HDL in CKD—The Devil Is in the Detail

Abstract

The picture of HDL cholesterol (HDL-C) as the “good” cholesterol has eroded. This is even more surprising because there exists strong evidence that HDL-C is associated with cardiovascular disease (CVD) in the general population as well as in patients with impairment of kidney function and/or progression of CKD. However, drugs that dramatically increase HDL-C have mostly failed to decrease CVD events. Furthermore, genetic studies took the same line, as genetic variants that have a pronounced influence on HDL-C concentrations did not show an association with cardiovascular risk. For many, this was not surprising, given that an HDL particle is highly complex and carries >80 proteins and several hundred lipid species. Simply measuring cholesterol might not reflect the variety of biologic effects of heterogeneous HDL particles. Therefore, functional studies and the involvement of HDL components in the reverse cholesterol transport, including the cholesterol efflux capacity, have become a further focus of study during recent years. As also observed for other aspects, CKD populations behave differently compared with non-CKD populations. Although clear disturbances have been observed for the “functionality” of HDL particles in patients with CKD, this did not necessarily translate into clear-cut associations with outcomes.

This review focuses on HDL in the context of CKD. It considers the epidemiologic as well as the genetic evidence and discusses the heterogeneity of HDL particles, as well as their components and their involvement in cholesterol efflux and reverse cholesterol transport. It points to the discrepancies between different findings in non-CKD populations, which can often not be confirmed in CKD populations, a phenomenon also well known from interventional studies in fields of study beyond lipid metabolism.

Normally, HDL particles have potent anti-inflammatory, antioxidative, and antithrombotic properties caused by several components carrying these properties. These are enzymes, apolipoproteins, complements, and other components. As reviewed extensively in the past,¹ certain conditions such as CKD with systemic oxidative stress and inflammation substantially reduce these capabilities of HDL particles and can transform them into the opposite pro-oxidant and proinflammatory direction.² This has consequences for the interrelated pathways and, for example, the ability of HDL to prevent oxidation of LDL,³ reduced monocyte chemotactic activity,⁴ or disturbances in the reverse cholesterol transport.

Furthermore, it is emphasized that some of the inconsistencies between studies in CKD populations in the literature might be caused by the heterogeneity of populations. HDL-C concentrations, composition of HDL particles, disturbances in the functionality, and especially the reverse cholesterol transport might be different between various stages of kidney impairment, especially between patients with and without nephrotic syndrome, as well as between patients with ESRD treated with hemodialysis, peritoneal dialysis, or kidney transplantation.

A Brief Introduction to Lipoprotein Pathways

Lipids such as cholesterol and triglycerides are not soluble in water and are therefore transported in blood and other body fluids together with proteins assembled to lipoproteins. The number of these proteins is quite large and >80 in the case of HDL particles.⁵ They contribute not only to the structure of lipoproteins but also to the metabolic control by activation or inhibition of enzymes, and interaction with lipoprotein receptors.

Figure 1 illustrates the two main lipoprotein pathways. The exogenous pathway transports dietary lipids from the intestine as triglyceride-rich chylomicrons. Chylomicrons are catabolized by lipoprotein lipase to chylomicron remnants, which are finally taken up by the liver. This process generates free fatty acids, which are absorbed mainly by liver, muscle, and adipose tissue for either energy production or storage. Other organs such as the kidney can also take them up. A cellular overload with free fatty acids results in an accumulation of toxic derivatives, promoting cellular dysfunction and damage, a metabolic disorder called lipotoxicity.⁶

Figure 1.

Schematic illustration of the major normal lipoprotein metabolic pathways. Blue arrows refer to points of action of the respective enzymes in blue. ABCA1, ATP-binding cassette transporter A1; ABCG1, ATP-binding cassette transporter G1; ApoA-I, apolipoprotein A-I; ApoA-IV, apolipoprotein A-IV; CE, cholesterol ester; CETP, cholesteryl ester transfer protein; FFA, free fatty acid; HTGL, hepatic triglyceride lipase; IDL, intermediate-density lipoprotein; LCAT, lecithin–cholesterol acyltransferase; LDL-R, LDL receptor; LPL, lipoprotein lipase; LRP, LDL-R–related protein; SR-B1, scavenger receptor B1; TG, triglyceride; VLDL, very low-density lipoprotein; VSMC, vascular smooth muscle cells. Reprinted and adapted from reference 7, with permission.

The assembly and secretion of triglyceride-rich VLDL particles characterize the endogenous pathway. They transport triglycerides from the liver to peripheral tissues and are hydrolyzed by the lipoprotein lipase to intermediate-density lipoproteins, which either can be taken up by the liver or can be further hydrolyzed to LDL particles. This conversion results in a depletion of triglycerides and a relative increase in cholesterol content. LDL transports cholesterol primarily to hepatocytes but also to peripheral tissues. During circulation, LDL can become chemically modified, oxidized, and taken up by macrophages and vascular smooth muscle cells. Macrophages overloaded with cholesteryl esters can be transformed into foam cells, a crucial step in atherosclerosis development.

HDL is a key player in the reverse cholesterol transport, shuttling cholesterol from peripheral cells to the liver and to steroidogenic organs for steroid hormone production. This important step relieves peripheral cells and especially macrophages from cholesterol burden. HDL particles are characterized by a very high protein content (roughly 45%), cholesterol (25%), phospholipids (25%), and triglycerides (5%). However, this percentage varies depending on the HDL particle size. HDL precursor particles are secreted by the liver and intestine and can absorb free cholesterol from cell membranes. This process is mediated by ATP-binding cassette (ABC) transporter A1, ABCG1, scavenger receptor class B type 1 (SR-B1), apoA-I, apoA-IV, and others. ApoA-I as the major apolipoprotein of HDL activates the enzyme lecithin–cholesterol acyltransferase, which esterifies the accepted free cholesterol. HDL₃ particles are transformed into larger HDL₂ particles by acquisition of additional apolipoproteins, cholesteryl esters, and triglycerides.⁷ Large HDL particles bind to SR-B1 in liver and steroidogenic tissues. It is believed that the HDL particle is not internalized and degraded as a whole particle, but docks to this receptor and releases the cholesteryl ester cargo to the cell.^8,9 Furthermore, cholesteryl esters can take an alternative route via transfer from HDL to triglyceride-rich lipoproteins such as chylomicron remnants or VLDL mediated by the cholesteryl ester transfer protein (CETP) in exchange for triglycerides.¹⁰

HDL-C and CVD in the General Population

A decade ago, everything was so much easier. There was bad cholesterol and good cholesterol, and even the general population was educated on this topic. The epidemiologic evidence was quite strong, with a meta-analysis of 68 prospective studies including >300,000 participants free of CVD at baseline and a follow-up time of almost 3 million person-years. This study described an adjusted 22% decrease in risk of coronary heart disease for each increase in HDL-C of 15 mg/dl.¹¹

This observation was a clear mandate for the pharmaceutical industry to develop drugs that increase HDL-C. The first and very obvious target was the inhibition of CETP, which plays a major role in the reverse cholesterol transport, relieving the peripheral cells from a cholesterol overload.¹² A handful of drugs with that target have been developed that have increased HDL-C by up to >100%. Phase 3 trials prematurely stopped three of the drugs because of either off-target effects with an increase in death rates, or lack of efficacy.¹³ Very recently, the results on anacetrapib have been reported as one of the last substances in a running phase 3 trial including 30,449 adults with atherosclerotic vascular disease who were already receiving an effective LDL cholesterol–lowering regimen at baseline. This trial has met its primary end point by significantly reducing major coronary events by 9% compared with placebo (hazard ratio, 0.91; 95% confidence interval, 0.85 to 0.97; P=0.004).¹⁴ Interestingly, this drug did not only increase HDL-C by 104% compared with placebo, but also decreased non–HDL-C values by 18%. From other interventional studies, this decrease in non–HDL-C alone has been expected to result in a 10% relative reduction in risk of coronary death and myocardial infarction. This was very similar to the observed changes.¹⁴ In addition, lipoprotein(a) [Lp(a)] was lowered by 25%. Furthermore, the increase in HDL-C by 46 mg/dl did not meet the expectations from associated effect sizes from observational studies by far. It is therefore likely that the observed effects on major coronary events might be caused by the effect on apoB-containing lipoproteins such as LDL cholesterol and Lp(a), rather than by effects on HDL-C.¹⁵

The enthusiasm toward the idea to increase HDL-C was dampened further by genetic studies that did not find an association for genetic variants with CVD, although these variants showed a pronounced association with HDL-C concentrations.^16–18 For example, observational epidemiologic studies revealed that an increase of HDL-C by 1 SD is associated with a 38% lower risk for myocardial infarction. However, no increase in risk was observed when the same increase in HDL-C concentrations was caused by a group of genetic variants that are exclusively associated with these higher HDL-C concentrations.¹⁹ Even the opposite can be the case: a loss-of-function variant in the SR-B1 gene impairs posttranslational processing of SR-B1 and abrogates selective HDL-C uptake. Heterozygous carriers of this variant have about 70% higher HDL-C concentrations and, interestingly, 2.8-fold higher HDL_2b concentrations. Contrary to expectations, carriers of the mutation have a 79% higher risk of CHD.²⁰ This might be interpreted as a severe disturbance in HDL particle uptake and the measured HDL-C concentration might not reflect the functionality of the HDL particles.

HDL-C and CVD in CKD

The evidence that HDL-C concentrations are associated with prevalence and incidence of CVD events as well as mortality is quite contrasting.^21–28 A recent study in a cohort of >33,000 patients on hemodialysis revealed a U-shaped association with an increased risk for total and CVD mortality in patients with HDL-C concentrations <30 mg/dl and >60 mg/dl.²⁹ Many studies searched for alternative tools to predict CVD in these high-risk populations,² such as various lipoprotein or apolipoprotein ratios.^30,31 With more sophisticated techniques, many other HDL subfractions (up to ten) were identified and quantitated,^32,33 again with more or less success concerning prediction of outcomes. We currently have an insufficient understanding of the physiologic and pathophysiologic role of these many subfractions. Therefore, the functionality of HDL particles came into the spotlight because of the increasing evidence that the proteome and lipidome of HDL particles is heavily disturbed not only in the uremic state, but also in very early stages of kidney impairment.³⁴ This functionality is thought to be heavily related to cholesterol efflux with the related enzymes and apolipoproteins, or the modulation of inflammatory and immune cells.^1,35 This could not only have an influence on CVD outcomes but also on CKD and CKD progression.

Epidemiologic Observations on HDL-C Concentrations in CKD

Several but not all epidemiologic studies pointed to an association between low HDL-C or a high triglyceride to HDL-C ratio and poor kidney function or progression of CKD.^31,36–48 The very recent and largest study by Bowe et al.⁴⁹ investigated the association between HDL-C levels and various CKD end points in a cohort of almost 2 million male veterans from the United States, with a median follow-up of 9 years. Individuals with HDL-C concentrations <30 mg/dl had a 10%–20% higher risk for CKD and/or progression of CKD compared with individuals with concentration ≥40 mg/dl, depending on the definition of the end point and how intensely they adjusted their results for confounders. The risk was still 8% higher for the most severe end point, defined by a composite outcome of ESRD, dialysis, or transplantation.⁴⁹

Although these results were quite strong, it must not conceal the fact that epidemiologic studies cannot disentangle with certainty cause or consequence of an association of HDL-C with CKD outcomes. Even in prospective studies, HDL-C could already be influenced by very early and hard to detect stages of CKD long before CKD or progression of CKD became evident. As discussed recently,⁵⁰ one possibility to see whether HDL-C influences kidney function would be to meta-analyze the data from trials that targeted HDL-C concentrations, and to investigate whether the intervention had a positive effect on eGFR and/or albumin-to-creatinine ratio during the observation period. This has been done in the most recent Randomized Evaluation of the Effects of Anacetrapib through Lipid Modification (REVEAL) trial and the results pointed in the opposite direction: the authors described a higher frequency of a decrease of eGFR <60 ml/min per 1.73 m² in the anacetrapib group during the observation period compared with placebo (11.5% versus 10.6%; P=0.04), and a tendency toward a higher frequency of a 40% decline in eGFR (2.9% versus 2.5%; P=0.07).¹⁴ A meta-analysis of all available studies might be necessary to obtain a sufficient statistical and reliable power because the effect size of these drugs on kidney function might be low, and because the trials were primarily not designed to study the effect of the drugs on kidney function. It is interesting to note that in the Study of Heart and Renal Protection (SHARP) trial, the intervention with statins did not only show no effect on progression of kidney disease, but the baseline HDL-C concentration also did not modify this negative finding.⁵¹

Genetic Observations on HDL-C and CKD

During the last decade, a certain type of genetic studies, so-called Mendelian randomization studies, became quite popular to collect supportive evidence of a causal association between a biomarker and an outcome of interest.^50,52 This principle relies on an instrumental variable in form of a genetic variant (e.g., single nucleotide polymorphisms [SNPs], copy number variants, etc.) that has a significant influence on a biomarker. If the biomarker is causally related to the outcome of interest, one would expect that the genetic variant is also associated with the outcome. In the case of low HDL-C concentrations and impaired kidney function, we would expect that a genetic variant that is associated with low HDL-C concentrations would also have an effect on eGFR if the association between HDL-C and kidney function was indeed causal.⁵⁰

The name Mendelian randomization stems from the random assortment of alleles from parents to offspring during meiosis. If one of the parents carries an allele from a certain gene that causes lower HDL-C levels and one allele that causes normal HDL-C values, it is randomly determined which of the two alleles is transmitted to the offspring. This also holds true for the alleles from the other parent (Figure 2). It becomes even more complex because the effect of a single allele of a certain gene on HDL-C concentrations is usually quite low, with effect sizes < 1 mg/dl per risk allele. However, many genes are known to additively contribute to HDL-C concentrations and if the random assortment of alleles results in a preponderance of HDL-C–lowering alleles, an offspring might receive a large number of HDL-C–lowering alleles, to which they are exposed for their whole lives (Figure 2). It will thereby be determined in this early stage whether someone is exposed all his life to lower or higher HDL-C concentrations. A large number of HDL-C–lowering alleles would also mean a larger risk for CVD or an impairment of kidney function if HDL-C causally influences CVD and/or kidney function. On a population level, often a large number of investigated individuals is mandatory, depending on the effect sizes of the genetic variants on, for example, the HDL-C concentrations as the biomarker of interest and depending on the biologic effects of this marker on the outcome (Figure 2). Core assumptions and threats that could interfere with Mendelian randomizations were recently discussed in detail by Sekula et al.⁵²

Figure 2.

Schematic illustration of a Mendelian randomization approach of HDL-C influencing genetic variants and outcomes of interest. Let us assume there exists a large number of genes that influence HDL-C concentrations (e.g., genes 1–89). For each gene, we know genetic variants that decrease and others that increase HDL-C, and others that are neutral. For each gene, it is determined randomly at the time of conception which of the two alleles from the father (F) and which of the two alleles from the mother (M) are transmitted to the offspring. Because many genes influence HDL-C concentrations, it is finally a question of whether an offspring has a preponderance of HDL-C–lowering alleles to which they are is exposed for their lifetime. If there is a causal effect of HDL-C concentrations on outcomes such as CVD or kidney function, one would expect that there will be also a preponderance of the HDL-C concentration-decreasing alleles in the patient groups with the outcome of interest compared with groups without that outcome. Usually, a large number of patients and controls have to be investigated to get reliable results.

Following this idea, a recent study used 68 genetic SNP variants that were found to be associated with HDL-C in a genome-wide association study (GWAS) with >188,000 individuals, and tested their association with eGFR using the results from another GWAS on kidney function including >133,000 individuals (Figure 3).⁵³ A larger number of HDL-C SNPs found was larger than could be accounted for by chance to be associated with eGFR (21% versus 5%). Most interestingly and in contrast to expectations, the genetic variants that showed the strongest associations with HDL-C levels (e.g., variants in the genes of CETP, LIPC, LIPG, or LPL) did not show any association with kidney function and vice versa. An extensive evaluation of pleiotropy indicated that the effects of some of the HDL-associated SNPs on eGFR cannot be mediated by HDL-C alone. That means for these SNPs, if there is indeed an association of HDL-C–associated SNPs with eGFR, then there must be additional mechanisms explaining this relationship than simply the cholesterol content of the HDL particle.⁵³ A further publication using the same data sets but a slightly different approach revealed very similar HDL-C SNP effects on kidney function which, however, were finally significant: a difference of 1 SD of HDL-C (approximately 17 mg/dl) was associated with an increase of eGFR of only 0.8% (P=0.004).⁵⁴ Even if the effect was significant, a causal effect of such a small magnitude would not explain the observed epidemiologic association of HDL-C on kidney function.⁴⁹

Figure 3.

Schematic overview on the study design, main findings, and interpretation of the results derived from two genome-wide association studies on HDL-C and eGFR. For explanation, see Genetic Observations on HDL-C and CKD. Reprinted from reference 53, with permission.

This observation is in line with the idea that we actually do not know the guilty party in the HDL particle that is responsible for the outcome. That means that HDL-C represents a surrogate marker that does not reflect the responsible culprit or the real HDL functionality. Particularly, the functionality of HDL particles has been a major focus of investigations over the last two decades, with inspiring but also contrasting results.

Cholesterol Efflux Capacity and HDL Functionality

A key component in the development of atherosclerosis is the overloading of macrophages with cholesterol in the arterial wall, resulting in foam cells. The relief of macrophages from cholesterol is thus a crucial atheroprotective step. Therefore, functional in vitro assays have been developed that measure the cholesterol efflux capacity. Figure 4 illustrates the principle of these assays. Mouse macrophage cell lines such as J774 are cultivated and used as donor cells. They are enriched with cholesterol by exposure to radiolabeled or fluorescence-labeled cholesterol and stimulated with cAMP to induce cholesterol transporters. The cholesterol acceptor, usually whole serum or plasma, apoB-depleted serum, or isolated HDL from the patient under investigation, is then added to the medium. ApoB-depleted serum, obtained by precipitation, is often preferred because this fast and simple procedure removes LDL, intermediate-density lipoproteins, and VLDL as well as Lp(a) from the serum in a far more gentle manner compared with ultracentrifugation, which might detach some of the many functionally important proteins from the HDL particle. The apoB-depleted serum abolishes the transfer of labeled cholesterol from cells to these apoB-containing lipoproteins, and exclusively analyzes the cholesterol transport capacity to HDL particles. After incubation, the efflux of labeled cholesterol from the cells to the medium is quantified and reflects the cholesterol transport mediated by ABCA1, ABCG1, SR-B1, and aqueous diffusion. The principle of the assay can be modified in different ways and thereby influences which of the transporters is more or less in the focus of the investigation. Strongly standardized criteria have to be applied to ensure a reliable measurement of the efflux capacity. It is not clear whether differences in the assay design might have caused some of the discrepant results in the literature (reviewed by Anastasius et al.⁵⁵).

Figure 4.

Principle of cholesterol efflux capacity assays. Macrophage cell lines are enriched with labeled cholesterol and stimulated with cAMP to induce cholesterol transporters. Then a cholesterol acceptor, such as apoB-depleted serum of the patient under investigation, is added to the medium. After incubation the efflux of labeled cholesterol from the cells to the medium is quantified and reflects the cholesterol transport mediated by ABCA1, ABCG1, SR-B1, and aqueous diffusion.

The situation in patients with CKD points to an impaired cholesterol efflux. Holzer et al.³⁴ studied the proteome and lipid composition of HDL from patients on hemodialysis. They found pronounced changes in HDL with increased amounts of serum amyloid A1, albumin, lipoprotein-associated phospholipase A2, apoC-III, triglycerides, and lysophospholipids, as well as decreased amounts of phospholipids. These changes in uremic HDL were associated with a decreased cholesterol efflux from macrophages³⁴ and a shift toward a proinflammatory phenotype. The anti-inflammatory and antiapoptotic functions were more severely suppressed in patients receiving hemodialysis rather than peritoneal dialysis.⁵⁶ Yamamoto et al. made similar observations in patients on hemodialysis: besides a profound efflux impairment, they described a reduced chemotactic ability and an increased macrophage cytokine response with increased concentrations of TNF-α, IL-6, and IL-1-β. When patients on hemodialysis received statins, they presented with a reduced inflammatory response but impaired cholesterol acceptor function did not improve.⁵⁷ In further studies, monocyte subsets were analyzed as predictors of CVD events in patients with CKD stages 2–4: higher CD14⁺⁺CD16⁺ monocyte counts independently predicted CVD events. These monocytes showed a preferential lipid accumulation and high CD36, CD68, and low ABCA1 expression, and finally, a low cholesterol efflux.⁵⁸ These findings are in line with a reduced ABCA1 mRNA in monocytes from patients with CKD of various etiologies with CKD stages 4–5 but not on dialysis when compared with control monocytes.⁵⁹

Many studies from the general population and in patients with CVD showed a clear association between decreased cholesterol efflux capacity and cardiovascular outcomes.^60–63 This association was usually independent of HDL-C concentrations.

Studies investigating the cholesterol efflux capacity in patients with CKD did not find evidence for an association with CVD. The CARE FOR HOMe study included 526 patients with CKD stages 2–4 at baseline. The cholesterol efflux capacity was not significantly different between patients with and without prevalent CVD at baseline. It also did not predict CVD events that occurred during the prospective observation period of an average of 4.6 years.⁶⁴ A post hoc analysis of the German Diabetes Dialysis Study (4D-study) on 1147 patients with diabetes who were receiving hemodialysis, with a median follow-up of 4.1 years, also did not find an association between efflux capacity and a combined end point (including a composite of cardiac death, nonfatal myocardial infarction. and stroke), with cardiac events and all-cause mortality.⁶⁵ A further study investigated a cohort of 495 kidney transplant recipients with a median follow-up of 7 years. Baseline cholesterol efflux capacity did not predict cardiovascular and total mortality. Interestingly, a high efflux capacity at baseline was associated with a lower risk for graft failure, independent of apoA-I, HDL-C, and creatinine clearance.⁶⁶

The reason of the discrepant results between non-CKD and CKD populations is not clear. One explanation might be that the CVD in patients with CKD is not as much of atherosclerotic nature with atherothrombotic events such as myocardial infarctions, but is probably more connected to arteriosclerotic changes of the smaller vessels with thickening of the medial arterial layer resulting in cardiac fibrosis and hypertrophy, heart failure, and arrhythmia.⁶⁷ Further, it is not clear why cholesterol efflux capacity was not associated with CVD in transplanted patients but with graft survival in the same cohort. It might reflect different vascular and inflammatory processes in the transplant vasculopathy with site-specific endothelial processes. It could also be that the cholesterol efflux from kidney cells such as mesangial cells or podocytes is of major importance for the homeostasis in the kidney allografts. Overall, we have to keep in mind that currently, only a few studies are available that were conducted on different stages of CKD. Clearly, larger and more extended studies with standardized experimental protocols are required.

HDL-Related Components

HDL particles are highly complex and carry >80 proteins and some hundred lipid species,⁵ as well as a dozen micro-RNAs.⁶⁸ Therefore, recent studies (reviewed by Annema and von Eckardstein^5,69) described dysfunctional HDL particles⁷⁰ that lose their atheroprotective effect, probably because of compositional modifications.⁶⁹ Table 1 gives an overview on the proteome of HDL particles but is far from being complete, and more details can be found at the HDL Proteome Watch (http://homepages.uc.edu/∼davidswm/HDLproteome.html). Only a few of them have been investigated in the context of CKD and especially of progression of CKD.

Table 1.

List of major components of the HDL proteome

Proteins	Function
Apolipoproteins
A-I	Major apolipoprotein of HDL; activates LCAT
A-II	Enhances hepatic triglyceride lipase activity
A-IV	Participates in reverse cholesterol transport; probably involved in the activation of LCAT and LPL
C-I	Activates LCAT; modulates CETP activity; inhibits hepatic lipase
C-II	Activator of LPL
C-III	Inhibitor of LPL and hepatic lipase
C-IV	Involved in triglyceride metabolism
D	Associated with LCAT
E	Ligand for several members of the LDL receptor gene family (especially LDL-R and LRP)
F	Also known as lipid transfer inhibitor protein (LTIP); inhibits CETP
H	Also known as β-2-glycoprotein 1; regulates platelet aggregation
J	Also called clusterin; binds hydrophobic molecules, interacts with cell receptors
L-1	Trypanolytic factor of human serum
M	Binding of small hydrophobic molecules
Enzymes
LCAT	Esterification of cholesterol to cholesteryl esters
PON1	Paraoxonase 1: hydrolysis of homocysteine thiolactone; hydrolyses a wide variety of substrates
PAF-AH (LpPLA2)	Lipoprotein-associated phospholipase A2: hydrolysis of short-chain oxidized phospholipids
GSPx-3	Plasma glutathione selenoperoxidase 3: reduction of hydroperoxides by glutathione
Lipid transfer proteins
PLTP	Phospholipid transfer protein: conversion of HDL into larger and smaller particles, transport of LPS; positive acute-phase reactant
CETP	Shuttles cholesteryl esters and triglycerides between HDL and apoB-containing lipoproteins.
Acute-phase proteins
SAA	Serum amyloid A; acute-phase protein
LBP	LPS binding protein; binds phospholipids, thereby acting as a lipid exchange protein similar to CETP and PLTP
α-1 acid glycoprotein 2	Also called orosomucoid-2; negative acute-phase reactant
α-2 HS glycoprotein	Also called fetuin-A; promotes endocytosis; possesses opsonic properties
Fibrinogen α-chain	Precursor of fibrin, cofactor in platelet aggregation
Complement components
C3	Complement activator through both classic and alternative activation pathways
C4	Activation of the classic pathway of the complement system
C4b-binding protein	Controls the classic pathway.
C9	Pore-forming subunit of the membrane attack complex
Vitronectin	Inhibitor of the membrane-damaging effect of the terminal cytolytic pathway
Proteinase inhibitors
α-1 antitrypsin	Inhibitor of serine proteinases
α-2 antiplasmin	Inhibits plasmin and trypsin and inactivates chymotrypsin
Haptoglobin-related protein	Decoy substrate to prevent proteolysis
Other proteins
Retinol-binding protein	Delivers retinol from liver to peripheral tissues
Transthyretin	Carries the thyroid hormone thyroxine (T4) and retinol-binding protein bound to retinol
Serotransferrin	Iron transport glycoprotein
Vitamin D-binding protein	Vitamin D binding and transport
α-1B glycoprotein	Function unknown
Hemopexin	Iron binding protein

This list is incomplete because only the proteins detected and confirmed by other studies are provided, according to Kontush et al. and Shal et al.^159,160 LCAT, lecithin–cholesterol acyltransferase; LPL, lipoprotein lipase; LDL-R, LDL-receptor; LRP, lipoprotein receptor-related protein. Modified from reference 159, with permission.

Usually HDL-binding peptides, proteins, and microRNAs (miRNAs) are small and the binding of these components to HDL components, especially under uremic conditions might be different compared with healthy controls. Unbound molecules are cleared by glomerular filtration and might therefore accumulate in plasma in the case of impaired GFR. An accumulation in plasma could therefore be an effect rather than a cause of GFR decline. However, when bound to HDL particles that are not filtered, these molecules are more attractive candidates as causative biomarkers.

ApoA-IV

ApoA-IV is the third most abundant HDL-associated protein and has similar properties as apoA-I. It is involved in the reverse cholesterol transport.^71,72 This glycoprotein is mainly produced in the small intestine⁷³ and is a structural protein of chylomicrons, VLDL, HDL, or is found free in plasma.^74,75 It activates the lipoprotein lipase,⁷⁶ lecithin–cholesterol acyltransferase,⁷⁷ and is involved in the CETP-mediated transfer of cholesteryl esters from HDL to LDL.⁷⁸ It has antiatherogenic and antioxidative properties.^{71,72,76,79,80} In line with that, mice overexpressing human or mouse apoA-IV showed less atherosclerotic lesions in the aorta.^79,80 A mouse model injected with lipid-free apoA-IV revealed a more stable plaque phenotype and a reduced incidence of acute plaque disruptions.⁸¹ This is in line with epidemiologic data in humans that observed lower apoA-IV concentrations in patients with CVD.^75,82–86 Low apoA-IV levels predicted an increased risk for all-cause mortality and sudden cardiac death in patients with CKD on hemodialysis.⁸⁷

The glomerular filtration capacity is one of the main determinants of apoA-IV concentration. Until recently, most of the data on apoA-IV and kidney function came from patients with primary CKD: apoA-IV already starts to increase in early stages of eGFR impairment.⁸⁶ However, the concentrations are modified by proteinuria and serum albumin levels in the sense that patients with a low eGFR have lower apoA-IV levels than expected when the serum albumin level is also low, compared with high serum albumin levels.⁸⁵ Patients with tubular proteinuria had significantly higher amounts of apoA-IV in their urine than those with a purely glomerular type of proteinuria. This has generated the idea that apoA-IV is glomerular-filtrated and subsequently reabsorbed in the proximal tubular cells under physiologic circumstances.⁸⁵ These observations are in line with a pronounced immunoreactivity mainly in the tubular system in histologic investigations in healthy human kidneys.⁸⁸ The concentrations of apoA-IV are highest in ESRD independent of whether treatment is performed by hemodialysis or peritoneal dialysis.⁸⁹

Besides several studies in patients with primary CKD,^{39,85–87,89–93} only one large study in the general population including roughly 6000 individuals has been performed.⁹⁴ There was an inverse correlation between apoA-IV concentrations and eGFR: individuals in the third and fourth quartile of apoA-IV concentrations had 1.8 and 5.1 ml/min per 1.73 m² lower eGFR values, respectively, compared with the first quartile of apoA-IV. The probability for the presence of an eGFR<60 ml/min per 1.73 m² was 1.46-fold and 3.47-fold higher in the third and fourth quartile, respectively, versus the first quartile of apoA-IV. At least in the general population, apoA-IV was not significantly associated with the presence of a urinary albumin-to-creatinine ratio ≥30 mg/g. This supports the idea that apoA-IV concentrations are influenced by the glomerular filtration capacity rather than by albuminuria in the general population.⁹⁴ It is currently not known, whether the higher apoA-IV concentrations in individuals with kidney impairment are also a response to the disease enhancing the reverse cholesterol transport to relieve mesangial cells from an overload of cholesterol.⁹⁴ The increase of antioxidative apoA-IV could also be a reaction to the high oxidative stress in CKD.⁹⁵

Genetic support for a causal association between kidney function and apoA-IV concentrations came from a recent GWAS.⁹⁶ This study used 53 SNPs that have been identified to be associated with kidney function.⁹⁷ It found that the more genetic eGFR-decreasing kidney function variants an individual carried, the higher their apoA-IV concentrations were.⁹⁶ However, the opposite analysis of whether SNPs that have a significant influence on apoA-IV concentrations also have an effect on kidney function, remained negative and did not find any of the three apoA-IV SNPs recently identified⁹⁶ as being associated with eGFR.⁹⁸

Finally, a prospective study in patients with primary nondiabetic mild to moderate CKD has investigated whether apoA-IV levels predict progression of CKD, defined as doubling of baseline serum creatinine and/or terminal renal failure necessitating RRT.³⁹ Each increase of apoA-IV by 10 mg/dl increased the risk for a progression by 62% independent of baseline GFR. Patients with apoA-IV levels above the median had a significantly faster progression compared with patients with concentrations below the median.³⁹ A recent report confirmed apoA-IV to predict rapidly declining kidney function in patients with type 2 diabetes mellitus independent of conventional clinical variables.⁹⁹

This finding of high apoA-IV concentrations and CKD progression was unexpected, considering the physiologic functions of apoA-IV. It has been speculated that apoA-IV is either not fully functionally active or that the increase of apoA-IV is an unsuccessful attempt to slow or to halt the progression. High apoA-IV levels could also reflect an aspect of renal impairment that is not parallel to GFR decline.³⁹

Apo L1

ApoL1 is a minor component of HDL-C particles in the more dense fractions HDL_3b/c.¹⁰⁰ ApoL1 concentrations have been suspected to only be a surrogate of other antioxidative and anti-inflammatory proteins, such as paraoxonases.¹⁰¹

In 2010, the description of an association of certain APOL1 genetic variants (called G1 and G2) with the susceptibility to nondiabetic CKD significantly contributed to the understanding of the 3–4-fold higher incidence rates for ESRD in blacks compared with whites.^102,103 About 13% of blacks carry high-risk genotypes. This frequency is drastically increased to around 70% in patients of this ethnicity with FSGS, HIV-associated nephropathy, and to 45% in patients with hypertension-associated kidney disease.^102,103 These risk alleles are practically nonexistent in nonblack populations.^104,105 Sequencing studies in that gene region did not reveal other independent risk variants than G1 and G2.^106,107

ApoL1 is highly interesting from an evolutionary standpoint and is kind of a Trojan horse, causing human resistance against Trypanosoma brucei brucei. When HDL particles are taken up as nutrients by these parasites, the HDL-attached apoL1 is inserted in the membranes of the parasites and results in pore formation and lysis of the parasites. Interestingly, Trypanosoma brucei rhodesiense has developed a serum resistance-associated protein (SRA) that binds to apoL1 and thereby avoids pore formation. This finally results in sleeping sickness. However, the two genetic variants G1 and G2 of the APOL1 gene are located in the SRA-interacting domain of apoL1. Their presence no longer allows the binding of SRA to apoL1 resulting in resistance against T. brucei rhodesiense.¹⁰⁴ On the other hand, carriers of these alleles have a markedly increased risk for nondiabetic CKD as well as kidney function decline.^108–112 The mechanism for this is still a matter of debate recently reviewed by Dummer et al.,¹⁰⁴ and might mainly be related to mechanisms such as lysosomal membrane permeabilization and autophagic cell death. Others observed that the risk variants increased cytoplasmic cell swelling mediated by an increased efflux of intracellular K⁺ and a subsequent influx of Na⁺ with Cl⁻ and H₂O.¹¹³

Because not all carriers of the risk genotypes develop CKD, a “second hit” theory has been developed that involves the innate immune system and the coagulation system.¹¹⁴ This is in line with a significant interaction between the APOL1 high-risk genotype and factor VIIIc and protein C for the development of ESRD.¹¹⁵ Recent data showed that soluble urokinase plasminogen activator receptor (suPAR), a marker of immune activation and systemic inflammation, modifies the association between APOL1 risk variants and decline in kidney function: the risk was attenuated in patients with lower suPAR, and enforced in those with higher suPAR levels. The authors identified high-affinity interactions between apoL1 and suPAR, as well as between apoL1 and αvβ3 integrin. They showed that the G1 and G2 risk variants synergized with suPAR in the activation of αvβ3 integrin on podocytes, which leads to the formation of autophagosomes, dysregulation of the actin cytoskeleton, and cell detachment.¹¹⁶

Recent studies also investigated the association of these high-risk variants G1 and G2 with CVD outcomes and the results were quite heterogeneous in the general population.^117–120 A cohort of black patients with hypertension-attributed CKD followed for up to 12 years found APOL1 risk variants not to be associated with an overall risk for CVD, although some signals for cardiovascular mortality were observed.¹²¹ Furthermore, these risk variants were associated with prevalent CVD in a cohort of patients with SLE.¹²²

Serum Amyloid A

A shotgun proteomics approach identified 49 HDL-associated proteins in a uremia-specific pattern. One of the most interesting ones was serum amyloid A (SAA) because it promoted inflammatory cytokine production and inversely correlated with the anti-inflammatory potency of HDL.¹²³ A follow-up study in transplanted patients revealed that after restoration of kidney function, many of these proteins do not change and the functional alterations of HDL in these patients might be contributing to the remaining high cardiovascular risk after transplantation.¹²⁴ Interestingly, one of the proteins that modified the effect of HDL on CVD is SAA. As long as concentrations of this acute phase protein are below the 80th percentile, HDL-C was inversely and independently associated with all-cause and cardiovascular mortality as seen in non-CKD populations. However, when SAA was above the 80th percentile, HDL-C lost its protective function and high HDL-C concentrations were even associated with an increased all-cause and cardiovascular mortality. This might be explained by the observation that excessive SAA can displace antiatherogenic apoA-I from the HDL surface as an inflammatory response, and in extreme circumstances, SAA can account for up to 80% of the HDL proteins.⁵ This risk-modifying effect was not only shown in the general population but also in a CVD group,¹²⁵ as well as in patients on hemodialysis with diabetes mellitus.¹²⁶ Cell culture studies revealed that HDL supplemented with SAA significantly inhibited the nitric oxide (NO) production and significantly increased the endothelial production of reactive oxygen species.¹²⁵

Asymmetric and Symmetric Dimethylarginine

Asymmetric dimethylarginine (ADMA) and symmetric dimethylarginine (SDMA) are structurally related to L-arginine and may interfere with L-arginine–related signaling. However, they differ in their functional effects on NO synthesis. ADMA is an established competitive inhibitor of the NO synthase, which is less the case for SDMA.¹²⁷ Both ADMA and SDMA are associated with CVD as well as total mortality.^128,129 Several studies described ADMA and SDMA to be connected to kidney impairment,¹³⁰ as well as to the progression of CKD.^131–140 Recent data identified SDMA within HDL particles of CKD patients but not in healthy controls, and SDMA was considered to be the culprit that modifies HDL and reverts the endothelial-protective properties. This modified HDL induces endothelial dysfunction, impairs endothelial repair, increases BP, and activates endothelial TLR-2 signaling, resulting in enhanced reactive oxygen species production and inhibition of endothelial NO bioavailability.² SDMA itself is considered as functionally inactive, and it was therefore hypothesized that SDMA may form a complex with apoA-I, the major apo of HDL particles: apoA-I inhibited endothelial NO production after supplementation with SDMA, whereas apoA-I without SDMA did not significantly affect endothelial NO production.² Interestingly, recent data found a pronounced correlation between serum SDMA and SDMA in HDL. A higher serum SDMA was independently associated with a lower HDL-associated cholesterol efflux capacity and an accumulation of SDMA in HDL abolished the anti-inflammatory and regenerative properties of HDL.¹⁴¹ Because this accumulation of SDMA in HDL seems to be quite specific for CKD and is already present in very early phases of kidney impairment, it might be an important contributor to the disturbances in cholesterol efflux in patients with CKD. These data are in line with the finding that HDL from children with CKD compared with those without CKD inhibited NO production, promoted superoxide production, increased the expression of vascular cell adhesion molecule 1 in human aortic endothelial cells, and reduced cholesterol efflux from macrophages.²¹ Interestingly, two recent meta-analyses with slightly different study inclusion criteria found a significant association of ADMA but not of SDMA with CVD outcomes in patients with kidney diseases.^128,129

Advanced Oxidation Protein Products

Advanced oxidation protein products (AOPPs) are markers of oxidative stress and are carried by oxidized plasma proteins such as oxidized albumin. They accumulate in renal disease as well as CVD.^142,143 AOPP-albumin binds with high affinity to SR-B1 and thereby blocks the binding of HDL to SR-B1 and the cholesterol ester uptake. Albumin isolated from patients on hemodialysis but not from healthy controls markedly inhibited SR-B1–mediated HDL-cholesteryl ester transfer in vitro, dependent on the AOPP content of albumin. It has been suggested that this results in an abnormal composition of HDL in CKD.¹⁴⁴ Studies in patients with IgA nephropathy described renal AOPPs as a predictor for progression of the disease.^145,146 Experimental studies demonstrated that AOPPs have an influence on mesangial cell perturbations as well as on podocyte apoptosis.^147,148

miRNAs

miRNAs are small, noncoding RNAs that have an influence on gene expression of their targeted genes by posttranscriptional regulation of mRNA stability and translation.¹⁴⁹ They are not only found in the intracellular but also in the extracellular space, where they participate in intercellular communication. To avoid neutralization by ribonucleases, they are protected by an association with protein and lipid carriers.¹⁵⁰ One of the important carriers are HDL lipoproteins.¹⁵¹ There is growing evidence that miRNAs control many genes involved in HDL metabolism, thereby regulating HDL biogenesis, cellular cholesterol efflux, HDL-C uptake in the liver, and many steps in reverse cholesterol transport.¹⁵² A number of miRNAs are already known to control key elements in this cascade, such as ABCA1, ABCG1, and SR-B1 (for review, see Canfrán-Duque et al.¹⁵²). A pronounced network of regulators at different sites result in a granular regulation of the key elements in HDL metabolism and reverse cholesterol transport. For example, miR-33 not only influences HDL biogenesis by controlling ABCA1, but also cholesterol efflux from peripheral cells by controlling ABCA1 and ABCG1, bile acid synthesis by regulating CYP7A1, as well as bile acid secretion by influencing ABCB11 or ATP8B1, and thereby bile excretion.^153,154 In addition to miR-33, ABCA1 and ABCG1 at the peripheral cells are regulated by miR26, miR-27, and miR-758. The uptake of cholesterol via SR-B1 is regulated by miR-185, miR-96, miR-223, and miR-125a.¹⁵² We are only just beginning to understand this extremely flexible regulation system. Epidemiologic studies of these miRNAs and their influence on metabolic or cardiovascular outcomes are starting to appear, but their results are contrasting (for review see Willeit et al.¹⁵⁵). Many methodological problems need to be solved before the reliability of miRNAs as potential biomarkers can be investigated in depth.

miRNAs seem to play an important role in renal physiology and pathophysiology.^156,157 It is suggested that miRNA processing in renal cells and especially podocytes is of utmost importance for renal morphology and function. They are involved in water homeostasis, osmoregulation, renin production, calcium sensing, sodium and potassium handling, and renal senescence.¹⁵⁸ Because of the wide range of involvement of miRNAs in pivotal processes of cell cycle, stem cell differentiation, and apoptosis, a further layer in the understanding of the regulation of lipoprotein metabolism and development of atherosclerosis has started to be introduced.¹⁵⁶ However, we are currently missing large studies that investigate the influence of this regulatory layer on outcomes in patients with CKD.

Conclusions

The field of HDL-C is full of discrepancies. Epidemiologic studies are not in line with genetic studies and intervention studies that increase HDL-C. The latter two are not in favor of HDL-C as a causal factor for CVD and/or CKD. However, the content of cholesterol in the HDL particle might not be the responsible culprit, considering that HDL particles contain a very large number of proteins and lipids. Therefore, during recent years, there has been a major shift of focus toward the functionality of HDL particles and the role of its components in the reverse cholesterol transport and especially on cholesterol efflux capacity. Many disturbances have been seen in patients with CKD and we are only at the beginning of this journey, for which further studies are required. The results have been heterogeneous, which might on the one hand be explained by the heterogeneity of kidney disease and the different stages of the disease itself, and on the other hand by the insufficient standardization and accuracy of the various assays. The associations between the observed disturbances and end points were often much stronger in non-CKD than in CKD populations. There is still much to learn before this knowledge can be effectively translated into potential diagnostic and therapeutic strategies.

Disclosures

The author declares no competing interests. The Medical University of Innsbruck holds patents on apoA-IV as a marker for kidney impairment and progression of CKD.