Personalized Profiling of Lipoprotein and Lipid Metabolism Based on 1018 Measures from Combined Quantitative NMR and LC-MS/MS Platforms

The goal of this study is to advance molecular epidemiology by integrating and analyzing large-scale population data from two complementary platforms: quantitative NMR spectroscopy for 209 lipoprotein measures and LC-MS/MS for 809 lipid classes and species. Together, these datasets encompass 1018 molecular measures across 7830 participants from two population cohorts. The study aims to elucidate the molecular and compositional relationships within the circulatory lipid transport system and their connections to physiological and clinical characteristics.

Through cluster analyses and unsupervised neural network techniques (self-organizing maps), the study identifies inherent links between deep molecular data and population health traits, including cardiometabolic risk factors such as lipoprotein lipids, apolipoprotein B, ceramides, and oxidized lipids. Notable findings include the consistent cardiometabolic risk associations of triglyceride species across the lipoprotein cascade and the intricate metabolic profiles of cholesteryl ester species, which vary in their risk implications. This work highlights the unprecedented detail and compatibility of combined omics methodologies, paving the way for comprehensive molecular insights and new opportunities in large-scale population health studies.

The original article

Personalized Profiling of Lipoprotein and Lipid Metabolism Based on 1018 Measures from Combined Quantitative NMR and LC-MS/MS Platforms

Siyu Zhao, Corey Giles, Kevin Huynh, Johannes Kettunen, Marjo-Riitta Järvelin, Mika Kähönen, Jorma Viikari, Terho Lehtimäki, Olli T. Raitakari, Peter J. Meikle, Ville-Petteri Mäkinen, Mika Ala-Korpela

Analytical Chemistry, Vol 96/Issue 52

https://doi.org/10.1021/acs.analchem.4c03229

licensed under CC-BY 4.0

Selected sections from the article follow. Formats and hyperlinks were adapted from the original.

Abstract

Applications of advanced omics methodologies are increasingly popular in biomedicine. However, large-scale studies aiming at clinical translation are typically siloed to single technologies. Here, we present the first comprehensive large-scale population data combining 209 lipoprotein measures from a quantitative NMR spectroscopy platform and 809 lipid classes and species from a quantitative LC-MS/MS platform. These data with 1018 molecular measures were analyzed in two population cohorts totaling 7830 participants. The association and cluster analyses revealed excellent coherence between the methodologically independent data domains and confirmed their quantitative compatibility and suitability for large-scale studies. The analyses elucidated the detailed molecular characteristics of the heterogeneous circulatory macromolecular lipid transport system and the underlying structural and compositional relationships. Unsupervised neural network analysis─the so-called self-organizing maps (SOMs)─revealed that these deep molecular and metabolic data are inherently related to key physiological and clinical population characteristics. The data-driven population subgroups uncovered marked differences in the population distribution of multiple cardiometabolic risk factors. These include, e.g., multiple lipoprotein lipids, apolipoprotein B, ceramides, and oxidized lipids. All 79 structurally unique triglyceride species showed similar associations over the entire lipoprotein cascade and indicated systematically increased risk for carotid intima media thickening and other atherosclerosis risk factors, including obesity and inflammation. The metabolic attributes for 27 individual cholesteryl ester species, which formed six distinct clusters, were more intricate with associations both with higher─e.g., CE(16:1)─and lower─e.g., CE(20:4)─cardiometabolic risk. The molecular details provided by these combined data are unprecedented for molecular epidemiology and demonstrate a new potential avenue for population studies.

Lipoprotein lipids have had a central role in clinical risk assessment for cardiometabolic diseases since the 1950s. (1) The realization that ¹H NMR spectroscopy would be inherently suited for lipoprotein analytics took place in the early 1990s. (2) In addition to the fundamental advantages of ¹H NMR spectroscopy in producing quantitative data, the spherical monolayered structure of lipoprotein particles provides a compelling attribute that the chemical shifts of lipid molecules transported by these particles are dependent on the particle diameter. This phenomenon arises from the orientational order of the surface lipids and the resulting anisotropy of the magnetic susceptibility, and leads to distinctive particle size-dependent frequency shifts. This facilitates comprehensive lipoprotein analytics that is pertinent to the metabolic constituents of the lipoprotein cascade. Practical applications and statistical considerations have indicated that 14 lipoprotein subclasses and their main lipid constituents can be robustly quantified. (3,4) Multiple commercial methodologies have also been developed. (5−7) Here we applied a platform that has been widely applied in large-scale epidemiology and genetics over the last 15 years with some 500 publications and 2 M samples analyzed, including all the 0.5 M serum samples in the UK Biobank. (3,5,8) This particular platform has also played a key role in yielding open access genetic data, with the most recent genome-wide association study meta-analysis being carried out in 136,000 individuals. (9)

Mass spectrometry based lipidomics methodologies have been developed independent of the above and are technologically more challenging than the applications of ¹H NMR spectroscopy. (10−12) Nevertheless, epidemiological applications have become feasible over the recent years and clinical translation has been suggested, for example, with respect to assessing cardiometabolic risk (13,14) and cognitive impairment. (15) Initial genetic studies have also been published. (16,17)

The NMR spectroscopy and MS lipidomics approaches to lipid characterization and quantification are fundamentally different (Figure 1). In NMR the serum sample is directly analyzed with the lipoprotein particles intact; this allows quantification of the comprehensive and specific data on the lipoprotein subclasses and lipids. (18) Unlike NMR, LC-MS/MS lipidomics calls for lipid extraction of the serum samples as an integral prerequisite for the analyses. (19) The lipidomics sample preparations thus represent pooled mixtures of all the lipid molecules from all the circulating lipoprotein particles. All the information from the specific lipoprotein origin of a certain lipid in the circulation is lost in the extraction phase. However, with LC-MS/MS hundreds of unique lipid species can be analyzed from dozens of different molecular lipid classes. (12,20)

Anal. Chem. 2024, 96, 52, 20362-20370: Figure 1. Schematic illustration of the methodological differences and analytical characteristics between 1H NMR and LC-MS/MS lipidomics and their molecular views into serum lipids. (A) The lipoprotein analysis with NMR is performed directly from the intact serum sample. (B) For the LC-MS/MS analysis all the lipids need to be extracted to a homogeneous lipid-soluble mixture. This allows LC-MS/MS to reach detailed molecular quantification, however, as a total serum concentration without any information on the originating lipoproteins. (C,D) The NMR analysis gives molecular information on 14 lipoprotein subclasses with the abundant core and surface lipids quantified. The percentages depict the compositional variation of these lipids in different lipoprotein particles.

Lipoprotein particles are the sole transport vehicles for all lipid molecules in serum, (18,21,22) though albumin contains the bulk of free fatty acids and acylcarnitines, and transports also various lyso-type of lipid molecules. (23,24) The NMR and the LC-MS/MS data on serum essentially represent the same lipids as noted by Ghorasaini et al. (25) The NMR provides the size-resolution and metabolic information on lipoprotein particles (209 measures) and the LC-MS/MS lipidomics opens up a prolific and detailed molecular view on all the circulating lipids (809 measures).

These different methodologies and molecular views to study serum lipids have independently advanced toward large-scale population studies. Here, we integrate these state-of-the-art data─over 1000 quantitative molecular lipid measures per individual─in two large-scale population cohorts of over 7800 participants. These data reveal the fundamental relationships between the lipoprotein subclass cascade and metabolism (NMR) and the plethora of specific molecular information (LC-MS/MS). The molecular and metabolic corollaries are numerous, and we elaborate on some pertinent findings. Among the various analyses we also demonstrate the admirable coherence within the independent experiments and data domains. It is also uncovered─via data-driven, unsupervised neural network analysis─that these molecular lipid measures per se can be utilized for metabolically and clinically meaningful subgrouping of populations. The results illustrate that these combined comprehensive data lead to improved molecular characterization of population-level metabolic health and cardiometabolic risk.

Materials and Methods

¹H NMR Spectroscopy

We applied an NMR platform that has been widely used in epidemiology and for which the general methodological issues have been published. (3,5,8,9,18) The separation of lipoprotein subclasses by proton NMR spectroscopy is based on particle size, (2) the platform resolution being 14 subclasses (Figure 2). (3,18) Independent verification for the number of subclasses has been published by Mihaleva and co-workers with an in-depth handling of statistical grounds. (4) Particle (P), phospholipid (PL), free cholesterol (FC), cholesteryl ester (CE), and triglyceride (TG) concentrations are quantified for all 14 subclasses, also allowing calculation of the relative lipid composition for each lipoprotein subclass. The platform also provides all key clinically used lipid measures (e.g., total TG, cholesterol, and HDL-C) as well as apolipoprotein B (apoB) and A-I (apoA-I) concentrations. In total 209 lipoprotein-related measures were analyzed (98 concentration and 70 compositional measures from 14 lipoprotein subclasses, 19 clinical lipid measures, and 22 measures related to fatty acids, average particle sizes, and apoA-I and apoB).

LC-MS/MS Mass Spectrometry

Serum lipids were extracted as previously described. (19) The LC-MS/MS profiling was performed using the standardized methodology described recently by Huynh et al. (12) and updated with a larger set of targets. The mass spectrometric details and parameters are given in the Supporting Information.

Conclusions

The presented combination of lipoprotein and lipid data from NMR spectroscopy and LC-MS/MS lipidomics led to an unprecedented metabolic and molecular detail with over 1000 measures per person in a population setting of over 7800 individuals. The analyses demonstrated notable coherence between the data from inherently different spectroscopic techniques. The data-driven analyses revealed multiple confirmatory and novel molecular results in relation to systemic lipid metabolism and illustrated the power of all-inclusive lipid data in understanding cardiometabolic population health, related clinical factors, and disease risk. The analyses emphasized the augmented value of interpreting lipidomics data (deep molecular phenotyping) together with lipoprotein data (deep metabolic phenotyping) and association heatmaps between all the lipidomics measures and the key measures characterizing lipoprotein subclass metabolism are provided. These unique motifs are useful as guidelines for the key lipoprotein mediators for various lipidomics associations in biomedical studies when comprehensive lipoprotein profiles are not available.