A Long Road Ahead for Discovering New HDL Metrics That Reflect Cardiovascular Disease Risk

Jay W. Heinecke, MD; Karin E. Bornfeldt, PHD


J Am Coll Cardiol. 2017;70(2):179-181. 

Clinical and epidemiological studies show a robust, inverse association between the standard clinical metric, high-density lipoprotein???cholesterol (HDL-C) level, with cardiovascular disease (CVD) risk.[1???3] This observation has triggered intense interest in targeting HDL-C for therapeutic intervention.

However, several lines of evidence suggest that the association between HDL-C levels and CVD status is indirect and that elevating HDL-C is not necessarily therapeutic. Indeed, genetic variations that associate with altered HDL-C do not strongly associate with altered CVD risk.[4,5] Trials of certain drugs that elevate HDL-C, such as cholesteryl ester transfer protein inhibitors and niacin, also have failed to reduce events in statin-treated subjects with established CVD.[3] Moreover, alterations in expression levels of certain proteins involved in murine HDL metabolism greatly increase both HDL-C levels and atherosclerosis.[6,7] Recent studies strongly supported the proposal that alterations in scavenger receptor B, the hepatic receptor that mediates uptake of HDL cholesteryl esters, likewise increase both HDL-C and the risk of CVD in humans.[8]

These observations indicated that HDL-C does not necessarily reflect HDL's cardioprotective effects in either humans or mice. However, genetically engineered deficiencies in proteins implicated in HDL's function markedly affected atherosclerosis in mice,[9] providing compelling evidence that HDL is in the causal pathway of accelerated atherogenesis in animal models. It is therefore critical to identify new HDL factors that modulate CVD risk in humans.

In this issue of the Journal, Padr?? et al.[10] investigate changes in HDL's composition and function induced by a hypercholesterolemic diet. They fed pigs a low-fat diet or a high-fat diet for 10 days, isolated HDL from blood by ultracentrifugation, and subjected the HDL to lipidomic and proteomic analysis, using mass spectrometric methods. They identified 255 molecular species of lipids and >500 protein spots using 2-dimensional gel electrophoresis. Then, using principal component analysis, they focused on 55 lipids that varied at least 1.5-fold between normocholesterolemic and hypercholesterolemic HDL. Protein spots identified as retinol binding protein-4, apolipoprotein M, and cellular retinoic acid binding protein-1 showed reduced expression in hypercholesterolemic HDL, which also had reduced antioxidant capacity and impaired cholesterol efflux capacity in functional studies. The authors concluded that hypercholesterolemia remodeled HDL lipids and proteins, inducing a dysfunctional state.

Using the same pig model, the group had previously shown that hypercholesterolemia impairs HDL's ability to protect the heart after experimental myocardial infarction.[11] Because pigs and humans share many similarities in lipoprotein metabolism, and given the translational relevance of cardiac ischemia, this approach has the potential to yield new insights into HDL's role in cardioprotection.

What were the strengths and weaknesses of the current study, and what challenges face investigators who want to identify new HDL metrics that are clinically relevant? Use of a clinically relevant animal model and the focus on HDL were clearly strengths of the work of Padr?? et al.[10] The emphasis on a global approach to HDL's lipids and proteins also had many advantages because of the complexity of the many subspecies of HDL. Moreover, the authors attempted to link these alterations in HDL's composition to impaired cholesterol efflux capacity. This was critical because multiple studies of large numbers of subjects have shown that impaired cholesterol efflux capacity of HDL strongly predicted incident and prevalent CVD risk,[12,13] although the molecular basis for altered cholesterol efflux capacity in humans with CVD is poorly understood. In fact, recent studies showed that reduced HDL particle number was a stronger predictor of incident CVD events than was HDL's cholesterol efflux capacity in some patient populations.[14] Finally, it would greatly have strengthened the results of the study to replicate the findings in a second group of animals.

It will be critical in future studies to confirm the differential levels of the lipids and proteins in hypercholesterolemic HDL of pigs, and to extend those observations in humans. One important concern, however, is the small number of pigs used in the study by Padr?? et al..[10] Only 7 pigs were placed on each of the 2 diets, yet the relative abundance of more than 200 lipid species in HDL was analyzed by principal component analysis. Although the authors adjusted the statistical analysis for multiple comparisons, it is difficult to interpret the significance of the results. There also were no data on the intra- and interassay coefficients of variation in the analysis. The issue of small numbers becomes even more challenging in the protein analysis, where more than 500 protein spots were quantified but only 4 analyses were performed for each group of pigs.

Rifai et al.[15] cogently laid out a strategy for discovering new and clinically useful biomarkers. Their approach involves 6 essential steps: candidate discovery, qualification, verification, assay optimization, clinical validation, and commercialization. The work of Padr?? et al.[10] falls into the first category: biomarker discovery.

Blood is a logical place to search for biomarkers because it is readily obtained at relatively low cost and with minimal invasiveness. A major impediment is the complexity of blood, which remains poorly characterized and is likely to contain thousands of core proteins and lipids. Two important advantages of focusing on HDL for biomarker discovery, as in the study by Padr?? et al.,[10] are HDL's clinical relevance as an established index of CVD risk and its vastly simpler protein and lipid repertoire relative to blood, plasma, or serum. A significant limitation is the necessity of isolating HDL for analysis. The clinical gold standard for HDL, ultracentrifugation, is time consuming and expensive.

The discovery of candidate biomarkers and confirmation of differential expression (qualification) generally involves small numbers of samples. The next key step in biomarker discovery is verification, the assessment of the specificity of candidate biomarkers and their relevance to specific diseases. This process typically involves the analysis of hundreds or even thousands of samples from multiple populations of subjects. Validation of assays and establishing the clinical relevance of a biomarker involve the analysis of even greater numbers of samples.

Padr?? et al.[10] have taken a first step towards identifying alterations in HDL that may be clinically relevant. Many other changes in HDL, ranging from antioxidant potential to the ability to regulate nitric oxide synthase activity in endothelial cells, have been proposed to be clinically relevant. But with the exception of the cholesterol efflux capacity assay, none of these functional effects have been quantified in large numbers of subjects by investigators from different institutions. The key challenge for identifying new HDL metrics is establishing the diagnostic accuracy of proposed biomarkers in large and diverse clinical populations, using validated reproducible assays with established sensitivity and specificity.