Early diabetes‐like phenotypes in germ‐free mice induced by gut microbiota from patients with type 2 diabetes Authors: Jianghao Feng, Xing Liu, Hua Zhu, Yun Yang, Wei Liu, Xiaoliang Jiang, Zhiwei Yang Published: 2026 ABSTRACT AbstractBackgroundThe contribution of gut microbiota to the early stage of type 2 diabetes (T2D) remains incompletely understood. This study established a germ‐free mouse model colonized with gut microbiota from donors with T2D to determine whether diabetes‐associated microbiota could induce early diabetes‐like phenotypes.MethodsHuman fecal samples were collected from 18 healthy donors and 14 donors with T2D for microbiota profiling. Based on these features, samples from 12 healthy donors and 3 donors with T2D were selected, pooled within each group, and transplanted into germ‐free mice, generating a healthy microbiota recipient group (HM, n = 14) and a diabetes‐associated microbiota recipient group (DM, n = 13). Glucose homeostasis was assessed over 10 weeks using fasting blood glucose and intraperitoneal glucose tolerance tests. Gut microbial succession was analyzed by 16S rRNA sequencing. Fecal and plasma metabolomics were performed to identify metabolic and their associations with microbial changes.ResultsDM mice developed an early diabetes‐like phenotype characterized by progressive impairment of glucose tolerance, reduced insulin levels, and mild renal alterations, without sustained overt fasting hyperglycemia. Microbial divergence emerged before stable metabolic dysfunction and was accompanied by persistent dysbiosis in DM mice. Integrated metabolomic analysis identified coordinated alterations in fecal and plasma metabolites, with cholic acid and L‐Dopa decreased in both compartments. L‐Dopa showed consistent associations with altered genera, particularly Ruminococcus and Sellimonas.ConclusionsTransplanting diabetes‐associated human gut microbiota effectively induced early glucose dysregulation in germ‐free recipient mice. This model provides a framework for studying microbiota‐associated events during early T2D‐related metabolic deterioration. INTRODUCTION Type 2 diabetes (T2D) is a chronic metabolic disease that is often accompanied by cardiovascular complications and accounts for a large proportion of premature deaths worldwide. 1 In 2021, about 529 million people were living with diabetes, and this number is expected to increase to 1.31 billion by 2050. 2 The development of T2D involves a complex interplay of metabolic dysregulation, inflammation, and immune dysfunction. 3 , 4 , 5 In recent years, the gut microbiota has emerged as an important regulator of host metabolism, immune balance, and intestinal homeostasis. 6 , 7 In patients with T2D, gut microbial dysbiosis has been consistently associated with reduced microbial diversity and altered community composition. 8 , 9 , 10 These alterations commonly involve reduced microbial richness and changes in overall community structure. 11 , 12 These observations suggest a close relationship between gut microbial dysbiosis and metabolic disease. Human observational studies do not readily determine whether gut microbiota alterations contribute directly to the early development of T2D or simply arise secondary to the disease state. Interpretation is further complicated by substantial inter‐individual variability, as well as by the effects of diet, medication use, and environmental exposure on gut microbial composition. 13 , 14 , 15 Fecal microbiota transplantation into germ‐free (GF) animals provides an experimental strategy for testing whether disease‐associated microbial communities can directly influence host phenotypes. 16 , 17 , 18 , 19 Pharmacologically induced diabetic models already exhibit marked metabolic abnormalities, which may obscure microbiota‐dependent effects. Antibiotic‐treated pseudo‐germ‐free models are also limited by incomplete microbial depletion and the potential confounding effects of antibiotic exposure. 20 , 21 By contrast, GF mice provide a cleaner background for evaluating the effects of transplanted microbial communities. 22 Several studies have used microbiota transplantation to investigate metabolic disease, but important questions remain unresolved. Many experimental studies have relied on murine‐to‐murine transplantation or conventional recipient models, which limit the evaluation of human disease‐associated microbial communities in a microbiota‐free setting. 23 Previous studies have focused on established metabolic dysfunction rather than the early stage of disease development. 24 As a result, it remains unclear whether diabetes‐associated microbial remodeling precedes overt metabolic impairment or instead emerges after host metabolic abnormalities are already present. A longitudinal GF mouse model colonized with human gut microbiota provides an experimental framework for addressing this question. In this study, we established a germ‐free mouse model of early diabetes‐like phenotypes by transplanting gut microbiota from patients with T2D. This model was used to examine whether T2D‐associated gut microbiota could induce early diabetes‐like changes in recipient mice, characterized mainly by impaired glucose tolerance and early metabolic dysfunction rather than overt diabetes. We further characterized the temporal dynamics of microbial succession after transplantation and integrated fecal and plasma metabolomic analyses to identify microbiota‐associated metabolic alterations during disease progression. Through this approach, we aimed to define the temporal relationship among donor‐associated microbial features, post‐transplant microbial remodeling, and the emergence of early metabolic dysfunction. MATERIALS AND METHODS Animals All animal protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of the Institute of Laboratory Animal Science, Chinese Academy of Medical Sciences (CAMS) and Peking Union Medical College (PUMC) (ZH23003). Germ‐free (GF) mice were generated as previously described. 25 Eight‐week‐old male GF C57BL/6J mice were randomly allocated into two groups: a healthy microbiota recipient group (HM, n = 14) and a diabetes‐associated microbiota recipient group (DM, n = 13). All mice were maintained in aseptic isolators throughout the study. The housing facility was maintained under a 12‐h light/dark cycle with controlled temperature and humidity. Mice were provided with free access to sterile food and water. To accurately monitor food and water intake, mice were housed in individual cages. Fresh fecal samples were collected from the germ‐free mice for sterility testing before transplantation. Sterility assessment was performed according to GB/T 14926.41‐2001, and the results of both bacterial and fungal examinations were negative. Microbiota preparation and transplantation Collection of human fecal specimens was approved by the ethics committee of the Institute of Laboratory Animal Science, Chinese Academy of Medical Sciences (CAMS) and Peking Union Medical College (PUMC) (YZW23001). A total of 32 human fecal samples were collected, including 18 samples from healthy individuals and 14 samples from patients with T2D. None of the donors had received antibiotics within 4 weeks of fecal sample collection. Healthy donors were selected from individuals without obesity, diabetes, fatty liver disease, other metabolic disorders, or gastrointestinal inflammation. Donors in the T2D group had a clinical diagnosis of T2D and had fasting blood glucose >7.0 mmol/L or glycated hemoglobin (HbA1c) ≥6.5%, without other major metabolic disorders. All eligible human fecal samples were included in the microbiota analysis. For the transplantation experiment, fecal samples from 12 healthy donors and 3 donors with T2D were selected and pooled within each group before transplantation. The T2D donors used for transplantation were newly diagnosed or early‐stage patients who had not received glucose‐lowering medication before sample collection. To prepare the microbiota suspension, 1 g of the frozen pooled stool was suspended in 100 mL of sterile 0.1 M phosphate‐buffered saline (PBS, pH 7.2) and vortexed thoroughly. The resulting fecal suspension was allowed to settle briefly, and the supernatant was collected for transplantation. Germ‐free mice received 0.4 mL of the corresponding fecal supernatant by oral gavage once daily for three consecutive days. Phenotypic characterization Body weight, food intake and water consumption were recorded weekly. Fasting blood glucose (FBG) was measured at weeks 0, 2, 4, 6, 8, and 10 using tail‐vein blood samples collected after an 8‐h fast. Intraperitoneal glucose tolerance tests (IPGTTs) were performed at weeks 2, 6, and 10 by intraperitoneal injection of glucose at 2 g/kg body weight following an 8‐h fast. Blood glucose levels were recorded at the indicated time points after glucose administration. At the end of the experiment, plasma samples were collected for biochemical analysis, which was performed by the Shanghai Model Organisms Center (Shanghai, China). Plasma insulin levels were measured using an Ultra Sensitive Mouse Insulin ELISA Kit (Crystal Chem, Cat. No. 90080). Insulin concentrations were calculated from the standard curve generated in each assay. The assay sensitivity was 0.05 ng/mL, and the detection range was 0.1–6.4 ng/mL (low range), 0.1–12.8 ng/mL (wide range), or 1–64 ng/mL (high range). HOMA‐IR and HOMA‐β were calculated from fasting blood glucose and fasting insulin levels. Insulin values measured in ng/mL were converted to μIU/mL using a factor of 25 before calculation. HOMA‐IR was calculated as fasting blood glucose × fasting insulin (ng/mL) × 25/22.5, and HOMA‐β as 20 × fasting insulin (ng/mL) × 25/(fasting blood glucose − 3.5). Gene expression analysis Total RNA was extracted using TRIzol reagent (Accurate Biology, AG21102) and reverse transcribed into cDNA using M‐MLV reverse transcriptase (Applied Biosystems, USA). Quantitative PCR (qPCR) was performed using SYBR Premix Ex Taq II (Takara, Tokyo, Japan) to determine relative gene expression levels. Gene expression was normalized to GAPDH and calculated using the 2−ΔΔCt method. The primer sequences used in this study are listed in Table 1. TABLE 1Primers of RT‐qPCR.GenesForward primers (5′‐3′)Reverse primers (5′‐3′)TNF‐αGCCTCTTCTCATTCCTGCTTTGGGAACTTCTCATCCCTTTGIL‐1βGCAACTGTTCCTGAACTCAACTATCTTTTGGGGTCCGTCAACTIL‐6TAGTCCTTCCTACCCCAATTTCCTTGGTCCTTAGCCACTCCTTCGAPDHCTGGGCTACACTGAGCACCAAGTGGTCGTTGAGGGCAATG Primers of RT‐qPCR. Western blot analysis Kidney tissues were lysed in RIPA buffer supplemented with protease and phosphatase inhibitors (Roche). Protein concentrations were determined using a BCA protein assay kit (Beyotime Biotechnology, Shanghai, China). Equal amounts of protein (20 μg per lane) were separated by SDS‐PAGE and transferred onto PVDF membranes. After blocking with 5% non‐fat milk for 1 h at room temperature, the membranes were incubated overnight at 4°C with primary antibodies against TNF‐α (1:1000; Santa Cruz Biotechnology, sc‐52746), IL‐6 (1:1000; Santa Cruz Biotechnology, sc‐57315), IL‐1β (1:2000; Abcam, ab234437), and GAPDH (1:5000; Proteintech, HRP‐60004). The membranes were then incubated with HRP‐conjugated secondary antibodies for 1 h at room temperature. Protein bands were visualized using an enhanced chemiluminescence (ECL) detection system (Beyotime Biotechnology) and quantified using ImageJ software (NIH, Bethesda, MD, USA). Protein expression levels were normalized to GAPDH. Histological analysis Kidney and pancreatic tissues were fixed in 4% neutral buffered formalin for 24 h, embedded in paraffin, and sectioned at 5 μm. Tissue sections were stained with hematoxylin and eosin (H&E) for histopathological evaluation. For renal histological analysis, 27 random microscopic fields containing glomeruli were selected from each group at ×40 magnification for assessment of glomerular morphology and mesangial matrix expansion. The mesangial matrix index (MMI) was quantified from H&E‐stained kidney sections using ImageJ software. For pancreatic histological analysis, 14 microscopic fields containing islets were selected from each group at ×40 magnification to evaluate islet morphology and islet density. Metabolomic analysis Fecal samples (50 mg) were extracted with 400 μL of methanol: water (4:1, v/v) containing 0.02 mg/mL L‐2‐chlorophenylalanine as an internal standard. Samples were homogenized using a Wonbio‐96c high‐throughput tissue crusher for 6 min at −10°C and 50 Hz, followed by ultrasonic extraction for 30 min at 5°C and 40 kHz. After incubation at −20°C for 30 min, the extracts were centrifuged at 13 000 g for 15 min at 4°C, and the supernatants were collected for analysis. Plasma samples were processed in parallel. A pooled quality control (QC) sample was prepared by mixing equal volumes of all samples and injected regularly during the analytical run. LC–MS/MS analysis was performed on a Thermo UHPLC‐Q Exactive HF‐X system equipped with an ACQUITY HSS T3 column (100 × 2.1 mm, 1.8 μm; Waters, USA). The mobile phases consisted of 0.1% formic acid in water/acetonitrile (95:5, v/v) (A) and 0.1% formic acid in acetonitrile/isopropanol/water (B). Mass spectrometric data were acquired in both positive and negative electrospray ionization modes using data‐dependent acquisition over a range of 70–1050 m/z. Raw data were processed using Progenesis QI for peak picking, alignment, and normalization. Metabolite annotation was performed by searching the HMDB, METLIN, and Majorbio databases. The data were analyzed using the free online Majorbio Cloud Platform (cloud.majorbio.com). Differential metabolites were defined as those with VIP >1 and p < 0.05. Microbiota analysis Microbiota profiling was performed on all human fecal samples, including 18 samples from healthy individuals and 14 from patients with T2D, to characterize donor‐associated microbial features. All eligible human fecal samples were included in this analysis. For the transplantation experiment, fecal samples from 12 healthy donors and 3 donors with T2D were selected and pooled within each group. The T2D donors used for transplantation were newly diagnosed or early‐stage patients who had not received glucose‐lowering medication before sample collection. In recipient mice, fecal samples were collected at weeks 2, 4, 6, 8, and 10 after transplantation for longitudinal analysis of gut microbial succession. Microbial genomic DNA was extracted from stool samples using the MagAttract PowerSoil Pro DNA Kit (Qiagen, Hilden, Germany) following the manufacturer's protocol. DNA integrity was evaluated by 1% agarose gel electrophoresis, while DNA concentration and purity were determined using a NanoDrop 2000 UV–vis spectrophotometer (Thermo Scientific, Wilmington, DE, USA). Subsequently, the V3‐V4 hypervariable region of the bacterial 16S rRNA gene was amplified using the universal primers 338F (5′‐ACTCCTACGGGAGGCAGCAG‐3′) and 806R (5′‐GGACTACHVGGGTWTCTAAT‐3′). The resulting PCR amplicons were purified via 2% agarose gel electrophoresis and quantified using a QuantiFluor™‐ST fluorometer (Promega, Madison, WI, USA). Finally, the purified amplicons were pooled in equimolar ratios and paired‐end sequenced on an Illumina MiSeq platform (Illumina, San Diego, CA, USA). Statistical analysis Microbiome and metabolome data were analyzed using the Majorbio Cloud platform (https://cloud.majorbio.com) and R software. Principal component analysis (PCA) was performed using the R package ropls (version 1.6.2). Longitudinal fasting blood glucose and IPGTT data were analyzed using mixed‐effects models to account for repeated measurements within individual mice. For IPGTT curves, post hoc comparisons between groups at individual time points were performed where appropriate, and the area under the curve (AUC) was analyzed separately. Longitudinal alpha diversity indices, including Sobs, ACE, Shannon, and Pielou_e, were analyzed using two‐way repeated‐measures ANOVA. Group comparisons for other data were conducted using either Student's t‐test or the Wilcoxon rank‐sum test, as indicated in the corresponding figure legends. Correlation analysis between microbial genera and metabolites was performed using Spearman's rank correlation and visualized using the R package pheatmap (version 1.0.8). The Microbial Dysbiosis Index (MDI) was calculated as the log10 ratio of the total relative abundance of genera increased in the T2D group to that of genera decreased in the T2D group. Bayesian network analysis was performed in R using the bnlearn package, with network structure learned by the hill‐climbing algorithm based on 500 bootstrap resamples and an averaged network threshold of 0.3, with metabolite‐to‐microbe edges prohibited by a blacklist constraint. RESULTS Human fecal samples from donors with type 2 diabetes show increased microbial dysbiosis Gut microbiota profiles were first analyzed in the full set of human fecal samples, including 18 samples from healthy donors and 14 samples from donors with T2D. Principal component analysis (PCA) at the genus level did not show clear separation between the two groups (Figure 1A). Alpha diversity, assessed by the Sobs index, was also comparable between groups (Figure 1B). The overall genus level community composition was broadly similar between healthy donors and donors with T2D (Figure 1C). In contrast, the Microbial Dysbiosis Index was significantly higher in the diabetic group (Figure 1D), indicating increased microbial dysbiosis despite limited global separation in community composition. Differential genus level analysis further showed enrichment of Ligilactobacillus, Megasphaera, DTU089, norank_f__Anaerovoracaceae and UCG‐008 in the diabetic group (Figure 1E). Gut microbial features in healthy donors and donors with T2D. (A) Principal component analysis (PCA) of gut microbial community structure in healthy donors and donors with T2D at genus level. (B) Comparison of the Sobs index between healthy donors and donors with T2D. (C) Genus level community composition in healthy donors and donors with T2D. (D) Comparison of the Microbial Dysbiosis Index (MDI) between healthy donors and donors with T2D. (E) Differential genera between healthy donors and donors with T2D at the genus level. Group differences in panel (A) were assessed by PERMANOVA. Statistical significance in panels (B) and (D) was analyzed using the Wilcoxon rank‐sum test. Differential genera in panel (E) were identified based on FDR‐adjusted p values. *p < 0.05, ** p < 0.01. Human gut microbiota colonizes germ‐free mice with partial retention of donor‐associated features Gut microbiota was first analyzed in the full human sample set to define the overall microbial features of healthy donors and donors with T2D. Based on these data, fecal samples from 12 healthy donors and 3 donors with T2D were selected for transplantation. All patients with T2D were included in the human microbiota analysis, whereas the T2D donors used for transplantation were newly diagnosed or early‐stage patients who had not received glucose‐lowering medication. The selected samples were pooled within each group and transplanted into germ‐free mice, followed by longitudinal fecal sampling and phenotypic analysis (Figure 2A). Healthy donors and HM mice shared 46 genera, whereas donors with T2D and DM mice shared 30 genera (Figure 2B,C). Heatmap analysis showed that HM and DM mice retained major microbial features of their corresponding donor groups (Figure 2D,E). The transplanted communities were also reshaped in the murine gut. At the phylum level, HM mice showed increased Bacillota and reduced Bacteroidota, Actinomycetota and Thermodesulfobacteriota relative to healthy donors (Figure 2F). DM mice showed increased Bacillota with reduced Bacteroidota and Actinomycetota relative to donors with T2D (Figure 2G). At the genus level, Escherichia‐Shigella and Blautia were increased in HM mice, whereas [Clostridium]_innocuum_group were increased in DM mice (Figure 2H,I). These results indicate that human gut microbiota successfully colonized germ‐free mice but underwent restructuring after engraftment. Colonization of human gut microbiota in germ‐free mice and donor–recipient microbial differences. (A) Experimental design for human microbiota transplantation and longitudinal analysis in recipient germ‐free mice. (B) Venn diagram showing shared and unique microbial features between healthy donors and HM mice. (C) Venn diagram showing shared and unique microbial features between donors with T2D and DM mice. (D) Heatmap showing major microbial features in healthy donors and HM mice. (E) Heatmap showing major microbial features in donors with T2D and DM mice. (F) Differential taxa between healthy donors and HM at the phylum level. (G) Differential taxa between donors with T2D and DM at the phylum level. (H) Differential genera between healthy donors and HM at the genus level. (I) Differential genera between donors with T2D and DM at the genus level. Statistical significance in panels (F–I) was analyzed using the Wilcoxon rank‐sum test with FDR correction. *p < 0.05, **p < 0.01, ***p < 0.001. Diabetes‐associated microbiota induces early diabetes‐like changes in recipient mice Recipient mice colonized with diabetes‐associated microbiota exhibited progressive metabolic abnormalities after transplantation. Fasting blood glucose was significantly higher in the DM group than in the HM group at week 4 (Figure S1A). Intraperitoneal glucose tolerance tests showed no clear difference between groups at week 2 (Figure 3A). Glucose intolerance became evident in DM mice at week 6 (Figure 3B) and was more pronounced at week 10, as indicated by a higher glucose peak, delayed glucose clearance, and an increased area under the curve (AUC) (Figures 3C and S1B). DM mice also showed reduced plasma insulin levels and increased non‐esterified fatty acid (NEFA) levels (Figure 3D,E). Both HOMA‐IR and HOMA‐β were lower in DM mice than in HM mice at week 10 (Figure 3F,G), together with reduced plasma insulin levels and impaired glucose tolerance. Early metabolic and renal phenotypes in recipient germ‐free mice. (A) Intraperitoneal glucose tolerance test (IPGTT) curves for HM and DM mice at week 2. (B) IPGTT curves for HM and DM mice at week 6. (C) IPGTT curves for HM and DM mice at week 10. (D) Plasma insulin levels in HM and DM mice. (E) Plasma non‐esterified fatty acid (NEFA) levels in HM and DM mice. (F) HOMA‐IR in HM and DM mice at week 10. (G) HOMA‐β in HM and DM mice at week 10. (H) Representative H&E staining of kidney sections from HM and DM mice. (I) Quantitative comparison of the Mesangial Matrix Index (MMI) in HM and DM mice. (J) Western blot images showing the protein expression of IL‐6, IL‐1β, and TNF‐α in kidney tissues. (K–M) Quantification of renal IL‐6 (K), TNF‐α (L), and IL‐1β (M) protein expression. (N–P) Relative mRNA expression levels of IL‐6 (N), TNF‐α (O), and IL‐1β (P) in kidney tissues. Statistical significance in panels (A–C) was determined using mixed‐effects models. Statistical significance in panels (D–G) and (I, K–P) was determined using an unpaired two‐tailed Student's t‐test. *p < 0.05, **p < 0.01, ***p < 0.001. Mild renal alterations were also detected in the DM group. Histological examination showed glomerular swelling and capillary narrowing in DM mice (Figure 3H), with a significant increase in mesangial matrix index (MMI) (Figure 3I). Renal expression of TNF‐α, IL‐6, and IL‐1β was also elevated at both the mRNA and protein levels (Figure 3J–P). Pancreatic histology did not show obvious structural abnormalities, and islet density was comparable between the two groups (Figure S1C). These results indicate that diabetes‐associated microbiota induced early glucose dysregulation accompanied by mild renal alterations in recipient mice. Transplanted diabetes‐associated microbiota progressively establishes a stable dysbiotic community The gut microbiota of recipient mice underwent progressive changes after transplantation. PCA at the genus level showed distinct temporal trajectories in the HM and DM groups, with significant separation among time points (Figure 4A). The Sobs and ACE indices, which reflect community richness, increased progressively from weeks 2 to 8 in both groups and declined slightly at week 10 (Figure 4B,C). Compared with the HM group, the DM group showed significantly lower Sobs and ACE indices at weeks 8 and 10, indicating reduced community richness at the later stage of colonization. The Shannon index, which reflects community diversity, showed no obvious difference between groups at the earlier time points, but was significantly lower in the DM group at week 10 (Figure 4D), indicating reduced community diversity at the late stage of colonization. The Pielou_e index, which reflects community evenness, changed only slightly at the earlier time points, but was also significantly lower in the DM group at week 10 (Figure 4E), indicating reduced community evenness at the late stage of colonization. The MDI was consistently higher in DM mice than in HM mice at weeks 2, 4, 6, 8, and 10 (Figure S1D), further supporting the progressive establishment of a dysbiotic microbial state in the DM group. These findings indicate that the transplanted diabetes‐associated microbiota progressively shifted toward a community with reduced richness, diversity, and evenness, accompanied by persistently elevated microbial dysbiosis after colonization. Temporal changes in gut microbial composition and alpha diversity in HM and DM mice. (A) PCA of gut microbiota community structure at weeks 2, 4, 6, 8, and 10. (B) Sobs index in HM and DM mice across weeks 2, 4, 6, 8, and 10. (C) ACE index in HM and DM mice across weeks 2, 4, 6, 8, and 10. (D) Shannon index in HM and DM mice across weeks 2, 4, 6, 8, and 10. (E) Pielou_e index in HM and DM mice across weeks 2, 4, 6, 8, and 10. Group differences in panel (A) were assessed by PERMANOVA. Longitudinal alpha diversity data in panels (B–E) were analyzed using two‐way repeated‐measures ANOVA, and statistical significance at individual time points represents post hoc comparisons between HM and DM mice. *p < 0.05, **p < 0.01. Genus‐level community profiles also showed distinct patterns of microbial assembly in the two groups. In HM mice, Escherichia‐Shigella became dominant by week 4 and increased from weeks 2 to 6, but declined during the later stage of colonization (Figure 5A). In DM mice, Hungatella emerged as a dominant genus by week 4 and continued to increase thereafter (Figure 5B). Differential genus‐level analysis at each time point further showed dynamic divergence between the two groups. At week 2, Escherichia‐Shigella was enriched in the HM group (Figure 5C). At weeks 4 and 6, Hungatella was increased and Romboutsia was reduced in DM mice (Figure 5D, E). At weeks 8 and 10, Eisenbergiella and Faecalitalea were reduced in DM mice (Figure 5F,G). These temporal changes were accompanied by progressive microbial divergence in the DM group. By the late stage of colonization, the diabetic microbiota had developed into a more stable dysbiotic state in recipient mice. Temporal genus‐level microbial dynamics in HM and DM mice after transplantation. (A) Community composition in HM mice across weeks 2, 4, 6, 8, and 10 at genus level. (B) Community composition in DM mice across weeks 2, 4, 6, 8, and 10 at genus level. (C) Differential genera between HM and DM mice at week 2. (D) Differential genera between HM and DM mice at week 4. (E) Differential genera between HM and DM mice at week 6. (F) Differential genera between HM and DM mice at week 8. (G) Differential genera between HM and DM mice at week 10. Statistical significance in panels (C–G) was analyzed using the Wilcoxon rank‐sum test with FDR correction. *p < 0.05, **p < 0.01. Diabetes‐associated microbiota reshapes fecal and systemic metabolism in recipient mice Metabolomic profiling showed broad metabolic differences between the HM and DM groups in both plasma and feces. A total of 89 differentially abundant metabolites were identified in plasma and 729 were identified in feces (Figure 6A,B). Six metabolites were shared between the two compartments. Among these shared metabolites, 3‐ethylphenyl sulfate, cholic acid, and L‐Dopa were significantly reduced in both feces and plasma from DM mice compared with HM mice (Figure 6C,D). N‐Acetylleucine and dodecanedioic acid were increased in both compartments in the DM group. Acetyl‐DL‐leucine showed opposite changes in the two compartments, with higher levels in plasma and lower levels in feces. These results show that diabetes‐associated microbiota induced coordinated metabolic alterations in both intestinal and circulating compartments. Fecal and plasma metabolite differences in HM and DM mice. (A) Differential metabolites identified in plasma from HM and DM mice. (B) Differential metabolites identified in fecal samples from HM and DM mice. (C) Relative abundance of shared differential metabolites in plasma. (D) Relative abundance of shared differential metabolites in feces. Statistical significance thresholds in panels (A) and (B) were screened using VIP >1 and p < 0.05. Statistical significance in panels (C) and (D) was analyzed using an unpaired two‐tailed Student's t‐test. *p < 0.05, **p < 0.01, ***p < 0.001. L‐Dopa shows consistent associations with altered genera in the microbe–metabolite network Correlation analysis at week 10 identified broad associations between the six shared differential metabolites and multiple altered genera in both plasma‐ and feces‐based datasets (Figure 7A,B). In both feces and plasma, several shared differential metabolites showed significant correlations with multiple altered genera. Among these metabolites, L‐Dopa showed the most consistent associations with altered genera across both plasma and feces. Significant negative associations were observed between L‐Dopa and Ruminococcus as well as Sellimonas, whereas positive associations were observed with Erysipelatoclostridium and Romboutsia. Network analysis further showed close links between L‐Dopa and altered genera in both plasma‐ and feces‐based networks, particularly the association with Ruminococcus (Figure 7C,D). These results suggest that L‐Dopa is a prominent metabolite associated with diabetes‐related microbial alterations in recipient mice. Correlation analysis between gut microbiota and shared metabolites in plasma and feces. (A) Correlation analysis between gut microbiota and plasma metabolites. (B) Correlation analysis between gut microbiota and fecal metabolites. (C) Network analysis of microbial‐metabolic interactions in plasma. (D) Network analysis of microbial‐metabolic interactions in feces. Correlations in panels (A) and (B) were assessed using Spearman's rank correlation. Network relationships in panels (C) and (D) were inferred using Bayesian network structure learning with the hill‐climbing algorithm in bnlearn, based on 500 bootstrap resamples and an averaged network threshold of 0.3. *p < 0.05, **p < 0.01, ***p < 0.001. DISCUSSION Gut microbiota from patients with T2D induced early diabetes‐like changes in germ‐free recipient mice. Mice receiving diabetes‐associated microbiota developed progressive glucose intolerance, reduced insulin levels, and mild renal alterations. Microbial divergence became evident before stable metabolic dysfunction and paralleled the progression of glucose intolerance. At the endpoint, altered microbial genera showed significant associations with changes in fecal and plasma metabolites. These findings support a contributory role of diabetes‐associated microbiota in the early development of metabolic dysfunction. We used human fecal microbiota transplantation in germ‐free mice and followed both microbial succession and host phenotypes over 10 weeks. Similar approaches have been used previously in metabolic disease research, but studies focused specifically on the early phase of T2D remain limited. 26 , 27 A germ‐free background reduces pre‐existing microbial interference and allows the effects of transplanted communities to be assessed more directly. 22 The longitudinal design further allowed us to examine the temporal sequence of microbial and metabolic changes during disease development. One of the main findings of this study was that microbial divergence became evident before stable metabolic dysfunction was established in recipient mice. In the DM group, glucose intolerance became apparent from week 6 and was more pronounced by week 10, whereas changes in microbial composition were already detectable at earlier time points. This temporal pattern suggests that diabetes‐associated microbial remodeling may precede overt metabolic impairment rather than simply being secondary to it. Previous experimental studies have shown that microbiota‐related alterations can contribute to host metabolic abnormalities in controlled models. 28 , 29 A similar temporal relationship between microbial alteration and disease progression has also been reported in other disease models. 30 In the present model, these findings support a close association between microbial succession and early metabolic deterioration. At the microbial level, the diabetes‐associated community progressively developed into a persistently dysbiotic state. In HM mice, dominant genera gradually declined and the composition of the community became more even over time. In DM mice, a limited number of genera showed continued expansion, together with depletion of others. Hungatella was prominently during the middle phase of colonization and has also been reported to be increased in diabetes‐related phenotypes. 31 , 32 Reduced abundance of Eisenbergiella and Faecalitalea became more evident at later stages; both genera have been linked to metabolic health in previous studies. 33 , 34 , 35 These findings suggest that the diabetes‐associated microbiota did not persist in recipient mice as a static community, but instead underwent ecological restructuring and developed into a stable dysbiotic configuration in the murine gut. This interpretation is consistent with previous reports showing that human fecal microbiota can engraft in germ‐free mice but undergo host‐dependent remodeling after transplantation. 36 The microbial alterations observed in recipient mice were also accompanied by changes in both fecal and plasma metabolites at the endpoint. Among the shared differential metabolites, cholic acid and L‐Dopa were of particular interest, and both were reduced in DM mice. Cholic acid is closely involved in host glucose and lipid homeostasis, and altered bile acid metabolism has been repeatedly linked to metabolic disease. 37 , 38 As an intermediate in catecholamine biosynthesis, L‐Dopa is related not only to neurotransmitter metabolism but also to peripheral metabolic regulation. 39 , 40 In our study, the consistent decrease in L‐Dopa in both feces and plasma suggests that it may represent a notable metabolic feature associated with early glucose dysregulation after transplantation of diabetic microbiota. L‐Dopa may originate from dietary sources and may also be synthesized by gut microorganisms. 41 , 42 However, because the present metabolomic analysis was untargeted and based on relative abundance, the absolute level and functional significance of this change remain to be determined. Among these metabolites, L‐Dopa showed the most consistent associations across the two compartments. Negative correlations were observed between L‐Dopa and Ruminococcus as well as Sellimonas, whereas positive correlations were observed with Erysipelatoclostridium and Romboutsia. The association with Sellimonas is also noteworthy, as previous work has suggested an association between this genus and T2D. 43 A close association between L‐Dopa and Ruminococcus was also apparent in the network structure, in agreement with previous observations in other diseases. 44 , 45 These findings suggest that L‐Dopa, particularly its association with Ruminococcus, represents a notable microbe–metabolite feature in this model. Previous work in Parkinson's disease‐related settings has shown that gut bacteria can directly metabolize L‐Dopa in the intestine, thereby influencing its local transformation and systemic availability. 41 , 46 In this context, the altered abundance of L‐Dopa observed in the present model may reflect broader host–microbiota metabolic interactions during diabetes‐associated microbial remodeling. At the same time, the current data are more consistent with an association‐based interpretation, and the precise functional role of L‐Dopa in this process will require further mechanistic investigation. Several limitations should be considered when interpreting the present findings. First, although microbiota profiling was performed on the full set of human fecal samples, the donor material used for transplantation was derived from only a subset of those samples. In particular, the T2D donors used for transplantation were newly diagnosed or early‐stage patients who had not received glucose‐lowering medication. This may limit the extent to which the donor inoculum reflects the full heterogeneity of the broader T2D population. Second, donor material used for transplantation was pooled within each group before administration to recipient mice, which limited donor‐level inference in the germ‐free mouse experiment. Pooling may also have influenced the magnitude of the transmitted phenotype. This approach may have attenuated extreme donor‐specific effects while allowing shared disease‐associated microbial features to be more consistently represented in recipient mice. Therefore, the pooled‐donor design was more suitable for evaluating whether T2D‐associated human microbiota, as a whole, could induce an early diabetes‐like phenotype than for assessing the effects of individual donors. Third, metabolomic profiling was performed only at the endpoint, which did not allow direct longitudinal comparison with the temporal microbial changes observed after transplantation. Fourth, the associations identified between altered genera and shared metabolites were correlative and do not establish causality. Further studies using defined microbial communities, mono‐colonization, or targeted metabolite intervention will be needed to determine whether specific taxa or metabolites play functional roles in disease progression. In summary, this study establishes a germ‐free mouse model of early diabetes‐like changes induced by gut microbiota from patients with T2D. Diabetes‐associated microbial remodeling became evident before stable metabolic dysfunction and progressed together with worsening glucose intolerance. Endpoint metabolomic analysis identified shared alterations in fecal and plasma metabolites and revealed L‐Dopa‐associated microbe–metabolite associations. These findings provide an experimental framework for studying microbiota‐associated events during the early stage of T2D‐related metabolic deterioration. AUTHOR CONTRIBUTIONS Yun Yang: Resources. Wei Liu: Data curation. Xiaoliang Jiang: Writing – review and editing. Hua Zhu: Investigation; methodology. Jianghao Feng: Methodology; formal analysis; investigation; data curation; writing – original draft. Xing Liu: Investigation; methodology. Zhiwei Yang: Conceptualization; data curation; formal analysis; validation; writing – review and editing; supervision. FUNDING INFORMATION This research was funded by the National Key Research and Development Program of China (2022YFF0710600), Chinese Academy of Medical Sciences Innovation Fund for Medical Sciences (2021‐1‐I2M‐035), Non‐profit Central Research Institute Fund of Chinese Academy of Medical Sciences (2023‐PT180‐01), National High‐level Innovation and Entrepreneurship Talents Program for Young Backbone Talents (SpringBuds Project) (20250333). CONFLICT OF INTEREST STATEMENT The authors have no conflict of interest related to this manuscript. ETHICS STATEMENT All animal protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of the Institute of Laboratory Animal Sciences, Chinese Academy of Medical Sciences (CAMS) and Peking Union Medical College (PUMC) (approval no.: ZH23003). Supporting information Figure S1.