"-omics" as tool to address the burden of non-communicable age-related disease in populations in epidemiological transition

Summary

Background and objectives

The objective of this doctoral thesis was to explore “‑omics” technologies as a tool to address the burden of non-communicable age-related disease in populations in epidemiological transition.  We choose diastolic left ventricular (LV) dysfunction, heart failure (HF) and chronic kidney disease (CKD) as disease models, because of their high prevalence in the general population, the global rise in their incidence as highlighted by the 2013 and 2015 Global Burden of Disease reports, and the unmet needs of millions of people with asymptomatic illness for whom non-invasive biomarker-based screening could make the difference between healthy ageing versus reduced longevity with low quality of life. 

Methods

Unless specified otherwise, we ran our research in the Flemish Study on Environment, Genes and Health Outcomes (FLEMENGHO), a randomly recruited family-based population cohort, in which data collection for this thesis took place at two time points: baseline (2005–2010) and follow-up (2009–2013).  We assessed diastolic LV function echocardiographically.  In analyses of diastolic LV function as continuous outcome, we considered that in early diastole a higher peak velocity of the mitral annulus (e’) and a lower ratio of the transmitral blood velocity to mitral annular velocity (E/e’) reflect a more performant diastolic LV function.  In categorical analyses, we defined diastolic LV dysfunction based on age-specific criteria capturing impaired relaxation, increased filling pressure (E/e’), or both.  We estimated the glomerular filtration rate from serum creatinine, using the Chronic Kidney Disease Epidemiology Collaboration equation.  Urinary proteomics were done by capillary electrophoresis coupled to mass spectrometry and tissue proteomics by liquid chromatography-tandem mass spectrometry.  Circulating metabolites were measured by nuclear magnetic resonance spectroscopy.  In multivariable-adjusted statistical analyses, we expressed effect sizes per 1‑SD in the explanatory variables, biomarkers in most instances.  In mechanistic studies, we accounted for multiple testing, where appropriate, and we applied partial least square analysis.  

Multidimensional urinary classifiers

In Chapter 2, we investigated the diagnostic performance of multidimensional urinary proteomic classifiers.  Chapter 2.1 was a proof-of-concept study, in which we extended the findings of a small case-control study to the population at large by demonstrating cross-sectional association of diastolic LV function with the urinary classifiers HF1 and HF2, respectively consisting of 85 and 671 peptide fragments.  With full adjustment, HF1 was associated with a 0.204 cm/s lower e’ peak velocity (p=0.007) and a 0.145 higher E/e’ ratio (p=0.020), while HF2 was associated with a 0.174 higher E/e’ ratio (p=0.008).  Of 745 participants, 9% had impaired LV relaxation and 13% had elevated E/e’.  The odds ratio of impaired relaxation associated with HF1 was 1.38 (p=0.043) and that of increased E/e’ in relation to HF2 was 1.38 (p=0.052).  Moving from a cross-sectional (Chapter 2.1) to a longitudinal (Chapter 2.2) approach, we assessed whether baseline HF1 could discriminate between normal and mildly impaired diastolic LV dysfunction 5 years later.  In continuous analyses, higher baseline HF1 was a forerunner of lower e’ and higher E/e’ at follow-up (p≤0.043).  In categorical analyses, higher HF1 at baseline predicted subclinical diastolic LV dysfunction at follow-up (p=0.013).  Over a 5‑year horizon, HF1 improved discrimination between people with normal versus mildly impaired diastolic LV function over and beyond other risk factors (p≤0.024).  In Chapter 2.3, we demonstrated in 791 FLEMENGHO participants followed up for 6.1 years that HF1 predicted 63 fatal plus non-fatal cardiovascular endpoints (hazard ratio [HR], 1.30; p=0.029) and 45 cardiac events (HR, 1.39; p=0.018).  For systolic pressure, the corresponding HRs were 0.97 (p=0.85) and 0.93 (p=0.66).  Prognostic accuracy improved (p≤0.006) by adding HF1 to models including other baseline predictors.  In Chapter 2.4, we left the treaded path of research into HF1 and HF2 in the Flemish population.  Following a case-control design, we developed and next validated a novel multidimensional urinary proteomic biomarker for HF, consisting of 96 urinary peptides, which we named Heart Failure Predictor (HFP).  The area under the receiver-operating characteristic curve was 0.70 (p=0.0047) for HFP and 0.57 (p=0.62) for NT-proBNP.  In Chapter 2.5, we moved to renal dysfunction.  CKD273 consists of 273 urinary peptides and, in advanced CKD, predicts further decline of renal function.  In line with the recommendation of the European Medicines Agency to use earlier and milder endpoints in CKD studies, in Chapter 2.5, we ran a subject-level meta-analysis of 2087 individuals from six cohorts, of whom 100% had a baseline eGFR ≥60 ml/min/1.73 m2, 73.5% had diabetes mellitus and 6.2% had microalbuminuria.  Over five follow-up visits, eGFR decreased more with higher baseline CKD273 than urinary albumin excretion rate (1.64 vs. 0.82 ml/min/m2; p<0.0001).  Over 4.6 years, 390 participants experienced a first renal endpoint (eGFR decrease by ≥10 to <60 ml/min/1.73 m2) and 172 a renal endpoint sustained over follow-up.  The risk of a first and sustained renal endpoint increased with the albumin excretion rate (p≤0.043) and CKD273 (p≤0.031), but only CKD273 (p=0.039) improved discrimination between progressors and non-progressors. 

Mechanistic studies

In Chapter 3, we investigated the underling pathophysiology of diastolic LV dysfunction, HF and CKD through identification of proteins based on proteomic approaches.  In the continuous and categorical cross-sectional analyses reported in Chapter 3.1, diastolic LV dysfunction was associated with higher urinary levels of collagen I fragments and lower levels of urinary collagen III degradation products.  Serum carboxyterminal propeptide of procollagen I (PICP) is a marker of collagen I synthesis.  Carboxyterminal telopeptide of collagen I (CITP) and tissue inhibitor of the matrix metalloproteinase type 1 (TIMP‑1) are markers of collagen 1 degradation.  The serum markers of collagen 1 synthesis and degradation were positively correlated with urinary collagen 1 fragments and negatively with urinary collagen III fragments.  Keeping consistence with the urinary data, diastolic LV performance decreased with higher serum levels of CITP and TIMP‑1. 

In Chapter 3.2, we moved from urinary to tissue proteomics on myocardial biopsies taken from explanted hearts from healthy donors and terminally ill HF patients selected for heart transplantation.  Oxidative phosphorylation and mitochondrial dysfunction were the two top canonical pathways in ischaemic and dilated cardiomyopathy.  In both disease entities, there was greater abundance of proteins involved in the organization of the extracellular matrix or derived from the circulating blood and lower abundance of mitochondrial proteins.  Proteins involved in cell cycle regulation, DNA repair, transcription, calcium fluxes, contractility, signal transduction, the cytoskeleton, protein scaffolding, trafficking and folding and cell migration were either down- or upregulated. 

In Chapter 3.3, we analysed 74 sequenced urinary peptides with a detectable signal in over 95% of participants, whose eGFR was measured at baseline (n=805) or at baseline and follow-up (n=597).  In cross-sectional analyses of the baseline data, eGFR decreased (p≤0.022) with urinary fragments of mucin‑1 (‑4.48 ml/min/1.73 m2), collagen III (‑2.84) and fibrinogen (‑1.70) and was bi-directionally associated (p≤0.0006) with two urinary collagen I fragments (+2.28 and ‑3.20).  eGFR changes over 5 years (follow-up minus baseline) resulted in consistent estimates (p≤0.025) for mucin‑1 (‑1.85), collagen (‑1.37 to 1.43) and fibrinogen (‑1.45) fragments.  Relative risk of having or progressing to CKD stage 3 was positively associated with mucin‑1.  This urinary biomarker predicted eGFR decline to less than 60 ml/min/1.73 m2 over and beyond microalbuminuria (p=0.011).  Overall, our Chapter 3 findings highlight that disturbance of collagen turnover leading to fibrosis and remodelling of the extracellular matrix are common mechanisms associated with LV and renal dysfunction, while a urinary fragment of the mucin‑1 subunit a is a forerunner of renal impairment. 

Metabolomic studies

In Chapter 4, we addressed the metabolic signature in the circulating blood of diastolic LV dysfunction.  We first ran a cross-sectional analysis of the baseline data obtained in 711 FLEMENGHO participants (Chapter 4.1) and next cross-sectional analyses of both the baseline and follow-up data and a longitudinal analysis predicting diastolic LV function at follow-up from metabolic data obtained at baseline in 570 people (Chapter 4.2).  In partial least square analyses combining the cross-sectional data at baseline and follow-up or considering the longitudinal results, serum markers consistently associated with better diastolic LV function included 2‑aminobutyrate, leucine and 4‑hydroxybutyrate and those consistently associated with worse function were glucose + glutamine and valerate.  Branched-chain amino-acid (BCAA) metabolism and aminoacyl tRNA biosynthesis were the top pathways associated with diastolic LV dysfunction. 

Conclusions and perspectives 

This dissertation makes a strong argument for continuing research allowing the introduction of multidimensional urinary classifiers into clinical practice.  HF1 correlated with diastolic LV dysfunction concurrently and over a 5‑year span and predicted the incidence of cardiovascular and cardiac complications.  HF1, HF2 and HFP are urinary biomarkers mainly consisting of dysregulated collagen fragments probably reflecting remodelling of the cardiac extracellular matrix.  CKD273 already being supported by FDA as a tool in CKD research is a diagnostic test foreshadowing progression to CKD stage 3 and is approved for clinical use in Germany.  From a mechanistic point of view, single urinary peptides provided insight in the pathophysiological processes reflecting cardiac and renal fibrosis.  Shedding of the mucin‑1 subunit a, an extracellular protein expressed at the luminal membrane of renal tubular cells, was a predictor of renal impairment and in this role outperformed urinary microalbumin in the FLEMENGHO population sample (Chapter 3.3) with replication in a predominantly diabetic cohort (Chapter 2.5, Table S1).  We showed consistency in the cross-sectional and longitudinal associations of diastolic LV function with circulating metabolic markers.  Additionally to disturbances in the balance between aerobic and anaerobic energy generation, we identified BCAA metabolism and aminoacyl tRNA biosynthesis as top metabolic pathways in early-stage diastolic LV dysfunction.  ”‑Omics” approaches for identification of novel drug targets and for the individualisation of lifestyle measures, notably the only guideline-endorsed intervention for the prevention of diastolic LV dysfunction, are rising on the horizon.  Reaching these goals would definitely pave the way for the clinical application of the biomarkers reported on in this dissertation.