Altered HDL proteome predicts incident CVD in chronic kidney disease patients

Abstract

Patients with chronic kidney disease (CKD) are at high risk for CVD. However, traditional lipid risk factors, including low HDL levels, cannot completely explain the increased risk. Altered HDL proteome is linked with both CVD and CKD, but the role of HDL proteins in incident CVD events in patients with CKD is unknown. In this prospective case-control study, we used targeted proteomics to quantify 31 HDL proteins in 92 subjects (46 incident new CVD and 46 one-to-one matched controls) at various stages of CKD. We tested associations of HDL proteins with incident CVD using matched logistic regression analysis. In the model fully adjusted for clinical confounders, lipid levels, C-reactive protein, and proteinuria, no significant associations were found for HDL-C, but we observed inverse associations between levels of HDL proteins paraoxonase/arylesterase 1 (PON1), paraoxonase/arylesterase 3 (PON3), and LCAT and incident CVD. Odds ratios (per 1 SD) were 0.38 (0.18-0.97, P = 0.042), 0.42 (0.20-0.92, P = 0.031), and 0.30 (0.11-0.83, P = 0.020) for PON1, PON3, and LCAT, respectively. Apolipoprotein A-IV remained associated with incident CVD in CKD patients in models adjusted for clinical confounders and lipid levels but lost significance with the addition of C-reactive protein and proteinuria to the model. In conclusion, levels of four HDL proteins, PON1, PON3, LCAT, and apolipoprotein A-IV, were found to be inversely associated with incident CVD events in CKD patients. Our observations indicate that HDLs' protein cargo, but not HDL-C levels, can serve as a marker-and perhaps mediator-for elevated CVD risk in CKD patients.

Keywords: CKD; CVD; HDL; HDL proteomics; HDL-C levels; MS; case-control study; matched logistic regression analysis; parallel reaction monitoring.

Fig. 1

PRM analysis of HDL proteins in CKD patients with or without incident CVD. A: APOA4, (B) LCAT, (C) PON1, and (D) PON3. HDL was isolated, and HDL proteins were analyzed as described in Materials and methods section. The average levels of proteins in HDL isolated from the control group were set at an arbitrary unit of one. The box plots show the distribution of the data of HDL proteins (median and interquartile ranges), whereas the dots represent individual data points. Crosses (x) represent 99% and 1% levels. The small squares within the boxes represent mean levels. P values are from a Mann-Whitney U test.

Fig. 2

ORs of HDL proteins for incident CVD. Unadjusted ORs, 95% confidence interval (CI), and P values were obtained from a one-to-one matched multinomial logistic regression analysis. ORs are per SD increase in differences of levels of HDL proteins between matched subjects.

Fig. 3

ORs of HDL proteins for incident CVD after adjustment for potential confounders. ORs, 95% confidence interval (CI), and P values were obtained from a one-to-one matched multinomial logistic regression analysis. Model 1: ORs after adjustment for clinical characteristics, including age, hypertension status, present smoker, statin use, BMI, and eGFR; model 2: ORs after adjustment for clinical characteristics (see model 1) and levels of lipids (HDL-C, LDL-C, and triglycerides); model 3: ORs after adjustment for clinical characteristics (see model 1), levels of lipids (see model 2), proteinuria, and CRP. ORs are per SD increase in differences in levels of HDL proteins between matched subjects. BMI, body mass index.

Fig. 4

OR of APOA4 by stratification for incident CVD. Unadjusted ORs of APOA4 in HDL predicting incident CVD are obtained from a logistic regression analysis in stratified subgroups to assess any differences in ORs across categories of subject characteristics, including diabetic status (yes vs. no), gender (female vs. male), age (<55 vs. >55), hypertension (yes vs. no), eGFR (<45 vs. >45 ml/min/1.73 m²), HDL-C (<45 vs. >45 mg/dl), CRP (<0.38 vs. >0.38 mg/dl), and UPC ratio (<1.25 vs. >1.25). P values for interaction were obtained from a multivariable logistic regression analysis with an interaction term (APOA4 times one clinical characteristic) added into the model. UPC ratio, urinary protein to creatinine ratio.