Systematic comparison of seven animal protein sources reveals more favorable glucose homeostasis in mice fed duck protein compared with pork protein, associated with gut microbiota and secondary bile acid metabolism

Abstract

Animal protein sources may differentially influence glucose homeostasis, yet the underlying mechanisms remain unclear. We investigated how casein, pork, beef, mutton, chicken, duck, and goose proteins differentially affected glucose homeostasis in mice and their associations with gut microbiota and bile acids. Mice fed duck protein exhibited more favorable glucose tolerance and insulin sensitivity than those fed pork protein. These differences were accompanied by coordinated alterations in gut microbiota and bile acid profiles, including enrichment of Clostridium, increased abundance of the baiE gene, and elevated secondary bile acids. Compared with pork protein, duck protein intake increased ileal fibroblast growth factor 15 expression and portal active glucagon-like peptide-1 concentrations, upregulated adipose thermogenic and lipid oxidation genes, while downregulating hepatic gluconeogenic genes. Correlation analysis revealed associations between these genes and glucose metabolic parameters. Collectively, alterations in the microbiota–bile acid axis may contribute to the favorable glucose homeostasis observed in duck protein-fed mice.

Graphical abstract

Highlights

• •Different meat protein sources differentially regulate glucose metabolism in mice.
• •Duck protein shows better glucose tolerance and insulin sensitivity than pork.
• •Duck protein enriches Clostridium and increases secondary bile acid metabolism.
• •Duck protein increases ileal FGF15 and GLP-1 responses.
• •Duck protein lowers hepatic gluconeogenic and raises adipose lipid oxidation genes.

1. Introduction

Impaired glucose homeostasis, characterized by glucose intolerance and insulin resistance, underlies the development of type 2 diabetes, which affects more than 11% of the adult population worldwide (Genitsaridi et al., 2025). Dietary protein has emerged as an important dietary factor influencing glucose homeostasis (Thomsen et al., 2022). Meat is a major source of high-quality protein, and its global consumption has nearly quintupled since the 1960s, with further growth projected to 2034 (González et al., 2020). However, accumulating evidence suggests that the associations between meat consumption and type 2 diabetes differ by meat source: excessive red meat intake is generally linked to an increased risk (C. Li et al., 2024; O'Connor et al., 2021), whereas the effects of poultry remain inconclusive (Damigou et al., 2022; Du et al., 2020). These differences are often attributed to potential effects of non-protein components, such as heme iron and processing-related compounds (Kim et al., 2015), while the role of proteins from different sources remains unclear. Addressing this question may improve our understanding of how meat protein source influences glucose homeostasis and the biological pathways associated with these effects.

The metabolic effects of different protein sources are increasingly recognized to be mediated, at least in part, through interactions with the gut microbiota (Blakeley-Ruiz et al., 2025; Nychyk et al., 2021). Different meat sources have been associated with distinct gut microbial composition, which may consequently influence microbial metabolite production (Larsson et al., 2025). Among these metabolites, secondary bile acids are of particular interest. They are generated by bacterial transformation of primary bile acids, primarily through 7α/β-dehydroxylation mediated by gut bacteria containing bile acid-inducible (bai) genes, such as Clostridium and Eubacterium (Collins et al., 2023; T. Li & Chiang, 2023; Wahlström et al., 2016). These secondary bile acids regulate glucose homeostasis through activation of Takeda G protein-coupled receptor 5 (TGR5) and farnesoid X receptor (FXR), which promote glucagon-like peptide-1 (GLP-1) secretion and fibroblast growth factor 15/19 (FGF15/19) expression, respectively (Ahmad & Haeusler, 2019; Potthoff et al., 2011; Thomas et al., 2009).

Emerging evidence suggests that different protein sources are associated with distinct gut microbiota composition and bile acid metabolism. For example, cecal delivery of casein hydrolysate increases the abundance of Eubacterium and enhances the production of deoxycholic acid (DCA) and lithocholic acid (LCA) in a porcine model (Pi et al., 2020). In addition, whey protein intake enhances ileal FXR expression while suppressing intestinal TGR5/GLP-1 signaling compared with soy protein β-conglycinin in mice (Nonogaki & Kaji, 2023). Collectively, these findings suggest that dietary proteins may differentially influence gut microbiota and bile acid metabolism. However, direct evidence linking different meat protein sources with gut microbiota, bile acid metabolism, and glucose homeostasis remains limited.

Therefore, this study evaluated the effects of several common animal protein sources (casein, pork, beef, mutton, chicken, duck, and goose) on glucose homeostasis in mice. We hypothesized that different dietary protein sources would differentially shape gut microbial communities and bile acid metabolism, which may contribute to differences in host glucose homeostasis. We found that compared with pork protein, duck protein was associated with enhanced gut microbial bile acid transformation and improved glucose homeostasis. These findings provide new insight into the potential role of the microbiota–bile acid axis in protein source-dependent differences in glucose homeostasis.

2. Materials and methods

2.1. Diet preparation

Meat proteins were extracted from the longissimus dorsi muscle of pork, beef, and mutton, as well as the pectoralis major muscle of chicken, duck, and goose, purchased from a local market in Nanjing, China (Xie et al., 2022). Briefly, after removal of visible fat and connective tissue, the meat was cut into small pieces and heated in a water bath at 72 °C until the internal temperature reached 70 ± 2 °C. After cooling, the samples were minced, freeze-dried, and subsequently defatted 4 times with 5 volumes of n-hexane. After evaporation, the resulting meat powders were passed through a 30-mesh screen and used as protein extracts. Crude protein and fat contents were measured by Kjeldahl nitrogen analysis (Kjeltec 8400; Foss, Hilleroed, Denmark) and Soxhlet extraction (E-816; Buchi, Flawil, Switzerland), respectively. The fat content of each defatted protein extract was less than 1%, while protein content exceeded 90%, with no significant differences among groups (Supplemental Fig. S1). Amino acid composition was measured using an automatic amino acid analyser (L-8900; Hitachi, Tokyo, Japan) (Yuan et al., 2023) and shown in Supplemental Table S1. Experimental diets were formulated based on AIN-93G (Reeves et al., 1993), with casein replaced by the respective meat protein extracts under isonitrogenous and isocaloric conditions (Supplemental Table S2).

2.2. Animal experiments and sample collection

All animal experiments were approved by the Animal Care and Use Committee of Nanjing Agricultural University (approval No. SYXK(Su)2021-0086). Six-week-old male C57BL/6J mice were obtained from Hangzhou Ziyuan Laboratory Animal Technology Co., Ltd. (Zhejiang, China). Mice were group-housed (3 per cage) in standard plastic cages under specific pathogen-free conditions (22 ± 2 °C, 12 h light/ dark cycle) with free access to food and water. Cages and wood shaving bedding were changed every week. After a 2-week acclimation, mice were randomly assigned to 7 groups (n = 9) and provided with diets containing casein, pork, beef, mutton, chicken, duck, or goose proteins for 10 weeks. Mice were marked for individual identification throughout the experimental period. Body weight and food intake were recorded every 2 days. Every two weeks, mice were fasted for 6 h beginning at 08:00, after which fasting blood glucose was measured using a handheld glucometer (AQ+; Sinocare, Changsha, China). Fecal samples were collected biweekly for 24 h for total bile acid (TBA) measurement using a commercial kit (E003–1-1; Jiancheng, Nanjing, China). Also, feces were collected for 72 h in metabolic cages at week 7. Dietary and fecal nitrogen were determined using the Kjeldahl method, while lipid content was quantified using the Soxhlet method. Apparent digestibility of nitrogen and fat was calculated using the formula: digestibility (%) = (intake (mg) − fecal output (mg))/(intake (mg)) × 100. Both intake and fecal output were determined on a dry matter basis. At the end of the test period, mice were anesthetized using urethane (1 g/kg body weight). Portal blood was collected into tubes containing heparin, aprotinin, and sitagliptin (final concentrations of 50 IU/mL, 500 KIU/mL, and 10 mmol/L, respectively) using an Omnican Insulin Syringe U40 (0.3 × 8 mm; B. Braun, Melsungen, Germany). Plasma was then separated by centrifugation at 2300 ×g for 10 min at 4 °C. The liver, adipose tissues, jejunal and ileal mucosa, cecal and colonic contents were then collected, immediately frozen in liquid nitrogen, and stored at −80 °C. Food deprivation was not performed before sacrifice.

2.3. Oral glucose tolerance test (OGTT) and insulin tolerance test (ITT)

OGTT and ITT were performed at weeks 8 and 9, respectively. For OGTT, mice were fasted overnight and administered a glucose solution (2 g/kg; 20% w/v in saline). Blood glucose concentrations were measured in the tail vein at 0, 15, 30, 60, and 120 min post-gavage using a handheld glucometer. Fasting insulin was measured using an ELISA Kit (E-EL-M1382; Elabscience, Wuhan, China). The homeostatic model assessment of insulin resistance (HOMA-IR) was calculated as follows: HOMA-IR = (fasting glucose (mmol/L) × fasting insulin (μU/mL))/22.5. For ITT, mice were fasted for 4 h and received recombinant human insulin (0.75 U/kg; 0.1 U/mL in saline) via intraperitoneal injection. Blood glucose was measured in the tail vein at 0, 15, 30, and 60 min following injection.

2.4. Biochemical analysis

Briefly, liver samples were rinsed with cold saline. Hepatic glycogen was measured using a commercial kit (A043-1-1; Jiancheng). Hepatic glucose-6-phosphatase (G6pase) was measured using a mouse ELISA kit (ml037561; Shanghai Enzyme-linked Biotechnology, Shanghai, China). The concentrations of active GLP-1 in portal plasma were measured using an ELISA kit (27700; Immuno-Biological Laboratories, Gunma, Japan).

2.5. Bile acid analysis

Bile acid analysis was adapted from a previous study with modifications (Bai et al., 2022). Briefly, 50 μL of portal plasma were mixed with 200 μL of acetonitrile/methanol (1:1), and 25 mg of colonic contents were mixed with 500 μL of acetonitrile/methanol/water (2:2:1). Both extraction solvents were supplemented with isotope-labeled bile acids as internal standards, including lithocholic acid‑d₄ (LCA-d4), deoxycholic acid‑d₆ (DCA-d6), glycocholic acid‑d₄ (GCA-d4), taurochenodeoxycholic acid‑d₄ (TCDCA-d4), lithocholic acid‑d₄–3-sulfate (LCA-d4–3S), and glycocholic acid‑d₄–3-sulfate (GCA-d4–3S). Portal plasma samples were sonicated for 10 min. Colonic content samples were homogenized at 35 Hz for 4 min, followed by sonication for 5 min. The homogenization-sonication procedure was performed in triplicate. Subsequently, both plasma and colonic content extracts were incubated at −40 °C for 1 h and centrifuged at 12000 ×g at 4 °C for 15 min. Supernatants were analyzed by ultra-high-performance liquid chromatography-mass spectrometry (UHPLC-MS).

UHPLC-MS was performed using a Vanquish UHPLC coupled to an Orbitrap Exploris 120 MS (Thermo Fisher Scientific, San Jose, CA, USA). Chromatographic separation was achieved on a BEH C18 column (150 × 2.1 mm, 1.7 μm; Waters, Milford, MA, USA) with 5 mmol/L ammonium acetate (A) and acetonitrile (B) as the mobile phases at 0.3 mL/min. The gradient elution was programmed as follows: 20% B for 8.5 min, increased to 26% B over 6.5 min, then to 46% B over 3.5 min, followed by a rapid increase to 99% B in 1.5 min with a 3 min hold. Then returned to 20% B in 0.4 min and equilibrated for 3.6 min. MS was conducted as follows: electrospray ionization in the negative mode; acquisition mode: parallel reaction monitoring (PRM); spray voltage: +3.5/−3.2 kV; sheath gas flow rate: 40 arb; aux gas flow rate: 15 arb; aux gas temperature: 350 °C; capillary temperature: 320 °C. Bile acid classification and abbreviations are listed in Supplemental Table S3.

2.6. Gut microbiota 16S rRNA sequencing analysis

Total microbial DNA was extracted from cecal contents using the Fecal Genome DNA Extraction Kit (AU46111–96; BioTeke, Wuxi, China). The V3–V4 region of the microbial 16S rRNA gene was amplified using 50 ng of DNA as template with primers 341F (5′-CCTACGGGNGGCWGCAG-3′) and 805R (5′-GACTACHVGGGTATCTAATCC-3′) (Logue et al., 2016). PCR products were purified with AMPure XT Beads (Beckman Coulter Genomics, Danvers, MA, USA) and quantified using a Qubit fluorometer (Thermo Fisher Scientific). Amplicons were sequenced using an Agilent 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA) with Illumina library quantitative kits (Kapa Biosciences, Woburn, MA, USA), pooled, and sequenced on an Illumina NovaSeq 6000 platform with paired-end 250 bp reads.

After quality filtration and assembly, sequences were denoised into amplicon sequence variants (ASVs) and analyzed using QIIME 2 (V.2019.7) with the SILVA and NT-16S reference databases. α-Diversity was assessed using the Kruskal-Wallis test. Bray-Curtis distance matrices were used to characterize β-diversity, with community structure visualized by principal coordinates analysis (PCoA). Differences in microbial community composition were tested using permutational multivariate analysis of variance (PERMANOVA) with 9,999 permutations. The taxa abundances were assessed at the phylum and genus levels. Differential features were identified using the Kruskal-Wallis test (P < 0.05). Potential biomarkers were analyzed using linear discriminant analysis effect size (LEfSe), with a linear discriminant analysis (LDA) score > 4 and P < 0.05. Given that mice were group-housed, individual-level microbiota analyses may be influenced by coprophagy. To verify the main findings, β-diversity and Clostridium relative abundance were reanalyzed at the cage level (n = 3 cages per group) by combining ASV abundance profiles from mice housed within the same cage.

2.7. Real-time quantitative PCR (qPCR)

Total RNA was extracted from tissue samples using FastPure Complex Tissue/Cell Total RNA Isolation Kit (Vazyme, Nanjing, China), followed by reverse transcription into cDNA using HiScript III RT SuperMix for qPCR (+gDNA wiper; Vazyme). qPCR was performed using ChamQ SYBR qPCR Master Mix (Low ROX Premixed; Vazyme) on a QuantStudio 6 flex real-time PCR system (Applied Biosystems, Waltham, MA, USA). Relative gene expression was quantified using the 2^-△△Ct method with glyceraldehyde-3-phosphate dehydrogenase (Gapdh) as the internal reference gene. Gapdh Ct values did not differ significantly among the dietary groups. The primers were synthesized by Sangon Biotech (Shanghai, China) and listed in Supplemental Table S4.

The abundance of the functional gene baiE, which encodes the bile acid 7α-dehydratase, was quantified according to a previously reported method with minor modifications (Hernández-Rocha et al., 2021). Microbial genomic DNA was extracted from cecal contents using the TIANamp Stool DNA Kit (DP328; TIANGEN, Beijing, China). qPCR was carried out as described above using DNA templates (5 ng/μL) and primers targeting Clostridium scindens ATCC 35704. The relative abundance of baiE was determined using the 2^-△△Ct method and normalized to the total bacterial 16S rRNA gene.

2.8. Western blotting

  Briefly, ileal mucosa (50 mg) was homogenized in 500 μL of RIPA lysis buffer (R0010; Solarbio, Beijing, China) supplemented with 1 mmol/L PMSF and disrupted by ultrasonication. Following centrifugation at 12,000g for 10 min at 4 °C, the supernatant was collected for protein quantification using a BCA Protein Assay Kit (P0010; Beyotime, Shanghai, China). Aliquots containing 50 μg of protein were separated on 4–20% SurePAGE™ precast gels (M00655; GenScript, Nanjing, China) and electrotransferred onto PVDF membranes. After blocking with 5% skim milk, the membranes were incubated overnight at 4 °C with primary antibodies against FGF15 (1:1000, ab319994; Abcam, Cambridge, UK) and GAPDH (1:70,000, A19056; Abclonal, Wuhan, China). After washing, the membranes were incubated with horseradish peroxidase-conjugated secondary antibody (1:5000, AS014; ABclonal) for 2 h at room temperature. Immunoreactive bands were visualized using an enhanced chemiluminescence reagent (34580; Thermo Fisher Scientific) and captured with a protein blot imaging system (Amersham ImageQuant 800, Cytiva, Tokyo, Japan). Band intensities were quantified using ImageJ (version 1.53a) and normalized to GAPDH.

2.9. Statistical analysis

Data are shown as mean ± SEM. For comparisons among multiple groups, one-way ANOVA with Tukey-Kramer post hoc test or, where appropriate, the Kruskal-Wallis test followed by Dunn's test with Benjamini-Hochberg correction was applied. Blood glucose measurements were evaluated using repeated measures ANOVA, with group, time, and their interaction as factors; when a significant interaction was observed, Tukey-Kramer's test was applied. The correlations among significantly altered bile acids and bacterial abundance, gene expression levels, and physiological indicators were evaluated using Spearman's correlation analysis, with the Benjamini-Hochberg method applied to control the false discovery rate for multiple comparisons.

3. Results

3.1. Effects of dietary proteins on growth performance and glucose metabolism

We evaluated the metabolic effects of different protein sources in mice (Fig. 1a). Food intake did not differ among groups (Fig. 1b). From day 34 onward, mice fed pork, beef, and mutton protein diets showed significantly higher body weights compared with casein-fed controls, whereas no differences were observed in mice fed chicken, duck, and goose protein diets (Fig. 1c). Organ weights were largely comparable, except for increased epididymal white adipose tissue (eWAT) mass in the pork protein-fed group relative to casein- and duck protein-fed groups (Supplemental Table S5). In addition, mice fed duck protein showed higher apparent nitrogen digestibility than those fed casein and goose protein, and higher apparent fat digestibility than those fed beef protein (Fig. 1d, e).

Fig. 1.

Different dietary protein feeding for 10 weeks alters body weight and glucose metabolism without affecting food intake in mice. (a) Schematic of the experimental design. (b) Cumulative food intake and (c) body weight during the feeding period. Apparent (d) nitrogen and (e) fat digestibility, calculated as follows: (intake - fecal output)/(intake) × 100. (f–h) Fasting blood glucose measured after a 6 h fast every two weeks. (i) Overnight fasting blood glucose (0 min before glucose administration during the OGTT). (j) Glucose response curves in the oral glucose tolerance test (OGTT) after 8 weeks of feeding. (k) Cumulative area under the curve (AUC) of the OGTT. (l) Glucose response curves in the insulin tolerance test (ITT) after 9 weeks of feeding. (m) AUC of ITT. (n) Plasma levels of fasting insulin after 8 weeks of feeding. (o) Homeostatic model assessment for insulin resistance (HOMA-IR). (p) Liver levels of glycogen. Data are expressed as mean ± SEM (n = 9). Different symbols and letters indicate significant differences among groups (P < 0.05, one-way or repeated-measures ANOVA followed by Tukey-Kramer's test). † pork vs casein; †† beef vs casein; ††† mutton vs casein. $ duck vs chicken; $$ casein vs chicken; # duck vs beef; ## duck vs pork; ### casein vs pork; #### goose vs pork. & duck vs pork; ‡ duck vs casein; § duck vs goose.

Fasting blood glucose was monitored every two weeks after 6 h of fasting during the dietary intervention. No significant difference was observed during the early intervention period. From week 6 onward, fasting blood glucose levels were lower in the duck protein group than in the pork protein group (Fig. 1f–i). At week 8, the duck protein group showed a tendency toward lower fasting blood glucose compared with the pork protein group (P = 0.06; Fig. 1i). The OGTT suggested differences in glucose tolerance among dietary protein groups. Blood glucose levels at 15 and 120 min were significantly lower in the duck protein group compared with the pork protein group, resulting in a reduced glucose AUC (Fig. 1i–k). Consistently, the ITT suggested greater insulin sensitivity in the duck protein group relative to the pork protein group, as evidenced by significantly lower relative blood glucose levels at 15, 30, and 60 min following insulin injection, along with a decreased glucose AUC (Fig. 1l, m). Neither duck nor pork differed significantly from casein in OGTT or ITT AUC. Fasting insulin was lower in the casein, beef, mutton, and duck protein groups compared with the chicken protein group (Fig. 1n). A trend toward reduced HOMA-IR was observed in mice fed duck protein compared with chicken protein (Fig. 1o). Hepatic glycogen content was significantly lower in mice fed the beef, chicken, duck, and goose proteins compared with those fed pork protein (Fig. 1p). Together, these results suggest that different dietary protein sources exerted divergent effects on glucose homeostasis, with mice fed duck protein showing more favorable glucose tolerance and insulin sensitivity than those fed pork protein.

3.2. Dietary proteins differentially alter intestinal and portal plasma bile acid profiles

To investigate the effect of dietary proteins on bile acid metabolism, fecal TBA levels were monitored biweekly, and bile acid profiles in colonic contents and portal plasma were analyzed at the end of the feeding period. Compared with mice fed other meat proteins, those receiving duck protein exhibited significantly higher fecal TBA levels from week 4 onward (Fig. 2a). At the end of the experiment, the colonic contents of mice in the duck protein group contained significantly higher concentrations of total bile acids, secondary bile acids, and 12α-hydroxylated (12αOH) bile acids than those of mice fed casein, pork, mutton, or chicken protein (Fig. 2b). Specifically, duck protein feeding increased DCA, isoDCA, and 12-oxoLCA (12oLCA) among 12αOH bile acids, as well as hyodeoxycholic acid (HDCA), chenodeoxycholic acid (CDCA), and LCA among non-12αOH bile acids (Fig. 2c, d). Consistently, the DCA/CA ratio, an index of microbial 7α-dehydroxylation, was significantly higher in the duck protein group than in the pork, beef, mutton, and chicken protein groups (Supplemental Fig. S2).

Fig. 2.

Alterations in bile acid profiles in mice fed different protein sources. (a) Changes in fecal total bile acid (TBA) levels during the feeding period. (b) Total concentrations of bile acids, primary bile acids, secondary bile acids, 12α-hydroxylated (12αOH) bile acids, and non-12αOH bile acids in colonic contents. Concentrations of individual (c) 12αOH and (d) non-12αOH bile acids in colonic contents. (e) Total concentrations of bile acids, primary bile acids, secondary bile acids, 12αOH bile acids, and non-12αOH bile acids in portal plasma. Concentrations of individual (f) 12αOH and (g) non-12αOH bile acids in portal plasma. Abbreviations: DCA, deoxycholic acid; CA3S, cholic acid-3-sulfate; CA, cholic acid; ACA, allocholic acid; isoDCA, isodeoxycholic acid; 12oLCA, 12-oxolithocholic acid; 7oDCA, 7-oxodeoxycholic acid; UCA, ursocholic acid; ωMCA, ω-muricholic acid; βMCA, β-muricholic acid; αMCA, α-muricholic acid; HDCA, hyodeoxycholic acid; CDCA, chenodeoxycholic acid; LCA, lithocholic acid; UDCA, ursodeoxycholic acid. Bile acids that account for <1% of the sum are not shown in the main figures but are comprehensively listed in Supplemental Table S6 and S7. Data are expressed as mean ± SEM (n = 9). Different letters indicate significant differences among groups (P < 0.05, one-way ANOVA followed by Tukey-Kramer's test or Kruskal-Wallis test followed by Dunn's test with Benjamini-Hochberg correction).

Dietary protein source also changed the bile acid profiles in portal plasma (Fig. 2e–g). The duck protein group showed significantly higher concentrations of total bile acids, primary bile acids, secondary bile acids, and 12αOH bile acids than other protein groups, while non-12αOH bile acids were also increased relative to the casein, pork, mutton, and chicken groups (Fig. 2e). At the individual bile acid level, portal concentrations of DCA, 12oLCA, and HDCA were significantly higher in the duck protein group compared with other protein groups (Fig. 2f, g). Collectively, these results suggest that dietary protein source differentially shapes bile acid composition in both the intestine and portal circulation, with the duck protein group exhibiting increased levels of several microbiota-derived secondary bile acids.

3.3. Dietary proteins differentially shape gut microbiota composition

We next investigated the effect of ingesting different proteins on gut microbiota composition. Mice fed duck protein showed higher Chao1 and Shannon α-diversity indices compared with those fed pork, mutton, and chicken proteins (Fig. 3a, b). PCoA analysis further suggested significant differences in the gut microbial β-diversity among the groups (Fig. 3c). PERMANOVA showed that the β-diversity of the duck protein group differed significantly from that of other proteins (Supplemental Table S8). At the phylum level, the gut microbial community across all groups was primarily composed of Firmicutes, Bacteroidota, Desulfobacterota, and Verrucomicrobiota. (Fig. 3d, Supplemental Fig. S3a–d). At the genus level, Desulfovibrionaceae_unclassified, Akkermansia, Muribaculaceae_unclassified, and Lachnospiraceae_NK4A136_group were abundant (Fig. 3e, Supplemental Fig. S3e–h). To further identify key microbial taxa, LEfSe analysis was conducted (Fig. 3f). Casein-fed mice were enriched with biomarkers including Desulfovibrionaceae_unclassified, Ileibacterium, Lachnospiraceae_unclassified, Clostridiales_unclassified, and Parabacteroides; the beef protein group with Clostridia_UCG-014_unclassified and Dorea; the chicken protein group with Akkermansia and Faecalibaculum; and the duck protein group was characterized by Muribaculaceae_unclassified, Desulfovibrio, and Clostridium. At the cage level, separation of microbial communities among dietary groups and differences in the relative abundance of Clostridium remained consistent with the individual mouse-level analysis (Supplemental Fig. S4). Collectively, these results suggest distinct gut microbial structures among different protein sources, with the duck protein group characterized by enrichment of specific microbial taxa, especially Clostridium.

Fig. 3.

Dietary proteins from different sources reshape gut microbiota composition in mice. (a, b) α-Diversity indices of the cecal microbiota. (c) β-Diversity assessed by principal coordinates analysis (PCoA) based on Bray-Curtis (See PERMANOVA results in Supplemental Table S8). PCoA1 and PCoA2 represent the percentage of variance explained by each coordinate. Each dot represents an individual sample, and ellipses represent 95% confidence intervals. Gut microbiota composition at the phylum (d) and genus (e) levels. (f) Differentially abundant taxa identified by linear discriminant analysis effect size (LEfSe). Taxa with LDA scores >4 are shown. Data are expressed as mean ± SEM (n = 9). Different letters indicate significant differences among groups (P < 0.05, Kruskal-Wallis test followed by Dunn's test with Benjamini-Hochberg correction).

3.4. Clostridium abundance positively correlates with intestinal secondary bile acid profiles

Associations between gut microbial taxa and significantly altered bile acids were examined using Spearman rank correlation analysis (Fig. 4a). Among the identified genera, Clostridium exhibited the strongest positive correlations with multiple secondary bile acids. Other genera, including Peptococcus, Murimonas, Anaerotruncus, Lachnospiraceae_UCG-006, and Tuzzerella, also showed strong positive correlations with these bile acids (r > 0.5, P < 0.05). All of these genera belong to the class Clostridia. Their relative abundances were significantly higher in duck protein-fed mice compared with those fed pork, beef, mutton, and chicken proteins (Fig. 4b–g). Additionally, goose protein feeding moderately elevated the abundances of some of these genera. Importantly, the relative abundance of baiE, a functional gene encoding bile acid 7α-dehydratase involved in secondary bile acid transformation in Clostridium scindens (Ridlon et al., 2016), was significantly higher in the duck protein group than in all other protein groups (Fig. 4h). Collectively, these findings suggest that the concurrent enrichment of Clostridium and increased abundance of baiE in the duck protein group were associated with elevated intestinal secondary bile acid levels.

Fig. 4.

Correlations between significantly altered bile acids and gut microbiota. (a) Spearman correlations between significantly altered bile acids and the relative abundances of bacterial genera. Asterisks indicate significant correlations. Genera highlighted in red and bold form had positive correlation coefficients with bile acids greater than 0.5. Relative abundance of (b) Clostridium, (c) Lachnospiraceae_UCG-006, (d) Murimonas, (e) Anaerotruncus, (f) Peptococcus, and (g) Tuzzerella. (h) Relative abundance of baiE in community DNA. Abbreviations: CDCA, chenodeoxycholic acid; LCA, lithocholic acid; isoDCA, isodeoxycholic acid; DCA, deoxycholic acid; HDCA, hyodeoxycholic acid; 12oLCA, 12-oxolithocholic acid. Data are expressed as mean ± SEM (n = 9). Different letters indicate significant differences among groups (P < 0.05, Kruskal-Wallis test followed by Dunn's test with Benjamini-Hochberg correction). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3.5. Dietary proteins differentially affect gene expression related to bile acid metabolism and signaling

We next investigated the effects of different dietary proteins on the hepatic and ileal expression of genes involved in bile acid synthesis, transport, and receptor signaling. Compared with casein and beef proteins, duck and goose proteins significantly upregulated mRNA expression of hepatic cholesterol 7α-hydroxylase (Cyp7a1), the rate-limiting enzyme initiating bile acid synthesis (Fig. 5a). Expression of sterol 12α-hydroxylase (Cyp8b1), responsible for generating 12αOH bile acids, was higher in mice fed beef, mutton, chicken, and duck proteins than in those fed casein, pork, and goose proteins. Compared with casein, duck protein feeding also upregulated the expression of 25-hydroxycholesterol 7α-hydroxylase (Cyp7b1), which yields non-12αOH bile acids.

Fig. 5.

Protein sources alter gene expression involved in bile acid synthesis, transport, and signaling in the liver and ileum. Relative mRNA expression of (a) bile acid synthesis enzymes and (b) bile acid nuclear receptors and their downstream targets. (c) Relative protein levels of FGF15 in the ileal mucosa. (d) Concentrations of active GLP-1 in the portal plasma. (e) Relative mRNA expression of bile acid transporters. Abbreviations: Cyp7a1, cholesterol 7α-hydroxylase; Cyp8b1, sterol 12α-hydroxylase; Cyp27a1, sterol 27-hydroxylase; Cyp7b1, 25-hydroxycholesterol 7α-hydroxylase; Fxr, farnesoid X receptor; Shp, short heterodimer partner; Fgf15, fibroblast growth factor 15; Tgr5, G protein-coupled bile acid receptor 1; Gcg, Preproglucagon; GLP-1, glucagon-like peptide-1; Ntcp, sodium-taurocholate cotransporting polypeptide; Bsep, bile salt export pump; Abcc3, ATP binding cassette subfamily C member 3; Abcc4, ATP binding cassette subfamily C member 4; Asbt, apical sodium-dependent bile acid transporter; Ibabp, intestinal bile acid binding protein; Osta/b, heteromeric organic solute transporter α/β. Data are expressed as mean ± SEM (n = 9). Different letters indicate significant differences among groups (P < 0.05, one-way ANOVA with Tukey-Kramer's test).

Ileal Fxr expression did not differ among the dietary groups, whereas its downstream effector, Fgf15, was markedly upregulated in mice fed duck protein compared with those fed casein, chicken, and goose proteins (Fig. 5b). Consistently, ileal FGF15 protein levels were significantly higher in the duck protein group than in the casein, pork, chicken, and goose protein groups (Fig. 5c). Although Tgr5 expression remained unchanged, expression of its downstream target preproglucagon (Gcg), which encodes the precursor of GLP-1, was significantly higher in the duck protein group relative to the pork, beef, and chicken groups (Fig. 5b). Consistently, portal plasma active GLP-1 concentrations were significantly higher in mice fed duck protein than in those fed other protein sources (Fig. 5d).

Expression of bile acid transporters varied by protein source (Fig. 5e, f). Hepatic expression of sodium-taurocholate cotransporting polypeptide (Ntcp) and bile salt export pump (Bsep) was significantly upregulated in the goose protein group. Ileal apical sodium-dependent bile acid transporter (Asbt) expression was increased in both the duck- and goose-fed mice compared with the casein- and pork-fed groups. Ileal expression of heteromeric organic solute transporter α/β (Osta/b) was elevated in mice fed beef, mutton, duck, and goose proteins compared with those fed casein and pork proteins. Collectively, these results suggest that protein sources differentially affected genes involved in bile acid synthesis and transport, as well as FGF15 protein levels and active GLP-1 concentrations.

3.6. Duck and pork proteins differentially affect gene expression related to glucose and fatty acid metabolism in the liver, jejunum, and adipose tissues

Because the most apparent difference in glucose tolerance and insulin sensitivity was observed between the duck and pork protein groups (Fig. 1 f–i), subsequent analyses focused on these two groups to further characterize the molecular changes associated with their distinct glucose metabolic phenotypes. In the liver, duck protein feeding significantly increased the mRNA levels of glucose transporter 2 (Glut2) and glycogen synthase (Gs), while reducing the expression of glucose-6-phosphatase (G6pase) and phosphoenolpyruvate carboxykinase 1 (Pck1) compared with pork protein (Fig. 6a). Consistently, protein levels of hepatic G6pase were also decreased (Fig. 6b). Duck protein feeding further upregulated the expression of genes involved in fatty acid oxidation, including peroxisome proliferator-activated receptor γ coactivator-1α (Pgc1a) and the medium- and long-chain acyl-CoA dehydrogenases (Acadm and Acadl). In the jejunum, duck protein feeding reduced the mRNA expression of Glut2 and G6pase compared with pork protein (Fig. 6c). In brown adipose tissue (BAT), duck protein significantly elevated the mRNA expression of Tgr5, Pgc1a, and uncoupling protein 1 (Ucp1), a marker of thermogenesis. Moreover, increased expression of genes involved in fatty acid oxidation, including Pgc1a, peroxisome proliferator-activated receptor α (Ppara), carnitine palmitoyltransferase 2 (Cpt2), Acadm, and Acadl, was detected in both eWAT and inguinal white adipose tissue (iWAT) of mice consuming the duck protein diet. Suppressed expression of adipose triglyceride lipase (Atgl), the rate-limiting enzyme of lipolysis, was also observed in eWAT in response to the duck protein diet. Collectively, compared with pork protein, duck protein feeding suppressed the expression of hepatic gluconeogenic genes while increasing fatty acid oxidation and thermogenic genes in adipose tissues.

Fig. 6.

Compared with pork protein, duck protein feeding alters gene expression involved in glucose and fatty acid metabolism in multiple tissues. (a) Relative mRNA expression of hepatic genes related to glucose transport, gluconeogenesis, glycogen synthesis, fatty acid oxidation, and lipogenesis. (b) G6pase concentration in the liver. (c) Relative mRNA expression of genes related to glucose transport and gluconeogenesis in the jejunum. Relative mRNA expression of bile acid receptors and genes related to thermogenesis, fatty acid oxidation, and lipogenesis in the (d) brown adipose tissue (BAT), (e) epididymal white adipose tissue (eWAT), and (f) inguinal white adipose tissue (iWAT). Abbreviations: Ampk, AMP-activated protein kinase; Pi3k, phosphoinositide 3-kinase; Akt, protein kinase B; Glut2, glucose transporter 2; Gsk3b, glycogen synthase kinase 3β; Gs, glycogen synthase; G6pase, glucose-6-phosphatase; Pck1, phosphoenolpyruvate carboxykinase 1; Pgc1a, peroxisome proliferator-activated receptor γ coactivator-1α; Ppara, peroxisome proliferator-activated receptor α; Acox1, acyl-Coenzyme A oxidase 1; Cpt1b, carnitine palmitoyltransferase 1b; Cpt2, carnitine palmitoyltransferase 2; Acadm, medium-chain acyl-CoA dehydrogenase; Acadl, long-chain acyl-CoA dehydrogenase; Pparg, peroxisome proliferator-activated receptor γ; Srebp1, sterol regulatory element-binding protein 1; Fasn, fatty acid synthase; Sglt1, sodium glucose cotransporter-1; Ucp1, uncoupling protein 1; Dio2, type 2 iodothyronine deiodinase; Glut4, glucose transporter 4; Atgl, adipose triglyceride lipase. Data are expressed as mean ± SEM (n = 9). Asterisks indicate significant differences between groups (P < 0.05, Student's t-test). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3.7. Associations among bile acids, metabolic gene expression, and glucose metabolic indices

Spearman correlation analysis revealed that bile acids were positively correlated with hepatic expression of genes involved in glucose transport, glycogen synthesis, and fatty acid oxidation, while showing negative correlations with gluconeogenesis-related genes in both the liver and jejunum (Fig. 7a, b). In adipose tissues, bile acids were positively correlated with the expression of Tgr5 and thermogenic genes in BAT, as well as genes involved in fatty acid oxidation in both eWAT and iWAT (Fig. 7c–e). Furthermore, glucose metabolic indices were positively correlated with hepatic and jejunal gluconeogenic gene expression (Fig. 7f, g), and negatively correlated with thermogenesis- and fatty acid oxidation-related gene expression in adipose tissues (Fig. 7h–j). Collectively, these findings suggest coordinated associations among bile acid profiles, glucose metabolic indices, and the expression of genes involved in gluconeogenesis, fatty acid oxidation, and thermogenesis.

Fig. 7.

Bile acids and glucose metabolic markers correlate with genes involved in glucose and fatty acid metabolism. Spearman correlations of bile acids (a–e) or glucose metabolic markers (f–j) with gene expression in liver, jejunum, brown adipose tissue (BAT), epididymal adipose tissue (eWAT), and inguinal adipose tissue (iWAT). Spearman's correlation coefficients and corresponding P values for the correlations are calculated. Yellow lines indicate positive correlations, and blue lines indicate negative correlations. Genes in red are statistically significant (P < 0.05). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

4. Discussion

The present study suggested that dietary meat protein sources differentially influenced glucose homeostasis, with duck protein feeding exhibiting more favorable glucose tolerance and insulin sensitivity than pork protein. These differences were accompanied by coordinated changes in gut microbial composition, bile acid metabolism, and FGF15 and GLP-1 responses. Compared with pork protein, duck protein feeding showed enrichment of Clostridium, a genus known to harbor bile acid-transforming capacity, increased abundance of the baiE gene encoding bile acid 7α-dehydratase, together with elevated levels of multiple secondary bile acids, suggesting an enhanced microbial capacity for secondary bile acid transformation. Consistently, portal plasma bile acid profiling revealed corresponding increases in total and secondary bile acids, indicating that intestinal bile acid alterations were detectable in the portal circulation. The altered bile acid profile coincided with enhanced ileal FGF15 expression at both the mRNA and protein levels, increased ileal Gcg expression, and higher portal plasma active GLP-1 concentrations, suggesting coordinated changes in intestinal bile acid-responsive signaling. These intestinal responses were paralleled by reduced hepatic expression of gluconeogenic genes and increased expression of genes involved in fatty acid oxidation and thermogenesis in adipose tissues. Collectively, these findings suggested that duck protein feeding was associated with coordinated alterations in gut microbiota, microbial bile acid metabolism, and FGF15 and GLP-1 responses, which may contribute to the more favorable glucose homeostasis observed relative to pork protein feeding.

Compared with pork protein, mice fed duck protein exhibited higher microbial diversity and enrichment of several bacterial genera, including Clostridium, Lachnospiraceae_UCG-006, Murimonas, Anaerotruncus, Peptococcus, and Tuzzerella (Fig. 3, Fig. 4). Recent studies have shown that meat proteins from different sources exhibit distinct digestion kinetics and peptide profiles (Diao et al., 2025; Zheng et al., 2026). Specifically, duck protein hydrolysates were reported to yield the lowest peptide content and the fewest identified peptides among six livestock/poultry meats, suggesting a peptide release pattern distinct from that of other meat sources during digestion (Zheng et al., 2026). Such differences in digestive properties may alter the spectrum of substrates available for microbial fermentation and potentially influence the growth or metabolic activity of specific gut microbes (Wu et al., 2022). Several genera enriched in the duck protein group, including Clostridium, Murimonas, Anaerotruncus, and Peptococcus, are known to ferment proteins and amino acids (Amaretti et al., 2019; Smith & Macfarlane, 1998). Their enrichment may therefore be related to differences in the availability of protein-derived substrates. However, whether the distinct peptide profile generated during duck protein digestion directly contributes to the enrichment of these genera remains to be determined. In addition, secondary bile acids possess hydrophobic and selective antimicrobial properties that can reshape microbial communities by suppressing bile acid-sensitive bacteria while favoring bile acid-tolerant taxa (Tian et al., 2020). Collectively, potential differences in substrate availability, together with ecological effects associated with altered secondary bile acid profiles, may help explain the microbial characteristics observed with duck protein intake.

We observed extensive correlations between multiple bacterial genera and secondary bile acids (Fig. 4), suggesting a functional linkage between duck protein-induced microbiota remodeling and bile acid transformation. These associations are biologically plausible in light of established microbial bile acid-transforming pathways. Specifically, DCA and LCA are generated from cholic acid (CA) and CDCA through 7α-dehydroxylation (Hirano et al., 1981), whereas isoDCA and 12oLCA arise from DCA by 3β-epimerization and 12α-oxidation, respectively (Cai et al., 2022). HDCA can be generated either from βMCA via 6β-epimerization followed by 7β-dehydroxylation (Wahlström et al., 2016) or from DCA through sequential 12α-dehydroxylation and 6α-hydroxylation (Lucas et al., 2021). Consistent with these pathways, the abundance of Clostridium, which includes species capable of bile acid transformations such as 7α/β-dehydroxylation, oxidation, and epimerization (Hirano et al., 1981; Wahlström et al., 2016), showed the strongest positive correlations with secondary bile acids. The increased abundance of the baiE gene in mice fed duck protein further suggests a greater potential for microbial 7α-dehydroxylation activity. In addition, cooperative interactions may further expand bile acid metabolic capacity. For example, Peptococcus has been implicated in bile acid desulfation, facilitating the transformation of bile acid sulfates into secondary bile acids (Collins et al., 2023). Notably, although hepatic Cyp8b1 mRNA expression was not the highest in the duck protein group, this group exhibited the greatest abundance of 12αOH bile acids (Fig. 2, Fig. 5a). This dissociation suggests that, in the present study, hepatic Cyp8b1 mRNA expression alone may not fully explain differences in intestinal 12αOH bile acid levels and further highlights the role of gut microbial biotransformation in shaping bile acid composition (Sayin et al., 2013). Together, our findings suggest that the enrichment of Clostridium, together with increased abundance of the baiE gene, may contribute to increased levels of secondary bile acids following duck protein feeding.

Differences in bile acid profiles among protein groups were accompanied by variations in bile acid receptor-related signaling (Fig. 5). Compared with pork protein, duck protein feeding markedly increased ileal Gcg mRNA expression and portal plasma active GLP-1 concentrations, implying an altered TGR5–GLP-1 signaling axis. TGR5 is a G protein-coupled bile acid receptor whose activation in intestinal L cells stimulates GLP-1 secretion and improves glucose tolerance (Kawamata et al., 2003; Thomas et al., 2009). Likewise, TGR5-agonistic bile acids, such as HDCA, have been shown to improve glucose tolerance in a TGR5-dependent manner (Makki et al., 2023). Recent evidence further supports a role for Clostridium-mediated bile acid metabolism in regulating GLP-1 signaling and glucose homeostasis (She et al., 2024). Therefore, we speculate that the enrichment of Clostridium and the concomitant expansion of secondary bile acids with TGR5 agonistic activity, including DCA, HDCA, and LCA (Fig. 2, Fig. 4), may have contributed to the enhanced GLP-1 secretion observed in the duck protein group. Additionally, activation of TGR5 signaling does not necessarily require increased receptor expression. Consistent with previous studies (Harach et al., 2012), ileal Tgr5 mRNA expression remained unchanged (Fig. 5b), suggesting that enhanced TGR5 signaling may be driven primarily by increased ligand availability rather than receptor abundance. Nevertheless, other G protein-coupled receptors expressed in intestinal L cells, such as GPR119, have also been implicated in nutrient-induced GLP-1 secretion (Higuchi et al., 2020), the contribution of receptors other than TGR5 cannot be excluded.

In parallel, increased CDCA levels may favor FXR activation (Makishima et al., 1999). Although FXR exerts divergent effects on glucose metabolism depending on physiological context (Jiang et al., 2015), intestinal FXR activation induces FGF15 expression, which suppresses hepatic gluconeogenesis and promotes glycogen synthesis (Kir et al., 2011; Potthoff et al., 2011). Correspondingly, duck protein feeding increased FGF15 expression at both the mRNA and protein levels compared with pork protein (Fig. 5b, c), while hepatic expression of the key gluconeogenic genes Pck1 and G6pase was reduced (Fig. 6a, b), consistent with the inhibitory effect of FGF15 on the hepatic gluconeogenesis pathway. Correlation analysis further revealed significant associations among secondary bile acids, hepatic G6pase expression, and glucose metabolic markers (Fig. 7a, f). As excessive hepatic gluconeogenesis contributes to hyperglycemia (Petersen et al., 2017), reduced expression of gluconeogenic genes may partly contribute to the more favorable glucose homeostasis observed in duck protein-fed mice compared with pork protein-fed mice. Collectively, these observations suggest that duck protein-induced alterations in microbial bile acid metabolism may enhance both FGF15 and GLP-1 responses, potentially contributing to the more favorable glucose metabolic phenotype observed in duck protein-fed mice compared with pork protein-fed mice.

Notably, despite increased hepatic Gs mRNA expression, duck protein-fed mice exhibited lower hepatic glycogen content than pork protein-fed mice (Figs. 1p, 6a). This discrepancy may reflect dynamic glycogen turnover, as glycogen content is determined by the balance between synthesis and utilization (Agius, 2015). Moreover, glycogen synthase activity is regulated primarily by reversible phosphorylation and therefore may not directly correspond to Gs mRNA expression (Roach et al., 2012). In addition, because mice were euthanized under ad libitum feeding conditions, hepatic glycogen content may also have been influenced by differences in recent feeding status.

Compared with pork protein, duck protein feeding increased the mRNA expression of Ucp1 and fatty acid oxidation-related genes, which were significantly correlated with secondary bile acids and glucose metabolic indices (Fig. 6, Fig. 7). These findings raise the possibility that bile acid signaling may contribute to the favorable glucose metabolic phenotype in the duck protein group by enhancing adipose thermogenic and fatty acid oxidation-related gene expression. Secondary bile acids are known activators of TGR5 signaling in adipose tissue, which elevate intracellular cAMP levels and stimulate thermogenic gene expression (Broeders et al., 2015). Consistently, adipocyte-overexpression of UCP1 has been shown to improve glucose homeostasis in association with upregulation of genes involved in thermogenesis and fatty acid β-oxidation (Park et al., 2025; Takahashi et al., 2017). Nevertheless, we did not assess energy expenditure or thermogenesis. Therefore, whether these transcriptional changes translate into increased energy expenditure remains to be determined. Our data imply a potential role for a bile acid–TGR5–thermogenesis axis in the favorable glucose homeostasis associated with duck protein feeding. Future studies incorporating measurements of energy expenditure and thermogenic capacity are warranted to clarify the functional significance of these molecular changes.

This study has several limitations. Although we provide strong correlational evidence linking duck protein intake with Clostridium enrichment, altered bile acid profiles, and host glucose regulation, additional studies employing targeted microbial manipulation or bile acid receptor antagonism are warranted to determine whether alterations in microbial bile acid metabolism mediate the beneficial effects of duck protein on glucose homeostasis. Moreover, although differences in gut microbial composition were observed among dietary protein groups, the present study did not assess digestion kinetics or peptide profiles. Further studies integrating digestion characterization and peptide profiling are needed to elucidate the biochemical basis underlying protein source-dependent microbial alterations. In addition, defatted meat proteins rather than whole meat were used for diet preparation, which allowed us to isolate the effects of protein itself; however, it remains unclear whether lipids and other components naturally present in duck meat would modify the responses. Finally, because mice were group-housed, microbiota analyses based on individual mice may be influenced by coprophagy. Cage-level analyses showed separation of microbial communities among dietary groups and differences in the relative abundance of Clostridium, consistent with the individual-level analyses. Future studies with increased cage replication are warranted to further validate these findings.

5. Conclusion

In conclusion, our data suggested that different animal protein sources exert distinct effects on glucose homeostasis. Compared with pork protein, duck protein consumption exhibited more favorable glucose tolerance and insulin sensitivity. Importantly, these differences in glucose homeostasis were accompanied by enrichment of Clostridium, increased abundance of the baiE gene, and elevated production of secondary bile acids. These microbial and bile acid alterations coincided with enhanced FGF15 and GLP-1 responses, reduced hepatic gluconeogenic gene expression, and increased expression of genes involved in adipose fatty acid oxidation and thermogenesis. Our findings support a potential role for gut microbial bile acid transformation and associated bile acid receptor signaling in the favorable glucose homeostasis observed following duck protein consumption. This study improves the understanding of protein source-dependent differences in glucose homeostasis under physiological conditions.

CRediT authorship contribution statement

Hongxia Liu: Conceptualization, Investigation, Formal analysis, Writing – original draft, Writing – review & editing, Funding acquisition. Haifeng Li: Investigation, Formal analysis, Writing – original draft. Yang Zhai: Investigation, Writing – review & editing. Feifan Zhang: Investigation. Jichao Huang: Investigation. Jingxin Sun: Investigation. Yan Lyu: Investigation. Ming Huang: Conceptualization, Supervision, Funding acquisition.

Ethics statement

All animal experiments were approved by the Animal Care and Use Committee of Nanjing Agricultural University (approval No. SYXK(Su)2021-0086) and conducted in accordance with the NIH (National Research Council) Guide for the Care and Use of Laboratory Animals and the ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary material

Data availability

Data will be made available on request.