Atherosclerosis negatively impacts the prognosis of renal transplant recipients (RTRs) in two ways, (i) in the form of pre-existing, mostly complex atherosclerotic lesions, which are the underlying pathology for CVD and (ii) as de novo atherosclerotic lesion formation in the graft, known as transplant vasculopathy (TV), the single major cause for chronic graft failure (GF) (1Stoumpos S. Jardine A.G. Mark P.B. Cardiovascular morbidity and mortality after kidney transplantation., 2Vascular remodeling in transplant vasculopathy., 3Rahmani M. Cruz R.P. Granville D.J. McManus B.M. Allograft vasculopathy versus atherosclerosis.). Patients receiving kidney grafts often have a long-standing history of end-stage renal disease with dialysis treatment, which is per se associated with a 30- to 40-fold increase in age-adjusted CVD mortality (4Wong M.G. Wanner C. Knight J. Perkovic V. Lowering cholesterol in chronic kidney disease: is it safe and effective?.). After transplantation, CVD risk is still 4- to 6-fold higher and is the leading cause of death in RTRs (4Wong M.G. Wanner C. Knight J. Perkovic V. Lowering cholesterol in chronic kidney disease: is it safe and effective?.). A decrease in renal function over time further contributes to the increased risk. The substrate for this chronic graft functional decline is TV, an atherosclerotic process in the vasculature of the transplanted organ, affecting 50% of allografts after 5 years and 90% after 10 years (2Vascular remodeling in transplant vasculopathy., 3Rahmani M. Cruz R.P. Granville D.J. McManus B.M. Allograft vasculopathy versus atherosclerosis.). Importantly, in RTRs, classical CVD risk factors, such as levels of LDL-C or HDL-C, fail to serve as predictive biomarkers either for CVD events or for de novo atherosclerosis leading to TV-mediated chronic GF (1Stoumpos S. Jardine A.G. Mark P.B. Cardiovascular morbidity and mortality after kidney transplantation., 3Rahmani M. Cruz R.P. Granville D.J. McManus B.M. Allograft vasculopathy versus atherosclerosis., 5Cardiovascular complications after renal transplantation and their prevention.). Thus, the identification of predictive biomarkers represents an unmet clinical need in this patient population. Recent studies indicated that assays capturing the functional properties of HDL lipoproteins might provide clinical information beyond cholesterol levels within these, exemplified by HDL cholesterol efflux capacity being able to predict CVD in the general population (6Ebtehaj S. Gruppen E.G. Bakker S.J.L. Dullaart R.P.F. Tietge U.J.F. HDL (high-density lipoprotein) cholesterol efflux capacity is associated with incident cardiovascular disease in the general population., 7Rohatgi A. Khera A. Berry J.D. Givens E.G. Ayers C.R. Wedin K.E. Neeland I.J. Yuhanna I.S. Rader D.R. de Lemos J.A. Shaul P.W. HDL cholesterol efflux capacity and incident cardiovascular events.) and chronic GF in RTRs (8Annema W. Dikkers A. Freark de Boer J. Dullaart R.P.F. Sanders J.-S.F. Bakker S.J.L. Tietge U.J.F. HDL cholesterol efflux predicts graft failure in renal transplant recipients.), independent of HDL-C levels. However, throughout the atherogenic process, especially LDL particles play a central and pivotal role (9Borén J. Chapman M.J. Krauss R.M. Packard C.J. Bentzon J.F. Binder C.J. Daemen M.J. Demer L.L. Hegele R.A. Nicholls S.J. Nordestgaard B.G. Watts G.F. Bruckert E. Fazio S. Ference B.A. et al.Low-density lipoproteins cause atherosclerotic cardiovascular disease: pathophysiological, genetic, and therapeutic insights: a consensus statement from the European Atherosclerosis Society consensus panel.). Specifically, binding of LDL particles to proteoglycans in the vessel wall is an early key event in the initiation of atherosclerotic lesions, as summarized in the now widely accepted response-to-retention hypothesis of atherogenesis (10Tabas I. Williams K.J. Borén J. Subendothelial lipoprotein retention as the initiating process in atherosclerosis: update and therapeutic implications.). However, thus far, the concept that measures of LDL functionality can be used as predictive biomarkers has not been widely explored. Especially, the susceptibility of LDL to bind to vascular proteoglycans appears promising in this respect. Therefore, in the present work, we investigated, whether in comparison to LDL-C levels, the lipoprotein-proteoglycan binding susceptibility (LPBS) of apoB-containing lipoproteins is prospectively associated with the two clinically relevant atherosclerosis-related outcomes in RTRs, CVD mortality on the one hand and chronic GF on the other.Materials and methods Study design and study populationFor this prospective study, an established and well-characterized patient cohort of adult RTR (TransplantLines) was used (8Annema W. Dikkers A. Freark de Boer J. Dullaart R.P.F. Sanders J.-S.F. Bakker S.J.L. Tietge U.J.F. HDL cholesterol efflux predicts graft failure in renal transplant recipients., 11van Ree R.M. de Vries A.P.J. Oterdoom L.H. The T.H. Gansevoort R.T. Homan van der Heide J.J. van Son W.J. Ploeg R.J. de Jong P.E. Gans R.O.B. Bakker S.J.L. Abdominal obesity and smoking are important determinants of C-reactive protein in renal transplant recipients.). Patients were recruited at the outpatient clinic of the University Medical Center Groningen between August 2001 and July 2003. To lessen the potential confounding effect of early immune-mediated rejection, patients were required to have, at inclusion, a functioning graft for at least 1 year after transplantation. Exclusion criteria were overt congestive heart failure, endocrine abnormalities other than diabetes mellitus, malignancies other than cured skin cancer, and suspected acute infection. 606 of 847 eligible patients volunteered to participate and were included in the cohort (72% consent rate). Nonparticipants were compared with participants with regard to age, sex, BMI, plasma creatinine, creatinine clearance, and proteinuria, and no significant differences were found (8Annema W. Dikkers A. Freark de Boer J. Dullaart R.P.F. Sanders J.-S.F. Bakker S.J.L. Tietge U.J.F. HDL cholesterol efflux predicts graft failure in renal transplant recipients., 11van Ree R.M. de Vries A.P.J. Oterdoom L.H. The T.H. Gansevoort R.T. Homan van der Heide J.J. van Son W.J. Ploeg R.J. de Jong P.E. Gans R.O.B. Bakker S.J.L. Abdominal obesity and smoking are important determinants of C-reactive protein in renal transplant recipients.). Written informed consent was obtained from all participants. The institutional review board approved the study protocol (METc2001/039), which complied with the Declaration of Helsinki. The study is registered at ClinicalTrials.gov with identifier NCT03272854 under the name “The TransplantLines Insulin Resistance and Inflammation Biobank and Cohort Study”. Outcome measures and end points of the studyThe main outcome measure of this study is LPBS. Primary end points are (i) death-censored GF, defined as return to dialysis therapy or retransplantation and (ii) cardiovascular mortality, defined as deaths with the principal cause of death being CV in nature, using codes 410–447 of the International Classification of Diseases, ninth revision (12World Health Organization
International Classification of Diseases : [9th] Ninth Revision, Basic Tabulation List with Alphabetic Index [Internet].). If the status of a patient was unknown, general practitioners or referring nephrologists were contacted. Subjects’ status regarding survival and GF was recorded until April 2012. Causes of death were available until May 2009. One subject was lost to follow-up. Baseline measurements and definitionsTransplant characteristics, such as subject demographics, previous history, as well as date and type of transplantation, were extracted from the Groningen Renal Transplant Database. Smoking status and CVD history (considered positive if participants previously had a myocardial infarction, transient ischemic attack, or cerebrovascular accident) were obtained using a self-report questionnaire at inclusion. Medical records were consulted for information on use of medication. Standard immunosuppressive regimen changed over the years: from 1968 to 1989, prednisolone and azathioprine (100 mg/day); from January 1989 to February 1993, cyclosporine standard formulation (10 mg/kg; trough levels of 175–200 μg/l in the first 3 months, 150 μg/l between 3 and 12 months after transplantation, and 100 μg/l thereafter) and prednisolone (initially 20 mg/day, rapidly tapered to 10 mg/day); from March 1993 to May 1997, cyclosporine microemulsion (10 mg/kg, trough levels as before) and prednisolone; and from May 1997 onward, mycophenolate mofetil (2 g/day) was added (13Sinkeler S.J. Zelle D.M. Homan van der Heide J.J. Gans R.O.B. Navis G. Bakker S.J.L. Endogenous plasma erythropoietin, cardiovascular mortality and all-cause mortality in renal transplant recipients.).Blood samples were drawn at time of inclusion after an 8–12 h overnight fasting period. No specific antioxidant was added at the time of blood draw; however, all samples contained EDTA, a known antioxidant, and were handled in an identical fashion. All standard laboratory measures, including hemoglobin A1C (HbA1c), were performed at time of inclusion. Concentrations of total cholesterol (TC) and HDL-C were determined using the cholesterol oxidase-phenol aminophenazone method. Concentration of LDL-C was calculated using the Martin Hopkins equation (14Martin S.S. Blaha M.J. Elshazly M.B. Toth P.P. Kwiterovich P.O. Blumenthal R.S. Jones S.A. Comparison of a novel method vs the friedewald equation for estimating low-density lipoprotein cholesterol levels from the standard lipid profile.). apoA-I, apoB, and lipoprotein (a) [LP(a)] were determined by nephelometry with reagents for Dade Behring nephelometer systems (BN II, Siemens, Marburg, Germany). The glycerol-3-phosphate oxidase-phenol aminophenazone method was used to measure plasma triglycerides. Levels of plasma high-sensitive C-reactive protein were assessed by ELISA (8Annema W. Dikkers A. Freark de Boer J. Dullaart R.P.F. Sanders J.-S.F. Bakker S.J.L. Tietge U.J.F. HDL cholesterol efflux predicts graft failure in renal transplant recipients.). Plasma glucose levels were determined using the glucose-oxidase method, and plasma insulin was measured using an AxSYM autoanalyzer. Levels of glycosylated HbA1c were assessed by high-performance liquid chromatography (8Annema W. Dikkers A. Freark de Boer J. Dullaart R.P.F. Sanders J.-S.F. Bakker S.J.L. Tietge U.J.F. HDL cholesterol efflux predicts graft failure in renal transplant recipients.). homeostasis model assessment of insulin resistance was used to determine insulin resistance by multiplying glucose (mmol/l) with insulin (μU/ml) and dividing the result by 22.5 (15Matthews D.R. Hosker J.P. Rudenski A.S. Naylor B.A. Treacher D.F. Turner R.C. Homeostasis model assessment: insulin resistance and β-cell function from fasting plasma glucose and insulin concentrations in man.). Creatinine concentrations in plasma and urine were determined using a modified version of the Jaffé method. The estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration formula (16Levey A.S. Stevens L.A. Schmid C.H. Zhang Y.L. Castro A.F. Feldman H.I. Kusek J.W. Eggers P. Van Lente F. Greene T. Coresh J. A new equation to estimate glomerular filtration rate.); all values included in the analyses were obtained at time point of inclusion. Creatinine clearance was calculated from 24-h urinary creatinine excretion and plasma creatinine. Total urinary protein concentration was analyzed using the Biuret reaction.Proteinuria was defined as urinary protein excretion ≥0.5 g per 24 h. The diagnosis of diabetes mellitus was made if antidiabetic medication was used or fasting plasma glucose concentration was ≥ 7.0 mmol/l or HbA1c was >6.5% (17American Diabetes Association
Diagnosis and classification of diabetes mellitus.). The BMI was calculated by dividing the weight in kilograms by the height in meters squared. Blood pressure was measured automatically in 1 min intervals after a 6 min rest in the supine position (Omron M4; Omron Europe BV, The Netherlands), and the mean of three measurements was taken.The LDL particle number and average size were determined in EDTA plasma (n = 158 participants) by 1H-NMR spectroscopy using a Vantera® NMR Clinical Analyzer as previously described (Labcorp, Raleigh, NC) (18Matyus S.P. Braun P.J. Wolak-Dinsmore J. Jeyarajah E.J. Shalaurova I. Xu Y. Warner S.M. Clement T.S. Connelly M.A. Fischer T.J. NMR measurement of LDL particle number using the Vantera Clinical Analyzer.). Plasma group IIA phospholipase A2 (sPLA2-IIA) was assessed in n = 40 participants with an ELISA (Cayman Chemical, Ann Arbor, MI) according to the manufacturer’s instructions. Conjugated dienes were determined in apoB-containing lipoproteins (n = 40 participants) by absorbance at 233 nm after lipid extraction with dichloromethane/methanol (2:1, v/v) (19Microsomal lipid peroxidation.) using precipitates generated by polyethylene glycol-6000 precipitation of apoB-containing lipoproteins as described (6Ebtehaj S. Gruppen E.G. Bakker S.J.L. Dullaart R.P.F. Tietge U.J.F. HDL (high-density lipoprotein) cholesterol efflux capacity is associated with incident cardiovascular disease in the general population.). Values were normalized for LDL-C levels. Laboratory analysis of lipoprotein binding to proteoglycans

EDTA plasma samples of 589 participants were available for the laboratory analysis of this study. The plasma samples were collected at baseline, placed on ice, centrifuged at 4°C, immediately stored at −80°C, and left unthawed until analysis. The samples were stored from time of inclusion until 2019.

To determine the binding susceptibility of lipoproteins to proteoglycans, human aortic proteoglycans were isolated from the intima-media of atherosclerotic human aortas (20Bancells C. Benítez S. Jauhiainen M. Ordóñez-Llanos J. Kovanen P.T. Villegas S. Sánchez-Quesada J.L. Oörni K. High binding affinity of electronegative LDL to human aortic proteoglycans depends on its aggregation level.), and the glycosaminoglycan content of proteoglycans was quantified as overall marker of proteoglycans (21Ahmed O. Littmann K. Gustafsson U. Pramfalk C. Öörni K. Larsson L. Minniti M.E. Sahlin S. Camejo G. Parini P. Eriksson M. Ezetimibe in combination with simvastatin reduces remnant cholesterol without affecting biliary lipid concentrations in gallstone patients.). Then, wells of polystyrene 96-well plates (Thermo Fisher Scientific) were coated with 100 μl of proteoglycans (50 μg/ml in PBS) by incubation at 4°C overnight. Wells were blocked with 1% bovine serum albumin in PBS for 1 h at 37°C. Wells without proteoglycan coating served as controls for unspecific binding. To measure lipoprotein binding to the immobilized proteoglycans, 1 μl of plasma (derived from RTRs at baseline) was added to the wells in a buffer containing 140 mmol/l NaCl, 2 mmol/l MgCl2, 5 mmol/l CaCl2, and 10 mmol/l MES, pH 5.5, and incubated for 1 h at 37°C. The wells were washed with 10 mmol/l MES-50 mmol/l NaCl, pH 5.5, and the amount of bound TC was determined using the Amplex Red cholesterol kit (Molecular Probes). Each sample was analyzed in duplicate and the nonspecific binding in a single well had been blocked by the blocking buffer. The nonspecific binding consistently accounts for about 5% of the binding to the PG-coated wells. The assay was performed over a duration of several weeks. The day-to-day variation of the measurement is 21Ahmed O. Littmann K. Gustafsson U. Pramfalk C. Öörni K. Larsson L. Minniti M.E. Sahlin S. Camejo G. Parini P. Eriksson M. Ezetimibe in combination with simvastatin reduces remnant cholesterol without affecting biliary lipid concentrations in gallstone patients., 22ApoB-100 lipoprotein complex formation with intima proteoglycans as a cause of atherosclerosis and its possible ex vivo evaluation as a disease biomarker.). Statistical analysis

Baseline characteristics of the study population were analyzed for gender-stratified tertiles of levels of bound TC/plasma LDL-C (low, medium, and high). Normally distributed continuous variables are depicted as the mean ± standard deviation, whereas continuous variables with a skewed distribution are given as the median [25th–75th percentile]. Categorical variables are summarized by absolute numbers (percentages).

Baseline characteristics were tested for differences among groups with low, medium, and high LPBS based on sex-stratified tertiles. Baseline characteristics for normally distributed continuous variables were tested for differences among groups with one-way ANOVA. The Kruskal-Wallis test was used to assess differences between groups for continuous variables with a skewed distribution. Group differences in categorical data were tested with Pearson chi-squared test.

LPBS of subjects who reached the respective end points was compared with values of subjects not reaching the end points of the study by independent samples t test. Similarly, the LPBS for males and females was computed using independent samples t test. Subsequently, all characteristics with a P < 0.10 across gender-stratified tertiles of LPBS were entered into a step-wise multivariable linear regression model with backward elimination (P < 0.05) to identify variables independently associated with LPBS.

Multivariable Cox regression was used to calculate hazard ratios (HRs) and 95% CI for the primary end points. Adjustment of potential confounders was used to assess the independent association of LPBS with the end points chronic GF and cardiovascular mortality. Potential confounders were determined as known risk factors of chronic GF and CVD in RTRs and included age, sex, eGFR, periods of acute rejection, number of human leukocyte antigen mismatches, primary renal disease, diabetes mellitus, BMI, dialysis time, type of transplantation, use of calcineurin inhibitors, use of proliferation inhibitors, use of statins, time between transplantation and baseline, and donor age. Validity of proportional hazard assumptions was tested using Schoenfeld residuals. Furthermore, subgroup analysis using interaction tests were performed in which HRs were determined across categories of baseline characteristics. For continuous variables, the median value was used as cutoff. For the end point chronic GF, the subject characteristics were sex (male vs. female), age (<52.1 vs. >52.1 years), use of statins (yes vs. no), eGFR (<46.7 and ≥46.7 ml/min per 1.73 m2), and period of acute rejection (yes vs. no).

To compare the relevance of the proposed novel functionality parameter with a traditional quantitative parameter, statistical analyses were repeated for plasma concentration of LDL-C at baseline and results were compared with those of LPBS.

Two-sided P-values <0.05 were considered to indicate statistical significance. All statistical analyses and visualization of data were conducted using STATA® Statistical Software, Release 15.1 (StataCorp, College Station, TX).

ResultsIn this longitudinal follow-up study, the LPBS was measured in 589 RTRs. Individual LPBS values were expressed as the ratio between proteoglycan-bound cholesterol and plasma LDL-C levels (nmol/mmol). Baseline characteristics according to sex-stratified tertiles of LPBS are summarized in Table 1. The concentrations of TC, LDL-C, apoA-I (each PP = 0.004), apoB (P = 0.012), and the LDL-C/apoB ratio (PP = 0.04) and BMI (P = 0.015) showed a significant inverse association with binding susceptibility, but for other potential cardiovascular risk factors including age, history of cardiovascular events, diabetes mellitus, tobacco abuse, plasma triglycerides, and HDL-C, no relationship with LPBS was evident.

Table 1Baseline characteristics according to sex-stratified tertiles of lipoprotein-proteoglycan binding susceptibility

CVA, cerebrovascular accident; eGFR, estimated glomerular filtration rate; HbA1c, hemoglobin A1c; HOMA-IR, Homeostatic Model Assessment for Insulin Resistance; HLA, human leukocyte antigen; Lp, lipoprotein; PG, proteoglycan; TC, total cholesterol; TIA, transient ischemic attack; Tx, transplantation

Normally distributed continuous variables are depicted as the mean ± standard deviation, continuous variables with a skewed distribution are given as the median [25th–75th percentile], and categorical variables are summarized by absolute numbers (percentages). Differences between tertiles of lipoprotein-proteoglycan binding susceptibility were tested using one-way ANOVA for normally distributed continuous variables, Kruskal-Wallis test for continuous variables with a skewed distribution, and Pearson's chi-squared test for categorical variables.

Subsequently, backward multiple linear regression analysis was used to assess which variables are determinants of LPBS in RTRs (Table 2). The concentration of TC (standardized β = −0.24, PPR2 was 0.10.

Table 2Predictors of lipoprotein binding susceptibility

All variables with P < 0.1 between tertiles were entered into a stepwise linear regression with backward elimination.

To further explore factors associated with LPBS, we first correlated the LDL-C/apoB ratio as an, allowedly, relatively crude but easy to calculate measure of the LDL size with LPBS. Previously, a smaller size of LDL particles had been identified as a determinant of increased binding to proteoglycans (23Sartipy P. Camejo G. Svensson L. Hurt-Camejo E. Phospholipase A(2) modification of low density lipoproteins forms small high density particles with increased affinity for proteoglycans and glycosaminoglycans.). Surprisingly, an overall significant positive correlation was observed in the RTR (r = 0.159, PP = 0.01), whereas medium- (r = −0.07, not significant) or large-sized LDL particles (r = 0.03, P = not significant) did not correlate with LPBS. Because oxidative modification had been shown to decrease binding of apoB-containing lipoproteins to proteoglycans (24Oörni K. Pentikäinen M.O. Annila A. Kovanen P.T. Oxidation of low density lipoprotein particles decreases their ability to bind to human aortic proteoglycans. Dependence on oxidative modification of the lysine residues.), we measured in a subset of our cohort (n = 40) the conjugated diene content in these particles and correlated the result with LPBS. Consistent with previous reports, also in RTRs, a significant negative correlation between conjugated dienes and LPBS was observed (r = −0.46, P = 0.003). In addition, enzymatic modification by sPLA2-IIA was reported to increase binding of LDL to proteoglycans (23Sartipy P. Camejo G. Svensson L. Hurt-Camejo E. Phospholipase A(2) modification of low density lipoproteins forms small high density particles with increased affinity for proteoglycans and glycosaminoglycans.). Also, this result could be replicated in our cohort because plasma levels of sPLA2-IIA correlated positively with LPBS (r = 0.42, P = 0.016, n = 40). Because it has been previously demonstrated that Lp(a) is increased among patients with a low eGFR, (25Hopewell J.C. Haynes R. Baigent C. The role of lipoprotein (a) in chronic kidney disease: thematic review series: lipoprotein (a): coming of age at last.) which in turn leads to a higher binding affinity to proteoglycans (26Lundstam U. Hurt-Camejo E. Olsson G. Sartipy P. Camejo G. Wiklund O. Proteoglycans contribution to association of Lp(a) and LDL with smooth muscle cell extracellular matrix.), we measured Lp(a) levels in our study population. Lp(a) was not correlated with LPBS (r = −0.06, P = 0.13). Furthermore, an adjusted Cox regression with Lp(a) as independent variable showed that there is no significant association between Lp(a) levels and GF (HR = 1.11, P = 0.25).

Data concerning specific causes of death were available for a median follow-up of 7 years. Of 130 (22%) patients who died in this period, 68 did so because of confirmed cardiovascular causes (12% of the total study population, 52% of the deceased patients). Furthermore, 29 (5% of the total study population, 22% of the deceased patients) patients died from malignancy, 23 (4% of the total study population, 18% of the deceased patients) from an infectious death, and 10 (2% of the total study population, 7% of the deceased patients) from other causes. During the median follow-up of 9.5 years for GF, a total of 73 (13%) subjects experienced this end point.

At baseline, LDL-C as well as LPBS were comparable between survivors and deceased RTRs, with respect to cardiovascular mortality (LDL-C: 3.90 ± 1.0 vs. 3.88 ± 1.0 mmol/l, P = 0.87; LPBS: 1.34 ± 0.42 vs. 1.29 ± 0.5 nmol/mmol, P = 0.33). While baseline LDL-C levels were also similar in patients developing GF or not (4.03 ± 1.38 vs. 3.88 ± 0.91 mmol/l, P = 0.22), LPBS was significantly higher in patients who subsequently developed GF than in those with a surviving graft (1.47 ± 0.63 vs. 1.32 ± 0.39 nmol/mmol, P = 0.003).

Cox regression analysis revealed that neither LPBS nor the classical biomarker LDL-C was associated with CVD mortality (Table 3); this conclusion remained valid after adjustment for a number of potential confounding factors in different statistical models. However, Cox regression analysis showed a prospective association between LPBS and chronic GF (HR, 1.87; 95% CI, 1.24–2.84; P = 0.003, Table 4, model 1). Adjusting for age and sex did not considerably reduce this association (HR, 1.84; 95% CI, 1.21–2.81; P = 0.004, Table 4, model 2).

Table 3Comparison between the association of either LDL function (LPBS) or mass levels of LDL-C with cardiovascular mortality

HR, hazard ratio; LPBS, lipoprotein-proteoglycan binding susceptibility.

Model 1: crude; model 2: adjusted for age and sex; model 3: model 2+ adjustment diabetes mellitus; model 4: model 2+ adjustment for BMI; model 5: model 2+ adjustment for dialysis time and time between transplantation and inclusion; model 6: model 2+ adjustment for type of transplantation and donor age; model 7: model 2+ adjustment for use of calcineurin inhibitors and proliferation inhibitors; model 8: model 2+ adjustment for use of statins.

Table 4Comparison between the association of either LDL function (LPBS) or mass levels of LDL-C with chronic graft failure

HR, hazard ratio; LPBS, lipoprotein-proteoglycan binding susceptibility.

Model 1: crude; model 2: adjusted for age and sex; model 3: model 2 + adjustment for use of statins; model 4: model 2 + adjustment for estimated glomerular filtration rate; model 5: model 2 + adjustment for period of acute rejection; model 6: model 2 + adjustment for number of human leukocyte antigen mismatches, primary renal disease and period of acute rejection; model 7: model 2+ adjustment for dialysis time and time between transplantation and baseline; model 8: model 2+ adjustment for type of transplantation and donor age.

Comparably, adjustment for a number of other potentially impacting factors, namely the use of statins, periods of acute rejection, number of HLA mismatches, primary renal disease, dialysis time, time between transplantation and inclusion, type of transplantation, and donor age did not appreciatively change the significance of the prospective association. However, after additional adjustments for eGFR, significance was lost (HR, 1.25, 95% CI, 0.85–1.82; P = 0.25, Table 4, model 7). We attempted to further delineate the relationship of LPBS and eGFR. Pearson’s correlation coefficients showed that there is no significant correlation between eGFR and LPBS (r = −0.03, P = 0.52). Then, an interaction term was computed and the association with GF was assessed (HR = 0.99, P = 0.40). This showed that there is no significant interaction between eGFR and LPBS.In contrast to these findings with respect to the functional read-out of LPBS, LDL-C levels were not associated with GF, neither in univariate nor in all computed multivariable Cox regression models (Table 4).Cox regression analyses were repeated with crude proteoglycan binding. The results were not substantially different with regard to the normalized marker of LPBS (supplemental Tables S1 and S2).As shown in Fig. 1, the association of LPBS with chronic GF was not different for males versus females (P for interaction = 0.12), subjects with high versus low age (P for interaction = 0.38), use of statins versus no use of statins (P for interaction = 0.67), high versus low eGFR (P for interaction = 0.17), or period of acute rejection versus no period of acute rejection (P for interaction = 0.45). However, for the association of LPBS with CVD mortality, there was an interaction with the use of calcineurin inhibitors versus no use of calcineurin inhibitors (P for interaction = 0.03), indicating that the relationship of LPBS with CVD risk is stronger in subjects that use calcineurin inhibitors.

Fig. 1Hazard ratios for the association of LDL function (LPBS) with incident graft failure, by several participant level characteristics. eGFR, estimated glomerular filtration rate; LPBS, lipoprotein-proteoglycan binding susceptibility.

DiscussionThe results of this study demonstrate that beyond the static measurement of circulating LDL-C levels, functional metrics determining lipoprotein retention to proteoglycans such as LPBS can provide useful clinical information. Specifically, we show that in RTRs, LPBS was prospectively associated with chronic GF, a manifestation of de novo atherosclerosis, but not with CVD death, largely a consequence of preexisting, complex atherosclerotic lesions. We interpret these findings as being consistent with the response-to-retention hypothesis of atherosclerotic lesion formation (10Tabas I. Williams K.J. Borén J. Subendothelial lipoprotein retention as the initiating process in atherosclerosis: update and therapeutic implications.