Gut microbiota dysbiosis and metabolic abnormalities promote oxidative stress and fibrosis in idiopathic pulmonary fibrosis

Abstract

Idiopathic pulmonary fibrosis (IPF) is a progressive lung disease marked by declining pulmonary function and excessive fibrosis. Recent studies suggest that gut microbiota and their metabolites influence systemic inflammation and fibrotic processes via the gut-lung axis. We conducted a comprehensive analysis involving pulmonary function tests, gut microbiota profiling, fecal metabolomics, and serum oxidative stress marker measurements in IPF patients and matched healthy controls. Additionally, a bleomycin-induced rat model of IPF were used to assess the effects of tryptophan-glycine (Trp-Gly) supplementation and Ruminococcus torques(R.torques) administration. IPF patients showed significant reductions in lung function (FVC%, FEV1%, TLC%, DLCO%) and distinct gut microbial alterations, including increased Ruminococcus abundance. Metabolomics revealed depletion of Trp-Gly and disrupted amino acid metabolism associated with microbial changes. Serum levels of inducible nitric oxide synthase (iNOS) were elevated, correlating negatively with Trp-Gly and positively with kynurenine, suggesting enhanced oxidative stress. In the animal model, Trp-Gly treatment mitigated fibrosis, oxidative stress, and TGF-β/Smad3 signaling activation, whereas R.torques aggravated these pathological features. Our findings uncover a microbiota-metabolite-oxidative stress axis implicating Ruminococcus-driven metabolic dysregulation in IPF pathogenesis. Targeting this axis provides promising avenues for novel diagnostic and therapeutic strategies in IPF.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-43366-2.

Introduction

Idiopathic pulmonary fibrosis (IPF) is a devastating interstitial lung disease typified by chronic, progressive fibrosis of lung parenchyma, resulting in irreversible loss of respiratory function and eventual respiratory failure¹. This condition primarily involves individuals in mid-to-late life, with its incidence rising in parallel with age². Clinically, patients are characterized by an insidious development of exertional dyspnea, a nonproductive cough, and a continuous deterioration in exercise capacity³. Despite advances, early diagnosis remains challenging and therapeutic options limited, with median survival ranging 3–5 years post-diagnosis⁴.

Traditional investigations have largely focused on local pulmonary processes⁵. However, emerging research reveals systemic metabolic and immunological dysregulation influenced by distal organ systems, notably the gut microbiome⁶. The human gut harbors trillions of microorganisms that engage in complex interactions with the host, producing diverse metabolites including short-chain fatty acids, amino acid derivatives, and bile acids⁷. These microbial products can traverse the gut-lung axis, modulating pulmonary immune responses and potentially influencing fibrotic processes⁸.

Despite growing knowledge, how gut microbiota composition is modified, together with shifts in metabolic patterns and their interactions in IPF, is still not well clarified. This study aims to provide a systematic characterization of intestinal microbiota and the fecal metabolome in individuals with IPF relative to matched healthy subjects, to identify critical microbe–metabolite relationships, and to investigate their links to systemic oxidative stress and pulmonary function decline. Our results provide comprehensive evidence supporting the gut-lung axis as a pivotal mechanism in IPF pathogenesis and highlight novel targets for diagnostic and therapeutic development.

Materials and methods

Participants and sample collection

The final study cohort comprised 17 individuals with IPF and an equal number of healthy counterparts matched for age and gender. Recruitment of the IPF group was contingent upon adherence to the official 2022 ATS/ERS/JRS/ALAT clinical practice guidelines for diagnosis⁹, showed a typical UIP pattern on HRCT¹⁰, presented with progressive dyspnea for more than 6 months or persistent dry cough, and were aged ≥ 18 years¹¹. Patients were excluded if they had ILD of known etiology (including connective-tissue disease–related ILD, drug- or toxin-induced ILD, or radiation injury), coexisting pulmonary diseases that could affect assessment (COPD or asthma), acute pulmonary events within the previous 3 months, malignancy, or severe systemic conditions. Healthy controls were adults without respiratory or chronic systemic diseases, without recent infection, and without recent use of antibiotics, immunomodulators, or proton pump inhibitors. Exclusion criteria for controls included COPD, asthma, connective-tissue diseases, malignancy, gastrointestinal disorders, recent gastrointestinal surgery, or any condition likely to influence gut microbiota. All controls underwent screening to confirm the absence of clinically relevant abnormalities. Approval for the study was obtained from the Ethics Committee of the Fifth Affiliated Hospital of Zhengzhou University, and informed consent was secured from every participant(KY2024033).

Fresh fecal samples were self-collected by participants using sterile containers, immediately placed on ice, and stored at -80 °C within 2 h to preserve microbial DNA and metabolites. Peripheral blood was drawn by venipuncture into serum separator tubes, spun at 3000 revolutions per minute for 20 min to isolate serum, divided into aliquots, and preserved at -80 °C for subsequent oxidative stress marker analysis. Pulmonary function tests (PFTs) were performed within one week of sample collection using standardized spirometry and body plethysmography according to ATS/ERS guidelines, measuring FVC, FEV1, TLC, and DLCO. Values were expressed as percentages of predicted normal.

16 S rRNA gene sequencing and microbiota analysis

Total DNA was isolated from 200 mg of stool using the QIAamp PowerFecal DNA Kit. This method employs physical bead-beating to facilitate the unbiased recovery of DNA from diverse microbial taxa with varying cell wall structures. To rule out reagent or environmental interference, extraction blanks consisting of sterile water were included in every extraction run. These negative controls yielded no detectable DNA and were carried through to PCR amplification, where they showed no visible bands, ensuring the absence of background contamination. DNA integrity and yield were evaluated by NanoDrop spectrophotometry and agarose gel electrophoresis.

Amplification of the bacterial 16 S rRNA gene’s V4 hypervariable region was performed using designated primers: 515 F (5’-GTGCCAGCMGCCGCGGTAA-3’) and 806R (5’-GGACTACHVGGGTWTCTAAT-3’) with Illumina adapter overhangs. PCR was performed under optimized thermal cycling conditions. Amplicons were purified, indexed using Nextera XT Index Kit, and pooled equimolarly. Sequencing was conducted on an Illumina NovaSeq 6000 platform with paired-end 250 bp reads.

Raw sequencing reads underwent quality filtering, trimming of primers and adapters, and chimera removal using QIIME2’s DADA2 plugin, generating amplicon sequence variants (ASVs). Taxonomic assignment was performed against the SILVA 138 database using a naïve Bayes classifier¹². Downstream analyses included calculation of alpha diversity metrics (Chao1 richness, Shannon diversity) to assess within-sample complexity, and beta diversity metrics (Bray-Curtis dissimilarity) to evaluate between-sample compositional differences. Visualization was performed using principal coordinate analysis (PCoA). Differential taxa were identified by LEfSe analysis with LDA score threshold > 2.0 and P < 0.05.

Untargeted metabolomics profiling

Approximately 50 mg of frozen fecal material was extracted with 1 mL of pre-chilled methanol–water (4:1) solution supplemented with 0.1% formic acid and the internal standard L-2-chlorophenylalanine. The mixture was thoroughly mixed, sonicated under ice-cooling, and then subjected to high-speed centrifugation (12,000 rpm, 15 min, 4 °C). The resulting supernatant was passed through a 0.22 μm membrane into LC-MS vials. Metabolic profiling was carried out using a UHPLC system (Thermo Scientific Vanquish) interfaced with a high-resolution Q Exactive HF-X mass spectrometer. Chromatographic separation employed an ACQUITY UPLC BEH Amide column (Waters) with a gradient of ammonium acetate and acetonitrile. Data acquisition was conducted in both positive and negative electrospray ionization modes, including full MS scans and data-dependent MS/MS.

Raw data preprocessing including peak detection, alignment, normalization, and annotation utilized Compound Discoverer and XCMS software. Metabolite identification relied on accurate mass¹³, MS/MS spectral matching, and database searches against HMDB. Metabolic variations between groups were assessed using multivariate approaches, such as principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA). Metabolites showing significant differences were selected based on VIP scores greater than 1, a fold change>1.5 or<0.67, and a p-value<0.05.

Serum biomarker quantification

Serum iNOS levels were quantified using a commercially available ELISA kit (Thermo Fisher Scientific, catalog no. EEL035), following the manufacturer’s instructions. Tryptophan and glycine concentrations in serum were assessed via HPLC with fluorescence detection following pre-column derivatization. Standard calibration curves were employed to ensure accurate quantification, and all samples were analyzed in duplicate.

To establish the optimal therapeutic dose of Trp-Gly, a dose-response evaluation was conducted using a bleomycin-induced idiopathic pulmonary fibrosis (IPF) rat model. Fibrosis was induced via intratracheal instillation of bleomycin (5 mg/kg). Twenty-four hours post-instillation, rats were randomized into five groups (n = 3 per group): (1) Control group (healthy rats administered vehicle, 0.2mL/day normal saline); (2) Model group (IPF rats administered vehicle); (3) Low-dose group (Trp-Gly 5 mg/kg/day); (4) Medium-dose group (Trp-Gly 20 mg/kg/day); and (5) High-dose group (Trp-Gly 50 mg/kg/day). Therapeutic interventions were delivered via intragastric administration on a daily basis for a two-week duration. Upon conclusion of the experiment (Day 14), the rats were sacrificed to obtain blood specimens. To evaluate the dose-dependency of the treatment, we prioritized the quantification of serum iNOS, a pivotal biomarker reflecting the oxidative stress burden in IPF progression.

Rats were administered a single oral bolus of Trp-Gly (20 mg/kg), reconstituted in normal saline, via gavage. Blood samples (200µL) were drawn from the tail vein at designated intervals ranging from 0.5 to 24 h post-dosing. Following collection, serum was isolated via centrifugation (3000×g, 10 min, 4 °C) and cryopreserved at -80 °C. Analyte concentrations were determined using a liquid chromatography-tandem mass spectrometry (LC-MS/MS) system equipped with an electrospray ionization (ESI) source operating in positive mode. Chromatographic resolution was performed on a Waters ACQUITY UPLC BEH C18 column (2.1 × 100 mm, 1.7 μm) thermostated at 40 °C. The mobile phase consisted of 0.1% formic acid in water (A) and acetonitrile (B) delivered at 0.3mL/min using a binary gradient. Trp-Gly was detected via multiple reaction monitoring (MRM) using the transition m/z 276.1→159.0. Calibration curves were established using spiked blank rat serum (0.5–500 ng/mL). Key pharmacokinetic parameters (C_max, T_max, AUC_0−24 h, and t_1/2) were computed via non-compartmental analysis using Phoenix WinNonlin (v8.3).

Resuscitation and cultivation of R.torques

R.torques strains were retrieved from frozen stocks (-80 °C) and plated onto Columbia blood agar media using sterile swabs. The cultures were maintained under anaerobic conditions for 48 h. Following incubation, colony morphology was evaluated within a biosafety cabinet. For downstream processing, bacteria were resuspended in sterile PBS (1 mL) in microcentrifuge tubes. Colonies of R.torques were scraped off the agar surface with a sterile swab and resuspended by vortexing and pressing the swab in the PBS to release bacteria. A spectrophotometer was turned on and set to a single wavelength of 600 nm. A cuvette containing 600 µL PBS was placed into the spectrophotometer and zeroed. Based on the turbidity of the bacterial suspension (typically, when visible turbidity is observed, the OD600 exceeds 1.0), the suspension was diluted approximately 10-fold to reach an OD600 of about 0.1, and then further diluted 20-fold. A new sterile swab was used to pick the diluted bacterial suspension for plating, which was then incubated anaerobically. Bacteria from the second passage were collected by swabbing and resuspended in PBS. The suspension was brought to a McFarland standard of 1 (~ 3 × 10⁸CFU/mL). Next, 1.5 mL was added to 13.5 mL of BHI broth supplemented with 10% FBS and cultured anaerobically at 37 °C on a shaker at 60 rpm for 16–18 h to achieve exponential growth. The bacterial suspension was adjusted to a concentration of 1 × 10⁹ CFU/mL for gavage, based on doses established in previous studies¹⁴.

IPF model establishment

Twelve healthy male Sprague-Dawley rats (6–8 weeks old, SPF grade; purchased from Speifu Co., Ltd.) were acclimated for 1 week under controlled environmental conditions (temperature 22 ± 2 °C, humidity 50–60%, 12-h light/dark cycle) with free access to standard chow and water. Before any experimental intervention, all rats were pretreated with a mixed antibiotic regimen to eliminate potential differences in baseline gut microbiota. A total of 500 mg ciprofloxacin, 2 g metronidazole, 1.5 g erythromycin, and 400 mg albendazole were dissolved in 40 mL of water and thoroughly mixed. Each rat received 200 µL of the antibiotic mixture by oral gavage once daily for 3 consecutive days.

After pretreatment, the rats were randomly divided into four groups (n = 3 per group): Control, Model, Trp-Gly, and R.torques. Except for the Control group, all rats were anesthetized and administered bleomycin A5 (5 mg/kg in 0.2 mL saline) via intratracheal instillation to induce IPF¹⁵ while the Controls received an equal volume of saline. To minimize stress, all manipulations were performed gently under deep anesthesia. Beginning the day after model induction, the Trp-Gly and R.torques groups received Trp-Gly (20 mg/kg, gavage once daily for 14 days) and R.torques (1 × 10⁹CFU, gavage once daily), respectively. The Control and Model groups received equal volumes of saline by gavage. Twenty-four hours after the final treatment, body weight was recorded, and the rats were euthanized via intraperitoneal injection of xylazine (12.5 mg/kg) and ketamine (87.5 mg/kg)¹⁶. Both lungs were harvested; lung weight was recorded and lung index was calculated (lung weight/body weight). Left lungs were snap-frozen in liquid nitrogen for Western blot analysis, and right lungs were fixed in 4% paraformaldehyde for histology. Each experiment was independently repeated three times. Animals were monitored daily for signs of distress. They were euthanized immediately if they showed body weight loss ≥ 20%, severe lethargy, labored breathing, inability to eat or drink, or any other signs of severe pain or illness. Based on our experience with bleomycin-induced pulmonary fibrosis (IPF) rat models, these criteria ensured reproducible results while minimizing animal suffering. Animal research has been approved by the Ethics Committee of the Fifth Affiliated Hospital of Zhengzhou University (KY-D-2025046).

Immunohistochemistry

Paraffin-embedded right lung sections (4 μm), derived from tissues fixed in 4% paraformaldehyde, were subjected to deparaffinization and microwave-based antigen retrieval (citrate buffer, pH 6.0). Endogenous peroxidase activity and non-specific binding sites were blocked using 3% H₂O₂ and 3% BSA, respectively. Subsequently, sections were incubated with anti-α-SMA (1:200, Abcam, ab5694) or anti-COL1A1 (1:1500, Abcam, ab138492) antibodies overnight at 4 °C. Following HRP-conjugated secondary antibody application at room temperature, positive staining was visualized via DAB reaction. The sections were finally counterstained with hematoxylin, dehydrated, and mounted for imaging.

Histological staining

For morphological assessment, sections underwent H&E staining involving hematoxylin application, acid-alcohol differentiation, and ammonia water-induced bluing. After eosin staining, samples were dehydrated through an ethanol series, cleared, and mounted. Collagen deposition was assessed via Masson’s trichrome staining. Dewaxed slides were processed strictly according to the manufacturer’s guidelines, involving sequential incubation with the provided staining solutions (A through F), followed by standard dehydration and mounting procedures.

Western blotting

Left lung tissues were lysed in RIPA buffer containing protease/phosphatase inhibitors and centrifuged (12,000×g, 15 min, 4 °C) to obtain supernatants. After determining protein content via BCA assay, aliquots (20–40 µg) were boiled, subjected to SDS-PAGE separation, and transferred to PVDF membranes. Blocking was performed using 5% milk or BSA, followed by incubation with primary antibodies against TGF-β (21898-1-AP), Smad3 (66516-1-Ig), GAPDH (10494-1-AP) (Wuhan Sanying) and Phospho-SMAD3 (C25A9; Cell Signaling Technology; #9520) at the indicated dilutions. Blots were then exposed to HRP-linked secondary antibodies (1:2000) for 1 h. Signals were detected using an ECL system, and band intensity was analyzed with ImageJ. In instances of duplicated sample loading, the redundant lane was excluded from the final presentation for clarity, without affecting the interpretation of the results.

Statistical analysis

Data were expressed as mean ± SD or median (IQR) based on normality assessed by the Shapiro-Wilk test. Two-group comparisons were performed using Student’s t-test or Mann-Whitney U test. Spearman correlation was used to evaluate associations between microbial taxa, metabolites, and clinical variables (P< 0.05). Analyses were conducted in R 4.3.2 and GraphPad Prism 9.

Results

Clinical characteristics and pulmonary function

A total of 34 subjects were included, comprising seventeen IPF patients and seventeen age- and sex-matched healthy controls. The mean ages were comparable between the IPF group (70.2 ± 7.4 years) and the control group (68.5 ± 6.9 years). Moreover, there were no significant differences between the two groups in terms of BMI (P = 0.84), smoking (P = 0.426), or alcohol consumption (P = 0.757). Analysis of pulmonary function revealed that IPF patients had significantly lower lung function parameters compared to controls. Specifically, FVC% was reduced to 83.2 ± 6.3% in IPF patients versus 88.9 ± 6.4% in controls (P = 0.00012). Similarly, FEV1% decreased to 82.1 ± 6.9% in the IPF group compared to 88.3 ± 6.3% in controls (P = 0.00018). TLC% was also reduced, measuring 70.2 ± 7.1% in IPF patients versus 79.3 ± 6.9% in controls (P = 0.00008). DLCO% demonstrated the most pronounced decline, with values of 44.8 ± 7.3% in IPF versus 83.4 ± 6.8% in controls (P = 0.00005) (Fig. 1). These findings confirm significant impairments in lung ventilation and gas exchange capacity in the IPF cohort, consistent with restrictive lung disease and disrupted alveolar-capillary function characteristic of IPF¹⁷.

Fig. 1.

Comparison of pulmonary function parameters between IPF patients and healthy controls. Boxplots depict the distributions of FVC%, FEV1%, TLC%, and DLCO% in the two groups. Each point represents an individual subject.

Global microbiota structure and metabolomic profiles

We performed principal component analysis (PCA) and Principal Coordinate Analysis (PCoA) to evaluate the overall structural changes in metabolites and microbiota, respectively. As shown in Fig. 2A, the metabolomic profiles of IPF patients formed a distinct cluster separate from healthy controls. Permutational multivariate analysis of variance (PERMANOVA) confirmed significant differences in the metabolome (P = 0.001, R²=0.099), indicating profound metabolic reprogramming in IPF. In contrast, the gut microbiota structure based on Bray-Curtis dissimilarity did not show global separation between groups in the PCoA plot (Fig. 2B), and PERMANOVA analysis yielded no statistically significant difference (P = 0.674, R²=0.026). This discrepancy suggests that while the overall microbial community structure remains relatively stable, specific functional perturbations or the alteration of key low-abundance taxa may drive the significant metabolic shifts observed. Consequently, we proceeded with LEfSe analysis to identify specific differential taxa that may contribute to these metabolic differences despite the lack of global structural separation.

Fig. 2.

Principal Component Analysis. (A): PCA of metabolites; (B): PCA of microbiota.

Microbial-metabolite interactions

The results showed significant differences in microbiota structure and metabolite composition between the control and IPF groups (Figs. 3 and 4). At the genus level, Ruminococcus, Parabacteroides, and Atopobium were significantly correlated with multiple metabolites (P < 0.001). Specifically, Ruminococcus was strongly negatively correlated with Trp-Gly (r=-0.640, P = 4.53E-05); Parabacteroides was negatively correlated with 15(R)-prostaglandin E1 (r=-0.638, P = 4.87E-05); and Atopobium was positively correlated with carnitine metabolites, including carnitine C8:1 (r = 0.614, P = 0.000111), carnitine C6:0 (r = 0.617, P = 0.000102), and carnitine C10:0 (r = 0.627, P = 7.17E-05)(Table 1). These findings suggest that Ruminococcus and Parabacteroides may participate in amino acid and prostaglandin metabolism, whereas Atopobium appears closely linked to carnitine-related fatty acid metabolism.

Fig. 3.

Heatmap of microbiota and metabolite abundances in samples. Asterisks indicate levels of statistical significance: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Fig. 4.

Correlation heatmap of microbiota and metabolite abundances in samples. Asterisks indicate levels of statistical significance: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Table 1.

Correlation table between genus-level microbiota and metabolites.

Index	Taxonomy	Correlation	P-value	Compounds
MEDN2007	Ruminococcus	-0.64036	4.53E-05	Trp-Gly
MEDN1865	Parabacteroides	-0.63832	4.87E-05	15(R)-Prostaglandin E1
MEDP1428	Atopobium	0.614242	0.000111	Carnitine C8:1
MEDP1289	Atopobium	0.616857	0.000102	Carnitine C6:0
MEDP1419	Atopobium	0.627316	7.17E-05	Carnitine C10:0

Although multiple fecal metabolites exhibited significant correlations with different genera, fecal Trp-Gly was prioritized because it showed the strongest association with Ruminococcus, the most differentially abundant genus between IPF and healthy controls. Biologically, Trp-Gly is directly linked to tryptophan metabolism, which has been implicated in immune regulation, oxidative stress, and epithelial injury in IPF. In contrast, other positively correlated metabolites, such as carnitine species, while significant, are less directly connected with known IPF mechanisms. Prior studies have also shown that alterations in tryptophan metabolism in IPF patients may lead to reduced anti-inflammatory capacity, increased levels of pro-inflammatory factors such as TNF-α and IL-6, and exacerbated oxidative stresss^18,19. Additionally, glycine, as a key anti-inflammatory and antioxidant molecule, when decreased, may cause excessive activation of inflammatory pathways such as NF-κB and impair glutathione (GSH) synthesis, further aggravating oxidative stress and IPF progression²⁰.

Based on these correlation and mechanistic considerations, Trp-Gly emerged as the metabolite most closely associated with Ruminococcus. We hypothesize that Ruminococcus may influence IPF progression by modulating Trp-Gly metabolism, although direct experimental validation, such as colonization models and assessment of gut permeability, is required. This study highlights the potential pathogenic role of Ruminococcus in IPF, offering insights into microbiota–metabolite–disease interactions and suggesting a candidate target for future therapeutic intervention.

Serum oxidative stress and amino acid levels

Based on the associations between Ruminococcus abundance, Trp-Gly levels, and oxidative stress, this study analyzed serum iNOS and Trp, Gly, and Kyn metabolite levels in IPF patients and healthy controls to investigate potential links between gut microbiota, metabolites, and systemic oxidative stress. Results demonstrated that IPF patients exhibited significantly elevated serum iNOS levels compared to controls (Fig. 5A), indicating increased oxidative stress in the disease state. Trp and Gly levels were significantly reduced in IPF patients and showed strong negative correlations with iNOS (Trp: Spearman’s r=-0.86, P = 8.6 × 10⁻⁶; Gly: r=-0.83, P = 3.6 × 10⁻⁵) (Fig. 5B). Additionally, serum Kyn-a principal downstream metabolite of Trp catabolism via the kynurenine pathway-was markedly increased in IPF patients and exhibited a positive correlation with iNOS levels (r = 0.72, P = 1.2 × 10⁻³) (Fig. 5B).

Fig. 5.

Alterations in serum oxidative stress marker and amino acid metabolites in IPF patients compared to healthy controls and their correlations. (A) Serum iNOS, Serum Trp, Gly and Serum Kyn levels between IPF patients and normal controls. (B) Correlation between serum Trp, Gly, Kyn and iNOS levels in IPF patients. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

These findings indicate heightened oxidative stress in IPF, which is associated with altered Trp metabolism and amino acid availability, including reduced Trp-Gly levels, and coincides with gut microbiota alterations such as increased Ruminococcus. While these observations are correlative, they support a hypothetical model in which dysbiosis-induced perturbations in amino acid metabolism may contribute to oxidative stress and IPF progression. Trp-Gly, as a dipeptide derived from tryptophan and glycine, may influence redox homeostasis through modulation of antioxidant pathways such as glutathione synthesis and NF-κB signaling, although direct evidence in IPF remains to be established. Further experimental studies, including functional validation and intervention models, are required to directly test this mechanistic link. Furthermore, to validate the feasibility of our therapeutic intervention, we assessed the bioavailability of Trp-Gly in rats. Following a single oral dose of 20 mg/kg, serum analysis revealed that Trp-Gly was rapidly absorbed, reaching a peak concentration (Cmax) of 15.4 ± 2.1 µg/mL at 1 h (Tmax). The peptide remained detectable in serum for up to 12 h, confirming that oral administration effectively delivers Trp-Gly into the systemic circulation. The relationship between three different concentration dosages and serum iNOS in rats is shown in Supplementary Fig. 1A&1B. The results indicate that 20 mg/kg has the best effect. Similarly, Supplementary Fig. 1C also shows that 20 mg/kg has better absorption and pharmacokinetic properties.

Species-level differential microbiota analysis

To identify meaningful bacterial taxa for subsequent experimental validation, we performed Linear Discriminant Analysis (LDA) scores (log10) on the microbiota data. Statistically significant differential taxa are presented in Fig. 6A. Analysis revealed a marked increase of the Ruminococcaceae and R.torques species in IPF patients relative to healthy controls. Furthermore, seven species within the Ruminococcus were detected. We conducted differential abundance analysis of these species between the IPF and normal groups, which are displayed as bar plots in Fig. 6B. Notably, R.torques was significantly enriched in the IPF group, consistent with the previous results, indicating it is the most differentially abundant species within the Ruminococcus. Further analysis at the species level identified six taxa with significant differences, as shown in Fig. 6C, including R.torques, Campylobacter gracilis, uncultured marine, Bacteroides caccae, Veillonella atypica, and Bifidobacterium pseudocatenulatum. Taken together, these findings suggest that among Ruminococcus species, R.torques is the key differentially abundant taxon that may play an important role in IPF. Analysis based on the SILVA database revealed significant enrichment of the Ruminococcus in IPF patients, with R.torques showing a statistically significant difference between groups (P < 0.05). Thus, R.torques may represent the primary driver of genus-level differences and serve as a representative species in subsequent animal experiments to elucidate its role in IPF pathogenesis.

Fig. 6.

Species-level differences in microbiota between IPF and control groups. (A) Bacterial taxa with statistically significant differences between the IPF and normal groups were identified by LDA score (log10) analysis and presented as a bar plot. (B) Differences in the abundance of seven detected species within the Ruminococcus between the IPF and normal groups are shown. (C) Six bacterial taxa with statistically significant differences at the species level are displayed. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Body weight and lung index

According to the IPF modeling method described in the Materials and Methods section, we established a rat model of IPF using bleomycin (Supplementary Fig. 2A-D). To assess the successful induction of bleomycin-induced IPF in rats and the initial effects of different interventions, we recorded the body and lung weights for each group. The lung index (lung weight/body weight), a marker of fibrosis severity (Fig. 7A), was markedly higher in the model group versus controls (P < 0.01), confirming the establishment of IPF. Trp-Gly treatment significantly lowered the lung index relative to the model group (P < 0.05), indicating partial improvement in lung injury, although it did not fully return to control levels (P < 0.05). In contrast, rats receiving R.torques displayed an increased lung index compared with the model group (P < 0.05), suggesting worsening of fibrosis. Supporting these findings, serum iNOS levels, an indicator of oxidative stress, were strongly elevated in the model group (P < 0.001). Administration of Trp-Gly reduced iNOS concentrations (P < 0.01), whereas R.torques further raised them (P < 0.05) (Fig. 7B), implying that R.torques may enhance IPF-related oxidative stress.

Fig. 7.

Changes in lung coefficient and iNOS levels among different treatment groups. (A) Changes in lung coefficient across the groups. (B) Changes in iNOS levels across the groups. *P < 0.05, **P < 0.01, ***P < 0.001.

H&E and masson staining

As shown in Fig. 8, HE staining demonstrated well-preserved lung architecture in the control group, with clear alveolar septa and no notable inflammatory infiltration. In the model group, alveolar walls were thickened, structures were disorganized, and there was extensive infiltration of lymphocytes, monocytes, and macrophages, reflecting a pronounced chronic inflammatory response. Trp-Gly treatment mitigated inflammatory infiltration and partially restored alveolar structure, although mild inflammation and septal thickening persisted. In contrast, R.torques led to more severe alveolar damage and inflammatory cell accumulation than the model group. Masson trichrome staining showed minimal collagen fibers in controls, while the model group exhibited marked collagen deposition, indicative of fibrosis progression. Trp-Gly reduced collagen accumulation relative to the model, demonstrating partial improvement, whereas R.torques further intensified collagen deposition, exacerbating pulmonary fibrosis.

Fig. 8.

HE staining and Masson’s trichrome staining of lung tissues from different treatment groups.

Immunohistochemistry (IHC)

To further elucidate myofibroblast activation and collagen deposition during pulmonary fibrosis, immunohistochemical analysis of α-smooth muscle actin (α-SMA) and type I collagen α1 chain (COL1A1) was performed on lung tissues. α-SMA is a classic marker for myofibroblasts, reflecting the transdifferentiation of fibroblasts into myofibroblasts, while COL1A1 expression directly indicates active collagen synthesis. The results indicated that both markers were expressed at low levels in lungs of normal controls, suggesting minimal fibrotic activity. In contrast, the model group exhibited marked upregulation of α-SMA and COL1A1, demonstrating extensive myofibroblast activation and abnormal collagen deposition during fibrosis. Trp-Gly treatment significantly suppressed the expression of these fibrosis markers, highlighting its potential anti-fibrotic effect. Notably, the R.torques group not only failed to inhibit but further promoted increased expression of α-SMA and COL1A1, suggesting that this bacterial strain may exacerbate fibrosis by enhancing myofibroblast activation and collagen production (Fig. 9). The results suggest a strong association of pulmonary fibrosis progression with gut microbiota, implying that dysbiosis may play a key role in promoting fibrosis.

Fig. 9.

Immunohistochemical analysis of α-SMA and COL1A1 expression in lung tissues.

Western blot

To investigate the effects of different interventions on fibrosis-related signaling, we examined the activation of TGF-β and its downstream effector Smad3, a key pathway driving fibroblast proliferation and collagen deposition in pulmonary fibrosis. In the model group, both TGF-β and phosphorylated Smad3 (p-Smad3) were markedly elevated, indicating pathway activation and fibrotic progression. Trp-Gly treatment significantly suppressed TGF-β and p-Smad3 levels, suggesting its anti-fibrotic effect involves inhibition of TGF-β/Smad3 signaling. Conversely, R.torques further increased their expression, supporting its fibrosis-promoting role, while total Smad3 remained unchanged, indicating that differences arose mainly from Smad3 phosphorylation (Fig. 10). The uncropped and original immunoblot data are provided in Supplementary Fig. 3. These findings reveal that Trp-Gly and specific gut microbiota modulate pulmonary fibrosis through the TGF-β/Smad3 pathway, providing a theoretical basis for targeted therapies and highlighting gut microbiota modulation as a potential adjuvant strategy.

Fig. 10.

Western blot analysis of fibrogenic signaling pathway proteins in lung tissues.

Discussion

IPF is a devastating interstitial lung disease characterized by progressive lung function decline, excessive extracellular matrix deposition, and architectural distortion of the alveoli²¹. Our clinical data reaffirmed the hallmark restrictive ventilatory defects and severely impaired gas exchange in IPF patients, as evidenced by significantly reduced FVC%, FEV1%, TLC%, and DLCO% compared with age- and sex-matched healthy controls²². These functional deficits highlight the importance of gaining deeper insight into the complex processes underlying IPF progression.

Emerging evidence increasingly implicates the gut microbiota as a pivotal modulator of systemic immunity and inflammation through the gut-lung axis, influencing respiratory health and disease²³. Although global beta-diversity metrics did not reveal a complete structural shift in the microbiome (P = 0.674), likely due to the high inter-individual variability inherent in human cohorts, LEfSe analysis successfully identified specific pathogenic signatures. Notably, enriched Ruminococcus²⁴ and Parabacteroides²⁵-genera previously linked to pro-inflammatory and pro-fibrotic phenotypes-coexisted with depletion of potentially protective taxa such as Enterococcus²⁵ and Streptococcus²⁵. These microbial alterations align with prior reports associating specific dysbiotic signatures with chronic pulmonary inflammation and fibrosis, suggesting that IPF is accompanied by subtle but biologically significant perturbations in gut microbial ecology.

Our integrative metabolomic profiling further revealed a striking depletion of the dipeptide Trp-Gly in IPF patients’ fecal samples, a metabolite closely correlated with Ruminococcus abundance. This finding is consistent with a model where microbial dysbiosis disrupts amino acid-derived metabolites²⁶ critical for maintaining immune tolerance²⁷ and redox homeostasis²⁸ in the lung microenvironment. Tryptophan metabolism, particularly via the kynurenine pathway, yields bioactive metabolites that regulate oxidative stress²⁹ and fibroblast activation³⁰—central events in IPF pathogenesis. Concurrently, glycine serves as a precursor for glutathione synthesis, the master intracellular antioxidant, and its depletion compromises cellular defenses against reactive oxygen species (ROS)³¹, thereby amplifying fibrotic signaling cascades³². The observed strong inverse correlation between Ruminococcus and Trp-Gly levels suggests that overgrowth of this genus may actively perturb these metabolic pathways, exacerbating disease progression. Contrary to Trp-Gly, supplementation with R.torques significantly exacerbated pulmonary fibrosis lesions. This was reflected by further increases in lung weight index and inducible nitric oxide synthase (iNOS) levels, aggravated lung tissue inflammation and collagen deposition, as well as upregulated expression of fibrosis-related proteins. Studies have demonstrated that R.torques possesses a strong capacity to degrade the intestinal mucus layer, which is primarily composed of mucins. In the gut, R.torques effectively utilizes and breaks down mucins by secreting various glycosidases and adhesion proteins, thereby disrupting the intestinal surface barrier³³. This process compromises the intestinal mucosal barrier function and increases the risk of translocation of luminal contents, including bacterial components and inflammatory mediators, into the host circulation, thus triggering local intestinal and systemic inflammatory responses³³. Further research has confirmed that in patients with inflammatory bowel disease (IBD), the abundance of mucin-degrading bacteria such as R.torques is significantly elevated, which can synergize with other bacterial species to enhance mucin degradation. This not only exacerbates intestinal barrier damage but also facilitates the entry of more harmful substances into the systemic circulation³⁴. These inflammatory signals and bacterial products may act on the lungs via the gut-lung axis, inducing or aggravating pulmonary inflammation and fibrosis.

Corroborating this mechanistic link, we detected significantly elevated serum iNOS levels in IPF patients, indicative of systemic oxidative stress³⁵. The robust negative correlations between serum iNOS and both tryptophan and glycine concentrations underscore the contribution of amino acid metabolic disturbances to oxidative burden. Moreover, increased kynurenine—a key metabolite downstream of tryptophan catabolism—positively correlated with iNOS³⁶, further implicating dysregulated tryptophan metabolism in oxidative and fibrotic lung injury. Collectively, these data highlight a microbiota-metabolite-oxidative stress axis that may constitute a central pathogenic mechanism in IPF.

To experimentally validate these clinical observations, we employed a bleomycin-induced rat model of IPF³⁷. Consistent with human data, bleomycin-treated rats exhibited increased lung index, collagen deposition, myofibroblast activation (α-SMA), and upregulation of the canonical profibrotic TGF-β/Smad3 signaling pathway³⁸. Therapeutic administration of Trp-Gly attenuated fibrosis and oxidative stress markers while suppressing fibrogenic signaling, demonstrating its potential as an antifibrotic agent. Conversely, supplementation with Ruminococcus exacerbated lung injury, heightened oxidative stress³⁹, and amplified profibrotic²⁴ signaling cascades, thereby reinforcing the pathogenic role inferred from microbiome and metabolome analyses. These findings align with growing literature indicating that gut microbiota and their metabolites can modulate pulmonary fibrosis⁴⁰ through immune⁴¹ and oxidative stress pathways⁴². Thus, R.torques may contribute to disease progression and represents a potential microbial driver that warrants further validation.

Nonetheless, our study has several limitations. Our study has several limitations. First, the clinical cohort size (n = 17 per group) is relatively modest. This limited sample size may account for the partial overlap observed in the PCA of microbial beta diversity, as human microbiome data typically requires larger cohorts to overcome inter-individual heterogeneity. However, despite the lack of complete global separation, the identification of robust differential taxa (LEfSe) and metabolites, validated by animal experiments, strengthens our conclusions. Second, the cross-sectional design limits causal inference in humans. Third, regarding the correlation analysis, we assessed associations between fecal Trp-Gly and serum iNOS. While this involves different biological matrices, it aligns with the concept of the gut-lung axis where gut-derived metabolites translocate into circulation; nevertheless, direct measurement of circulating Trp-Gly levels would further strengthen this link. In the animal experiments, only three rats were used per group, which may reduce the statistical power and reliability of the conclusions. Although we observed a negative correlation between Ruminococcus abundance and Trp‑Gly levels in clinical samples, and confirmed the biological effects of R.torques and Trp‑Gly in animals, we did not directly measure Trp‑Gly levels in feces or serum after R.torques administration. Therefore, the specific molecular mechanism by which R.torques consumes Trp‑Gly remains unverified in vivo. Future studies employing larger sample sizes, longitudinal sampling, microbiota transplantation, targeted metabolite modulation, and quantitative assessment of Trp‑Gly after bacterial intervention are needed to confirm causality and further elucidate the underlying mechanisms.

Conclusion

In summary, our integrative clinical and experimental data illuminate a novel microbiota-metabolite-oxidative stress axis in IPF, whereby dysbiosis—especially Ruminococcus expansion—disrupts amino acid metabolism, promotes oxidative stress, and activates profibrotic signaling. Targeting this axis holds promise for early diagnosis and innovative therapies aimed at halting or reversing IPF. Future studies should focus on clarifying the causal relationship between R.torques and fibrosis, including the use of colonization models and assessments of intestinal permeability.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

IPF

Idiopathic Pulmonary Fibrosis

FVC

Forced Vital Capacity

FEV

Forced Expiratory Volume

TLC

Total Lung Capacity

DLCO

Diffusion Capacity for Carbon monoxide

Author contributions

Bing Bai: Conceptualization, Formal analysis, Writing-original draft; Fazhan Li: Methodology, Data curation; Pengyuan Zheng: Supervision, Writing review & editing; Xiaochen Li: Validation, Supervision; Di Guo: Software, Writing review & editing, Wenfei Zhao: Project administration, Funding acquisition. Chenfeng Hua: Writing review & editing, Visualization. All authors contributed to and approved the submitted manuscript.

Funding

This study was supported by the Henan Provincial Medical Science and Technology Research Program Joint Construction Project (grant number LHGJ20210482).

Data availability

The 16 S rRNA sequencing dataset generated during the current study has been deposited in the NCBI Sequence Read Archive (SRA) under accession number PRJNA1337890 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1337890). Additional data supporting the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

This study was approved by the Ethics Committee of The Fifth Affiliated Hospital of Zhengzhou University. Human studies were conducted in accordance with the Declaration of Helsinki, with approval number KY2024033. Written informed consent was obtained from all participants. Animal experiments were carried out following the Guide for the Care and Use of Laboratory Animals and were approved under ethical approval number KY-D-2025046. The study was conducted in accordance with relevant guidelines and regulations and reported in line with the ARRIVE guidelines.