Association Between Body Fat Response to
Exercise Training and Multilocus ADR
Genotypes
Dana A. Phares,* Amy A. Halverstadt,* Alan R. Shuldiner,† Robert E. Ferrell,‡ Larry W. Douglass,§
Alice S. Ryan,¶ Andrew P. Goldberg,¶ and James M. Hagberg*
Abstract
PHARES, DANA A., AMY A. HALVERSTADT, ALAN
R. SHULDINER, ROBERT E. FERRELL, LARRY W.
DOUGLASS, ALICE S. RYAN, ANDREW P.
GOLDBERG, AND JAMES M. HAGBERG. Association
between body fat response to exercise training and
multilocus adrenergic receptor genotypes. Obes Res. 2004;
12:807– 815.
Objective: To examine the contribution of adrenergic recep-
tor (ADR) gene polymorphisms and their gene-gene inter-
actions to the variability of exercise training-induced body
fat response.
Research Methods and Procedures: This was an interven-
tion study that used a volunteer sample of 70 healthy,
sedentary men (n ⫽ 29) and postmenopausal women (n ⫽
41) 50 to 75 years of age, with a BMI ⱕ37 kg/m2, from the
Washington, DC, metropolitan area. Participants completed
6 weeks of dietary stabilization (American Heart Associa-
tion diet) before 24 weeks of supervised aerobic exercise
training. Diet was maintained throughout the intervention.
Change in percent total body fat, percent trunk fat, and fat
mass by DXA in ADR genotype groups (Glu12/Glu9 ␣2b-
ADR, Trp64Arg ␤3-ADR, and Gln27Glu ␤2-ADR) at base-
Received for review May 27, 2003.
Accepted in final form March 1, 2004.
The costs of publication of this article were defrayed, in part, by the payment of page
charges. This article must, therefore, be hereby marked “advertisement” in accordance with
18 U.S.C. Section 1734 solely to indicate this fact.
*Department of Kinesiology, University of Maryland, College Park, Maryland; †Division of
Endocrinology, Diabetes and Nutrition, University of Maryland School of Medicine, Bal-
timore, Maryland; ‡Department of Human Genetics, University of Pittsburgh, Graduate
School of Public Health, Pittsburgh, Pennsylvania; §Biometrics Program, Department of
Animal and Avian Sciences, University of Maryland, College Park, Maryland; and ¶Divi-
sion of Gerontology, Baltimore Geriatric, Research, Education and Clinical Center, Univer-
sity of Maryland School of Medicine, Baltimore, Maryland.
Address correspondence to Dana A. Phares, Department of Kinesiology, University of
Maryland, College Park, MD 20742-2611.
E-mail: db240@umail.umd.edu
Copyright © 2004 NAASO
line and after 24 weeks of aerobic exercise training was
measured.
Results: In multivariate analysis (covariates: age, gender,
and baseline value of phenotype), best fit models for percent
total body and trunk fat response to exercise training re-
tained main effects of all three ADR gene loci and the
effects of each gene-gene interaction (p ⫽ 0.009 and 0.003,
respectively). Similarly, there was a trend for the fat mass
response model (p ⫽ 0.03). The combined genetic factors
explained 17.5% of the overall model variability for percent
total body fat, 22% for percent trunk fat, and 10% for fat
mass.
Discussion: The body fat response to exercise training in
older adults is associated with the combined effects of the
Glu12/Glu9 ␣2b-, Trp64Arg ␤3-, and Gln27Glu ␤2-ADR
gene variants and their gene-gene interactions.
Key words: body fat phenotype, association, aerobic
exercise, body fat regulation
Introduction
One of the long-term goals of genetics research has been
to translate the discovery of susceptibility genes into effec-
tive, personalized medical care. However, any single gene
variant may not impact health care substantially because the
predictive ability, and thus, the clinical relevance, may be
minimal for complex traits (1). Indeed, the complexity of a
phenotype such as body fat, which is undoubtedly affected
by numerous genes and environmental factors, has con-
founded research efforts to arrive at unequivocal associa-
tions, and perhaps, fuels the skepticism regarding the even-
tual ability of genetic diagnostics to optimize clinical
treatment for common multifactorial conditions. It is well
established that there is great interindividual variation in
responses to exercise, with some individuals experiencing
large beneficial changes in body fat and others experiencing
OBESITY RESEARCH Vol. 12 No. 5 May 2004 807
Multilocus ADR Genotypes and Body Fat, Phares et al.
little or no change. These large interindividual differences
are likely to be determined, in part, by genetic influences.
From this perspective, given that multiple genetic loci are
likely to influence body fat variability (2), one current
challenge for genetics is to account for the collective effects
of multiple gene and gene-gene interactions when assessing
the potential for a treatment intervention such as exercise
training to be successful for an individual.
Adrenergic receptors (ADRs),1 known to be expressed in
adipose tissue and to regulate lipolytic responses, include
the ␤3-ADR and ␤2-ADR (both stimulatory) and the ␣2b-
ADR (inhibitory). Each of these ADRs have genetic variants
(Gln27Glu ␤2-, Trp64Arg ␤3-, and Glu12/Glu9 ␣2b-ADR).
Several studies have reported independent association of
these ADR variants with body fat distribution, obesity,
and/or altered lipolytic function (3). However, a number of
studies have reported no ADR gene associations with body
fat phenotypes (3). Several factors may explain the incon-
sistencies among results, including the likely possibility that
there are no major genes for typical overweight and obesity,
but rather, there are a number of genes each having only a
minor effect when considered alone (3,4).
Exercise training promotes energy expenditure by several
mechanisms, one of which is by enhancing catecholamin-
ergic ␤-ADR stimulation of lipolytic activity and concom-
itantly blunting ␣2-ADR antilipolytic activity, therefore
stimulating body fat mobilization. Nonetheless, only one
study has assessed the independent impact of a ␤-ADR
polymorphism on body composition changes with exercise
training, and there was no relationship (5). Two other stud-
ies reported significant associations of single gene variants
[UCP3(GAIVS6) and LPL (S447X)] with body composi-
tion responses to exercise training (6,7). However, the in-
dependent effect of these single loci, even when the rela-
tionship was significant, accounted for ⬍5% of the
interindividual variation in the body fat response to exercise
(6,7). More recently, interaction between Arg16Gly ␤2- and
Trp64Arg ␤3-ADR genotypes have been reported to be
associated with changes in total and regional adiposity in
response to endurance training in black subjects (8).
Based on the polygenic nature of body fat regulation and the
probability that the multiple genetic determinants of body fat
are interactive, we hypothesized that candidate ADR gene
variants and their interactions would be associated with the
response of body fat phenotypes to exercise training.
Research Methods and Procedures
Participants
Men and women recruited from the Washington, DC
metropolitan area underwent a telephone interview to de-
1
Nonstandard abbreviations: ADR, adrenergic receptor; CVD, cardiovascular disease; HRT,
hormone replacement therapy; VO2max, volume of oxygen at maximal exercise effort;
HERITAGE, Health, Risk Factors, Exercise Training and Genetics.
808 OBESITY RESEARCH Vol. 12 No. 5 May 2004
termine initial eligibility. Subjects were 50 to 75 years of
age, nonsmoking, nondiabetic, and sedentary (⬍20 minutes
of continuous aerobic exercise ⱕ2 times/wk) and had no
prior history of cardiovascular disease (CVD), had a BMI
ⱕ37 kg/m2, and, if female, were ⬎2 years past menopause.
Women maintained their hormone replacement therapy
(HRT) status (either on or not on HRT) for the duration of
the study. The use of any medication that affected lipid or
glucose metabolism was an exclusion. The University of
Maryland College Park and University of Maryland at Bal-
timore Institutional Review Boards approved the study, and
all subjects provided their written consent.
Screening
During the first screening visit, health history and phys-
ical activity questionnaires were reviewed. Height and
weight were measured. A blood sample was drawn from an
antecubital vein to isolate leukocyte DNA for genotyping
analyses. Fasting blood samplings and 2-hour oral glucose
tolerance tests were conducted. Only subjects who showed
no diabetes (fasting glucose ⬍ 126 mg/dL; 2-hour postpran-
dial ⬍ 200 mg/dL), were apparently healthy based on blood
chemistries, had a fasting total cholesterol ⱖ200 mg/dL,
and/or a high-density lipoprotein-cholesterol ⬍35 mg/dL
qualified for the study. Subjects also underwent a physical
examination and completed a treadmill exercise test with
blood pressure and electrocardiographic monitoring to in-
sure they were free from detectable CVD, pulmonary dis-
ease, or other chronic conditions that would prohibit partic-
ipation in exercise training.
Study Design
Qualified subjects underwent a 6-week dietary stabiliza-
tion phase. After completing dietary stabilization and before
the start of exercise training, each subject completed body
composition and volume of oxygen at maximum effort of
exercise (VO2max) baseline assessments. Subjects repeated
the same tests after the 24-week exercise training interven-
tion.
Dietary Stabilization and Monitoring
Subjects met with the study dietitian for 1 hour, 2 d/wk,
for 6 weeks to receive instruction on the American Heart
Association Step 1 diet (⬍30% of total caloric intake from
fat, 55% from carbohydrates, 15% from protein, and cho-
lesterol intake ⬍300 mg/d) (9). Subjects completed 7-day
food records at weeks 1, 8, 16, and 24 of the exercise inter-
vention and met with the study dietitian once every 2 weeks
throughout the course of the exercise intervention to ensure
maintenance of the diet. Subjects who lost more weight than
could be expected from the exercise energy expenditure were
counseled against caloric restriction. Food records were ana-
lyzed using Computrition 3.1 (Computrition, Inc., Chatsworth,
CA).
Multilocus ADR Genotypes and Body Fat, Phares et al.
Baseline Testing
Body Composition. DXA was used to assess total and
regional body composition (DPX-L or DPX-IQ; Lunar
Corp., Madison, WI) while subjects maintained a relaxed
supine position.
Aerobic Capacity. VO2max was measured during a modi-
fied treadmill protocol (10 –12). Expired gases were contin-
uously analyzed by a computer-based metabolic analysis
system. True VO2max was considered to be achieved based
on standard physiological criteria (11).
Genotyping. The three candidate ADR gene polymor-
phisms (Glu12/Glu9 ␣2b-ADR, Trp64Arg ␤3-ADR, and
Gln27Glu ␤2-ADR) were analyzed by polymerase chain
reaction-restriction fragment length polymorphism methods
(13,14). Forward and reverse oligonucleotide DNA primers
were used to amplify leukocyte DNA, and the appropriate
restriction enzymes digested the amplification products into
characteristic fragments using standard procedures. All am-
plified fragments were analyzed by electrophoresis through
a 4% agarose gel (Metaphor agarose; FMC Bioproducts,
Rockland, ME) and visualized by ultraviolet transillumina-
tion after staining with ethidium bromide. Blind duplicate
genotyping was employed to maintain genotyping quality
control. Genotypes were defined as noncarrier or carrier of
the Glu9 ␣2b-ADR, the Arg64 ␤3-ADR, and the Glu27
␤2-ADR alleles.
Exercise Training
The training intervention consisted of endurance exercise
3 d/wk for 24 weeks, under the supervision of study per-
sonnel (11). Exercise intensity and duration progressed
from 50% to 70% VO2max and from 20 to 40 minutes during
the first 10 weeks. Subjects trained for 40 minutes at 70%
VO2max for the remaining 14 weeks and added a 45- to
60-minute walk at home on the weekend. Only subjects who
completed ⬎75% of all training sessions were included in
the final data analyses.
Statistics
Data are presented as mean ⫾ SE. Baseline subject char-
acteristics and body fat variables were compared between
genders and also between ADR genotype groups at the three
loci using Student’s t tests. Paired Student’s t tests were
used to assess VO2max, body weight, BMI, percent total body
fat, percent trunk fat, and fat mass (in kilograms) responses
to exercise training for the entire group. A ␹2 analysis was
used to determine whether genotype distribution deviated
from expected Hardy-Weinberg equilibrium and to test for
gender or HRT use differences among genotype groups,
including gene-gene multilocus genotype groups.
Changes in body fat variables with training (change ⫽
final ⫺ baseline) were examined using the SAS multivariate
general linear model procedure, with gender, age, and base-
line value of the phenotype as covariates. In general, the
modeling approach was similar to the backward elimination
procedure and included reducing the saturated model (con-
taining all variables) to a model consisting of a single
independent variable. More specifically, the initial model
included the main effect for each of the ADR gene variants
(Glu12/Glu9 ␣2b-ADR, Trp64Arg ␤3-ADR, and Gln27Glu
␤2-ADR) and their gene-gene interactions (Glu12/Glu9 ␣2b-
ADR ⫻ Trp64Arg ␤3-ADR, Glu12/Glu9 ␣2b-ADR ⫻
Gln27Glu ␤2-ADR, Trp64Arg ␤3-ADR ⫻ Gln27Glu ␤2-
ADR). With the least significant F-statistic being the deter-
mining criterion, the least important explanatory sources of
variation were eliminated from the model in an iterative
fashion. An ADR gene variant main effect (Glu12/Glu9 ␣2b-
ADR, Trp64Arg ␤3-ADR, or Gln27Glu ␤2-ADR) was only
eligible for removal from the model when it was no longer
present in any of the remaining gene-gene interaction terms.
On examination of all elimination steps, the best fit model
was determined as the model where mean square error was
minimized and adjusted R2 was maximized. This modeling
approach helps eliminate the correlation among explanatory
variables in a way that is most likely to retain the most
significant sources of variation. Because sample sizes for
each genotype are a function of different allelic frequencies,
the experiment was unequally replicated (unbalanced). Un-
balanced data causes statistical nonindependence among the
explanatory variables. The backward elimination procedure
reduced the consequences of this lack of independence.
Adjusted R2 values are used to report explained variability.
Statistical procedures were analyzed on SPSS 10.0 (SPSS
Inc., Chicago, IL) and SAS 8e (SAS Institute, Cary, NC)
software. The threshold of significance was p ⱕ 0.01.
Results
Seventy subjects (29 men and 41 women), who were
primarily white (83%), completed exercise training. When
the data are analyzed among only whites (n ⫽ 48), the
results were comparable to those reported for the entire
group. Glu12/Glu9 ␣2b-ADR, Trp64Arg ␤3-ADR, and
Gln27Glu ␤2-ADR allele frequencies were comparable with
those reported previously (13–20), did not differ between
gender groups, and were in Hardy-Weinberg equilibrium.
Allele and genotype frequencies are shown in Table 1.
There was no difference in the frequency of men and
women between any noncarrier and carrier ADR genotype
group. Women were evenly split between users and nonus-
ers of HRT (49% and 51%, respectively), and there was no
difference in the frequency of users and nonusers of HRT
between any noncarrier and carrier ADR genotype group.
Furthermore, there were no significant differences among
multilocus genotype groups for frequencies of men/women
or for nonusers/users of HRT (␹2 ⬎ 0.19 for all).
There were no differences in age or BMI between men
and women at baseline; however, men had higher body
weight and VO2max, whereas women had higher total body
OBESITY RESEARCH Vol. 12 No. 5 May 2004 809
Multilocus ADR Genotypes and Body Fat, Phares et al.
Table 1. ADR gene polymorphism allele frequencies and genotype distributions
ADR gene
polymorphism Allele frequency
Glu12/Glu9 ␣2b-ADR Glu12 Glu9
0.71 0.29
Trp64Arg ␤3-ADR Trp64 Arg64
0.89 0.11
Gln27Glu ␤2-ADR Gln27 Glu27
0.61 0.39
* Values in parentheses are the sample sizes.
fat (Table 2). Body weight, BMI, VO2max, percent total body
fat, percent trunk fat, and fat mass at baseline did not differ
among genotype groups at any of the three ADR gene loci or
their gene-gene interaction groups (data not shown).
VO2max (milliliters per kilogram per minute) increased by
16%, whereas body weight decreased slightly but signifi-
cantly in response to exercise training in the entire group
(Table 2). There was no significant difference in VO2max or
body weight response to training among genders or among
users or nonusers of HRT. VO2max and body weight re-
sponses were similar among genotype groups (p ⬎ 0.01 for
all tests) when each of the ADR polymorphisms was exam-
ined individually. Exercise training decreased percent total
body fat, percent trunk fat, and total fat mass among all
subjects (Table 2). Changes in percent total body fat, per-
cent trunk fat, and fat mass were significant within (p ⬍
0.01) and similar between (p ⬎ 0.01) genders; thus, results
were combined for all subsequent analyses using gender as
a covariate.
Total Body Fat
Based on the multivariate linear model procedure, the
best fit model for change in percent total body fat with
exercise training included as integral explanatory variables
all three ADR gene loci main effects and each of their
gene-gene interaction effects (Table 3). The probabilities for
the ␣2b ADR ⫻ ␤3 ADR and the ␤3 ADR ⫻ ␤2 ADR
interactions were statistically significant, although the ADR
gene variant main effects were not. Overall, explained vari-
ability for the response of percent total body fat to exercise
training accounted for by the final model was 19%, with the
genetic factors accounting for 18% of the overall variability.
The ADR gene variants and their gene-gene interactions
accounted for significant portions of the interindividual
variance in body fat response to exercise training when all
were included in the analysis model. For example, whereas
both Trp64Arg ␤3-ADR genotype groups showed a signif-
icant percent total body fat loss with exercise training (p ⬍
810 OBESITY RESEARCH Vol. 12 No. 5 May 2004
Genotype frequency*
Glu12/Glu12 Glu12/Glu9 Glu9/Glu9
0.53 (37) 0.36 (25) 0.11 (8)
Trp/Trp Trp/Arg Arg/Arg
0.81 (57) 0.16 (11) 0.03 (2)
Gln/Gln Gln/Glu Glu/Glu
0.37 (26) 0.47 (33) 0.16 (11)
0.0001 for both groups), Arg64 ␤3-ADR carriers had a
2-fold greater loss of percent total body fat compared with
Arg64 ␤3-ADR noncarriers, which approached significance
(⫺2.7 ⫾ 0.5% vs. ⫺1.3 ⫾ 0.3%; p ⫽ 0.027). Similarly,
Glu27 ␤2-ADR carriers had a tendency for a greater loss of
percent total body fat compared with noncarriers (⫺2.7 ⫾
0.4% vs. ⫺1.3 ⫾ 0.4%, p ⫽ 0.015). Furthermore, percent
total body fat response was significantly associated with
␣2b-ADR ⫹ ␤3-ADR and ␤3-ADR ⫹ ␤2-ADR multilocus
genotypes. Glu12/Glu12 homozygotes who simultaneously
carried the Arg64 ␤3-ADR allele lost significantly more
total body fat (⫺4.0 ⫾ 0.9%) compared with all other
␣2b-ADR ⫹ ␤3-ADR multilocus genotype groups (p ⫽
0.009; Figure 1A). The Arg64 ␤3-ADR carrier ⫹ Glu27
␤2-ADR carrier genotype group showed a percent total body
fat reduction (⫺4.2 ⫾ 0.8%) that was at least three times
greater than all other ␤3-ADR ⫹ ␤2-ADR multilocus geno-
type groups (all p ⬍ 0.005; Figure 1C).
Trunk Fat
All three ADR gene loci main effects and gene-gene
interaction effects remained in the best fit model as integral
explanatory variables for the change in percent trunk fat
with training (Table 3). The genetic factors in the best fit
model for response of percent trunk fat to exercise training
accounted for 22% of the interindividual variability.
Based on this multivariate linear analysis model, the
Arg64 ␤3-ADR carriers reduced percent trunk fat twice as
much as Arg64 ␤3-ADR noncarriers, with the responses
among groups approaching significance (⫺3.1 ⫾ 0.6% vs.
⫺1.6 ⫾ 0.3%, respectively; p ⫽ 0.03). Similarly, Glu27
␤2-ADR carriers showed a trend toward a significantly
greater reduction than Glu27 ␤2-ADR noncarriers (⫺3.2 ⫾
0.5% vs. ⫺1.5 ⫾ 0.5%; p ⫽ 0.02). Furthermore, Glu9
␣2b-ADR noncarriers ⫹ Arg64 ␤3-ADR carriers had signif-
icantly greater percent trunk fat reduction (⫺4.8 ⫾ 1.1%)
than the other ␣2b-ADR ⫹ ␤3-ADR multilocus genotype
groups, whose changes ranged from ⫺2.7% to ⫺0.5% (p ⬍
Multilocus ADR Genotypes and Body Fat, Phares et al.

Table 2. Initial subject characteristics and changes with exercise training

Gender group
Characteristics Total (n ⴝ 70) Men (n ⴝ 29) Women (n ⴝ 41) p*
Age (years) 57.8 ⫾ 0.7 57.7 ⫾ 1.1 57.9 ⫾ 0.9 0.886
Height (cm) 168.8 ⫾ 1.2 177.3 ⫾ 1.1 162.9 ⫾ 1.1 ⬍0.001
Weight (kg)
Initial 80.1 ⫾ 1.8 87.4 ⫾ 2.6 74.9 ⫾ 2.1 ⬍0.001
Change with training ⫺1.2 ⫾ 0.3† ⫺1.7 ⫾ 0.3† ⫺0.8 ⫾ 0.4 0.065
BMI (kg/m2)
Initial 28.0 ⫾ 0.5 27.8 ⫾ 0.7 28.2 ⫾ 0.7 0.715
Change with training ⫺0.4 ⫾ 0.1† ⫺0.5 ⫾ 0.1† ⫺0.3 ⫾ 0.1† 0.190
VO2max (L/min)
Initial 2.0 ⫾ 0.1 2.5 ⫾ 0.1 1.7 ⫾ 0.04 ⬍0.001
Change with training 0.3 ⫾ 0.03† 0.4 ⫾ 0.1† 0.2 ⫾ 0.02† 0.001
VO2max (mL/kg/min)
Initial 25.3 ⫾ 0.6 28.7 ⫾ 0.9 22.9 ⫾ 0.5 ⬍0.001
Change with training 4.1 ⫾ 0.4† 5.2 ⫾ 0.7† 3.4 ⫾ 0.4† 0.018
Total body fat (%)
Initial 35.3 ⫾ 1.1 27.4 ⫾ 1.1 40.8 ⫾ 1.1 ⬍0.001
Change with training ⫺1.3 ⫾ 0.2† ⫺1.4 ⫾ 0.4† ⫺1.2 ⫾ 0.3† 0.642
Trunk fat (%)
Initial 36.2 ⫾ 1.0 30.6 ⫾ 1.2 40.2 ⫾ 1.0 ⬍0.001
Change with training ⫺1.4 ⫾ 0.3† ⫺1.6 ⫾ 0.5† ⫺1.3 ⫾ 0.3† 0.611
Fat mass (kg)
Initial 28.4 ⫾ 1.2 24.6 ⫾ 1.6 31.1 ⫾ 1.5 0.006
Change with training ⫺1.4 ⫾ 0.2† ⫺1.6 ⫾ 0.3† ⫺1.2 ⫾ 0.4† 0.393

Values are expressed as means ⫾ SE.

* Probability for Student’s t test comparing gender groups.
† Indicates significant change within group after training p ⬍ 0.01.

0.01 for all). The Glu9 ␣2b-ADR noncarrier ⫹ Glu27 ␤2-
ADR carrier group showed a percent trunk fat change with
training that was two to four times greater than any other
␣2b-ADR ⫹ ␤3-ADR multilocus genotype group; however,
these changes approached, yet were not, significant (all p ⬍
0.04; Figure 1A). Subjects who carried both the Arg64
␤3-ADR and the Glu27 ␤2-ADR alleles showed the greatest
reduction in percent trunk fat compared with all other ␤3-
ADR ⫹ ␤2-ADR multilocus genotype groups (all p ⬍ 0.01;
Figure 1C).
Fat Mass
All three ADR gene variant main effects and their gene-
gene interaction effects remained in the best fit multivariate
linear analysis model for change in fat mass with exercise
training (Table 3). The probability values for ␤3-ADR,
␣2b-ADR ⫻ ␤3-ADR, and ␤3-ADR ⫻ ␤2-ADR in the final
model approached significance for fat mass response to
training (all p ⬍ 0.04). The explained variability of the best
fit model for fat mass response to exercise training was
15%, 10% of which was attributable to the genetic factors.
Arg64 ␤3-ADR carriers demonstrated a 2-fold greater fat
mass loss compared with noncarriers (⫺2.8 ⫾ 0.6 vs.
⫺1.4 ⫾ 0.3 kg); however, the difference was not significant
(p ⫽ 0.037). Similarly, Glu9 ␣2b-ADR noncarriers ⫹ Glu27
␤2-ADR carriers lost 2-fold greater fat mass (⫺3.2 ⫾ 0.8
kg) compared with those not carrying either variant (Glu9
␣2b-ADR noncarriers ⫹ Glu27 ␤2-ADR noncarrier; ⫺1.4 ⫾
0.6 kg, p ⫽ 0.031). The Glu9 ␣2b-ADR noncarrier ⫹ Arg64
␤3-ADR carrier multilocus genotype group reduced fat mass
significantly more than subjects not carrying either variant
(Glu9 ␣2b-ADR noncarriers ⫹ Arg64 ␤3-ADR noncarrier;
⫺3.8 ⫾ 1.0 vs. ⫺0.8 ⫾ 0.3 kg, respectively; p ⫽ 0.007).

OBESITY RESEARCH Vol. 12 No. 5 May 2004 811

Multilocus ADR Genotypes and Body Fat, Phares et al.

Table 3. Probability values for the multivariate asso-

ciations between the selected ADR variants and body
fat phenotype responses to exercise training (N ⫽ 70)
⌬ Total body ⌬ Trunk ⌬ Fat mass
ADR variant fat (%) fat (%) (kg)
␣2b 0.2547 0.3267 0.5627
␤3 0.0265 0.0311 0.0368
␤2 0.0154 0.0154 0.1013
␣2b ⫻ ␤3 0.0015 0.0004 0.0193
␣2b ⫻ ␤2 0.0964 0.0137 0.1004
␤3 ⫻ ␤2 0.0073 0.0134 0.0270
Full model 0.0092 0.0033 0.0258

Full model also includes age, gender, and baseline value of that
body fat phenotype.
␣2b-ADR, Glu12/Glu9 polymorphism; ␤3-ADR, Trp64Arg poly-
morphism; ␤2-ADR, Gln27Glu polymorphism.

Discussion
In common conditions such as overweight and obesity,
the prevailing doctrine has been that we are products of our
individual genome interacting with our particular environ-
ment. Thus, we examined the possibility that the coexist-
ence and interaction of multiple ADR gene variants may
confer specific synergy relative to the response of body fat
to exercise training. We found that the combined effects of
the Glu12/Glu9 ␣2b-ADR, Trp64Arg ␤3-ADR, and
Gln27Glu ␤2-ADR gene polymorphisms and their gene-
gene interactions contribute significantly to explaining in-
terindividual variability in body fat responses to exercise
training. In addition, our results suggest that carrying the
Glu9 ␣2b-ADR, Arg64 ␤3-ADR, and Glu27 ␤2-ADR alleles,
which previously have been shown to be associated with
overweight and obesity, seems to be associated with en-
hanced body fat loss with endurance exercise training in
older men and women.
Multiple lines of evidence raise the possibility that ADRs
expressed in adipocytes and their associated candidate gene
polymorphisms may have independent, as well as interac-
tive, effects on body fat responses to exercise training.
Although the Arg64 ␤3-ADR variant itself has been asso-
ciated with various obesity phenotypes, including reduced
resting energy expenditure (21,22), greater visceral obesity
(20), and reduced ability to mobilize visceral fat (21), the
effects are modest and inconsistent among studies (23).
Similarly, there seems to be an inconsistent association
between the Gln27Glu ␤2ADR polymorphism and body fat
phenotypes (3). These discrepancies may be partially ex-
plained by the interaction of other genetic variants with the
Trp64Arg ␤3-ADR or Gln27Glu ␤2-ADR allele to affect
Figure 1: Change in percent total body and trunk fat after exercise
training among ADR multilocus genotype groups. Values for pair-
wise comparisons are reported as least square means ⫾ SE. *p ⫽
0.009 vs. Glu9-/Arg64-; †p ⫽ 0.007 vs. Glu9-/Arg64-; ‡p ⫽ 0.006
vs. Glu9-/Glu27-; §p ⫽ 0.003 vs. Arg64-/Glu27-, p ⫽ 0.001 vs.
Arg64-/glu27⫹, p ⫽ 0.005 vs. Arg64⫹/Glu27-; 储p ⫽ 0.0005 vs.
Glu9-/Arg64-, p ⫽ 0.01 vs. Glu9⫹/Arg64⫹; ¶p ⫽ 0.0008 vs.
Glu9-/Arg64-; **p ⫽ 0.001 vs. Glu9-/Glu27-; and ††p ⫽ 0.003 vs.
Arg64-/Glu27-, p ⫽ 0.002 vs. Arg64-/Glu27⫹, p ⫽ 0.007 vs.
Arg64⫹/Glu27-. ⫹, carrier; ⫺, noncarrier.
body fat phenotypes. In this study, the decision to examine
a family of ADR gene polymorphisms stems from the idea
that adipocyte ADRs collectively modulate lipolytic activ-
ities and a certain amount of functional redundancy or
compensation would be expected among the different adi-
pocyte ␤-ADRs. Evidence of functional redundancy and
interactive effects among adipocyte ␤-ADRs was reported
recently by Bachman et al. (24), who examined mice lack-
ing the three known ␤-ADRs (␤-less mice) and compared
them with mice lacking ␤1- and ␤3-ADRs (␤1,3-less mice),
and to wild-type mice. The ␤-less mice developed massive
obesity because of an inability to trigger dietary-induced

812 OBESITY RESEARCH Vol. 12 No. 5 May 2004

Multilocus ADR Genotypes and Body Fat, Phares et al.
thermogenesis, whereas ␤1,3-less mice had normal weight,
strongly suggesting functional redundancy among the
␤-ADRs (24,25). Furthermore, it has been well established
that the prevailing action of catecholamines in human fat
cells depends on the balance between lipolytic ␤-ADR and
antilipolytic ␣-ADR functional activities, raising intriguing
questions regarding potential gene-gene interactions in
genes involved in the reciprocal regulation of lipolysis. This
polygenic interplay was recently elucidated by Dionne et al.
(13), who reported an interactive effect of the Glu12/Glu9
␣2b-ADR and the Trp64Arg ␤3-ADR polymorphisms on
obesity-related phenotypes in healthy white women.
Whereas the Glu12/Glu9 ␣2b-ADR polymorphism alone did
not associate with obesity-related phenotypes, subjects car-
rying both the Arg64 ␤3-ADR and Glu9 ␣2b-ADR variants
had 9.3 kg greater fat mass and 4.8% greater percent body
fat compared with subjects carrying only the Arg64 ␤3-
ADR variant. Data from the Quebec Family Study also
suggests that an ␣2-ADR and ␤3-ADR interaction affects
abdominal fat levels and plasma lipids (16). Likewise, our
results lend strong support to the importance of ADR gene-
gene interactions. We found that the ␣2b-ADR ⫻ ␤3-ADR
interaction was the most significant source of variation for
change in total body fat, trunk fat, and fat mass with
exercise training.
Investigation of gene-environment interactions, specifi-
cally the gene-exercise training interaction, a discipline we
have termed kinesiogenomics, has begun to provide com-
pelling evidence that genetic variation among individuals
may influence the response to exercise (26 –28). To date,
only the Health, Risk Factors, Exercise Training and Ge-
netics (HERITAGE) study has reported on the association
of an ADR gene variant with the body composition response
to exercise training (5). The results suggest that there is no
association between the Trp64Arg ␤3-ADR polymorphism
and training-induced BMI, fat mass, percent total body fat,
or abdominal fat area responses in men or women (5).
Likewise, univariate analyses of our data showed no signif-
icant association of the Trp64Arg ␤3-ADR polymorphism
with any of the body composition phenotype responses we
examined. Finding no association in analyses examining a
single ADR polymorphism and exercise training response
underscores that the modest effect of a single variant on the
response of a common phenotype to an intervention may be
unremarkable without simultaneously analyzing the inde-
pendent and interactive effects of other candidate loci
(6,29,30). This is emphasized by the fact that, when multi-
ple gene variants were entered into our model, strong asso-
ciations emerged. Likewise, a recent report from the HER-
ITAGE study suggests that the interaction between the
Arg16Gly ␤2-ADR and Trp64Arg ␤3-ADR gene variants is
associated with total body fat mass and abdominal fat re-
sponse to exercise training in black subjects, despite finding
no association for either polymorphism independently (8).
One of the strengths of our study is that it takes into
account a multilocus profile that may be considered a pu-
tative candidate for conferring risk and investigates the
simultaneous effect of a well-controlled environmental
change, in this case, exercise training. The regression model
used in this study, by having a relatively reduced error term
compared with a single gene-gene interaction model, pro-
duced explained variabilities that are substantially higher
than those previously reported in association studies. Ex-
plained variabilities for the genetic effects from our best fit
models were 10% for fat mass, 18% for percent total body
fat, and 22% for percent trunk fat. Thus, the potential to
identify individuals at risk for obesity or fat distribution-
related disorders who might be ideal targets for an exercise
intervention as a preventive health strategy, and who could
be predicted to respond to an extent that is clinically rele-
vant, is enhanced by the use of these multiple gene models.
On a practical level, it seems that the susceptibility con-
ferred by a gene variant (or combination of gene variants) in
a sedentary state may be markedly attenuated in the phys-
ically active state. In 1999, Meirhaeghe et al. (15) reported
that a significant association between the Gln27Glu ␤2-
ADR gene variant and obesity was only evident in physi-
cally inactive men. This concept is further supported by a
recent report that suggests that the Trp64Arg ␤3-ADR poly-
morphism is a risk factor for obesity among sedentary, but
not more active, people (31). According to our data, subjects
who were simultaneously Arg64 ␤3-ADR carriers and
Glu27 ␤2-ADR carriers showed significantly greater loss of
total percent body fat, trunk fat, and fat mass with training
compared with all other multilocus Trp64Arg ␤3-ADR ⫹
Gln27Glu ␤2-ADR genotypes groups. Thus, the possibility
exists that a gene variant that confers disadvantageous body
fat characteristics in a sedentary population may be associ-
ated with an enhanced response to an exercise training
intervention. The HERITAGE study has shown similar
findings for a different body composition-related gene locus
(uncoupling protein 3), reporting that the variant allele of
the GAIVS6 polymorphism was associated with an en-
hanced body composition response to exercise training (7).
These results converge with the theory that human genes
evolved in an environment where humans were regularly
physically active, and in such an environment, the variants
would not be detrimental; however, the comparatively in-
active environment of modern Westernized culture pro-
vokes the susceptibility of gene variants, creating a permis-
sive effect on genotypic expression of a fat phenotype (32).
Whereas the results of this study suggest that multilocus
ADR genotype is associated with body fat change with
exercise training, the specific physiological influences of
the ADR polymorphisms and their genotype-genotype inter-
actions on a body fat phenotype in response to long-term
exercise training has not been determined. Because it is well
accepted that chronic exercise training enhances cat-
OBESITY RESEARCH Vol. 12 No. 5 May 2004 813
Multilocus ADR Genotypes and Body Fat, Phares et al.
echolaminergic lipolytic activity, and net lipolytic activity
in an individual depends on the balance between ␣-ADR
antilipolytic and ␤-ADR lipolytic activities, it is plausible
that certain combinations of ADR polymorphisms may shift
the balance in antilipolytic vs. lipolytic activity to favor
lipolysis over storage. Future studies are needed to clarify
the functional mechanisms (such as changes in receptor
number, sensitivity, responsiveness, mRNA expression)
that may explain the response of certain multilocus ADR
genotypes to long-term exercise training.
Sample size is a potential limitation of this study; conse-
quently, the data should be interpreted with caution.
Clearly, the number of subjects in each genotype or mul-
tilocus genotype group differs depending on the frequency
of the variant allele. Furthermore, the current analysis is
based on genotype groupings representing carriers vs. non-
carriers of specific alleles. Therefore, future research with
more subjects would allow for allele-based groupings (0, 1,
or 2 variant alleles) and could uncover important allelic
dose responses.
The prospective nature of this study provided an oppor-
tunity to examine the response of a phenotype to a stan-
dardized exercise intervention in terms of a multilocus
genotype. Our findings provide preliminary evidence that
the adrenergic receptors in shared biological pathways,
which contain variants with functional consequences, act
independently and interactively to modulate the response of
body fat phenotypes to exercise training. The results from
this study support the view that a multivariate approach can
improve the predictive ability regarding the variability in
the body fat response to exercise training in comparison
with the results obtained in studies that use univariate gene
variant associations. The finding that specific ADR geno-
types often deemed disadvantageous in terms of body fat in
a sedentary individual may actually be associated with a
better response to exercise training in terms of clinically
relevant body fat changes with exercise training is provoc-
ative and may have important clinical implications.
Acknowledgments
After acceptance of this manuscript, we completed our
intervention on an additional 60 subjects, now giving us a
total of 130 subjects. We found that, generally, the same
results were evident with the larger, 130 subject, group.
This study was supported by National Institutes of Health
Grants AG15389 (to J.M.H.), AG00268 (to A.A.H.), and
DK46204 (to R.E.F.); the Geriatric Research, Education,
and Clinical Center and Medical Research Service of the
Department of Veterans Affairs (to A.P.G.); and the Uni-
versity of Maryland Claude D. Pepper OAIC-2P60-
AG12583-06A1 (to A.P.G.). We thank the participants and
staff of Team Gene of the Gene Exercise Research Study.
We also thank Kathie Brandt and the staff of the Baltimore
Geriatric Research, Education and Clinical Center for care
814 OBESITY RESEARCH Vol. 12 No. 5 May 2004
of the subjects during medical testing and Demian Lewis
(University of Maryland, School of Medicine) for genotype
analysis expertise.
References
1. Finley Austin MJ, Kreiner T. Integrating genomics technol-
ogies in health care: practice and policy challenges and op-
portunities. Physiol Genomics. 2002;8:33– 40.
2. Nelson M, Kardia S, Ferrell RE, Sing C. A combinatorial
partitioning method to identify multilocus genotypic partitions
that predict quantitative trait variation. Genome Res. 2001;11:
458 –70.
3. Perusse L, Chagnon YC, Weisnagel SJ, et al. The human
obesity gene map: the 2000 update. Obes Res. 2001;9:135– 69.
4. Shuldiner AR, Sabra M. Trp64Arg beta3-adrenoceptor:
when does a candidate gene become a disease-susceptibility
gene? Obes Res. 2001;9:806 –9.
5. Garenc C, Perusse L, Rankinen T, et al. The Trp64Arg
polymorphism of the beta3-adrenergic receptor gene is not
associated with training-induced changes in body composi-
tion: the HERITAGE Family Study. Obes Res. 2001;9:337–
41.
6. Garenc C, Perusse L, Bergeron J, et al. Evidence of LPL
gene-exercise interaction for body fat and LPL activity: the
HERITAGE Family Study. J Appl Physiol. 2001;91:1334 – 40.
7. Lanouette CM, Chagnon YC, Rice T, et al. Uncoupling
protein 3 gene is associated with body composition changes
with training in HERITAGE study. J Appl Physiol. 2002;92:
1111– 8.
8. Ukkola O, Rankinen T, Rice T, et al. Interactions among the
B2- and B3-adrenergic receptor genes and total body fat and
abdominal fat level in the HERITAGE Family Study. Int J
Obes Relat Metab Disord. 2002;27:389 –93.
9. American Heart Association. Dietary guidelines for healthy
American adults. A statement for physicians and health pro-
fessionals by the nutrition committee. Circulation. 1988;77:
721A– 4A.
10. Dengel DR, Hagberg JM, Coon PJ, Drinkwater DT, Gold-
berg AP. Effects of weight loss by diet alone or combined
with aerobic exercise on body composition in older obese
men. Metabolism. 1994;43:867–71.
11. Wilund K, Ferrell RE, Phares DA, Goldberg A, Hagberg
JM. Changes in high-density lipoprotein-cholesterol subfrac-
tions with exercise training may be dependent on cholesteryl
ester transfer protein (CETP) genotype. Metabolism. 2002;51:
774 – 8.
12. Seals DR, Hagberg JM, Hurley BF, Ehsani AA, Holloszy
JO. Endurance training in older men and women. J Appl
Physiol. 1984;57:1024 –9.
13. Dionne IJ, Turner A, Tchernof A, et al. Identification of an
interactive effect of ␤3- and ␣2b-adrenoceptor gene polymor-
phisms on fat mass in Caucasian women. Diabetes. 2001;50:
91–5.
14. Large V, Hellstrom L, Reynisdottir S, et al. Human beta-2
adrenoceptor gene polymorphisms are highly frequent in obe-
sity and associate with altered adipocyte beta-2 adrenoceptor
function. J Clin Invest. 1997;100:3005–13.
Multilocus ADR Genotypes and Body Fat, Phares et al.
15. Meirhaeghe A, Helbecque N, Cottel D, Amouyel P. ␤2-
adreneoceptor gene polymorphism, body weight, and physical
activity. Lancet. 1999;353:896.
16. Ukkola O, Rankinen T, Weisnagel SJ, et al. Interactions
among the alpha 2-, beta 2-, and beta 3-adrenergic receptor
genes and obesity-related phenotypes in the Quebec Family
Study. Metabolism. 2000;49:1063–70.
17. Baldwin CT, Schwartz F, Baima J, et al. Identification of a
polymorphic glutamic acid stretch in the alpha2B-adrenergic
receptor and lack of linkage with essential hypertension. Am J
Hypertens. 1999;12:853–7.
18. Heinonen P, Koulu M, Pesonen U, et al. Identification of a
three-amino acid deletion in the alpha2B-adrenergic receptor
that is associated with reduced basal metabolic rate in obese
subjects. J Clin Endocrin Metab. 1999;84:2429 –33.
19. Small KM, Brown KM, Forbes SL, Liggett SB. Polymor-
phic deletion of three intracellular acidic residues of the alpha
2b-adrenergic receptor decreases G protein-coupled receptor
kinase-mediated phosphorylation and desensitization. J Biol
Chem. 2000;276:4917–22.
20. Widen E, Lehto M, Kanninen T, Walston J, Shuldiner AR,
Groop LC. Association of a polymorphism in the beta 3-ad-
renergic-receptor gene with features of the insulin resistance
syndrome in Finns. N Engl J Med. 1995;333:348 –51.
21. Tchernof A, Starling RD, Turner A, et al. Impaired capacity
to lose visceral adipose tissue during weight reduction in
obese postmenopausal women with the Trp64Arg ␤3- adre-
noceptor gene variant. Diabetes. 2000;49:1709 –13.
22. Tchernof A, Starling RD, Walston JD, et al. Obesity-related
phenotypes and the ␤3-adrenoceptor gene variant in post-
menopausal women. Diabetes. 1999;48:1425– 8.
23. Allison DB, Heo M, Faith MS, Pietrobelli A. Meta-analysis
of the association of the Trp64Arg polymorphism in the beta3
adrenergic receptor with body mass index. Int J Obes Relat
Metab Disord. 1998;22:559 – 66.
24. Bachman ES, Dhillon H, Zhang C, et al. Beta-AR signaling
required for diet-induced thermogenesis and obesity resis-
tance. Science. 2002;297:843–5.
25. Dulloo AG. Sympathetic defense against obesity. Science.
2002;297:780 –1.
26. Perusse L, Bouchard C. Genotype-environment interaction
in human obesity. Nutr Rev. 1999;57:S31–7.
27. Bouchard C, Tremblay A. Genetic influences on the re-
sponse of body fat and fat distribution to positive and negative
energy balances in human identical twins. J Nutr. 1997;127:
943S–7S.
28. Bray MS. Genomics, genes, and environmental interaction:
the role of exercise. J Appl Physiol. 2000;88:788 –92.
29. Ott J, Hoh J. Statistical multilocus methods for disequilib-
rium analysis in complex traits. Hum Mutat. 2001;17:285– 8.
30. Williams S, Addy JH, Phillips J III, et al. Combinations of
variations in multiple genes are associated with hypertension.
Hypertension. 2000;36:2– 6.
31. Marti del Moral A, Corbalan MS, Martinez-Gonzalez MA,
Martinez JA. TRP64ARG polymorphism of the beta 3-ad-
renergic receptor gene and obesity risk: effect modification by
a sedentary lifestyle. Diab Obes Metab. 2002;4:428.
32. Booth FW, Chakravarthy MV, Spagenburg EE. Exercise
and gene expression: physiological regulation of the human
genome through physical activity. J Physiol. 2003;543.2:399 –
411.
OBESITY RESEARCH Vol. 12 No. 5 May 2004 815