Multiomic Biomarkers Associated with Early Traits of Atherosclerosis and Osteoporosis

Abstract

Atherosclerosis and osteoporosis are complex multifactorial diseases, each contributing to a significant disease burden and related health costs globally. Several studies have shown that these diseases share risk factors and pathophysiological mechanisms suggesting that they are co/multimorbid conditions. Follow-up studies of the diseases for in-depth understanding of common underlying molecular mechanisms are essential for joint prevention and therapeutics of these diseases. However, despite the strong evidence for co/multimorbidity hypothesis of the diseases, studies investigating the underlying molecular mechanisms of the diseases using omics data, such as transcriptomics and lipidomics, are lacking.

The aims of the study were, study I: Identification of lipidome-wide molecular lipid co-expression modules jointly associated with early markers of atherosclerosis and osteoporosis, study II: identification of transcriptome-wide gene co-expression modules jointly associated with early markers of atherosclerosis and osteoporosis, study III: lipidome-wide multivariate association analysis of early markers of both the diseases to identify molecular lipid species with potentially distinct biological role in the comorbidity/multimorbidity, study IV: identification of the shared biological processes underlying atherosclerosis-osteoporosis co/multimorbidity.

All the studies (I-IV) in this thesis are based on the data collected from ongoing Cardiovascular Risk in Young Finns Study (YFS). The YFS is a prospective multicenter follow-up study initiated in 1980 with 3596 children and adolescents aged 3 to 18 years randomly selected from the areas of five university hospitals in Finland, investigating cardiovascular risk factors from childhood to adulthood. The participants have been since followed up for over 40 years with regular intervals. Early markers of atherosclerosis used in this study included bulbus and carotid intima media thickness (CIMT) measured with high-resolution ultrasound during 2007 follow-up. Similarly, early markers of osteoporosis included indices of bone mineral density (BMD) and content, measured using peripheral quantitative computer tomography from the distal and shaft sites of tibia and radius during YFS 2007 follow-up at age 30-45 years.

In studies I and II, weighted co-expression network analyses of these early markers of atherosclerosis and osteoporosis were performed using lipidomic data profiled during 2007 using liquid chromatography-tandem mass spectrometry (study I) and genome-wide transcriptomic data profiled during 2011 from whole blood of YFS participants using Illumina HumanHT-12 version 4 Expression BeadChip (study II). The network analyses were performed to identify densely interconnected networks or modules of lipid species or genes that are associated with early markers of both the diseases. Joint association of the identified lipid or gene modules with the early markers of both diseases was assessed by performing multivariate analysis of variance (MANOVA). In study III, we performed MANOVA of the lipidome data to identify joint associations between each of the 437 individual molecular plasma lipid species and selected early markers of both diseases. In study IV, we performed gene set analysis (GSA) of the transcriptomic data to identify biological processes shared by the early markers of these two diseases. The GSA was done in case-control setting. Participants with high CIMT (>90th percentile) were defined as cases for subclinical atherosclerosis. In case of osteoporosis, study population-based T-scores for BMD were calculated and T-score ≤ −1 was used for the definition of low BMD cases.

In study I, we identified one plasma lipidomic module containing 105 lipid species that was significantly and jointly associated with early markers of both atherosclerosis and osteoporosis. Majority of lipid species in the module belonged to classes of glycerolipid (n=60), glycerophospholipid (n=13) and sphingolipid (n=29). Twenty of the lipid species belonging to class sphingolipid were plasma ceramides. In study II, we identified two gene modules significantly associated with the early markers of both diseases. The three most significant genes in the two identified gene modules were NOSIP, GXYLT2 and TRIM. In study III, we identified four lipid species significantly associated with the early markers of both diseases. The four lipid species and the classes they belong to are TAG(18:0/18:0/18:1) (glycerolipid), PC(40:3) (glycerophospholipid), Gb3(d18:1/22:0) (sphingolipid) and Gb3(d18:1/24:0) (sphingolipid). In study IV, we did not identify any biological processes jointly associated with the early markers of both diseases. However, we identified three novel biological processes associated with high CIMT and replicated 234 gene sets significantly associated with high CIMT with false discovery rate (FDR) ≤ 0.01. These three novel biological processes were copper homeostasis, neural crest cell migration and nicotinate and nicotinamide metabolism.

The results from studies I-III support the atherosclerosis-osteoporosis co/multimorbidity hypothesis. They reveal a set of joint lipidomic and transcriptomic biomarkers for early signs of both the diseases which might imply their important role in developing dual-purpose prevention and/or treatment methods. The results of study IV show three gene sets representing three novel biological processes associated high CIMT, which might explain the transcriptomic link between these biological processes and atherosclerosis and serve as biomarkers of atherosclerosis. Methodologically, these studies (I-IV) highlight the importance of bioinformatic approaches such as weighted co-expression network analysis and gene set analysis for exploratory investigation of omics based shared structures and molecular features for co/multimorbidities.