Novel Findings From a Metabolomics Study of Left Ventricular Diastolic Function: The Bogalusa Heart Study

Alexander C Razavi; Lydia A Bazzano; Jiang He; Camilo Fernandez; Seamus P Whelton; Marie Krousel‐Wood; Shengxu Li; Jovia L Nierenberg; Mengyao Shi; Changwei Li; Xuenan Mi; Jason Kinchen; Tanika N Kelly

doi:10.1161/JAHA.119.015118

. 2020 Jan 29;9(3):e015118. doi: 10.1161/JAHA.119.015118

Novel Findings From a Metabolomics Study of Left Ventricular Diastolic Function: The Bogalusa Heart Study

Alexander C Razavi ^1,², Lydia A Bazzano ^1,², Jiang He ^1,², Camilo Fernandez ^1,², Seamus P Whelton ³, Marie Krousel‐Wood ^1,², Shengxu Li ⁴, Jovia L Nierenberg ¹, Mengyao Shi ¹, Changwei Li ⁵, Xuenan Mi ¹, Jason Kinchen ⁶, Tanika N Kelly ^1,^✉

PMCID: PMC7033875 PMID: 31992159

Abstract

Background

Diastolic dysfunction is one important causal factor for heart failure with preserved ejection fraction, yet the metabolic signature associated with this subclinical phenotype remains unknown.

Methods and Results

Ultra‐high‐performance liquid chromatography–tandem mass spectroscopy was used to conduct untargeted metabolomic analysis of fasting serum samples in 1050 white and black participants of the BHS (Bogalusa Heart Study). After quality control, 1202 metabolites were individually tested for association with 5 echocardiographic measures of left ventricular diastolic function using multivariable‐adjusted linear regression. Measures of left ventricular diastolic function included the ratio of peak early filling velocity to peak late filling velocity, ratio of peak early filling velocity to mitral annular velocity, deceleration time, isovolumic relaxation time, and left atrial maximum volume index (LAVI). Analyses adjusted for multiple cardiovascular disease risk factors and used Bonferroni‐corrected alpha thresholds. Eight metabolites robustly associated with left ventricular diastolic function in the overall population and demonstrated consistent associations in white and black study participants. N‐formylmethionine (B=0.05; P=1.50×10⁻⁷); 1‐methylhistidine (B=0.05; P=1.60×10⁻⁷); formiminoglutamate (B=0.07; P=5.60×10⁻⁷); N2, N5‐diacetylornithine (B=0.05; P=1.30×10⁻⁷); N‐trimethyl 5‐aminovalerate (B=0.04; P=5.10×10⁻⁶); 5‐methylthioadenosine (B=0.04; P=1.40×10⁻⁵); and methionine sulfoxide (B=0.04; P=3.80×10⁻⁶) were significantly associated with the natural log of the ratio of peak early filling velocity to mitral annular velocity. Butyrylcarnitine (B=3.18; P=2.10×10⁻⁶) was significantly associated with isovolumic relaxation time.

Conclusions

The current study identified novel findings of metabolite associations with left ventricular diastolic function, suggesting that the serum metabolome, and its underlying biological pathways, may be implicated in heart failure with preserved ejection fraction pathogenesis.

Keywords: biomarker, diastolic function, epidemiology, heart failure, metabolomics

Subject Categories: Heart Failure, Biomarkers, Metabolism, Epidemiology, Echocardiography

Clinical Perspective

What Is New?

The present study is the first population‐based, untargeted metabolomic analysis of left ventricular diastolic function in a biracial (35% black, 65% white) community‐based sample of middle‐aged adults.
We found that lipid‐ and amino acid–derived metabolites independently associate with patterns of mitral inflow and left ventricular relaxation in people without heart failure.

What Are the Clinical Implications?

This hypothesis‐generating study serves as an original groundwork to use metabolite biomarkers for the detection of subclinical heart failure in the general population.

Introduction

The prevalence of heart failure in the United States is projected to increase by over 40% to affect >8 million individuals by 2030.1, 2 Together with global epidemiologic trends, this estimated rise in prevalence classifies heart failure as the most rapidly increasing form of cardiovascular disease (CVD) in the United States and globally.3 Among all heart failure cases, heart failure with preserved ejection fraction (HFpEF) is becoming the major disease subtype, presently responsible for at least 50% of all heart failure cases.4, 5 Such data underline a crucial need to characterize and study subclinical HFpEF phenotypes, which may not only allow for early HFpEF risk prediction but also lead to the development of low‐risk interventions to delay or prevent heart failure onset, improving quality of life in an aging population.

Left ventricular diastolic dysfunction (LVDD), defined by impaired relaxation, decreased compliance, and/or increased left ventricular filling pressures, is one key subclinical precursor for HFpEF.6 This upstream phenotype represents a potential inflection point on the HFpEF causal disease pathway, as evidence suggests that LVDD is reversible.7, 8 Molecular phenotyping may represent a cost‐effective approach for identifying individuals with asymptomatic LVDD, while also implicating novel biological pathways in heart failure pathogenesis. Because metabolites reflect the collective output of both genetic and environmental factors9 and share a proximal relationship with clinical phenotypes, metabolomics profiling represents a particularly powerful molecular tool for such precision medicine initiatives. Unfortunately, few studies have applied this methodology to the study of LVDD.10, 11 Given the fact that both impaired systemic and myocardial metabolism have been implicated in HFpEF pathogenesis,12 there is a need to determine the role of systemic metabolites in left ventricular diastolic function (LVDF) to improve heart failure prevention.

The current study used metabolomics profiling and echocardiography to conduct a metabolome‐wide association study of LVDF in a cross‐sectional analysis of BHS (Bogalusa Heart Study) participants. We hypothesized that serum metabolites, particularly those involved in biological pathways involving fatty acid metabolism, would be associated with diastolic function.

Methods

The authors declare that all supporting data are available within the article and its online supplementary files.

Study Population

The BHS, seated in Bogalusa, Louisiana, is a population‐based study examining the natural course of CVD across the life span. Between 1973 and 2016, 7 surveys of children aged 4 to 17 as well as 10 surveys of adults, who had been previously observed as children, were completed. The current BHS cohort includes 1298 participants, born between January 1959 and June 1979, who were examined twice in both childhood and adulthood for CVD risk factors and outcomes. Among these 1298 BHS participants, we selected the 1050 individuals who had echocardiographic, metabolomic, and respective covariable data measured contemporaneously at the most recent BHS study examination taking place from 2013 to 2016 (Figure S1). The current study sample (n=1050) was highly similar to the overall BHS sample (n=1298) with respect to sociodemographic characteristics and traditional CVD risk factors (Table S1). All study participants provided written informed consent at each examination, and study protocols were approved the Institutional Review Board of the Tulane University Health Sciences Center.

Echocardiography Protocol

Two‐dimensional and tissue Doppler echocardiography were performed by trained cardiac sonographers at the BHS field office. Participants were placed in a partial left lateral decubitus position for echocardiographic assessment. Ten cycles of each 2‐dimensional and Doppler signal were recorded. Echocardiographic recordings were accomplished using an Aplio 300 ultrasound instrument (Toshiba America Medical Systems, Tustin, CA) with a linear array transducer of 7.5 mHz using a standard protocol.13, 14 Diastolic function parameters assessed included early left ventricular filling peak velocity (E), late left ventricular filling peak velocity (A), medial mitral annular velocity (e’), deceleration time of the E wave (DT), isovolumic relaxation time (IVRT), and left atrial maximum volume index.15 The apical 4‐chamber view was used to assess left atrial volume as well as patterns of mitral inflow including, DT, E wave, A wave, and medial e’ wave velocities. The E and A wave velocities were assessed using pulsed‐wave Doppler echocardiography of transmitral flow at the mitral valve leaflet tips, while medial e’ velocity was measured using pulsed‐wave tissue Doppler echocardiography of the mitral annulus. The apical 5‐chamber view was used to measure IVRT, specifically by placing the sample volume in the left ventricular outflow tract to concurrently evaluate aortic ejection and the onset of mitral inflow. Doppler sample volumes were placed between the mitral leaflet tips to measure DT. Left atrial volume was quantified by observing windows in which the left atrial length and transverse diameter were maximized. Left atrial volume was then divided by body surface area to estimate left atrial maximum volume index. Heart rate was assessed during the echocardiography protocol.

Metabolomic Profiling

Fasting serum samples were obtained between 2013 and 2016 from BHS participants and stored at −80°C prior to metabolite quantification. Metabolomic analysis of serum samples was consequently performed at Metabolon Inc (Durham, NC). Metabolites were quantified using ultra‐high‐performance liquid chromatography–tandem mass spectroscopy,16 providing a comprehensive analysis of serum metabolites. Metabolites were identified by automated comparison of ion characteristics to a reference library of chemical standard records that include retention time, molecular weight (m/z), preferred adducts, and in‐source fragments, as well as associated mass spectroscopy spectra. The majority of missing data in mass spectroscopy–based metabolomics experiments is due to detection limitations.17 Thus, if a metabolite value was missing, we assumed that it was below the detection limit and assigned it the lowest detectable value in the data set for the corresponding metabolite. Recent metabolomics studies have employed a similar technique.18, 19

Untargeted metabolomic analysis yielded 1466 metabolites, including 956 known biochemical compounds. These metabolites belonged to pathways related to amino acids (n=184), lipids (n=408), nucleotides (n=41), peptides (n=35), carbohydrates (n=25), cofactors and vitamins (n=34), xenobiotics (n=220), and energy (n=9). Furthermore, an additional 510 unnamed compounds presently lacking chemical standards were also identified. We excluded metabolites (n=213) detected in <20% of participants or with limited intra‐assay reliability, defined by a reliability coefficient <0.30 (n=51) in duplicate samples. After completion of quality control procedures, 1202 metabolites remained for analysis. Among these 1202 metabolites, 167 were missing or under the detection limit in 50% to 80% of the samples. These latter metabolites were categorized and analyzed according to 3 mutually exclusive groups: (1) missing or below the detection limit; (2) below the median of measured values; and (3) greater than or equal to the median.18, 19 For the remaining 1035 metabolites, the minimum observed value of each analyte was imputed for below‐the‐detection‐limit or missing values, with these metabolites examined as continuous variables.

Covariable Assessment

Rigorous protocols were employed to collect clinical data on BHS participants.20 Validated questionnaires were used to obtain demographic and lifestyle variables, specifically, age, race, sex, cigarette smoking, alcohol consumption, and antihypertensive medications. Fasting measures of low‐density lipoprotein cholesterol, high‐density lipoprotein cholesterol, triglycerides, and hemoglobin A_1c were collected using standardized methods.20 Blood pressure was measured in triplicate, while height and weight were measured in duplicate at the time of the physical exam. Weight in kilograms was divided by height in meters squared to calculate body mass index. Serum creatinine was measured using the Kinetic Jaffe method. Serum creatinine was used to calculate estimated glomerular filtration rate via the Chronic Kidney Disease Epidemiology Collaboration equation.21

Statistical Analysis

BHS study sample characteristics were presented as means and SDs for continuous variables, and categorical variables were expressed as percentages. The Student t test and chi‐square test were used to examine differences in normally distributed continuous variables and categorical variables, respectively, between blacks and whites. Normality of distribution was assessed via the Kolmogorov–Smirnov test. Both the E/A and E/e’ ratios underwent natural logarithmic transformation to normalize the distribution of these outcome variables before the analysis.

Before association analyses, we standardized the concentrations of metabolites by dividing each metabolite by its respective SD. Each metabolite was individually tested for association with the diastolic function parameters using multivariable‐adjusted linear regression models. In overall and race‐stratified multivariable models, 16 covariables were adjusted for, including age, sex, race (when appropriate), education, cigarette smoking, alcohol drinking, average systolic blood pressure, body mass index, low‐density lipoprotein cholesterol, high‐density lipoprotein cholesterol, serum triglycerides, hemoglobin A_1c, estimated glomerular filtration rate, heart rate, and antihypertensive medications. The analyses were corrected for multiple testing via the Bonferroni approach, with an alpha threshold of 4.16×10⁻⁵ (0.05/1202) used for determining statistical significance. To further reduce false‐positive associations, only findings significant in the overall analysis and in at least 1 race group, while displaying consistent effect direction and nominal significance (P<0.05) in the other race group, were reported in the current analysis. We conducted Pearson pairwise correlation tests to assess the relationships between metabolites that achieved significance after Bonferroni correction (P<4.16×10⁻⁵).

Weighted correlation network analysis (WGCNA)22 was employed to discover networks of highly correlated serum metabolites among BHS participants. WCGNA is an unsupervised data reduction technique, but unlike principal components analysis, it allows correlations between components, and thus may more effectively capture biological pathways underlying identified metabolites.22, 23 Previous metabolomic studies have used this data reduction technique.22, 24 In short, an adjacency matrix was generated to represent the metabolite network based on weighted pairwise‐correlations of all metabolites.25 Densely interconnected metabolites, also referred to as modules, were identified using an unsupervised hierarchical clustering approach.26 An eigenmetabolite was created for each metabolite module, representing the module's first principal component and can be interpreted as a weighted average value of correlated metabolites. Because there was similar metabolite clustering across ethnic groups, metabolite modules were generated among all study participants using data derived from the 1202 metabolites that passed quality control. To identify biological pathways underlying each module, metabolites most highly correlated with each module's eigenmetabolite were identified (r>0.70). We then used the subpathways associated with such metabolites to describe each module. For weighted correlation network analysis analyses, adjusted LVDF measures were created using the residual values generated by regressing each raw LVDF parameter on age, sex, race (when appropriate), education, cigarette smoking, alcohol drinking, average systolic blood pressure, body mass index, low‐density lipoprotein cholesterol, high‐density lipoprotein cholesterol, serum triglycerides, hemoglobin A_1c, estimated glomerular filtration rate, heart rate, and antihypertensive medications. We then examined the Pearson correlations among each of the 9 metabolite modules and the residual LVDF values, which were independent of the covariable effects. We used a Bonferroni‐corrected alpha threshold of 5.56×10⁻³ (0.05/9) to identify significant correlations among the 9 eigenmetabolite modules with the residual of each LVDF phenotype. Analogous to the single‐metabolite analysis, modules were considered significant upon achieving significance in the overall population and in at least 1 race group, with nominal significance (P<0.05) and consistent effect direction in the other race group. Statistical analyses were performed in SAS (version 9.4; SAS Institute, Cary, NC) and R (version 3.4.3) software.

Sensitivity Analysis

Four post hoc sensitivity analyses were performed, one of which assessed the relationship of metabolites with diastolic function additionally adjusting for left ventricular mass, a variable hypothesized to be on the causal pathway of diastolic dysfunction. A second sensitivity analysis similarly evaluated the relationship of metabolites with diastolic function, after removing BHS participants (n=12) who self‐reported previous myocardial infarction and/or a diagnosis of congestive heart failure. We also performed a sensitivity analysis additionally adjusting for gait speed, one surrogate marker of skeletal muscle breakdown, to help elucidate if significant endogenous metabolites related to myocyte turnover were of skeletal or cardiac origin. Finally, we conducted a sensitivity analysis adjusting for lifetime mean systolic blood pressure, as opposed to the systolic blood pressure measurement at the most recent examination, in a subsample of study participants. Study participants (n=804) who had at least 4 measures of systolic blood pressure from previous BHS exams were included, and these 4 measures were used to calculate a mean lifetime systolic blood pressure, which was included as a covariate in linear regression models.

Results

Table 1 presents the characteristics of the 1050 participants, including information on sociodemographic, lifestyle, medication, metabolic, and hemodynamic factors. The average age of the study cohort was 48 years, composed of 58% women and over one third black individuals. The study sample, on average, was obese (body mass index=31.38±7.80 kg/m²) and prediabetic (hemoglobin A_1c=5.87±1.19), with evidence of dyslipidemia (low‐density lipoprotein cholesterol=115.42±35.89 mg/dL), elevated systolic blood pressure (systolic blood pressure=123.32±16.69 mm Hg), and preserved kidney function (estimated glomerular filtration rate=93.34±17.18 mL/min per 1.73 m²). There were significant differences between blacks and whites with respect to all covariables except for sex and resting heart rate.

Table 1.

Characteristics of 1050 Bogalusa Metabolome Study Participants

Variable	All (n=1050)	Whites (n=684)	Blacks (n=366)	P Value^a
Sociodemographic
Female, %	57.81	56.14	60.93	0.1668
Post–high school education, %	49.90	56.58	37.43	<0.0001
Age, y, mean (SD)	48.13 (5.32)	48.46 (5.09)	47.52 (5.70)	0.0084
Lifestyle
Smoking, n (%)				0.0352
Never	529 (50.38)	353 (51.61)	176 (48.09)
Former	301 (28.67)	204 (29.82)	97 (26.50)
Current	220 (20.95)	127 (18.57)	93 (25.41)
Drinking, n (%)				0.0138
Never	116 (11.05)	60 (8.77)	56 (15.30)
Former	332 (31.62)	219 (32.02)	113 (30.87)
Current	602 (57.33)	405 (59.21)	197 (53.83)
Metabolic
BMI, kg/m², mean (SD)	31.38 (7.80)	30.4 (6.94)	33.34 (8.86)	<0.0001
LDL cholesterol, mg/dL, mean (SD)	115.42 (35.89)	117.20 (33.93)	112.08 (39.12)	0.0350
HDL cholesterol, mg/dL, mean (SD)	51.88 (16.30)	50.97 (16.36)	53.57 (16.06)	0.0133
Hemoglobin A_1c, %, mean (SD)	5.87 (1.19)	5.72 (0.91)	6.15 (1.54)	<0.0001
Serum triglycerides, mg/dL, mean (SD)	126.76 (68.15)	134.83 (71.48)	111.69 (58.61)	<0.0001
Hemodynamic
Systolic BP, mm Hg, mean (SD)	123.32 (16.69)	121.13 (14.78)	127.42 (19.14)	<0.0001
Antihypertensive medication, n (%)	359 (34.19)	190 (27.78)	169 (46.17)	<0.001
eGFR, mL/min per 1.73 m², mean (SD)	93.34 (17.18)	90.71 (15.02)	98.26 (19.71)	<0.0001
Cardiac
Resting heart rate, beats/min, mean (SD)	71.67 (11.42)	71.39 (11.09)	72.19 (12.01)	0.2782
E/A ratio (SD)^b	1.12 (1.34)	1.14 (1.32)	1.08 (1.35)	0.0227
E/e’ ratio (SD)^b	7.54 (1.32)	7.24 (1.30)	8.00 (1.36)	<0.0001
Isovolumic relaxation time, ms (SD)	93.49 (22.36)	89.86 (21.26)	100.26 (22.79)	<0.0001
Deceleration time, ms (SD)	186.61 (45.15)	188.81 (44.28)	184.76 (42.72)	0.0418
Left atrial maximum volume index (SD)	28.85 (8.25)	28.10 (7.90)	30.26 (8.70)	<0.0001

Open in a new tab

BMI indicates body mass index; BP, blood pressure; E/A, peak early filling velocity to peak late filling velocity; E/e’, peak early filling velocity to mitral annular velocity eGFR, estimated glomerular filtration rate; HDL, high‐density lipoprotein; LDL, low‐density lipoprotein.

P value for comparison between white vs. black participants.

Presented as geometric mean and SD.

LVDF of the study sample is also shown in Table 1. Compared with blacks, whites had statistically significant preserved diastolic function for patterns of mitral inflow, E/A (P=0.0227) and E/e’ ratio (P<0.0001), left atrial maximum volume index (P<0.0001), and left ventricular relaxation, via IVRT (P<0.0001). On the other hand, whites had significantly higher values for DT (P=0.0418), compared with blacks.

There were 8 serum metabolites, 7 amino acid derived and 1 lipid derived, that were associated with diastolic function (Figure 1; Table 2). The 7 amino acid–derived metabolites were associated with the log‐transformed E/e’ ratio phenotype and represented several different biological pathways, including histidine, lysine, methionine, polyamine metabolism, and the urea cycle. Among E/e’ ratio–related metabolites, formiminoglutamate (B=0.07; P=5.60×10⁻⁷) had the largest effect size, followed by N‐formylmethionine (B=0.05; P=1.50×10⁻⁷); 1‐methylhistidine (B=0.05; P=1.60×10⁻⁷); N2, N5‐diacetylornithine (B=0.05; P=1.30×10⁻⁷); methionine sulfoxide (B=0.04; P=3.80×10⁻⁶); 5‐methylthioadenosine (B=0.04; P=1.40×10⁻⁷); and N‐trimethyl 5‐amino‐valerate (B=0.04; P=5.10×10⁻⁶). Butyrylcarnitine, a lipid‐derived metabolite, was significantly associated with IVRT (B=3.18; P=2.10×10⁻⁶). Of note and with the exception of methionine sulfoxide, all significant associations of metabolites with the log‐transformed E/e’ ratio were driven by significance in blacks, with nominal significance in whites (Table S2). The reverse of this latter pattern was observed for the diastolic function parameter of IVRT.

Proposed biological relationships between identified metabolites and left ventricular diastolic dysfunction. Arrows indicate proposed direction(s) of biological relationships.

Table 2.

Novel Metabolites Significantly Associated With Echocardiographic Measures of Diastolic Function

Metabolite	Sub‐Pathway	Ln E/e’ Ratio		Ln E/A Ratio		IVRT (ms)		DT (ms)		LAVI (mL/m²)
Metabolite	Sub‐Pathway	β (SE)	P Value	β (SE)	P Value	β (SE)	P Value	β (SE)	P Value	β (SE)	P Value
Amino acid
1‐methylhistidine	Histidine metabolism	0.05 (0.01)	1.60E‐07^a	0.03 (0.01)	1.10E‐03^b	1.41 (0.75)	6.10E‐02	1.35 (1.63)	4.10E‐01	0.40 (0.32)	2.20E‐01
Formiminoglutamate	Histidine metabolism	0.07 (0.01)	5.60E‐07^a	0.03 (0.01)	4.50E‐02^b	2.68 (1.13)	1.90E‐02^b	0.94 (2.46)	7.00E‐01	−0.09 (0.47)	8.40E‐01
5‐methylthioadenosine	Poylamine metabolism	0.04 (0.01)	1.40E‐05^a	0.03 (0.01)	1.10E‐03^b	0.10 (0.74)	8.90E‐01	2.18 (1.60)	1.70E‐01	0.26 (0.28)	3.60E‐01
N‐trimethyl 5‐amino‐valerate	Lysine metabolism	0.04 (0.01)	5.10E‐06^a	0.02 (0.01)	2.00E‐02^b	1.91 (0.75)	1.10E‐02^b	0.15 (1.62)	9.30E‐01	0.19 (0.28)	5.00E‐01
N2, N5‐diacetylornithine	Urea cycle; arginine and proline metabolism	0.05 (0.01)	1.30E‐07^a	0.03 (0.01)	8.50E‐04^b	1.62 (0.70)	2.10E‐02^b	1.67 (1.51)	2.70E‐01	0.16 (0.31)	6.00E‐01
N‐formylmethionine	Methionine, cysteine, SAM, and taurine metabolism	0.05 (0.01)	1.50E‐07^a	0.04 (0.01)	1.30E‐05^a	1.00 (0.78)	2.00E‐01	2.10 (1.69)	2.10E‐01	0.58 (0.29)	4.70E‐02^b
Methionine sulfoxide	Methionine, cysteine, SAM, and taurine metabolism	0.04 (0.01)	3.80E‐06^a	0.03 (0.01)	1.30E‐03^b	0.55 (0.67)	4.10E‐01	0.90 (1.44)	5.30E‐01	0.55 (0.24)	2.50E‐02^b
Lipid
Butyrylcarnitine	Fatty acid metabolism, BCAA metabolism	0.03 (0.01)	1.40E‐04^b	−0.01 (0.01)	2.60E‐01	3.18 (0.67)	2.10E‐06^a	−1.69 (1.46)	2.50E‐01	−0.33 (0.25)	1.80E‐01

Open in a new tab

Adjusted for age, race, sex, education, smoking, alcohol drinking, antihypertensive medications, heart rate, estimated glomerular filtration rate, body mass index, systolic blood pressure, low‐density lipoprotein cholesterol, high‐density lipoprotein cholesterol, serum triglycerides, and hemoglobin A_1c. E indicates peak velocity blood flow in early diastole due to gravity; e’, medial mitral annular velocity; IVRT, isovolumic relaxation time; Ln, natural logarithm; SAM, s‐adenosylmethionine.

Statistically significant P value (P<4.16E‐05).

Nominally significant P value (P<0.05).

Associations of each metabolite with the diastolic function phenotypes, according to super pathway, are displayed in Figures S2 through S6. Such plots are depicted for the overall sample, blacks, and whites for each respective diastolic function outcome. These plots also provide the names of metabolites that were significant for diastolic function parameters in blacks or whites, but not in the overall sample, and thus shed light on potential race‐specific associations of metabolites with LVDF.

Several metabolites that met stringent Bonferroni significance‐level criteria for their association with E/e’ ratio had nominally significant associations with other diastolic function phenotypes (Table 2). N‐formylmethionine shared the most robust associations with LVDF, as this amino acid–derived metabolite was significantly associated with the log‐transformed E/A ratio (B=0.04; P=1.30×10⁻⁵), and had a nominally significant association with left atrial maximum volume index (B=0.58; P=4.70×10⁻²). Methionine sulfoxide also had nominally significant associations with the log‐transformed E/A ratio (B=0.03; P=1.30×10⁻³) and left atrial maximum volume index (B=0.55; 2.50×10⁻²), although with lower parameter estimates when compared to N‐formylmethionine. Finally, 3 metabolites, formiminoglutamate, N‐trimethyl 5‐amino‐valerate, and N‐trimethyl 5‐amino‐valerate, demonstrated nominally significant associations with both IVRT and the log‐transformed E/A ratio.

Significant metabolites demonstrated modest to moderate intercorrelation (Figure 2). The highest pairwise correlation coefficients were observed between N‐formylmethionine and 1‐methylhistidine (r=0.64) and between N‐formylmethionine and 5‐methylthioadenosine (r=0.53). With respect to biological pathways, the 9 metabolite modules identified among study participants are presented in Tables S3 through S7. A module containing metabolites involved in primary and secondary bile acid metabolism demonstrated a significant positive correlation with E/e’ in the overall population (r=0.16; P<0.0001) and in blacks (r=0.27; P<0.0001), with a consistent effect direction in whites (r=0.06; P=0.09) (Table S3). Figure S7 presents the Pearson correlations of the 8 metabolites that were most highly correlated with the primary and secondary bile acid metabolism eigenmetabolite value (r>0.70).

Heat map displaying pairwise correlation coefficients for metabolites significantly associated left ventricular diastolic function. 1‐MH indicates 1‐methylhistidine; 5‐MTA, 5‐methylthioadenosine; BC, butyrylcarnitine; FG, formiminoglutamate; MSO, methionine sulfoxide; N2, N5‐DAO, N2, N5‐diacetylornithine; N‐FM, N‐formylmethionine; TAV, N‐trimethyl 5‐aminovalerate.

Results from sensitivity analyses are presented in Table S8, excluding individuals with self‐report myocardial infarction or congestive heart failure; Table S9, adding left ventricular mass to the multivariable regression model; Table S10, adding gait speed to the multivariable regression model; and Table S11, adjusting for lifetime mean systolic blood pressure. Beta estimates and P values in all 3 sensitivity analyses were consistent with primary study results depicted in Table 2. N‐formylmethionine maintained its effect size with the log‐transformed E/e’ phenotype when excluding individuals with self‐report myocardial infarction or congestive heart failure (B=0.05; P=2.60×10⁻⁷), adjusting for mean lifetime systolic blood pressure (B=0.04; P=4.10×10⁻⁴), and when adding left ventricular mass (B=0.05 P=4.80×10⁻⁶) or gait speed (B=0.04; P=3.90×10⁻⁵) to the multivariable regression model. For each standardized unit increase in butyrylcarnitine, we observed a significant 3‐ms higher IVRT across all sensitivity analyses, except for the sensitivity analysis that controlled for mean lifetime systolic blood pressure.

Discussion

While the increasing availability of noninvasive imaging methods has allowed for the assessment of LVDF in broader populations, the biochemical profile associated with this upstream heart failure (HF) phenotype remains largely unexplored. To address this critical gap in CVD prevention knowledge, we conducted one of the first epidemiologic studies that assesses the role of serum metabolites in LVDF, via an untargeted approach. Formiminoglutamate; 1‐methylhistidine; N2, N5‐diacetylornithine; N‐trimethyl 5‐aminovalerate; N‐formylmethionine; 5‐methylthioadenosine; and methionine sulfoxide were positively associated with the E/e’ ratio, while butyrylcarnitine had a significant positive association with IVRT. In addition to single‐metabolite analyses, a network of correlated metabolites involved in primary and secondary bile metabolism was associated with the E/e’ ratio. These results build on previous smaller, targeted metabolomic studies of diastolic function10, 11 and suggest that a higher concentration of both lipid and amino acid–derived metabolites, particularly those associated with mitochondrial impairment, may serve as early biomarkers for preclinical LVDD. Given the diagnostic and therapeutic challenges associated with HFpEF,27, 28, 29 findings from the current study may serve as a platform to develop more precise biological assays and risk stratification protocols for HFpEF.

Findings from our analysis suggest that amino acid–derived metabolites hold a close relationship with the E/e’ ratio, one indicator of left ventricular filling.30 The E/e’ ratio has emerged as one promising echocardiographic parameter to help facilitate the diagnosis of HFpEF,31 as evident by its presence in the H₂FpEF (F=filling pressure, Doppler echocardiographic E/e’ >9 equates to 1 point) score clinically recommended to differentiate cardiac versus noncardiac causes of heart failure.32 While IVRT has not been incorporated into diagnostic scoring criteria for HFpEF, increases in this cardiac time interval have been independently associated with risk of future ischemic heart disease, peripheral artery disease, and stroke.33 A similar relationship has also been described for increases in E/e’ and the risk of incident primary cardiac events.34 Below, we discuss the relevance of each of the 8 robustly identified metabolites in LVDD and outline potential pathophysiological mechanisms underlying such associations by referencing previous work in this research space.

Butyrylcarnitine and N‐Formylmethionine: Mitochondrial Impairment

We speculate that mitochondrial uncoupling is driving metabolic alterations in subclinical HF, such that asymptomatic LVDD is associated with mitochondrial impairment and leads to elevations in serum butyrylcarnitine and N‐formylmethionine. In the present study, butyrylcarnitine and N‐formylmethionine were positively associated with IVRT and E/e’, respectively. Butyrylcarnitine is a member of the acylcarnitine family, a group of metabolites functioning in fatty acid metabolism, particularly in mitochondrial fatty acid beta oxidation.35 Likewise, N‐formylmethionine is an initiator of protein synthesis in the mitochondria of all eukaryotic cells.36 Unlike other tissues and organs, fatty acids produce between 70% and 90% of ATP needed for heart function and metabolism.37, 38 Heart failure pathophysiology attributable to mitochondrial dysfunction is associated with both impaired fatty acid and glucose oxidation.37, 38, 39 Our results support previous work in this research space, which has demonstrated that individuals with HF have a higher venous concentration of mitochondrial‐dependent metabolites, succinate, 3‐hydroxybutyrate, and acetone, compared with controls.40 Notably, succinate, 3‐hydroxybutyrate, and acetone share a positive association of myocardial energy metabolism, further underlining the central role of altered mitochondrial metabolism in HF.40

Mitochondrial impairment may also trigger apoptosis, causing cardiac myocyte N‐formylmethionine to leak into circulation. These hypotheses are in part supported by research demonstrating that individuals with HFpEF have higher systemic concentrations of acylcarnitines compared with age‐matched non–heart failure controls.41, 42 Similarly, a 4‐component metabolite panel, including butyrylcarnitine, was found to better predict heart failure prognosis in individuals with stage B and stage C, when compared with brain natriuretic peptide.43 Systemic N‐formylmethionine release may occur after myocardial infarction,44 suggesting that ischemia may be one contributing factor in cardiomyocyte mitochondrial impairment in the setting of LVDD. These latter results, together with the findings of the present study, highlight the potential diagnostic value of serum butyrylcarnitine and N‐formylmethionine in LVDD and HFpEF.

1‐Methylhistidine and N‐Trimethyl 5‐Aminovalerate: Diet

Exogenous environmental factors, specifically diet, have significant effects on the human metabolome.45 Our findings demonstrated that 2 diet‐associated metabolites, 1‐methylhistidine and N‐trimethyl 5‐aminovalerate, were positively correlated with the E/e’ ratio. Derived from dietary anserine sources, 1‐methylhistidine has been reported as a marker of meat consumption, specifically red meat and chicken. Several previous studies have found robust linear relationships between meat intake and urinary 1‐methylhistidine.46, 47 On the other hand, external evidence suggests that N‐trimethyl 5‐aminovalerate may be a metabolite indicator of dairy milk intake.48, 49 Importantly, restriction of both dairy and red meat consumption, via the Dietary Approaches to Stop Hypertension and Mediterranean diets, are associated with a reduced risk of developing heart failure.50, 51, 52 While systolic blood pressure reduction may be the primary mechanism of the protective HF effects of these 2 diets, it is also important to consider other CVD risk factors in the setting of our metabolite findings. For example, N‐trimethyl 5‐aminovalerate positively associates with abdominal adiposity and intramuscular adiposity,53 risk factors for insulin resistance, type 2 diabetes mellitus, and coronary microvascular dysfunction. Notably, type 2 diabetes mellitus is a common comorbid disease that is present in up to one half of all patients with HFpEF.54 Thus, our work serves as preliminary biological evidence that may help explain the potential benefits of animal protein restriction in atherosclerosis and HF risk reduction. However, future longitudinal studies that combine nutritional and metabolome data with diastolic function are undoubtedly required to understand and test such hypotheses and biological phenomena.

N2, N5‐Diacetylornithine and 5‐Methylthioadenosine: Endothelial Dysfunction

N2, N5‐diacetylornithine is a metabolite by‐product of the urea cycle, an important biological pathway to consider in the setting of HFpEF and diastolic function. Compared with individuals with HF with reduced ejection fraction, those with HFpEF have a higher chronic kidney disease–associated mortality rate.54 Urea‐to‐creatinine ratio, and urinary albumin‐to‐creatinine ratio, biochemical indices of renal function, associate with parameters of LVDD.55 Although BHS participants in the current study had preserved kidney function, it is possible that N2, N5‐diacetylornithine levels begin to accumulate due to urea cycle perturbations in subclinical renal disease, which may be accompanied by systemic microvascular and endothelial dysfunction.56

Findings from our study lead to the postulation that subclinical LVDD attributable to underlying endothelial dysfunction is associated with impaired nitric oxide synthesis and may lead to an elevation of systemic polyamines particularly, 5‐methylthioadenosine. Present in all mammalian tissues, 5‐methylthioadenosine is a polyamine and sulfur‐based nucleoside.57 The 5‐methylthioadenosine/methionine and polyamine pathways have previously been implicated in heart failure among a sample of 515 individuals.43 Additionally, animal models of ventricular hypertrophy and aortic stenosis demonstrate increases in systemic polyamines after tissue remodeling.58, 59 Nitric oxide, a vasodilatory and protective metabolite in CVD, inhibits polyamine production.60

Methionine Sulfoxide: Oxidative Stress

Methionine sulfoxide reductase is a powerful antioxidant that reduces methionine sulfoxide61 and has been implicated in vascular disease and cardiac ischemia.62 We observed a significant positive association between methionine sulfoxide and left ventricular filling. Methionine oxide results from the oxidation of methionine and is produced from activated neutrophils. Present in both the cytoplasm and mitochondria, methionine sulfoxide may be released from cardiac myocyte turnover in the presence of LVDD or systemic endothelial dysfunction.

Formiminoglutamate: Impaired Liver Metabolism

Formiminoglutamate is an intermediate of histidine metabolism, and high levels of this metabolite have been traditionally used to identify vitamin B₁₂ and/or folate deficiency, as well as underlying liver disease.63, 64 Such nutritional deficiencies have been hypothesized to contribute to anemia observed in patients with congestive HF, yet scarce data exist. Megaloblastic anemia caused by vitamin B₁₂ and/or folate deficiency are relatively rare in heart failure65, 66; therefore, the positive relationship identified between formiminoglutamate and E/e’ may point more toward co‐occurrence of subclinical liver disease, as opposed to anemia, in diastolic dysfunction and HF. Folate and vitamin B₁₂ deficiency do not currently predict prognosis in heart failure66; therefore, our results should be considered only as hypothesis‐generating observations and exploratory in nature. Furthermore, though our analyses adjusted for alcohol intake in BHS participants, regular alcohol consumption associates with inferior diastolic function67 and leads to higher urinary excretion of formiminoglutamate.64

In addition to single‐metabolite analyses, aggregation approaches also provided novel insights. Among metabolites with a significant relationship with LVDF, N‐formylmethionine and 1‐methylhistidine, as well as N‐formylmethionine and 5‐methylthioadenosine demonstrated moderate pairwise correlation. These results suggest that there may be an interaction between exogenous and endogenous pathways that influences the role of metabolites in diastolic dysfunction. Likewise, a correlated network of metabolites involved in primary and secondary bile acid metabolism significantly was associated with E/e’ in the overall population and in blacks and had a consistent effect direction in whites. In addition to implicating the role of bile acid metabolites, such as glycohyocholate and taurocholate, with E/e’, our results may further underline the importance of E/e’ and left ventricular filling in the natural course of diastolic dysfunction and HFpEF. Our findings are consistent with and build on current knowledge regarding the role of bile acids in HF. The ratio of secondary to primary bile acids is increased in individuals with HF and this pattern has been associated with a reduced HF survival time.68 Likewise, an increased level of serum bile acids, together with diastolic dysfunction has been well described in cirrhotic cardiomyopathy.69

Previous Mechanistic Studies: Branched Chain Amino Acid Metabolism

Much of the existing literature regarding the metabolomics of HF has assumed a biomarker‐based approach, assessing the ability of metabolites to diagnose and predict disease prognosis. The identification of biomarkers associated with HF phenotypes, however, has also provided valuable mechanistic information on the pathophysiology of HFpEF. On the whole, impaired fatty acid and glucose oxidation, urea cycle dysfunction, and anabolic‐catabolic imbalance are main pathophysiological processes underlying HF.70 Ventricular filling is an energy‐intensive process,71 and recent mechanistic studies have implicated branched chain amino acids (BCAAs), leucine, isoleucine, and valine, in the modulation of left ventricular function.72, 73, 74, 75 Branched‐chain keto acid accumulation in cardiomyocytes has been documented in mitochondrial targeted 2C‐type serine/threonine protein phosphatase knockout mice75 as well as in humans with dilated cardiomyopathy.74 On the other hand, animal models also suggest that supplementation with BCAA may improve cardiomyocyte survival and preserve ventricular function at middle age.73 Thus, while a broad physiological range of circulating BCAA concentrations may exist, these amino acids may become detrimental when BCAA catabolism is impaired in subclinical or overt HF. Hypertrophic hearts in particular exhibit a reduction in mitochondrial targeted 2C‐type serine/threonine protein phosphatase mRNA and protein expression,75 and future longitudinal omics studies of mitochondrial targeted 2C‐type serine/threonine protein phosphatase and LVDD may be valuable to further elucidate the biological relationships of BCAA metabolism, mitochondrial function, and diastolic function. We found that butyrylcarnitine, a lipid metabolite involved in BCAA metabolism, had a positive association with IVRT; an association that warrants further investigation.

Conclusions

This study had several important strengths. To our knowledge, this is one of the first metabolome‐wide association studies of LVDF. Furthermore, identified metabolites demonstrated consistent associations with diastolic function across race groups and in sensitivity analyses adjusting for left ventricular mass and gait speed and excluding individuals with self‐reported myocardial infarction and/or congestive HF. On the other hand, our study was not without limitations. Its cross‐sectional design prohibits temporal inferences. Nevertheless, metabolites reported here may serve as biomarkers for LVDD, making them particularly relevant for future metabolomics studies involving HFpEF. Furthermore, to minimize false‐positive findings, only metabolites that demonstrated consistent associations with diastolic function across race groups were reported here. Therefore, our study may miss race‐specific metabolite–diastolic function associations. To address this limitation, we reported all significant race‐specific metabolites in Figures S2 through S4, to inform future studies and research regarding the potential effect modifying role of race on the association of metabolites and LVDF.

In conclusion, we have observed that lipid‐ and amino acid–derived metabolites associate with LVDF in a racially diverse cohort of middle‐aged adults. Butyrylcarnitine was positively correlated with left ventricular relaxation, while N‐formylmethionine; formiminoglutamate; N2, N5‐diacetylornithine; N‐trimethyl 5‐aminovalerate; 5‐methylthioadenosine; and methionine sulfoxide had positive associations with left ventricular filling pressure. Such metabolites may serve as early biomarkers for subclinical HFpEF in the general population. Future replication and longitudinal studies are required to further understand the role of serum metabolites in LVDF and progression to HFpEF.

Sources of Funding

This research was supported by the National Institute on Aging as well as the National Heart, Lung, and Blood Institute of the National Institutes of Health under grant numbers R21AG051914 (Principal Investigator, Kelly), 5R01AG041200‐05 (Principal Investigator, Bazzano), and P20GM109036 (Principal Investigator, He). Dr Razavi is currently funded through a fellowship training grant supported by the National Heart, Lung, and Blood Institute of the National Institutes of Health under grant number F30HL147486.

Disclosures

Dr Kinchen, who contributed to metabolomics data collection, is employed by Metabolon Inc and did not participate in the study design, statistical analysis, or scientific interpretation of the current research. The remaining authors have no disclosures to report.

Supporting information

Table S1. Comparison of Bogalusa Heart Study Samples

Table S2. Novel Metabolites Significantly Associated With Echocardiographic Measures of Diastolic Function, Stratified by Race

Table S3. Correlations of Nine Metabolite Modules From WGCNA Analysis With the E/e’ Ratio

Table S4. Correlations of Nine Metabolite Modules From WGCNA Analysis With the E/A Ratio

Table S5. Correlations of Nine Metabolites Modules From WGCNA Analysis With Isovolumic Relaxation Time (IVRT)

Table S6. Correlations of Nine Metabolite Modules From WGCNA Analysis With Deceleration Time (DT)

Table S7. Correlations of Nine Metabolite Modules With Left Atrial Maximum Volume Index (LAVI)

Table S8. Novel Metabolites Significantly Associated With Echocardiographic Measures of Diastolic Function in 1035 BHS Participants, Excluding Individuals With CHF or MI

Table S9. Novel Metabolites Significantly Associated With Echocardiographic Measures of Diastolic Function in 1050 BHS Participants, Additionally Adjusting for Left Ventricular Mass

Table S10. Novel Metabolites Significantly Associated With Echocardiographic Measures of Diastolic Function in 1050 BHS Participants, Additionally Adjusting for Gait Speed

Table S11. Novel Metabolites Significantly Associated With Echocardiographic Measures of Diastolic Function in 804 BHS Participants, Adjusting for Mean Lifetime Systolic Blood Pressure

Figure S1. Flow diagram of study participant inclusion.

Figure S2. −Log10(P) probability plot of the multivariable‐adjusted association of serum metabolites with the natural logarithm of the E/A ratio: all participants (left), African Americans (middle), whites (right).

Figure S3. −Log10(P) probability plot of the multivariable‐adjusted association of serum metabolites with the natural logarithm of the E/e’ ratio: all particpants (left), African Americans (middle), whites (right).

Figure S4. −Log10(P) probability plot of the multivariable‐adjusted association of serum metabolites with isovolumic relaxation time: all participants (left), African Americans (middle), whites (right).

Figure S5. −Log10(P) probability plot of the multivariable‐adjusted association of serum metabolites with deceleration time: all participants (left), African Americans (middle), whites (right).

Figure S6. −Log10(P) probability plot of the multivariable‐adjusted association of serum metabolites with left atrial volume index: all participants (left), African Americans (middle), whites (right).

Figure S7. Heat map displaying pairwise correlation coefficients for the network of metabolites underlying primary and secondary bile acid metabolism pathways.

Click here for additional data file.^{(1.7MB, pdf)}

Acknowledgments

The authors thank all staff members and study personnel who help conduct, sustain, and continue the BHS. The authors are especially grateful to the BHS study participants.

(J Am Heart Assoc. 2020;9:e015118 DOI: 10.1161/JAHA.119.015118.)

References

1. Ezekowitz BJ. Forecasting the future of cardiovascular disease in the United States: a policy statement from the American Heart Association. Circulation. 2011;123:933–944. [DOI] [PubMed] [Google Scholar]
2. Vigen R, Maddox TM, Allen LA. Aging of the United States population: impact on heart failure. Curr Heart Fail Rep. 2012;9:369–374. [DOI] [PMC free article] [PubMed] [Google Scholar]
3. Ziaeian B, Fonarow GC. Epidemiology and aetiology of heart failure. Nat Rev Cardiol. 2016;13:368. [DOI] [PMC free article] [PubMed] [Google Scholar]
4. Dunlay SM, Roger VL, Redfield MM. Epidemiology of heart failure with preserved ejection fraction. Nat Rev Cardiol. 2017;14:591. [DOI] [PubMed] [Google Scholar]
5. Oktay AA, Rich JD, Shah SJ. The emerging epidemic of heart failure with preserved ejection fraction. Curr Heart Fail Rep. 2013;10:401–410. [DOI] [PMC free article] [PubMed] [Google Scholar]
6. Zile MR, Baicu CF, Bonnema DD. Diastolic heart failure: definitions and terminology. Prog Cardiovasc Dis. 2005;47:307–313. [DOI] [PubMed] [Google Scholar]
7. Lee CH, Hung KC, Chang SH, Lin FC, Hsieh MJ, Chen CC, Chu CM, Hsieh IC, Wen MS, Wu D. Reversible left ventricular diastolic dysfunction on Doppler tissue imaging predicts a more favorable prognosis in chronic heart failure. Circ J. 2012;76:1145–1150. [DOI] [PubMed] [Google Scholar]
8. Gorcsan J III, Diana P, Lee J, Katz WE, Hattler BG. Reversible diastolic dysfunction after successful coronary artery bypass surgery: assessment by transesophageal Doppler echocardiography. Chest. 1994;106:1364–1369. [DOI] [PubMed] [Google Scholar]
9. Deidda M, Piras C, Bassareo PP, Dessalvi CC, Mercuro G. Metabolomics, a promising approach to translational research in cardiology. IJC Metab Endocr. 2015;9:31–38. [Google Scholar]
10. Zhang ZY, Marrachelli VG, Yang WY, Trenson S, Huang QF, Wei FF, Thijs L, Van Keer J, Monleon D, Verhamme P, Voigt JU. Diastolic left ventricular function in relation to circulating metabolic biomarkers in a population study. Eur J Prev Cardiol. 2019;26:22–32. [DOI] [PubMed] [Google Scholar]
11. Zhang ZY, Marrachelli VG, Thijs L, Yang WY, Wei FF, Monleon D, Jacobs L, Nawrot T, Verhamme P, Voigt JU, Kuznetsova T. Diastolic left ventricular function in relation to circulating metabolic biomarkers in a general population. J Am Heart Assoc. 2016;5:e002681 DOI: 10.1161/JAHA.115.002681. [DOI] [PMC free article] [PubMed] [Google Scholar]
12. Albert CL, Tang WW. Metabolic biomarkers in heart failure. Heart Fail Clin. 2018;14:109–118. [DOI] [PMC free article] [PubMed] [Google Scholar]
13. Bazzano LA, Belame SN, Patel DA, Chen W, Srinivasan S, McIlwain E, Berenson GS. Obesity and left ventricular dilatation in young adulthood: the Bogalusa Heart Study. Clin Cardiol. 2011;34:153–159. [DOI] [PMC free article] [PubMed] [Google Scholar]
14. Sahn DJ, DeMaria AN, Kisslo JO, Weyman AF. Recommendations regarding quantitation in M‐mode echocardiography: results of a survey of echocardiographic measurements. Circulation. 1978;58:1072–1083. [DOI] [PubMed] [Google Scholar]
15. Nagueh SF, Smiseth OA, Appleton CP, Byrd BF, Dokainish H, Edvardsen T, Flachskampf FA, Gillebert TC, Klein AL, Lancellotti P, Marino P. Recommendations for the evaluation of left ventricular diastolic function by echocardiography: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging. Eur J Echocardiogr. 2016;17:1321–1360. [DOI] [PubMed] [Google Scholar]
16. Evans AM, DeHaven CD, Barrett T, Mitchell M, Milgram E. Integrated, nontargeted ultrahigh performance liquid chromatography/electrospray ionization tandem mass spectrometry platform for the identification and relative quantification of the small‐molecule complement of biological systems. Anal Chem. 2009;81:6656–6667. [DOI] [PubMed] [Google Scholar]
17. Xia J, Mandal R, Sinelnikov IV, Broadhurst D, Wishart DS. MetaboAnalyst 2.0—a comprehensive server for metabolomic data analysis. Nucleic Acids Res. 2012;40:W127–W133. [DOI] [PMC free article] [PubMed] [Google Scholar]
18. Zheng Y, Yu B, Alexander D, Mosley TH, Heiss G, Nettleton JA, Boerwinkle E. Metabolomics and incident hypertension among blacks: the Atherosclerosis Risk in Communities study. Hypertension. 2013;62:398–403. [DOI] [PMC free article] [PubMed] [Google Scholar]
19. Li C, He J, Li S, Chen W, Bazzano L, Sun X, Shen L, Liang L, Shen Y, Gu X, Kelly TN. Novel metabolites are associated with augmentation index and pulse wave velocity: findings from the Bogalusa Heart Study. Am J Hypertens. 2019;32:547–556. [DOI] [PMC free article] [PubMed] [Google Scholar]
20. Foster TA, Berenson GS. Measurement error and reliability in four pediatric cross‐sectional surveys of cardiovascular disease risk factor variables—the Bogalusa Heart Study. J Chronic Dis. 1987;40:13–21. [DOI] [PubMed] [Google Scholar]
21. Levey AS, Stevens LA. Estimating GFR using the CKD Epidemiology Collaboration (CKD‐EPI) creatinine equation: more accurate GFR estimates, lower CKD prevalence estimates, and better risk predictions. Am J Kidney Dis. 2010;55:622. [DOI] [PMC free article] [PubMed] [Google Scholar]
22. Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics. 2008;9:559. [DOI] [PMC free article] [PubMed] [Google Scholar]
23. Worley B, Powers R. Multivariate analysis in metabolomics. Curr Metabolomics. 2013;1:92–107. [DOI] [PMC free article] [PubMed] [Google Scholar]
24. Pedersen HK, Gudmundsdottir V, Nielsen HB, Hyotylainen T, Nielsen T, Jensen BA, Forslund K, Hildebrand F, Prifti E, Falony G, Le Chatelier E; MetaHIT Consortium , Kristiansen K, Brix S, Raes J, Wang J, Hansen T, Bork P, Brunak S, Oresic M, Ehrlich SD, Pedersen O. Human gut microbes impact host serum metabolome and insulin sensitivity. Nature. 2016;535:376–381. [DOI] [PubMed] [Google Scholar]
25. Yip AM, Horvath S. Gene network interconnectedness and the generalized topological overlap measure. BMC Bioinformatics. 2007;8:22. [DOI] [PMC free article] [PubMed] [Google Scholar]
26. Zhang B, Horvath S. A general framework for weighted gene co‐expression network analysis. Stat Appl Genet Mol Biol. 2005;4:Article17. [DOI] [PubMed] [Google Scholar]
27. Grewal J, McKelvie RS, Persson H, Tait P, Carlsson J, Swedberg K, Ostergren J, Lonn E. Usefulness of N‐terminal pro–brain natriuretic peptide and brain natriuretic peptide to predict cardiovascular outcomes in patients with heart failure and preserved left ventricular ejection fraction. Am J Cardiol. 2008;102:733–737. [DOI] [PubMed] [Google Scholar]
28. Owan TE, Hodge DO, Herges RM, Jacobsen SJ, Roger VL, Redfield MM. Trends in prevalence and outcome of heart failure with preserved ejection fraction. N Engl J Med. 2006;355:251–259. [DOI] [PubMed] [Google Scholar]
29. Takase H, Dohi Y. Kidney function crucially affects B‐type natriuretic peptide (BNP), N‐terminal pro BNP and their relationship. Eur J Clin Invest. 2014;44:303–308. [DOI] [PubMed] [Google Scholar]
30. Park JH, Marwick TH. Use and limitations of E/e’ to assess left ventricular filling pressure by echocardiography. J Cardiovasc Ultrasound. 2011;19:169–173. [DOI] [PMC free article] [PubMed] [Google Scholar]
31. Paulus WJ, Tschöpe C, Sanderson JE, Rusconi C, Flachskampf FA, Rademakers FE, Marino P, Smiseth OA, De Keulenaer G, Leite‐Moreira AF, Borbély A. How to diagnose diastolic heart failure: a consensus statement on the diagnosis of heart failure with normal left ventricular ejection fraction by the Heart Failure and Echocardiography Associations of the European Society of Cardiology. Eur Heart J. 2007;28:2539–2550. [DOI] [PubMed] [Google Scholar]
32. Mohammed SF, Borlaug BA, Roger VL, Mirzoyev SA, Rodeheffer RJ, Chirinos JA, Redfield MM. Comorbidity and ventricular and vascular structure and function in heart failure with preserved ejection fraction: a community‐based study. Circ Heart Fail. 2012;5:710–719. [DOI] [PMC free article] [PubMed] [Google Scholar]
33. Biering‐Sørensen T, Mogelvang R, Schnohr P, Jensen JS. Cardiac time intervals measured by tissue Doppler imaging M‐mode: association with hypertension, left ventricular geometry, and future ischemic cardiovascular diseases. J Am Heart Assoc. 2016;5:e002687 DOI: 10.1161/JAHA.115.002687. [DOI] [PMC free article] [PubMed] [Google Scholar]
34. Sharp AS, Tapp RJ, Thom SA, Francis DP, Hughes AD, Stanton AV, Zambanini A, O'Brien E, Chaturvedi N, Lyons S, Byrd S. Tissue Doppler E/E′ ratio is a powerful predictor of primary cardiac events in a hypertensive population: an ASCOT substudy. European Heart Journal. 2009;31:747–752. [DOI] [PubMed] [Google Scholar]
35. Schooneman MG, Vaz FM, Houten SM, Soeters MR. Acylcarnitines: reflecting or inflicting insulin resistance? Diabetes. 2013;62:1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
36. Pollard TD, Earnshaw WC, Lippincott‐Schwartz J, Johnson G. Cell Biology E‐Book. Philadelphia, PA: Elsevier Health Sciences; 2016. [Google Scholar]
37. De Jong KA, Lopaschuk GD. Complex energy metabolic changes in heart failure with preserved ejection fraction and heart failure with reduced ejection fraction. Can J Cardiol. 2017;33:860–871. [DOI] [PubMed] [Google Scholar]
38. Ikegami R, Shimizu I, Yoshida Y, Minamino T. Metabolomic analysis in heart failure. Circ J. 2017;82:10–16. [DOI] [PubMed] [Google Scholar]
39. Doenst T, Nguyen TD, Abel ED. Cardiac metabolism in heart failure: implications beyond ATP production. Circ Res. 2013;113:709–724. [DOI] [PMC free article] [PubMed] [Google Scholar]
40. Du Z, Shen A, Huang Y, Su L, Lai W, Wang P, Xie Z, Xie Z, Zeng Q, Ren H, Xu D. 1H‐NMR‐based metabolic analysis of human serum reveals novel markers of myocardial energy expenditure in heart failure patients. PLoS One. 2014;9:e88102. [DOI] [PMC free article] [PubMed] [Google Scholar]
41. Zordoky BN, Sung MM, Ezekowitz J, Mandal R, Han B, Bjorndahl TC, Bouatra S, Anderson T, Oudit GY, Wishart DS, Dyck JR. Metabolomic fingerprint of heart failure with preserved ejection fraction. PLoS One. 2015;10:e0124844. [DOI] [PMC free article] [PubMed] [Google Scholar]
42. Hunter WG, Kelly JP, McGarrah RW III, Khouri MG, Craig D, Haynes C, Ilkayeva O, Stevens RD, Bain JR, Muehlbauer MJ, Newgard CB. Metabolomic profiling identifies novel circulating biomarkers of mitochondrial dysfunction differentially elevated in heart failure with preserved versus reduced ejection fraction: evidence for shared metabolic impairments in clinical heart failure. J Am Heart Assoc. 2016;5:e003190 DOI: 10.1161/JAHA.115.003190. [DOI] [PMC free article] [PubMed] [Google Scholar]
43. Cheng ML, Wang CH, Shiao MS, Liu MH, Huang YY, Huang CY, Mao CT, Lin JF, Ho HY, Yang NI. Metabolic disturbances identified in plasma are associated with outcomes in patients with heart failure: diagnostic and prognostic value of metabolomics. J Am Coll Cardiol. 2015;65:1509–1520. [DOI] [PubMed] [Google Scholar]
44. Sansbury BE, DeMartino AM, Xie Z, Brooks AC, Brainard RE, Watson LJ, DeFilippis AP, Cummins TD, Harbeson MA, Brittian KR, Prabhu SD. Metabolomic analysis of pressure‐overloaded and infarcted mouse hearts. Circ Heart Fail. 2014;7:634–642. [DOI] [PMC free article] [PubMed] [Google Scholar]
45. Esko T, Hirschhorn JN, Feldman HA, Hsu YH, Deik AA, Clish CB, Ebbeling CB, Ludwig DS. Metabolomic profiles as reliable biomarkers of dietary composition. Am J Clin Nutr. 2017;105:547–554. [DOI] [PMC free article] [PubMed] [Google Scholar]
46. Sjölin J, Hjort G, Friman G, Hambraeus L. Urinary excretion of 1‐methylhistidine: a qualitative indicator of exogenous 3‐methylhistidine and intake of meats from various sources. Metabolism. 1987;36:1175–1184. [DOI] [PubMed] [Google Scholar]
47. Khodorova NV, Rutledge DN, Oberli M, Mathiron D, Marcelo P, Benamouzig R, Tomé D, Gaudichon C, Pilard S. Urinary metabolomics profiles associated to bovine meat ingestion in humans. Mol Nutr Food Res. 2019;63:1700834. [DOI] [PubMed] [Google Scholar]
48. Pallister T, Haller T, Thorand B, Altmaier E, Cassidy A, Martin T, Jennings A, Mohney RP, Gieger C, MacGregor A, Kastenmüller G. Metabolites of milk intake: a metabolomic approach in UK twins with findings replicated in two European cohorts. Eur J Nutr. 2017;56:2379–2391. [DOI] [PMC free article] [PubMed] [Google Scholar]
49. Pallister T, Jennings A, Mohney RP, Yarand D, Mangino M, Cassidy A, MacGregor A, Spector TD, Menni C. Characterizing blood metabolomics profiles associated with self‐reported food intakes in female twins. PLoS One. 2016;11:e0158568. [DOI] [PMC free article] [PubMed] [Google Scholar]
50. Levitan EB, Wolk A, Mittleman MA. Consistency with the DASH diet and incidence of heart failure. Arch Intern Med. 2009;169:851–857. [DOI] [PMC free article] [PubMed] [Google Scholar]
51. Fitó M, Estruch R, Salas‐Salvadó J, Martínez‐Gonzalez MA, Arós F, Vila J, Corella D, Díaz O, Sáez G, De La Torre R, Mitjavila MT. Effect of the Mediterranean diet on heart failure biomarkers: a randomized sample from the PREDIMED trial. Eur J Heart Fail. 2014;16:543–550. [DOI] [PubMed] [Google Scholar]
52. Hummel SL, Seymour EM, Brook RD, Sheth SS, Ghosh E, Zhu S, Weder AB, Kovács SJ, Kolias TJ. Low‐sodium DASH diet improves diastolic function and ventricular–arterial coupling in hypertensive heart failure with preserved ejection fraction. Circ Heart Fail. 2013;6:1165–1171. [DOI] [PMC free article] [PubMed] [Google Scholar]
53. Lustgarten MS, Price LL, Phillips EM, Fielding RA. Serum glycine is associated with regional body fat and insulin resistance in functionally‐limited older adults. PLoS One. 2013;8:e84034. [DOI] [PMC free article] [PubMed] [Google Scholar]
54. Mentz RJ, Kelly JP, von Lueder TG, Voors AA, Lam CS, Cowie MR, Kjeldsen K, Jankowska EA, Atar D, Butler J, Fiuzat M. Noncardiac comorbidities in heart failure with reduced versus preserved ejection fraction. J Am Coll Cardiol. 2014;64:2281–2293. [DOI] [PMC free article] [PubMed] [Google Scholar]
55. Li PJ, Jin T, Luo DH, Shen T, Mai DM, Hu WH, Mo HY. Effect of prolonged radiotherapy treatment time on survival outcomes after intensity‐modulated radiation therapy in nasopharyngeal carcinoma. PLoS One. 2015;10:e0141332. [DOI] [PMC free article] [PubMed] [Google Scholar]
56. Malyszko J. Mechanism of endothelial dysfunction in chronic kidney disease. Clin Chim Acta. 2010;411:1412–1420. [DOI] [PubMed] [Google Scholar]
57. Avilaa MA, Garcıa‐Trevijanoa ER, Lub SC, Corralesa FJ, Matoc JM. Molecules in focus. Int J Biochem Cell Biol. 2004;36:2125–2130. [DOI] [PubMed] [Google Scholar]
58. Pegg AE, Hibasami HI. Polyamine metabolism during cardiac hypertrophy. Am J Physiol Endocrinol Metab. 1980;239:E372–E378. [DOI] [PubMed] [Google Scholar]
59. Bartolome J, Huguenard J, Slotkin TA. Role of ornithine decarboxylase in cardiac growth and hypertrophy. Science. 1980;210:793–794. [DOI] [PubMed] [Google Scholar]
60. Durante W, Johnson FK, Johnson RA. Arginase: a critical regulator of nitric oxide synthesis and vascular function. Clin Exp Pharmacol Physiol. 2007;34:906–911. [DOI] [PMC free article] [PubMed] [Google Scholar]
61. Klutho PJ, Pennington SM, Scott JA, Wilson KM, Gu SX, Doddapattar P, Xie L, Venema AN, Zhu LJ, Chauhan AK, Lentz SR. Deletion of methionine sulfoxide reductase A does not affect atherothrombosis but promotes neointimal hyperplasia and extracellular signal‐regulated kinase 1/2 signaling. Arterioscler Thromb Vasc Biol. 2015;35:2594–2604. [DOI] [PMC free article] [PubMed] [Google Scholar]
62. Picot CR, Perichon M, Lundberg KC, Friguet B, Szweda LI, Petropoulos I. Alterations in mitochondrial and cytosolic methionine sulfoxide reductase activity during cardiac ischemia and reperfusion. Exp Gerontol. 2006;41:663–667. [DOI] [PubMed] [Google Scholar]
63. Lascelles PT, Donaldson D. Formiminoglutamic acid (FIGLU) test In: Springer ed. Diagnostic Function Tests in Chemical Pathology. Dordrecht: Springer; 1989:59–60. [Google Scholar]
64. Carter FC, Heller P, Schaffner G, Korn RJ. Formiminoglutamic acid (FIGLU) excretion in hepatic cirrhosis. Arch Intern Med. 1961;108:41–46. [DOI] [PubMed] [Google Scholar]
65. Witte KK, Desilva R, Chattopadhyay S, Ghosh J, Cleland JG, Clark AL. Are hematinic deficiencies the cause of anemia in chronic heart failure? Am Heart J. 2004;147:924–930. [DOI] [PubMed] [Google Scholar]
66. van der Wal HH, Comin‐Colet J, Klip IT, Enjuanes C, Beverborg NG, Voors AA, Banasiak W, van Veldhuisen DJ, Bruguera J, Ponikowski P, Jankowska EA. Vitamin B12 and folate deficiency in chronic heart failure. Heart. 2015;101:302–310. [DOI] [PubMed] [Google Scholar]
67. Catena C, Colussi G, Verheyen ND, Novello M, Fagotto V, Soardo G, Sechi LA. Moderate alcohol consumption is associated with left ventricular diastolic dysfunction in nonalcoholic hypertensive patients. Hypertension. 2016;68:1208–1216. [DOI] [PubMed] [Google Scholar]
68. Mayerhofer CC, Ueland T, Broch K, Vincent RP, Cross GF, Dahl CP, Aukrust P, Gullestad L, Hov JR, Trøseid M. Increased secondary/primary bile acid ratio in chronic heart failure. J Card Fail. 2017;23:666–671. [DOI] [PubMed] [Google Scholar]
69. Møller S, Wiese S, Halgreen H, Hove JD. Diastolic dysfunction in cirrhosis. Heart Fail Rev. 2016;21:599–610. [DOI] [PubMed] [Google Scholar]
70. Hunter WG, Kelly JP, McGarrah RW, Kraus WE, Shah SH. Metabolic dysfunction in heart failure: diagnostic, prognostic, and pathophysiologic insights from metabolomic profiling. Curr Heart Fail Rep. 2016;13:119–131. [DOI] [PMC free article] [PubMed] [Google Scholar]
71. Bayeva M, Gheorghiade M, Ardehali H. Mitochondria as a therapeutic target in heart failure. J Am Coll Cardiol. 2013;61:599–610. [DOI] [PMC free article] [PubMed] [Google Scholar]
72. Tanada Y, Shioi T, Kato T, Kawamoto A, Okuda J, Kimura T. Branched‐chain amino acids ameliorate heart failure with cardiac cachexia in rats. Life Sci. 2015;137:20–27. [DOI] [PubMed] [Google Scholar]
73. D'Antona G, Ragni M, Cardile A, Tedesco L, Dossena M, Bruttini F, Caliaro F, Corsetti G, Bottinelli R, Carruba MO, Valerio A. Branched‐chain amino acid supplementation promotes survival and supports cardiac and skeletal muscle mitochondrial biogenesis in middle‐aged mice. Cell Metab. 2010;12:362–372. [DOI] [PubMed] [Google Scholar]
74. Sun H, Olson KC, Gao C, Prosdocimo DA, Zhou M, Wang Z, Jeyaraj D, Youn JY, Ren S, Liu Y, Rau CD. Catabolic defect of branched‐chain amino acids promotes heart failure. Circulation. 2016;133:2038–2049. [DOI] [PMC free article] [PubMed] [Google Scholar]
75. Lu G, Sun H, She P, Youn JY, Warburton S, Ping P, Vondriska TM, Cai H, Lynch CJ, Wang Y. Protein phosphatase 2Cm is a critical regulator of branched‐chain amino acid catabolism in mice and cultured cells. J Clin Invest. 2009;119:1678–1687. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials