Cory Bystrom (Presenter)
Cleveland HeartLab, Inc.
Authorship: Timothy S. Collier (1), Darlene L.Y. Dai (2), Zsuzsanna Hollander (2), Raymond Ng (2), Bruce McManus (2), Marc Penn (1), and Cory Bystrom (1)
(1) Cleveland HeartLab Inc., Cleveland, OH (2) PROOF Centre of Excellence, Vancouver, BC
| Short Abstract High density lipoprotein (HDL) plays many roles in vascular biology. Cholesterol efflux capacity (CEC) is one functional role that has been linked to cardiovascular disease in cell-based studies. Here we present a targeted method for the quantitation of 21 HDL-associated proteins from which a mathematical model for the prediction of CEC was discovered. This model was further refined for the stratification of patients with coronary artery disease and healthy controls, and confirmed in a separate experiment. The results of this work is a rapid, scalable, precise prototype assay that enables further exploration of the relationship between HDL and vascular biology. |
Long Abstract
Introduction
High density lipoprotein (HDL) plays a number of roles in vascular biology, including lipid and cholesterol transport, anti-oxidative and anti-thrombotic activity, endothelial regulation, and immune response regulation. Cholesterol efflux capacity (CEC), the transport of intracellular cholesterol from macrophages to an accepting HDL particle, is the first HDL function to be conclusively associated with cardiovascular disease risk in human clinical studies. CEC is traditionally assessed in a cell-based assay using radio-isotope or fluorescently labeled cholesterol, and is subject to limitations in speed, throughput, scalability, and precision that limit its potential as a clinical tool. Given the close association of ApoA-I, HDL’s primary structural protein, with the processes of CEC, we hypothesized that other HDL-associated proteins would likely affect CEC, providing a means of CEC prediction through compositional analysis. Using a novel HDL isolation technique coupled to mass spectrometry-based proteomics, we developed a targeted quantitative assay for 21 HDL-associated proteins and built correlative models associating HDL protein composition with cholesterol efflux capacity. The best predictive model, applied to a clinical population, demonstrated an inverse relationship between predicted CEC and diagnosed cardiovascular disease.
Methods
All sample preparation and data analysis was performed on randomized and blinded samples.
Cohorts of 70 training samples and 35 test samples were used for predictive model development. Evaluation of predictive models was performed in a sample set from the Fairbanks Institute for Healthy Communities Biobank consisting of 74 samples with diagnosed coronary artery disease (CAD), 83 samples with diagnosed CAD and a major adverse cardiovascular event, and 74 matched controls.
Recombinant His-tagged ApoA-I was added to serum samples and incubated at 37 deg C. Purification was performed using an automated liquid handler utilizing pipet tips containing an IMAC column bed. For proteomic analysis, eluted HDL was denatured with DTT and digested with the addition of Lys-C and incubation at 37 deg C for up to 4 hours. A stable isotope-labeled peptide mixture was added to each sample post-digestion.
Peptides were separated by reversed-phase chromatography utilizing a 5-minute gradient and detected using an Agilent 6490 triple-quadrupole mass spectrometer operating in dynamic MRM mode. Transitions were selected to be unique to the peptide targeted in the sample. Two transitions were monitored per peptide, and up to two peptides were targeted per protein. All peaks were manually reviewed using qualifier ion ratios and internal standard peaks.
The cell-based assay to assess CEC was performed externally by Vascular Strategies, Inc. on training and test samples in addition to a subset of clinical specimens. Serum samples were depleted of LDL by PEG precipitation and efflux measured from J774 macrophages primed with 3H-labeled cholesterol.
Statistical analysis of the data was performed using R, with normalized peak intensities subjected to univariate robust linear regression with elastic net filtering and multivariate analyses including partial linear regression, random forest, and logistic regressions with 5-fold cross validation to produce predictive models with maximum spearman correlation and low median absolute difference (cost) as key metrics for model selection.
Results
Targeted proteomic analysis of 21 HDL-associated proteins in training and test data sets yielded a 6-protein model for the prediction of cholesterol efflux capacity (Spearman r=0.65, P<0.001). When used to discriminate between healthy specimens and those with diagnosed CAD, the predictive CEC algorithm showed modest performance (ROC curve AUC=0.62, P=0.02). However, reweighting the coefficients of the predictive model’s component proteins to optimize diagnosis of CAD yielded model with significantly greater diagnostic performance (ROC curve AUC=0.73, P<0.001). The model’s performance was confirmed in a second sampling of different individuals from the same specimen bank (n=92 Control, n=92 CAD).
Conclusions
The proteomic analysis of HDL from a serum sample to predict cholesterol efflux capacity is rapid, scalable, precise and provides a new avenue to further explore the structural and functional relationships between HDL and cardiovascular disease processes.
References & Acknowledgements:
The authors thank Alan Swick, Brenda Lee Hudson, Roberta Ambuhel for invaluable laboratory support and extensive consultation regarding specimens and clinically associated information. This project was supported, in part, with samples from the Fairbanks Institute for Healthy Communities funded by the Richard M. Fairbanks Foundation and the Indiana Clinical and Translational Sciences Institute funded, in part, by Grant Number TR000006 from the National Institutes of Health, National Center for Advancing Translational Sciences, Clinical and Translational Sciences Award. The authors also wish to acknowledge that this work was completed in collaboration with the Regenstrief Institute, Inc.
| Description | Y/N | Source |
| Grants | no | |
| Salary | yes | Cleveland HeartLab, Inc. |
| Board Member | no | |
| Stock | yes | Cleveland HeartLab, Inc. |
| Expenses | no |
IP Royalty: no
| Planning to mention or discuss specific products or technology of the company(ies) listed above: | no |