Patients
Stool samples from a total of 52 patients with varying stages of CKD were collected in this study: CKD3A (n = 12), CKD3B (n = 11), CKD4 (n = 15), CKD5 (n = 4) and ESRD (n = 10) (Table 1). Patients’ characteristics are summarized in Table 1. Among 52 patients, 31 were reported to have Type 2 diabetes mellitus and 7 patients were reported to have human immunodeficiency virus (HIV) infection. As expected, urine protein creatinine ratio, serum creatinine and blood urea nitrogen level increased with progressing stages of CKD (CKD 3A to ESRD). There was no significant difference in fat, protein, carbohydrates, dietary fiber and calorie intake between CKD patients with different stages (Supplementary Table S1).
Alpha and beta-diversity
Richness and Shannon index were not significantly different between different patient groups, meanwhile the CKD5 group showed a significant decrease in Simpson diversity compared with CKD 3A (FDR < 0.06), CKD3B (FDR < 0.06) and CKD4 (FDR < 0.07) groups (Fig. 1A). The variation within patient groups was greater than variation across groups, and the overall gut microbiota composition of different CKD patient groups was not statistically different (Adonis test p-value = 0.057, Fig. 1B) even though ESRD patients tended to group separately from the patients with earlier stage patients (CKD 3A/3B, Fig. 1B). Mean Bray–Curtis distance within the cohorts was shown to be higher in the later stages of CKD (CDK 4/5 and ESRD) when compared with samples from subjects with early stage disease (CKD 3A/3B) (Fig. 1C). Mean Bray–Curtis distance of samples from the earliest stage group (CKD 3A) showed that the microbial composition of the subjects in the ESRD group is farthest away from patients in CKD 3A stage (Fig. 1D). Spearman correlation analysis showed that the dissimilarity in BMI between the subjects did not show any significant correlation with the taxonomic dissimilarity between the subjects (p-value = 0.182, Supplementary Fig. S1). In addition, the HIV status of the subjects did not have any significant effect on the taxonomic dissimilarity (ANOVA FDR > 0.19, Supplementary Fig. S2).
Alpha and beta diversity. (A) Richness, Shannon and Simpson index. (B) Ordination plot using Bray–Curtis dissimilarity score. Adonis test p-value = 0.057. (C) Mean Bray–Curtis distance within the cohorts. P-values were calculated by ANOVA following by Tukey post-hoc test. CKD 3B vs. CKD 3A p = 0.99; CKD 4 vs. CKD 3A p < 0.001; CKD 5 vs. CKD 3A p < 0.001; ESRD vs. CKD 3A p < 0.01; CKD 4 vs. CKD 3B p < 0.001; CKD 5 vs. CKD 3B p < 0.001; ESRD vs. CKD 3B p < 0.01; CKD 5 vs. CKD 4 p = 0.36; ESRD vs. CKD 4 p = 0.97; ESRD vs. CKD 5 p = 0.56. (D) Mean Bray–Curtis distance from the earliest stage CKD (CKD 3A).
Gut microbiota composition
A total of 24 microbial species were significantly different between early stage patients (CKD 3A/B) and ESRD patients (FDR < 0.1, Fig. 2A). A total of 57 microbial species were significantly associated with the disease severity (Supplementary Table S2). The top six positive and negative associations are shown in Fig. 2B, among which Eubacterium rectale and species belonging to Collinsealla genera demonstrated the strongest association. At genus level, 19 microbial genera were significantly associated with the severity of CKD (Supplementary Table S3), and the top six positive and negative correlations are shown in Fig. 2C.
Microbial composition. (A) Significant microbial species between early stage CKD (n = 23) and ESRD patients (n = 10). Wilcox-rank sum test was used to calculate p-values and multiple testing was corrected using FDR. Relative abundance of microbial species > 1% in one of the subjects and with FDR < 0.1 were shown. (B) Microbial species significantly associated with the severity of CKD. Association was calculated by Maaslin2. Top six species showing negative and positive association with disease severity were shown. (C) Microbial genera significantly associated with the severity of CKD. Association was calculated by Maaslin2. Top six species showing negative and positive association with disease severity were shown. ***FDR < 0.0005; **FDR < 0.005; *FDR < 0.05.
Further, we identified 22 microbial species harboring genes encoding butyrate kinase and butyryl-coA transferase in patients. The butyrate synthesis associated species composition was estimated by summing the relative composition of 22 known butyrate synthesis species identified in the subjects at > 1% relative composition. The summed up relative composition from these species represents the butyrate synthesis capacity. The Linear mixed effect model showed the butyrate synthesis associated taxa was negatively associated with the disease severity (p-value = 6e−04) and ESRD patients had the lowest level of butyrate synthesis associated taxa (Fig. 3A). Spearman correlation analysis showed that the serum creatinine levels were significantly correlated with butyrate synthesis associated species composition (Fig. 3B, Left). The albumin/creatinine ratio in the serum showed significantly negative association with butyrate synthesis associated species composition (Fig. 3B, Right). The level of Bifidobacterium was increased with disease severity, meanwhile lactobacillus was decreased with disease severity (Fig. 3C). Interestingly, Methanobacteria increased from early to late stage CKD patients, but it was not detected in patients with ESRD (Fig. 3D).
Key microbial taxa. (A) Butyrate synthesis associated taxa. P-value was calculated using linear mixed effect modelling. (B) The serum creatinine level shows statistically significant (p-value = 0.01) negative association with butyrate synthesis associated species composition (Left). The albumin/creatinine ratio in the serum shows statistically significant (p-value = 3.74e−5) negative association with butyrate synthesis associated species composition (Right). The spearman correlation was used for estimating association. (C) Bifidobacterium and Lactobacillus. (D) Methanobacteria. ***FDR < 0.0005; **FDR < 0.005; *FDR < 0.05.
Next, we identified five microbial genera and eight microbial species that were different between the diabetic patients and non-diabetic patients (FDR < 0.01, Fig. 4A,B). However, the diabetic status didn’t significantly affect the gut microbial community structure (Adonis test p-value > 0.874, Supplementary Fig. S3). Eight microbial genera and 11 microbial species were different between patients with high (albumin/creatinine ratio > 300) and low (albumin/creatinine ratio < 30) albuminuria (FDR < 0.01, Fig. 4C,D).
Different microbial species between different groups. (A) Comparison between diabetic patients and non-diabetic patients at genera level. (B) Comparison between diabetic patients and non-diabetic patients at species level. (C) Comparison between patients with high albuminuria and low albuminuria at genera level. (D) Comparison between patients with high albuminuria and low albuminuria at species level.
Functional metagenome
A total of 38 Kyoto Encyclopedia of Genes and Genomes (KEGG) modules were different between early stage patients (CKD 3A/B) compared with ESRD patients (FDR < 0.1, Fig. 5A). Significant association was found between 17 KEGG modules and the disease severity (FDR < 0.1, Fig. 5B, Supplementary Table S4), among which 14 modules were negatively correlated with disease severity, while the remainder were positively correlated with the disease severity including dipeptide transport system and two-component regulatory systems involved in cell wall stress response and potassium transport. A total of 35 Cluster of Orthologous groups (KOG) enzymes were significantly different between early stage patients (CKD 3A/B) compared with ESRD patients (FDR < 0.1, Fig. 6A, Supplementary Table S5). Significant association was found between six KOG enzymes and the disease severity (FDR < 0.1, Fig. 6B, Supplementary Table S6), among which five were negatively correlated with the disease severity and capreomycidine synthase was positively correlated with the disease severity.
KEGG modules. (A) Significant KEGG modules between early stage CKD (n = 23) and ESRD patients (n = 10). Wilcox-rank sum test was used to calculate p-values of top 100 modules and multiple testing was corrected using FDR (FDR < 0.05 were shown). (B) KEGG modules significantly associated with the severity of CKD (FDR < 0.05). Association was calculated by Maaslin2. KEGG definitions: Supplementary Table S4. ***FDR < 0.0005; **FDR < 0.005; *FDR < 0.05.
KOG enzymes. (A) Significant KOG enzymes between early stage CKD 3A/B (n = 23) and ESRD patients (n = 10). Wilcox-rank sum test was used to calculate p-values of top 100 KOG enzymes and multiple testing was corrected using FDR (FDR < 0.05 were shown). KOG definitions: Supplementary Table S5. (B) KOG families significantly associated with the severity of CKD (FDR < 0.05). Association was calculated by Maaslin2. K20712 3-(hydroxyamino)phenol mutase; K21029 molybdopterin-synthase adenylyltransferase; K20816 streptothricin hydrolase; K20615 capreomycidine synthase; K20866 glucose-1-phosphatase; K20452 dimethylmaleate hydratase large subunit. ***FDR < 0.0005; **FDR < 0.005; *FDR < 0.05.
Stool butyrate level
To confirm the impact of butyrate synthesis associated species on butyrate production, we measured the butyrate concentration in stool samples. Due to the sample availability, stool samples were collected from 8 patients with stage 3A, 6 patients with stage 3B, and 8 patients with stage 4 and 5. Stool butyrate level was significantly reduced in CDK patients with stage 4 and 5 compared with patients with stage 3A (Fig. 7, ANOVA overall p-value = 0.024, Tukey’s multiple comparison test adjusted p-value = 0.0216).
Stool butyrate level in CKD patients. Overall p-value was calculated using one-way ANOVA (p-value = 0.0240). Adjusted p-value was calculated using Tukey’s multiple comparison test (Stage 3A vs. Stage 3B adjusted p-value = 0.2962; Stage 3A vs. Stage 4 and 5 adjusted p-value = 0.0216; Stage 3B vs. Stage 4 and 5 adjusted p-value = 0.4743).
Source: Ecology - nature.com
