Metabolomics, a promising approach to translational research in cardiology

IJC Metabolic & Endocrine 9 (2015) 31–38

Contents lists available at ScienceDirect

IJC Metabolic & Endocrine journal homepage: http://www.journals.elsevier.com/ijc-metabolic-and-endocrine

Metabolomics, a promising approach to translational research in cardiology Martino Deidda a,⁎,1, Cristina Piras b,1, Pier Paolo Bassareo a, Christian Cadeddu Dessalvi a, Giuseppe Mercuro a a b

Dept. of Medical Sciences “M. Aresu”, University of Cagliari, Asse didattico Medicina, SS Sestu KM 0,700, Monserrato, 09042 Cagliari, Sardinia, Italy Dept. of Biomedical Sciences, University of Cagliari, Cittadella Universitaria, SS Sestu KM 0,700, Monserrato, 09042 Cagliari, Sardinia, Italy

a r t i c l e

i n f o

Article history: Received 20 February 2015 Received in revised form 28 September 2015 Accepted 2 October 2015 Available online 5 October 2015 Keywords: Metabolomics Cardiology Heart failure Coronary artery disease Hypertension Diabetes Obesity Dislypidemia Pediatric cardiology

a b s t r a c t The metabolome is the complete set of metabolites found in a biological cell, tissue, organ or organism, representing the end products of cellular processes. Metabolomics is the systematic study of small-molecule metabolite profiles produced by specific cellular processes. mRNA gene expression data and proteomic analyses do not show the complexity of physiopathological processes that occur in a cell, tissue or organism. Metabolic profiling, in contrast, represents a paradigm shift in medical research from approaches that focus on a limited number of enzymatic reactions or single pathways, with the goal of capturing the complexity of metabolic networks. In this article, we will provide a description of metabolomics in comparison with other, better known “omics” disciplines such as genomics and proteomics. In addition, we will review the current rationale for the implementation of metabolomics in cardiology, its basic methodology and the available data from human studies in this discipline. The topics covered will delineate the importance of being able to use the metabolomic information to understand the mechanisms of diseases from the perspective of systems biology, and as a non-invasive approach to the diagnosis, grading and treatment of cardiovascular diseases. © 2015 The Authors. Published by Elsevier Ireland Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction The “omics” sciences refer to a group of analytical methodologies that aim to achieve the collective characterization and quantification of pools of biological molecules, such as genes, transcripts, proteins and metabolites, which translate into the structure, function and dynamics of cells, tissues or organisms. Genomics can be described as a comprehensive analysis of DNA structure and function [1]. Understanding biological diversity at the whole genome level will provide insight into the origins of individual traits and disease susceptibility. Proteomics involves the systematic study of proteins to provide a comprehensive view of the structure, function and regulation of biological systems [2]. Genomics not only involves the study of single-nucleotide polymorphisms (SNP) and mutations in DNA, but it also includes, through the sub-discipline of transcriptomics, the complete analysis of gene expression in a cell. However, mRNA gene expression data and proteomics analyses do not show the complexity of the physiological and/or pathological processes that occur in a cell, tissue or organism. In contrast, the metabolomic approach represents a paradigm revolution

⁎ Corresponding author at: Department of Medical Sciences “M. Aresu”, University of Cagliari, Asse didattico Medicina, SS Sestu KM 0,700, Monserrato, 09042 Cagliari, Sardinia, Italy. E-mail address: [email protected] (M. Deidda). 1 Equally contributing authors.

in medical research, from a perspective that focuses on single pathways towards a holistic investigation that is able to capture the complexity of entire metabolic networks [3,4]. See Fig. 1 Metabolomics (or metabonomics) is the study of the metabolic profile of small molecules in a biological organism. The metabolome, in turn, is defined as the complete set of metabolites present in a cell, tissue, organ or organism [5] and represents all of the end products of cellular processes [3,4]. Metabolomics provides a functional view of an organism, as determined by the sum of its genes, RNA, proteins, and environmental factors including, for example, nutrition, medications and treatments. It reflects the true functional endpoints of biological events and is extensively used as a “functional-genomic” tool and in systems biology [3,4]. The philosophy underlying functional genomics is that a given “biologic” event should not be viewed as an isolated phenomenon but as a part of a complex network of changes, capable to interfere each other, within an organism; corollary of this affirmation is that in order to understand a biological event in all its complexity is necessary to measure as many entities as possible at any level of organization [3]. Specifically, metabolomics is the study of “… the complete set of metabolites/ low-molecular-weight intermediates, which are context dependent, varying according to the physiology, developmental or pathological state of the cell, tissue, organ or organism…”, while metabonomics has been defined as the “quantitative measurement of the dynamic multiparametric metabolic response of living systems to pathophysiological stimuli or genetic modification”, [6–8]. These two terms are

http://dx.doi.org/10.1016/j.ijcme.2015.10.001 2214-7624/© 2015 The Authors. Published by Elsevier Ireland Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

32

M. Deidda et al. / IJC Metabolic & Endocrine 9 (2015) 31–38

Fig. 1. From gene expression to phenotype. What “omics” tell us and how they interact.

currently used interchangeably and are becoming synonymous (See Fig. 2). In this paper, we prefer to use the term metabolomics to refer to either of the aforementioned meanings. 1.1. General aspects As mentioned above, metabolomics is the measurement of multiple small-molecule metabolites in biological samples, including body fluids (urine, blood, cerebrospinal fluid, saliva, among others), tissues (heart, liver, kidney, brain, among others) and exhaled breath [9]. This technology is able to provide a comprehensive profile of the metabolic state of the whole organism, and it is linked to genetic settings and the influence of endogenous (i.e., hormones) and exogenous (i.e., food, drugs) substances or pathophysiological conditions. These stimuli induce a series of biochemical reactions that generate a wide and complex array of metabolites, many of which will be released in body fluids and tissues. Metabolomic analytical technologies can identify molecules with a relatively low molecular weight (b1000 Da), including amino and fatty acids, nucleic acids, carbohydrates, organic acids, vitamins, polyphenols, lipids, intermediates of many biochemical pathways, and inorganic and elemental species. The analysis of human biological samples provides characteristic numbers of metabolites, depending on the biofluid or tissue under examination and the analytical instrumentation used, and they can reach values in the thousands. The Human Metabolome Project [10], that has the aim to identify, quantify, catalog and record all metabolites potentially found in human specimens, have already

Fig. 2. The increasing number of publications covering the following search terms: Metabolomic* or Metabonomic*.

identified over 40,000 metabolites, stored in the Human Metabolome Database [11]. The metabolomic approach consists of two sequential steps: • The sample is analyzed using an experimental technique that is suitable for providing a detailed molecular fingerprint, i.e., the full complement of low-molecular-weight metabolites; and 2) the data analysis is performed to identify and quantify the correlation between the identified species and the metabolic processes under investigation.

1.2. Experimental techniques Among many different technological platforms, such as infrared and fluorescence spectroscopy and Coulombic arrays, nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry (MS) have been successfully used in the metabolomics field (Table 1), but the most commonly used are certainly the last two [3]. The NMR technique provides a detailed molecular picture of the biological sample. It simultaneously detects, in a non-targeted manner, signals from many different compounds, such as carbohydrates, amino acids, organic and fatty acids, amines, and lipids, without any initial sample pretreatment. In the NMR spectrum (See Fig. 3), each metabolite generates its own specific signals (often referred to as resonances). Each resonance displays a fine structure that can be directly related to the relative position of the different chemical groups in the chemical structure of the metabolite. The resonance intensity is proportional to the number of nuclei under observation generating that specific resonance, and to the molar concentration of the metabolite. Mass spectrometry, which is typically combined with a high-resolution separation technique like gas chromatography (GC) or liquid chromatography (LC), discriminates the molecules present in a biological sample according to their mass-to-charge ratio. Each of the peaks in the resulting chromatogram is associated with the characteristic mass spectrum of the metabolite (See Fig. 4), allowing their identification. NMR and MS have different advantages and limitations, and they are often regarded as complementary techniques. NMR provides a relatively fast but detailed analysis of the molecular composition of the sample with a (usually) rather simple preparation. The experiments are highly reproducible; the technique is non-destructive and represents a universal detector for all of the molecules containing NMR-active nuclei. However, NMR spectroscopy has some major drawbacks, mainly concerning the relatively low sensitivity when compared to other analytical techniques and the high management costs. On the other hand, MS techniques are highly specific and

M. Deidda et al. / IJC Metabolic & Endocrine 9 (2015) 31–38

33

Table 1 Technique of analysis used in Metabolomics (modified from Griffin et al. [3]). Techique

Cost

Throughput

Advantages

Disadvantages

High-resolution NMR spectroscopy

Low per sample

~10 min

Poor sensitivity Co-resonances for 1D-NMR spectroscopy 2D-NMR spectroscopy is time consuming

In vivo NMR spectroscopy

High

High-resolution magic-anglespinning NMR

Low

~30 min with preparation time ~15 min

Direct-infusion MS

Low

3–4 min

Simultaneous detection of many different compounds, such as carbohydrates, amino acids, organic and fatty acids, amines, and lipids without any initial sample pre-treatment Technique non-destructive Good libraries of spectra Easy to process Possibility to observe metabolism of the working heart Imaging can map metabolite distributions Possibility to monitor cellular environment (e. g., compartmentation) in intact tissue Tissue can be chilled to reduce post- mortem effects Has been used to profile both aqueous and lipophilic metabolites in various studies Minimal carry-over, as no chromatography involved Good reproducibility Simple to optimize

GC–MS

Low–medium 20–30 min for fatty acids 30–45 min for aqueous metabolites

LC–MS

Medium

~15–30 min

Triple–quadrupole (targeted) MS

Medium to high,

15 min per chromatography run ~60 min for more comprehensive screens

Chromatography is robust and reproducible Metabolite identification is aided by adoption of common ionization parameters in electron impact and wide availability of the Can be quantitative Chromatography reduces the effect of ion suppression and can separate isobaric species Suitable for measuring intact lipids, dipeptides, tripeptides, and other macromolecules Highly sensitive Highly quantitative Targeted Results readily transferable because concentrations can be measured

sensitive, but they require preprocessing of the sample preparation (i.e., derivatization), which implies the consumption of specimens.

1.3. Data analysis and interpretation In metabolomic studies, different approaches are used for the collection, processing and interpretation of the data depending on the specific problem: metabolic profiling, metabolomics and metabolic fingerprinting. Metabolic profiling seeks to identify and quantify a selected number of pre-defined metabolites in a given sample, whereas metabolomics allows the complete analysis of all of the metabolites in a sample, which are quantified and identified (targeted analyses). Finally, metabolic fingerprinting is aimed at measuring the global profile of the metabolites characterizing the sample, without specific identification and quantification (untargeted analyses). In the metabolomics field, many hundreds of samples are routinely analyzed, and a minimum of several hundreds of metabolites are usually detected. However, the detection of metabolites, in the case of biological

Very poor sensitivity (hyperpolarization or higher field strengths could improve it) Tissue cannot be perfused, so their viability is limited Ion suppression can be a substantial problem Identification can require chromatography, e. g. for isobaric species Metabolite identification is a major challenge and requires MSn acquisitions Semiquantitative at best Metabolites need derivatization and not all metabolites are suitable for derivatization

Chromatography can drift during a sample run, which makes data processing difficult Metabolite identification is a major challenge Targeted, so discovery of novel biomarkers unlikely Time consuming to set up quantitative assays

samples, is not the final step; all the data obtained from the collection stage should be further analyzed using statistical multivariate methods (chemometrics) to extract biological, physiological and clinically relevant information. Before performing the statistical analysis, the data are subjected to several processing steps: 1) the spectrum requires different processing depending on the specific analytical techniques employed; 2) a data matrix is produced from the analytical measurements with m rows (observations, samples) and n columns (variables, frequencies, integrals); 3) data normalization and scaling are performed; and 4) multivariate statistical modeling of the data is completed.

1.4. Statistical multivariate analysis In metabolomic studies, the complexity of the spectroscopic data due to the commonly high number of samples and variables (i.e., metabolites) inevitably requires the use of data reduction techniques to extract latent metabolic information and enable sample classification and biomarker identification. The statistical multivariate methods can be distinguished

Fig. 3. NMR spectra of a HF patient.

34

M. Deidda et al. / IJC Metabolic & Endocrine 9 (2015) 31–38

Fig. 4. MS spectra of alanine.

using unsupervised and supervised approaches. A classic example of an unsupervised method is principal component analysis (PCA), which provides an initial overview of all of the data for the detection of outliers, trends, patterns, or clusters. However, the supervised approaches include the partial least squares discriminant analysis (PLS-DA) or orthogonal partial least squares discriminant analysis (OPLS-DA), for instance, which classify samples into different categories or classes based on a priori knowledge. This a priori information, e.g., the object membership, constitutes an additional data table, Y, which describes different sample properties that can be either discrete or continuous to be used together with the experimental data table, X, within the multivariate analysis. The N-dimensional results are typically presented as easier to read twodimensional (2D) projections. One of these is referred to as the scores plot, which reveals any inherent clustering of data, trends or outliers. The other is referred to as the loadings plot, which shows the importance (weight) of the different variables (i.e., the metabolites) in defining the model. 1.5. Metabolomics in cardiology During the last decade, both animal and human studies have facilitated the rapid development of metabolomics, and a combination of targeted and untargeted approaches has been applied for cardiovascular research [12]. Specific metabolic profiles have been identified for several cardiovascular risk factors and diseases, such as lipid disorders, diabetes, hypertension, myocardial ischemia and heart failure [13–19]. Selected human studies are discussed in the following sections. 1.5.1. Metabolic disorders related to cardiovascular risk: obesity, dyslipidemia and diabetes Glucose homeostasis is a complex physiologic process involving multiple organ systems and is strictly related to lipid metabolism/storage. Alterations in glucose and lipid metabolism interfere each other and have important effects on body composition and all these are able to modify the CV risk profile. The spectrum of biochemical changes associated with the transition from normal glucose homeostasis to insulin resistance and diabetes is largely unknown. An approach based on MS to simultaneously measure 191 metabolites has been used to study the insulin-dependent homeostatic program triggered by the ingestion of glucose [20]. In healthy subjects, the dynamics of the metabolites revealed the actions of insulin along four main axes - proteolysis, lipolysis, ketogenesis and glycolysis reflecting the transition from catabolism to anabolism. In pre-diabetics, a blunt response, which correlated with insulin resistance in all four axes, was observed. In details, a decline in glycerol and leucine/iso-leucine (markers of lipolysis and proteolysis, respectively) was found and seems to be the strongest predictor of insulin sensitivity. Authors

concluded that different axes could be responsible for the same elevation of fasting insulin and that exist two forms of insulin resistance: in one form, subjects are selectively resistant to insulin's suppression of proteolysis whereas in the other to insulin's suppression of lipolysis. On the other hand, genetic, environmental and lifestyle factors are involved in the etiology of glucose metabolism alterations such as of obesity. Obesity, for its part, is capable to increase the risk of cardiovascular diseases and diabetes. Pietilainen and coll. carried out a metabolomics investigation on young adult monozygotic twin pairs, discordant or concordant for obesity, by means of liquid chromatography coupled to mass spectrometry. The earliest stage of acquired obesity, independent of genetic influences, was found to be primarily associated with an increase in lysophosphatidylcholines, lipids found in proinflammatory and pro-atherogenic conditions, and with a decrease in ether phospholipids, which are known to have antioxidant properties. Moreover, these lipid changes were associated with insulin resistance [21]. Changes in lipid metabolism affect CV risk and several clinical trials [22–24] have demonstrated the beneficial effects of statin treatment in primary and secondary prevention of CV diseases. A lipidomic analyses of variation in response to simvastatin conducted on plasma samples of 944 subjects selected form the cholesterol and pharmacogenetics study revealed a correlation of baseline cholesterol esters and phospholipid metabolites with the response to treatment of low-density-lipoprotein cholesterol (LDL) [25]. In addition, the response of C-reactive protein (CRP) to therapy correlated with basal plasmalogens, which are lipids that are involved in inflammation. This modulation appeared to be independent of the variation in plasma lipid levels, which correlated with the LDL-C and CRP responses, thus suggesting that distinct metabolic pathways regulate the effects of statins on these two biomarkers. Metabolomics has also been applied to the evaluation of metabolic changes induced by rosiglitazone, which is currently under investigation due to its adverse effects on cardiovascular outcomes, in type 2 diabetic patients with coronary artery disease (CAD). Badeau and coll. showed that short-term treatment with rosiglitazone resulted in minor improvements in metabolism: serum lactate and glutamine concentrations changed, reflecting improvements in insulin sensitivity, and circulating lactate concentrations proved to be inversely correlated with the increase in myocardial glucose uptake [26]. More recently, Hung and coll. used a metabolomics approach in order to evaluate whether plasma metabolic profile during the oral glucose tolerance test (OGTT) could improve the diagnosis of CAD in patients with positive non-invasive stress tests. The study was conducted on 117 subjects with positive stress test results who received coronary angiography; plasma samples obtained in fasting, 30 min, and 120 min after OGTT were analyzed using liquid chromatography combined with mass spectrometry. Results showed that the fasting plasma metabolome, but not its changes during the oral glucose tolerance test, could improve the diagnosis of CAD in this setting of patients, regardless of the clinical risk factors [27]. Noteworthy, authors suggested that tryptophan metabolism, particularly about indolelactic acid and glutaric acid, may have a causal relationship with the development of atherosclerosis. 1.5.2. Hypertension Hypertension is a major risk factor for CV disease, stroke, and renal failure and its pathogenesis is significantly influenced by genetic, lifestyle, and environmental factors; however, the underlying pathways remain still to be completely clarified. On these bases, a GC_MS-based metabolomics study was recently conducted using plasma samples from 20 young hypertensive men and 20 age-matched healthy controls. The authors discovered a distinct metabolic fingerprint of young patients that was strongly related to amino acid biosynthesis. The correlations of GC-MS metabolomics data were visualized as networks based on Pearson's correlation coefficient (threshold = 0.6), and the variation of several amino acids – glycine, lysine and cysteine –,

M. Deidda et al. / IJC Metabolic & Endocrine 9 (2015) 31–38

screened as hub metabolites, seemed to have a major impact on the metabolic change in young hypertensive men, thus suggesting that disorders in amino acid metabolism could play an important role in the development of hypertension [28]. In line with these data seems to be the results of an NMR study conducted between 2009 and 2012 on 157 hypertensive patients and 99 healthy subjects, all of whom were Uygur people, one of the 55 ethnic minorities of the People's Republic of China. The recruitment criteria were systolic blood pressure (SBP) ≥ 140 mm Hg and/or diastolic blood pressure (DBP) ≥90 mm Hg at rest; patients with secondary hypertension and previous cardiovascular diseases were excluded from the study. NMR data/metabolic information were analyzed with PCA, PLS-DA and OPLS-DA. The OPLS-DA results indicated that the metabolic fingerprints were significantly different between the two groups, and 12 different metabolites were identified. In comparison with healthy subjects, hypertensive patients showed lower levels of several amino acids, including valine, alanine, pyroracemic acid, inose, p-hydroxyphenylalanine, and methylhistidine, with a significant increase in VLDL, LDL-C, lactic acid, and acetone. However, a possible limitation of this study could be that, although clinical data did not show significant differences between the two groups with respect to age, weight, height and some lipid indices including triglycerides and LDL-C, fasting plasma glucose (FPG), SBP and DBP, HDL and total cholesterol (TC) were significantly different [29]; basal differences in FPG, HDL and TC levels could have been related per se to metabolic alterations, leading to a potential misinterpretation of the observed clustering and/or of the metabolic fingerprint. Another study targeted to investigate the molecular basis of hypertension was conducted in 2013 by Zheng and coll. They designed a study aimed to evaluate human serum metabolome before the onset of hypertension in a well-defined prospective cohort setting. Metabolomic analysis was performed on serum of 896 black normotensive people from the Atherosclerosis Risk in Communities Study (565 women aged 45–64 years), the blood samples of whom were collected at the enrollment examination. During a 10-year follow-up, 38% of the baseline normotensive subjects developed hypertension (n = 344). After adjustment for traditional risk factors and the estimated glomerular filtration rate, each + 1SD difference in baseline 4-hydroxyhippurate, which is a product of gut microbial fermentation, was associated with a 17% increased risk of hypertension. This correlation remained significant after adjusting for both baseline SBP and DBP. Moreover, the PCA revealed a sex steroid pattern associated with the risk of incident hypertension that was consistent in both genders after a stratified analysis [30]. Metabolomic analysis was also applied in order to verify whether it was possible to distinguish between different BP categories based on the metabolic profile. Sixty-four individuals were classified into three BP categories [low/normal (≤130 mmHg), borderline (131–149 mm Hg) and high systolic BP (SBP; N150 mm Hg)] and studied. After an orthogonal signal correction of the NMR data, the PCA scores revealed several similarities in serum metabolic profiles between patients in the borderline and high SBP categories, whereas regions of the NMR spectra attributed to different lipids [in particular, lipoproteins associated with very-low-density cholesterol (VLDL) and LDL-C] clearly distinguished the borderline/high BP categories from the low/normal BP category (higher lipid level were detected in the higher BP categories). However a limitation of this study was that the enrollment and classification of the participants were based on a single BP measurement taken in a hospital environment, and thus, there was the potential for misclassification of the true BP values [31].

1.5.3. CAD CAD is one of the prominent causes of morbidity and mortality worldwide and identification of myocardial ischemia is crucial both for diagnostic aims and for the evaluation of the response to therapeutic interventions. For these reasons, most of the metabolomic studies

35

carried out on CAD were addressed to identify biomarkers of myocardial ischemia or to discover possible new diagnostic tools. In a pioneering clinical application of metabolomics, Brindle and coll. [32] used the pattern-recognition techniques applied to NMR spectra of human sera to diagnose not only the presence but also the severity of CAD. The authors argued that in their analysis, N 90% of the patients with stenosis of all three main coronary vessels were distinct from subjects with angiographically normal coronary arteries, with a specificity of N 90%. The difference between the sample groups, which was mainly due to subtle changes in lipoprotein composition, supported the potential of an NMR-based metabolomics approach in predicting the various stages of CAD. However, Kirschenlohr and coll. later identified a number of confounding factors for a diagnosis primarily based on the lipid composition; in particular, the use of statins may have biased the results of the study reported by Brindle [33]. In another study, blood samples were obtained before and after exercise stress testing from 36 patients, 18 of whom showed inducible ischemia (cases) and 18 who did not (controls) [34]. The variations in 6 metabolites, including citric acid, demonstrated a significant discordant regulation between the 2 groups of individuals, differentiating the cases from controls with a high degree of precision. In 2008 Barba conducted a similar study on 31 patients with suspected effort angina without previous myocardial infarction that underwent a stress single-photon emission computed tomography. NMR-based metabolomics were able to predict exercise-induced ischemia and the main contributors to discrimination were lactate and glucose, as well as methyl and methylene moieties of lipids and longchain amino acids [35]. Progressively, metabolomics was employed in studies with design of increasing complexity, in order to achieve better-controlled experimental conditions and, then, more accurate results. In 2008 Lewis examined metabolic profiles of individuals undergoing planned or spontaneous myocardial ischemia to identify early biomarkers of cardiac damage. Thirty-six patients undergoing alcohol septal ablation treatment for hypertrophic obstructive cardiomyopathy, a human model of planned myocardial infarction (PMI), were examined and compared with subjects undergoing elective diagnostic coronary angiography and patients with spontaneous myocardial infarction, that served as negative and positive controls, respectively; moreover, coronary sinus sampling was used to discriminate cardiac-derived from peripheral metabolic changes. A PMI-derived metabolic footprint, consisting of aconitic acid, hypoxanthine, trimethylamine N-oxide, and threonine, resulted able to distinguish PMI patients from those undergoing diagnostic coronary angiography with high accuracy and abnormalities in circulating metabolites were detected b 10 min after myocardial injury. It is note of worthy that observed metabolite changes are able to modulate the response to hypoxic injury in vitro, suggesting a possible role of these metabolites in evolving response to ischemia in vivo [36]. Another metabolic footprint of acute myocardial ischemia (MIS) was identified by Bodi and coll. in a study of 2014, fundamental because the highly controlled conditions of its design. Investigators performed an NMR analysis of peripheral blood serum obtained from swine and patients undergoing angioplasty balloon-induced transient coronary occlusion [37]. MIS induced 1) an increase in circulating glucose, lactate, glutamine, glycine, glycerol, phenylalanine, tyrosine, and phosphoethanolamine; 2) a decrease in cholinecontaining compounds and triacylglycerols; and 3) a change in the pattern of total, esterified, and nonesterified fatty acids. Moreover, the authors performed a preliminary clinical validation of the developed metabolomic biosignature in patients with spontaneous acute chest pain, achieving a negative predictive value of 95%. To evaluate the discriminative capabilities of metabolites for CAD, an MS-based profiling of 69 metabolites in subjects from the CATHGEN biorepository was performed by Shah and coll. Two independent case/ control groups were constructed and profiled; moreover, cases of both

36

M. Deidda et al. / IJC Metabolic & Endocrine 9 (2015) 31–38

groups were combined in order to evaluate the capability of identified metabolic fingerprint to predict cardiovascular events. Finally, a third independent group was profiled as “event-replication” group. Two principal component analysis-derived factors resulted associated with CAD: the first comprised branched-chain amino acid metabolites, and the second consisted of urea cycle metabolites. Furthermore, a factor composed of dicarboxylacylcarnitines was able to predict death/myocardial infarction and resulted associated with CV events in the third group [38]. Two years later, the same Shah published another study based on MS profiling of 69 metabolites of 2.023 consecutive patients undergoing cardiac catheterization. Statistical analysis of metabolomics data was conducted using PCA, whereas independent relationships between factors and time-to-clinical events were assessed using Cox modeling. Subsequently, clinical and metabolomic models were compared. Five of 13 metabolite factors were independently associated with mortality (medium-chain acylcarnitines, short-chain dicarboxylacylcarnitine, long-chain dicarboxylacylcarnitines, branched-chain amino acids and fatty acids), and three were associated with death/MI (factors 2, 3 and 12). In detail, 27% and 11% of the intermediate-risk patients were correctly reclassified in terms of mortality and death/MI, respectively [39]. More recently, Rizza and coll., with the aim to create an integrated clinical-metabolomic model able to improve the prediction of incident CV events in aged people, performed a targeted MS-based profiling of 49 metabolites in a group of very elderly patients (n = 67, mean age of 85 ± 3 years) with previous cardiovascular diseases and showed that mitochondrial dysfunction detected by metabolomic analysis was associated with MACEs after adjustment for clinical CV covariates. The two metabolites that were independently associated with cardiac complications were medium/long-chain acylcarnitines and alanine. Moreover, when compared with the Framingham Recurring-CoronaryHeart-Disease-Score, medium/long-chain acylcarnitines showed to determine a significant improvement in discrimination with a cNRI (Net Reclassification Improvement) = 0.79 [40]. Despite establishment of secondary prevention strategies and the advent of increasing technology coronary stents, restenosis after stent implantation remains a major topic in cardiology. With this background, Hasokawa and coll. conducted in 2012 a study on 86 patients who were hospitalized for a 6-month follow-up coronary angiography after stent implantation (23 with major restenotic lesions, 47 with minor restenotic lesions and 16 with de novo atherosclerotic lesions), identifying a cluster of 8 metabolites - isobutylamine, sarcosine, homoserine, ribulose, taurine, glutamine, glucose, and tryptophan - that were significantly different in patients with major restenosis in comparison with those with minor restenosis [41]. Interesting findings about ischemia-induced metabolic derangements derive from a study of Turer and coll. in the setting of cardiac surgery. An MS-based platform was applied to profile 63 intermediate metabolites in blood obtained from serial paired peripheral arterial and coronary sinus of 37 patients undergoing cardiac surgery. After reperfusion, metabolomics profiling revealed distinct patterns of myocardial substrate utilization, with significant differences between patients with and without CAD. In particular, significantly lower extraction ratios of most substrates, as well as the significant release of 2 specific acylcarnitine species; moreover, these alterations were more evident in patients with impaired ventricular function, suggesting limited myocardial metabolic reserve and flexibility in this group of patients. Over the physiopathologic issues, authors highlighted that alterations in metabolic profiles after ischemia/reperfusion are associated with postoperative hemodynamic course and could have a role in role for optimization of perioperative care of cardiac surgical patients [42]. 1.5.4. Arrhythmias Atrial fibrillation (AF) is the most common arrhythmia in clinical practice. Despite similar CV risk factors, different proportion of patients with paroxysmal AF develops persistent or permanent AF.

Combined metabolomic and proteomic analyses were used to evaluate whether metabolic derangements might “modulate” the progression of AF. The contribution of electrical, structural, and contractile remodeling processes to the self-perpetuating nature of this arrhythmia is now well known. However, the role of metabolic changes in the myocardium has only been partially investigated. High-resolution proton NMR spectroscopy was used to analyze human atrial appendage tissues from matched cohorts of patients undergoing cardiac surgery in sinus rhythm, in persistent atrial fibrillation o who developed AF postoperatively [43]. The persistent AF patients presented a rise in betahydroxybutyrate, the major substrate in ketone body metabolism, along with an increase in ketogenic amino acids and glycine. These metabolomic findings were substantiated by proteomic experiments demonstrating the differential expression of 3-oxoacid transferase, the key enzyme in ketolytic energy production. Noteworthy, patients susceptible to post-operative AF showed a deregulation of energy metabolites in comparison with patients in sinus rhythm and may be correctly identified using metabolic profile in more than 80% of cases. 1.5.5. Pediatric cardiology A small minority of metabolomic investigations have been devoted to cardiovascular risk factors and diseases in pediatric patients [44]. However, an unfavorable perinatal environment (maternal diseases such as diabetes, placental insufficiency, malnutrition, among others) may adversely affect the future evolution and function of the cardiovascular system, although adverse events may not occur until adulthood [45]. In addition, cardiac and renals function appear to be increasingly interdependent, and renal dysfunction in preterm or underweight infants can predict future CV events [46,47]. Metabolomic analysis of urine samples was performed in 40 children who underwent cardiac surgery for correction of congenital cardiac defects [48]. Homovanillic acid sulfate, a dopamine metabolite, was able to discriminate patients who developed acute kidney injury from those who did not and was proposed as a new, sensitive, and predictive early biomarker of kidney injury. Moreover, a metabolic signature for patent ductus arteriosus in preterm infants was identified using NMR-based metabolomic analysis of urine samples [49]. Recently, our group published a study of young adults who were born preterm with extremely low birth weights (b1000 g; ex-ELBW; n = 19) compared with a control group of subjects who were born at term with a weight appropriate for their gestational age (AGA; n = 13) to evaluate the presence of differences in the urinary metabolic profile and in the hematic asymmetric dimethylarginine (ADMA) level. To achieve this goal, urine samples were analyzed by NMR spectroscopy, and then submitted to an unsupervised and a supervised multivariate analysis. Blood samples were collected, and the ADMA concentration was assessed with high-performance liquid chromatography. Using a PLS-DA model, it was possible to discriminate between ex-ELBW and AGA. Statistically significant differences in ADMA levels were detected between ex-ELBW and AGA. Moreover, the metabolic profile correlated with the ADMA concentrations in ex-ELBW but not in AGA, thus suggesting the presence of a subclinical cardio-renal impairment in ex-ELBW subjects [47]. 1.5.6. Heart failure Heart failure (HF) is characterized by metabolic remodeling and an increasing number of studies highlights the importance of the impairment in cardiac energy metabolism in the progression of HF [50]. During this metabolic remodeling, changes in substrate utilization, oxidative phosphorylation, and high-energy phosphate metabolism occur. Although metabolomics has had a substantial impact on the investigation of various cardiovascular diseases [3,4], recent published studies on its application in HF has produced conflicting results. The first study addressed to apply metabolimics to heart failure has been published in 2011 by Kang and coll. who carried out an NMRbased profiling on urine of 15 HF patients and 20 healthy subjects. The

M. Deidda et al. / IJC Metabolic & Endocrine 9 (2015) 31–38

results, obtained with the application of an OPLS-DA, showed higher urinary levels of acetate, acetone, cytosine, methylmalonate and phenylacetylglycine and decreased of 1-methylnicotinamide in HF patients compared to the healthy controls. These findings suggest that Krebs cycle metabolites and fatty acid metabolism were modified during HF, confirming an altered energy metabolism in these patients [51]. In 2013, Tenori and coll. conducted an NMR-based study considering, in addition to the presence of HF, its relationship with BNP and the NYHA classes. To this aim they enrolled 185 patients with stable chronic HF and under a state-of-the-art therapy, whereas 111 age and sex matched healthy subjects constituted the control group. After applying an OPLS-DA, they concluded that the metabolism in HF patients shows a disease-related profile in comparison with healthy subjects (higher serum concentrations of phenylalanine, tyrosine, isoleucine, and creatine, and lower levels of serum lactate, citrate, lysine and L-dopa), behaving like an on/off phenomenon regardless of the severity of the disease [52]. In contrast, Du and coll. found that the serum metabolite profiles of patients with different degrees of systolic dysfunction were associated with a progressive impairment of myocardial energy expenditure (MEE), a parameter that is related to the left ventricular ejection fraction (LVEF). Accordingly, based on the MEE levels, the study population was classified into three groups with progressive disease severity. Each group was characterized by a specific metabolic fingerprint, which was obtained after the Orthogonal Signal Correction and PLS regression analysis. The Variable Important in Prediction (VIP) related to increased levels of MEE were 3-hydroxybutyrate, acetone and succinate [53]. Unfortunately, the MEE as well as all of the parameters derived from regional and global LV functions were not very sensitive because they were highly dependent on loading and likely to produce abnormal results if they were obtained under unusual load conditions as in HF [54]. On the other hand, the group of Padeletti carried out a prospective study in which were collected blood and urine samples from 32 HF patients who underwent cardiac resynchronization therapy (CRT) and from 39 age and sex-matched healthy subjects. Clinical, electrocardiography and echocardiographic evaluations were performed in each patient before CRT and after 6 months of follow-up (6FU). The metabolomic analysis demonstrated a different metabolic fingerprint in HF patients compared to healthy controls, both at enrollment and at 6FU. It is noteworthy that the 6FU metabolic fingerprint was different from the baseline fingerprint, independent of the different causes of the disease, but was not predictive of the response to CRT [55]. More recently, a key study aimed to assess the diagnostic and prognostic values of metabolomics in HF was published by Cheng and coll.; they performed a mass spectrometry-based profiling of plasma metabolites in over 400 HF patients and 114 control subjects, divided in two groups, the first used in order to build a metabolomic model of the disease and identify the relative metabolic fingerprint, the second for the validation of the model. Moreover a third, independent group of Class C HF patients who showed a recovery to the I NHYA class at 6 and 12 months follow-up was enrolled and analyzed. A fingerprint constituted by histidine, phenylalanine, spermidine, and phosphatidylcholine C34:4, showed to have a diagnostic value comparable to BNP. Furthermore, when prognostic value was evaluated using the combined endpoints of death or HF-related re-hospitalization, a metabolite panel, which included asymmetric methylarginine/arginine ratio, butyrylcarnitine, spermidine, and the total amount of essential amino acids, was identified and its prognostic values resulted independent of BNP and traditional risk factors and was better than that of BNP [56]. Our preliminary data seem to confirm that metabolomics is able to differentiate HF patients with different degrees of systolic dysfunction using the Speckle Tracking-derived longitudinal Strain Rate as the Y variable in a multivariate analysis. The PLS-DA showed a clear clustering of three a priori classes (LVEF b35%, LVEF N35% and b50%, healthy subjects), with a progressive modification of the metabolic fingerprint of healthy subjects towards HF patients with severe impairment of systolic

37

function; the most important metabolites implicated in the clustering were 2-hydroxybutyrate, glycine, methylmalonate, and myo-inositol [57].

2. Conclusions Cardiovascular diseases are the leading cause of death worldwide. However, their molecular etiology and pathophysiological progression remain incompletely understood, probably because they result from both genetic and environmental factors; for this reason, the system biology approach, that consider each biologic event as a part of a complex network of interactions, could be the way to dramatically improve our knowledge of mechanism that underlie the development of disease status. In fact, one of the challenges of translational research is exactly to integrate, in a holistic approach, clinical data with proteomic, transcriptomic and metabolomic information. In this regard, the sensitivity shown by metabolomics in the detection of molecular changes resulting from physical-pathological processes “proximal” to the disease could represent a great opportunity. The success of this approach will depend on how the different data will be integrated, as exemplified in presented studies. Finally, the association of metabolomics – and all the “omics” sciences – with clinical and instrumental data could improve our understanding of the molecular basis of cardiovascular diseases, prediction of therapeutic responses and risk of adverse effects during treatment, and final implementation of new tools towards the goal of more personalized medicine.

References [1] W.S. Klug, M.R. Cummings, C.A. Spencer, M.A. Palladino, Concepts of genetics Pearson Higher Ed, 2011. [2] N.L. Anderson, N.G. Anderson, Proteome and proteomics: new technologies, new concepts, and new words, Electrophoresis 19 (1998) 1853–1861. [3] J.L. Griffin, H. Atherton, J. Shockcor, L. Atzori, Metabolomics as a tool for cardiac research, Nat. Rev. Cardiol. 8 (2011) 630–643. [4] G. Mercuro, P.P. Bassareo, M. Deidda, C. Cadeddu, L. Barberini, et al., Metabolomics: a new era in cardiology? J. Cardiovasc. Med. (Hagerstown) 12 (2011) 800–805. [5] S.G. Oliver, M.K. Winson, D.B. Kell, F. Baganz, Systematic functional analysis of the yeast genome, Trends Biotechnol. 16 (1998) 373–378. [6] J.K. Nicholson, J.C. Lindon, E. Holmes, ‘Metabonomics’: understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data, Xenobiotica 29 (1999) 1181–1189. [7] J.K. Nicholson, J.C. Lindon, Systems biology: metabonomics, Nature 455 (2008) 1054–1056. [8] S.G. Oliver, Functional genomics: lessons from yeast, Philos. Trans. R. Soc. Lond. Ser. B Biol. Sci. 357 (2002) 17–23. [9] M. Coen, E. Holmes, J.C. Lindon, J.K. Nicholson, NMR-based metabolic profiling and metabonomic approaches to problems in molecular toxicology, Chem. Res. Toxicol. 21 (2008) 9–27. [10] D.S. Wishart, C. Knox, A.C. Guo, R. Eisner, N. Young, et al., HMDB: a knowledgebase for the human metabolome, Nucleic Acids Res. 37 (2009) D603–D610. [11] D.S. Wishart, T. Jewison, A.C. Guo, M. Wilson, C. Knox, et al., HMDB 3.0–The Human Metabolome Database in 2013, Nucleic Acids Res. 41 (2013) D801–D807. [12] S.J. Bruce, I. Tavazzi, V. Parisod, S. Rezzi, S. Kochhar, et al., Investigation of human blood plasma sample preparation for performing metabolomics using ultrahigh performance liquid chromatography/mass spectrometry, Anal. Chem. 81 (2009) 3285–3296. [13] J.P. Cooke, A.M. Wilson, Biomarkers of peripheral arterial disease, J. Am. Coll. Cardiol. 55 (2010) 2017–2023. [14] W.B. Dunn, D.I. Broadhurst, S.M. Deepak, M.H. Buch, G. McDowell, et al., Serum metabolomics reveals many novel metabolic markers of heart failure, including pseudouridine and 2-oxoglutarate, Metabolomics 3 (2007) 413–426. [15] H. Lin, J. Zhang, P. Gao, Silent myocardial ischemia is associated with altered plasma phospholipids, J. Clin. Lab. Anal. 23 (2009) 45–50. [16] E. Holmes, R.L. Loo, J. Stamler, M. Bictash, I.K. Yap, et al., Human metabolic phenotype diversity and its association with diet and blood pressure, Nature 453 (2008) 396–400. [17] M.S. Hodavance, S.L. Ralston, I. Pelczer, Beyond blood sugar: the potential of NMRbased metabonomics for type 2 human diabetes, and the horse as a possible model, Anal. Bioanal. Chem. 387 (2007) 533–537. [18] V.P. Mäkinen, P. Soininen, C. Forsblom, M. Parkkonen, P. Ingman, et al., 1 H NMR metabonomics approach to the disease continuum of diabetic complications and premature death, Mol. Syst. Biol. 4 (2008) 167. [19] A.D. Watson, Thematic review series: systems biology approaches to metabolic and cardiovascular disorders. Lipidomics: a global approach to lipid analysis in biological systems, J. Lipid Res. 47 (2006) 2101–2111.

38

M. Deidda et al. / IJC Metabolic & Endocrine 9 (2015) 31–38

[20] O. Shaham, R. Wei, T.J. Wang, C. Ricciardi, G.D. Lewis, et al., Metabolic profiling of the human response to a glucose challenge reveals distinct axes of insulin sensitivity, Mol. Syst. Biol. 4 (2008) 214. [21] K.H. Pietiläinen, M. Sysi-Aho, A. Rissanen, T. Seppänen-Laakso, H. Yki-Järvinen, et al., Acquired obesity is associated with changes in the serum lipidomic profile independent of genetic effects–a monozygotic twin study, PLoS ONE 2 (2007), e218. [22] Z. Zhou, E. Rahme, L. Pilote, Are statins created equal? Evidence from randomized trials of pravastatin, simvastatin, and atorvastatin for cardiovascular disease prevention, Am. Heart J. 151 (2006) 273–281. [23] P.M. Ridker, E. Danielson, F.A. Fonseca, J. Genest, A.M. Gotto, et al., Rosuvastatin to prevent vascular events in men and women with elevated C-reactive protein, N. Engl. J. Med. 359 (2008) 2195–2207. [24] C. Baigent, A. Keech, P.M. Kearney, L. Blackwell, G. Buck, et al., Efficacy and safety of cholesterol-lowering treatment: prospective meta-analysis of data from 90,056 participants in 14 randomised trials of statins, Lancet 366 (2005) 1267–1278. [25] R. Kaddurah-Daouk, R.A. Baillie, H. Zhu, Z.B. Zeng, M.M. Wiest, et al., Lipidomic analysis of variation in response to simvastatin in the cholesterol and pharmacogenetics study, Metabolomics 6 (2010) 191–201. [26] R.M. Badeau, M.J. Honka, R. Lautamäki, M. Stewart, A.J. Kangas, et al., Systemic metabolic markers and myocardial glucose uptake in type 2 diabetic and coronary artery disease patients treated for 16 weeks with rosiglitazone, a PPARγ agonist, Ann. Med. 46 (2014) 18–23. [27] C.S. Hung, H.Y. Li, C.H. Kuo, M.S. Lin, T.C. Kuo, et al., Fasting but not changes of plasma metabolome during oral glucose tolerance tests improve the diagnosis of severe coronary arterial stenosis, Clin. Endocrinol. (2015). [28] L. Wang, E. Hou, L. Wang, Y. Wang, L. Yang, et al., Reconstruction and analysis of correlation networks based on GC-MS metabolomics data for young hypertensive men, Anal. Chim. Acta 854 (2015) 95–105. [29] L. Zhong, J.P. Zhang, A.G. Nuermaimaiti, K.X. Yunusi, Study on plasmatic metabolomics of Uygur patients with essential hypertension based on nuclear magnetic resonance technique, Eur. Rev. Med. Pharmacol. Sci. 18 (2014) 3673–3680. [30] Y. Zheng, B. Yu, D. Alexander, T.H. Mosley, G. Heiss, et al., Metabolomics and incident hypertension among blacks: the atherosclerosis risk in communities study, Hypertension 62 (2013) 398–403. [31] J.T. Brindle, J.K. Nicholson, P.M. Schofield, D.J. Grainger, E. Holmes, Application of chemometrics to 1 H NMR spectroscopic data to investigate a relationship between human serum metabolic profiles and hypertension, Analyst 128 (2003) 32–36. [32] J.T. Brindle, H. Antti, E. Holmes, G. Tranter, J.K. Nicholson, et al., Rapid and noninvasive diagnosis of the presence and severity of coronary heart disease using 1 H-NMR-based metabonomics, Nat. Med. 8 (2002) 1439–1444. [33] H.L. Kirschenlohr, J.L. Griffin, S.C. Clarke, R. Rhydwen, A.A. Grace, et al., Proton NMR analysis of plasma is a weak predictor of coronary artery disease, Nat. Med. 12 (2006) 705–710. [34] M.S. Sabatine, E. Liu, D.A. Morrow, E. Heller, R. McCarroll, et al., Metabolomic identification of novel biomarkers of myocardial ischemia, Circulation 112 (2005) 3868–3875. [35] I. Barba, G. de León, E. Martín, A. Cuevas, S. Aguade, et al., Nuclear magnetic resonance-based metabolomics predicts exercise-induced ischemia in patients with suspected coronary artery disease, Magn. Reson. Med. 60 (2008) 27–32. [36] G.D. Lewis, R. Wei, E. Liu, E. Yang, X. Shi, et al., Metabolite profiling of blood from individuals undergoing planned myocardial infarction reveals early markers of myocardial injury, J. Clin. Invest. 118 (2008) 3503–3512. [37] V. Bodi, J. Sanchis, J.M. Morales, V.G. Marrachelli, J. Nunez, et al., Metabolomic profile of human myocardial ischemia by nuclear magnetic resonance spectroscopy of peripheral blood serum: a translational study based on transient coronary occlusion models, J. Am. Coll. Cardiol. 59 (2012) 1629–1641. [38] S.H. Shah, J.R. Bain, M.J. Muehlbauer, R.D. Stevens, D.R. Crosslin, et al., Association of a peripheral blood metabolic profile with coronary artery disease and risk of subsequent cardiovascular events, Circ. Cardiovasc. Genet. 3 (2010) 207–214.

[39] S.H. Shah, J.L. Sun, R.D. Stevens, J.R. Bain, M.J. Muehlbauer, et al., Baseline metabolomic profiles predict cardiovascular events in patients at risk for coronary artery disease, Am. Heart J. 163 (2012) 844–850 e1. [40] S. Rizza, M. Copetti, C. Rossi, M.A. Cianfarani, M. Zucchelli, et al., Metabolomics signature improves the prediction of cardiovascular events in elderly subjects, Atherosclerosis 232 (2014) 260–264. [41] M. Hasokawa, M. Shinohara, H. Tsugawa, T. Bamba, E. Fukusaki, et al., Identification of biomarkers of stent restenosis with serum metabolomic profiling using gas chromatography/mass spectrometry, Circ. J. 76 (2012) 1864–1873. [42] A.T. Turer, R.D. Stevens, J.R. Bain, M.J. Muehlbauer, J. van der Westhuizen, et al., Metabolomic profiling reveals distinct patterns of myocardial substrate use in humans with coronary artery disease or left ventricular dysfunction during surgical ischemia/reperfusion, Circulation 119 (2009) 1736–1746. [43] M. Mayr, S. Yusuf, G. Weir, Y.L. Chung, U. Mayr, et al., Combined metabolomic and proteomic analysis of human atrial fibrillation, J. Am. Coll. Cardiol. 51 (2008) 585–594. [44] P.P. Bassareo, V. Fanos, M. Deidda, L. Barberini, G. Mercuro, Metabolomic approach to foetal and neonatal heart, J. Matern. Fetal Neonatal Med. 25 (2012) 19–21. [45] G. Mercuro, P.P. Bassareo, G. Flore, V. Fanos, I. Dentamaro, et al., Prematurity and low weight at birth as new conditions predisposing to an increased cardiovascular risk, Eur. J. Prev. Cardiol. 20 (2013) 357–367. [46] P.P. Bassareo, V. Fanos, M. Mussap, G. Flore, A. Noto, et al., Urinary NGAL and hematic ADMA levels: an early sign of cardio-renal syndrome in young adults born preterm? J. Matern. Fetal Neonatal Med. 26 (Suppl 2) (2013) 80–83. [47] P.P. Bassareo, V. Fanos, A. Noto, P. Solla, L. Barberini, et al., Clinical metabolomics and hematic ADMA predict the future onset of cardiorenal syndrome in young grownup subjects who were born preterm, Clin. Biochem. 47 (2014) 423–426. [48] R.D. Beger, R.D. Holland, J. Sun, L.K. Schnackenberg, P.C. Moore, et al., Metabonomics of acute kidney injury in children after cardiac surgery, Pediatr. Nephrol. 23 (2008) 977–984. [49] L. Atzori, L. Barberini, M. Lussu, F. Murgia, A. Noto, et al., Metabolomics & patent ductus arteriosus diagnosis: is 1 H-NMR (nuclear magnetic resonance) spectroscopy of urine at birth predictive as ultrasound, J. Matern. Fetal Neonatal Med. 24 (2011). [50] S. Neubauer, The failing heart—an engine out of fuel, N. Engl. J. Med. 356 (2007) 1140–1151. [51] S.M. Kang, J.C. Park, M.J. Shin, H. Lee, J. Oh, et al., 1H nuclear magnetic resonance based metabolic urinary profiling of patients with ischemic heart failure, Clin. Biochem. 44 (2011) 293–299. [52] L. Tenori, X. Hu, P. Pantaleo, B. Alterini, G. Castelli, et al., Metabolomic fingerprint of heart failure in humans: a nuclear magnetic resonance spectroscopy analysis, Int. J. Cardiol. 168 (2013) e113–e115. [53] Z. Du, A. Shen, Y. Huang, L. Su, W. Lai, et al., 1 H-NMR-based metabolic analysis of human serum reveals novel markers of myocardial energy expenditure in heart failure patients, PLoS ONE 9 (2014) e88102. [54] T.H. Marwick, Should we be evaluating the ventricle or the myocardium? Advances in tissue characterization, J. Am. Soc. Echocardiogr. 17 (2004) 168–172. [55] L. Padeletti, P.A. Modesti, S. Cartei, L. Checchi, G. Ricciardi, et al., Metabolomic does not predict response to cardiac resynchronization therapy in patients with heart failure, J. Cardiovasc. Med. (Hagerstown) 15 (2014) 295–300. [56] M.L. Cheng, C.H. Wang, M.S. Shiao, M.H. Liu, Y.Y. Huang, et al., Metabolic disturbances identified in plasma are associated with outcomes in patients with heart failure: diagnostic and prognostic value of metabolomics, J. Am. Coll. Cardiol. 65 (2015) 1509–1520. [57] M. Deidda, C. Piras, C.C. Dessalvi, E. Locci, L. Barberini, F. Torri, F. Ascedu, L. Atzori, G. Mercuro, Metabolomic approach to profile functional and metabolic changes in heart failure, J. Transl. Med. 13 (1) (2015) 297, http://dx.doi.org/10.1186/s12967015-0661-3.