiota–host interactions.

A primary result of our study is the establishment of a causal relationship between the gut microbiota and constipation risk, rather than the prior associative observations. Using bidirectional MR across 196 gut taxa from the MiBioGen consortium, we found strong evidence that 11 taxa exert a causal effect on constipation risk without reverse causation. Victivallis and Holdemanella were linked to a higher risk of constipation. Consistent with our findings, prior work reported higher fecal Holdemanella in FC than in controls, with a decline after successful lactulose therapy [11]. Although Victivallis has not been directly studied in constipation, its abundance is reduced in patients with diarrhea-predominant IBS (IBS-D) [34]. Among the remaining taxa showing protective associations, the Eubacterium rectale group, Eubacterium brachy group, and Coprococcus 2 align with their established roles as key butyrate producers, a metabolite known to support colonic health and motility [35, 36]. More intriguingly, we identified protective associations for Haemophilus and for the class Betaproteobacteria, even though the phylum Proteobacteria is often considered a marker of dysbiosis [37, 38]. Such discrepancies highlight the value of MR in revealing causal pathways that can challenge conventional associations.

Integrating the MR results with fecal 16S rRNA profiles from patients with FC, Roseburia emerged as the sole overlapping taxon: its abundance was significantly reduced in FC while MR indicated a protective causal effect on constipation risk. This genetic-level causal inference was strongly corroborated by our clinical and preclinical data. As a core butyrate-producing, anti-inflammatory genus, Roseburia is regarded as a key indicator of intestinal health [39]. Multiple studies have reported significantly lower Roseburia levels in FC—most notably in slow-transit constipation (STC) [8, 40]. Moving beyond descriptive associations, interventional evidence provides functional support: in a psyllium supplementation trial in constipated patients, increases in Roseburia tracked with higher stool water content [41]. Crucially, in our cohort the abundance of RI was inversely associated with constipation symptom severity and quality-of-life impairment. This tight linkage between bacterial abundance and clinical phenotype underscores the relevance of our findings to human disease. In constipation mice models, gavage with live RI rescued the constipation phenotype, significantly improving defecation frequency, fecal water content, and whole-gut transit. Beyond constipation, reduced abundance of RI has also been consistently documented in inflammatory bowel diseases (IBD), where it correlates with epithelial barrier disruption, mucosal inflammation, and diarrhea [15]. Experimental models further demonstrate that administration of RI ameliorates colitis by modulating intestinal barrier, microbiome, and inflammatory responses [42, 43] .These findings suggest that RI exerts dual functions in gut health: enhancing motility and relieving constipation on the one hand, and providing anti-inflammatory and barrier-protective benefits in diarrheal disease contexts such as IBD on the other.

Although the Roseburia are renowned as prolific producers of the short-chain fatty acid butyrate—widely recognized for strengthening the epithelial barrier and dampening mucosal inflammation [44]—their effects on gut motility remain disputed [45]. Our investigation reveals a novel, butyrate-independent mechanism of action. Through a combination of untargeted metabolomics on the RI supernatant and fecal samples from FC patients, we identified the purine nucleoside inosine as a key candidate effector. Inosine was significantly enriched in the RI supernatant and was depleted in the feces of FC patients, mirroring the depletion of the bacterium itself. This discovery situates our work within the broader field of purinergic signaling, which plays a fundamental role in regulating gastrointestinal function, including motility, secretion, inflammation and neuro-inflammation [46]. While ATP and adenosine are the most studied purinergic signaling molecules, inosine, a stable metabolite of adenosine, is emerging as a critical bioactive compound with potent immunomodulatory and tissue-protective effects [47, 48]. Recent studies have shown that intestinal B. pseudolongum modulated enhanced immunotherapy response through production of the metabolite inosine [31]. Additional studies have linked gut microbiota-derived inosine to the attenuation of colitis, often acting through the adenosine A2A receptor (A2AR) or PPARγ signaling to bolster mucosal barrier function [49]. The same work also reported that exogenous inosine increases stool frequency and shortens gastrointestinal transit in animal models, indicating a pro-motility effect [49]. Our study expands the functional repertoire of inosine, demonstrating for the first time that its administration alone is sufficient to recapitulate the pro-motility effects of RI, rescuing the constipation phenotype in our mouse model in a dose-dependent manner. This identifies inosine as a novel “postbiotic” with therapeutic potential for motility disorders. Given that bioactivity resided exclusively in the < 3 kDa ultrafiltrate—a widely used operational cutoff to separate small metabolites from macromolecules, we interpret the active factor as a low–molecular–weight metabolite rather than a protein or peptide. At the same time, we acknowledge that this approach does not definitively exclude bioactive oligopeptides < 3 kDa; therefore, orthogonal confirmation will be pursued in future work, including optimized heat-inactivation series and proteinase K digestion assays to more conclusively verify the molecular nature of the active component. Colonic RNA-seq in RI-treated mice pointed to the enrichment of the TGF-β pathway, a finding we validated at the protein level by demonstrating that both RI and inosine administration led to increased levels of TGF-β1 and p-Smad3, the canonical downstream effector. Thus, TGF-β signaling emerges as a positive regulator of intestinal motility in our study. This discovery is inconsistent with some published research, as TGF-β signaling in the gut is overwhelmingly associated with pathology [50]. It is known to promote invasion and metastasis in later stages of cancer [51] and, due to its pro-fibrotic effects, can lead to intestinal wall thickening, stiffness, and motility disorders, seemingly contradicting our results [52]. However, these pathological conditions typically arise from strong, prolonged activation of the pathway in disease states. In contrast, FC is not inherently characterized by strong inflammation. Therefore, the signaling we observe likely represents a physiological regulatory mechanism for maintaining homeostasis, which is fundamentally different from the chronic, high-level TGF-β signaling that drives pathological remodeling in inflammation and cancer. Moreover, our data show that the RI-inosine axis targets colonic smooth muscle cells to regulate the expression of the contractile protein transgelin, which acts as a key downstream target of TGF-β to promote intestinal motility [4, 53]. Furthermore, TGF-β signaling is known to be critical for the development and maintenance of the ENS, the primary regulator of gut motility [54]. While not directly tested, it remains possible that the RI-inosine-TGF-β1 axis also acts on enteric neurons, representing a parallel pathway for motility regulation. Our findings thus do not refute the established pathological roles of TGF-β but rather uncover a novel, beneficial, and physiological function in the direct regulation of intestinal contractility.

We identified transgelin, an actin-binding protein, as the terminal effector linking a microbial cue to the host contractile apparatus. Transgelin’s core function is to organize the actin cytoskeleton and maintain the contractile phenotype [4, 55].Our prior work has showed that deglycosylated azithromycin (Deg-AZM) act through transgelin to increase F-actin/G-actin and stress fibers, thereby enhancing smooth muscle contractility and peristalsis, and positioned transgelin as a target for slow-transit constipation [4]. Previous work has established that Tagln expression can be regulated by TGF-β [53]. Smad3 can bind rapidly and specifically to the Tagln promoter, identifying transgelin as a direct TGF-β/Smad3 target that drives epithelial cell migration in lung fibrosis [29]. In addition, TGF-β3 upregulates multiple contractile genes—including Tagln—and thereby enhances the contractile function of human intestinal smooth muscle cells (SMCs), implicating TGF-β–dependent control of transgelin in the maintenance of gut motility [56]. Our study is the first to demonstrate that this regulation can be initiated by a microbial metabolite to restore gut function. The definitive role of this pathway was confirmed by our inhibitor experiments, where administration of the TGF-β receptor I inhibitor SB431542 completely abolished the upregulation of transgelin and the pro-motility benefits conferred by both RI and inosine. Although inosine is not a known ligand for TGF-β receptors, it likely signals through its own receptor like A2AR or A2B on intestinal epithelial or immune cells, inducing active TGF-β1 production [57, 58]. Our additional in vivo experiment showed that A2A receptor blockade with ZM241385 did not abolish inosine-induced activation of the TGF-β1/p-Smad3/transgelin pathway, suggesting that A2A receptor is unlikely to mediate this effect. Interestingly, prior studies have indicated that A2B receptor antagonism can interfere with TGF-β/Smad2/3 and NF-κB signaling in podocytes and protect against phenotypical transformation in diabetic glomerulopathy [59], raising the possibility that A2B receptor may contribute despite its lower affinity for inosine. What’s more, inosine also possess neurotrophic and neuroprotective properties, and enteric neurons are also a source of TGF-β1 [60]. Secreted TGF-β1 likely acted in a paracrine manner on colonic SMCs to activate Smad3, increase transgelin, and augment contractility. This model represents an indirect, intercellular crosstalk between purinergic and TGF-β signaling. However, it remains a hypothesis and further studies are required to delineate the molecular intermediates that mediate this crosstalk in the gut.

The primary strength of this study lies in its multimodal, integrated design—spanning causal genetic inference, clinical relevance, and molecular validation—which established a protective role for RI in intestinal motility. Multi-omics delineated a coherent mechanism, identified inosine as the key bacterial effector, and revealed a novel, seemingly paradoxical, pro-motility TGF-β1/p-Smad3/transgelin cascade. The convergence of evidence from MR, human cohort studies, animal models, multi-omics, and targeted molecular biology provides a robust foundation for our conclusions. Nevertheless, we acknowledge several limitations that also chart important directions for future work. First, our MR analyses relied on GWAS summary statistics derived predominantly from individuals of European ancestry, which may limit the generalizability of the causal inferences to other populations. Future replication across PanUK, EMBL-EBI, and other reliable GWAS datasets will be necessary to confirm generalizability. Moreover, while our MR analysis identified the Roseburia genus, our subsequent experiments focused on RI. It is plausible that other species within this genus also contribute to the observed protective effect. Third, the modest sample size represents a constraint on the broader applicability of our conclusions, and larger, well-powered cohorts will be essential for confirmation. While the consistency of key outcomes provided reassurance, concerns about external validity remain, particularly with regard to transcriptomic and biopsy-based analyses. Moving forward, we intend to expand enrollment across multiple centers, adopt standardized protocols for tissue acquisition, and develop complementary fecal and mucosal biomarker assays to enable validation in larger and less invasive populations. Fourth, although our data support a coherent mechanism, the pathway has not yet been fully validated in gene-knockout animal models or in-vitro systems, leaving certain molecular details unresolved. Although we attempted to culture primary colonic primary colonic smooth muscle cells (SMCs), their fragility, low survival, and poor transfection efficiency—particularly after pharmacologic stimulation—precluded reliable readouts. Future research should incorporate tools such as gene knockout mice and human or commercial colonic SMCs to further elucidate the underlying mechanisms. Finally, although identifying inosine as an activator of TGF-β1 signaling represented a major advance, further evaluation is needed to explore the specific cell-surface receptor that initially binds inosine to trigger TGF-β1 production.

Conclusion

In summary, this study established a causal protective role for RI in functional constipation by integrating bidirectional MR analysis with 16S rRNA sequencing results. We subsequently uncovered the novel signaling pathway responsible for this effect, demonstrating that RI and its derived purine metabolite, inosine, promote intestinal motility by activating the TGF-β1/p-Smad3/Transgelin axis in colonic smooth muscle. These findings clarified the roles of inosine and TGF-β signaling in gut host-microbe communication, providing a new mechanistic framework to develop probiotics and postbiotic therapies for functional constipation.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

FC

Functional constipation

ENS

Enteric nervous system

MR

Mendelian randomization

GWAS

Genome-wide association studies

CRC

Colorectal cancer

IVs

Instrumental variables

UVMR

Univariate Mendelian randomization

IVW

Inverse variance weighting

ORs

Odds ratios

CIs

Confidence intervals

IBS

Irritable bowel syndrome

PAC-SYM

Patient assessment of constipation symptoms

BSS

Bristol Stool Scale

PAC-QOL

Patient assessment of constipation quality of life

ATCC

American Type Culture Collection

MWCO

Molecular weight cut-off

SPF

Specific pathogen-free

H&E

Hematoxylin and eosin

AB-PAS

Alcian blue–periodic acid–Schiff

GITT

Gastrointestinal transit time

IPR

Intestinal propulsion rate

RCM

Modified reinforced clostridial medium

ADA

Adenosine deaminase

PNP

Purine nucleoside phosphorylase

SMCs

Smooth muscle cells

Author contributions

Conceptualization XD, HR, BW, WZ. Data curation XD, HZ, HL. Formal analysis XD, HZ, MC, MS. Methodology XD, HZ, HL, MS. Software XD, HL, MC. Validation HZ, MC. Funding acquisition KJ, BW. Supervision KJ, HR, BW, WZ. Writing original draft XD, HZ, MC, MS. Revision for critical important content XD, MS, KJ, HR, BW, WZ. All authors read and approved the final manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (Grant nos. 8217032098, 82370545), Traditional Chinese Medicine and Integrated Traditional Chinese and Western Medicine Research Project from the Tianjin Municipal Health Commission(Grant nos. 2025185), Tianjin Health Research Project (Grant nos. TJWJ2024XK004), and National key research and development program (2022YFC2504004).

Data availability

All the data supporting the conclusions of this article are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This human study was approved by Medical Ethics Committee of the Tianjin Medical University General Hospital (Ethical No. IRB2024-YX-560-01). All participants know the aim of this study and the informed consent was signed by oneself or legal guardian. All animal experimental procedures were approved by the Animal Ethics Committee of the Tianjin Medical University General Hospital (Ethical No. IRB2021-DWFL-086).

Consent for publication

All authors have agreed to publish this manuscript.

Competing interests

The authors declare no competing interests.