How gut bacteria and tryptophan processing interact in type 2 diabetic mice treated with metformin

Chromatographic parameters were established using gradient inversion, and the appropriate analytical peaks were obtained. The chromatograms of the 16 compounds in positive ion mode are shown in Fig. 1; Fig. S1. No interference peaks of any of the analytes or the isotope-labeled internal standard (ISTD) were observed under the selected conditions in the serum and fecal samples, and there was no interference between the analytes and the ISTD in the blank water.

Standard solutions of different concentrations were mixed with the ISTD and analyzed by LC-MS/MS. The linear relationship between the analyte area ratio (analyte area/ISTD area) and its nominal concentration was analyzed. Each standard curve consisted of a minimum of five points with different concentrations to ensure accurate quantification. The method exhibited good selectivity. Table 1 summarizes the calibration curve, coefficient of determination, the limit of quantitation (LOQ), the limit of detection (LOD), and the intra- and inter-day precision of this determination. The correlation coefficients for the calibration curves of each compound were >0.99, achieving a linear range with R² > 0.996. The LOQ (signal-to-noise ratio, S/N = 10) for the 16 analytes ranged from 0.29 to 69.444 nmol/L, while the LOD (S/N = 3) ranged from 0.087 to 20.833 nmol/L. The relative standard deviation (RSD, %) values of inter- and intra-day repeatability and precision were below 20%, except for indole-3-pyruvic acid (IPyA) in blood samples and 3-skatole (3-methylindole, 3MI) in stool samples. The analyte recovery rate was evaluated by adding indole-3-acetic-2,2-d2 as the ISTD during the initial sample extraction, with retention time and mass in Table 2. The recovery rates in serum and feces were 88.63% and 94.54%, respectively (Table 1).

Metformin partially rectified the concentration of TRP metabolites in the serum, colon contents, and urine of T2DM mice (Fig. 2A). Figure 2A illustrates the procedure for establishing a T2DM mouse model and metformin intervention. As displayed in Fig. 2B, the concentrations of ILA, indole-3-acetaldehyde (IAAld), and KYN were significantly elevated (P < 0.05) in the serum samples of the T2DM group compared to the CON group, while the concentrations of IPA and IA were significantly reduced (P < 0.05). Treatment with metformin effectively decreased ILA levels while significantly boosting IPA levels back to normal levels in T2DM mice.

In the colon contents (Fig. 3C), the concentrations of tryptophol (IEt) and KYN were significantly higher (P < 0.05) in the T2DM group in comparison to the healthy controls, while the levels of IAA, tryptamine (TrA), 3MI, IPA, IA, and IAld were significantly lower (P < 0.05) than those observed in the CON group. Additionally, although the observed trend did not reach statistical significance, there was a tendency toward an increase in IAAld (P = 0.0535) and ILA (P = 0.0628) in the T2DM group. After metformin treatment, the concentrations of IAA, TrA, 3MI, IPA, and IAld in colon contents significantly increased, accompanied by a significant reduction in the levels of IEt, IAAld, ILA, and KYN. These findings suggest that metformin intervention effectively mitigates dysregulated TRP metabolism within the gut microbiota.

Furthermore, metformin treatment significantly reduced the increased KYN concentration in the urine samples of T2DM mice, but did not lead to significant improvements in the decreased levels of ILA, IAAld, IAA, TRP, Ind, IAld, and IPyA (Fig. 2D). Notably, metformin treatment did not result in any significant alterations in TRP metabolism in fecal samples of T2DM mice (Fig. S2A).

In addition, the KYN/TRP ratio in the serum, colon contents, and urine samples of the T2DM group was significantly increased (P < 0.01). While metformin therapy restored normal KYN/TRP ratios in the serum and urine (Fig. 2E and G), it did not fully normalize the KYN/TRP ratios in colon contents (Fig. 2F). The KYN/TRP ratio is frequently used to assess the activity of the extrahepatic Trp-degrading enzyme indoleamine 2,3 dioxygenase (IDO). Although KYN and TRP values varied, the increase of KYN/TRP ratio in T2DM indicated an increase in host KYN pathway metabolism (dominated by IDO) within the TRP metabolic pathways (14), while the corresponding microbiota-dominated indole metabolic pathway decreased (6). Treatment with metformin showed benefits in correcting the imbalance between host and bacterial TRP metabolism in T2DM. Interestingly, the ILA/IPA ratio in the serum and colon contents samples was also significantly increased in the T2DM group (P < 0.01), and this ratio was restored after metformin treatment (Fig. 2H and I).

Principal component analysis (PCA) was carried out to elucidate the alterations in the TRP metabolite profiles in serum, colonic contents, and urine among the three groups. ANOSIM analysis, employing Bray-Curtis distances, unveiled notable distinctions in the TRP metabolite profiles found in serum (R = 0.96, P < 0.01), colon contents (R = 0.72, P < 0.01), and urine (R = 0.83, P < 0.01) among the CON, T2DM, and MET groups (Fig. 2J, K and L; Fig.S2C). Remarkably, the position of the confidence ellipse in the MET group fell between that of the CON and MET groups, suggesting a partial restoration of the TRP metabolic profile post-metformin treatment.

Partial least squares discriminant analysis (PLS-DA) was conducted to analyze metabolites from both serum samples and colon content samples (Fig. S3A and C). We employed PLS-DA variable importance projection scores to assess the significance of various TRP metabolites in the context of metformin treatment for T2DM. Notably, IPA, IAAld, and ILA emerged as crucial variables among the serum metabolites (Fig. S3B). Conversely, IAA, 3MI, and IPA were identified as key variables in the colon contents (Fig. S3D). These findings highlight the potential of these metabolites as biomarkers for predicting the onset and progression of T2DM.

The rarefaction curve for the analysis of alpha diversity using amplicon sequence variants (ASVs) in different groups confirms the adequate sequencing depth of each sample (Fig. S4). The ACE, Chao1, Shannon, and Simpson indices (Fig. 3A through D) demonstrated a significant decrease in the richness and diversity of the gut microbiota in the T2DM group compared to the CON group. However, metformin treatment notably restored these indices.

Principal coordinates analysis was conducted at the amplicon ASV level to assess the distinct patterns of bacterial community structures within the three different groups (Fig. 3E). ANOSIM analysis, utilizing Bray-Curtis distances, unveiled significant distinctions among the CON, T2DM, and MET groups (R = 0.58, P < 0.01). The hierarchical clustering based on Bray-Curtis distance indicated that the microbial composition of the MET group exhibited higher similarity to the CON group compared to the samples from the T2DM group (Fig. 3F).

Based on the annotated 16S rDNA sequencing results from the SILVA database, Fig. 4A and B illustrated the phyla and genera respectively, with abundances surpassing 1%. To identify differentially abundant taxa among the groups, we employed linear discriminant analysis effect size (LEfSe). Figure 4C and D demonstrate that the T2DM group exhibited lower abundance of SCFA-producing bacteria, including Faecalibaculum, Lachnospiraceae NK4A136 group, Alistipes, Roseburia, and Turicibacter. Notably, after metformin intervention, the abundance of Allisteria was restored, while Faecalibaculum and Turicibacter experienced partial restoration.

Utilizing the PICRUST2 method to predict the functional profile of the sequencing results, we observed a significant enrichment of TRP metabolism-related pathways following metformin treatment (Fig. S5). To better understand the intricate interplay between the gut microbiota and microbial-derived indole derivatives (Fig. 5A), we conducted Spearman correlation analysis to examine the association between TRP metabolism and the abundance of gut microbiota. As shown in Fig. 5B and C, we found a positive correlation between Lactobacillaceae, Lactotobacillus, and Dubosiella (which were enriched in the T2DM group) with KYN, while displaying a negative correlation with 3MI and IPA. Additionally, we observed positive correlations between Faecalibaculum, Turicibacter, and Alistipes (which decreased in the T2DM group and were restored by metformin treatment) with IPA, IAA, IAld, TrA, and 3MI. Conversely, these genera displayed negative correlations with ILA, KYN, and IAAld.

To investigate the exclusive origins of indole derivatives from the gut microbiota, we analyzed the levels of TRP metabolites in serum and colonic content from germ-free (GF) and conventional (CV) mice (Fig. 6A). Surprisingly, we successfully detected TRP, 5-HT, IPyA, ILA, KYN, 3MI, IAA, and IA in the serum of both GF and CV mice (Fig. 6B). However, CV mice exclusively exhibited the presence of IPA and IAAld. In the colonic content, only five compounds (TRP, IPyA, IEt, IA, and KYN) were detected in the colonic content of GF mice (Fig. 6C), while the concentrations of IAA, TrA, 3MI, IPA, IAld, IAld, and ILA in CV mice also exceeded the detection limit.

We successfully developed an LC-MS/MS method to quantify TRP and its 15 catabolites in biological samples. This method was validated and applied to assess TRP metabolism profiles in various biological samples from T2DM mice. As shown in Fig. 7, our findings unveiled perturbations in TRP metabolism in T2DM mice, predominantly characterized by increased ILA levels and decreased IPA levels in peripheral blood and intestinal contents, thus significantly disrupting the ILA/IPA ratio. Noteworthy, the ILA/IPA ratio reverted to normalcy post-treatment with metformin, suggesting the potential of ILA/IPA as a diagnostic marker or therapeutic target for T2DM. The regulation of TRP metabolism by metformin might involve the restoration of the relative abundance of Turicibacter, Enterorhabdus, RF39, Clostridia_UCG-014, and Alistipes. Additionally, the detection of indole metabolites in the serum of GF mice suggested the existence of endogenous TRP metabolic pathways in the host. Thus, the possibility of metformin modulating endogenous TRP metabolism should not be disregarded. Further investigations are required to unravel the intricate relationship between the gut microbiota, the host, and TRP metabolism.

TRP, KYN, 3MI, TrA, ILA, Ind, IAld, IA, IPA, IAA, and 2-OX were obtained from Aladdin Bio-Technology Co., Ltd. (Shanghai, China). Indole-3-acetamide (IAM), IEt, and IPyA were purchased from Yuanye Bio-Technology Co., Ltd. (Shanghai, China). 5-HT and ISTD: Indole-3-acetic-2,2-d2 was purchased from Sigma-Aldrich (Shanghai, China). IAAld was purchased from Bide Pharmatech Ltd. (Shanghai, China). All reagents were obtained with a minimum purity of 97%, except IPyA (95%) and IAAld (90%). The chemical structures of the TRP metabolites are shown in Fig. S6. All organic solvents and water used in the sample and mobile phase preparations were of LC-MS grade and obtained from Sigma-Aldrich (Shanghai, China) and Watsons (Guangzhou, China).

All 16 chemicals and the ISTD were weighed using a BT224S electronic balance (Sartorius AG, Göttingen, Germany). A stock solution of 10 mM in N, N-dimethylformamide was prepared for each standard, and the stock solution was diluted with 20% acetonitrile in the eight gradients (). All the solutions of each standard were stored at −20°C until use. Table S1

This study comprised three batches of mice experiments. (i) Five mice were housed in each cage and maintained at a temperature of 20 ± 2°C, in a 12-h diurnal cycle. They were provided with ad libitum access to food and water. At the conclusion of the experiment, fresh stool and urine samples were collected using sterile centrifuge tubes. Whole blood was kept at 24°C for 2 h, then centrifuged at 2,400 rpm at 4°C for 15 min to collect the supernatant. The mice were euthanized by cervical dislocation, and colonic contents were sampled. All samples were stored at −80°C. These SPF mouse samples were utilized for evaluating the precision and accuracy of the LC-MS/MS method.

Serum samples were extracted as described by Xu et al. (73), with minor modifications. Then, 100 µL of the serum samples was mixed with 100 µL of the IS (1 µM) and 400 µL of methanol in a 1.5 mL centrifuge tube, vortexed for 10 s, and centrifuged at 4°C and 17,800 × g for 5 min. Subsequently, 450 µL of the supernatant was transferred into a new 1.5 mL centrifuge tube, and an additional 450 µL of methanol was added to the original centrifuge tube containing the precipitate. The mixture was vortexed, sonicated for 5 min, and then centrifuged at 4°C and 14,000 rpm for 5 min. The supernatant was collected and combined with the previous 450 µL supernatant solution. Finally, after centrifugation at 4°C and 14,000 rpm for 10 min, all the supernatant was aspirated, and the solution was filtered through a membrane.

The extraction method for the intestinal content and fecal samples was improved according to the method described by Zou et al. (20). Before sample processing, fresh samples were freeze-dried and ground into a powder. The sample (weighing 25 mg) was dissolved in a mixed solution of 250 µL water, 250 µL methanol, and 100 µL IS (1 µM), vortexed for 10 s, sonicated for 5 min, and centrifuged at 4°C and 17,800 × g for 5 min. Subsequently, 450 µL of the supernatant was transferred into a new 1.5 mL centrifuge tube, and an additional 450 µL of methanol was added to the original centrifuge tube containing the precipitate. The mixture was vortexed, sonicated for 5 min, and then centrifuged at 4°C and 14,000 rpm for 5 min. The supernatant was collected and combined with the previous 500 µL supernatant solution. Finally, after centrifugation at 4°C and 14,000 rpm for 10 min, all the supernatant was aspirated, and the solution was filtered through a membrane.

The extraction method for urine samples was slightly modified from that described by Zou et al. (20). In a 1.5 mL centrifuge tube, the sample (200 µL) was aspirated and mixed with 100 µL IS (1 µM) and 700 µL methanol. The solution was vortexed for 10 s, sonicated for 5 min, and placed at −20°C for 10 min. The solution was then centrifuged for 10 min at 4°C and 17,800 × g. Finally, 500 µL of supernatant was mixed with 500 µL of water, vortexed for 10 s, and observed to see whether it was clear. If not, centrifugation was repeated until the liquid was clear, after which the solution was filtered through a membrane.

LC-MS/MS analysis was performed using an AQUITY ultra-performance liquid chromatography system, thermostatic autosampler, ultra-high-performance binary pump (I-class, Waters, MA, USA), and a QTRAP 6500 tandem mass spectrometer (Sciex, Framingham, MA, USA). The controlling software was Analyst 1.6.2. Chromatographic separation was achieved on an ACQUITY PREMIER BEH C18 column (1.7 µm, 2.1 × 150 mm², 1/pk, Waters, Milford, DE, USA) at 45°C. Considering the structural diversity and lipophilic range of the 16 compounds, mass spectrometric detection was performed using multiple reaction monitoring (MRM) with an electrospray ionization source in positive mode to minimize ion elution. For high-performance liquid chromatography, solvent A was 0.1% (vol/vol) formic acid in water, and solvent B was 0.1% (vol/vol) formic acid in ACN at a flow rate of 0.3 mL/min. The gradient elution procedure was as follows: 0–0.8 min, 85% A; 0.8–10.5 min, 85%–5% A; 10.5–11.4 min, 5% A; 11.4–11.5 min, 5%–85% A; and 11.5–12 min, 85% A.

The mass spectrometer was set to the positive electrospray ionization mode with MRM. The IonSpray voltage was set at 5,000 V and the temperature at 300°C. Curtain gas, ion source gas 1, ion source gas 2, and collision gas were set to 25, 10, 10, and medium, respectively. The entrance potential was set to 10 V. MRM transitions, declustering potential, collision energy, and collision cell exit potential were optimized using a syringe infusion pump. Table 1 summarizes the compound-specific chromatographic and mass spectrometry parameters, including retention time, MRM transitions, declustering potential, collision energy, and collision cell exit potential. Data acquisition and processing were performed using the Analyst software (version 1.6) from AB SCIEX.

Sixteen analyte standard stock solutions were mixed with ISTD stock solutions, and the linearity of the calibration curves for each compound was evaluated under eight concentration gradients. Calibration curves were established using a linear regression equation and a linear correlation coefficient with the concentration and peak area. The LOD was defined as an S/N of 3:1, and the LOQ was defined as an S/N of 10:1.

samples were prepared by mixing the ISTD and serum, feces from 10 specific pathogen-free mice to assess the recovery rate and precision. The intra-day and inter-day precision was evaluated by repeating the analysis of the QC samples on the same day and on three separate days. A daily calibration curve was used to calculate the concentration of each analyte in the sample, and the accuracy was expressed as the RSD.

SPSS software (version 26.0) was used for the statistical analysis. Statistical differences between the two groups were analyzed using an independent-sample t test, and data among the three groups were analyzed using one-way ANOVA with post hoc Bonferroni comparisons or Kruskal-Wallis with post hoc Bonferroni comparisons; P values <0.05 and <0.01 were considered statistically significant.