Caffeine intake and CYP1A2 variants associated with high caffeine intake protect non-smokers from hypertension

Abstract

The 15q24.1 locus, including CYP1A2, is associated with blood pressure (BP). The CYP1A2 rs762551 C allele is associated with lower CYP1A2 enzyme activity. CYP1A2 metabolizes caffeine and is induced by smoking. The association of caffeine consumption with hypertension remains controversial. We explored the effects of CYP1A2 variants and CYP1A2 enzyme activity on BP, focusing on caffeine as the potential mediator of CYP1A2 effects. Four observational (n = 16 719) and one quasi-experimental studies (n = 106) including European adults were conducted. Outcome measures were BP, caffeine intake, CYP1A2 activity and polymorphisms rs762551, rs1133323 and rs1378942. CYP1A2 variants were associated with hypertension in non-smokers, but not in smokers (CYP1A2-smoking interaction P = 0.01). Odds ratios (95% CIs) for hypertension for rs762551 CC, CA and AA genotypes were 1 (reference), 0.78 (0.59-1.02) and 0.66 (0.50-0.86), respectively, P = 0.004. Results were similar for the other variants. Higher CYP1A2 activity was linearly associated with lower BP after quitting smoking (P = 0.049 and P = 0.02 for systolic and diastolic BP, respectively), but not while smoking. In non-smokers, the CYP1A2 variants were associated with higher reported caffeine intake, which in turn was associated with lower odds of hypertension and lower BP (P = 0.01). In Mendelian randomization analyses using rs1133323 as instrument, each cup of caffeinated beverage was negatively associated with systolic BP [-9.57 (-16.22, -2.91) mmHg]. The associations of CYP1A2 variants with BP were modified by reported caffeine intake. These observational and quasi-experimental results strongly support a causal role of CYP1A2 in BP control via caffeine intake.

Full text

Caffeine intake and CYP1A2 variants associated
with high caffeine intake protect non-smokers
from hypertension
Idris Guessous1,2, Maria Dobrinas4, Zolta
´n Kutalik5,11, Menno Pruijm6, Georg Ehret1,3,
Marc Maillard6, Sven Bergmann5, Jacques S. Beckmann5,7, Daniele Cusi12, Federica Rizzi13,
Franco Cappuccio14, Jacques Cornuz10, Fred Paccaud1, Vincent Mooser8,15,
Jean-Michel Gaspoz2,Ge
´rard Waeber9, Michel Burnier6, Peter Vollenweider9,
Chin B Eap3,16 and Murielle Bochud1,∗
1
Community Prevention Unit, University Institute of Social and Preventive Medicine (IUMSP), Lausanne University
Hospital, Lausanne, Switzerland,
2
Unit of Population Epidemiology, Division of Primary Care Medicine,
Department of Community Medicine and Primary Care and Emergency Medicine and
3
Cardiology, Department of
Speciality Medicine, Geneva University Hospital, Geneva, Switzerland,
4
Unit of Pharmacogenetics and Clinical
Psychopharmacology, Centre for Psychiatric Neurosciences, Department of Psychiatry, Centre Hospitalier
Universitaire Vaudois, University of Lausanne, Hospital of Cery, Prilly, Switzerland,
5
Department of Medical Genetics,
6
Service of Nephrology, Department of Medicine, Centre Hospitalier Universitaire Vaudois
7
Service of Medical
Genetics, Centre Hospitalier Universitaire Vaudois,
8
Department of Laboratories, Service of Biomedicine, Centre
Hospitalier Universitaire Vaudois,
9
Department of Medicine, Internal Medicine, Centre Hospitalier Universitaire
Vaudois and
10
Community Medicine and Public Health, University Outpatient Clinic, University of Lausanne,
Lausanne, Switzerland,
11
Swiss Institute of Bioinformatics, Lausanne, Switzerland,
12
Department of Medicine,
Surgery and Dentistry, Graduate School of Nephrology, Division of Nephrology, University of Milano, San Paolo
Hospital, Milano, Italy,
13
KOS Genetic srl, Milano, Italy,
14
University of Warwick, Warwick Medical School, Coventry,
UK,
15
Genetics Division, GlaxoSmithKline R&D King of Prussia, PA 19406, USA and
16
School of Pharmaceutical
Sciences, University of Geneva, University of Lausanne, Geneva, Switzerland
Received October 26, 2011; Revised February 22, 2012; Accepted April 2, 2012
The 15q24.1 locus, including CYP1A2, is associated with blood pressure (BP). The CYP1A2 rs762551 C allele
is associated with lower CYP1A2 enzyme activity. CYP1A2 metabolizes caffeine and is induced by smoking.
The association of caffeine consumption with hypertension remains controversial. We explored the effects of
CYP1A2 variants and CYP1A2 enzyme activity on BP, focusing on caffeine as the potential mediator of
CYP1A2 effects. Four observational (n516 719) and one quasi-experimental studies (n5106) including
European adults were conducted. Outcome measures were BP, caffeine intake, CYP1A2 activity and poly-
morphisms rs762551, rs1133323 and rs1378942. CYP1A2 variants were associated with hypertension in
non-smokers, but not in smokers (CYP1A2-smoking interaction P50.01). Odds ratios (95% CIs) for hyperten-
sion for rs762551 CC, CA and AA genotypes were 1 (reference), 0.78 (0.59 – 1.02) and 0.66 (0.50– 0.86), respect-
ively, P50.004. Results were similar for the other variants. Higher CYP1A2 activity was linearly associated
with lower BP after quitting smoking (P50.049 and P50.02 for systolic and diastolic BP, respectively),
but not while smoking. In non-smokers, the CYP1A2 variants were associated with higher reported caffeine
intake, which in turn was associated with lower odds of hypertension and lower BP (P50.01). In Mendelian
randomization analyses using rs1133323 as instrument, each cup of caffeinated beverage was negatively
∗
To whom correspondence should be addressed at: Institute of Social and Preventive Medicine, Route de la Corniche 2, 1066 Epalinges, Switzerland.
Tel: +41 213140899; Fax: +41 213147373; Email: murielle.bochud@chuv.ch
#The Author 2012. Published by Oxford University Press. All rights reserved.
For Permissions, please email: journals.permissions@oup.com
Human Molecular Genetics, 2012, Vol. 21, No. 14 3283–3292
doi:10.1093/hmg/dds137
Advance Access published on April 5, 2012
at University of Warwick on August 27, 2012http://hmg.oxfordjournals.org/Downloaded from

associated with systolic BP [29.57 (216.22, 22.91) mmHg]. The associations of CYP1A2 variants with
BP were modified by reported caffeine intake. These observational and quasi-experimental results strongly
support a causal role of CYP1A2 in BP control via caffeine intake.
INTRODUCTION
Recent meta-analyses of genome-wide association studies
(GWASs) have identified the CYP1A2 locus, on chromosome
15q24.1, as being robustly associated with blood pressure (BP)
and hypertension (1,2).
In humans, CYP1A2, encoded by the CYP1A2 gene, is re-
sponsible for 13% of the cytochrome P450 activity of the
liver (3). CYP1A2 is the main enzyme responsible for the me-
tabolism of caffeine (1,3,7-trimethylxanthine, 137X), a purine
alkaloid that occurs naturally in coffee beans. Caffeine intake
is heritable (4), and this heritability appears to be quite specific
to caffeine (5). Recently, a GWAS has identified the CYP1A2
gene as being associated with caffeine consumption (6). Al-
though several CYP1A2 genetic variants have been identified
(http://www.cypalleles.ki.se/, last accessed date 1 August
2011), their effects on the CYP1A2 enzyme activity are not
clear (7).
On the short term (i.e. ,3 months), regular coffee or caf-
feine intake increases BP (8). Also, acute consumption of caf-
feine at dietary levels appears to raise BP (9). Yet, a tolerance
to the acute cardiovascular effects of caffeine has been
described (10) and there is no clear evidence that regular caf-
feine intake over long periods of time increases the incidence
of hypertension, as reflected by an absence of significant posi-
tive association in large-scale prospective studies (11 –13) and
inconsistent results in cross-sectional studies (14 –16).
Smoking is a well-known inducer of CYP1A2 activity (17)
and quitting smoking decreases CYP1A2 activity (18,19). The
aims of this study were to analyze the associations of CYP1A2
variants with BP and hypertension in the general adult popu-
lation, exploring the potential modification of these associa-
tions by smoking, and focusing on caffeine as the potential
mediator of CYP1A2 effects on BP.
RESULTS
Table 1lists the characteristics of the participants to the obser-
vational (n¼16 719 independent people) and the experimen-
tal study (n¼106) including European adults. There were no
evidence of departure from Hardy – Weinberg proportions
(P¼0.94, P¼0.58 and P¼0.86, for rs762551, rs1133323
and rs1378942, respectively). Adjusting for population stratifi-
cation did not alter any of the genetic associations presented in
what follows (data not shown).
Association of CYP1A2 variants with hypertension,
by smoking status
Table 2shows adjusted associations of CYP1A2 variants with
hypertension, by smoking status. Among non-smokers, the
three single-nucleotide polymorphisms (SNPs) were linearly
associated with hypertension. The protective allele is T for
rs1133323 and A for rs1378942 and rs762551. Individuals
with genotypes CT and TT at rs1133323 were 24 and 35%
less likely to have hypertension compared with individuals
with a CC genotype (P¼0.0006). Rs1378942 CA and AA
individuals were 18 and 33% less likely to have hypertension
than rs1378942 CC individuals (P¼0.002), and rs762551 AC
and AA individuals were 22 and 34% less likely to have
hypertension than rs762551 CC individuals (P¼0.004). Inter-
actions between smoking and each CYP1A2 variant on hyper-
tension were significant (P-values for interaction: P¼0.009,
P¼0.01 and P¼0.01 for rs1133323, rs1378942 and
rs762551). CYP1A2 genotypes were not associated with
hypertension among smokers. Results were similar among
never and ex-smokers (interaction tests were not significant,
data not shown). Associations and trends of CYP1A2 variants
with hypertension were confirmed in HYPERGENES, with
statistically significant results for rs1133323 and rs1378942
among non-smokers only (Supplementary Material, Table S1).
However, the statistical interactions were not significant.
Association of CYP1A2 activity with BP before
and after smoking cessation
Figure 1shows the adjusted mean systolic (SBP) and diastolic
BP (DBP) by tertiles of CYP1A2 activity before and after
smoking cessation. No linear relationship was seen before
smoking cessation. After smoking cessation, the adjusted
mean SBP decreased from 131.5 (standard error, SE, 2.4) to
125.0 (2.4), and to 124.7 (2.4) mmHg, with increasing tertiles
of CYP1A2 activity (P-value for trend ¼0.049). The adjusted
mean DBP decreased from 82.7 (1.8) to 79.5 (1.7), and to 76.6
(1.8) mmHg, with increasing tertiles of CYP1A2 activity
(P-value for trend ¼0.02).
CYP1A2 variants and caffeine consumption
Among non-smokers, the three SNPs were associated with
high reported caffeine intake. Rs1133323 CT and TT indivi-
duals were, respectively, 40 and 60% more likely than
rs1133323 CC individuals to report high caffeine intake
(P-value ¼0.0001). CYP1A2 genotypes were not associated
with high reported caffeine intake in smokers (Supplementary
Material, Table S2).
Reported caffeine intake and hypertension or BP
Table 3shows associations of reported caffeine intake with
hypertension, by smoking status. Among non-smokers,
reported caffeine intake showed a negative dose – effect rela-
tionship with hypertension. Compared with subjects who
report 0 cup/day of caffeine intake, individuals who reported
1– 3 cups/day, 4 – 6 cups/day and .6 cups/day were 13, 22
and 41% less likely to have hypertension (P-value for
3284 Human Molecular Genetics, 2012, Vol. 21, No. 14
at University of Warwick on August 27, 2012http://hmg.oxfordjournals.org/Downloaded from

trends ¼0.03). Reported caffeine intake was not associated
with hypertension among smokers. Associations and trends
were similar in never smokers and ex-smokers (Supplemen-
tary Material, Table S3). These results were confirmed in
the independent population-based Bus Sante´ study (Supple-
mentary Material, Table S4). Supplementary Material,
Figure S1 illustrates the adjusted mean SBP and DBP in
smokers and non-smokers, by number of reported caffein-
ated cups/day. The mean adjusted SBP and DBP decreased
with the number of reported caffeinated cups/day in non-
smokers (both P-value for trends,0.05), but not in
smokers.
Association of reported caffeine intake and BP
using a Mendelian randomization approach
with instrumental variables
All three SNPs were appropriate instrumental variables (F.
10 in the first-stage regression) (Table 4). The rs1133323
variant was the best instrument (F¼15). The negative
Table 1. Demographic and risk factor characteristics for all participants, by study (n¼16 719)
CoLaus (N¼6127) Bus Sante´(N¼7573) HYPERGENES:
controls (N¼1396)
HYPERGENES:
cases (N¼1517)
GenSmoke (N¼106)
Age (years), mean (SD) 53.1 (10.8) 56.2 (11.4) 63.8 (12.01) 49.0 (9.43) 40.9 (10.7)
Men, n(%) 2909 (47.5) 3793 (50.1) 812 (58.17) 1001 (65.99) 55 (51.9)
High reported caffeine intake
(4+cups/day), n(%)
1764 (28.8) 3928 (51.9)
a
NA
b
NA NA
Current smokers, n(%) 1647 (26.9) 1386 (18.3) 408 (29.23) 451 (29.73) 106 (100.0)
c
Current alcohol consumption, n(%) 1552 (25.3) 2339 (30.9) NA NA NA
Diabetes, n(%) 386 (6.3) 390 (5.2) NA NA NA
Contraceptive use (women), n(%) 261 (8.1) NA NA NA 23 (45.1)
Body mass index (kg/m
2
), mean (SD) 25.8 (4.5) 25.5 (4.2) 25.5 (3.55) 27.2 (3.94) 25.1 (4.2)
Hypertension, n(%) 2197 (35.9) 2519 (33.3) 0 (0) 1517 (100) 35 (33.0)
SBP (mmHg), mean (SD) 128.3 (17.9) 128.3 (18.8) 123.3 (9.44) 153.7 (13.94) 128.4 (15.8)
DBP (mmHg), mean (SD) 79.3 (10.8) 75.7 (10.9) 77.2 (6.45) 99.1 (8.51) 81.6 (10.6)
eGFR CKD-epi
d
(ml/min
per 1.73 m
2
), mean (SD)
85.7 (15.2) NA 82.0 (14.90) 88.7 (16.95) NA
Triglycerides (mmol/l),
mean (SD)
1.40 (1.2) 1.3 (0.9) 1.4 (0.65) 1.4(0.71) NA
Total cholesterol (mmol/l),
mean (SD)
5.6 (1.0) 5.6 (1.0) 5.6 (1.01) 5.5(1.00) NA
a
High reported caffeine intake defined as 2+cups/day.
b
NA, not available.
c
All subjects are smokersat baseline.
d
eGFR, glomerular filtration rate estimated using the CKD-EPI formula.
Table 2. Association of CYP1A2 variants with hypertension, by smoking status, odds ratio (95%CI), in the CoLaus study
CYP1A2 variants Non-smokers Smokers
rs1133323 genotype CC CT TT P-value CC CT TT P-value
N1046 1789 767 387 668 253
Unadjusted Ref 0.81 (0.74– 0.95) 0.71 (0.59–0.87) 0.002 Ref 1.18 (0.89– 1.56) 1.16 (0.82 – 1.65) 0.49
Model 1
a
Ref 0.76 (00.63– 0.91) 0.64 (0.51 – 0.80) ,0.001 Ref 1.27 (0.92 – 1.74) 1.21 (0.81–1.81) 0.33
Model 2
b
Ref 0.76 (0.63–0.92) 0.65 (0.52 – 0.82) ,0.001 Ref 1.27 (0.93 – 1.75) 1.22 (0.81–1.82) 0.32
Model 3
c
Ref 0.76 (0.63–0.92) 0.65 (0.52 – 0.82) ,0.001 Ref 1.26 (0.92 – 1.74) 1.21 (0.81–1.81) 0.35
rs1378942 genotype CC CA AA P-value CC CA AA P-value
N497 1826 1712 197 696 553
Unadjusted Ref 0.86 (0.71– 1.05) 0.79 (0.64–0.97) 0.06 Ref 0.85 (0.60 – 1.20) 1.06 (0.74 – 1.50) 0.19
Model 1
a
Ref 0.80 (0.63–1.02) 0.66 (0.52 – 0.84) 0.002 Ref 0.82 (0.55 – 1.21) 1.08 (0.72 – 1.62) 0.12
Model 2
b
Ref 0.81 (0.64–1.03) 0.67 (0.53 – 0.86) 0.003 Ref 0.82 (0.55 – 1.22) 1.09 (0.73 – 1.63) 0.13
Model 3
c
Ref 0.82 (0.64–1.04) 0.67 (0.53 – 0.85) 0.002 Ref 0.81 (0.55 – 1.20) 1.09 (0.72 – 1.62) 0.12
rs762551 genotype CC CA AA P-value CC CA AA P-value
N366 1693 1958 148 623 664
Unadjusted Ref 0.86 (0.69– 1.09) 0.80 (0.64–1.01) 0.14 Ref 0.90 (0.61 – 1.34) 1.06 (0.72 – 1.57) 0.42
Model 1
a
Ref 0.77 (0.59–1.01) 0.66 (0.50 – 0.86) 0.004 Ref 0.80 (0.51 – 1.25) 1.06 (0.68 – 1.65) 0.13
Model 2
b
Ref 0.78 (0.60–1.03) 0.67 (0.51 – 0.88) 0.006 Ref 0.80 (0.51 – 1.26) 1.07 (0.68 – 1.66) 0.13
Model 3
c
Ref 0.78 (0.59–1.02) 0.66 (0.50 – 0.86) 0.004 Ref 0.79 (0.50 – 1.24) 1.06 (0.68 – 1.65) 0.11
a
Model 1 was adjusted for age, sex, BMI, contraceptive use, cholesterol, triglycerides, diabetes, alcohol and CKD-EPI.
b
Model 2 was adjusted as full model 1 +reported caffeine intake.
c
Model 3 was adjusted as full model 2 +menopause.
Model 3-adjusted P-values for interaction tests: P¼0.009, P¼0.01 and P¼0.01 for rs1133323, rs1378942 and rs762551, respectively.
Human Molecular Genetics, 2012, Vol. 21, No. 14 3285
at University of Warwick on August 27, 2012http://hmg.oxfordjournals.org/Downloaded from

association of reported cup of caffeine intake with BP
observed in ordinary least squares (OLS) analyses was con-
firmed using two-stage least squares (2SLS). For both SBP
and DBP, the 2SLS-based associations were stronger than the
OLS-based associations—SBP OLS versus 2SLS: 20.48
(20.76, 20.21) versus 29.57 (216.22, 22.91), 20.48
(20.74, 20.22) versus 29.23 (216.12, 22.30) and 20.44
(20.79, 20.18) versus 26.55 (212.77, 20.33) for
rs1133323, rs1378942 and rs762551, respectively (Table 4);
DBP OLS versus 2SLS: 20.96 ( 20.54, 20.18) versus 25.47
(29.57,21.38), 20.37 (20.54, 20.20) versus 27.83
(213.02, 22.64) and 20.33 (20.50, 20.16) versus 26.00
(210.61, 21.38) for rs1133323, rs1378942, and rs762551, re-
spectively. The three SNPs were not appropriate instrumental
variables in smokers (Supplementary Material, Table S5).
These results support a negative causal relation between caf-
feine intake and BP.
Association of CYP1A2 variants with BP among
non-smokers, by caffeine intake
Significant negative associations of CYP1A2 variants with
both SBP and DBP were seen only in the presence of caffeine
Table 3. Association of reported caffeine intake with hypertension, by smoking status, odds ratio (95% CI), in the CoLaus study
Non-smokers Smokers
Reported caffeine
intake
0 cups/day 1–3 cups/day 4– 6 cups/day .6 cups/day P-value for
trend
0 cups/day 1–3 cups/day 4– 6 cups/day .6 cups/day P-value
for trend
N326 2999 1006 149 69 969 465 144
Model 1 Ref 0.85 (0.65– 1.13) 0.76 (0.56–1.03) 0.64 (0.40– 1.04) 0.025 Ref 1.53 (0.80– 2.93) 1.56 (0.80–3.05) 0.98 (0.45– 2.12) 0.16
Model 2 Ref 0.86 (0.62– 1.18) 0.77 (0.54–1.09) 0.61 (0.35– 1.05) 0.040 Ref 1.84 (0.84– 4.03) 1.82 (0.81–4.08) 0.97 (0.38– 2.44) 0.31
Model 3 Ref 0.87 (0.63– 1.20) 0.78 (0.54–1.11) 0.59 (0.34– 1.02) 0.033 Ref 1.83 (0.84– 4.01) 1.82 (0.81–4.07) 0.96 (0.38– 2.42) 0.31
Model 1 was adjusted for age, sex, BMI, contraceptive use, total cholesterol, triglycerides, diabetes, alcohol and CKD-EPI.
Model 2 was adjusted as model 1 +CYP1A2 variants.
Model 3 was adjusted as full model 2 +menopause.
Model 3-adjusted P-value for interaction test: P¼0.19.
Figure 1. Before and after smoking cessation mean (SE) SBP and DBP, by
tertiles of CYP1A2 activity in GenSmoke study (n¼106). Adjusted for
age, sex, BMI, number of cigarettes smoked at baseline, smoking cessation
treatment and contraceptive use.
3286 Human Molecular Genetics, 2012, Vol. 21, No. 14
at University of Warwick on August 27, 2012http://hmg.oxfordjournals.org/Downloaded from

intake (Supplementary Material, Tables S6 and S7). In the
presence of reported caffeinated beverage intake, the coeffi-
cients for SBP were 21.83 (22.63, 21.03), 21.52 (22.31,
20.73) and 21.44 (22.27, 20.62) for rs1133323 T,
rs1378942 A and rs762551 A alleles, respectively (Supple-
mentary Material, Table S6). The coefficients for DBP
were 21.03 (21.56, 20.50), 21.23 (21.75, 20.71) and
21.09 (21.64, 20.55) for rs1133323 T, rs1378942 A and
rs762551 A alleles, respectively (Supplementary Material,
Table S7). In the absence of caffeine intake, regression coeffi-
cients were positive (yet close to zero) and tended to differ sig-
nificantly from coefficients obtained in the presence of
reported caffeine intake (adjusted P-value for caffeine
intake– CYP1A2 interaction were ,0.10 for rs1378942 and
rs762551). Given the small number of participants reporting
no intake of caffeinated beverage, we have low power for
these associations and cannot exclude a small negative effect
of rs1133323 T, rs1378942 A and rs762551 A variants on
BP. With respect to hypertension, the P-values for caffeine –
CYP1A2 variants interactions for rs1133323, rs1378942 and
rs762551 were 0.25, 0.01 and 0.08, respectively, in models
that included covariates, which suggest that the effects of
CYP1A2 variants on hypertension in the presence differ from
those in the absence of reported caffeine intake. We therefore
provide evidence that reported caffeine intake modifies the
effects of CYP1A2 variants on BP and hypertension.
Our results point toward caffeine intake as a likely mechan-
ism by which the CYP1A2 gene and enzyme activity may in-
fluence BP and risk of hypertension (Supplementary Material,
Fig. S2).
DISCUSSION
Our results suggest that alleles in three CYP1A2 variants
(rs762551, rs1133323 and rs1378942) may drive, at least in
part, the robust association of the 15q24.1 locus with BP
and hypertension (1,2). We found that smoking, a well-known
CYP1A2 inducer, modified the association of CYP1A2 var-
iants with hypertension. In the CoLaus study, non-smokers
carrying the AA CYP1A2 genotype, which is associated with
increased CYP1A2 activity, were 35% less likely to have
hypertension than non-smokers carrying the reference
CYP1A2 genotype. We did not observe such associations in
current smokers. The strength of the association of CYP1A2
variants with BP reported previously may therefore have
been underestimated because results were not stratified by
smoking status (1,2). Importantly, using a quasi-experimental
design, we found CYP1A2 activity to be negatively associated
with SBP and DBP in 106 ex-smokers, but the linear associ-
ation was absent before these same 106 subjects quit
smoking. This supports a causal role of CYP1A2 variants on
BP and highlights the potential functional role of CYP1A2 ac-
tivity in BP control.
We found that the three selected CYP1A2 variants, which
are not highly correlated with each other, are strongly asso-
ciated with reported caffeine intake. The latter confirms the
results of recent GWASs on caffeine intake (20,21). The iden-
tification of CYP1A2 genetic markers associated with caffeine
intake is in line with both the activity of the CYP1A2 enzyme
Table 4. Change in SBP and DBP (mmHg) by reported daily caffeinated beverage cups using an instrumental variable approach (rs1133323, rs1378942, rs762551) (the CoLaus study)
SNP Model SBP DBP F-value
(first stage)OLS 2SLS P
diff
OLS 2SLS P
diff
Beta (95% CI) P-value Beta (95% CI) P-value Beta (95% CI) P-value Beta (95% CI) P-value
rs1133323
T allele
Unadjusted
(N¼4910)
20.71 (21.04, 20.38) ,0.001 29.58 (217.02, 22.17) 0.019 0.018 20.38 (20.57, 20.018) ,0.001 25.87 (210.37, 21.36) 0.011 0.017 15.26
Adjusted
a
(N¼4887)
20.48 (20.76, 20.21) 0.001 29.57 (216.22, 22.91) 0.005 0.008 20.96 (20.54, 20.18) ,0.001 25.47 (29.57, 21.38) 0.009 0.015
rs1378942
A allele
Unadjusted
(N¼5481)
20.64 (20.96, 20.33) ,0.001 26.88 (214.31, 0.55) 0.69 0.099 20.36 (20.54, 20.17) ,0.001 27.79 ( 213.45, 22.13) 0.007 0.010 11.00
Adjusted
a
(N¼5454)
20.48 (20.74, 20.22) ,0.001 29.23 (216.12, 22.30) 0.009 0.013 20.37 (20.54, 20.20) ,0.001 27.83 (213.02, 22.64) 0.003 0.005
rs762551
A allele
Unadjusted
(N¼5454)
20.62 (20.92, 20.30) ,0.001 23.45 (210.00, 3.09) 0.301 0.396 20.32 (20.51, 20.13) 0.001 25.21 (29.80, 20.61) 0.026 0.036 13.44
Adjusted
a
(N¼5426)
20.44 (20.79, 20.18) 0.001 26.55 (212.77, 20.33) 0.039 0.054 20.33 (20.50, 20.16) ,0.001 26.00 (210.61, 21.38) 0.011 0.016
a
Adjusted for age, sex, BMI, contraceptive use, cholesterol, triglyceride, diabetes, alcohol, eGFR (CKD-EPI) and menopause.
OLS, ordinary least squares; 2-SLS, two-stage least squares.
Human Molecular Genetics, 2012, Vol. 21, No. 14 3287
at University of Warwick on August 27, 2012http://hmg.oxfordjournals.org/Downloaded from

and the significant heritability for caffeine use, toxicity, toler-
ance and withdrawal symptoms (4). CYP1A2 variants may,
therefore, represent interesting proxies for reported caffeine
intake.
High reported caffeine intake was associated with a lower
prevalence of hypertension only in non-smokers in two inde-
pendent population-based studies: CoLaus and Bus Sante´.
Smoking appears to blunt the association of caffeine intake
with hypertension, possibly via CYP1A2 induction. Trends
were similar in CoLaus participants not aware of having
hypertension (4493, 73%), implying that our results do not
merely reflect a reduction in caffeine intake following hyper-
tension diagnosis (i.e. reverse causality). Overall, our findings
suggest that high caffeine intake might protect non-smokers
against hypertension. Failure to account for the modifying
effect of smoking may explain why no clear association of
regular caffeine intake with the incidence of hypertension
has been found in large-scale prospective studies (11,12) and
in cross-sectional studies (14 –16).
We found that the association of CYP1A2 variants with BP
was significant and negative only in the presence of reported
caffeine intake, although we have low power to explore this
association in the absence of reported caffeine intake. These
results suggest that caffeine intake may mediate the effect of
CYP1A2 on BP. To further explore the causal role of caffeine
intake on BP, we conducted a Mendelian randomization ana-
lysis, using CYP1A2 variants as genetic instruments. We
found convincing evidence that caffeine intake is causally
negatively associated with SBP and DBP and may therefore
protect against the risk of developing hypertension. The
results of the Mendelian randomization approach are compat-
ible with a clinically relevant effect of caffeine intake on BP
(up to 9 and 7 mmHg of SBP and DBP by cup of caffeinated
beverages per day, respectively).
Our findings are in line with the results of two recent
large-scale prospective studies, showing that high reported
coffee consumption may reduce the risk of cerebral infarction
(22) or stroke (23,24). In 3837 type 2 diabetic patients, high
coffee intake at baseline tended to be associated, although
not significantly, with lower risk of stroke after 20 years of
follow-up (25). Previous studies showing either no association
(26,27) or a positive association (28) between coffee intake
and stroke included a much smaller number of strokes
(,100 events) than more recent studies (.1000 events)
(22,23). Because hypertension is the single most important
modifiable risk factor for ischemic stroke (29), a protective
effect of caffeine intake on hypertension could explain part
of the observed reduction in stroke in people with elevated
coffee intake (22,23). In both studies (22,23), BP and/or
hypertension-unadjusted prevalences decreased with higher
reported coffee intake. In the Nurse’s Health Study (23), the
protective effect of long-term coffee consumption on stroke
was present in never and past smokers but not in smokers,
compatible with the hypothesis that the protective effect of
caffeine intake against hypertension-related CV outcomes is
obscured in the presence of smoking.
The wide inter-individual variability in CYP1A2 activity
can be due to both environmental (e.g. smoking, caffeine
intake) and genetic (e.g. CYP1A2 variants) factors (30,31).
In the GenSmoke study, individual change in CYP1A2 activity
after smoking cessation ranged from 1.0-fold (no change) to
7.3-fold decrease (19). Yet, none of the currently identified
CYP1A2 polymorphisms seems to explain the large inter-
individual variability in CYP1A2 activity (32). Therefore, un-
identified genetic variations in the CYP1A2 gene and/or in
other genes controlling CYP1A2 activity, such as the aryl
hydrocarbon receptor (33) or the cytochrome P450 oxidore-
ductase (34), could be responsible for the observed differences
in CYP1A2 enzymatic activity.
One of the mechanisms by which CYP1A2 variants, and
thus CYP1A2 activity, could influence BP is via the effect
of caffeine on renal segmental tubular sodium handling. Caf-
feine and its metabolite, paraxanthine, have known diuretic
and natriuretic effects (35 –37) and belong to the group of
methylxanthines that are nonselective adenosine receptor
antagonists (38). Caffeine exerts its natriuretic action via the
adenosine A1 receptors blockade (39–43), leading to
decreased proximal tubular sodium re-absorption (40,41,43).
Clinical and policy implications
Our results suggest that there is no evidence that patients with
high BP need to refrain from caffeinated beverages. In con-
trary, in non-smokers, caffeinated beverages are associated
with lower risk of high BP. Given the observed direct link
between the CYP1A2 enzyme and BP, factors that modify
CYP1A2 activity should be considered in the management
of hypertension. These include drugs (e.g. omeprazole, cloza-
pine), habits (smoking, caffeine) and dietary factors (crucifer-
ous vegetables, charcoal-broiled meat) (3).
Considering the widespread use of caffeine and the high
prevalence of hypertension, our results may also have large
public health implications. There is currently no specific rec-
ommendation regarding caffeine intake in hypertension guide-
lines (44,45). Our analyses suggest that the reported protective
effect of caffeine intake on stroke could be mediated via the
inverse association between caffeine intake and hypertension—
the major modifiable risk factor for stroke. This could guide
recommendation on the appropriateness of caffeinated bever-
age consumption in the context of stroke prevention, which
does not currently mention caffeine intake (46).
Limitations and strengths
We did not measure serum caffeine levels, but used reported
caffeine intake. Although .70% of caffeine is provided by
coffee consumption, our results are not generalizable to
coffee consumption. Yet, results were similar in two independ-
ent population-based studies with different questionnaire data.
Although the cross-sectional nature of our study limits causal
inference for non-genetic associations, genetic associations
provide information on cumulative risk even in cross-sectional
designs. Oral contraceptive and cigarette smoking have been
reported to, respectively, inhibit and increase CYP1A2 activity
(47), and we account for that by adjusting and stratifying our
analyses accordingly. Given that heavy coffee consumption
can also increase CYP1A2 activity (48), we adjusted for
reported caffeine intake when appropriate. There are,
however, numerous other drugs that are metabolized by
CYP1A2 to an extent suggesting clinical relevance (49). To
3288 Human Molecular Genetics, 2012, Vol. 21, No. 14
at University of Warwick on August 27, 2012http://hmg.oxfordjournals.org/Downloaded from

account for this, we also restricted our analyses to CoLaus
individuals who were not taking any drugs (n¼2539, 41%).
This did not alter the results materially (data not shown).
Finally, the effect of coffee compounds other than caffeine
on CYP1A2 is unlikely given that caffeine is the only
known coffee compound to be detoxified by CYP1A2 (50).
As few SNPs are associated with altered CYP1A2 activity
(7), we also measured the CYP1A2 activity, using a gold
standard method.
There are commonly acknowledged necessary conditions
for Mendelian randomization to provide causal inference in
observational epidemiology (51). Results for the Mendelian
randomization approach should therefore be interpreted cau-
tiously. For example, although the instruments (i.e. CYP1A2
variants) were clearly correlated with caffeine intake, one con-
dition is that the genetic instruments affect BP in no other way
than through caffeine intake. In addition, our Mendelian ran-
domization analyses resulted in wide confidence intervals
and low precision as is usually the case for genetic instruments
in common complex human disease. Also, we cannot exclude
that the true causal variant may be in linkage disequilibrium
with these alleles. Finally, the absence of association among
smokers could be real or due to a lack of power. If the
effect were the same among smokers and non-smokers, the
power to detect it among smokers in this study would be
,50%. However, (i) the different direction of the associations
in smokers and non-smokers, (ii) the absence of clear trends in
smokers, (iii) the presence of statistical interactions between
smoking status and CYP1A2 variants and (iv) the presence
of significant association of CYP1A2 activity with BP after,
but not before, smoking cessation in the same subjects strong-
ly suggest that the effects of CYP1A2, CYP1A2 activity and
caffeine intake on BP and hypertension differ in non-smokers
and in smokers.
In summary, our results based on gene – environment inter-
action, quasi-experimental data and the Mendelian randomiza-
tion approach provide strong evidence that caffeine mediates
the effect of CYP1A2 on BP and hypertension, and that
smoking modifies these associations. The associations we
found are strong, biologically credible, with dose– response
relationships and, for the genetic ones, with unambiguous tem-
poral sequence. Overall, our findings may lead to a new area
of research for the prevention and treatment of hypertension.
MATERIALS AND METHODS
Details are available in the Supplementary Material.
The CoLaus study: CYP1A2 variants, reported caffeine
intake, BP and hypertension
The CoLaus study complied with the Declaration of Helsinki
and was approved by the local Institutional Ethics Committee.
All participants gave written informed consent. The sampling
procedure of the CoLaus study was population based, with
participants aged 35– 75 years, and details have been
described previously (52).
Assessment process, clinical and biological data
Recruitment began in June 2003 and ended in May 2006. BP
was measured three times on the left arm after at least 10 min
rest in the seated position, using a clinically validated auto-
mated oscillometric device (Omron
w
HEM-907, Matsusaka,
Japan) with a standard cuff, or a large cuff if arm circumfer-
ence was ≥33 cm (53). The average of the last two BP read-
ings was used for analyses. Hypertension was defined as mean
SBP ≥140 mmHg or mean DBP ≥90 mmHg or presence of
anti-hypertensive medication. Participants self-reported their
consumption of caffeinated beverages as follows: 0 cup/day,
1– 3 cups/day, 4 –6 cups/day and .6 cups/day. Smoking
was defined as present if a participant reported to be a
current smoker at the time of examination, regular alcohol
consumption was defined as present for participants reporting
to drink alcohol at least once a day, and contraceptive pill use
was self-reported. Diabetes was defined as a fasting glucose
≥7 mmol/l and/or the presence of antidiabetic drug treatment
(insulin or oral drugs). Additional information can be found in
Supplementary Material.
Genotyping and CYP1A2 variants
Nuclear DNA was extracted from whole blood for whole-
genome scan analysis, and genotyping was performed using
the Affimetrix 500 K SNP chip, as recommended by the manu-
facturer. Overall, 91 single-nucleotide polymorphisms (SNPs)
were genotyped or imputed within or near the CYP1A2 gene
(Methods in Supplementary Material). Among these, 55 had
minor allele frequency .10%. rs762551, a polymorphism
shown to have a main effect on CYP1A2 activity (http://www.
snpedia.com/index.php/Rs762551), was not among the geno-
typed SNPs but was imputed with good quality (r
2
-hat ¼
0.92). We selected the CYP1A2 SNP that is most strongly
associated with (i) DBP in the GWAS (rs1378942) (2), (ii)
reported caffeine intake in the CoLaus study (rs1133323)
and (iii) CYP1A2 enzyme activity (rs762551) (54). Linkage
disequilibrium (r
2
) of these SNPs were as follows in the
CoLaus study: rs762551– rs1133323, r
2
¼0.23; rs762551–
rs1378942, r
2
¼0.57; rs1133323– rs1378942, r
2
¼0.43.
Allele frequencies were estimated by the gene counting
method, and departures from Hardy– Weinberg equilibrium
were tested using a x
2
test.
The GenSmoke study: CYP1A2 phenotyping
(experimental study)
GenSmoke is a longitudinal study conducted at the Centre for
Psychiatric Neurosciences of the Department of Psychiatry
and at the University Outpatient Clinic of Lausanne, Switzer-
land. The study was approved by the local Institutional Ethics
Committee. Written informed consent was obtained from all
the participants. The study primarily aimed at assessing the
inter-individual variability of the induction of CYP1A2 by
smoking, as described elsewhere (19). CYP1A2 activity was
determined before and 4 weeks after smoking cessation in
volunteers. The 4-week duration was chosen because the in-
ductive effect of smoking on CYP1A2 is expected to disappear
within 4 weeks (55), which was confirmed in GenSmoke (19).
Smoking abstinence was assessed by self-declaration and
by measuring expired carbon monoxide levels (Micro
Human Molecular Genetics, 2012, Vol. 21, No. 14 3289
at University of Warwick on August 27, 2012http://hmg.oxfordjournals.org/Downloaded from

Smokerlyzer, Bedfont Scientific Ltd, Rochester, UK), and all
comedications were recorded. Additional information can be
found in Supplementary Material.
Individuals were asked to refrain from caffeine-containing
beverages and foods on the night before the day of the sched-
uled test. Blood was collected 6 h after the intake of a 200 mg
caffeine capsule. Before caffeine intake (hour 0) and again
before blood sampling (at 6 h), compliance regarding caffeine re-
striction was assessed by self-declaration and a new test
was programmed and performed if compliance was doubtful
(n¼2). The paraxanthine (17X) and caffeine (137X) plasma
levels were measured by gas chromatography/mass spectrometry,
using a previously described method (56). The 17X/137X ratio,
which is a valid marker of CYP1A2 activity (49,57), was calcu-
lated for all individuals. BP was determined by a single measure
using the Omron HEM-907 Digital Blood Pressure Monitor
machine, in the seated position, after at least 5 min of rest.
Statistical analyses
The CoLaus study
Continuous variables were described with means and standard
deviations and categorical variables with percentages. We
used multiple logistic regression to test the association
between (i) CYP1A2 genotypes and hypertension, (ii)
CYP1A2 genotypes and high reported caffeine intake (i.e. 4
cups of more per day) and (iii) reported caffeine intake (i.e.
0 cup/day, 1– 3 cups/day, 4 –6 cups/day and .6 cups/day)
and hypertension, while adjusting for potential confounding
factors. We used multiple linear regression to test the associ-
ation between SBP/DBP and reported caffeine intake while
adjusting for potential confounding factors. For genetic associ-
ation analyses, we adjusted SBP/DBP for antihypertensive
treatment by adding a 15/10 mmHg constant, as suggested
(58). We used multiple logistic or linear regressions to test
for trends. Interactions between smoking and CYP1A2 variants,
and smoking and high reported caffeine intake, were tested
using likelihood ratio tests. To ensure the robustness of our find-
ings, we conducted additional analyses in participants without
any medication and in participants not aware of having hyper-
tension, adjusting for population stratification principle compo-
nents. Only those individuals for whom all covariates of interest
for the purpose of this study were available were included in the
analysis (99.1% of the overall cohort).
The GenSmoke study
To test the association between CYP1A2 activity and BP, we
used multiple linear regression. Analyses were adjusted for
age (age-squared for DBP), sex, BMI, contraceptive use,
smoking cessation treatment and number of cigarettes
smoked at baseline as covariates in the models. To illustrate
the results graphically, we used dummy variables coding for
tertiles of CYP1A2 activity. We tested the association of
CYP1A2 activity with BP before and after smoking cessation.
Mendelian randomization
To explore the potential causal effect of caffeine intake on BP,
we applied a Mendelian randomization approach using genetic
instrumental variables (59,60). We used the number of caffein-
ated beverage cups as our exposure variable. The number of
caffeinated beverage cups were coded as 0, 2, 5 and 7 for
the 0 cup/day, 1– 3 cups/day, 4 – 6 cups/day and .6 cups/
day categories, respectively. In a first stage, we regressed
the number of cups on our instrument (genotypes at
rs1133323, rs1378942, rs762551). In a second stage, we
regressed the SBP (similarly the DBP) on the fitted values
from the first-stage regression. The regression coefficient in
this second stage can be interpreted as a causal effect of caf-
feine intake on BP. We ensured that the instrument was suffi-
ciently strong by checking that the F-value obtained in the
first-stage regression was .10 (59,60). For each association
of interest, we conducted both OLS regression and 2SLS re-
gression, using the ivregress function in Stata (Stata Corpor-
ation, College Station, TX, USA). We compared OLS and
2SLS estimates using the Durbin – Hausman test (61).
Statistic methods for HYPERGENES and Bus Sante´ are in
Supplementary Material. All analyses were conducted using
Stata, version 11.0 (StataCorp LP, College Station,
TX, USA). Statistical significances for association/trend
tests and interaction tests were set at P-value ,0.05 and
,0.10, respectively.
SUPPLEMENTARY MATERIAL
Supplementary Material is available at HMG online.
ACKNOWLEDGEMENTS
The CoLaus study: The authors express their gratitude to the
participants in the Lausanne CoLaus study and to the investi-
gators who have contributed to the recruitment, in particular
Yolande Barreau, Anne-Lise Bastian, Binasa Ramic, Martine
Moranville, Martine Baumer, Marcy Sagette, Jeanne
Ecoffey, and Sylvie Mermoud for data collection. HYPER-
GENES: Regarding the present work, cases and controls
were recruited within specific cohorts/networks: FLEMEN-
GHO/EPOGH (Coordinator J. Staessen); Wandsworth
Heart & Stroke Study (Coordinator F. Cappuccio); IMMI-
DIET (Coordinator L. Iacoviello); Milano-Sassari Cohort
(D. Cusi); SOPHIA (Coordinator N. Glorioso).
Conflict of Interest statement. None declared.
FUNDING
HYPERGENES is a large cooperative project funded by EU
within the FP7 (HEALTH-F4-2007-201550). The Bus Sante´
study is supported by the Geneva University Hospitals. I.G.,
M.P. and G.E. are supported by a Swiss National Science
Foundation grant (SNF 33CM30-124087/1). The CoLaus study
was supported by research grants from GlaxoSmithKline, the
Faculty of Biology and Medicine of Lausanne, Switzerland and
the Swiss National Science Foundation (33CSCO-122661). The
GenSmoke study was supported by the Swiss Federal Office of
Public Health – Tobacco Prevention Funding (06.004879).
M.Bo. is supported by the Swiss School of Public Health Plus
(SSPH+). The fundershad no role in study design, data collection
and analysis, decision to publish or in preparation of the manu-
script.
3290 Human Molecular Genetics, 2012, Vol. 21, No. 14
at University of Warwick on August 27, 2012http://hmg.oxfordjournals.org/Downloaded from

REFERENCES
1. Levy, D., Ehret, G.B., Rice, K., Verwoert, G.C., Launer, L.J., Dehghan,
A., Glazer, N.L., Morrison, A.C., Johnson, A.D., Aspelund, T. et al.
(2009) Genome-wide association study of blood pressure and
hypertension. Nat. Genet.,41, 677– 687.
2. Newton-Cheh, C., Johnson, T., Gateva, V., Tobin, M.D., Bochud, M.,
Coin, L., Najjar, S.S., Zhao, J.H., Heath, S.C., Eyheramendy, S. et al.
(2009) Genome-wide association study identifies eight loci associated
with blood pressure. Nat. Genet.,41, 666– 676.
3. Zhou, S.F., Chan, E., Zhou, Z.W., Xue, C.C., Lai, X. and Duan, W. (2009)
Insights into the structure, function, and regulation of human cytochrome
P450 1A2. Curr. Drug Metab.,10, 713–729.
4. Laitala, V.S., Kaprio, J. and Silventoinen, K. (2008) Genetics of coffee
consumption and its stability. Addiction,103, 2054– 2061.
5. Yang, A., Palmer, A.A. and de Wit, H. (2010) Genetics of caffeine
consumption and responses to caffeine. Psychopharmacology (Berl.),211,
245–257.
6. Cornelis, M.C., Monda, K.L., Yu, K., Paynter, N., Azzato, E.M., Bennett,
S.N., Berndt, S.I., Boerwinkle, E., Chanock, S., Chatterjee, N. et al.
(2011) Genome-wide meta-analysis identifies regions on 7p21 (AHR) and
15q24 (CYP1A2) as determinants of habitual caffeine consumption. PLoS
Genet.,7, e1002033.
7. Djordjevic, N., Ghotbi, R., Jankovic, S. and Aklillu, E. (2010) Induction
of CYP1A2 by heavy coffee consumption is associated with the
CYP1A2 -163C.A polymorphism. Eur. J. Clin. Pharmacol.,
66, 697– 703.
8. Noordzij, M., Uiterwaal, C.S., Arends, L.R., Kok, F.J., Grobbee, D.E. and
Geleijnse, J.M. (2005) Blood pressure response to chronic intake of coffee
and caffeine: a meta-analysis of randomized controlled trials.
J. Hypertension,23, 921– 928.
9. James, J.E. (2004) Critical review of dietary caffeine and blood pressure: a
relationship that should be taken more seriously. Psychosoc. Med.,66,
63– 71.
10. Robertson, D., Wade, D., Workman, R., Woosley, R.L. and Oates, J.A.
(1981) Tolerance to the humoral and hemodynamic effects of caffeine in
man. J. Clin. Invest.,67, 1111– 1117.
11. Klag, M.J., Wang, N.Y., Meoni, L.A., Brancati, F.L., Cooper, L.A., Liang,
K.Y., Young, J.H. and Ford, D.E. (2002) Coffee intake and risk of
hypertension: the Johns Hopkins precursors study. Arch Intern. Med.,162,
657–662.
12. Winkelmayer, W.C., Stampfer, M.J., Willett, W.C. and Curhan, G.C.
(2005) Habitual caffeine intake and the risk of hypertension in women.
JAMA,294, 2330– 2335.
13. Uiterwaal, C.S., Verschuren, W.M., Bueno-de-Mesquita, H.B., Ocke, M.,
Geleijnse, J.M., Boshuizen, H.C., Peeters, P.H., Feskens, E.J. and
Grobbee, D.E. (2007) Coffee intake and incidence of hypertension. Am. J.
Clin. Nutr.,85, 718– 723.
14. Bertrand, C.A., Pomper, I., Hillman, G., Duffy, J.C. and Micheli, I. (1978)
No relation between coffee and blood pressure. NEJM,299,315– 316.
15. Lang, T., Degoulet, P., Aime, F., Fouriaud, C., Jacquinet-Salord, M.C.,
Laprugne, J., Main, J., Oeconomos, J., Phalente, J. and Prades, A. (1983)
Relation between coffee drinking and blood pressure: analysis of 6,321
subjects in the Paris region. Am. J. Cardiol.,52, 1238–1242.
16. Stensvold, I., Tverdal, A. and Foss, O.P. (1989) The effect of coffee on
blood lipids and blood pressure. Results from a Norwegian cross-sectional
study, men and women, 40 – 42 years. J. Clin. Epidemiol,42, 877 – 884.
17. Zhou, S.F., Yang, L.P., Zhou, Z.W., Liu, Y.H. and Chan, E. (2009)
Insights into the substrate specificity, inhibitors, regulation, and
polymorphisms and the clinical impact of human cytochrome P450 1A2.
AAPS J.,11, 481– 494.
18. MacLeod, S., Sinha, R., Kadlubar, F.F. and Lang, N.P. (1997)
Polymorphisms of CYP1A1 and GSTM1 influence the in vivo function of
CYP1A2. Mutat. Res.,376, 135– 142.
19. Dobrinas, M., Cornuz, J., Oneda, B., Kohler Serra, M., Puhl, M. and Eap,
C.B. (2011) Impact of smoking, smoking cessation, and genetic
polymorphisms on CYP1A2 activity and inducibility. Clin. Pharmacol.
Ther.,90, 117– 125.
20. Cornelis, M.C., Monda, K.L., Yu, K., Paynter, N., Azzato, E.M., Bennett,
S.N., Berndt, S.I., Boerwinkle, E., Chanock, S., Chatterjee, N. et al.
(2011) Genome-wide meta-analysis identifies regions on 7p21 (AHR) and
15q24 (CYP1A2) as determinants of habitual caffeine consumption. PLoS
Genet.,7, e1002033.
21. Sulem, P., Gudbjartsson, D.F., Geller, F., Prokopenko, I., Feenstra, B.,
Aben, K.K., Franke, B., den Heijer, M., Kovacs, P., Stumvoll, M. et al.
(2011) Sequence variants at CYP1A1-CYP1A2 and AHR associate with
coffee consumption. Hum. Mol. Genet.,20, 2071–2077.
22. Larsson, S.C., Mannisto, S., Virtanen, M.J., Kontto, J., Albanes, D. and
Virtamo, J. (2008) Coffee and tea consumption and risk of stroke subtypes
in male smokers. Stroke,39, 1681–1687.
23. Lopez-Garcia, E., Rodriguez-Artalejo, F., Rexrode, K.M., Logroscino, G.,
Hu, F.B. and van Dam, R.M. (2009) Coffee consumption and risk of
stroke in women. Circulation,119, 1116– 1123.
24. Larsson, S.C., Virtamo, J. and Wolk, A. (2011) Coffee consumption and
risk of stroke in women. Stroke,42, 908–912.
25. Bidel, S., Hu, G., Qiao, Q., Jousilahti, P., Antikainen, R. and Tuomilehto,
J. (2006) Coffee consumption and risk of total and cardiovascular
mortality among patients with type 2 diabetes. Diabetologia,49,
2618–2626.
26. Grobbee, D.E., Rimm, E.B., Giovannucci, E., Colditz, G., Stampfer, M.
and Willett, W. (1990) Coffee, caffeine, and cardiovascular disease in
men. N. Eng. J. Med.,323, 1026–1032.
27. Greenberg, J.A., Dunbar, C.C., Schnoll, R., Kokolis, R., Kokolis, S. and
Kassotis, J. (2007) Caffeinated beverage intake and the risk of heart
disease mortality in the elderly: a prospective analysis. Am. J. Clin. Nutr.,
85, 392– 398.
28. Hakim, A.A., Ross, G.W., Curb, J.D., Rodriguez, B.L., Burchfiel, C.M.,
Sharp, D.S., Yano, K. and Abbott, R.D. (1998) Coffee consumption in
hypertensive men in older middle-age and the risk of stroke: the Honolulu
Heart Program. J. Clin. Epidemiol.,51,487– 494.
29. Sacco, R.L. (1997) Risk factors, outcomes, and stroke subtypes for
ischemic stroke. Neurology,49, S39– S44.
30. Gunes, A. and Dahl, M.L. (2008) Variation in CYP1A2 activity and its
clinical implications: influence of environmental factors and genetic
polymorphisms. Pharmacogenomics,9, 625– 637.
31. Ingelman-Sundberg, M., Sim, S.C., Gomez, A. and Rodriguez-Antona, C.
(2007) Influence of cytochrome P450 polymorphisms on drug therapies:
pharmacogenetic, pharmacoepigenetic and clinical aspects. Pharmacol.
Ther.,116, 496– 526.
32. Jiang, Z., Dragin, N., Jorge-Nebert, L.F., Martin, M.V., Guengerich, F.P.,
Aklillu, E., Ingelman-Sundberg, M., Hammons, G.J., Lyn-Cook, B.D.,
Kadlubar, F.F. et al. (2006) Search for an association between the human
CYP1A2 genotype and CYP1A2 metabolic phenotype. Pharmacogenet.
Genomics,16, 359– 367.
33. Cauchi, S., Stucker, I., Solas, C., Laurent-Puig, P., Cenee, S., Hemon, D.,
Jacquet, M., Kremers, P., Beaune, P. and Massaad-Massade, L. (2001)
Polymorphisms of human aryl hydrocarbon receptor (AhR) gene in a
French population: relationship with CYP1A1 inducibility and lung
cancer. Carcinogenesis,22, 1819– 1824.
34. Gomes, A.M., Winter, S., Klein, K., Turpeinen, M., Schaeffeler, E.,
Schwab, M. and Zanger, U.M. (2009) Pharmacogenomics of human liver
cytochrome P450 oxidoreductase: multifactorial analysis and impact on
microsomal drug oxidation. Pharmacogenomics,10, 579– 599.
35. Nussberger, J., Mooser, V., Maridor, G., Juillerat, L., Waeber, B. and
Brunner, H.R. (1990) Caffeine-induced diuresis and atrial natriuretic
peptides. J. Cardiovasc. Pharmacol.,15, 685– 691.
36. Passmore, A.P., Kondowe, G.B. and Johnston, G.D. (1987) Renal and
cardiovascular effects of caffeine: a dose-response study. Clin. Sci.
(Lond.),72, 749– 756.
37. Biaggioni, I., Paul, S., Puckett, A. and Arzubiaga, C. (1991) Caffeine and
theophylline as adenosine receptor antagonists in humans. J. Pharmacol.
Exp. Ther.,258, 588– 593.
38. Fredholm, B.B., Ijzerman, A.P., Jacobson, K.A., Klotz, K.N. and Linden,
J. (2001) International Union of Pharmacology. XXV. Nomenclature and
classification of adenosine receptors. Pharmacol. Rev.,53, 527–552.
39. Rieg, T., Steigele, H., Schnermann, J., Richter, K., Osswald, H. and
Vallon, V. (2005) Requirement of intact adenosine A1 receptors for the
diuretic and natriuretic action of the methylxanthines theophylline and
caffeine. J. Pharmacol. Exp. Ther.,313, 403–409.
40. Knight, R.J., Bowmer, C.J. and Yates, M.S. (1993) The diuretic action of
8-cyclopentyl-1,3-dipropylxanthine, a selective A1 adenosine receptor
antagonist. Br. J. Pharmacol.,109, 271– 277.
41. Wilcox, C.S., Welch, W.J., Schreiner, G.F. and Belardinelli, L. (1999)
Natriuretic and diuretic actions of a highly selective adenosine A1
receptor antagonist. J. Am. Soc. Nephrol.,10, 714– 720.
Human Molecular Genetics, 2012, Vol. 21, No. 14 3291
at University of Warwick on August 27, 2012http://hmg.oxfordjournals.org/Downloaded from

42. Brater, D.C., Kaojarern, S. and Chennavasin, P. (1983)
Pharmacodynamics of the diuretic effects of aminophylline and
acetazolamide alone and combined with furosemide in normal subjects.
J. Pharmacol. Exp. Ther.,227, 92–97.
43. van Buren, M., Bijlsma, J.A., Boer, P., van Rijn, H.J. and Koomans, H.A.
(1993) Natriuretic and hypotensive effect of adenosine-1 blockade in
essential hypertension. Hypertension,22, 728– 734.
44. Mancia, G., Laurent, S., Agabiti-Rosei, E., Ambrosioni, E., Burnier, M.,
Caulfield, M.J., Cifkova, R., Clement, D., Coca, A., Dominiczak, A. et al.
(2009) Reappraisal of European guidelines on hypertension management:
a European Society of Hypertension Task Force document. Blood Press.,
18, 308– 347.
45. Chobanian, A.V., Bakris, G.L., Black, H.R., Cushman, W.C., Green, L.A.,
Izzo, J.L. Jr, Jones, D.W., Materson, B.J., Oparil, S., Wright, J.T. Jr et al.
(2003) Seventh report of the Joint National Committee on Prevention,
Detection, Evaluation, and Treatment of High Blood Pressure.
Hypertension,42, 1206– 1252.
46. Goldstein, L.B., Bushnell, C.D., Adams, R.J., Appel, L.J., Braun, L.T.,
Chaturvedi, S., Creager, M.A., Culebras, A., Eckel, R.H., Hart, R.G. et al.
(2011) Guidelines for the primary prevention of stroke: a guideline for
healthcare professionals from the American Heart Association/American
Stroke Association. Stroke,42, 517– 584.
47. Zhou, S.F., Wang, B., Yang, L.P. and Liu, J.P. (2010) Structure, function,
regulation and polymorphism and the clinical significance of human
cytochrome P450 1A2. Drug Metab. Rev.,42, 268–354.
48. Djordjevic, N., Ghotbi, R., Bertilsson, L., Jankovic, S. and Aklillu, E.
(2008) Induction of CYP1A2 by heavy coffee consumption in Serbs and
Swedes. Eur. J. Clin. Pharmacol.,64, 381– 385.
49. Faber, M.S., Jetter, A. and Fuhr, U. (2005) Assessment of CYP1A2
activity in clinical practice: why, how, and when? Basic Clin. Pharmacol.
Toxicol.,97, 125– 134.
50. Cornelis, M.C., El-Sohemy, A., Kabagambe, E.K. and Campos, H. (2006)
Coffee, CYP1A2 genotype, and risk of myocardial infarction. JAMA,295,
1135– 1141.
51. Bochud, M., Chiolero, A., Elston, R.C. and Paccaud, F. (2008) A
cautionary note on the use of Mendelian randomization to infer causation
in observational epidemiology. Int. J. Epidemiol.,37, 414– 416; author
reply 416–417.
52. Firmann, M., Mayor, V., Vidal, P.M., Bochud, M., Pecoud, A., Hayoz, D.,
Paccaud, F., Preisig, M., Song, K.S., Yuan, X. et al. (2008) The CoLaus
study: a population-based study to investigate the epidemiology and
genetic determinants of cardiovascular risk factors and metabolic
syndrome. BMC Cardiovasc. Disord.,8,6.
53. El Assaad, M.A., Topouchian, J.A., Darne, B.M. and Asmar, R.G. (2002)
Validation of the Omron HEM-907 device for blood pressure
measurement. Blood Press. Monit.,7, 237– 241.
54. Sachse, C., Brockmoller, J., Bauer, S. and Roots, I. (1999) Functional
significance of a C A polymorphism in intron 1 of the cytochrome
P450 CYP1A2 gene tested with caffeine. Br. J. Clin. Pharmacol.,47,
445–449.
55. Faber, M.S. and Fuhr, U. (2004) Time response of cytochrome P450 1A2
activity on cessation of heavy smoking. Clin. Pharmacol. Ther.,76,
178–184.
56. Eap, C.B., Bender, S., Jaquenoud Sirot, E., Cucchia, G., Jonzier-Perey,
M., Baumann, P., Allorge, D. and Broly, F. (2004) Nonresponse to
clozapine and ultrarapid CYP1A2 activity: clinical data and analysis of
CYP1A2 gene. J. Clin. Psychopharmacol.,24, 214– 219.
57. Fuhr, U. and Rost, K.L. (1994) Simple and reliable CYP1A2 phenotyping
by the paraxanthine/caffeine ratio in plasma and in saliva.
Pharmacogenetics,4, 109– 116.
58. Tobin, M.D., Sheehan, N.A., Scurrah, K.J. and Burton, P.R. (2005)
Adjusting for treatment effects in studies of quantitative traits:
antihypertensive therapy and systolic blood pressure. Stat. Med.,24,
2911– 2935.
59. Bochud, M. (2008) On the use of Mendelian randomization to infer
causality in observational epidemiology. Eur. Heart J.,29, 2456–2457.
60. Bochud, M. and Rousson, V. (2010) Usefulness of Mendelian
randomization in observational epidemiology. Int. J. Environ. Res. Public
Health,7, 711– 728.
61. Lawlor, D.A., Harbord, R.M., Sterne, J.A., Timpson, N. and Davey Smith,
G. (2008) Mendelian randomization: using genes as instruments for
making causal inferences in epidemiology. Stat. Med.,27, 1133–1163.
3292 Human Molecular Genetics, 2012, Vol. 21, No. 14
at University of Warwick on August 27, 2012http://hmg.oxfordjournals.org/Downloaded from