Metabolomic profiling of BAs in plasma, urine and fecal samples

The demographic characteristics of the study population were extracted from the medical record system of hospital and are presented in Supplementary Table 6. Of 50 kinds of BAs in the established UPLC-MS/MS method (Fig. 1A), twenty-three kinds in the plasma, twenty kinds in the feces, and twenty-nine kinds in the urine, with >30% of the measurements below the lower limit of quantitation (LOQ), were excluded. Within the detected BAs, thirteen kinds in the plasma, seventeen kinds in the feces, and three kinds in the urine were significantly altered in DKD group than in T2DM group or CON group (Fig. 1B).

Fig. 1: Plasma, fecal and urine profiling of BAs in the step-wise alterations of DKD patients.

A Chromatogram of 50 BAs in MRM mode. B The heatmap of preliminarily selected differential metabolites in the plasma, feces and urine samples of the DKD, T2DM and CON groups. C The OPLS-DA analyses of the metabolomic profiling of the plasma, feces and urine samples from the DKD, T2DM and CON groups. D The Venn diagrams for overlap of key BAs with P < 0.05 and VIP value > 1.2 in the plasma, feces and urine samples of the DKD, T2DM and CON groups. E The loading plots of the key BAs with both P < 0.05 and VIP value > 1.2 in the plasma, feces and urine samples of the DKD, T2DM and CON groups. The detailed information relating to the chromatogram of each BA is listed in Supplementary Table S4.

The OPLS-DA analysis was then performed, highlighting the optimal discriminations among the three groups in the plasma (Fig. 1C; R2X = 0.366, R2Y = 0.724, Q2 = 0.607) and feces (R2X = 0.392, R2Y = 0.787, Q2 = 0.675), yet non-optimal discrimination in the urine (R2X = 0.239, R2Y = 0.356, Q2 = 0.285), with optimal binary classifier, validity and degree of overfitting (Supplementary Fig. 1). According to the significance of P < 0.05 in the Student’s t-test and VIP > 1.2 in the OPLS-DA model, eight kinds of BAs, including Glycochenodeoxycholic acid (GCDCA), Glycohyocholic acid (GHCA), Glycocholic acid (GCA), Taurocholic acid (TCA), Taurochenodeoxycholic acid (TCDCA), Glycodeoxycholic acid (GDCA), Taurodeoxycholic acid (TDCA), and Norcholic acid (NCA), were the key BAs in the plasma contributing to the group difference among three groups (Fig. 1D, E). Eight kinds of BAs in the feces, including TDCA, Glycolithocholic acid (GLCA), GCA, Cholic acid (CA), Chenodeoxycholic acid-3-β-D-glucuronide (CDCA-3Gln), 7-Ketodeoxycholic acid (7-KDCA), Chenodeoxycholic acid (CDCA) and Isolithocholic acid (ILCA), as well as Taurolithocholic acid-3-sulfate (TLCA-3S), Glycolithocholic acid-3-sulfate (GLCA-3S) and 7-KDCA in the urine, were the major BAs contributing to the group differences.

The separated metabolomic profiling of plasma BAs

For the BA profiling in plasma, six kinds of the primary conjugated BAs, including CDCA-3Gln, GCDCA, GCA, TCDCA, TCA and NCA, and four kinds of the secondary conjugated BAs, including GHCA, GDCA, ursodeoxycholic acid (UDCA) and TDCA, as well as two kinds of the secondary unconjugated BAs, namely deoxycholic acid (DCA) and 3β-deoxycholic acid (3β-DCA), were significantly altered in the patients with DKD (Supplementary Table 7). When exploring the detailed BA metabolism in the patients with T2DM or DKD, or patients in the diabetic condition (either T2DM or DKD), or patients in Non-DKD condition (either CON or T2DM), the OPLS-DA model was employed, visualizing the optimal separation between the DKD group and CON group (R2X = 0.365, R2Y = 0.866, Q2 = 0.761, Fig. 2A). The SUS-plots and ROC analyses identified GCDCA, GHCA, GCA, TCA, CDCA-3Gln, NCA and TCDCA as key BAs discriminating the DKD group from CON group (Fig. 2B, C). The OPLS-DA analysis also visualized the separation between the DKD group and T2DM group (R2X = 0.339, R2Y = 0.771, Q2 = 0.589, Fig. 2D). The SUS-plots and ROC analysis identifying GCDCA, GCA, TCA, GHCA and UDCA as key BAs discriminating the DKD group from T2DM group (Fig. 2E, F). In addition, it also indicated that GCDCA, GHCA, GCA and TCA were the key BAs discriminating the DKD group from the Non-DKD group (R2X = 0.315, R2Y = 0.784, Q2 = 0.675, Fig. 2G–I), while the GHCA, TCDCA, TDCA, NCA, GCDCA, GDCA, and GCA were the key BAs identifying the patients in diabetic condition from the CON group (R2X = 0.231, R2Y = 0.771, Q2 = 0.677, Fig. 2J−L). Accompanied with the significant alteration of these six kinds of BAs in plasma (Fig. 2M–R), the combined results indicated that GCDCA, GCA, TCA and GHCA were significantly increased in the DKD group and strongly correlated with the progression of DKD from either T2DM or healthy status.

Fig. 2: The separated metabolomic profiling of the key BAs in the plasma of the patients with DKD and T2DM, and healthy individuals.

A The OPLS-DA score plot between the DKD group and the CON group. The SUS-plot (B) and ROC curve analyses (C) of the key BAs between the DKD group and the CON group. D The OPLS-DA score plot between the DKD group and the T2DM group. The SUS-plot (E) and ROC curve analyses (F) of the key BAs between the DKD group and the T2DM group. G The OPLS-DA score plot between the DKD group and the Non-DKD group (both T2DM group and CON group). The SUS-plot (H) and ROC curve analyses (I) of the key BAs between the DKD group and the Non-DKD group. J The OPLS-DA score plot between the group in diabetic condition (both DKD and T2DM group) and the CON group. The SUS-plot (K) and ROC curve analyses (L) of the key BAs between the group in diabetic condition and CON group. The plasma levels of the key BAs, including M GCDCA, N GHCA, O GCA, P TCA. *P < 0.05, **P < 0.01.

The separated metabolomic profiling of fecal BAs

For the BA profiling in feces, the significant alterations of seventeen kinds of BAs presented in the DKD group, including (1) two kinds of primary unconjugated BAs, namely CA and CDCA, (2) five kinds of primary conjugated BAs, including TCA, TCDCA, CDCA-3Gln, GCDCA and ω-Muricholic acid (ω-MCA), (3) four kinds of secondary unconjugated BAs, namely 7-Ketolithocholic acid (7-KLCA), 12-KLCA, ILCA and 7-KDCA, and (4) six kinds of secondary conjugated BAs, including 3β-Hyodeoxycholic acid (3β-HDCA), LCA-3S, Tauroursodeoxycholic acid (TUDCA), TDCA, GLCA and Glycoursodeoxycholic acid (GUDCA) (Supplementary Table 8). The separated OPLS-DA analysis showed the optimal separation between the DKD group and the CON group (R2X = 0.291, R2Y = 0.915, Q2 = 0.868, Fig. 3A), with GLCA, CDCA, LCA-3S, 7-KDCA, ILCA, ω-MCA, TDCA and CDCA-3Gln as the key BAs to group difference (Fig. 3B, C). The OPLS-DA analysis between the DKD group and the T2DM group (R2X = 0.274, R2Y = 0.827, Q2 = 0.723, Fig. 3D) also showed the ideal modeling with GLCA, CDCA, 7-KDCA, CDCA-3Gln, GCA, ω-MCA and CA as the key BAs contributing to the group difference (Fig. 3E, F). In addition, it also indicated that GLCA, CDCA, 7-KDCA, ILCA, ω-MCA and CDCA-3Gln as the key BAs discriminating the DKD group from the Non-DKD group (R2X = 0.251, R2Y = 0.833, Q2 = 0.759, Fig. 3G–I), while GCA, TDCA, CA, 7-KLCA, TCDCA, TUDCA, ILCA and 7-KDCA were the key BAs identifying the patients in diabetic condition from the CON group (R2X = 0.246, R2Y = 0.779, Q2 = 0.705, Fig. 3J–L). Combing the significant increase of their fecal levels, GLCA, 7-KDCA, ILCA, ω-MCA and CDCA-3Gln were positively correlated with the progression of DKD from either T2DM or healthy status (Fig. 3M–Q).

Fig. 3: The separated metabolomic profiling of the key BAs in the feces of the patients with DKD and T2DM, and healthy individuals.

A The OPLS-DA score plot between the DKD group and the CON group. The SUS-plot (B) and ROC curve analyses (C) of the key BAs between the DKD group and the CON group. D The OPLS-DA score plot between the DKD group and the T2DM group. The SUS-plot (E) and ROC curve analyses (F) of the key BAs between the DKD group and the T2DM group. G The OPLS-DA score plot between the DKD group and the Non-DKD group (both T2DM group and CON group). The SUS-plot (H) and ROC curve analyses (I) of the key BAs between the DKD group and the Non-DKD group. J The OPLS-DA score plot between the group in diabetic condition (both DKD and T2DM group) and the CON group. The SUS-plot (K) and ROC curve analyses (L) of the key BAs between the group in diabetic condition and CON group. The levels of the key fecal BAs, including M GLCA, N 7-KDCA, O ILCA, P ω-MCA, and Q CDCA-3Gln. *P < 0.05, **P < 0.01, NS, no significance.

The separated metabolomic profiling of urinary BAs

When analyzing the BA levels in the urine, it presented the decrease of GLCA-3S, TLCA-3S and 7-KDCA in the DKD group (Supplementary Table 9). The OPLS-DA models showed the poor performance between the DKD group and CON group (R2X = 0.248, R2Y = 0.754, Q2 = 0.591), the DKD group and T2DM group (R2X = 0.179, R2Y = 0.593, Q2 = 0.326), the DKD group and Non-DKD group (R2X = 0.234, R2Y = 0.451, Q2 = 0.271), and the patients in diabetic condition from the CON group (R2X = 0.237, R2Y = 0.672, Q2 = 0.529), with the validation of the significance of TLCA-3S, GLCA-3S, and 7-KDCA in urine that contribute to the DKD identification by the SUS-plots and ROC analyses (Fig. 4A–L). It also showed step-wise decrease of these three BAs in the progression of DKD from T2DM and CON. These results indicated that the urine GLCA-3S, TLCA-3S and 7-KDCA may significantly involve in the progression of DKD (Fig. 4M–O).

Fig. 4: The separated metabolomic profiling of the key BAs in the urine samples of the patients with DKD and T2DM, and healthy individuals.

A The OPLS-DA score plot between the DKD group and the CON group. The SUS-plot (B) and ROC curve analyses (C) of the key BAs between the DKD group and the CON group. D The OPLS-DA score plot between the DKD group and the T2DM group. The SUS-plot (E) and ROC curve analyses (F) of the key BAs between the DKD group and the T2DM group. G The OPLS-DA score plot between the DKD group and the Non-DKD group (both T2DM group and CON group). The SUS-plot (H) and ROC curve analyses (I) of the key BAs between the DKD group and the NonDKD group. J The OPLS-DA score plot between the group in diabetic condition (both DKD and T2DM group) and the CON group. The SUS-plot (K) and ROC curve analyses (L) of the key BAs between the group in diabetic condition and CON group. The levels of the key fecal BAs, including M 7-KDCA, N TLCA-3S and O GLCA-3S. *P < 0.05, **P < 0.01.

The ratios between conjugated- and unconjugated- BAs

The ratios between conjugated BAs and unconjugated BAs detected in the plasma, feces and urine of patients with T2DM, patients with DKD and healthy individuals, were calculated, showing the increased conjugated/unconjugated ratios of CA, DCA, CDCA, UDCA, and HCA in the plasma, and of CA, CDCA and LCA in the feces (Table 1). These findings suggested that the patients with DKD have increased transition from unconjugated- to conjugated- BAs.

Table 1 The calculated ratios between some conjugated and unconjugated bile acids in the study participants.The correlation matrix of BAs

The correlation matrix for BAs in study participants were performed, showing the positive correlation of the key BAs in the feces and the negative correlation of the key BAs in the urine with the key BAs in the plasma. Moreover, it also presented the strong correlation of plasma GCDCA, GHCA and GCA between each other (R2 > 0.7, Fig. 5A). When analyzing their correlation with the clinical indicators reflecting DKD progression, including hemoglobin, serum albumin, serum creatinine, blood glucose, serum triglyceride, total cholesterol, low-density lipoprotein, high-density lipoprotein, parathyroid hormone, glycated hemoglobin, eGFR and urinary protein and urinary microalbumin in 24 h, it presented the strong correlation of GCDCA with eGFR (R2 = 0.825, Fig. 5B), 24 h urinary protein (R2 = 0.863, Fig. 5C) and 24 h urinary microalbumin (R2 = 0.852, Fig. 5D). In addition, the fecal levels of GLCA, 7-KDCA and CDCA-3Gln also had the strong correlations with eGFR (R2 > 0.7, Fig. 5E–G). These results indicated the optimal correlation of BAs within the plasma, feces and urine, and the levels of GCDCA in plasma, and the levels of GLCA, 7-KDCA and CDCA-3Gln in the feces were the key BAs reflecting the impairment of renal function in patients with DKD.

Fig. 5: Correlation matrix for the BAs in the study participants.

A The correlation heat map of the key BAs in the plasma, fecal and urine samples of the study participants. The blue area shows strong positive correlation, the red area shows strong negative correlation. The specific Pearson’s r value is presented in each cell. The cells in white were defined as R=0. Scatter plots show the correlation of plasma levels of GCDCA with B eGFR, C 24-h urinary protein, and D 24-h urinary microprotein. Scatter plots show the correlation of eGFR with the fecal levels of E GLCAA, F 7-KDCA, and G CDCA-3Gln. The green dots, blue dots and red dots represent the samples in the CON group, T2DM group and DKD group, respectively.