Ci cula ing B ain-De i ed Neu o ophic Fac o Concen a ions and he
Risk o Ca dio ascula Disease in he Communi y
Be nha d M. Kaess, MD; Sa ah R. P eis, ScD; Wol gang Lieb, MD, MSc; Alexa S. Beise , PhD; Qiong Yang, PhD; Tai C. Chen, PhD;
Ch is ian Hengs enbe g, MD; Jeane e E dmann, PhD; He ibe Schunke , MD; Sudha Seshad i, MD; Ramachand an S. Vasan, MD;
on behal o CARDIoGRAM*
Backg ound-—B ain-de i ed neu o ophic ac o (BDNF) is a pleio opic pep ide in ol ed in main aining endo helial in eg i y. I is
unknown i ci cula ing BDNF le els a e associa ed wi h isk o ca dio ascula disease (CVD).
Me hods and Resul s-—We p ospec i ely in es iga ed he associa ion o ci cula ing BDNF le els wi h ca dio ascula e en s and
mo ali y in 3687 pa icipan s (mean age 65 yea s, 2068 women) om he F amingham Hea S udy (FHS). Using a common
nonsynonomous single nucleo ide polymo phism (SNP) in he BDNF gene ( s6265), we hen pe o med a Mendelian andomiza ion
expe imen in he CARDIoGRAM (Co ona y AR e y DIsease Genome-Wide Replica ion And Me a-Analysis) conso ium (>22 000
co ona y a e y disease [CAD] cases, >60 000 con ols) o in es iga e whe he SNP s6265 was associa ed wi h CAD in
CARDIoGRAM and, i so, whe he he e ec es ima e di e ed om ha p edic ed based on FHS da a. On ollow-up (median
8.9 yea s), 467 indi iduals (261 women) in FHS expe ienced a CVD e en , and 835 (430 women) died. In mul i a iable-adjus ed
Cox eg ession, se um BDNF was associa ed in e sely wi h CVD isk (haza d a io [HR] pe 1-SD inc ease 0.88, 95% CI 0.80 o
0.97, P=0.01) and wi h mo ali y (HR 0.87, 95% CI 0.80 o 0.93, P=0.0002). SNP s6265 was associa ed wi h BDNF concen a ions
(0.772 ng/mL inc ease pe mino allele copy) in FHS. In CARDIoGRAM, SNP s6265 was associa ed wi h CAD (odds a io 0.957,
95% CI 0.923 o 0.992), a magni ude consis en wi h he p edic ed e ec (HR pe mino allele copy 0.99, 95% CI 0.98 o 1.0;
P=0.06 o di e ence be ween p edic ed and obse ed e ec ).
Conclusion-—Highe se um BDNF is associa ed wi h a dec eased isk o CVD and mo ali y. Mendelian andomiza ion sugges s a causal
p o ec i e ole o BDNF in he pa hogenesis o CVD. (J Am Hea Assoc. 2015;4:e001544 doi: 10.1161/JAHA.114.001544)
Key Wo ds: ca dio ascula disease •g ow h ac o s •Mendelian andomiza ion •mo ali y • isk ac o s
Ca dio ascula disease (CVD) is he leading cause o
dea h in he de eloped coun ies, and clinical isk ac o s
o CVD (eg, obesi y, dyslipidemia, diabe es, and a seden a y
li es yle) ha e been known o decades. Ne e heless, he
molecula basis o CVD is complex and linked o a b oad ange
o biological pa hways, including lipid and glucose me abolism,
inflamma ion, ascula epai , and angiogenesis. B ain-de i ed
neu o ophic ac o (BDNF) is a pleio opic pep ide media o
in ol ed in he egula ion o appe i e and physical ac i i y and in
neu oplas ici y.
1
He e ozygous BDNF knockou mice consume
almos 50% mo e ood han do hei wild- ype li e ma es and
a e obese.
2
Independen o i s e ec on ene gy homeos asis,
BDNF has been implica ed in angiogenesis and he main e-
nance o ascula in eg i y in ecen epo s.
3
Thus, condi ional
BDNF knockou mice ha e g ea e myoca dial damage a e
expe imen al in a c ion compa ed wi h wild- ype mice.
4
These
F om he Na ional Hea , Blood, and Lung Ins i u e’s F amingham Hea S udy, F amingham, MA (B.M.K., S.R.P., A.S.B., Q.Y., S.S., R.S.V.); Deu sches He zzen um M€
unchen,
Technische Uni e si €
a M€
unchen, Ge many (B.M.K., C.H., H.S.); DZHK (Ge manCen e o Ca dio ascula Resea ch), pa ne si e Munich Hea Alliance, M€
unchen, Ge many
(B.M.K., C.H., H.S.); Depa men o Bios a is ics, Bos on Uni e si y School o Public Heal h, Bos on, MA (S.R.P., A.S.B., Q.Y.); Ins i u e o Epidemiology, Ch is ian-Alb ech s-
Uni e si y, Kiel, Ge many (W.L.); Depa men o Neu ology (A.S.B.,S.S.), Sec ions o Endoc inology, Diabe es, and Nu i ion (T.C.C.)and Sec ions o P e en i e Medicine and
Epidemiology and Ca diology (R.S.V.), Bos on Uni e si y School o Medicine, Bos on, MA; Ins i u €
u In eg a i e und Expe imen elle Genomik, Uni e si €
a zu L€
ubeck,
Ge many (J.E.); DZHK (Ge man Resea ch Cen e o Ca dio ascula Resea ch), pa ne si e Hambu g/L€
ubeck/Kiel, L€
ubeck, Ge many (J.E.).
*An accompanying Appendix S1, which lis he membe s o he CARDIoGRAM Conso ium is a ailable a h p://jaha.ahajou nals.o g/con en /4/3/e001544/
suppl/DC1
Co espondence o: Ramachand an S. Vasan, MD, F amingham Hea S udy, 73 M Way e A e, Sui e 2, F amingham, MA 01702-5803. E-mail: [email protected]
Be nha d M. Kaess, MD, Deu sches He zzen um M€
unchen, Laza e s . 36, 80363 Munich, Ge many. E-mail: [email p o ec ed]
Recei ed No embe 5, 2014; accep ed Janua y 30, 2015.
ª2015 The Au ho s. Published on behal o he Ame ican Hea Associa ion, Inc., by Wiley Blackwell. This is an open access a icle unde he e ms o he C ea i e
Commons A ibu ion-NonComme cial License, which pe mi s use, dis ibu ion and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed and is
no used o comme cial pu poses.
DOI: 10.1161/JAHA.114.001544 Jou nal o he Ame ican Hea Associa ion 1
ORIGINAL RESEARCH
obse a ions aise he possibili y ha BDNF may play a ole in
he pa hogenesis o CVD.
Indeed, limi ed epidemiological da a sugges ha highe
ci cula ing BDNF le els may be associa ed wi h a lowe
p e alence o ca dio ascula isk ac o s and wi h a lowe
mo ali y in small samples.
5,6
Con e sely, pa ien s wi h
uns able angina ha e been epo ed o ha e inc eased BDNF
le els in he co ona y ci cula ion compa ed wi h indi iduals
wi h s able angina, sugges ing ha BDNF may de imen ally
influence plaque s abili y.
7
Taken oge he , epidemiological
da a on he ela ion be ween BDNF and CVD a e spa se and
conflic ing. Thus, we p ospec i ely in es iga ed he associa-
ion o ci cula ing BDNF le els wi h ca dio ascula e en s and
mo ali y in he communi y-based F amingham Hea S udy
(FHS) coho . To assess a po en ially causal ela ion be ween
BDNF le els and CVD, we pe o med a Mendelian andomi-
za ion analysis by using he CARDIoGRAM (Co ona y AR e y
DIsease Genome-Wide Replica ion And Me a-Analysis) me a-
sample.
Me hods
S udy Samples
F amingham Hea S udy (FHS)
The FHS sample o he p esen in es iga ion was de i ed
om he O iginal and O sp ing coho s
8,9
and is based on
he 23 d FHS O iginal coho examina ion (1992–1996) and
he 7 h O sp ing coho examina ion (1998–2001). The FHS
s a ed in 1948 wi h he ec ui men o 5209 indi iduals
om he gene al popula ion esiding in he F amingham, MA,
a ea, e e ed o as he O iginal coho . Pa icipan s in his
coho a e e alua ed a he FHS clinic app oxima ely e e y
2 yea s, including a de ailed medical his o y, physical
examina ion, blood p essu e assessmen , ECG, and phlebo -
omy o he measu emen o ascula isk ac o s. The
F amingham O sp ing s udy began in 1971 wi h he
ec ui men o 5124 indi iduals who we e he o sp ing o
he O iginal Coho o he o sp ings’spouses.
9
Follow-up
examina ions in he s udy clinic a e pe o med app oxima ely
once e e y 4 yea s.
O 1026 O iginal coho pa icipan s a ending examina ion
cycle 23, we excluded 357 who did no ha e a ailable BDNF
measu emen s (>225 pa icipan s had isi s a home and
blood was no d awn; addi ionally, se um samples d awn in
he fi s 4 mon hs o he examina ion cycle we e no a ailable
o his assay due o s o age- ela ed p oblems). O 3539
O sp ing coho pa icipan s a ending examina ion cycle 7,
we excluded 3 pa icipan s o lack o ollow-up da a and an
addi ional 518 who did no ha e a ailable BDNF measu e-
men s (due o lack o su ficien amoun s o se um aliquo s),
esul ing in a final sample size o 3687 (669 O iginal coho ,
3018 O sp ing coho pa icipan s).
Physical ac i i y was assessed (in he O sp ing coho
only) by using a s uc u ed ques ionnai e and hen quan ified
as a physical ac i i y index, which is calcula ed based on
weigh ed equencies and du a ion o he epo ed ac i i ies
as de ailed p e iously.
10
Dep ession was assessed (in he
O sp ing coho only) using he Cen e o Epidemiological
S udies–Dep ession Scale (CESD), which is an accep ed ool
o sc een o dep essi e symp oms in obse a ional coho
se ings.
11
The s udy was app o ed by he Bos on Uni e si y Medical
Cen e Ins i u ional Re iew Boa d; w i en in o med consen
was ob ained om all pa icipan s.
Bioma ke measu emen s in he FHS coho
Se um BDNF le els we e de e mined by using Quan ikine
ELISA ki s ob ained om R&D Sys ems a e a 20- old dilu ion
in polyp opylene ubes wi h Calib a o Diluen RD6P included
in he ki s. The sensi i i y o he assay is es ima ed o be a o
below 0.0625 ng/mL. The in a- and in e -assay coe ficien s
o a ia ion we e 3.8% o 6.2% and 7.6% o 11.3%, espec-
i ely. High-sensi i y C- eac i e p o ein (hsCRP) was mea-
su ed (only in he O sp ing coho ) by using a nephelome e
(Dade Beh ing).
12
B ain na iu e ic pep ide (BNP) was mea-
su ed only in he O sp ing coho a examina ion cycle 6 (no
7), by using an immuno adiome ic assay (Shionogi) as
p e iously desc ibed.
13
Geno yping in FHS
Single nucleo ide polymo phism (SNP) s6265 was di ec ly
geno yped on he Illumina Beads a ion 500G geno yping
sys em by using Illumina Fas T ack Se ices and he Golden
Ga e assay wo kflow p o ocol. The sample success a e was
99.44%, locus success a e was 97.4%, geno ype call a e
was 99.72%, ep oducibili y was 100%, and Mendelian
consis ency 99.99%. Mino allele equency o s6265
was 0.19 in FHS. Geno ypes a e a ailable a h p://
www.ncbi.nlm.nih.go /gap.
Ou come e en s in FHS
A“ca dio ascula e en ”was defined as angina pec o is,
co ona y insu ficiency (p olonged angina wi h documen ed
ECG changes), myoca dial in a c ion, s oke (ischemic o
hemo hagic) o ansien ischemic a ack, inciden hea
ailu e, in e mi en claudica ion, o dea h seconda y o
CVD. All ca dio ascula e en s we e adjudica ed by a panel
o 2 o 3 in es iga o s on e iew o hospi al eco ds,
medical o fice no es, and F amingham clinic isi no es, by
using s anda dized c i e ia ha ha e been desc ibed
p e iously.
14
DOI: 10.1161/JAHA.114.001544 Jou nal o he Ame ican Hea Associa ion 2
Ci cula ing BDNF and Ca dio ascula Disease Kaess e al
ORIGINAL RESEARCH
Mendelian andomiza ion in he CARDIoGRAM me a-
sample
We assessed he associa ion o he majo unc ional gene ic
a ian wi hin he BDNF gene ( s6265) wi h co ona y a e y
disease (CAD) in he CARDIoGRAM conso ium.
15
CARDIo-
GRAM is a me a-analysis o genome-wide associa ion da a o
CAD including myoca dial in a c ion and consis s o 14
popula ion-based coho s and case-con ol samples, including
in o al >22 000 cases and >60 000 con ols.
15
We chose
CARDIoGRAM o ou Mendelian andomiza ion expe imen
because i is he la ges conso ium o CAD ou comes wi h
a ailabili y o gene ic da a o he BDNF locus, he eby
maximizing ou s a is ical powe . We a e unawa e o la ge
conso ia ha ha e e alua ed a composi e ou come o CVD
(pa alleling ou p ima y ou come in FHS) o conso ia ha
ha e assessed all-cause mo ali y in a genome-wide con ex
wi h a ailabili y o BDNF SNP da a. Mino allele equency o
SNP s6265 was 0.18 in CARDIoGRAM.
S a is ical Analyses
BDNF le els we e no mally dis ibu ed. Age- and sex-adjus ed
incidence a es o CVD and all-cause mo ali y ac oss qua iles
o se um BDNF we e calcula ed by using di ec s anda diza ion
by sex and age s a a o <60, 60 o 74 and ≥75 yea s. We hen
es ima ed Cox p opo ional haza ds eg ession models o
u he explo e he ela ion o BDNF wi h CVD incidence and all-
cause mo ali y a e confi ming he assump ion o p opo ion-
ali y o haza ds o bo h ou comes. We assessed 2 models: (1)
adjus ing only o age and sex and (2) addi ionally adjus ing o
s anda d CVD isk ac o s
16
(ie, age, sex, smoking, sys olic
blood p essu e, hype ension ea men , o al and high-densi y
lipop o ein choles e ol, diabe es melli us, and body mass index
[BMI]). All models we e s a ified by s udy coho (O iginal
e sus O sp ing). We adjus ed o BMI (al hough i has been an
inconsis en componen o CVD isk p edic ion algo i hms)
because o he po en ial associa ion o BDNF wi h BMI.
17
Using
in e ac ion e ms, we assessed po en ial e ec modifica ion by
age, BMI, o s udy coho in ou mul i a iable models. Howe e ,
hese e ms we e s a is ically nonsignifican (all P>0.05 o
bo h ou comes). To u he explo e any po en ial nonlinea
ela ion o BDNF le els wi h CVD e en s and mo ali y, we
cons uc ed mul i a iable-adjus ed (co a ia es o model 2)
nonpa ame ic es ic ed cubic splines (wi h 3 kno s a he
qua iles o he BDNF dis ibu ion).
18
To u he assess a
po en ial ole o smoking, we adjus ed ou mul i a iable models
o li e ime smoking exposu e in pack-yea s (ins ead o cu en
smoking s a us). We also in es iga ed whe he cu en smoking
modifies he ela ion be ween se um BDNF and e en s by
in oducing an in e ac ion e m (smoking9BDNF) in o ou
mul i a iable models.
In explo a o y analyses, we addi ionally adjus ed ou Cox
models o o he no el CVD bioma ke s (ie, hsCRP and
BNP) and o physical ac i i y (by using he physical ac i i y
index) and dep ession (using he CESD sco e). These
analyses we e limi ed o he O sp ing coho because
hese a iables we e measu ed only in his coho . All
analyses we e pe o med by using SAS e sion 9.2 (SAS
Ins i u e). A P alue o <0.05 was conside ed s a is ically
significan .
To assess he con ibu ion o BDNF le els o CVD isk
p edic ion, we compa ed pe o mance me ics o a model
inco po a ing s anda d CVD isk ac o s (mul i a iable Cox
eg ession model 2) o a model ha addi ionally included
se um BDNF. We assessed inc emen in he c-s a is ic, he
in eg a ed disc imina ion imp o emen (IDI), and he con in-
uous ne eclassifica ion imp o emen (NRI), all calcula ed as
p e iously desc ibed.
19,20
Table 1. Cha ac e is ics o he S udy Sample
Cha ac e is ics To al (N=3687)
Age, y 6511
Women, n (%) 2068 (56.1)
Body mass index, kg/m
2
27.95.2
Sys olic blood p essu e, mm Hg 13020
Dias olic blood p essu e, mm Hg 7310
An ihype ensi e medica ion, n (%) 1354 (36.8)
To al choles e ol, mg/dL 20137
HDL choles e ol, mg/dL 5317
S a in medica ion, n (%) 618 (16.8)
Cu en smoke s, n (%) 422 (11.5)
Type 2 diabe es, n (%) 421 (11.5)
P e alen ca dio ascula disease, n (%) 586 (15.9)
Physical ac i i y index* 37.76.4
CESD sco e* 5.36.7
High sensi i i y CRP, mg/L, median
(in e qua ile ange)*
2.0 (1.0, 4.9)
BNP, ng/L, median (in e qua ile ange)* 7.7 (4.0, 16.5)
Se um BDNF, ng/mL 23.5 (8.3)
Qua ile 1, mean ( ange) 1.7–17.8
Qua ile 2, mean ( ange) 17.9–23.2
Qua ile 3, mean ( ange) 23.2–28.8
Qua ile 4, mean ( ange) 28.8–64.9
Values a e meanSD unless indica ed o he wise. HDL indica es high-densi y lipop o ein;
CESD, Cen e o Epidemiological S udy–Dep ession Scale; CRP, C- eac i e p o ein; BNP,
B- ype na iu e ic pep ide; BDNF indica es b ain-de i ed neu o ophic ac o .
*O sp ing coho only (n=3018).
DOI: 10.1161/JAHA.114.001544 Jou nal o he Ame ican Hea Associa ion 3
Ci cula ing BDNF and Ca dio ascula Disease Kaess e al
ORIGINAL RESEARCH
Mendelian andomiza ion expe imen
Mendelian andomiza ion is based on he ac ha alleles o
gene ic a ian s (he e, s6265, a known unc ional a ian in
exon 2 o he BDNF gene) a e ansmi ed andomly du ing
concep ion om he pa en s o hei o sp ing.
21
Because
s6265 al e s BDNF le els, indi iduals a e andomly assigned
o gene ically ele a ed o educed BDNF le els.
21,22
Impo -
an ly, he andom alloca ion o alleles du ing meiosis is
independen o adi ional CVD isk ac o s and he CVD
disease p ocess i sel . The e o e, es ima es om Mendelian
andomiza ion s udies a e insensi i e o (nongene ic) con-
ounde s and e e se causali y.
23
Thus, i lowe BDNF le els
cause CVD, one can es ima e he p edic ed ela i e CVD isk
inc ease ha goes along wi h gene ically educed BDNF
le els, based on (1) he associa ion o s6265 wi h ci cula ing
BDNF le els and (2) he associa ion o ci cula ing BDNF le els
wi h inciden CVD. Acco dingly, fi s we ela ed he SNP
s6265 o ci cula ing BDNF le els by using a linea eg ession
model (adjus ed o age and sex) assuming an addi i e
gene ic model. Nex , based on he associa ion o he SNP
wi h BDNF le els, and in u n he associa ion be ween
ci cula ing BDNF le els and inciden CVD, we es ima ed he
p edic ed s eng h o associa ion be ween SNP s6265 and
CVD in age- and sex-adjus ed Cox eg ession models. Thi d,
we e alua ed he s eng h o associa ion o he SNP s6265
wi h CAD in he CARDIoGRAM conso ium.
15
These analyses
we e again pe o med adjus ing o age and sex (p io o
me a-analysis). Las , we compa ed he p edic ed isk o CVD
(based on FHS da a) o his SNP wi h he obse ed ela i e
isk o CAD in CARDIoGRAM by using he 2-sample es .
Resul s
S udy Sample
The cha ac e is ics o ou s udy sample a e p o ided in
Table 1. Ou sample was middle-aged o elde ly, wi h a
mode a e ca dio ascula isk p ofile. Clinical cha ac e is ics,
s a ified by la e CVD e en s a us, a e lis ed in Table 2.
Pa icipan s wi h CVD e en s we e olde and had a mo e
ad e se ca dio ascula isk p ofile han hose wi hou a
u u e e en . The age- and sex-adjus ed co ela ions o se um
BDNF le els wi h clinical ca dio ascula isk ac o s a e gi en
in Table 3.
Table 2. Baseline Cha ac e is ics o he S udy Sample,
S a ified by Inciden CVD S a us Du ing Follow-up
Cha ac e is ic a Baseline
Inciden CVD Du ing Follow-up
No (n=2574) Yes (n=467)
Age, y 6111 7210
Body mass index, kg/m
2
27.85.3 28.25.3
Sys olic blood p essu e, mm Hg 12719 13920
Dias olic blood p essu e, mm Hg 7410 7310
An ihype ensi e medica ion, n (%) 724 (28) 222 (47)
To al choles e ol, mg/dL 20336 20337
HDL choles e ol, mg/dL 5517 5116
S a in medica ion, n (%) 313 (12) 79 (17)
Cu en smoke s, n (%) 296 (12) 59 (13)
Type 2 diabe es, n (%) 207 (8) 71 (15)
Values a e meanSD unless indica ed o he wise. CVD indica es ca dio ascula disease;
HDL, high-densi y lipop o ein.
Table 3. Co ela ion Be ween Se um BDNF and
Ca dio ascula Risk Fac o s
T ai Co ela ion
Body mass index 0.011
Sys olic blood p essu e 0.013
Dias olic blood p essu e 0.048*
To al choles e ol 0.086
†
HDL choles e ol 0.006
Log CRP
‡
0.002
Log BNP
‡
0.062
†
Da a a e age- and sex-adjus ed pa ial Pea son co ela ion coe ficien s. BDNF indica es
b ain-de i ed neu o ophic ac o ; HDL, high-densi y lipop o ein; CRP, C- eac i e p o ein;
BNP, b ain na iu e ic pep ide.
*P<0.01,
†
P<0.001.
‡
O sp ing coho only (n=3018).
Table 4. BDNF, Ca dio ascula E en s, and Mo ali y—E en
Ra es by BDNF Qua iles
BDNF
No. o E en s/
No. a Risk
Pe son-yea s
Follow-up
Incidence Ra e pe 1000
Pe son-yea s (SE)
Ca dio ascula e en s
Qua ile 1 133/737 5471 26.0 (2.3)
Qua ile 2 129/746 5901 22.7 (2.1)
Qua ile 3 109/757 6215 20.4 (2.0)
Qua ile 4 96/801 6419 20.3 (2.2)
All-cause mo ali y
Qua ile 1 257/921 7676 31.0 (1.9)
Qua ile 2 237/922 8153 26.3 (1.7)
Qua ile 3 182/922 8286 23.0 (1.7)
Qua ile 4 159/922 8195 23.7 (1.9)
BDNF indica es b ain-de i ed neu o ophic ac o .
S anda dized o he sex-and age-g oup (<60, 60–74, >=75 yea s) dis ibu ion o he
s udy sample.
DOI: 10.1161/JAHA.114.001544 Jou nal o he Ame ican Hea Associa ion 4
Ci cula ing BDNF and Ca dio ascula Disease Kaess e al
ORIGINAL RESEARCH
Se um BDNF and Incidence o CVD E en s
The age- and sex-adjus ed incidence a es o ca dio ascula
e en s du ing a median ollow-up o 8.5 yea s, s a ified by
qua iles o BDNF, a e p o ided in Table 4 (uppe hal ). CVD
incidence dec eased mono onically ac oss qua iles o
BDNF. The esul s o he p opo ional haza ds eg ession
models ela ing BDNF le els o CVD incidence a e shown in
Table 5 (uppe hal ) and Figu e 1 ( op panel). Each 1-SD
inc ease in se um BDNF was associa ed wi h a 9% o 12%
educed haza d o CVD e en s. Subjec s in he highes
qua ile o BDNF had a 27% o 32% lowe isk o
ca dio ascula e en s compa ed wi h hose in he lowes
BDNF qua ile. Explo a o y analyses addi ionally adjus ing
o hsCRP and BNP showed simila esul s (Table 6). Also,
addi ional adjus men o physical ac i i y index and dep es-
si e symp oms (using a CESD sco e ≥16 o define
dep ession) did no al e ou esul s (haza d a io o
inciden CVD pe 1-SD inc ease in BDNF 0.86, 95% CI 0.76
o 0.97; P=0.01). To u he assess he ole o smoking
beha io , we adjus ed ou mul i a iable models o li e ime
smoking exposu e in pack-yea s, which did no influence ou
findings (da a no shown). Simila ly, e ec es ima es did no
di e be ween smoke s and nonsmoke s (P>0.05 o
in e ac ion; da a no shown).
In mul i a iable-adjus ed splines (Figu e 2, op panel),
se um BDNF was linea ly and in e sely associa ed wi h CVD
isk.
Se um BDNF and All-Cause Mo ali y
The age- and sex-adjus ed incidence a es o all-cause
mo ali y du ing a median ollow-up o 8.9 yea s, s a ified by
qua iles o he BDNF dis ibu ion, a e shown in Table 4 (lowe
hal ). The esul s o he Cox eg ession models a e displayed
in Table 5 (lowe hal ) and Figu e 1 (bo om panel). Each 1-SD
inc ease in BDNF was associa ed wi h a 13% o 14% educed
haza d o dea h. Indi iduals in he highes qua ile o BDNF
had a 26% o 29% lowe isk o dea h compa ed wi h hose in
Table 5. Se um BDNF, Ca dio ascula E en s, and Mo ali y—Cox Reg ession Models
Ca dio ascula E en s*
Age- and Sex-Adjus ed Model Mul i a iable Model
†
Haza d Ra io (95% CI) PValue Haza d Ra io (95% CI) PValue
Con inuous ai
1-SD inc ease 0.91 (0.82 o 1.00) 0.045 0.88 (0.80 o 0.97) 0.01
Qua iles
Qua ile 1 1.00 Re e en 1.00 Re e en
Qua ile 2 0.90 (0.70 o 1.15) 0.39 0.92 (0.72 o 1.17) 0.48
Qua ile 3 0.78 (0.60 o 1.00) 0.049 0.73 (0.56 o 0.95) 0.02
Qua ile 4 0.73 (0.56 o 0.95) 0.02 0.68 (0.52 o 0.89) 0.005
T end ac oss qua iles 0.01 0.001
All-Cause Mo ali y
Age- and Sex-Adjus ed Model Mul i a iable Model
‡
Haza d Ra io (95% CI) PValue Haza d Ra io (95% CI) PValue
Con inuous ai
1-SD inc ease 0.86 (0.79 o 0.92) <0.0001 0.87 (0.80 o 0.93) 0.0002
Qua iles
Qua ile 1 1.00 Re e en 1.00 Re e en
Qua ile 2 0.83 (0.69 o 0.99) 0.03 0.85 (0.71 o 1.02) 0.08
Qua ile 3 0.69 (0.57 o 0.84) 0.0002 0.71 (0.58 o 0.86) 0.0006
Qua ile 4 0.71 (0.59 o 0.87) 0.0009 0.74 (0.60 o 0.91) 0.004
T end ac oss qua iles <0.0001 0.0005
BDNF indica es b ain-de i ed neu o ophic ac o , ca dio ascula disease.
*Models de i ed om 3041 indi iduals wi hou CVD a baseline.
†
Adjus ed o age, sex, smoking, sys olic blood p essu e, hype ension ea men , o al and high-densi y lipop o ein choles e ol, diabe es melli us, and body mass index and s a ified by
coho s a us.
‡
Adjus ed o age, sex, smoking, sys olic blood p essu e, hype ension ea men , o al and high-densi y lipop o ein choles e ol, diabe es melli us, body mass index, and CVD a baseline
and s a ified by coho s a us.
DOI: 10.1161/JAHA.114.001544 Jou nal o he Ame ican Hea Associa ion 5
Ci cula ing BDNF and Ca dio ascula Disease Kaess e al
ORIGINAL RESEARCH

he lowes qua ile. Ou findings emained obus in models
addi ionally adjus ing o hsCRP and BNP (Table 6) and o
physical ac i i y index and he CESD sco e (haza d a io o
dea h pe 1-SD inc ease in BDNF=0.87, 95% CI 0.77 o 0.97;
P=0.02). Again, adjus men o li e- ime smoking exposu e did
no al e findings (da a no shown), no did e ec es ima es
di e be ween smoke s and nonsmoke s (P>0.05 o in e ac-
ion).
Inc easing se um BDNF showed a linea ela ionship wi h
lowe all-cause mo ali y h ough mos o he dis ibu ion o
alues, al hough he ela ionship pla eaued a he uppe end
o he dis ibu ion (Figu e 2, bo om panel).
BDNF and CVD Risk P edic ion
BDNF did no significan ly inc ease he c-s a is ic o ou
s anda d isk ac o model o ca dio ascula e en s (a ea
unde he cu e 0.756 e sus 0.759, P=0.99). Howe e , bo h
IDI (0.004, 95% CI 0.001 o 0.006, P<0.05) and ca ego y- ee
NRI (0.198, 95% CI 0.049 o 0.352, P<0.05) we e s a is ically
significan . Simila ly, se um BDNF did no imp o e he
c-s a is ic o ou s anda d isk ac o model o mo ali y
(a ea unde he cu e 0.793 e sus 0.810, P=0.97); howe e ,
IDI (0.005, 95% CI 0.002 o 0.008, P<0.05) and ca ego y- ee
NRI (0.225, 95% CI 0.120 o 0.325, P<0.05) we e s a is ically
significan .
Mendelian Randomiza ion
To assess whe he he associa ion be ween ci cula ing
BDNF le els and CVD may be causal, we used a known
unc ional gene ic a ian in he BDNF gene ( s6265) and
pe o med a Mendelian andomiza ion expe imen as
de ailed ea lie .
21
We geno yped s6265 in 3089 F amingham pa icipan s
and assessed i s associa ion wi h ci cula ing BDNF le els. By
assuming an addi i e gene ic model and adjus ing o age
and sex, each copy o he mino T allele was associa ed wi h
0.772 ng/mL highe BDNF le els (95% CI 0.242 o 1.303,
P=0.0043; Figu e 3). Based on he e ec es ima e o he
associa ion be ween ci cula ing BDNF le els and inciden
CVD epo ed ea lie he e (Table 5), an inc ease o
0.772 ng/mL in BDNF le els ansla es in o a p edic ed
haza d a io o CVD o 0.991 (95% CI 0.982 o 1.000,
P=0.045) pe T allele. The ac ual obse ed age- and sex-
adjus ed odds a io o CAD associa ed wi h 1 T allele o
s6265 in CARDIoGRAM was 0.957 (0.923 o 0.992,
P=0.016) and did no di e significan ly om he p edic ed
e ec size de i ed om he F amingham coho (P=0.062 o
di e ence be ween obse ed and p edic ed e ec size;
Figu e 3), sugges ing ha he associa ion be ween BDNF
and CVD isk is likely causal.
Discussion
We p ospec i ely in es iga ed he ela ion o se um BDNF
concen a ions wi h CVD e en s and mo ali y in a la ge
communi y-based sample. We obse ed ha highe le els o
BDNF a e associa ed wi h lowe isk o bo h CVD e en s and
dea h, independen o s anda d isk ac o s, including ma ke s
o low-g ade inflamma ion, BMI, physical ac i i y, and dep es-
sion. Indi iduals in he highes qua ile o BDNF had an 25%
o 30% lowe adjus ed isk o u u e CVD e en s and dea h
Figu e 1. Age- and sex-adjus ed cumula i e incidence o ca -
dio ascula e en s ( op panel) and all-cause mo ali y (bo om
panel), s a ified by qua iles o se um BDNF. BDNF indica es
b ain-de i ed neu o ophic ac o ; CVD, ca dio ascula disease.
DOI: 10.1161/JAHA.114.001544 Jou nal o he Ame ican Hea Associa ion 6
Ci cula ing BDNF and Ca dio ascula Disease Kaess e al
ORIGINAL RESEARCH
compa ed wi h hose in he lowes qua ile. The biologic
plausibili y, demons a ion o a dose- esponse ela ion (in
qua ile-based analyses and splines), consis ency o associ-
a ion ac oss mul iple analyses, he empo al ela ion (BDNF
le els we e measu ed be o e he ou comes o in e es ), and
he esul s o he Mendelian andomiza ion expe imen a e
consis en wi h he no ion ha he associa ion is indeed a
causal one.
24
Epidemiological Da a on BDNF and CVD
Exis ing epidemiological da a ega ding he ela ion o ci cu-
la ing BDNF concen a ions wi h CVD and mo ali y a e spa se
and inconclusi e. A small case-con ol s udy o 31 pa ien s
wi h acu e co ona y synd ome and 19 heal hy con ols
conduc ed by Manni and collegues obse ed lowe plasma
le els o BDNF in he acu e co ona y synd ome cases
compa ed wi h he con ols, possibly sugges ing a p o ec i e
ole o BDNF.
25
K abbe and collegues epo ed ha plasma
BDNF was in e sely associa ed wi h mo ali y in a sample o
188 elde ly women.
6
Recen ly, Jian and collegues in es iga ed
plasma BDNF concen a ions in 885 pa ien s wi h angina.
5
They obse ed a c oss-sec ional associa ion o BDNF le els
wi h CVD isk ac o s in a p o ec i e di ec ion. Fu he mo e,
highe BDNF le els we e p ospec i ely associa ed wi h a lowe
incidence o majo co ona y e en s in hei high- isk e e al
sample.
5
Recen ly, Hallden and collegues epo ed in a la ge
sample o smoke s ha a polymo phism in he BDNF
( s4923461) gene is associa ed wi h all-cause and ca dio as-
cula mo ali y bu no wi h inciden CVD.
26
Se um BDNF was
no measu ed in hei s udy.
To ou knowledge, ou s udy is he fi s p ospec i e
analysis in a la ge communi y-based coho demons a ing
ha highe se um BDNF le els a e associa ed wi h a
dec eased isk o u u e CVD e en s and mo ali y.
Possible Mechanisms
BDNF, a membe o he neu o ophin amily, is in ol ed in he
egula ion o a b oad ange o physiological unc ions including
neu onal de elopmen and plas ici y, ood in ake, and physical
ac i i y.
27
As no ed ea lie , he e ozygous BDNF knockou mice
consume mo e ood han do hei wild- ype li e ma es and a e
obese.
2
A he same ime, hese BDNF knockou mice a e
hype ac i e, no le ha gic, despi e hei obesi y. This obse a-
ion has led o he concep ha BDNF may be eleased in
esponse o a challenging en i onmen wi h high locomo o and
in ellec ual equi emen s and sca ce ood.
27
Al e ed BDNF
le els ha e been epo ed in a ious neu opsychia ic diseases;
o example, dep ession appea s o be associa ed wi h
dec eased se um BDNF concen a ions.
28
Besides i s impo an ole in neu obiology, he e is
inc easing e idence ha BDNF is also in ol ed in ca dio as-
cula heal h and a he oscle osis. BDNF is exp essed in
ascula endo helial cells,
29
mac ophages, and smoo h
muscle cells o a he oscle o ic co ona y a e ies.
7
In one
s udy, pa ien s wi h uns able angina ha e been epo ed o
ha e inc eased BDNF le els in he co ona y ci cula ion
compa ed wi h indi iduals wi h s able angina, possibly
sugges ing ha BDNF may be in ol ed in s abili y o he
a he oscle o ic plaque.
7
In he same s udy, BDNF adminis-
a ion inc eased NAPD(H) oxidase ac i i y o human co ona y
a e y smoo h muscle cells in i o, sugges ing ha BDNF may
indeed ha e a de imen al e ec on plaque s abili y. Ne e -
heless, he clinical ele ance o hese obse a ions emains
o be elucida ed.
Table 6. Se um BDNF, Ca dio ascula E en s, and Mo ali y—Addi ional Adjus men o hsCRP and BNP
CVD Incidence* Mo ali y
†
Haza ds Ra io (95% CI) PValue Haza ds Ra io (95% CI) PValue
Con inuous ai
1-SD inc ease 0.85 (0.75 o 0.96) 0.007 0.84 (0.75 o 0.94) 0.002
Qua iles
Qua ile 1 1.00 Re e en 1.00 Re e en
Qua ile 2 0.95 (0.69 o 1.31) 0.76 0.81 (0.61 o 1.09) 0.17
Qua ile 3 0.74 (0.52 o 1.03) 0.08 0.68 (0.49 o 0.94) 0.02
Qua ile 4 0.60 (0.43 o 0.86) 0.005 0.69 (0.50 o 0.94) 0.02
T end ac oss qua iles 0.002 0.009
BDNF indica es b ain-de i ed neu o ophic ac o ; BNP, b ain na iu e ic pep ide; hsCRP, high-sensi i y C- eac i e p o ein; CVD, ca dio ascula disease.
*Based on 270 e en s in 2439 pa icipan s o he F amingham O sp ing coho ee o CVD a baseline. Models adjus ed o age, sex, smoking, sys olic blood p essu e, hype ension
ea men , o al and high-densi y lipop o ein choles e ol, diabe es melli us, body mass index, hsCRP, and BNP.
†
Based on 311 e en s in 2811 pa icipan s o he F amingham O sp ing coho . Models adjus ed o age, sex, smoking, sys olic blood p essu e, hype ension ea men , o al and high-
densi y lipop o ein choles e ol, diabe es melli us, body mass index, p e alen CVD, hsCRP, and BNP.
DOI: 10.1161/JAHA.114.001544 Jou nal o he Ame ican Hea Associa ion 7
Ci cula ing BDNF and Ca dio ascula Disease Kaess e al
ORIGINAL RESEARCH
Con e sely, BDNF has been shown o play a c ucial ole in
he de elopmen and main enance o he ca diac ascula u e.
BDNF deficiency in mice impai s he su i al o endo helial
cells in myoca dial a e ies and capilla ies, leading o ea ly
pos na al dea h.
3
Condi ional BDNF knockou mice wi h
expe imen al myoca dial in a c ion showed la ge a eas o
fib osis and impai ed e ascula iza ion and, as a conse-
quence, g ea e impai men o sys olic unc ion han
con ols.
4
These di e ences appea ed consequen o al e ed
exp ession o BDNF in he b ain (no in he hea ) in esponse
o he ischemic e en , s imula ed ia ca diac a e en
au onomous ne es a e myoca dial in a c ion.
O no e, polymo phisms in he BDNF gene ha e been
associa ed wi h BMI in ecen genome-wide associa ion
s udies.
30
We, he e o e, adjus ed ou mul i a iable models
o BMI and diabe es, which did no al e ou findings,
ende ing obesi y an unlikely explana ion o associa ions
obse ed in ou s udy. Simila ly, polymo phisms in he BDNF
gene ha e been associa ed wi h smoking beha io , wi h he
A/C Me allele a s6265 associa ed wi h bo h lowe BDNF
le els and a sligh ly inc eased isk o being a smoke .
26,31
We he e o e ex ensi ely in es iga ed he po en ial impac o
smoking on ou findings using di e en s a is ical
app oaches. We did no find e idence o a ele an ole
o smoking in he ela ion be ween BDNF and CVD o
mo ali y.
Taken oge he , i seems easible ha ci cula ing BDNF
le els a e p o ec i e by main aining ascula in eg i y and ha
hese p o ec i e sys emic e ec s ou weigh po en ial de i-
men al local e ec s on plaque s abili y.
We obse ed ha highe BDNF le els a e linea ly associ-
a ed wi h lowe isk o CVD e en s, whe eas he p o ec i e
associa ion wi h all-cause mo ali y aba es a highe BDNF
le els. This may sugges ha he p o ec i e ca dio ascula
e ec s o high BDNF le els may be coun e balanced by
po en ial de imen al e ec s on non-CVD; o example, high
BDNF le els may p omo e cance cell g ow h.
32
Clinical Implica ions
Ou da a indica e ha measu ing BDNF only ma ginally
imp o es ca dio ascula isk p edic ion. BDNF measu emen s
did no imp o e he a ea unde he ecei e ope a ing
cha ac e is ic cu e ( o ca dio ascula e en s); only IDI and
NRI we e nominally significan . Thus, i is unlikely ha se um
BDNF is o ele an clinical alue o diagnosing o p edic ing
CVD. Howe e , ou findings may be o pa hophysiological
in e es . Indeed, modula ing BDNF signaling may be a
p omising he apeu ic concep . O no e, he ela i ely modes
e ec o SNP s6265 on BDNF le els and CVD isk does no
necessa ily ansla e in o a i ial ole o he BDNF p o ein in
he pa hogenesis o CVD. In ac , mos common SNPs
epo ed in ecen genome-wide associa ion s udies ha e only
Figu e 2. Rela ions o se um BDNF wi h ca dio ascula e en s
( op panel) and all-cause mo ali y (bo om panel)—mul i a iable
adjus ed spline eg ession. BDNF indica es b ain-de i ed neu o-
ophic ac o ; CVD, ca dio ascula disease.
Figu e 3. Mendelian andomiza ion expe imen . BDNF indica es
b ain-de i ed neu o ophic ac o ; CAD, co ona y a e y disease;
CVD, ca dio ascula disease; HR, haza d a io.
DOI: 10.1161/JAHA.114.001544 Jou nal o he Ame ican Hea Associa ion 8
Ci cula ing BDNF and Ca dio ascula Disease Kaess e al
ORIGINAL RESEARCH
modes e ec sizes, bu he apeu ic modula ion o he
espec i e p o eins may ha e a a g ea e impac . Fo
example, SNP s2479409 in PCSK9 is associa ed wi h an
2 mg/dL al e a ion in LDL choles e ol le els (pe allele), bu
PCSK9 an ibodies lowe LDL choles e ol by almos 60 mg/
dL.
33,34
Gi en he pi o al ole o BDNF in neu oplas ici y, un il now
he apeu ic applica ions o BDNF (agonis s) ha e mainly been
mo i a ed by neu o ascula and neu odegene a i e diseases.
In his con ex , an o al selec i e BDNF agonis de i ed om
plan fla onoids showed p omising esul s in mouse models o
s oke and Pa kinson’s disease.
35
Fu he in es iga ions a e
wa an ed o cla i y whe he BDNF agonis s o pha macolog-
ical in e en ions aimed a inc easing ci cula ing BDNF may
be applicable o educing ca dio ascula isk. In pa icula , i
emains o be shown whe he he po en ial ca diop o ec i e
e ec s o BDNF ou weigh he heo e ical po en ial o p omo e
cance g ow h.
Limi a ions
Some limi a ions o ou in es iga ion should be discussed.
Ou FHS sample is p ima ily whi e o Eu opean ances y and
middle-aged o elde ly. We es ima ed associa ion o SNP
s6265 wi h se um BDNF in his homogeneous bu age- and
e hnici y- es ic ed sample, whe eas he associa ion o he
SNP wi h CAD was explo ed in he a la ge , and ela i ely
mo e he e ogeneous, CARDIoGRAM conso ium. Measu ing
se um BDNF in CARDIoGRAM was no easible. The e ec size
o he SNP was somewha s onge in CARDioGRAM ela i e
o ha es ima ed in FHS (al hough he 2 we e no s a is ically
di e en ). This may be due o he possibili y ha he SNP
cap u es e ec s a a issue le el (as opposed o blood le els)
and likely eflec s li elong exposu e as opposed o a single-
occasion pe iphe al blood measu emen ha may be mo e
suscep ible o eg ession dilu ion bias. Mo eo e , ou FHS
s udy sample is o limi ed size. We he e o e a p io i decided
o analyze a e y b oad end poin o maximize s a is ical
powe . Ou s udy sample was no su ficien ly powe ed o
de i e meaning ul esul s o specific ca dio ascula end
poin s. La ge analyses will be necessa y o add ess hese
ques ions.
Conclusions
Highe se um BDNF is associa ed wi h a dec eased isk o
ca dio ascula e en s and mo ali y. Mendelian andomiza ion
suppo s a causal p o ec i e e ec o BDNF on CVD. I
confi med, ou findings aise he possibili y ha pha maco-
logical in e en ions aimed a inc easing sys emic BDNF
signaling may be a p omising ool o modula ing ca dio as-
cula isk.
Sou ces o Funding
This wo k was suppo ed by he Na ional Hea , Lung, and
Blood Ins i u e’s F amingham Hea S udy (N01-HC-25195).
BDNF measu emen s we e suppo ed by he Na ional
Ins i u e on Aging (AG031287; Seshad i). Funding sou ces
o he CARDIoGRAM conso ium a e de ailed in Appendix
S1.
Disclosu es
D s Kaess, P eis, Lieb, Beise , Yang, Chen, Hengs enbe g,
E dmann, Schunke , Seshad i, and Vasan epo no conflic s
o in e es . Disclosu es o membe s o he CARDIoG am
conso ium a e lis ed in Appendix S1.
Re e ences
1. Capo ali A, Emanueli C. Ca dio ascula ac ions o neu o ophins. Physiol Re .
2009;89:279–308.
2. Ke nie SG, Liebl DJ, Pa ada LF. BDNF egula es ea ing beha io and locomo o
ac i i y in mice. EMBO J. 2000;19:1290–1300.
3. Dono an MJ, Lin MI, Wiegn P, Rings ed T, K aeme R, Hahn R, Wang S, Ibanez
CF, Rafii S, Hemps ead BL. B ain de i ed neu o ophic ac o is an endo helial
cell su i al ac o equi ed o in amyoca dial essel s abiliza ion. De elop-
men . 2000;127:4531–4540.
4. Okada S, Yokoyama M, Toko H, Ta eno K, Mo iya J, Shimizu I, Nojima A, I o T,
Yoshida Y, Kobayashi Y, Ka agi i H, Minamino T, Komu o I. B ain-de i ed
neu o ophic ac o p o ec s agains ca diac dys unc ion a e myoca dial
in a c ion ia a cen al ne ous sys em-media ed pa hway. A e ioscle Th omb
Vasc Biol. 2012;32:1902–1909.
5. Jiang H, Liu Y, Zhang Y, Chen ZY. Associa ion o plasma b ain-de i ed
neu o ophic ac o and ca dio ascula isk ac o s and p ognosis in angina
pec o is. Biochem Biophys Res Commun. 2011;415:99–103.
6. K abbe KS, Mo ensen EL, A lund K, Pede sen AN, Pede sen BK, Jo gensen T,
B uunsgaa d H. B ain-de i ed neu o ophic ac o p edic s mo ali y isk in
olde women. J Am Ge ia Soc. 2009;57:1447–1452.
7. Eji i J, Inoue N, Kobayashi S, Shi aki R, O sui K, Honjo T, Takahashi M, Ohashi
Y, Ichikawa S, Te ashima M, Mo i T, Awano K, Shinke T, Shi e J, Hi a a K,
Yokozaki H, Kawashima S, Yokoyama M. Possible ole o b ain-de i ed
neu o ophic ac o in he pa hogenesis o co ona y a e y disease. Ci cula ion.
2005;112:2114–2120.
8. Kannel WB, Dawbe TR, Kagan A, Re o skie N, S okes J III. Fac o s o isk in he
de elopmen o co ona y hea disease–six yea ollow-up expe ience. The
F amingham S udy. Ann In e n Med. 1961;55:33–50.
9. Kannel WB, Feinleib M, McNama a PM, Ga ison RJ, Cas elli WP. An
in es iga ion o co ona y hea disease in amilies. The F amingham o sp ing
s udy. Am J Epidemiol. 1979;110:281–290.
10. Kannel WB, So lie P. Some heal h benefi s o physical ac i i y. The
F amingham S udy. A ch In e n Med. 1979;139:857–861.
11. Gee lings SW, Beekman AT, Deeg DJ, Twisk JW, Van Tilbu g W. Du a ion and
se e i y o dep ession p edic mo ali y in olde adul s in he communi y.
Psychol Med. 2002;32:609–618.
12. Upadhyay A, La son MG, Guo CY, Vasan RS, Lipinska I, O’Donnell CJ,
Ka hi esan S, Meigs JB, Keaney JF J , Rong J, Benjamin EJ, Fox CS.
Inflamma ion, kidney unc ion and albuminu ia in he F amingham O sp ing
coho . Neph ol Dial T ansplan . 2011;26:920–926.
13. Wang TJ, La son MG, Le y D, Benjamin EJ, Leip EP, Wilson PW, Vasan RS.
Impac o obesi y on plasma na iu e ic pep ide le els. Ci cula ion.
2004;109:594–600.
14. F iedman GD, Kannel WB, Dawbe TR, McNama a PM. An e alua ion o ollow-
up me hods in he F amingham Hea S udy. Am J Public Heal h Na ions Heal h.
1967;57:1015–1024.
15. P euss M, Konig IR, Thompson JR, E dmann J, Abshe D, Assimes TL,
Blankenbe g S, Boe winkle E, Chen L, Cupples LA, Hall AS, Halpe in E,
Hengs enbe g C, Holm H, Laaksonen R, Li M, Ma z W, McPhe son R, Musunu u
DOI: 10.1161/JAHA.114.001544 Jou nal o he Ame ican Hea Associa ion 9
Ci cula ing BDNF and Ca dio ascula Disease Kaess e al
ORIGINAL RESEARCH