Vous êtes sur la page 1sur 10

Research Article

Lifestyle Genomics Received: July 3, 2018

Accepted: October 8, 2018
DOI: 10.1159/000494331 Published online: November 23, 2018

Variants in APOA5 and ADIPOQ Moderate

Improvements in Metabolic Syndrome
during a One-Year Lifestyle Intervention
Dana E. Lowry a Peri H. Fenwick a Kaitlin Roke a Khursheed Jeejeebhoy b

Rupinder Dhaliwal c Paula Brauer d Dawna Royall d Angelo Tremblay e


Doug Klein f David M. Mutch a

a Department of Human Health and Nutritional Sciences, University of Guelph, Guelph, ON, Canada; b Emeritus

Professor of Medicine and Physician, St. Michael’s Hospital, Toronto, ON, Canada; c Metabolic Syndrome Canada,

Kingston, ON, Canada; d Department of Family Relations and Applied Nutrition, University of Guelph, Guelph, ON,

Canada; e Department of Kinesiology, Faculty of Medicine, Université Laval, Québec City, QC, Canada; f Department

of Family Medicine, University of Alberta, Edmonton, AB, Canada

Keywords vention. Associations between SNPs and the continuous

Metabolic syndrome · Lifestyle intervention · Continuous MetS (cMetS) score, as well as individual MetS components,
MetS score · Single nucleotide polymorphism · Gene were examined. Results: Reductions in cMetS score at both
association · Nutrigenomics 3 months and 1 year were significantly associated with 2 vari-
ants: rs662799 (A/G) in apolipoprotein A5 (APOA5) and
rs1501299 (G/T) in adiponectin (ADIPOQ). Individuals carry-
Abstract ing a minor T allele in rs1501299 experienced a greater re-
Background: Metabolic syndrome (MetS) comprises a clus- duction in cMetS score at both 3 months and 1 year, where-
ter of risk factors including central obesity, hypertension, as major allele AA homozygotes in rs662799 experienced
dyslipidemia, and impaired glucose homeostasis. Lifestyle greater reductions in cMetS score during the intervention.
interventions that promote improvements in diet quality No associations were identified between the aforemen-
and physical activity represent a first line of therapy for MetS. tioned SNPs and individual components of MetS. Both un-
However, varying responses to lifestyle interventions are weighted and weighted genetic risk scores (GRS) using these
well documented and may be partially explained by underly- 2 SNPs revealed that individuals carrying none of the risk al-
ing genetic differences. The aim of this study was to investi- leles experienced significantly greater reductions in cMetS
gate if variants in genes previously associated with MetS in- score after 1 year. Conclusions: The findings from the cur-
fluence the magnitude of change in MetS risk during a 1-year rent study suggest that individuals with certain genotypes
lifestyle intervention. Methods: The present study used data may benefit more from a lifestyle intervention for MetS and
collected from the Canadian Health Advanced by Nutrition that specific variants, either independently or as part of a
and Graded Exercise study cohort (n = 159 men and women) GRS, could be used as a nutrigenomic tool to tailor the inter-
to investigate the effect of 17 candidate single nucleotide vention to reduce the risk of MetS. © 2018 S. Karger AG, Basel
polymorphisms (SNPs) on response to a 1-year lifestyle inter- - 11/26/2018 9:26:22 PM

© 2018 S. Karger AG, Basel David M. Mutch

Stockholm University Library

Department of Human Health and Nutritional Sciences

University of Guelph, 50 Stone Road East
E-Mail karger@karger.com
Guelph, ON N1G 2W1 (Canada)
Downloaded by:

E-Mail dmutch @ uoguelph.ca

Introduction in MetS risk. The primary objective of the present study
was to examine the influence of SNPs previously found to
Metabolic syndrome (MetS) encompasses a cluster of be associated with MetS on changes in cMetS score, as
factors that contribute to the risk for developing type 2 well as individual MetS components, during a 1-year life-
diabetes (T2D) and cardiovascular disease [1]. The prev- style intervention. A secondary objective was to develop
alence of MetS is estimated to be as high as 40% in some un-weighted (uw) and weighted (w) genetic risk scores
populations, particularly in developed countries and (GRS) to examine their impact on changes in cMetS score.
older generations [2]. Although nuances exist regarding We hypothesized that genetic variants would moderate
the definitions for MetS proposed by various organiza- improvements in cMetS score in response to a lifestyle
tions, the condition is generally defined as the clinical intervention. We anticipate that our findings will help to
manifestation of 3 or more of the following 5 risk factors: further personalize lifestyle interventions that aim to re-
central obesity (defined using waist circumference), el- duce MetS risk.
evated fasting blood glucose, hypertension, elevated tri-
glycerides, and reduced high-density lipoprotein choles-
terol (HDL-C) [3]. Diagnosing MetS is generally done by Materials and Methods
examining whether the five aforementioned risk factors
CHANGE Feasibility Study Overview
lie above or below defined threshold values. However,
The data used for the current analyses corresponds to the sub-
the use of dichotomous assessments is not without criti- set of individuals enrolled in the CHANGE feasibility study who
cism, which prompted the development of a continuous provided a DNA sample (Fig. 1). For a thorough description of the
MetS (cMetS) score [4]. Specifically, the cMetS score was CHANGE feasibility study, as well as inclusion and exclusion cri-
developed by Hillier and colleagues as a practical meth- teria, please refer to Jeejeebhoy et al. [10]. The CHANGE feasibil-
ity study was a prospective, longitudinal before-after demonstra-
od to evaluate MetS risk using principal component
tion study carried out at 3 primary healthcare clinics across Cana-
analysis. These authors demonstrated that each standard da, recruiting individuals with MetS between 2012 and 2014. The
deviation (SD) increase in the cMetS score was associ- primary objective of the CHANGE study was to reverse MetS and
ated with an increased risk of T2D and cardiovascular improve its components by the end of the intervention period.
disease. This year-long personalized lifestyle intervention used a team-
based approach comprising a primary care physician, a kinesiolo-
Initial treatment for MetS typically includes a lifestyle
gist, and a dietitian. The goal of the intervention was to create sus-
intervention comprising changes in diet and physical ac- tainable behavioral changes within the cohort by adhering to an
tivity, which has been shown to reduce both the severity individual diet plan broadly based on the Mediterranean diet prin-
of MetS and dependency on pharmacological therapy [5]. ciples and a combination of weekly aerobic, resistance, and flexi-
While lifestyle interventions for MetS are generally suc- bility exercises [11, 12]. Diet quality was assessed using the Cana-
dian Healthy Eating Index (HEI-C), which provides a score rang-
cessful, a wide range of responses are observed between
ing from 0 to 100 based on a person’s adherence to Canada’s Food
people, with some individuals showing great improve- Guide recommendations [13]. Dieticians met with participants to
ments, while others show little-to-no improvement [6– create individualized dietary plans based on a care map that inte-
8]. These variable responses to lifestyle interventions may grated evidence-based dietary advice with principles from behav-
stem, in part, from underlying genetic differences [6]. In- ior change models [12]. Participants also met with kinesiologists
to develop individualized fitness plans that included supervised
deed, hundreds of single nucleotide polymorphisms
and unsupervised activities to improve aerobic, resistance, and
(SNPs) have been associated with MetS as a whole, as well flexibility training [11]. Aerobic capacity was evaluated using max-
as with the individual MetS components [1]. Moreover, imal oxygen consumption (VO2max) [14]. All participants attended
many of these SNPs have also been reported to influence weekly visits (over the first 3 months) and monthly visits (over the
a person’s response to lifestyle interventions [6]. Conse- final 9 months) with a dietitian and kinesiologist, and were fol-
lowed by a primary physician quarterly. Standard blood clinical
quently, it stands to reason that SNPs may influence the measurements were completed at baseline, 3- and 12-months to
degree of improvement experienced by individuals un- monitor common biomarkers of metabolic health (e.g., lipids, glu-
dergoing lifestyle modifications for MetS [9]. cose). The cMetS score was calculated by combining the weighted
The Canadian Health Advanced by Nutrition and effects of waist circumference, triglycerides, blood glucose, and
Graded Exercise (CHANGE) feasibility study evaluated a systolic blood pressure using the approach developed by Hillier
and colleagues [4]. Written and oral informed consent was ob-
year-long tailored lifestyle intervention for individuals tained from all eligible patients before inclusion. This protocol was
with MetS [10]. We previously reported that the CHANGE approved by a Research Ethics Board of the participating clinics
study resulted in a 19% reversal of MetS [10]; however, it and affiliated universities.
is unknown if genetic variants moderate improvements - 11/26/2018 9:26:22 PM

2 Lifestyle Genomics Lowry et al.

Stockholm University Library

DOI: 10.1159/000494331
Downloaded by:
Table 1. Candidate SNPs associated with individual MetS components selected

MetS component Gene SNP Major/ MAF based on MAF in CHANGE Ref.
minor allele 1000 Genomes study participants

Blood pressure ATP2B1 rs17249754 G/A 0.2095 0.1572 15

ACE rs4343 A/G 0.3568 0.4840 16
Fasting blood glucose GLUT2 rs5400 G/A 0.2153 0.1604 17
TCF7L2 rs12255372 G/T 0.2139 0.2830 6
rs7903146 C/T 0.2278 0.3050 1
ADIPOQ rs1501299 G/T 0.3003 0.2642 18
Dyslipidemia FADS1 rs174537 G/T 0.3029 0.3584 19
CETP rs1800775 A/C 0.4535 0.4748 20
rs247617 C/A 0.2680 0.2893 21
rs5882 A/G 0.4661 0.2987 22
APOC3 rs2854117 C/T 0.4992 0.2642 23
APOA5 rs662799 A/G 0.1629 0.0912 6
rs964184 C/G 0.2222 0.1792 6
APOA1 rs670 C/T 0.1885 0.1950 24
Waist circumference MC4R rs12970134 G/A 0.2075 0.2421 1
rs17782313 T/C 0.2400 0.3073 6
FTO rs9939609 T/A 0.3401 0.4308 6

MAF, minor allele frequency.

DNA Extraction and Genotyping ity control. Twelve DNA samples were randomly selected for rep-
Blood samples were drawn at baseline, 3 months and 12 months lication and 100% concordance was achieved. Hardy-Weinberg
and stored at –80 ° C until analysis. DNA was extracted from whole
     equilibrium was evaluated for all SNPs using a χ2 test.
blood using the Qiagen PAXgene Blood DNA kit, according to
manufacturer instructions (Qiagen, Toronto, ON, Canada). DNA Genetic Risk Score
quantity was determined using a NanoDrop 2000c (Fisher Scien- The SNPs used to build un-weighted (uw) and weighted (w)
tific, Waltham, MA, USA) and quality was visually assessed on a GRSs were identified based on associations between individual
1% agarose gel. DNA samples were collected from each participant SNPs and change in cMetS score. An uwGRS ranging from 0 to 2
at baseline and 12 months; therefore, the sample with the highest was built, where a participant’s GRS was calculated by summing
quality was used for genotyping. the number of risk alleles for 2 SNPs: rs662799 (G) and rs1501299
(G). A GRS = 0 corresponds to a participant carrying no risk alleles,
SNP Selection GRS = 1 corresponds to a participant carrying one risk allele, and
SNPs of interest were selected based on an extensive search of a GRS = 2 corresponds to a participants carrying both risk alleles.
the existing literature. The search criteria included SNPs that have The wGRS was built by taking into account the effect size (β/stan-
been previously associated with MetS in genome-wide association dard error) of each SNP, as revealed from the individual additive
studies, lifestyle intervention studies, and/or meta-analyses. Only models for rs662799 and rs1501299.
SNPs with a minor allele frequency > 10% according to the 1000
Genomes Project (http://www.internationalgenome.org/; ac- Statistical Analyses
cessed 09/2016) were selected. A panel of 17 SNPs corresponding Linear regressions were used to investigate associations be-
to 12 genes was chosen for the present analysis (Table 1) [1, 6, tween SNPs and cMetS score, as well as between SNPs and indi-
15–24]. vidual MetS components. Models accounted for the following co-
variates: age, sex, BMI, ethnicity, change in HEI-C, participating
SNP Analysis site, and baseline medication. Medication was treated as a dichot-
All DNA samples were diluted to a concentration of 20 ng/µL. omous variable (Y/N), where participants using any medication
Genotyping was performed at the Centre for Applied Genomics related to a MetS component were considered “Y”. BMI was not
(The Hospital for Sick Children, Toronto, ON, Canada) using the included as a covariate when investigating cMetS score and waist
Sequenom MassARRAY platform, which is based on detection circumference. We constructed both dominant (MM vs. Mm +
through MALDI-TOF MS (Mass Array, Sequenom, San Diego, mm) and additive (MM vs. Mm. vs. mm) genetic models for all
CA, USA). Positive and negative controls, made up of a Yoruban regression analyses. The term “minor allele carrier” is used
HapMap trio and water samples, respectively, were used for qual- throughout the manuscript to refer to participants who are either - 11/26/2018 9:26:22 PM

Lifestyle-Gene Interactions and MetS Lifestyle Genomics 3

Stockholm University Library

DOI: 10.1159/000494331
Downloaded by:
haran African/Mediterranean/Arab ethnicity, 10 (6.3%)
Enrolled in the individuals of Asian/South Central American ethnicity,
CHANGE program and 2 (1.3%) individuals of unknown ethnicity.
(n = 305)
Participants were eligible at screening Baseline, 3- and 12-month characteristics for the 159
but did not meet definition of MetS at participants are presented in Table 2. Consistent with the
baseline assessment (n = 12 excluded)
Included in original CHANGE feasibility study, in this subset of pa-
CHANGE baseline tients both the HEI-C and estimated VO2max improved
analysis (n = 293)
significantly by 3 months and these improvements were
Participants without DNA samples
available (n = 53 excluded) maintained at 12 months (p < 0.0001 for both). The cMetS
Genotyped for score decreased at 3 months and the reduction was main-
SNPs of interest tained at 12 months (p < 0.0001). Individual MetS com-
(n = 240)
ponents significantly improved across the duration of the
12-month cMetS score unavailable
(n = 81 excluded) study in our subset of participants. Blood lipid profiles
Included in current were improved at 3 months and maintained at 12 months.
gene-cMetS score Specifically, triglyceride levels were reduced (p = 0.0006)
analysis (n = 159)
and HDL-C levels were increased (p < 0.0001). These im-
provements aligned with trends in other blood lipid
markers, that is, a reduction in LDL-C (p = 0.07) and an
Fig. 1. CHANGE CONSORT flow diagram.
increase in APOA1 (p = 0.07). Waist circumference de-
creased significantly after 3 months (p < 0.0001), which
reflected reductions in BMI. Both systolic and diastolic
heterozygous or minor homozygous carriers for a given SNP. The blood pressure were reduced after 3 months (both p <
change in cMetS score at 3 and 12 months was calculated as “cMetS 0.0001) and these reductions persisted at 12 months. Fast-
score at 3-months – cMetS score at baseline” and “cMetS score at ing blood glucose (p = 0.3) was the only MetS component
12-months – cMetS score at baseline” respectively. Since we chose
to not account for multiple testing, we only considered associa-
that did not change during the intervention.
tions that were statistically significant at both time points, and in
both dominant and additive models, to reduce the risk of reporting SNPs Influence the Magnitude of Change in the cMetS
false positives. Score
Prior to analyses, data were assessed for normality using a Sha- Seventeen candidate SNPs previously reported in the
piro-Wilk test. A Friedman’s repeated measures test was used
when analyzing continuous data in study participants at baseline, literature to be associated with individual MetS compo-
3- and 12-months. A χ2 test was used to compare categorical data nents were analyzed in the present investigation. All SNPs
at baseline, 3- and 12-months. All data are presented as mean ± SE. but one (rs17782313) were in Hardy-Weinburg equilib-
Linear regressions were performed with JMP 13 Statistical Soft- rium. No significant associations were observed between
ware (SAS Institute, Cary, NC, USA). All other statistical tests were the 17 candidate SNPs and the cMetS score at baseline
performed using GraphPad 6 Prism (GraphPad Software, Inc., CA,
USA). A p < 0.05 was considered statistically significant. (data not shown). However, examining the change in
cMetS score at both 3 and 12 months revealed several sta-
tistically significant and consistent associations with var-
ious SNPs (Table 3). In particular, rs662799 (A/G) in
Results APOA5 and rs1501299 (G/T) in ADIPOQ were associated
with change in cMetS score at both 3- and 12-months.
Characterization of the CHANGE Cohort Furthermore, these associations were significant in both
Participants who did not meet our criteria for MetS at dominant and additive genetic models. Individuals ho-
baseline, or were missing either a DNA sample or mozygous for the major A allele for rs662799 experienced
12-month cMetS data, were excluded from the present greater reductions in cMetS score compared to AG + GG
analysis (Fig.  1). The final sample size for the present carriers at both time points (Fig.  2a). In contrast, indi-
analyses was 159 (77 males and 82 females). The mean age viduals carrying a minor T allele in rs1501299 (GT + TT)
of participants was 60.7 ± 0.73 years (median of 62; range experienced greater reductions in cMetS score at both
18–75). The majority of the study population was Cauca- time points (Fig.  2b). Significant associations were ob-
sian (n = 131, 82.4%), while the remainder of the cohort served between other SNPs and the change in cMetS
comprised of 16 (10%) individuals of European/Sub-Sa- score; however, these associations were not consistent - 11/26/2018 9:26:22 PM

4 Lifestyle Genomics Lowry et al.

Stockholm University Library

DOI: 10.1159/000494331
Downloaded by:
Table 2. Characteristics of participants at baseline, 3, and 12 months

Characteristics Baseline 3 months 12 months p value

(n = 159) (n = 148) (n = 159)

Age, years 60.7±0.73 – – –

Female, n (%) 82 (51.6) – – –
BMI, kg/m2 31.0±0.27a 30.4±0.27b 30.1±0.29b <0.0001
Current smoker, n (%) 14 (8.8) 13 (8.8) 12 (7.55) 0.9
HEI-C 58.2±1.2a 68.4±1.0b 68.6±1.1b <0.0001
VO2max, mL/kg/min 32.2±0.57a 34.9±0.57b 35.2±0.55b <0.0001
LDL-C, mmol/L 2.54±0.08 2.49±0.09 2.49±0.08 0.07
APOA1, mmol/L* 1.43±0.02 Not measured 1.46±0.02 0.06
cMetS score 2.36±0.08a 1.79±0.08b 1.94±0.09b <0.0001
Metabolic syndrome criteria#
1. Elevated blood pressure or using pharmacotherapy, n (%) 144 (90.6) 135 (84.9) 131 (82.4) 0.1
Systolic blood pressure, mm Hg 133.9±1.2a 126.6±1.1b 130.2±1.1c <0.0001
Diastolic blood pressure, mm Hg 79.9±0.72a 76.7±0.72b 77.1±0.66b <0.0001
Received pharmacotherapy for elevated blood pressure, n (%) 128 (80.5) 117 (79.1) 127(79.9) 0.95
2. Elevated fasting blood glucose or using pharmacotherapy, n (%) 135 (84.9) 122 (76.7) 125 (78.6) 0.16
Blood glucose, mmol/L 6.39±0.09 6.26±0.09 6.44±0.11 0.3
Received pharmacotherapy for elevated blood glucose levels, n (%) 71 (44.7) 65 (43.9) 70 (44.0) 0.99
3. Elevated triglycerides or using pharmacotherapy, n (%) 98 (61.6) 78 (49.1) 83 (52.2) 0.06
Triglyceride level, mmol/L 2.22±0.16a 1.83±0.1b 1.88±0.07b 0.0006
Pharmacotherapy for dyslipidemia, n (%) 6 (3.7) 6 (4.1) 6 (3.7) 0.99
4. Reduced HDL-C, n (%) 79 (50.0) 83 (52.2) 60 (37.7) 0.02
HDL-C, mmol/L 1.19±0.02a 1.17±0.02 a 1.25±0.02b <0.0001
5. Large waist circumference, n (%) 149 (93.7) 135 (84.9) 129 (81.1) 0.003
Waist circumference, cm 106.1±0.68a 103.4±0.70b 102.1±0.79b <0.0001

Values denote mean ± SE. Continuous data were analyzed using a Friedman’s repeated measures test, while categorical data were
compared using a χ2 test. Values within a row that have a different superscript letter are statistically significantly different from one
another (p < 0.05).
BMI, body mass index; HDL-C, high-density lipoprotein cholesterol; HEI-C, Healthy Eating Index-Canadian; LDL-C, low-density
lipoprotein cholesterol; VO2max, maximal oxygen consumption.
* APOA1 was not measured at 3 months; therefore, a Wilcoxon repeated measures test was used. #Metabolic syndrome criteria were
defined as follows: blood pressure ≥130/85 mm Hg or receiving pharmacotherapy; fasting blood glucose ≥5.6 mmol/L or receiving
pharmacotherapy; triglyceride level ≥1.7 mmol/L or receiving pharmacotherapy; male patients with an HDL-C level <1.0 mmol/L or
female patients with an HDL-C level <1.3 mmol/L; waist circumference as determined by a prespecified technique (Europid, white, sub-
Saharan African, Mediterranean, middle eastern [Arab] patients ≥94 cm for men, 80 cm for women; Asian and South Central American
patients ≥90 cm for men, 80 cm for women; white American and Canadian patients ≥102 cm for men, 88 cm for women.

across time points and/or models. Therefore, these SNPs shown). In contrast, the 12-month change in cMetS score
were not considered in the construction of GRSs. was significantly associated with uwGRS (p = 0.0006;
Fig. 3), with a similar trend seen at 3 months (p = 0.005;
Genetic Risk Score and Change in the cMetS Score data not shown). Individuals with an uwGRS of 0 (i.e., car-
We established GRSs corresponding to the 2 aforemen- rying no risk alleles) showed greater reductions in their
tioned SNPs: rs662799 and rs1501299. These 2 SNPs were cMetS score compared to individuals with at least 1 risk
associated with a change in cMetS score at both 3- and allele (0 vs. 1, p = 0.05; 0 vs. 2, p = 0.004). Although we did
12-months using both dominant and additive genetic not detect statistical differences between individuals car-
models. At baseline, no associations were identified be- rying 1 or 2 risk alleles, it appears that the change in cMetS
tween the uwGRS (or wGRS) and cMetS score (data not score is reduced as the number of risk alleles increases. - 11/26/2018 9:26:22 PM

Lifestyle-Gene Interactions and MetS Lifestyle Genomics 5

Stockholm University Library

DOI: 10.1159/000494331
Downloaded by:
pgenotype = 0.002 pgenotype = 0.002
ptime = 0.06 ptime = 0.06
pinteraction = 0.52 pinteraction = 0.99
0 0
Change in cMetS score

Change in cMetS score

from baseline

from baseline
rs662799 –0.4 rs1501299
–0.4 ■ AA ■ GG
■ AG + GG ■ GT + TT


–0.8 –1.0
3 12 3 12
a Time point, months b Time point, months

Fig. 2. Associations between individual SNPs and change in cMetS score. Data obtained using a dominant ge-
netic model is presented for rs662799 in APOA5 (a) and rs1501299 in ADIPOQ (b). All linear regressions were
adjusted for sex, age, ethnicity, HEI-C, participating site and baseline medication. A 2-way ANOVA was used to
examine the main effects of Genotype and Time, as well as the Genotype × Time interaction.

Table 3. Associations between candidate SNPs and change in cMetS score at 3 and 12 months

Gene SNP 3-month change in cMetS score 12-month change in cMetS score
additive model, dominant model, additive model, dominant model,
p, β value p, β value p, β value p, β value

ATP2B1 rs17249754 0.99, 0.01 0.67, 0.04 0.27, 0.09 0.36, –0.07
ACE rs4343 0.81, 0.10 0.46, –0.06 0.56, 0.05 0.02, –0.19
GLUT2 rs5400 0.86, 0.01 0.85, 0.02 0.74, –0.03 0.61, –0.04
TCF7L2 rs12255372 0.05, –0.17 0.07, –0.15 0.24, –0.10 0.33, –0.08
rs7903146 0.10, –0.14 0.21, –0.11 0.25, –0.10 0.35, –0.08
ADIPOQ rs1501299 0.02, 0.2 0.05, 0.16 0.006, 0.22 0.01, 0.21
FADS1 rs174537 0.76, 0.03 0.14, 0.12 0.66, –0.04 0.56, 0.05
CETP rs1800775 0.83, 0.02 0.55, –0.05 0.44, –0.07 0.28, –0.09
rs247617 0.96, 0.01 0.87, 0.01 0.09, 0.14 0.14, 0.13
rs5882 0.8, –0.02 0.77, –0.02 0.31, –0.08 0.38, –0.07
APOC3 rs2854117 0.06, –0.16 0.19, –0.11 0.02, –0.19 0.13, –0.12
APOA5 rs662799 0.02, 0.19 0.03, 0.18 0.004, 0.23 0.02, 0.19
rs964184 0.23, –0.1 0.27, –0.09 0.1, –0.13 0.15, –0.12
APOA1 rs670 0.21, 0.1 0.05, 0.16 0.13, 0.12 0.02, 0.19
MC4R rs12970134 0.98, –0.01 0.96, 0.01 0.7, 0.03 0.43, 0.06
rs17782313 0.84, –0.02 0.80, –0.02 0.71, 0.03 0.75, 0.03
FTO rs9939609 0.83, 0.02 0.5, –0.06 0.18, –0.11 0.32, –0.08

Data corresponding to dominant and additive genetic models for all linear regressions are presented. Models
accounted for the following covariates: HEI-C, BMI, age, ethnicity, gender, recruitment site, and baseline
Statistically significant associations are highlighted with bold font (p < 0.05). - 11/26/2018 9:26:22 PM

6 Lifestyle Genomics Lowry et al.

Stockholm University Library

DOI: 10.1159/000494331
Downloaded by:
viduals mirror the findings from the entire CHANGE
0.5 study cohort as previously reported [10]. The decrease in
puwGRS = 0.0006
cMetS score, increase in HEI-C and VO2max, as well as
Change in cMetS score

improvements in all MetS components except fasting

b blood glucose, highlight the effectiveness of the lifestyle

intervention used in the CHANGE feasibility study. The

current investigation revealed that the magnitude of the
–0.5 b
reduction in cMetS score in response to the lifestyle inter-
vention is moderated by common SNPs in 2 genes.
–1.0 APOA-V is an important regulator of circulating tri-
0 1 2 glyceride levels through 2 primary mechanisms. First, cir-
uwGRS culating APOA-V can interact with lipoprotein lipase on
the luminal side of endothelial cells to promote the hy-
drolysis of triglyceride-rich lipoproteins. Second, APOA-
Fig. 3. Un-weighted (uw) genetic risk score (uwGRS) is associated
with change in cMetS score at 12 months. The uwGRS was calcu- V, can bind the hepatic LDL-receptor to encourage the
lated by summing the number of risk alleles in the 2 SNPs (rs662799 clearance of lipoprotein remnants [25]. Previous reports
and rs1501299) that were independently associated with change in have shown that MetS prevalence and the risk of MetS is
cMetS score, where 0 corresponds to individuals carrying no risk increased in individuals carrying the minor G allele in
alleles, 1 corresponds to individuals carrying one risk allele, and so rs662799 (also known as –1131T>C), a SNP located in the
on. Bars sharing letters are not significantly different from one an-
other (p < 0.05). The number of individuals in each group: GRS = upstream promoter region of the APOA5 gene [26–29].
0 (n = 56), GRS = 1 (n = 87), and GRS = 2 (n = 16). Specifically, people carrying the minor allele have been
shown to have elevated triglyceride levels and, in some
instances, reduced HDL-C levels in both healthy and un-
healthy individuals [29, 30].
In the present study, minor allele carriers showed
The wGRS showed similar results, with significant asso- smaller reductions in cMetS score in response to the life-
ciations seen with the 12-month change in cMetS (p = style intervention compared to major allele homozygotes.
0.003) and the 3-month change in cMetS (p = 0.002). These findings align with previous reports investigating
the modifying effect of this SNP on triglyceride levels dur-
ing lifestyle and/or dietary interventions. For example, a
Discussion 3-year trial in which Korean individuals with impaired
fasting glucose or newly diagnosed T2D replaced all re-
The present study examined whether genetic variants fined white rice with whole grains, barley, or legumes and
influence the magnitude of the reduction in cMetS score increased vegetable intake, showed that those carrying the
during a year-long lifestyle intervention in a subset of minor allele had lower plasma APOA-V levels and higher
participants from the CHANGE feasibility study. The pri- triglyceride levels at both baseline and following the inter-
mary findings of this investigation were: (1) 2 SNPs were vention compared to individuals carrying the major allele
consistently associated with a change in cMetS score at [31]. Additionally, a 3-month intervention consisting of
both 3 and 12 months (rs662799 [A/G] in APOA5 and diet quality improvements and regular walking in indi-
rs1501299 [G/T] in ADIPOQ), and (2) participants carry- viduals with hypertriglyceridemia showed greater im-
ing none of the risk alleles in the 2 aforementioned genes provements in major allele carriers for both APOA-V and
(GRS = 0) showed the greatest reduction in cMetS score triglyceride levels compared to minor allele carriers, de-
during the intervention. Together, this study identified spite a similar degree of compliance [32]. In contrast, the
novel lifestyle-gene interactions and showed that individ- aforementioned genotype effect in response to a lifestyle
uals carrying neither risk alleles in APOA5 and ADIPOQ intervention was not observed in a Japanese cohort [33].
may experience an improved response to a lifestyle inter- Collectively, these studies suggest a potential moderating
vention for MetS. effect for rs662799 on blood triglyceride levels in response
Although not the primary objective of this genetic- to longer-term lifestyle interventions. Interestingly, previ-
based study, the improvements in biochemical measures ous reports suggest that the frequency of the minor “risk”
and dietary behaviors observed within our subset of indi- allele is more common in the Asian population compared - 11/26/2018 9:26:22 PM

Lifestyle-Gene Interactions and MetS Lifestyle Genomics 7

Stockholm University Library

DOI: 10.1159/000494331
Downloaded by:
to those of European descent [31]. This suggests that the reduction in cMetS score seen in individuals carrying no
moderating effect of this SNP in response to a lifestyle in- risk alleles (GRS = 0) compared to individuals with at least
tervention should be investigated more closely in distinct 1 risk allele highlights the need to consider overall risk
subgroups of the general population. scores rather than individual SNPs. To the best of our
Adiponectin is a well-known adipokine encoded by knowledge, this is one of the first studies to examine GRSs
the ADIPOQ gene. Circulating adiponectin levels have in the context of a lifestyle intervention for MetS. A recent
been associated with improvements in both MetS and in- study by San-Cristobal et al. [39] used a GRS comprised
sulin resistance, and genetic variants in ADIPOQ have of 14 SNPs to explore associations between Mediterra-
been shown to influence circulating levels [34]. For nean Diet adherence, blood biomarkers, and genetic back-
example, individuals carrying the minor T allele in ground in volunteers of the Food4Me study. Individuals
rs1501299 were reported to have higher adiponectin lev- were categorized into “low” and “high” GRS groups, with
els in both Asian and European populations [35, 36]. The those having a low GRS experiencing a greater reduction
present study demonstrated that rs1501299 genotype is in total cholesterol levels after 6-months compared to
associated with reductions in cMetS score. Specifically, those with a high GRS. The SNPs used to create a GRS in
individuals carrying the minor T allele (GT + TT) showed this past study are different than those used in the present
a greater response to the lifestyle intervention compared study; however, the results align to suggest that individu-
to those homozygous for the major G allele. This is in- als with a greater number of “risk” alleles would benefit
triguing, as evidence suggests that individuals homozy- from more tailored strategies to improve health outcomes.
gous for the G allele are at greater metabolic risk than Conversely, an unfavorable genetic profile may predict an
T-allele carriers. For example, a recent meta-analysis of adverse response to what would be normally expected
Chinese Han populations reported a greater G allele fre- from healthy eating and exercise training.
quency in individuals with MetS [37]. Further, individu- The present study has several limitations and strengths
als homozygous for the GG genotype showed impaired to be considered. Primary limitations include the rela-
glucose tolerance in Spanish subjects [35]. We did not tively small sample size (n = 159), minor differences in
identify any associations between rs1501299 and fasting diet, and exercise alterations between individuals due to
glucose levels at baseline or during the intervention; personalized advice (as opposed to a standardized life-
therefore, we are unable to hypothesize why the change style intervention), lack of information regarding alcohol
in cMetS score is influenced by the rs1501299 genotype. intake, and the fact that all subjects had MetS at recruit-
Shin et al. [38] examined the association between ment (i.e., no control group) may have limited our ability
rs1501299 and circulating adiponectin and insulin resis- to detect certain gene associations (e.g., rs670 and HDL-
tance in response to a 12-week weight loss intervention. C). Future studies to validate the genetic findings on sub-
In contrast to our findings, these authors showed that GG sequent MetS intervention populations are warranted.
homozygotes showed significant improvements in insu- However, a major strength of the CHANGE study was the
lin sensitivity and increases in adiponectin following the duration of the intervention and data collection at mul-
weight-loss intervention, with little-to-no change seen in tiple time points. Furthermore, the use of 3 diverse par-
carriers of the T allele. It is unclear why our results, which ticipating sites across Canada means our results are gen-
show that T allele carriers experienced greater improve- eralizable, at minimum across North America. Moreover,
ments in metabolic health in response to the lifestyle in- having a physician, dietitian, and kinesiologist monitor
tervention, do not align with those reported by Shin et al. each participant’s progress at regular intervals during the
[38]. Measuring plasma adiponectin levels may have pro- study, and make adjustments if necessary, was optimal to
vided some insight to help understand this apparent dis- minimize risk of non-compliance. Finally, we were over-
crepancy; however, differences in the lifestyle interven- ly conservative in regard to the SNPs used in our GRS
tion (i.e., caloric deficit versus improvement in diet qual- analyses. Specifically, both SNPs were significantly asso-
ity), length of time of the intervention (3 vs 12 months), ciated with changes in cMetS at both 3 and 12 months in
and population make-up should be acknowledged. Thus, both dominant and additive genetic models. This conser-
further investigations of this particular SNP as a modera- vative approach reinforces the significant findings uncov-
tor of response to lifestyle interventions are necessary. ered with our uwGRS and wGRS models.
In addition to individual SNP associations, GRSs (both In conclusion, the present study demonstrated that
uwGRS and wGRS) provided strong correlation with the specific genetic variants, both alone and when combined
lifestyle intervention outcomes. The significantly greater into a GRS, influence the magnitude of change in cMetS - 11/26/2018 9:26:22 PM

8 Lifestyle Genomics Lowry et al.

Stockholm University Library

DOI: 10.1159/000494331
Downloaded by:
score in response to a lifestyle intervention. These results Disclosure Statement
reinforce the potential value of assessing genetic risk to
Metabolic Syndrome Canada is a not-for-profit charitable or-
better tailor health management and goal-setting during ganization that funded the current study. R.D. was paid for her
an intervention. Moreover, knowledge of gene-lifestyle work on the study by Queen’s University from this grant. R.D. be-
interactions may contribute to the development and/or came an employee of Metabolic Syndrome Canada after the com-
refinement of nutrigenomic advice for healthcare practi- pletion of study enrolment. D.K. received a grant as a participating
tioners and direct-to-consumer genetic companies alike. site for patient enrolment and data collection from Metabolic Syn-
drome Canada. P.B., D.R., D.M.M., and A.T. received grants for
program development from Metabolic Syndrome Canada. K.J. is
on the board of directors for Metabolic Syndrome Canada and will
Acknowledgments be involved in discussions about fundraising for this non-profit
Metabolic Syndrome Canada is a not-for-profit charitable or-
ganization that partially supported this work.

  1 Brown AE, Walker M. Genetics of Insulin Re-  8 Bouchard C, Blair SN, Church TS, Earnest 17 Gaulton KJ, Willer CJ, Li Y, Scott LJ, Con-
sistance and the Metabolic Syndrome. Curr CP, Hagberg JM, Häkkinen K, et al. Adverse neely KN, Jackson AU, et al. Comprehensive
Cardiol Rep. 2016 Aug;18(8):75. metabolic response to regular exercise: is it a association study of type 2 diabetes and re-
  2 O’Neill S, O’Driscoll L. Metabolic syndrome: rare or common occurrence? PLoS One. 2012; lated quantitative traits with 222 candidate
a closer look at the growing epidemic and its 7(5):e37887. genes. Diabetes. 2008 Nov;57(11):3136–44.
associated pathologies. Obes Rev. 2015 Jan;   9 Horne J, Madill J, O’Connor C, Shelley J, Gil- 18 Hou H, Ge S, Zhao L, Wang C, Wang W, Zhao
16(1):1–12. liland J. A Systematic Review of Genetic Test- X, et al. An Updated Systematic Review and
  3 Alberti KG, Eckel RH, Grundy SM, Zimmet ing and Lifestyle Behaviour Change: Are We Meta-analysis of Association Between Adipo-
PZ, Cleeman JI, Donato KA, et al.; Interna- Using High-Quality Genetic Interventions nectin Gene Polymorphisms and Coronary
tional Diabetes Federation Task Force on Ep- and Considering Behaviour Change Theory? Artery Disease. OMICS. 2017 Jun;21(6):340–
idemiology and Prevention; Hational Heart, Lifestyle Genomics. 2018;11(1):49–63. 51.
Lung, and Blood Institute; American Heart 10 Jeejeebhoy K, Dhaliwal R, Heyland DK, 19 Tanaka T, Shen J, Abecasis GR, Kisialiou A,
Association; World Heart Federation; Inter- Leung R, Day AG, Brauer P, et al. Family phy- Ordovas JM, Guralnik JM, et al. Genome-
national Atherosclerosis Society; Internation- sician-led, team-based, lifestyle intervention wide association study of plasma polyunsatu-
al Association for the Study of Obesity. Har- in patients with metabolic syndrome: results rated fatty acids in the InCHIANTI Study.
monizing the metabolic syndrome: a joint in- of a multicentre feasibility project. CMAJ PLoS Genet. 2009 Jan;5(1):e1000338.
terim statement of the International Diabetes Open. 2017 Mar;5(1):E229–36. 20 Hou H, Ma R, Guo H, He J, Hu Y, Mu L, et al.
Federation Task Force on Epidemiology and 11 Klein D, Jeejeebhoy K, Tremblay A, Kallio M, Association between Six CETP Polymor-
Prevention; National Heart, Lung, and Blood Rheaume C, Humphries S, et al. The phisms and Metabolic Syndrome in Uyghur
Institute; American Heart Association; World CHANGE program: exercise intervention in Adults from Xinjiang, China. Int J Environ
Heart Federation; International Atheroscle- primary care. Can Fam Physician. 2017 Jul; Res Public Health. 2017 Jun;14(6):14.
rosis Society; and International Association 63(7):546–52. 21 Emamian M, Avan A, Pasdar A, Mirhafez SR,
for the Study of Obesity. Circulation. 2009 12 Royall D, Brauer P, Bjorklund L, O’Young O, Sadeghzadeh M, Moghadam MS, et al. The li-
Oct;120(16):1640–5. Tremblay A, Jeejeebhoy K, et al. Development poprotein lipase S447X and cholesteryl ester
  4 Hillier TA, Rousseau A, Lange C, Lépinay P, of a Dietary Management Care Map for Met- transfer protein rs5882 polymorphisms and
Cailleau M, Novak M, et al.; D.E.S.I.R. Co- abolic Syndrome. Can J Diet Pract Res. 2014 their relationship with lipid profile in human
hort. Practical way to assess metabolic syn- Sep;75(3):132–9. serum of obese individuals. Gene. 2015 Mar;
drome using a continuous score obtained 13 Garriguet D. Diet quality in Canada. Health 558(2):195–9.
from principal components analysis. Diabe- Rep. 2009 Sep;20(3):41–52. 22 Paththinige CS, Sirisena ND, Dissanayake V.
tologia. 2006 Jul;49(7):1528–35. 14 Ebbeling CB, Ward A, Puleo EM, Widrick J, Genetic determinants of inherited suscepti-
 5 Pérez-Martínez P, Mikhailidis DP, Athyros Rippe JM. Development of a single-stage sub- bility to hypercholesterolemia - a comprehen-
VG, Bullo M, Couture P, Covas MI, et al. Life- maximal treadmill walking test. Med Sci sive literature review. Lipids Health Dis. 2017
style recommendations for the prevention Sports Exerc. 1991 Aug;23(8):966–73. Jun;16(1):103.
and management of metabolic syndrome: an 15 Xi B, Tang W, Wang Q. Polymorphism near 23 Petersen KF, Dufour S, Hariri A, Nelson-Wil-
international panel recommendation. Nutr the ATP2B1 gene is associated with hyperten- liams C, Foo JN, Zhang XM, et al. Apolipo-
Rev. 2017 May;75(5):307–26. sion risk in East Asians: a meta-analysis in- protein C3 gene variants in nonalcoholic fatty
  6 Fenwick PH, Jeejeebhoy K, Dhaliwal R, Royall volving 15 909 cases and 18 529 controls. liver disease. N Engl J Med. 2010 Mar;362(12):
D, Brauer P, Tremblay A, et al. Lifestyle ge- Blood Press. 2012 Apr;21(2):134–8. 1082–9.
nomics and the metabolic syndrome: A re- 16 Niu W, Qi Y, Gao P, Zhu D. Review: associa- 24 de Luis DA, Izaola O, Primo D, Aller R. Im-
view of genetic variants that influence re- tion between angiotensin converting enzyme plication of the rs670 variant of APOA1 gene
sponse to diet and exercise interventions. Crit G2350A polymorphism and hypertension with lipid profile, serum adipokine levels and
Rev Food Sci Nutr. 2018 Feb;1–12. risk: a meta-analysis. J Renin Angiotensin Al- components of metabolic syndrome in adult
  7 Wagh A, Stone NJ. Treatment of metabolic dosterone Syst. 2011 Mar;12(1):8–14. obese subjects. Clin Nutr. 2017 Dec;
syndrome. Expert Rev Cardiovasc Ther. 2004 S0261-5614(17)31427-9.
Mar;2(2):213–28. - 11/26/2018 9:26:22 PM

Lifestyle-Gene Interactions and MetS Lifestyle Genomics 9

Stockholm University Library

DOI: 10.1159/000494331
Downloaded by:
25 Guardiola M, Ribalta J, Gómez-Coronado D, 30 Lin E, Kuo PH, Liu YL, Yang AC, Tsai SJ. De- 35 González-Sánchez JL, Zabena CA, Martínez-
Lasunción MA, de Oya M, Garcés C. The apo- tection of susceptibility loci on APOA5 and Larrad MT, Fernández-Pérez C, Pérez-Barba
lipoprotein A5 (APOA5) gene predisposes COLEC12 associated with metabolic syn- M, Laakso M, et al. An SNP in the adiponectin
Caucasian children to elevated triglycerides drome using a genome-wide association gene is associated with decreased serum adi-
and vitamin E (Four Provinces Study). Ath- study in a Taiwanese population. Oncotarget. ponectin levels and risk for impaired glucose
erosclerosis. 2010 Oct;212(2):543–7. 2017 Sep;8(55):93349–59. tolerance. Obes Res. 2005 May;13(5):807–12.
26 Xu C, Bai R, Zhang D, Li Z, Zhu H, Lai M, et 31 Kim M, Chae JS, Kim M, Lee SH, Lee JH. Ef- 36 Li P, Jiang R, Li L, Liu C, Yang F, Qiu Y. Cor-
al. Effects of APOA5 -1131T[{GT}]C fects of a 3-year dietary intervention on age- relation of serum adiponectin and adiponec-
(rs662799) on fasting plasma lipids and risk of related changes in triglyceride and apolipo- tin gene polymorphism with metabolic syn-
metabolic syndrome: evidence from a case- protein A-V levels in patients with impaired drome in Chinese adolescents. Eur J Clin
control study in China and a meta-analysis. fasting glucose or new-onset type 2 diabetes Nutr. 2015 Jan;69(1):62–7.
PLoS One. 2013;8(2):e56216. as a function of the APOA5 -1131 T [{GT}] C 37 Gao M, Ding D, Huang J, Qu Y, Wang Y,
27 Ong KL, Jiang CQ, Liu B, Jin YL, Tso AW, polymorphism. Nutr J. 2014 Apr;13(1):40. Huang Q. Association of genetic variants in
Tam S, et al. Association of a genetic variant 32 Jang Y, Chae JS, Kim OY, Park HJ, Kim JY, the adiponectin gene with metabolic syn-
in the apolipoprotein A5 gene with the meta- Paik JK, et al. APOA5-1131T[{GT}]C geno- drome: a case-control study and a systematic
bolic syndrome in Chinese. Clin Endocrinol type effects on apolipoprotein A5 and triglyc- meta-analysis in the Chinese population.
(Oxf). 2011 Feb;74(2):206–13. eride levels in response to dietary intervention PLoS One. 2013 Apr;8(4):e58412.
28 Zhang Z, Peng B, Gong RR, Gao LB, Du J, and regular exercise (DIRE) in hypertriglyc- 38 Shin MJ, Jang Y, Koh SJ, Chae JS, Kim OY, Lee
Fang DZ, et al. Apolipoprotein A5 polymor- eridemic subjects. Atherosclerosis. 2010 Aug; JE, et al. The association of SNP276G[{GT}]T
phisms and risk of coronary artery disease: a 211(2):512–9. at adiponectin gene with circulating adipo-
meta-analysis. Biosci Trends. 2011 Aug; 5(4): 33 Yamasaki M, Mutombo PB, Iwamoto M, nectin and insulin resistance in response to
165–72. Nogi A, Hashimoto M, Nabika T, et al. The mild weight loss. Int J Obes. 2006 Dec;30(12):
29 Sarwar N, Sandhu MS, Ricketts SL, Butter- interaction of Apolipoprotein A5 gene pro- 1702–8.
worth AS, Di Angelantonio E, Boekholdt SM, moter region T-1131C polymorphism 39 San-Cristobal R, Navas-Carretero S, Living-
et al.; Triglyceride Coronary Disease Genetics (rs12286037) and lifestyle modification on stone KM, Celis-Morales C, Macready AL,
Consortium and Emerging Risk Factors Col- plasma triglyceride levels in Japanese. Nutr Fallaize R, et al. Mediterranean Diet Adher-
laboration. Triglyceride-mediated pathways Res Pract. 2015 Aug;9(4):379–84. ence and Genetic Background Roles within a
and coronary disease: collaborative analysis 34 Menzaghi C, Trischitta V, Doria A. Genetic Web-Based Nutritional Intervention: The
of 101 studies. Lancet. 2010 May; 375(9726): influences of adiponectin on insulin resis- Food4Me Study. Nutrients. 2017 Oct;9(10):9.
1634–9. tance, type 2 diabetes, and cardiovascular dis-
ease. Diabetes. 2007 May;56(5):1198–209. - 11/26/2018 9:26:22 PM

10 Lifestyle Genomics Lowry et al.

Stockholm University Library

DOI: 10.1159/000494331
Downloaded by: