Surface fitting of 2D diffusion-edited ¹H NMR spectroscopy data for the characterisation of human plasma lipoproteins

Determining the concentration and size of lipoprotein complexes is very important due to their role in cardiovascular diseases and metabolic disorders. However, standard methods for lipoprotein fractionation are manual and time consuming and cannot be used as standard diagnostic tools. Because different subclasses of lipoproteins have different radii and, hence, different diffusion velocities, we propose a fast and reliable method that uses 2D diffusion-edited ¹H NMR spectroscopy to acquire a set of 2D spectra of plasma samples, followed by a surface fitting algorithm based on Lorentzian functions to estimate the sizes and the relative proportions of different lipoprotein subclasses. We were able to demonstrate that the derived sizes and positions related to the Lorentzian functions follow an exponential relationship for normolipidaemic and dislipaemic samples with coefficients of determination (r²) of 0.85 and 0.81, respectively. Moreover, we found a linear relationship between the width and size of the Lorentzian functions for normolipidaemic samples (r² = 0.88) while for dislipaemic samples this relation was nonlinear (r² = 0.62). Dividing our samples set into four different lipoprotein profiles (normal lipid values, low HDL/LDL ratio, high triglycerides values and both risk factors) and using principal component analysis (PCA) followed by multivariate analysis of variance (MANOVA), our method was able to statistically discriminate between those groups, with p-values of 0.0016, 0.0006, <1e^-4 and 0. 0035, respectively. These parameters are characteristic and indicative of different lipoprotein profiles and can be used to distinguish between normolipidaemic, hypercholesterolaemic, hypertriglyceridaemic and chylomicronaemic profiles.