Serum metabolomics reveals lipid metabolism

1 downloads 0 Views 373KB Size Report
Apr 12, 2013 - Serum metabolomics reveals lipid metabolism variation between coronary artery disease and congestive heart failure: a pilot study.
http://informahealthcare.com/bmk ISSN: 1354-750X (print), 1366-5804 (electronic) Biomarkers, Early Online: 1–8 ! 2013 Informa UK Ltd. DOI: 10.3109/1354750X.2013.781222

ORIGINAL ARTICLE

Serum metabolomics reveals lipid metabolism variation between coronary artery disease and congestive heart failure: a pilot study Hemi Luan1*, Xiaomin Chen1*, Shilong Zhong2,3*, Xune Yuan1*, Nan Meng1, Jianfeng Zhang1, Jin Fu1, Ran Xu1, Connie Lee4, Siyuan Song1, Hui Jiang1, and Xun Xu1 BGI-Shenzhen, Shenzhen, China, 2Medical Research Center, Guangdong General Hospital, Guangzhou, Guangdong, China, 3Guangdong Cardiovascular Institute, Guangzhou, Guangdong, China, and 4Fu School of Engineering and Applied Science, Columbia University, New York, USA Abstract

Keywords

The aim of this pilot study is to find discriminating signals from the patient’s congestive heart failure (HF) caused by coronary artery disease (CAD) through a non-target metabolomics method and test their usefulness in progress of human HF diseases. Multivariate data analysis was used to identify the discriminating signals. Interestingly, 12 metabolites contributing to the complete separation of HF from matched CAD were identified. Metabolic pathways including free fatty acids, sphingolipids and amino acid derivatives were found to be disturbed in HF patients compared with CAD patients. Lipid molecules associated with energy metabolism and signaling pathways may play key roles in the development of failing heart.

Congestive heart failure, coronary artery disease, lipid metabolism, metabolomics

Introduction Congestive heart failure (HF) describes an inadequacy of the heart’s pumping function (Katz, 1975). Due to this dysfunction, HF can cause lots of symptoms including shortness of breath, leg swelling and a decreased tolerance to exercise. This is recognized as one of the most important public health problems (Gottdiener et al., 2002; Lainscak & Keber, 2003; Mancini et al., 1992). Heart failure is a common, costly, disabling and potentially deadly condition (Eriksson, 1995). Approximately, 23 million people worldwide are afflicted with congestive heart failure, and 2 million new cases of HF are diagnosed each year worldwide (Tang et al., 2012). In developed countries, about 2% of adults suffer from heart failure and this increases to 6%–10% for those over the age of 65 (Campos et al., 2011). The estimated prevalence of HF in China is 4 million patients aged 35–74 years (Jiang & Ge, 2009). Other cardiovascular disorders that have actually declined during the past few decades, in contrast, the incidence of heart failure is on the rise. In fact, it is the most rapidly growing cardiovascular disorder in the world (Braunwald, 2012; Gheorghiade & Bonow, 1998). Better living standards and longer lifespans is accompanied by dietary changes (increased fat and salt intake), cigarette smoking, less exercise and obesity. Coronary artery disease (CAD), diabetes and hypertension are the initial

*These authors contributed equally to this work. Address for correspondence: Prof. Xun Xu, BGI, Building No. 11, Beishan Industrial Zone, Yantian District, Shenzhen 518083, China. Fax: +86 755 2235 4236 (Office). Tel: +86 755 2232 5193 (Office). E-mail: [email protected]

History Received 14 December 2012 Accepted 25 February 2013 Published online 12 April 2013

consequences then heart failure follows. HF is the end product of myocardial damage caused by these disease processes either along or in combination with other factors (Boonmande Winter et al., 2012; Segers et al., 2012). In recent decades, more and more people realize that CAD has been the most common cause of HF, not hypertension or valvular heart disease. In 13 multicenter heart failure trials reported over the past 10 years, involving about 20 000 patients, CAD was the underlying cause of heart failure in nearly 70% of the patients (Colucci et al., 1996; Kjekshus et al., 1992). CAD is the result of the accumulation of atheromatous plaques within the walls of the coronary arteries that supply the myocardium with oxygen and nutrients. Limited blood flow to the heart caused by atheromatous plaques results in ischemia of the myocardial cells. Myocardial cells may die from a lack of oxygen and nutrients leading to heart muscle damage (Falk, 1983). Recurrent episodes of reversible myocardial ischemia producing repetitive myocardial stunning may contribute to the happening and worsening of left ventricular dysfunction and heart failure symptoms (Bolli, 1990). Another mechanism for left ventricular dysfunction involves the coronary endothelium which plays an important role in the physiological modulation of myocardial structure and function. Disordered endothelial function stimulates vasoconstriction, smooth muscle migration and proliferation, promoting myocardial ischemia, which further contributes to left ventricular dysfunction (Lerman & Zeiher, 2005). At present, there is a growing evidence to support the theory that alterations in substrate metabolism, seen in HF, contribute to contractile dysfunction and to the progression of left ventricular remodeling (Stanley et al., 2005). Cardiac metabolism may be a target for the treatment of HF

20 13

Biomarkers Downloaded from informahealthcare.com by University of Illinois on 05/22/13 For personal use only.

1

2

H. Luan et al.

Biomarkers, Early Online: 1–8

Table 1. Demographic and clinical chemistry characteristics of human subjects.

Samples number Age (median, range), year Male/Female number BMI (median, range), kg/m2 GLUC (median, range), mmol/L TRIG (median, range), mmol/L LDLC (median, range), mmol/L CHOL (median, range), mmol/L HDLC (median, range), mmol/L APOA (median, range), g/L LPa (median, range), mg/L LVEF (median, range), %

HF patients

CAD patients

Healthy controls

23 54, 42–75 23/0 19.13, 17.04–22.48 4.93, 3.55–15.68 1.15, 0.56–2.56 2.62, 1.59–2.00D 4.40, 3.21–6.62D 0.93, 0.48–1.50 1.02, 0.79–2.24 128.72, 46.00–1962.39 36.00, 18.00–67.00a

23 58, 44–84 23/0 18.69, 17.43–21.36 5.40, 3.84–14.39 1.18, 0.75–3.42 1.97, 1.33–3.19* 3.44, 2.80–5.77* 0.86, 0.66–1.20* 1.01, 0.79–1.41 167.53, 38.00–1158.61 60.00, 19.00–69.00

23 55, 38–69 23/0 26.77, 19.65–62.8 4.94, 3.13–8.97 1.33, 0.64–3.36 3.395, 1.71–7.12 5.57, 3.62–9.11 1.205, 0.68–1.71 0.96, 0.7–2.09 144, 19–1179 –

Biomarkers Downloaded from informahealthcare.com by University of Illinois on 05/22/13 For personal use only.

BMI, body mass index; GLUC, Glucose; TRIG, Triglyceride; LDLC, low-density lipoprotein; HDLC, high-density lipoprotein; CHOL, Cholesterol; APOA, Apolipoprotein (a); LPa, Lipoprotein (a); LVEF, Left ventricular ejection fractions; –, No detected. a Calculated by Student’s t-test and indicated significant result between CAD and HF (less than 0.05); *, calculated by Student’s t-test and indicated significant result between CAD and Healthy controls (less than 0.05).

(Taegtmeyer, 2004). The heart has a very high energy demand and must continually generate ATP at a higher rate to sustain contractile function. The fatty acid oxidation and carbohydrate metabolism is the heart’s chief energy supply. At the end stages of HF, myocardium has low ATP content due to a decreased ability to generate ATP by an oxidative metabolism and inability to effectively transfer chemical energy (Almeida et al., 2012; Fillmore & Lopaschuk, 2013). When the heart is acutely stressed, the dominance of fatty-acid metabolism rise to the term: ‘‘glucose–fatty-acid cycle’’. Energy metabolism readily switches from fat metabolism to carbohydrate metabolism to fuel oxidative energy production (Almeida et al., 2012; Jaswal et al., 2011). Although metabolism and heart function are inextricably linked and interest in this area is growing, metabolic dysfunction in HF and CAD are poorly understood, and the correlation of HF and CAD unknown. Metabolomics, a method for qualitation and quantitation of all small molecule metabolites in biological matrices following genetic mutation or exogenous changes, has become an important part of systems biology, complementing genomics and proteomics. Metabolomic analysis has been used for the diagnosis of a number of pathological elucidations of clinical-pathogenesis for various diseases assessing exposure of biological systems to xenobiotics due to high sensitivity and the capacity to quantitatively measure the entire composition of metabolites in a given biological specimen (Kaddurah-Daouk et al., 2008). Shah et al. have reported the mass spectrometry-based profiling of 69 metabolites in the patients with CAD and found the association of blood metabolic profile with CAD. Some potential biomarkers of heart failure were also found by the serum metabolomics. However, to our knowledge, there were few reports about the association between CAD and HF in the metabolic layer. Nuclear magnetic resonance (NMR) spectroscopy and gas chromatographic-mass spectrometry (GC-MS) were usually used in the cardiovascular disease research. The metabolites with low concentration and temperature sensitive could be undetected by above methods (Dunn et al., 2007; Shah et al., 2010; Xia et al., 2012). In the present study, we used high performance liquid chromatography-mass spectrometry (HPLC-MS) and computational methods to characterize the

metabolic changes between CAD and HF, identify characteristic metabolites and then interpret these changes in terms of metabolic pathways.

Methods Participants and specimens Patients diagnosed with HF (n ¼ 23) and CAD (n ¼ 23) were recruited from the Guangdong General Hospital and categorized according to pathological features. The clinical diagnosis and pathological reports of all the patients were obtained from the hospital. Body mass indices (BMI) for each patient were also assessed. The control group was comprised of 23 healthy men, of similar age, who underwent physical examination in the same hospital. Then, a second set of HF (n ¼ 15), CAD (n ¼ 15) and control subjects (n ¼ 15) were blindly selected and tested using our approach. The clinical diagnosis and pathological reports were obtained from the hospital. The participants’ clinical information is provided in Table 1. The complete ethical approval has been obtained, and all the patients gave written informed consent. The study was approved by the Institutional Review Board of Guangdong General Hospital. Reagents and sample preparation Acetonitrile, HPLC grade, was obtained from Merck (Darmstadt, Germany). Methanol and formic acid (HPLC grade) was purchased from Fisher Scientific Corporation (Loughborough, UK). Water was produced by a Milli-Q Ultra-pure water system (Millipore, Billerica, MA). Palmitic acid, sphinganine, phytosphingosine, arachidic acid, dodecanoic acid and myristic acid used were of HPLC grade and purchased from Sigma-Aldrich (St. Louis, MO) or Avanti Polar Lipids (Alabaster, AL). All other reagents were of HPLC grade. The blood samples were all collected from patients at Guangdong General Hospital. Serums were separated from vein blood and stored at 80  C until use in the assay (Lu et al., 2012). A ‘‘quality control’’ (QC) sample was prepared by mixing equal volumes (10 mL) from each of the 79 samples as they were being aliquoted for the analysis. This

DOI: 10.3109/1354750X.2013.781222

‘‘pooled’’ serum was used to provide a representative ‘‘mean’’ sample containing all the analytes encountered during the analysis (Zelena et al., 2009).

Biomarkers Downloaded from informahealthcare.com by University of Illinois on 05/22/13 For personal use only.

Analysis of serum samples by LC-MS The low-molecular-weight metabolites (51000 Da) in serum samples were isolated and treated as previously reported, with a few modifications (Lu et al., 2012). A total of 50 mL of thawed serum samples were collected and then precipitated by 200 mL of methanol. After centrifugation at 14 167 g for 10 min at 4  C, the supernatant was transferred to a 1.5 mL polypropylene tube, 10 mL of the supernatant were injected into the HPLC-MS. LC-MS data was acquired using a LTQ Orbitrap instrument (Thermo Fisher Scientific, Waltham, MA) set at 30 000 resolution. Sample analysis was carried out under positive ion mode. The mass scanning range was 50–1500 m/z and the capillary temperature was 350  C. Nitrogen sheath gas was set at a flow rate of 30 L/min. Nitrogen auxiliary gas was set at a flow rate of 10 L/min. Spray voltage was set to 4.5 kV. The LC-MS system was run in a binary gradient mode. Solvent A was 0.1% (v/v) formic acid/water and solvent B was 0.1% (v/ v) formic acid/methanol. The flow rate was 0.2 mL/min. An Agilent ZORBAX ODS C-18 column (Agilent Technologies, Santa Clara, CA) (150  2.1 mm, 3.5 mm) was used for all analysis. The gradient was as follows: 5% B at 0 min, 5% B at 5 min, 100% B at 8 min, 100% at 9 min, 5% B at 18 min and 5% B at 20 min. The pooled ‘‘QC’’ sample was injected five times at the beginning of the run to ensure the system equilibrium and then every five samples to further monitor the stability of the analysis (Lu et al., 2012).

LC-MS data processing and statistics Data pre-treatment including peak picking, peak grouping, retention time correction and second peak grouping was achieved using the XCMS software (http://metlin.scripps.edu/ download/) implemented with the freely available R statistical language (v 2.13.1) (http://www.r-project.org/). LC-MS raw data files were initially converted into netCDF format, then directly processed by the XCMS toolbox (http://metlin.scripps.edu/xcms/) (Smith et al., 2006). A list of the ion intensities of each peak detected was generated using retention time (RT) and the m/z data pairs as identifiers for each ion. The resulting three-dimensional matrix contained arbitrarily assigned peak indices (retention time–m/z pairs), sample names (observations) and ion intensity information (variables). To obtain consistent variables, the resulting matrix was further reduced by removing peaks with 80% missing values (those with ion intensity ¼ 0) and those with isotope ions from both the HF and CAD groups. For each metabolite peak reproducibly detected in two groups, the null hypothesis that the means of the HF and CAD sample populations were equal was tested using a two-tailed Student’s t-test results after false discovery rate (FDR) correction at p50.05 (Ling et al., 2009). Orthogonal projection to a latent structures discriminant analysis (OPLS-DA) was carried out to discriminate between HF and CAD patients (Wiklund et al., 2008).

Lipid metabolism variation between CAD and congestive HF

3

The optimal number of latent factors used in the OPLSDA model was selected using stratified 7-fold crossvalidation and the model quality was assessed using standard R2. R2, the squared correlation coefficient between the dependent variable and the PLS-DA prediction, measures ‘‘goodness of fit’’ (a value between 0 and 1, where 1 is a perfect correlation) using all the available data to build a given OPLS-DA model. The robustness of the final OPLSDA model was further validated by comparing the R2 value to a reference distribution of all of the possible models using permutation testing (n ¼ 1000), this follows standard protocol for metabolomic studies (Taylor & Mackinnon, 2012). A reference R2 distribution is obtained by calculating all the possible OPLS-DA models under random reassignment of the HF/CAD labels for each measured metabolic profile. If the correctly labeled model’s R2 value is close to the center of the reference distribution, then the model performs no better than a randomly assigned model and is, therefore, invalid. For all the OPLS-DA models described here, the associated reference distribution plots are provided and an estimate of the probability of the candidate model randomly occurring can be estimated. On the basis of a variable importance in the projection (VIP), a threshold of 2 from the 7-fold cross-validated OPLS-DA model, a number of metabolites responsible for the difference in the metabolic profiles of HF and CAD can be obtained. The metabolites identified by the OPLS-DA model were validated at a univariate level using the FDR test from the R statistical toolbox with the critical p value set to not higher than 0.05. The corresponding fold changes show how these selected differential metabolites varied between the HF and CAD groups. Discriminatory metabolites of interest were extracted from the combining VIP value, p value and fold change. For each discriminatory metabolites, an ROC curve was determined to assess each metabolite’s effectiveness as a univariate discriminatory biomarker (Ndrepepa et al., 2007). Exact molecular mass data from redundant m/z peaks used the online HMDB database (http://www.hmdb.ca/), METLIN (http://metlin.scripps.edu/) and KEGG (www.genome.jp/ kegg/) for database metabolite lookups. A metabolite name was reported when a match with a mass difference between observed and theoretical mass was 510 ppm. The isotopic distribution measurement was used to further validate the metabolite molecular formula of matched metabolites (Xu et al., 2010). The identities of the specific metabolites were confirmed by a comparison of their mass spectra and chromatographic retention times with those obtained using commercially available reference standards.

Results Using the optimized HPLC-MS analysis protocol and subsequent processes, such as baseline correction, peak deconvolution, alignment and normalization, we obtained a threedimensional matrix, including data filename, retention-time exact mass pair and normalized peak areas. There were 2326 retention time-exact mass pairs determined in each sample profile. Principal Component Analysis (PCA) shows that no shift was observed in the PCA space of data obtained on QC

Biomarkers Downloaded from informahealthcare.com by University of Illinois on 05/22/13 For personal use only.

4

H. Luan et al.

Biomarkers, Early Online: 1–8

Figure 1. (A) The scores plot for an OPLS-DA model using the optimal number of latent vectors (n ¼ 2) for data taken from the HF-CAD study (the right indicates HF; the left, CAD). Model construction was performed using a 7-fold cross validation resulting in an R2 of 0.935 and Q2 of 0.439. (B) The R2 distribution plot shows that the chosen model’s R2 value is significantly distant from the H0 randomly classified permutation distribution (n ¼ 1000); thus, the probability of the presented model randomly occurring is 50.001.

samples as shown in Figure S1. Thus, the metabolic features demonstrated acceptable reproducibility and stability in LC-MS profiling analyses. All variables were used in the following analysis. This PCA result was also calculated for the purpose of obtaining an overview of the data. The model explained 57.4% of the variation in the metabolic profiling [R2(X) ¼ 57.4%] with a predictability of 31.0% [Q2(X) ¼ 31.0%]. No outliers were observed in the data. The OPLS-DA model demonstrated satisfactory modeling and prediction using two orthogonal components (R2cum ¼ 0.935, Q2cum ¼ 0.439), achieving a distinct separation between the metabolite profiles of the two groups (Figure 1A). Model selection was performed using a 7-fold cross-validation and the final model was further validated with a permutation testing. The final model used two latent factors and the probability of this model randomly occurring was 50.001 (Figure 1B). Figure 1 shows the OPLS-DA scores plot and the permutation tests. From the corresponding loading plots and according to the variable’s importance in the projection (VIP42), the 114 ions farthest away from the origin contribute significantly to the separation between HF and CAD groups and may therefore be regarded as the differentiating metabolites for HF and CAD groups (Figure 2A). As shown in Figure 2(B), a volcano plot combines a statistical test (p value50.05, fold change 42 or 50.5) with the magnitude of the change enabling quick visual identification of those metabolites that display large-magnitude changes (Cui & Churchill, 2003). The volcano plot showed the top 31 significant features of the metabolites based on the statistical testing. Of the 2326 candidate peaks detected by the HPLC-MS, 12 were ultimately selected as discriminating metabolites by combining their VIP values and the volcano plot results (Table 2). These relative standard deviation (RSD) values of metabolites in QC samples were also shown in Table 2, which

were acceptable in no-targeted research (RSD%530%) (Dunn et al., 2012). Six metabolites were identified using MS spectral databases and four were confirmed using reference standards. These were free fatty acids, free sphingolipids, amino acid derivatives and phosphatides, etc. Among the identified metabolites, free fatty acids, such as dodecanoic acid, myristic acid, palmitic acid, arachidic acid were the serum metabolites found to have decreased in the HF patients, and compared to CAD, showing the greatest fold change (0.41, 0.41, 0.42 and 0.38, respectively) and area under the ROC curve (AUC) of 0.85, 0.89, 0.90 and 0.92 respectively. Free sphingolipids, such as sphinganine and phytosphingosine were also the metabolites that decreased (fold change: 0.36 and 0.38 respectively; AUC: 0.96 and 0.92, respectively) in HF patients. S-Allyl-L-cysteine and PrenylL -cysteine, derivatives of cysteine, were significantly decreased levels of amino acid derivatives (p value ¼ 0.0002 and p value ¼ 0.0003, respectively; fold changes: 0.46 and 0.43, respectively; AUC: 0.90 and 0.91, respectively) (Tables 2, 3 and Figure S2). For validation purposes, 12 metabolites were also detected in the validation study. An OPLS-DA model using the 12 metabolites (two latent factors) proved to be predictive, with R2 of 0.69, Q2 of 0.51. This OPLS-DA model could be achieved for discrimination of all the patients in the second set. The model was further validated using a permutation testing (Figure S3). An analysis was also made of the relationship between the levels of 12 metabolites and the left ventricular ejection fractions (LEVF) values of patients. LEVF is used to assess the function of the heart and diagnose heart failure. There was a significant correlation between LEVF values and S-allyl-L-cysteine (r ¼ 0.49; p ¼ 0.001, Spearman rank correlation), Prenyl-L-cysteine (r ¼ 0.57; p ¼ 0.0001, Spearman rank correlation), sphinganine (r ¼ 0.38; p ¼ 0.02, Spearman rank correlation), arachidic acid (r ¼ 0.35; p ¼ 0.03,

Lipid metabolism variation between CAD and congestive HF

Biomarkers Downloaded from informahealthcare.com by University of Illinois on 05/22/13 For personal use only.

DOI: 10.3109/1354750X.2013.781222

5

Figure 2. (A) The loading plot of OPLS-DA model. The 114 ions farthest away from the origin contribute significantly to the separation between HF and CAD groups (VIP42). (B) The volcano plot showed the top 31 significant features of the metabolites based on statistical testing (p value50.05, fold change 42 or 50.5). Twelve metabolites were selected as the discriminatory metabolites by combining the VIP values and the results from the volcano plot (names was marked).

Table 2. Discriminating metabolites of interest in the HF patients and CAD patients. m/z 162.0582 190.0893 218.2105 246.2410 274.2717 302.3035 318.2976 330.3350 346.3297 362.3235 176.0739 202.0895

Rt (min)

a

FC (HF/CHD)b

Adjusted p valuec

VIP

Adduct ion

Formula

Metabolites

0.46 0.43 0.41 0.41 0.42 0.36 0.38 0.38 0.36 0.36 0.41 0.41

0.0002 0.0003 0.0060 0.0002 0.0010 0.0001 0.0003 0.0001 0.0002 0.0002 0.0003 0.0009

3.9 2.0 3.0 2.6 5.3 3.8 4.16 2.23 2.69 2.26 4.19 2.07

[M þ H]þ [M þ H]þ [M þ NH4]þ [M þ NH4]þ [M þ NH4]þ [M þ H]þ [M þ H]þ [M þ NH4]þ [M þ NH4]þ [M þ NH4]þ

C6H11NO2S C8H15NO2S C12H24O2 C14H128O2 C16H32O2 C18H39NO2 C18H39NO3 C20H40O2 C20H40O3 C20H40O4

S-Allyl-L-cysteiney Prenyl-L-cysteiney Dodecanoic acid* Myristic acid * Palmitic acid* Sphinganine * Phytosphingosine * Arachidic acid* 2-hydroxyphytanic acidy MG(17:0/0:0/0:0)y

6.5 9.7 10.1 10.7 11.0 11.2 11.0 11.4 11.2 11.0 9.0 9.4

Metabolites are annotated using the following: *, available reference standards; y, accurate mass and isotopic distribution measurement with the aid of web-based resources, such as the Human Metabolome Database (http://www.hmdb.ca/) and METLIN (http://metlin.scripps.edu). a Retention time; bfold change; cCalculated by the two-tailed Student’s t-test results after false discovery rate correction.

Table 3. The AUC and odd ratio of discriminating metabolites of interest in the HF patients and CAD patients. m/z

AUC

162.0582 190.0893 218.2105 246.2410 274.2717 302.3035 318.2976 330.3350 346.3297 362.3235 176.0739 202.0895

0.90 0.91 0.85 0.89 0.90 0.96 0.92 0.93 0.95 0.93 0.42 0.89

Odd ratio (95% CI)a 13.12 16.74 21.49 26.99 25.6 228.6 46.67 1419.67 63.58 51.31 0.67 7.98

(3.21–53.60) (3.59–78.02) (2.99–154.14) (3.72–195.42) (3.92–168.33) (8.5–6147.88) (4.76–457.21) (6.99–28813) (6.48–623.53) (4.71–558.92) (0.35–1.31) (2.72–23.42)

p Valueb

RSD% (QC samples)

3.00E–04 3.00E–04 2.00E–3 1.00E–03 7.00E–04 1.00E–03 9.00E–04 7.00E–03 3.00E–04 1.00E–03 2.45E–01 2.00E–04

17.34 18.60 3.12 4.93 4.37 10.58 6.15 29.65 9.44 6.80 27.45 15.96

Calculated by logistic model; bcalculated by Fisher’s exact test and indicated an increased relative risk (less than 0.05). RSD: relative standard deviation.

a

6

H. Luan et al.

Spearman rank correlation) and 202.0895 p ¼ 0.0001, Spearman rank correlation).

Biomarkers, Early Online: 1–8

(r ¼ 0.57;

Biomarkers Downloaded from informahealthcare.com by University of Illinois on 05/22/13 For personal use only.

Discussion HF and CAD are both complex cardiovascular syndromes. They involve multiple biological pathways. CAD, one of the leading causes of heart failure, can ultimately lead to a decrease in the reserve of heart output and start a decline into heart failure. Despite intensive work, the pathogenesis of the cardiac abnormalities that result from HF are still not completely understood. Number of structural and biochemical cardiac abnormalities were shown to be associated with HF, such as defects in mitochondria and abnormal energy signal transduction (Rosca et al., 2012). Metabolic pathway abnormalities in the failing heart that result in decreased energy production, energy transfer and energy utilization have been proved to be the important factors that contribute to heart failure. To get insight into the metabolic mechanism of CAD and HF, we take a non-target metabolomics approach to identify metabolic signatures in serum that is predictive of subsequent HF caused by CAD. The serum metabolic profiling and subsequent multivariate analysis clearly distinguished HF patients from matching the CAD group. As shown in Figure 1, OPLS-DA revealed a clear and statistically significant separation between the HF and CAD samples. Results indicate that HF-related metabolites play an important role in lipid metabolism, sphingolipids metabolism and amino acid metabolism and more. The adult heart has a very high energy requirement and must continually take ATP at a high rate to sustain contractile function, basal metabolic processes and ionic homeostasis. To sustain sufficient ATP generation, the heart uses a variety of different carbon substrates as energy sources if available. Normally, fatty acid beta-oxidation can provide about 50%–70% of the ATP needed for the adult heart (Oliveira et al., 2012). Circulating fatty acids levels can dramatically vary during severe stress, such as myocardial ischemia and CAD, compared to healthy controls. Consistent with previous reports, we have also detected increasing fatty acids levels in the serum of CAD patients compared to healthy controls (Wang et al., 2003). However, compared with CAD patients, the fatty acids levels in the serum of the HF group have significantly decreased (Figure 3). Amounts of saturated and unsaturated fatty acids in the CAD patients were proven higher than those of the healthy controls (Merry et al., 2012). Fatty acids are supplied to the heart as either free fatty acids bound to albumin or as fatty acids released from triacylglycerol contained in chylomicrons or very-low-density lipoproteins. Plasma fatty acid concentration can be regulated by the net release from lipid–protein complexes. Lipoprotein lipase activity plays a key role in the free fatty acid levels of patients with CAD (Fabian et al., 1971). Current evidence suggests that the myocardial capacity for fatty acid beta-oxidant is relatively normal during the early development of heart failure, while there is a clear decrease in fatty acid betaoxidant capacity in the end stages. The decrease in fatty acid b-oxidation and parallel increase in glucose oxidation may be an indication of impaired energy efficiency in the failing heart. On the other hand, during heart failure, myocardial

Figure 3. Relative expression levels of four free fatty acids in the serum of healthy controls, CAD and HF (means  SE). *p Value50.05.

Figure 4. Relative expression levels of two sphingolipids in the serum of healthy controls, CAD and HF (means  SE). *p Value50.05.

fatty acid uptake rates are higher than expected in the normal heart (Lopaschuk et al., 2010). The variation of myocardial fatty acid uptake capacity and lipoprotein lipase activity are likely important reasons for the significant differences between HF and CAD patients. Sphingolipids play a significant role in membrane structures, signal transduction and regulation of a host of cellular processes, such as cell proliferation, differentiation and apoptosis (Egom et al., 2012). The regulating neurotransmitter binding and signal transducing functions in cardiovascular cell biology and pathology are of interest to many researchers. The SM-ceramide signaling pathways are considered to be one of the pathways activated in response to myocardial

DOI: 10.3109/1354750X.2013.781222

Lipid metabolism variation between CAD and congestive HF

7

Biomarkers Downloaded from informahealthcare.com by University of Illinois on 05/22/13 For personal use only.

(Figure S5). This result indicates that the adjuvant therapy, by elevating the carnitine in blood, could be beneficial in HF treatment. In summary, our data provide the new insight about major differences of metabolic profiles between CAD and HF. Metabolite profiles were associated with subsequent heart failure caused by CAD. The lipid metabolism, sphingolipids metabolism and amino acid metabolism variation may indicate the development of failing heart and provide significant implication for etiology and pathology of heart failure. We show that the method of HPLC-MS-based metabolomics is a reliable analytical system and improve the metabolite coverage for serum metabolomics in cardiovascular diseases. This HPLC-MS method is very suitable to discover the potential biomarkers with low concentration and temperature-sensitive in cardiovascular diseases, complementing to the GC-MS and NMR spectroscopy.

Conclusions Figure 5. Relative expression levels of two amino acid derivatives in serum of healthy controls, CAD and HF (means  SE). *p Value50.05.

ischemia/reperfusion (Van Brocklyn & Williams, 2012). Phytosphingosine and sphinganine, two key precursors of ceramide in the ceramide synthesis pathway, have decreased concentrations in the HF patients compared with the CAD group, indicating a disorder in sphingolipids metabolism (Figure 4). Decreased serum sphingolipids base levels in HF patients that were observed in our study could be a result of either their reduced release to the circulation or increased rate of their phosphorylation in blood cells (Knapp et al., 2012). Decreased levels of S-Allyl-L-cysteine and PrenylL-cysteine in the HF group have been detected, but there are no apparent differences between the CAD and HF patients (Figure 5). These metabolites belong to cysteine metabolism and may be a useful indicator for classifying the HF patients from two other groups. Many specific metabolites in the cysteine metabolism pathway have been associated with HF, such as homocysteine, taurine and cysteine (Diercks et al., 2012; Vacek et al., 2012). Besides the discriminatory metabolites selected above, the metabolites associated with CAD and HF were also detected. Phospholipids in the human circulation system are composed of about 70% phosphatidylcholines (PC). They, play a preventive role in the development of atherosclerosis (Park et al., 2004). PC(16:1/0:0), PC(18:2/ 0:0), PC(18:1/0:0), PC(18:0/0:0) was found to display significant differences between the CAD and healthy controls, but was lacking notable differences between HF and CAD patients. The decreased levels of PC may be an important risk factor in development of CAD and HF (Figure S4). Carnitine and acetylcarnitine have the positive effects on the heart function and glucose and lipid metabolism in heart failure patients with coronary heart disease and type-2 diabetes. L-carnitine has even been used as an adjuvant therapy to improve heart function in heart failure patients (Omori et al., 2012). Compared to the healthy controls and CAD patients, the HF patients had slightly lowered carnitine levels in the serum, but there are no other significant differences

The present study is a metabolic profiling screen detailing two cardiovascular diseases. A significant separation using the OPLS-DA analysis was obtained for the difference between HF patients versus CAD patients. The finding of discriminatory metabolites in serum preceding HF and CAD offers insight into disease pathogenesis. Associated energy lipid molecules metabolism and signaling pathways may reflect the development of failing heart.

Declaration of interest The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article. This work was supported by grants from the National Key Basic Research Development Program (Grant No. 2013CB945204) and National Science Foundation of China of China (Grant No. 81072701). References Almeida AS, Queiroga CS, Sousa MF, et al. (2012). Carbon monoxide modulates apoptosis by reinforcing oxidative metabolism in astrocytes: role of Bcl-2. J Biol Chem 287:10761–70. Bolli R. (1990). Mechanism of myocardial ‘‘stunning’’. Circulation 82: 723–38. Boonman-de Winter LJ, Rutten FH, Cramer MJ, et al. (2012). High prevalence of previously unknown heart failure and left ventricular dysfunction in patients with type 2 diabetes. Diabetologia 55: 2154–62. Braunwald E. (2012). The rise of cardiovascular medicine. Eur Heart J 33:838–45, 45a. Campos LA, Bader M, Baltatu OC. (2011). Brain renin-angiotensin system in hypertension, cardiac hypertrophy, and heart failure. Front Physiol 2:115.1–5. Colucci WS, Packer M, Bristow MR, et al. (1996). Carvedilol inhibits clinical progression in patients with mild symptoms of heart failure. US Carvedilol Heart Failure Study Group. Circulation 94:2800–6. Cui X, Churchill GA. (2003). Statistical tests for differential expression in cDNA microarray experiments. Genome Biol 4:210.1–10. Diercks DB, Owen K, Tolstikov V, Sutter M. (2012). Urinary metabolomic analysis for the identification of renal injury in patients with acute heart failure. Acad Emerg Med 19:18–23. Dunn W, Broadhurst D, Deepak S, et al. (2007). Serum metabolomics reveals many novel metabolic markers of heart failure, including pseudouridine and 2-oxoglutarate. Metabolomics 3:413–26.

Biomarkers Downloaded from informahealthcare.com by University of Illinois on 05/22/13 For personal use only.

8

H. Luan et al.

Dunn WB, Wilson ID, Nicholls AW, Broadhurst D. (2012). The importance of experimental design and QC samples in large-scale and MS-driven untargeted metabolomic studies of humans. Bioanalysis 4: 2249–64. Egom EE, Mamas MA, Clark AL. (2012). The potential role of sphingolipid-mediated cell signaling in the interaction between hyperglycemia, acute myocardial infarction and heart failure. Expert Opin Ther Targets 16:791–800. Eriksson H. (1995). Heart failure: a growing public health problem. J Intern Med 237:135–41. Fabian E, Havranek F, Stork A, Fabianova J. (1971). Lipoprotein lipase and postheparin esterase activity and postheparin increase in free fatty acids after the short-term administration of a sequential contraceptive. Am J Obstet Gynecol 109:1212–13. Falk E. (1983). Plaque rupture with severe pre-existing stenosis precipitating coronary thrombosis. Characteristics of coronary atherosclerotic plaques underlying fatal occlusive thrombi. Br Heart J 50:127–34. Fillmore N, Lopaschuk GD. (2013). Targeting mitochondrial oxidative metabolism as an approach to treat heart failure. Biochim Biophys Acta 1833:857–65. Gheorghiade M, Bonow RO. (1998). Chronic heart failure in the United States: a manifestation of coronary artery disease. Circulation 97: 282–9. Gottdiener JS, McClelland RL, Marshall R, et al. (2002). Outcome of congestive heart failure in elderly persons: influence of left ventricular systolic function. The Cardiovascular Health Study. Ann Intern Med 137:631–9. Jaswal JS, Keung W, Wang W, et al. (2011). Targeting fatty acid and carbohydrate oxidation – a novel therapeutic intervention in the ischemic and failing heart. Biochim Biophys Acta 1813:1333–50. Jiang H, Ge J. (2009). Epidemiology and clinical management of cardiomyopathies and heart failure in China. Heart 95:1727–31. Kaddurah-Daouk R, Kristal BS, Weinshilboum RM. (2008). Metabolomics: a global biochemical approach to drug response and disease. Annu Rev Pharmacol Toxicol 48:653–83. Katz AM. (1975). Congestive heart failure: role of altered myocardial cellular control. N Engl J Med 293:1184–91. Kjekshus J, Swedberg K, Snapinn S. (1992). Effects of enalapril on longterm mortality in severe congestive heart failure. CONSENSUS Trial Group. Am J Cardiol 69:103–7. Knapp M, Baranowski M, Lisowska A, Musial W. (2012). Decreased free sphingoid base concentration in the plasma of patients with chronic systolic heart failure. Adv Med Sci 57:100–5. Lainscak M, Keber I. (2003). Patient’s view of heart failure: from the understanding to the quality of life. Eur J Cardiovasc Nurs 2:275–81. Lerman A, Zeiher AM. (2005). Endothelial function: cardiac events. Circulation 111:363–8. Ling XB, Cohen H, Jin J, et al. (2009). FDR made easy in differential feature discovery and correlation analyses. Bioinformatics 25:1461–2. Lopaschuk GD, Ussher JR, Folmes CD, et al. (2010). Myocardial fatty acid metabolism in health and disease. Physiol Rev 90:207–58. Lu K, Knutson CG, Wishnok JS, et al. (2012). Serum metabolomics in a Helicobacter hepaticus mouse model of Inflammatory Bowel Disease reveals important changes in the microbiome, serum peptides, and intermediary metabolism. J Proteome Res 11:4916–26. Mancini DM, Walter G, Reichek N, et al. (1992). Contribution of skeletal muscle atrophy to exercise intolerance and altered muscle metabolism in heart failure. Circulation 85:1364–73. Merry AH, Erkens PM, Boer JM, et al. (2012). Co-occurrence of metabolic factors and the risk of coronary heart disease: a prospective cohort study in the Netherlands. Int J Cardiol 155:223–9.

Biomarkers, Early Online: 1–8

Ndrepepa G, Braun S, Kastrati A, Schomig A. (2007). Area under ROC curve, sensitivity, specificity of N-terminal probrain natriuretic peptide in predicting mortality in various subsets of patients with ischemic heart disease. Clin Res Cardiol 96:763–5. Oliveira PJ, Carvalho RA, Portincasa P, et al. (2012). Fatty acid oxidation and cardiovascular risk during menopause: a mitochondrial connection? J Lipids 2012:1–12 Omori Y, Ohtani T, Sakata Y, et al. (2012). L-Carnitine prevents the development of ventricular fibrosis and heart failure with preserved ejection fraction in hypertensive heart disease. J Hypertens 30: 1834–44. Park TS, Panek RL, Mueller SB, et al. (2004). Inhibition of sphingomyelin synthesis reduces atherogenesis in apolipoprotein E-knockout mice. Circulation 110:3465–71. Rosca MG, Tandler B, Hoppel CL. (2012). Mitochondria in cardiac hypertrophy and heart failure. J Mol Cell Cardiol 55:31–41. Segers VF, Brutsaert DL, De Keulenaer GW. (2012). Pulmonary hypertension and right heart failure in heart failure with preserved left ventricular ejection fraction: pathophysiology and natural history. Curr Opin Cardiol 27:273–80. Shah SH, Bain JR, Muehlbauer MJ, et al. (2010). Association of a peripheral blood metabolic profile with coronary artery disease and risk of subsequent cardiovascular events. Circ Cardiovasc Genet 3: 207–14. Smith CA, Want EJ, O’Maille G, et al. (2006). XCMS: processing mass spectrometry data for metabolite profiling using nonlinear peak alignment, matching, and identification. Anal Chem 78:779–87. Stanley WC, Recchia FA, Lopaschuk GD. (2005). Myocardial substrate metabolism in the normal and failing heart. Physiol Rev 85:1093–129. Taegtmeyer H. (2004). Cardiac metabolism as a target for the treatment of heart failure. Circulation 110:894–6. Tang T, Gao MH, Hammond HK. (2012). Prospects for gene transfer for clinical heart failure. Gene Ther 19:606–12. Taylor AB, Mackinnon DP. (2012). Four applications of permutation methods to testing a single-mediator model. Behav Res Methods 44: 806–44. Vacek TP, Vacek JC, Tyagi N, Tyagi SC. (2012). Autophagy and heart failure: a possible role for homocysteine. Cell Biochem Biophys 62: 1–11. Van Brocklyn JR, Williams JB. (2012). The control of the balance between ceramide and sphingosine-L-phosphate by sphingosine kinase: oxidative stress and the seesaw of cell survival and death. Comp Biochem Physiol B Biochem Mol Biol 163:26–36. Wang L, Folsom AR, Eckfeldt JH. (2003). Plasma fatty acid composition and incidence of coronary heart disease in middle aged adults: the Atherosclerosis Risk in Communities (ARIC) Study. Nutr Metab Cardiovasc Dis 13:256–66. Wiklund S, Johansson E, Sjostrom L, et al. (2008). Visualization of GC/TOF-MS-based metabolomics data for identification of biochemically interesting compounds using OPLS class models. Anal Chem 80:115–22. Xia J, Yi L, Liu N, et al. (2012). Human plasma metabolic profiles of coronary heart disease by gas chromatography-mass spectrometry with Monte Carlo tree approach. Analyt Lett 45:2185–97. Xu Y, Heilier JF, Madalinski G, et al. (2010). Evaluation of accurate mass and relative isotopic abundance measurements in the LTQOrbitrap mass spectrometer for further metabolomics database building. Anal Chem 82:5490–501. Zelena E, Dunn WB, Broadhurst D, et al. (2009). Development of a robust and repeatable UPLC-MS method for the long-term metabolomic study of human serum. Anal Chem 81:1357–64.