Global Plasma Metabolomics to Identify Potential Biomarkers of Blood Pressure Progression.

Arterioscler Thromb Vasc Biol is available at www.ahajournals.org/journal/atvb

Correspondence to: Johan Sundström, MD, PhD, Department of Medical Sciences, Uppsala University, Akademiska sjukhuset, Ing 40, 5 tr SE-751 85 Uppsala, Sweden. Email johan.sundstrom@medsci.uu.se

The Data Supplement is available with this article at https://www.ahajournals.org/doi/suppl/10.1161/ATVBAHA.120.314356. For Disclosures, see page e235.

© 2020 American Heart Association, Inc.

CLINICAL AND POPULATION STUDIES

Global Plasma Metabolomics to Identify Potential

Biomarkers of Blood Pressure Progression

Yi-Ting Lin, Samira Salihovic, Tove Fall, Ulf Hammar, Erik Ingelsson, Johan Ärnlöv, Lars Lind, Johan Sundström

OBJECTIVE:

The pathophysiology of hypertension remains incompletely understood. We investigated associations of circulating

metabolites with longitudinal blood pressure (BP) changes in the Prospective Investigation of the Vasculature in Uppsala

Seniors cohort and validated the findings in the Uppsala Longitudinal Study of Adult Men cohort.

APPROACH AND RESULTS:

Circulating metabolite levels were assessed with liquid- and gas-chromatography coupled to mass

spectrometry among persons without BP-lowering medication at baseline. We studied associations of baseline levels of metabolites

with changes in BP levels and the clinical BP stage between baseline and a follow-up examination 5 years later. In the discovery

cohort, we investigated 504 individuals that contributed with 757 observations of paired BP measurements. The mean baseline

systolic and diastolic BPs were 144 (19.7)/76 (9.7) mm Hg, and change in systolic and diastolic BPs were 3.7 (15.8)/−0.5 (8.6)

mm Hg over 5 years. The metabolites associated with diastolic BP change were ceramide, triacylglycerol, total glycerolipids, oleic

acid, and cholesterylester. No associations with longitudinal changes in systolic BP or BP stage were observed. Metabolites with

similar structures to the 5 top findings in the discovery cohort were investigated in the validation cohort. Diacylglycerol (36:2) and

monoacylglycerol (18:0), 2 glycerolipids, were associated with diastolic BP change in the validation cohort.

CONCLUSIONS:

Circulating baseline levels of ceramide, triacylglycerol, total glycerolipids, and oleic acid were positively

associated with longitudinal diastolic BP change, whereas cholesterylester levels were inversely associated with longitudinal

diastolic BP change. Two glycerolipids were validated in an independent cohort. These metabolites may point towards

pathophysiological pathways of hypertension.

GRAPHIC ABSTRACT:

A

graphic abstract

is available for this article.

Key Words:

blood pressure

◼

hypertension

◼

metabolomics

H

ypertension is the leading risk factor for premature

deaths worldwide,

_{and specific causes for the high}

blood pressure are found in <1 out of 10.

_Hence,

clarifying the pathophysiology in the hope of effective

prevention and treatment of hypertension is of utmost

importance. Circulating metabolites reflect both

endog-enous and exogendog-enous metabolic pathway activities and

can, therefore, provide a dynamic, coordinated

under-standing of disease pathophysiology. Dysregulation of

metabolites have been associated with hypertension in

several animal models, such as nephrectomized or

spon-taneously hypertensive rats.

_{However, differences in}

metabolomics studies between humans and rodents have

been reported.

_{Differences in diet, essential nutrients, and}

metabolic pathways can cause significant metabolomic

dif-ferences between species.

_{Previous human studies using}

metabolomics approaches to explore the pathophysiology

of hypertension have suggested several potential

path-ways, including inflammatory processes, oxidative stress,

lipid profile, and gut microflora.

_{But previous studies}

in the field are scarce,

_{findings are inconsistent, and}

most studies are limited by small sample sizes and

cross-sectional study designs. Only one longitudinal study has

reported on associations of lipid metabolites with risk of

See accompanying editorial on page 1801

CLINICAL AND POPULATION

STUDIES - AL

incident hypertension.

_{We are unaware of any previous}

studies that have investigated associations of circulating

metabolites with longitudinal blood pressure (BP) changes

in normotensive persons. We hypothesize that circulating

metabolites associated with such longitudinal BP changes

could reflect pathways involved in the pathophysiology of

hypertension. We, therefore, explored associations of the

metabolome with subsequent 5-year BP change in the

PIVUS (Prospective Investigation of the Vasculature in

Uppsala Seniors Study) cohort and validated in the Uppsala

Longitudinal Study of Adult Men (ULSAM) cohort.

MATERIALS AND METHODS

The data that support the findings of this study are available

from the corresponding author on reasonable request.

Study Sample

Discovery Cohort

The PIVUS cohort was used as the discovery cohort. All

70-year-old men and women living in Uppsala, Sweden, between 2001

and 2004 were invited to the PIVUS study (PIVUS-70).

_In

total, 1016 people (507 women and 499 men) took part in the

investigation (participation rate=50.1%). A second examination

cycle was performed in 2006 to 2009 when the participants

were 75 years old (PIVUS-75), and a third cycle in 2011 to

2014, when the participants were 80 years old (PIVUS-80).

Metabolomics analyses of plasma samples from the first 2

cycles, and BP measurements from all 3 cycles, were used in

this study to maximize statistical power.

Participants with heart failure, myocardial infarction, or

stroke history at baseline or follow-up were excluded from this

study. We also excluded those with any missing in BP

mea-surements at baseline or follow-up. Because antihypertensive

drugs have effects on plasma metabolomic profiles,

partici-pants with antihypertensive treatment at the baseline of each

observation period were also excluded from this study.

In the observation periods between 70 and

PIVUS-75, there were 498 subjects with complete BP measurements

at both examinations, and 476 of them had metabolomics data

at baseline (PIVUS-70); in the observation periods between

PIVUS-75 and PIVUS-80, 263 had BP measurements at both

examinations, of which 259 had metabolomics data at baseline

(PIVUS-75). Hence, 253 persons were observed twice (one

observation period between PIVUS-70 and PIVUS-75 and

another between PIVUS-75 and PIVUS-80) and 251 subjects

were observed once (either from PIVUS-70 to PIVUS-75 or

from PIVUS-75 to PIVUS-80). In sum, 504 individuals

contrib-uted with 757 observation periods in this study (Figure 1).

All participants provided written informed consent, Uppsala

University’s ethics committee approved the study, and the study

was conducted in accordance with the Helsinki Declaration.

Validation Cohort

The ULSAM study was initiated in 1970. All 50-year-old men

born between 1920 and 1924 and living in Uppsala, Sweden

were invited to a health survey, focusing on identifying

cardio-vascular risk factors (described in detail at www.pubcare.uu.se/

ULSAM).

_{The participants were, thereafter, invited to}

examina-tions at age 60, 70, 77, 82, and 88 years. Plasma samples were

available for 1138 participants at 70 years of age (1991–1994).

The present study used the fourth examination cycle as the

baseline, when the participants were about 77 years old (1998–

2001). We used the fifth examination cycle (2003–2005) as a

follow-up examination when participants were ≈82 years old.

Blood samples for metabolomics profiling were frozen

imme-diately after separation of plasma and stored at −80°C until

analysis. Participants without complete BP records or receiving

antihypertensive treatment at baseline were excluded from the

study, resulting in 222 individuals included in the current study.

Baseline and Follow-Up Examinations

All participants were examined in the early morning after an

overnight fast without any medication or smoking. A

question-naire surveying smoking habits, previous medical history, and

current regular medication was completed by each participant.

Height, weight, and body mass index (weight [kg]/height

[m]

_{), were measured under standardized conditions. Waist}

Nonstandard Abbreviations and Acronyms

BP

blood pressure

COX-1 cyclooxygenase-1

DBP

diastolic blood pressure

iPLA2

calcium-independent phospholipase A2

LDL

low-density lipoprotein

PIVUS

Prospective Investigation of the

Vascula-ture in Uppsala Seniors Study

SBP

systolic blood pressure

TXAS

thromboxane synthase

ULSAM

Uppsala Longitudinal Study of Adult Men

Highlights

• We investigated associations of a large number of

metabolites, assessed using gas

chromatography-mass spectrometry and liquid chromatography-tandem

mass spectrometry, with blood pressure progression

in a population-based cohort and validated in another

independent cohort. This is the first study to

evalu-ate association of metabolites with longitudinal blood

pressure progression using global metabolomics.

• Higher circulating baseline levels of ceramide,

triac-ylglycerol, total glycerolipids, and oleic acid, and lower

cholesterylester levels, were associated with

longitu-dinal diastolic blood pressure increase. Two

glycero-lipids were validated in an independent cohort.

• We found novel associations of metabolites with

longitudinal blood pressure increase using a

com-bined nontargeted and targeted mass

spectrom-etry metabolomics approach in a population-based

cohort. Unravelling the cause of hypertension is

important given the immense public health burden.

CLINICAL AND POPULATION

STUDIES - AL

circumference was measured at the umbilical level. Fasting

blood glucose and lipids were measured by standard

tech-niques. Serum cystatin C was measured by latex-enhanced

reagent (N Latex Cystatin C, Dade Behring) with a Behring BN

ProSpec analyzer (Dade Behring). Estimated glomerular

filtra-tion rate was calculated from serum cystatin C concentrafiltra-tions

(milligrams per liter) by the formula: y=77.24×cystatin C

_.

Diabetes mellitus was defined as P-glucose ≥7.0 mmol/L or

use of oral glucose-lowering agents or insulin.

In the PIVUS cohort, BP was measured to the nearest 1

mm Hg after at least 30 minutes of rest in a supine position,

and the average of 3 recordings was used. In the ULSAM

cohort, a nurse or physician measured BP twice in the right arm

to the nearest even number after a 10-minute rest in the supine

position, and the mean value was calculated. For patients

tak-ing any pharmaceutical BP-lowertak-ing treatment at follow-up,

BP values at follow-up were imputed by adding 10 mm Hg to

the actual systolic BP (SBP) and 5 mm Hg to the diastolic BP

(DBP) measurements at follow-up.

_{Sensitivity analyses of}

imputation by adding 15 mm Hg to SBP were also done. The

3 outcomes investigated were change in BP stage (defined

as JNC7 guideline BP categories, as in previous studies

₎

and change in continuous SBP and DBP between baseline and

follow-up.

Metabolomics Profiling

Metabolomics analyses in PIVUS were performed by

Metanomics GmbH (Berlin, Germany). Gas

chromatography-mass spectrometry (Agilent 6890 GC coupled to an Agilent

5973 MS-System, Agilent, Waldbronn, Germany) and liquid

chromatography-tandem mass spectrometry (Agilent 1100

high-performance liquid chromatography system [Agilent,

Waldbronn, Germany] coupled to an Applied Biosystems

API 4000 triple quadrupole mass spectrometer [Applied

Biosystems, Darmstadt, Germany]) were used for metabolic

profiling, as described in detail elsewhere.

_{Proteins were}

precipitated from plasma samples using 3 volumes of

ace-tonitrile, and polar and nonpolar fractions were separated by

adding water and a mixture of ethanol and dichloromethane

(2:1, v/v). For gas chromatography-mass spectrometry

anal-ysis, the nonpolar fraction was treated with methanol under

Figure 1.

Study design; 504 individuals contributed with 757 observations with paired blood pressure measurements in the

discovery cohort and 222 individuals in the validation cohort.

CLINICAL AND POPULATION

STUDIES - AL

acidic conditions to yield the fatty acid methyl esters derived

from both free fatty acids and hydrolyzed complex lipids. The

polar and nonpolar fractions were further derivatized with

O-methyl-hydroxylamine hydrochloride to convert oxo-groups

to O-methyl-oximes, and subsequently with a

N-methyl-N-(trimethylsilyl) trifluoroacetamide before analysis. For liquid

chromatography-tandem mass spectrometry analysis, both

fractions were reconstituted in appropriate solvent mixtures,

and high-performance liquid chromatography was performed

by gradient elution using methanol/water/formic acid on

reversed phase separation columns. Mass spectrometric

detection was performed with in targeted mode with

repeti-tive cycles of multiple reaction monitoring transitions for

pre-selected metabolites followed by nontargeted mode with a full

scan from a mass-to-charge ratio of 100 to 1000. The

instru-ment was operated in positive electrospray ionization mode

for metabolites in the nonpolar fraction (lipid fraction) and in

negative ionization mode for metabolites in the polar fraction.

Internal standards were added to increase precision, to quality

control the entire analytical process, and to monitor the

stabil-ity of the measurement. The internal standards used within this

study were either isotope-labeled compounds or chemicals

representing different chemical structures and polarities such

as amino acids, carbohydrates, cofactors, and lipids.

Metabolite identification was done by comparing mass to

charge ratio (M/z), retention time, and fragmentation to

authen-tic standards. Metabolite normalization and quantification were

calculated by determining metabolite levels in each study

sample relative to metabolite concentrations in reference pool

samples that were formed from aliquots of all study samples. To

allow an experiment-comprehensive alignment of data sets, the

semi-quantitative data were further normalized to the median

of MxPool samples representing a pool of commercial human

EDTA plasma containing >2000 different metabolites of known

concentrations. A one-point calibration was used to quantify

those metabolites that are present in the MxPool. Both types of

pooled reference samples were run in parallel through the entire

process. Metabolites that could not be quantified by MxPool

method were analyzed semi-quantitatively. Quality control of the

metabolomics dataset at the laboratory site comprised quality

checks on peak, analyte, and sample level. Furthermore, quality

assessment of plasma samples was performed using the MxP

Biofluids Quality Control assay (Metanomics Health), only those

metabolites that met specific quality criteria were included in

further statistical analyses (Table I in the

Data Supplement

) A

total of 563 named metabolites and 638 unknown analytes

were detected at baseline. Of these, 955 had low detectability

and were excluded from further analysis. Metabolite

identifica-tion was done by comparison to authentic standards first. Then

several extra experiments were done to further elucidate

struc-ture, such as high-resolution measurements using Fourier

trans-form ion cyclotron resonance MS (Bruker Solarix), the addition

of salts, peak purification via fractionation, and MS experiments.

These procedures provided more information on metabolite

structures, such as atmospheric pressure chemical ionization.

In total, 246 high-quality metabolites could be analyzed in the

present metabolomics data set (including quantitative results for

171 metabolites and semi-quantitative data for 75 metabolites).

Before statistical analysis, we excluded metabolites with >15%

samples below the limit of detection, and subjects with >5%

missing metabolites values were also excluded. We further

excluded a few metabolites that showed a strong bimodal

distri-bution at age 75, possibly related to batch effects undetected in

initial quality control. After quality control process, 220

metabo-lites remained and are listed in Table I in the

Data Supplement

.

To ensure the quality of the measurement, the comparison on

2 metabolites from the PIVUS cohort with those of established

clinical chemistry measurements was done and found a

moder-ate to high correlation (glucose, Spearman ρ=0.69, P<0.001;

creatinine, Spearman ρ=0.86, P<0.001).

Metabolomics analyses in ULSAM was performed using

ultra-performance liquid chromatography on a Waters Acquity

ultra-performance liquid chromatography system coupled to a

quadrupole time-of-flight mass spectrometer (Xevo G2 Q-TOF

MS; Waters Corporation, Milford, MA) platform at the Proteomics

and Metabolomics Facility of Colorado State University (Fort

Collins, CO), as previously described.

_{Nonconsecutive}

dupli-cate sample aliquots of 1 µL were injected onto an Acquity

ultra-performance liquid chromatography C8 column (1.8

µmol/L, 1.0×100 mm) analytical column held at 50°C using a

gradient from solvent A (95% water, 5% methanol, 0.1% formic

acid) to solvent B (95% methanol, 5% water, 0.1% formic acid).

Injections were made in 100% A, which was held for 0.1 minute,

ramped to 40% B in 0.9 minutes, to 70% B over 2 minutes,

and to 100% B over 8 minutes. The mobile phase was held at

100% B for 6 minutes, returned to starting conditions over 0.1

minutes, and allowed to re-equilibrate at for 5.9 minutes. The

flow rate was constant at 140 µL/min for the duration of the

run. The column was held at 50°C, while samples were held

at 10°C. Data acquisition in the positive electrospray ion mode

with a mass-to-charge ratio (m/z) range of 50 to 1200 at 5 Hz

was alternately performed in MS mode at a collision energy of

6 V and in indiscriminate MS/MS mode using higher collision

energy (15–30 V). Calibration was performed before sample

analysis via infusion of sodium formate solution, with mass

accuracy within 1 ppm. The capillary voltage was held at 2200

V, the source temp at 150°C, and the desolvation temperature

at 350°C at a nitrogen desolvation gas flow rate of 800 L/h.

The quadrupole was held at a collision energy of 6 volts. The

acquired raw data (chromatograms and mass spectra) were

taken further into a metabolomics data processing workflow by

XCMS package (Scripps Center for Metabolomics and Mass

Spectrometry, La Jolla, CA).

_{The XCMS data processing}

per-forms peak detection, alignment, grouping, and imputation of the

metabolic features characterized by a unique m/z and retention

time. In total, 10 162 features were detected and adjusted for

factors of external variability (plate/batch effect, analysis date,

retention, time drift, and sample collection) by ANOVA-type

standardization and log-transformation; by removal of spectra

with abnormal intensities and low inter-duplicate correlations

and retention times. For each feature, retention time, m/z, and

fragmentation patterns were compared with in-house and public

database reference libraries (>950 reference standards) and

matched according to Metabolomics Standard Initiative

guide-lines

_{to annotate spectra to metabolite names.}

Power Analysis

Because of the limited sample size in the present study, we

assessed the statistical power to detect associations using

1000 Monte Carlo simulations. We assumed that BP stage

progression would range between −3 and 3, with 1% in the

CLINICAL AND POPULATION

STUDIES - AL

outer categories, 4% in the categories −2 and 2 respectively,

20% in −1 and 1 and 50% in 0 category. The assumptions of

shifts between BP stages were based on the observed data.

We tested univariate associations of 220 metabolites, of which

1 was truly associated with the outcome by applying a

mixed-effects ordinal regression model with a Benjamini-Hochberg

correction for multiple testing, with a false discovery rate of

<5%.

_{All metabolites were assumed to be normalized to a}

standard normal distribution. For a sample size of 500

individu-als, with half of them having one and half of them 2 measures

of BP stage progression, we would have 80% power to detect

an odds ratio of 1.45 per SD. Using a standard threshold of

0.05, 4.9% of nonassociated metabolites were declared

signifi-cant. After false discovery rate-correction, there was on

aver-age 0.09 false positive findings per simulation (corresponding

to a false positive rate of 0.04% per metabolite).

Statistical Analysis

The study design is described in Figure 1.The mean (SD) and

count (percent) were presented in Table 1. Metabolite variables

were log

-transformed to obtain an approximately normal

distribution.

The PIVUS cohort was used in the discovery phase, and the

ULSAM cohort was used for replication. In the discovery phase,

the associations between the 220 metabolites and BP change

over 5 years were investigated. Because we had more than one

observation per individual, we analyzed associations of baseline

metabolites in SD units with longitudinal BP change (BP

dif-ference between baseline and follow-up 5 years later) using

mixed ordered logistic regression for BP stage progression and

mixed linear regression for continuous SBP and DBP change.

We assumed fixed effects for metabolites, age, sex and

base-line SBP and DBP, and a random intercept for subject ID. In

the mixed-effects regression, we assumed an independent

cor-relation structure which allows a distinct variance for each

ran-dom effect of subject and all covariances are 0. This handled

the fact that some individuals contributed with 2 observations.

Volcano plots for ORs or β coefficients and corresponding P

values from the corresponding ordered logistic or linear mixed

models were presented.

Adjustment for multiple comparisons was performed,

con-sidering an false discovery rate <5% as metabolome-wide

significant for each outcome (longitudinal change in SBP and

DBP, and BP stage progression). False discovery rate was

calculated according to the original version of Benjamini and

Hochberg from 1995.

_{The rationale for this significance}

threshold was that we wanted to find a reasonable balance

between false positive and false negative findings. We ranked

the metabolites by ascending P value and provided

bootstrap-obtained CIs around the ranks to quantify the uncertainty of the

importance of each metabolite.

To validate the findings from the discovery phase,

struc-turally similar metabolites as those found associated with BP

progression in PIVUS were investigated in the validation

sam-ple, using the same covariates with the exception of sex, as

ULSAM is all-male.

In a subsequent phase, we sought to investigate potential

causality of any findings using mixed models with multivariable

adjustment. The choice of variables for the adjusted models was

based on a causal diagram assisted by DAGitty,

_{version 2.2,}

software (www.dagitty.net; Figure I in the

Data Supplement

).

Causal directed acyclic graphs are useful to help identify

suf-ficient and minimum covariates to produce bias-minimized

models and to provide an overview of the causal assumptions

made. The model included fixed effects for covariates age, sex,

baseline BPs, body mass index, waist circumference, smoking,

diabetes mellitus, LDL (low-density lipoprotein), fasting

glu-cose, estimated glomerular filtration rate, and physical activity,

all assessed at baseline. Covariates that play important roles in

BP tracking were chosen based on previous knowledge.

The nature of the associations of the top metabolites with

DBP change were investigated using restricted cubic splines

with 3 knots. Sensitivity analyses of participants with and

with-out antihypertensive drug during follow-up were also done.

All the statistical methods were performed using Stata

(ver-sion 15, College Station, TX).

RESULTS

Baseline Characteristics

Baseline characteristics of the discovery and validation

cohorts are shown in Table 1. The mean SBP/DBP are

144 (20)/76 (10) at baseline and 148 (19)/75 (9) at

follow-up in the discovery cohort and 148 (19)/80 (9)

at baseline and 146 (16)/81 (9) at follow-up in the

vali-dation cohort. Distributions of the changes in SBP and

DBP are presented in Figure II in the

Data Supplement

.

Associations of Baseline Metabolites With

Longitudinal BP Change

In the discovery phase, estimates of the associations of

the 220 metabolites with change in BP stage and

con-tinuous BP, and corresponding P values, are displayed in

Figure 2 and Tables II, III, and IV in the

Data Supplement

.

Among the 220 analyzed metabolites, no metabolite

was associated with BP stage progression or

continu-ous SBP change (Figure 2A and 2B). Levels of ceramide

(d18:1,C24:0), triacylglycerol (C16:0,C16:1), total

glyc-erolipids, and oleic acid (C18:cis[9]1) were positively

associated with continuous DBP change, and

choles-terylester C16:0 was negatively associated with DBP

change (Figure 2C). Sensitivity analyses with imputation

of 15 mm Hg to SBP for treated produced similar results.

Ranking of the associations (with 95%

bootstrap-obtained CIs) of metabolites with BP stage progression

(Figure IV in the

Data Supplement

), continuous SBP

(Fig-ure V in the

Data Supplement

), and DBP change (Figure

VI in the

Data Supplement

) are graphically presented in

the

Data Supplement

. The ceramide (d18:1,C24:0),

cho-lesterylester C16:0, triacylglycerol (C16:0,C16:1), total

glycerolipids, and oleic acid (C18:cis[9]1) were within

the top 10 hits, but with wide confidence intervals of the

ranking, as expected.

In the subsequent phase using mixed models with

multivariate adjustment, ceramide (d18:1,C24:0; β, 1.06

CLINICAL AND POPULATION

STUDIES - AL

[95% CI, 0.50–1.61]), triacylglycerol (C16:0,C16:1; β,

1.00 [95% CI, 0.43–1.57]), total glycerolipids (β, 0.95

[95% CI, 0.40–1.50]), and oleic acid (C18:cis[9]1; β,

0.92 [95% CI, 0.37–1.47]) were independently

asso-ciated with continuous DBP change (adjusting for

baseline age, sex, diastolic BP, body mass index, waist

circumference, smoking, diabetes mellitus, LDL, fasting

glucose, estimated glomerular filtration rate, and

physi-cal activity; Table 2). Cholesterylester (C16:0; β, −1.06

[95% CI, −1.65 to −0.47]) was independently

nega-tively associated with continuous DBP change in these

models.

In the validation phase, 11 metabolites available in

the validation cohort that had similar structure to the 5

top findings in the discovery cohort were investigated.

Diacylglycerol (36:2; β, 1.52 [95% CI, 0.24–2.80]) and

monoacylglycerol (18:0; β, 1.74 [95% CI, 0.31–3.18), 2

glycerolipids, were significantly associated with diastolic

BP change (Table 2, Figures VII and VIII in the

Data

Supplement

).

In secondary analyses in the discovery cohort, cubic

spline models demonstrated chiefly linear association

of metabolites on diastolic BP among ceramide (d18:1,

C24:0), triacylglycerol (C16:0, C16:1), glycerolipids, and

cholesterylester (C16:0), except oleic acid (C18:cis[9]1;

Figure III in the

Data Supplement

). In sensitivity analyses,

the heterogeneity between participants with or without

antihypertensive treatment during follow-up was low (I

square=0, Figure IX in the

Data Supplement

).

DISCUSSION

Principal Observations

In this study, we investigated associations of a large

number of circulating metabolites with longitudinal BP

progression in a population-based cohort. Accounting for

multiple testing, and in mixed models with multivariable

adjustment, baseline levels of 4 metabolites (ceramide

[d18:1, C24:0], triacylglycerol [C16:0, C16:1], total

glyc-erolipids, and oleic acid [C18:cis[9]1]) were positively

associated with longitudinal change in DBP. One

metab-olite (cholesterylester C16:0) was negatively associated

with DBP change. In the validation study, diacylglycerol

(36:2) and monoacylglycerol (18:0), 2 glycerolipids, were

associated with diastolic BP change.

The Evidence for the Role of Lipidomics in

Hypertension

Observational studies have suggested that

dyslipid-emia, by way of endothelial dysfunction,

_may

con-tribute to the development of hypertension.

_{In a}

longitudinal analysis of the San Antonio Family Heart

Table 1.

Baseline Characteristics of 757 Observations of 504 Participants in the Discovery Cohort (Prospective Investigation of

the Vasculature in Uppsala Seniors) and Validation Cohort (Uppsala Longitudinal Study of Adult Men)

Characteristic

Discovery Cohort Validation Cohort Observation Periods From PIVUS-70 to PIVUS-75 (n=498) Observation Periods From PIVUS-75 to PIVUS-80 (n=259)

All Observation Periods (n=757)

The ULSAM Cohort (n=222)

Age, y 70.2 (0.2) 75.3 (0.2) 71.9 (2.4) 77.5 (0.8)

Men, n (%) 262 (52.6) 139 (53.7) 401 (53.0) 222 (100)

Smoker, n (%) 45 (9.0) 15 (5.8) 60 (7.9) 17 (7.7)

Systolic blood pressure at baseline, mm Hg 144 (20.7) 144.09 (17.7) 144 (19.7) 148 (19.0) Diastolic blood pressure at baseline, mm Hg 77 (9.9) 75 (9.1) 76 (9.7) 80 (9.2) Systolic blood pressure at follow-up, mm Hg 149 (19.6) 146 (18.9) 148 (19.4) 146 (16.1) Diastolic blood pressure at follow-up, mm Hg 76 (9.4) 74 (8.8) 75 (9.3) 81 (9.1) Body mass index, kg/m2 _{26.45 (3.85) 25.81 (3.71) 26.23 (3.82) 25.67 (2.96)} Waist circumference, cm 89.32 (11.3) 91.44 (10.3) 90 (11.0) 92.88 (9.42)

Diabetes mellitus, n (%) 28 (5.6) 12 (4.6) 40 (5.3) 16 (7.2)

Total cholesterol, mmol/L 5.53 (0.98) 5.73 (0.99) 5.60 (0.99) 5.53 (0.88) Low-density lipoprotein cholesterol, mmol/L 3.45 (0.84) 3.61 (0.86) 3.50 (0.85) 3.92 (0.86) High-density lipoprotein cholesterol, mmol/L 1.58 (0.45) 1.61 (0.46) 1.59 (0.45) 1.37 (0.35)

Triglycerides, mmol/L 1.20 (0.57) 1.21 (0.49) 1.20 (0.54) 1.38 (0.75)

Fasting glucose, mmol/L 5.72 (1.22) 5.42 (0.80) 5.62 (1.11) 5.75 (1.72) Estimated glomerular filtration rate, mL/min per 1.73 m2 _{70.12 (36.15) 67.98 (15.23) 69.36 (30.47) 78.44 (14.65)} Antihypertensive treatment at follow-up (%) 119 (23.9) 63 (24.3) 182 (24.0) 59 (30.9) Increased blood pressure stage at follow-up, n (%) 168 (33.7) 77 (29.7) 245 (32.4) 54 (24.3) Decreased blood pressure stage at follow-up, n (%) 74 (14.9) 61 (23.6) 135 (17.8) 65 (29.3)

PIVUS indicates Prospective Investigation of the Vasculature in Uppsala Seniors Study.

CLINICAL AND POPULATION

STUDIES - AL

Study, phosphatidylethanolamine 40:6, diacylglycerols

16:0/22:5 and 16:0/226 were associated with risk of

incident hypertension.

_{In cross-sectional analyses, a}

dicarboxylic long-chain fatty acid was associated with

BPs in the Twin UK study and replicated in the KORA

(Cooperative Health Research in the Region of

Augs-burg) and Hertfordshire studies,

_{and 14 lipid species}

were associated with BPs in the San Antonio Family

Heart Study.

_{No study has addressed associations of}

lipid metabolites with longitudinal change in

continu-ous BP change among untreated. In the present study,4

metabolites (ceramide, triacylglycerol, total glycerolipids,

and oleic acid) positively associated with longitudinal

change in DBP and confirmed the potential importance

of lipid metabolites in hypertension development.

Derangement of lipid metabolism is one of the key

characteristics of hypertension.

_{Lipids are typically}

subdivided into 8 categories: sphingolipids,

glycerolip-ids, fatty acyls, phospholipglycerolip-ids, sterol lipglycerolip-ids, prenol lipglycerolip-ids,

saccharolipids, and polyketides.

_{In the present study,}

representatives of 3 of these categories—sphingolipids

(ceramide), glycerolipids (triacylglycerol), and fatty acyls

(oleic acid, cholesterylester)—were associated with DBP

change. Associations of metabolites with DBP change

were found, but not with SBP change, although

metabo-lites associated with DBP change were also among the

highest ranked for associations with SBP change. Power

deficit may be one explanation for the insignificant SBP

results; and regression to the mean in repeated BP

mea-surements may affect SBP more than DBP due to higher

variability.

_{Potential mechanistic underpinnings}

include that DBP may be more affected by hormonal

pathways mediated by metabolites; and that

atheroscle-rosis and stiff arteries and arterioles explain increasingly

more of SBP with increasing age, drowning out other

pathways. The specific potential roles of 3 lipid

catego-ries are described below.

The Role of Sphingolipids (ie, Ceramide) in

Hypertension

Previous observational studies mostly correspond to our

finding of a positive association of ceramide with

longi-tudinal DBP change. Plasma ceramides C16:0, C22:0,

C24:0, and C24:1 were elevated in both

spontane-ously hypertensive rats and treatment-naïve patients

with stage 1 to 3 hypertension in one study.

_{In the}

PREDIMED trial (Prevención con Dieta Mediterránea),

a higher ceramide concentration was associated with

higher DBP in a cross-sectional setting.

The sphingolipid system was suggested to be involved

in BP regulation in 2 experimental studies.

_{The central}

intermediate of the sphingolipid biosynthetic pathway is

ceramide. Ceramides can mediate vascular dysfunction

by inhibiting the endothelial nitric oxide synthase-serine/

threonine protein kinases-heat shock protein 90

signal-ing complex.

_{Besides, ceramides lead to}

endothelium-dependent arterial contraction by inducing the release of

thromboxane A2 via a iPLA2 (calcium-independent

phos-pholipase A2), COX-1 (cyclooxygenase-1), and TXAS

(thromboxane synthase) dependent pathway.

_{In addition,}

vascular ceramide levels are sensitive to antihypertensive

Figure 2.

The relationship between baseline metabolites and

blood pressure outcomes.

Volcano plot of associations of baseline metabolites with (A) blood

pressure stage progression, (B) systolic blood pressure change, and

(C) diastolic blood pressure change over 5 years (red line set false

discovery rate 5% corrected P value), using mixed ordered logistic

(for blood pressure stage progression) or mixed linear (for change in

continuous blood pressures) models, with fixed effects for age, sex,

and baseline blood pressures.

CLINICAL AND POPULATION

STUDIES - AL

therapy. Losartan lowers vascular ceramide levels and

improves endothelial function via inhibition of

ceramide-mediated endothelium-dependent vasoconstriction.

Thus, sphingolipids contribute to endothelial dysfunction

by inhibiting endothelium-derived relaxing factors and

producing endothelium-derived contracting factors.

The Role of Glycerolipids (ie, Triacylglycerol) in

Hypertension

Insulin resistance may play an important role in the

asso-ciation of glycerolipids with hypertension. Lower

lysophos-phatidylcholines and higher triacylglycerols have been

suggested as a lipidomic profile of insulin resistance.

Hence, the association of triacylglycerols with longitudinal

DBP change in our study may involve insulin resistance.

Insulin resistance has been proposed as a cause of

hyper-tension,

_{and this relationship was independent of body}

mass index.

_{Studies suggested insulin resistance}

was correlated to hypertension because of abnormalities

in vasodilatation, blood flow, the

renin-angiotensin-aldo-sterone system, and over-activity of the sympathetic

ner-vous system.

_{Thus, insulin resistance may be considered}

as a molecular marker of multiple metabolic abnormalities

frequently associated with hypertension.

The Role of Fatty Acyls (ie, Oleic Acid,

Cholesterylester) in Hypertension

Hypertensive obesity patients have elevated plasma

nonesterified fatty acids, including oleic acid.

Circulat-ing oleic acid level has been observed to be higher in

patients with hypertension,

_{positively associated with}

DBP,

_{and associated with circadian BP disturbances.}

The possible underlying pathological process suggested

by previous studies could be that oleic acid increases

the production of mitochondrial reactive oxygen species,

decreases the activity of endothelial nitric oxide synthesis

activity,

_{and regulates the α and β adrenergic receptors}

involved in controlling the central and peripheral BP.

Studies investigating associations of cholesteryl esters

with hypertension are scarce. Cholesterol esters are

cho-lesterol molecules with long-chain fatty acids linked to

the hydroxyl group. Cholesterol esters accumulate in the

fatty lesions of atherosclerotic plaques and are major

constituents of the lipoprotein particles carried in blood

(HDL, LDL, VLDL).

_{Cholesteryl esters are involved in}

atherosclerosis,

_{which is linked to arterial stiffness,}

known to increase systolic pressure (because of reduced

capacitance) and lower diastolic pressure (because of

less elastic recoil) in older age.

_{These phenomena}

cor-respond with our finding of inverse associations of

cir-culating cholesteryl esters with DBP change. In addition,

one small cohort study reported lower cholesteryl ester

levels in patients with hypertension,

_{in support of our}

findings.

Strengths and Limitations

Strengths of this study include the application of

metab-olomics to a well-defined population-based

longitudi-nal study with minimal loss to follow-up, correction for

multiple testing, and adjustment for potential

confound-ers. The investigated metabolites have been previously

validated, and we applied a predefined quality control

scheme for all metabolites. Moreover, the validation study

partially confirmed the results from the discovery cohort.

We also estimated the uncertainty of the ranking of all

metabolite associations.

Table 2.

Associations in the Discovery Cohort of Baseline Levels of 5 Metabolites With Change in Diastolic Blood Pressure

Between Baseline and Follow-Up 5 Years, and Associations of Structurally Similar Metabolites in the Validation Cohort

Discovery Phase (the PIVUS Cohort) Validation Phase (the ULSAM Cohort)

Metabolites β Coefficient 95% CI Metabolites β Coefficient 95% CI

Glycerolipids 0.95 0.40 to 1.50 Diacylglycerol (36:2) 1.52 0.24 to 2.80 Diacylglycerol (34:1) 1.11 −0.19 to 2.42 Monoacylglycerol (14:0) −0.08 −1.47 to 1.30 Monoacylglycerol (16:0) 0.22 −1.13 to 1.58 Monoacylglycerol (16:1) 0.44 −1.07 to 1.96 Monoacylglycerol (18:0) 1.74 0.31 to 3.18 Monoacylglycerol (18:1) 0.47 −0.88 to 1.81 Monoacylglycerol (18:2) 0.26 −1.13 to 1.66 Ceramide (d18:1,C24:0) 1.06 0.50 to 1.61 Ceramide phosphoethanolamine (36:2) −0.49 −1.88 to 0.90 Triacylglycerol (C16:0,C16:1) 1.00 0.43 to 1.57 Triacylglycerol (total) 1.44 −0.88 to 3.76 Oleic acid (C18:cis[9]1) 0.92 0.37 to 1.47 Oleic acid −0.70 −2.05 to 0.66

Cholesterylester (C16:0) −1.06 −1.65 to −0.47 … … …

β coefficients express the associations of baseline metabolites (per SD) with diastolic blood pressure change, using mixed linear regression including fixed effects for the metabolite and covariates age, sex, diastolic blood pressure, body mass index, waist circumference, smoking, diabetes mellitus, low-density lipoprotein, fasting glucose, estimated glomerular filtration rate, and physical activity, all assessed at baseline, and random intercept for subject ID, in the discovery phase. Linear regression controlling for the same confounders were used in the validation phase. PIVUS indicates Prospective Investigation of the Vasculature in Uppsala Seniors Study; and ULSAM, Uppsala Longitudinal Study of Adult Men.

CLINICAL AND POPULATION

STUDIES - AL

Several limitations need to be acknowledged. First, we

only included elderly people living in Sweden, so

general-izability to other populations and age groups is unknown.

We excluded persons with comorbidities and use of

antihypertensive drugs, which limits the generalizability

somewhat. Besides, participants whose data were used

in the analyses might be different from those whose data

were not included or who dropped out. We do not know

whether this missingness is informative or not. Second,

a healthy cohort effect may be in play because only

par-ticipants surviving without myocardial infarction or stroke

during follow-up were included, and their BP tracking

may differ from those who died. Third, some potential

confounders, such as baseline BPs, may also partly be

on the causal pathway to BP progression (Figure V in

the

Data Supplement

). Fourth, glycerolipids were the

only metabolites that could be validated in the replication

cohort. The replication efforts were limited by low

statis-tical power due to a small sample, and the fact that the

validation sample was all-male. Finally, no proper

analy-sis of causality using Mendelian randomization could be

done because of lack of appropriate instruments for the

investigated metabolites and lack of large studies of BP

progression. Although previous studies provide some

support for the biological plausibility of our findings, more

proof of causality are necessary.

Conclusions

A comprehensive metabolomic approach can help to

elu-cidate the molecular mechanisms underpinning

hyper-tension. We discovered 4 metabolites (ceramide [d18:1,

C24:0], triacylglycerol, glycerolipids, and oleic acid

[C18:cis[9]1]) that were directly associated with

subse-quent longitudinal DBP change in a population-based

cohort, and one (cholesterylester C16:0) inversely

associ-ated with DBP change. Two glycerolipid associations were

validated in an independent cohort. These metabolites

could point towards pathophysiological pathways of

hyper-tension; elucidating those pathways may lead to

under-standing of potentially treatable causes of hypertension.

ARTICLE INFORMATION

Received January 1, 2020; accepted April 29, 2020.

Affiliations

From the Department of Medical Sciences, Uppsala University, Sweden (Y.-T.L., S.S., T.F., U.H., E.I., L.L., J.S.); Department of Family Medicine, Kaohsiung Medical University Hospital (Y.-T.L.) and Faculty of Medicine, College of Medicine (Y.-T.L.), Kaohsiung Medical University, Taiwan; School of Medical Sciences (S.S.) and School of Science and Technology (S.S.), Örebro University, Sweden; Division of Cardiovascular Medicine, Department of Medicine (E.I.), Stanford Cardiovascular Institute (E.I.), and Stanford Diabetes Research Center (E.I.), Stanford University School of Medicine, CA; Division of Family Medicine and Primary Care, Depart-ment of Neurobiology, Care Science and Society, Karolinska Institutet, Huddinge, Sweden (J.Ä.); School of Health and Social Studies, Dalarna University, Falun, Sweden (J.Ä.); and The George Institute for Global Health, University of New South Wales, Sydney, NSW, Australia (J.S.).

Acknowledgments

J. Sundström and Y.-T. Lin are the guarantors of this work, had full access to all the data, and take full responsibility for the integrity of the data and the accuracy of data analysis.

Disclosures

The company Metanomics Health GmbH had no influence over design, analysis, or interpretation of data in the present study and did not provide any funding for the study. J. Sundström is on an advisory board for Itrim. The other authors report no conflicts.

REFERENCE

1. Forouzanfar MH, Liu P, Roth GA, Ng M, Biryukov S, Marczak L, Alexander L, Estep K, Hassen Abate K, Akinyemiju TF, et al. Global Burden of Hyperten-sion and Systolic Blood Pressure of at Least 110 to 115 mm Hg, 1990-2015. JAMA. 2017;317:165–182. doi: 10.1001/jama.2016.19043 2. Chen S. Essential hypertension: perspectives and future directions. J

Hyper-tens. 2012;30:42–45. doi: 10.1097/HJH.0b013e32834ee23c

3. Hanifa MA, Skott M, Maltesen RG, Rasmussen BS, Nielsen S, Frøkiær J, Ring T, Wimmer R. Tissue, urine and blood metabolite signatures of chronic kidney disease in the 5/6 nephrectomy rat model. Metabolomics. 2019;15:112. doi: 10.1007/s11306-019-1569-3

4. Feng YL, Cao G, Chen DQ, Vaziri ND, Chen L, Zhang J, Wang M, Guo Y, Zhao YY. Microbiome-metabolomics reveals gut microbiota asso-ciated with glycine-conjugated metabolites and polyamine metabolism in chronic kidney disease. Cell Mol Life Sci. 2019;76:4961–4978. doi: 10.1007/s00018-019-03155-9

5. Zhang ZH, Chen H, Vaziri ND, Mao JR, Zhang L, Bai X, Zhao YY. Metab-olomic Signatures of Chronic Kidney Disease of Diverse Etiologies in the Rats and Humans. J Proteome Res. 2016;15:3802–3812. doi: 10.1021/acs.jproteome.6b00583

6. Liu X, Wang Y, Gao R, Xing Y, Li X, Wang Z. Serum metabolomic response to exercise training in spontaneously hypertensive rats. J Am Soc Hypertens. 2017;11:428–436. doi: 10.1016/j.jash.2017.05.003

7. James EL, Parkinson EK. Serum metabolomics in animal models and human disease. Curr Opin Clin Nutr Metab Care. 2015;18:478–483. doi: 10.1097/MCO.0000000000000200

8. Jiye A, Huang Q, Wang G, Zha W, Yan B, Ren H, Gu S, Zhang Y, Zhang Q, Shao F, et al. Global analysis of metabolites in rat and human urine based on gas chromatography/time-of-flight mass spectrometry. Anal Biochem. 2008;379:20–26. doi: 10.1016/j.ab.2008.04.025

9. Tzoulaki I, Iliou A, Mikros E, Elliott P. An Overview of Metabolic Phenotyp-ing in Blood Pressure Research. Curr Hypertens Rep. 2018;20:78. doi: 10.1007/s11906-018-0877-8

10. Nikolic SB, Sharman JE, Adams MJ, Edwards LM. Metabolomics in hypertension. J Hypertens. 2014;32:1159–1169. doi: 10.1097/HJH. 0000000000000168

11. Kulkarni H, Meikle PJ, Mamtani M, Weir JM, Barlow CK, Jowett JB, Bellis C, Dyer TD, Johnson MP, Rainwater DL, et al. Plasma lipi-domic profile signature of hypertension in Mexican American families: specific role of diacylglycerols. Hypertension. 2013;62:621–626. doi: 10.1161/HYPERTENSIONAHA.113.01396

12. Zheng Y, Yu B, Alexander D, Mosley TH, Heiss G, Nettleton JA, Boerwinkle E. Metabolomics and incident hypertension among blacks: the atheroscle-rosis risk in communities study. Hypertension. 2013;62:398–403. doi: 10.1161/HYPERTENSIONAHA.113.01166

13. Menni C, Graham D, Kastenmüller G, Alharbi NH, Alsanosi SM, McBride M, Mangino M, Titcombe P, Shin SY, Psatha M, et al. Metabolo-mic identification of a novel pathway of blood pressure regulation involv-ing hexadecanedioate. Hypertension. 2015;66:422–429. doi: 10.1161/ HYPERTENSIONAHA.115.05544

14. Dietrich S, Floegel A, Weikert C, Prehn C, Adamski J, Pischon T, Boeing H, Drogan D. Identification of Serum Metabolites Associated With Incident Hypertension in the European Prospective Investigation into Can-cer and Nutrition-Potsdam Study. Hypertension. 2016;68:471–477. doi: 10.1161/HYPERTENSIONAHA.116.07292

15. Lind L, Fors N, Hall J, Marttala K, Stenborg A. A comparison of three dif-ferent methods to evaluate endothelium-dependent vasodilation in the elderly: the Prospective Investigation of the Vasculature in Uppsala Seniors (PIVUS) study. Arterioscler Thromb Vasc Biol. 2005;25:2368–2375. doi: 10.1161/01.ATV.0000184769.22061.da

CLINICAL AND POPULATION

STUDIES - AL

16. Hiltunen TP, Rimpelä JM, Mohney RP, Stirdivant SM, Kontula KK. Effects of four different antihypertensive drugs on plasma metabolomic profiles in patients with essential hypertension. PLoS One. 2017;12:e0187729. doi: 10.1371/journal.pone.0187729

17. Hedstrand H. A study of middle-aged men with particular reference to risk factors for cardiovascular disease. Ups J Med Sci Suppl. 1975;19:1–61. 18. Larsson A, Malm J, Grubb A, Hansson LO. Calculation of glomerular filtration

rate expressed in mL/min from plasma cystatin C values in mg/L. Scand J Clin Lab Invest. 2004;64:25–30. doi: 10.1080/00365510410003723 19. Tobin MD, Sheehan NA, Scurrah KJ, Burton PR. Adjusting for treatment

effects in studies of quantitative traits: antihypertensive therapy and systolic blood pressure. Stat Med. 2005;24:2911–2935. doi: 10.1002/sim.2165 20. Lytsy P, Lind L, Sundström J. Endothelial function and risk of

hyperten-sion and blood pressure progreshyperten-sion: the prospective investigation of the vasculature in Uppsala seniors. J Hypertens. 2013;31:936–939. doi: 10.1097/HJH.0b013e32835ed5a0

21. Lytsy P, Ingelsson E, Lind L, Arnlöv J, Sundström J. Interplay of over-weight and insulin resistance on hypertension development. J Hypertens. 2014;32:834–839. doi: 10.1097/HJH.0000000000000081

22. Nishimura RA, Otto CM, Bonow RO, Carabello BA, Erwin JP III, Fleisher LA, Jneid H, Mack MJ, McLeod CJ, O’Gara PT, et al. 2017 AHA/ACC Focused Update of the 2014 AHA/ACC Guideline for the Management of Patients With Valvular Heart Disease: A Report of the American College of Cardiol-ogy/American Heart Association Task Force on Clinical Practice Guidelines. J Am Coll Cardiol. 2017;70:252–289. doi: 10.1016/j.jacc.2017.03.011 23. Kamlage B, Maldonado SG, Bethan B, Peter E, Schmitz O, Liebenberg V,

Schatz P. Quality markers addressing preanalytical variations of blood and plasma processing identified by broad and targeted metabolite profiling. Clin Chem. 2014;60:399–412. doi: 10.1373/clinchem.2013.211979 24. Wagner-Golbs A, Neuber S, Kamlage B, Christiansen N, Bethan B,

Rennefahrt U, Schatz P, Lind L. Effects of long-term storage at -80 degrees c on the human plasma metabolome. Metabolites. 2019;9:99. doi: 10.3390/ metabo9050099

25. Kamlage B, Neuber S, Bethan B, Gonzalez Maldonado S, Wagner-Golbs A, Peter E, Schmitz O, Schatz P. Impact of prolonged blood incubation and extended serum storage at room temperature on the human serum metabo-lome. Metabolites. 2018;8:6

26. Fall T, Salihovic S, Brandmaier S, Nowak C, Ganna A, Gustafsson S, Broeckling CD, Prenni JE, Kastenmüller G, Peters A, et al. Non-targeted metabolomics combined with genetic analyses identifies bile acid synthesis and phospholipid metabolism as being associated with incident type 2 diabe-tes. Diabetologia. 2016;59:2114–2124. doi: 10.1007/s00125-016-4041-1 27. Nowak C, Salihovic S, Ganna A, Brandmaier S, Tukiainen T, Broeckling CD,

Magnusson PK, Prenni JE, Wang-Sattler R, Peters A, et al. Effect of Insulin Resistance on Monounsaturated Fatty Acid Levels: A Multi-cohort Non-targeted Metabolomics and Mendelian Randomization Study. PLoS Genet. 2016;12:e1006379. doi: 10.1371/journal.pgen.1006379

28. Smith CA, Want EJ, O’Maille G, Abagyan R, Siuzdak G. XCMS: process-ing mass spectrometry data for metabolite profilprocess-ing usprocess-ing nonlinear peak alignment, matching, and identification. Anal Chem. 2006;78:779–787. doi: 10.1021/ac051437y

29. Sumner LW, Amberg A, Barrett D, Beale MH, Beger R, Daykin CA, Fan TW, Fiehn O, Goodacre R, Griffin JL, et al. Proposed minimum reporting stan-dards for chemical analysis Chemical Analysis Working Group (CAWG) Metabolomics Standards Initiative (MSI). Metabolomics. 2007;3:211–221. doi: 10.1007/s11306-007-0082-2

30. Laird NM, Ware JH. Random-effects models for longitudinal data. Biomet-rics. 1982;38:963–974.

31. Bruyndonckx R, Hens N, Aerts M. Simulation-based evaluation of the linear-mixed model in the presence of an increasing proportion of singletons. Biom J. 2018;60:49–65. doi: 10.1002/bimj.201700025

32. Benjamini Yoav HY. Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society. Series B (Methodological). 1995;57(1):12

33. Harrell FE. Regression modeling strategies: With applications to linear models, logistic regression, and survival analysis. New York: Springer; 2001. 34. Greenland S, Pearl J, Robins JM. Causal diagrams for epidemiologic

research. Epidemiology. 1999;10:37–48.

35. Textor J, van der Zander B, Gilthorpe MS, Liskiewicz M, Ellison GT. Robust causal inference using directed acyclic graphs: the R package ‘dagitty’. Int J Epidemiol. 2016;45:1887–1894. doi: 10.1093/ije/dyw341

36. Sundström J, Sullivan L, D’Agostino RB, Levy D, Kannel WB, Vasan RS. Relations of serum uric acid to longitudinal blood pressure

tracking and hypertension incidence. Hypertension. 2005;45:28–33. doi: 10.1161/01.HYP.0000150784.92944.9a

37. Nickenig G. Central role of the AT(1)-receptor in atherosclerosis. J Hum Hypertens. 2002;16(Suppl 3):S26–S33. doi: 10.1038/sj.jhh.1001436 38. Graessler J, Schwudke D, Schwarz PE, Herzog R, Shevchenko A,

Bornstein SR. Top-down lipidomics reveals ether lipid deficiency in blood plasma of hypertensive patients. PLoS One. 2009;4:e6261. doi: 10.1371/journal. pone.0006261

39. Halperin RO, Sesso HD, Ma J, Buring JE, Stampfer MJ, Gaziano JM. Dyslipidemia and the risk of incident hypertension in men. Hypertension. 2006;47:45–50. doi: 10.1161/01.HYP.0000196306.42418.0e 40. Allemann Y, Horber FF, Colombo M, Ferrari P, Shaw S, Jaeger P, Weidmann P.

Insulin sensitivity and body fat distribution in normotensive offspring of hypertensive parents. Lancet. 1993;341:327–331. doi: 10.1016/0140- 6736(93)90135-4

41. Ferrari P, Weidmann P, Shaw S, Giachino D, Riesen W, Allemann Y, Heynen G. Altered insulin sensitivity, hyperinsulinemia, and dyslipidemia in individuals with a hypertensive parent. Am J Med. 1991;91:589–596. doi: 10.1016/0002-9343(91)90211-f

42. Yanai H, Tomono Y, Ito K, Furutani N, Yoshida H, Tada N. The underlying mechanisms for development of hypertension in the metabolic syndrome. Nutr J. 2008;7:10. doi: 10.1186/1475-2891-7-10

43. Fahy E, Subramaniam S, Brown HA, Glass CK, Merrill AH Jr, Murphy RC, Raetz CR, Russell DW, Seyama Y, Shaw W, et al. A compre-hensive classification system for lipids. J Lipid Res. 2005;46:839–861. doi: 10.1194/jlr.E400004-JLR200

44. Fahy E, Subramaniam S, Murphy RC, Nishijima M, Raetz CR, Shimizu T, Spener F, van Meer G, Wakelam MJ, Dennis EA. Update of the LIPID MAPS comprehensive classification system for lipids. J Lipid Res. 2009;50(Suppl) :S9–14. doi: 10.1194/jlr.R800095-JLR200

45. Pocock SJ, Bakris G, Bhatt DL, Brar S, Fahy M, Gersh BJ. Regres-sion to the Mean in SYMPLICITY HTN-3: Implications for Design and Reporting of Future Trials. J Am Coll Cardiol. 2016;68:2016–2025. doi: 10.1016/j.jacc.2016.07.775

46. Moore MN, Atkins ER, Salam A, Callisaya ML, Hare JL, Marwick TH, Nelson MR, Wright L, Sharman JE, Rodgers A. Regression to the mean of repeated ambulatory blood pressure monitoring in five studies. J Hypertens. 2019;37:24–29. doi: 10.1097/HJH.0000000000001977

47. Howard JP, Cole GD, Sievert H, Bhatt DL, Papademetriou V, Kandzari DE, Davies JE, Francis DP. Unintentional overestimation of an expected antihy-pertensive effect in drug and device trials: mechanisms and solutions. Int J Cardiol. 2014;172:29–35. doi: 10.1016/j.ijcard.2013.12.183

48. Spijkers LJ, van den Akker RF, Janssen BJ, Debets JJ, De Mey JG, Stroes ES, van den Born BJ, Wijesinghe DS, Chalfant CE, MacAleese L, et al. Hypertension is associated with marked alterations in sphingolipid biology: a potential role for ceramide. PLoS One. 2011;6:e21817. doi: 10.1371/journal.pone.0021817

49. Wang DD, Toledo E, Hruby A, Rosner BA, Willett WC, Sun Q, Razquin C, Zheng Y, Ruiz-Canela M, Guasch-Ferré M, et al. Plasma Ceramides, Medi-terranean Diet, and Incident Cardiovascular Disease in the PREDIMED Trial (Prevención con Dieta Mediterránea). Circulation. 2017;135:2028–2040. doi: 10.1161/CIRCULATIONAHA.116.024261

50. Fenger M, Linneberg A, Jørgensen T, Madsbad S, Søbye K, Eugen-Olsen J, Jeppesen J. Genetics of the ceramide/sphingosine-1-phosphate rheostat in blood pressure regulation and hypertension. BMC Genet. 2011;12:44. doi: 10.1186/1471-2156-12-44

51. Yogi A, Callera GE, Aranha AB, Antunes TT, Graham D, McBride M, Dominiczak A, Touyz RM. Sphingosine-1-phosphate-induced inflammation involves receptor tyrosine kinase transactivation in vascular cells: upregu-lation in hypertension. Hypertension. 2011;57:809–818. doi: 10.1161/ HYPERTENSIONAHA.110.162719

52. Jiang XC, Goldberg IJ, Park TS. Sphingolipids and cardiovascular diseases: lipoprotein metabolism, atherosclerosis and cardiomyopathy. Adv Exp Med Biol. 2011;721:19–39. doi: 10.1007/978-1-4614-0650-1_2

53. Spijkers LJ, Janssen BJ, Nelissen J, Meens MJ, Wijesinghe D, Chalfant CE, De Mey JG, Alewijnse AE, Peters SL. Antihypertensive treatment differ-entially affects vascular sphingolipid biology in spontaneously hypertensive rats. PLoS One. 2011;6:e29222. doi: 10.1371/journal.pone.0029222 54. Frangioudakis G, Diakanastasis B, Liao BQ, Saville JT, Hoffman NJ,

Mitchell TW, Schmitz-Peiffer C. Ceramide accumulation in L6 skeletal muscle cells due to increased activity of ceramide synthase isoforms has opposing effects on insulin action to those caused by palmitate treatment. Diabetologia. 2013;56:2697–2701. doi: 10.1007/s00125- 013-3035-5

CLINICAL AND POPULATION

STUDIES - AL

55. Frangioudakis G, Garrard J, Raddatz K, Nadler JL, Mitchell TW, Schmitz- Peiffer C. Saturated- and n-6 polyunsaturated-fat diets each induce ceramide accumulation in mouse skeletal muscle: reversal and improve-ment of glucose tolerance by lipid metabolism inhibitors. Endocrinology. 2010;151:4187–4196. doi: 10.1210/en.2010-0250

56. Sung KC, Lim S, Rosenson RS. Hyperinsulinemia and homeostasis model assessment of insulin resistance as predictors of hypertension: a 5-year follow-up study of Korean sample. Am J Hypertens. 2011;24:1041–1045. doi: 10.1038/ajh.2011.89

57. Zhou MS, Wang A, Yu H. Link between insulin resistance and hyperten-sion: What is the evidence from evolutionary biology? Diabetol Metab Syndr. 2014;6:12. doi: 10.1186/1758-5996-6-12

58. Skarfors ET, Lithell HO, Selinus I. Risk factors for the development of hyper-tension: a 10-year longitudinal study in middle-aged men. J Hypertens. 1991;9:217–223. doi: 10.1097/00004872-199103000-00004 59. Haffner SM, Miettinen H, Gaskill SP, Stern MP. Metabolic

precur-sors of hypertension. The San Antonio Heart Study. Arch Intern Med. 1996;156:1994–2001.

60. Sinaiko AR, Steinberger J, Moran A, Hong CP, Prineas RJ, Jacobs DR Jr. Influence of insulin resistance and body mass index at age 13 on systolic blood pressure, triglycerides, and high-density lipo-protein cholesterol at age 19. Hypertension. 2006;48:730–736. doi: 10.1161/01.HYP.0000237863.24000.50

61. Ormazabal V, Nair S, Elfeky O, Aguayo C, Salomon C, Zuñiga FA. Association between insulin resistance and the development of car-diovascular disease. Cardiovasc Diabetol. 2018;17:122. doi: 10.1186/s12933- 018-0762-4

62. Greene EL, Lu G, Zhang D, Egan BM. Signaling events mediat-ing the additive effects of oleic acid and angiotensin II on vascular smooth muscle cell migration. Hypertension. 2001;37:308–312. doi: 10.1161/01.hyp.37.2.308

63. Liu Y, Chen T, Qiu Y, Cheng Y, Cao Y, Zhao A, Jia W. An ultrasonication-assisted extraction and derivatization protocol for GC/TOFMS-based metabolite profiling. Anal Bioanal Chem. 2011;400:1405–1417. doi: 10.1007/s00216-011-4880-z

64. Kim SR, Jeon SY, Lee SM. The association of cardiovascular risk factors with saturated fatty acids and fatty acid desaturase indices in erythro-cyte in middle-aged Korean adults. Lipids Health Dis. 2015;14:133. doi: 10.1186/s12944-015-0135-x

65. Yang M, Yu Z, Deng S, Chen X, Chen L, Guo Z, Zheng H, Chen L, Cai D, Wen B, et al. A Targeted Metabolomics MRM-MS Study on Identifying Potential Hypertension Biomarkers in Human Plasma and Evaluating Acu-puncture Effects. Sci Rep. 2016;6:25871. doi: 10.1038/srep25871 66. Gremmels H, Bevers LM, Fledderus JO, Braam B, van Zonneveld AJ,

Verhaar MC, Joles JA. Oleic acid increases mitochondrial reactive oxygen species production and decreases endothelial nitric oxide synthase activ-ity in cultured endothelial cells. Eur J Pharmacol. 2015;751:67–72. doi: 10.1016/j.ejphar.2015.01.005

67. Yang Q, Alemany R, Casas J, Kitajka K, Lanier SM, Escribá PV. Influence of the membrane lipid structure on signal processing via G protein-coupled recep-tors. Mol Pharmacol. 2005;68:210–217. doi: 10.1124/mol.105.011692 68. Funari SS, Barceló F, Escribá PV. Effects of oleic acid and its

conge-ners, elaidic and stearic acids, on the structural properties of phospha-tidylethanolamine membranes. J Lipid Res. 2003;44:567–575. doi: 10.1194/jlr.M200356-JLR200

69. Information. NCfB. 17:0 cholesteryl ester, cid=24779605. PubChem Database. Accessed August 8, 2019. https://pubchem.ncbi.nlm.nih.gov/ compound/17_0-Cholesteryl-ester.

70. Ghosh S, Zhao B, Bie J, Song J. Macrophage cholesteryl ester mobi-lization and atherosclerosis. Vascul Pharmacol. 2010;52:1–10. doi: 10.1016/j.vph.2009.10.002

71. van Popele NM, Grobbee DE, Bots ML, Asmar R, Topouchian J, Reneman RS, Hoeks AP, van der Kuip DA, Hofman A, Witteman JC. Asso-ciation between arterial stiffness and atherosclerosis: the Rotterdam Study. Stroke. 2001;32:454–460. doi: 10.1161/01.str.32.2.454

72. Gavish B, Izzo JL Jr. Arterial Stiffness: Going a Step Beyond. Am J Hyper-tens. 2016;29:1223–1233. doi: 10.1093/ajh/hpw061

73. Hu C, Kong H, Qu F, Li Y, Yu Z, Gao P, Peng S, Xu G. Application of plasma lipidomics in studying the response of patients with essential hypertension to antihypertensive drug therapy. Mol Biosyst. 2011;7:3271–3279. doi: 10.1039/c1mb05342f