Clinical and biochemical characteristics of participants
Among 438 patients with biopsy-confirmed MAFLD were included in the analysis, mean age was 41±14 years, 72.6% were men, mean eGFRCKD-EPI was 114.4±20 mL/min/1.73 m2, and 11.2% had abnormal albuminuria (defined as ACR ≥30 mg/g). According to their eGFRCKD-EPI levels, 11.2% of enrolled patients had CKD stage 1 (defined as eGFR ³90 mL/min/1.73 m2), and 2.5% had CKD stage 2 (eGFR 89-60 mL/min/1.73 m2). Only two patients had CKD stage ≥3 (eGFR <60 mL/min/1.73 m2). Among 15 healthy controls who were also included, age was 40±3 years and 60% were men, and none of them had abnormal albuminuria and kidney injury.
Patients with MAFLD were further divided into two groups according to the presence of abnormal albuminuria. As shown in Supplementary Figure 1, healthy controls were defined as HC, MAFLD patients with normal albuminuria were defined as MAFLD-ACR0, while the MAFLD patients with abnormal albuminuria were defined as MAFLD-ACR1. Compared with HC, all MAFLD patients had significantly higher levels of blood pressure, serum transaminases, LDL-C and lower HDL-C concentrations. Conversely, the three groups were similar for age and sex.
As shown in Table 1, compared to the MAFLD-ACR0 group, patients with MAFLD and abnormal albuminuria (MAFLD-ACR1) were more likely to be men, older, and to have higher blood pressure and hemoglobin A1c, and lower eGFR levels (that were within the normal range in most of our patients, 8.5% had CKD stage 1 and 0.5% had CKD stage 2 in the MAFLD-ACR0 group). Patients with MAFLD and abnormal albuminuria also had a higher prevalence of pre-existing type 2 diabetes and hypertension compared to those belonging to the MAFLD-ACR0 group. Liver histology features (steatosis, hepatocyte ballooning, and lobular inflammation) were not significantly different according to albuminuria group. Interestingly, there was a trend toward a higher proportion of subjects with liver fibrosis in the MAFLD-ACR1 group (Table 1).
Overall BA profiles in MAFLD patients stratified by abnormal albuminuria
Overall, glycine conjugated primary BAs (Gcon PBAs), unconjugated secondary BAs (uncon SBAs), unconjugated primary BAs (uncon PBAs), and glycine conjugated secondary BAs (Gcon SBAs) comprised more than 80% of BA pool in serum (Figure 1A). Compared to the HC group, BA profiles changed significantly in MAFLD patients, exhibiting higher plasma levels of conjugated BAs (con BAs), glycine conjugated BAs (Gcon BAs), conjugated primary BAs (Con PBAs), Gcon PBAs, and taurine-conjugated primary BAs (Tcon PBAs). To further investigate the differences in BA profiles, the circulating levels of these BAs, con BAs, Gcon BAs, Con PBAs (including both Tcon and Gcon PBAs) were markedly up-regulated in the MAFLD-ACR1 group compared with the MAFLD-ACR0 group. In addition, we found significantly higher ratios of con BAs to total BAs (con BAs/TBAs), Gcon BAs to TBAs (Gcon BAs/TBAs), and con BAs to unconjugated BAs (con BAs/ Uncon BAs) in the MAFLD-ACR1 group compared with the MAFLD-ACR0 group (Figure 1).
Association between NorCA levels and risk of abnormal albuminuria in MAFLD
The circus plot showed differential BAs and their fold changes by comparing MAFLD-ACR0 vs. MALFD-ACR1 in Figure 2A. Compared to the MAFLD-ACR0 group, we found that the MAFLD-ACR1 group had significantly higher levels of plasma GCDCA, TCDCA, and NorCA (Figure 2B-D).
Next, we analysed 38 BA biomarkers and 7 clinical biomarkers (i.e., age, sex, BMI, diabetes, hypertension, eGFR and ACR) in the heatmap (Figure 3A). Firstly, 6 primary BAs (i.e., HCA, THCA, TCA, GHCA, GCA and GCDCA) showed significant inverse associations with age, while 7 secondary BAs (6-ketoLCA, UDCA, HDCA, LCA, 7-ketoLCA, 12-ketoLCA and TLCA) and a sulfated BA, the LCA-3S, showed positive associations with age. Nine BAs (TCA, TCDCA, GCA, GCDCA, 6-ketoLCA, TLCA, TDCA, CDCA-3Glu and NorCA) were significantly associated with BMI. As regards to the presence of type 2 diabetes, we found that GHCA was inversely associated with diabetes, while 8 BAs (CA, CDCA, GCDCA, 6-ketoLCA, UDCA, 7-DHCA, 7-ketoLCA and NorCA) were positively associated with diabetes. The presence of hypertension was also associated with 8 BAs (CA, UCA, 6-ketoLCA, UDCA, 7-DHCA, 7-ketoLCA, TLCA and NorCA). Finally, we showed that 5 primary BAs (HCA, TCA, GHCA, GCA and GCDCA) were positively associated with eGFR levels. More importantly, 2 primary BAs (THCA, GHCA), 9 secondary BAs (βUCA, UCA, 6-ketoLCA, UDCA, 7-DHCA, βCA, dehydroLCA, 7-ketoLCA, TLCA) and NorCA showed significant associations with ACR, respectively. Further, the Spearman’s correlation analysis showed that NorCA levels markedly increased with increasing ACR values (Figure 3B).
To further study the positive association between levels of NorCA and albuminuria, we performed logistic regression analyses. As shown in Table 2, in univariable logistic regression, there was a significant association between higher NorCA levels and risk of abnormal albuminuria (unadjusted-OR=1.580, 95%CI 1.21-2.06; P=0.001). After adjustment for age, diabetes, hypertension, this association remained statistically significant (adjusted-OR=1.534, 95%CI 1.16-2.03). Notably, the significant association between higher NorCA levels and risk of abnormal albuminuria was not attenuated after further adjustment for sex, BMI, eGFR, presence of steatohepatitis, and histologic stage of liver fibrosis (adjusted-OR=1.525, 95%CI 1.14-2.04; P=0.004).
Biomarker discovery and prediction for abnormal albuminuria in MAFLD
Through BA and clinical biomarkers selection by predictive modelling in logistic regression analysis, the combination of serum BAs and clinical biomarkers enabled us to obtain optimal models for predicting the presence of abnormal albuminuria in patients with MAFLD. Firstly, we conducted 7 predictive models for BA biomarkers including GCDCA, NorCA, TLCA, 7-ketoLCA, TDCA, CA, TCDCA, with AUC values of 0.62 (95%CI 0.54-0.70), 0.61(95%CI 0.52-0.69), 0.53(95%CI 0.44-0.62), 0.51(95%CI 0.42-0.61), 0.49(95%CI 0.39-0.58), 0.46(95%CI 0.39-0.54) and 0.39(95%CI 0.32-0.46), respectively (Figure 4A). Furthermore, 7 clinical biomarkers, and the presence of type 2 diabetes and hypertension, CysC, eGFR, age, BMI and ALT, yielded an AUC of 0.67 (95%CI 0.61-0.74), 0.61 (95%CI 0.55-0.67), 0.66 (95%CI 0.58-0.74), 0.63 (95%CI 0.56-0.71), 0.59 (95%CI 0.52-0.67), 0.58 (95%CI 0.50-0.67) and 0.55 (95%CI 0.47-0.63), respectively (Figure 4B). Next, we tested widely known existing non-invasive fibrosis scores that are used in clinical practice for diagnosing high probability of advanced liver fibrosis, i.e., NAFLD fibrosis score (NFS), FIB-4 index, and fibrosis stage with AUCs of 0.65 (95%CI 0.57-0.72), 0.60 (95%CI 0.52-0.68), and 0.57 (95%CI 0.49-0.66) respectively, in predicting the presence of abnormal albuminuria in patients with MAFLD. However, univariable logistic regression analyses failed to show good performance in most of MAFLD patients. Furthermore, we added BA biomarkers to clinical biomarkers and obtained a predictive model for NorCA plus age, diabetes, and hypertension. This predictive model showed good diagnostic performance for predicting the presence of abnormal albuminuria in MAFLD, with an AUC of 0.74 (95%CI 0.67-0.81) (Figure 4D). More importantly, another predictive model including 4 BA biomarkers (NorCA, TLCA, TDCA and CA) plus 3 clinical biomarkers (diabetes, hypertension and BMI) had even better performance for predicting abnormal albuminuria in patients with MAFLD, with AUC of 0.80 (95%CI 0.74-0.86) (Figure 4E).