2.1. Demographic and baseline characteristics of subjects
A total of 90 subjects were enrolled in this study based on the inclusion and exclusion criteria, categorized into three groups: 30 patients with DKD, 30 patients with DM, and 30 healthy individuals as NCs.
The NC group comprised 14 males and 16 females, with a median age of 54.00 years (48.00-59.25). The DM group included 15 males and 15 females, with a median age of 57.00 years (54.00-58.00). The DKD group consisted of 17 males and 13 females, with a median age of 60.50 years (53.50-71.00). Gender distribution did not significantly differ among the three groups (P > 0.05). The SCr level in the DKD group [79.95 (58.75-121.83)μmol/L] was significantly elevated compared to both the DM [57.00 (46.75-65.25)μmol/L] and NC groups [61.80 (54.13-76.60)μmol/L], (P<0.001, P = 0.014, respectively). The estimated eGFR in the DKD group (76.44 ± 31.83 ml/min/1.73m²) was lower than that of the DM group (105.77 ± 6.54 ml/min/1.73m²) (P<0.001). BUN levels were higher in the DKD group [6.98 (5.62-9.29)mmol/L] compared to both the DM [4.83 (4.49-5.75)mmol/L] and NC groups [4.87 (4.29-5.70)mmol/L] (P<0.001 for both). Detailed general and clinical data are presented in Table 1.
Table 1. The general information and laboratory indicators of the subjects.
|
NC (n=30)
|
DM (n=30)
|
DKD (n=30)
|
DKD early (eGFR≥ 90ml/min/1.73m2)(n=12)
|
DKD middle (eGFR< 90ml/min/1.73m2) (n=18)
|
P value: (NC vs DM vs DKD) or (DM vs DKD)
|
P value: DKD early vs DKD middle
|
Age(years)
|
54.00(48.00, 59.25)
|
57.00(54.00, 58.00)
|
60.50(53.50, 71.00)
|
54.50(50.50, 61.50)
|
65.50(57.00, 72.75)
|
0.029*
|
0.006**
|
Gender, male(n, %)
|
14(46.7%)
|
15(50%)
|
17(56.7%)
|
5(41.7%)
|
12(66.7%)
|
0.733
|
0.176
|
Duration of the disease(years)
|
--
|
3.00(1.00, 10.00)
|
10.00(9.50, 16.00)
|
10.00(8.50,12.00)
|
11.50(9.00, 20.00)
|
<0.001***
|
0.247
|
BMI(Kg/m2)
|
--
|
23.56±2.61
|
25.13±3.13
|
26.33±3.68
|
24.33±2.52
|
0.039*
|
0.088
|
Smoking ≥14Y(n,%)
|
--
|
30%
|
30%
|
33.3%
|
27.8%
|
1.00
|
0.745
|
Metformin(n, %)
|
--
|
90%
|
40%
|
66.7%
|
22.2%
|
<0.001***
|
0.015*
|
SGLT-2 inhibitor
(n, %)
|
--
|
16.70%
|
56.70%
|
58.3%
|
55.6%
|
0.001**
|
0.880
|
TC(mmol/L)
|
5.29±0.88
|
4.75±0.94
|
4.86±1.21
|
5.24±1.19
|
4.61±1.19
|
0.107
|
0.164
|
TG(mmol/L)
|
1.18(0.84,1.63)
|
1.59(0.95,2.06)
|
1.76(1.34,2.21)
|
1.49(1.32,2.42)
|
1.79(1.38,2.21)
|
0.005**
|
0.498
|
LDL(mmol/L)
|
3.15±0.63
|
2.81±0.71
|
2.65±0.92
|
2.83±0.82
|
2.54±0.99
|
0.040*
|
0.409
|
HDL(mmol/L)
|
--
|
1.13(1.00, 1.33)
|
1.05(0.95, 1.22)
|
1.09(0.96, 1.33)
|
1.02(0.95, 1.21)
|
0.267
|
0.511
|
CRP(mg/L)
|
--
|
2.00(1.10, 4.85)
|
2.73(0.91, 3.93)
|
2.18(1.13, 3.60)
|
3.15(0.71, 4.25)
|
0.853
|
0.966
|
FPG(mmol/L)
|
5.16±0.49
|
10.21±3.30
|
7.29±2.04
|
8.11±2.28
|
6.74±1.72
|
<0.001***
|
0.070
|
PPG(mmol/L)
|
--
|
15.20(13.78,18.80)
|
12.15(11.50,15.57)
|
14.83(12.07,18.85)
|
12.00(9.66,12.80)
|
0.001
|
0.006**
|
HbA1c(%)
|
--
|
10.50(9.18, 12.35)
|
8.05(7.10, 9.33)
|
9.10(7.60, 11.05)
|
7.45(6.80, 8.90)
|
<0.001***
|
0.047*
|
UA(μmol/L)
|
313.90±79.90
|
290.20±61.66
|
354.72±105.93
|
315.33±118.76
|
380.98±90.50
|
0.014*
|
0.097
|
BUN(mmol/L)
|
4.87(4.29, 5.70)
|
4.83(4.49, 5.75)
|
6.98(5.62, 9.29)
|
6.46(5.48, 7.45)
|
7.36(5.76,11.56)
|
<0.001***
|
0.099
|
SCr(μmol/L)
|
61.80(54.13, 76.60)
|
57.00(46.75, 65.25)
|
79.95(58.75, 121.83)
|
55.50(43.00, 75.70)
|
110.50(83.25,166.75)
|
<0.001***
|
<0.001***
|
eGFR
(ml/min/1.73m2)
|
--
|
105.77±6.54
|
76.44±31.83
|
105.46±13.47
|
57.09±24.91
|
<0.001***
|
<0.001***
|
UACR(mg/g)
|
--
|
14.12(9.99,21.01)
|
942.61(593.16,1768.03)
|
704.75(466.36,1085.94)
|
1009.45(751.70,2204.73)
|
<0.001***
|
0.075
|
Hb(g/L)
|
142.13±13.69
|
146.07±14.15
|
125.93±17.00
|
137.75±14.19
|
118.06±14.09
|
<0.001***
|
0.001**
|
Alb(g/L)
|
45.40(43.45,47.10)
|
42.15(39.10,45.98)
|
38.25(34.95,40.13)
|
39.10(36.58,39.90)
|
36.10(32.58,40.93)
|
<0.001***
|
0.138
|
Abbreviations: NC, normal controls; DM, diabetic mellitus; DKD, diabetic kidney disease; BMI, body mass index; SGLT-2 , sodium-dependent glucose transporters 2; TC, total cholesterol; TG, triglyceride; LDL, low-density lipoprotein; HDL, high-density lipoprotein; CRP, C-reactive protein; FPG, fasting plasma glucose; PPG, postprandial plasma glucose; HbA1c, hemoglobin A1c; UA, uric acid; BUN, blood urea nitrogen; SCr, serum creatinine; eGFR, estimated glomerular filtration rate; UACR, urine albumin creatinine ratio; Hb, hemoglobin; Alb, albumin. *P<0.05, **P<0.01, ***P<0.001.
2.2. Serum metabolomic analysis in patients with DKD
2.2.1. Qualification and quantification of serum metabolites
We conducted comprehensive analysis of 180 serum metabolites across three groups of patients with DKD. These metabolites included 41 fatty acids (FAs), 39 amino acids (AAs), 25 organic acids (OAs), 21 carnitines, 14 bile acids (BAs), 12 carbohydrates, 7 short-chain fatty acids (SCFAs), 5 phenylpropanoic acids (PAs), 4 indoles, 4 benzoic acids (BAs), 2 peptides, 2 phenols, 1 pyridine, 1 imidazole, 1 nucleotide, and 1 benzenoid. Subsequently, these metabolites were utilized for biomarker screening.
2.2.2. Comparison of serum metabolic profiles
A principal component analysis (PCA) model was used to assess the serum metabolite profiles of the subjects. The results revealed distinct separation trends among the DKD, DM, and NC groups (P=0.027), indicating significant differences in their overall serum metabolic profiles, as depicted in Figure 1.
Building on the PCA model, we employed partial least squares discriminant analysis (PLS-DA) to delve deeper into the serum metabolic profiles of the three groups. This approach improved differentiation between the groups, highlighting substantial metabolic variations (P<0.001), as illustrated in Figure 2
Based on the PLS-DA model, an orthogonal partial least-squares discrimination analysis (OPLS-DA) model was further established for multi-dimensional analysis to preliminarily screen metabolites, contributing significantly to the differences in metabolic profiles among groups. Model reliability was validated via a 1000-time random permutation test, confirming robustness and significance. R2Y and Q2Y values were calculated for each permuted model, with results depicted in Figure 3 illustrating clear separation trends between DKD and DM groups, DKD and NC groups, and among DKD subgroups. All models demonstrated Q2Y values > 0.2 and Y-axis intercepts < 0 in permutation test curves (Figure S1), indicating non-overfitting and statistically significant differences in serum metabolic profiles among groups. These findings underscore distinct metabolic differences between DKD and DM groups, DKD and NC groups, and among DKD subgroups.
2.2.3. Correlation analysis between different serum metabolites and clinical indicators
We utilized multidimensional analysis (OPLS-DA model) with a VIP threshold > 1 and P < 0.01 to identify 60 significantly different metabolites between the DKD and NC groups. Among these, 28 metabolites were decreased in the DKD group, while 32 were increased, predominantly belonging to the carnitine class (Table S1). Correlation analysis revealed that 15 of the elevated serum metabolites in DKD showed a positive correlation with SCr levels, whereas 7 of the decreased metabolites showed a negative correlation (Figure S2). When comparing the DKD and DM groups, we identified 39 significantly different metabolites (VIP > 1, P < 0.01), mainly amino acids. Among these, 3 metabolites were decreased in DKD, while 36, including serum α-hydroxyisobutyric acid, were increased (Table S2). Analysis showed that 28 of the elevated metabolites in DKD had a negative correlation with eGFR, and 29 showed a positive correlation with UACR, whereas 1 of the decreased metabolites correlated positively with eGFR and 2 negatively with UACR (Figure 4). Further analysis within DKD subgroups revealed 23 significantly different metabolites between the DKD middle and DKD early groups (VIP > 1, P < 0.01), predominantly amino acids (Table S3). Among these, 5 metabolites were decreased in the DKD middle group, while 18 were increased. Analysis indicated that 13 of the elevated metabolites in the DKD middle group negatively correlated with eGFR, and 4 positively correlated with UACR, whereas 5 of the decreased metabolites showed a positive correlation with eGFR and 5 a negative correlation with UACR (Figure S3).
2.3. Analysis of gut microbiota
The rarefaction curves of all three study groups showed a plateauing trend, indicating that the sequencing depth of each sample closely matched the expected level. This suggests the adequacy and reliability of the sequencing data volume for subsequent analyses (Figure S4).
2.3.1. Analysis of alpha diversity
We analyzed the microbial diversity of the three groups. Significant differences were observed in the Simpson, Chao1, and ACE indices among the groups (P < 0.05) for alpha diversity. Specifically, the Simpson index was significantly lower in the DKD group compared to the NC group (P < 0.05) (Figure 5A).
Comparing the DKD middle group to the DKD early group, the ACE, Chao1, and Shannon indices were significantly lower (P < 0.05, P < 0.05, P < 0.05), while no significant difference was observed in the Simpson index (P = 0.091) (Figure S5A).
2.3.2. Analysis of beta diversity
To characterize the overall microbial features of the three groups, beta diversity comparison was performed using PERMANOVA. Principal coordinates analysis (PCoA) based on Bray–Curtis distance revealed significant differences in overall bacterial community structure among the groups (PERMANOVA test, DKD vs DM vs NC: P = 0.001) (Figure 5B). Adonis analysis indicated significant differences between the DKD and NC groups (F = 3.241, P = 0.001), but not between the DKD and DM groups (F = 1.311, P = 0.076). Non-metric Multidimensional Scaling (NMDS) analysis based on Weighted Unifrac distance also confirmed significant differences among the groups (Stress = 0.076) (Figure 5C). Adonis analysis further supported significant differences between the DKD and NC groups (F = 8.112, P = 0.003), while no significant differences were observed between the DKD and DM groups (F = 1.797, P = 0.119).
PCoA indicated no significant difference between the DKD middle and DKD early groups (P = 0.133). However, NMDS analysis revealed significant differences (Stress = 0.050) (Figure S5C). Despite this, Adonis analysis did not show significant differences (F = 0.512, P = 0.642).
2.3.3. Taxonomic changes in microbial composition
Next, we analysed the microbial composition at different taxonomic levels. The microbial composition at the phylum and genus levels were shown in (Figure 5D-E) (Figure S5D-E).
LEfSe analysis identified differentially abundant microbial features among NCs, DKD patients, and DM patients. Specifically, 24 species were differentially abundant between DKD patients and DM patients (Figure 6A), 32 species between DKD patients and NCs (Figure 6B), and 39 species between DKD middle patients and DKD early patients (Figure 6C) (LDA value > 2, P < 0.05) (Table S (4-6)).
Additionally, we identified functional alterations in the gut microbiota of DKD patients. Pathways such as “Valine, leucine, and isoleucine degradation”, “Biofilm formation Vibrio cholerae”, “Glyoxylate and dicarboxylate metabolism”, and “Tryptophan metabolism” were significantly enriched in DKD compared to NCs and DM patients (LDA > 2, P < 0.05) (Figure S6(A-B)). Further analysis within DKD subgroups revealed significant enrichment in pathways including "β Lactam resistance", "Folate biosynthesis", and "Lipopolysaccharide biosynthesis" in the DKD middle group compared to the DKD early group (LDA > 2, P < 0.05) (Figure S6C).
2.3.4. Correlation analysis between gut microbiota and clinical indicators
The correlation analysis of the 24 differential gut microbiota between the DKD and DM groups and the clinical indicators of the patients revealed significant associations. Specifically, g_Rikenella showed a positive correlation with UACR (r = 0.44, P < 0.001) and a negative correlation with eGFR (r = -0.41, P < 0.01). Conversely, g_Prevotella, g_Agathobacter, and g_Haemophilus exhibited a strong negative correlation with UACR (r = -0.33, P < 0.01; r = -0.36, P < 0.01; r = -0.33, P < 0.01) and a positive correlation with eGFR (r = 0.41, P < 0.01; r = 0.35, P < 0.01; r = 0.34, P < 0.01). Additionally, g_T34, f_Pasteurellaceae, o_Pasteurellales, o_Oscillospirales, and f_Ruminococcaceae were positively correlated with eGFR (r = 0.30, P = 0.02; r = 0.28, P = 0.03; r = 0.28, P = 0.03; r = 0.32, P = 0.01; r = 0.30, P = 0.02) and negatively correlated with UACR (r = -0.31, P = 0.02; r = -0.29, P = 0.02; r = -0.27, P = 0.02; r = -0.34, P = 0.03; r = -0.36, P < 0.01) (Figure 7).
2.4. Construction of disease prediction model using serum metabolites and gut microbiota
2.4.1. Screening and identification of predictive serum metabolic markers
The specific screening process is illustrated in Figure 8 Using both univariate analysis (t-test or Mann-Whitney U test, P < 0.05) and multivariate analysis (OPLS-DA, VIP > 1, P < 0.05), we identified 10 common metabolites out of 180 serum metabolites. These 10 metabolites were subsequently subjected to the LASSO algorithm. The top three serum metabolites were selected based on the ranking of non-zero LASSO coefficients: Imidazolepropionic acid, Adipoylcarnitine, and 1-Methylhistidine.
We opted for a logistic regression model and applied 10-fold cross-validation (CV) to assess the classification performance of the model on the subject cohort. The evaluation metrics used were CV-area under the receiver operating characteristic curve (CV-AUROC) and CV-area under the precision-recall curve (CV-AUPR).
The disease prediction model using these three metabolites exhibited robust discriminatory capability for diagnosing patients with DKD from those with DM (AUROC = 0.9, AUPR = 0.883), distinguishing between patients with DKD and NCs (AUROC = 0.841, AUPR = 0.725), and differentiating DKD early patients from DKD middle patients (AUROC = 0.894, AUPR = 0.933) (Figure 9 (A-C)).
2.4.2. Prediction of DKD and its stages based on differential gut microbiota
The logistic regression model evaluated the discriminatory power of differential bacterial genera between groups (DKD vs. DM, DKD vs. NC): g_Prevotella and g_Faecalibacterium (LDA ≥ 4, P < 0.05), and a significantly different bacterial genus between subgroups: g_Klebsiella (LDA ≥ 4, P < 0.05). The disease prediction model based on these three bacterial genera demonstrated strong discriminatory power to diagnose patients with DKD from patients with DM (AUROC = 0.69, AUPR = 0.772), to discriminate between patients with DKD and NCs (AUROC = 0.95, AUPR = 0.953), and to differentiate DKD early patients from DKD middle patients (AUROC = 0.759, AUPR = 0.837) (Figure 9 (A-C)).
2.4.3. Prediction of DKD and its stages based on serum metabolites and bacterial genera
The disease prediction model, a logistic regression model, based on the selected three serum metabolites combined with three bacterial genera, exhibited excellent discriminatory ability to diagnose patients with DKD from patients with DM (AUROC = 0.939, AUPR = 0.940), discriminate between patients with DKD and NCs (AUROC = 0.976, AUPR = 0.973), and differentiate DKD early patients from DKD middle patients (AUROC = 1.000, AUPR = 1.000) (Figure 9 (A-C)).
2.5. Correlation analysis between serum metabolites and gut microbiota
2.5.1. Pathway analysis of differential metabolites between DKD and DM groups
We conducted pathway analysis on differential metabolites between the DKD and DM groups (36 differential metabolites identified through both univariate and multivariate analyses: univariate analysis P < 0.05, and VIP > 1 and P < 0.05 in the OPLS-DA model for multivariate analysis). These metabolites were analyzed using MetaboAnalyst 5.0 (http://www.metaboanalyst.ca), referencing the KEGG database to generate pathway diagrams. Pathways with a P< 0.05 or an impact value ≥ 0.1 were considered significantly altered (Figure S7). Seven pathways met these criteria, including glycine, serine, and threonine metabolism; tryptophan metabolism; citrate cycle (TCA cycle); alanine, aspartate, and glutamate metabolism; phenylalanine metabolism; arginine and proline metabolism; and pentose phosphate pathway. These pathways play critical roles in the progression of DKD.
A related network (Figure 10) illustrated alterations in seven metabolic pathways and associated serum metabolites in DKD patients. Compared to the DM group, DKD patients showed elevated levels of 5-aminolevulinic acid, pyruvic acid, and dimethylglycine in the glycine, serine, and threonine metabolism pathway, alongside lower tryptophan levels. The tryptophan metabolism pathway exhibited enrichment with indolelactic acid and kynurenine. The citrate cycle was enriched with pyruvic acid and isocitric acid. The phenylalanine metabolism pathway demonstrated higher levels of phenylacetylglycine, hippuric acid, and pyruvic acid. The alanine, aspartate, and glutamate metabolism pathway showed enrichment with N-acetylaspartic acid and pyruvic acid. The arginine and proline metabolism pathway was enriched with citrulline and 4-hydroxyproline. Additionally, the pentose phosphate pathway exhibited increased levels of gluconolactone and pyruvic acid.
2.5.2. Integrating multi-omics analysis (correlation analysis between differential microbiota and metabolites in DKD and DM groups)
We further explored the association between gut microbiota and serum metabolites in relation to DKD using Spearman correlation coefficients. We selected 24 differentially abundant microbiota at various taxonomic levels between the DKD and DM groups (LDA > 2, P < 0.05) and identified 39 significantly different metabolites between the two groups using the OPLS-DA model (VIP > 1, P < 0.01). A Spearman correlation analysis was performed between these differential gut microbiota and serum metabolites, visualized as a correlation coefficient matrix heatmap (Figure S8).
Based on impact values and p values from KEGG analysis, glycine, serine, and threonine metabolism, tryptophan metabolism, citrate cycle, and phenylalanine metabolism emerged as pivotal pathways in the progression from DM to DKD. A metabolic network (Figure 11) was created using enriched serum metabolites within these pathways and 11 genus-level microbiota selected from the 24 differentially abundant microbiota.
Three metabolites (5-aminolevulinic acid, pyruvate, and dimethylglycine) were found at higher concentrations in the serum of the DKD group within the glycine, serine, and threonine metabolism pathway, while tryptophan levels were lower. Among the 11 differentially abundant microbiota at the genus level, g_Rikenella showed a positive correlation with 5-aminolevulinic acid (r = 0.29, P = 0.02), g_Eubacterium_hallii_group exhibited a positive correlation with pyruvate (r = 0.33, P = 0.01), and both g_Muribaculaceae and g_Rikenella demonstrated a positive correlation with dimethylglycine (r = 0.32, P = 0.01; r = 0.28, P = 0.03). Conversely, g_Agathobacter, g_Faecalibacterium, and g_Haemophilus were positively correlated with tryptophan (r = 0.30, P = 0.02; r = 0.33, P < 0.01; r = 0.26, P = 0.04).
Two metabolites enriched in the tryptophan metabolism pathway, indole lactic acid and kynurenine, were found at higher levels in the DKD group, while tryptophan levels were lower. Among the 11 differentially abundant microbiota at the genus level, g_Eubacterium_hallii_group exhibited a positive correlation with pyruvate (r = 0.33, P = 0.01).
Three metabolites enriched in the phenylalanine metabolism (phenylacetyglutamine, hippuric acid, and pyruvate) were found at higher levels in the DKD group. Among the 11 differentially abundant microbiota at the genus level, g_ Parabacteroides, g_ Eubacterium_hallii_group, and g_Muribaculacee exhibited a positive correlation with hippuric acid (r = 0.46, P < 0.01; r = 0.35, P = 0.02; r = 0.33, P = 0.03), while g_Romboutsia showed a negative correlation (r = -0.35, P = 0.02). Additionally, g_Muribaculaceae, g_Rikenella, g_Parabacteroides, and g_Eubacterium_hallii_group were positively correlated with phenylacetylglutamine (r = 0.36, P < 0.01; r = 0.37, P < 0.01; r = 0.33, P = 0.01; r = 0.27, P = 0.03), whereas g_Faecalibacterium and g_Haemophilus demonstrated a negative correlation (r = -0.30, P = 0.02; r = -0.33, P = 0.01).