E idence o Posi i ely Selec ed Si es in Mammalian a-De ensins
Da id J. Lynn,*And ew T. Lloyd,*àMa io A. Fa es,§and Cliona O’Fa elly*k
*Educa ion and Resea ch Cen e, S . Vincen ’s Uni e si y Hospi al, Dublin, I eland; Depa men o Medicine, Uni e si y College
Dublin, Dublin, I eland; àDepa men o Gene ics, T ini y College Dublin, Dublin, I eland; §Biology Depa men ,
Na ional Uni e si y o I eland, Maynoo h, I eland; and kConway Ins i u e, Uni e si y College Dublin, Dublin, I eland
a-De ensins a e a amily o mammalian an imic obial pep ides ha exhibi a iable ac i i y agains a panel o mic obes,
including bac e ia, ungi, and en eloped i uses. We ha e employed a maximum-likelihood app oach o de ec e idence
o posi i e selec ion (adap i e e olu ion) in he e olu ion o hese impo an molecules o he inna e immune esponse.
We ha e iden i ied 14 amino acid si es ha a e p edic ed o be subjec o posi i e selec ion. Fu he mo e, we show ha all
hese si es a e loca ed in he ma u e an imic obial pep ide and no in he p ep opep ide egion o he molecule, implying
ha hey a e o unc ional impo ance. These esul s sugges ha mammalian a-de ensins ha e been unde selec i e
p essu e o e ol e in esponse o po en ially in ec ious challenges by as -e ol ing mic obes.
In oduc ion
De ensins a e ca ionic an imic obial pep ides ha
show a b oad spec um o an imic obial ac i i y agains
g am-nega i e and g am-posi i e bac e ia, ungi, and
en eloped i uses. They a e gene ally hough o ac by
dis up ing he memb ane in eg i y o mic obes (Kagan
e al. 1990; Sa chell e al. 2003). In mammals, a-de ensins
and b-de ensins a e wo s uc u ally dis inc g oups, which
di e in size and in hei spacing o a six-cys eine mo i . a-
De ensins we e i s disco e ed as a amily o pep ides in
abbi mac ophages (Sels ed, Szkla ek, and Leh e 1984)
and we e subsequen ly iden i ied in he neu ophils o
humans, macaques, a s, abbi s, guinea pigs, and hams e s
and in human and mouse small in es ine Pane h cells (Raj
and Den ino 2002).
In humans, six a-de ensins ha e been iden i ied and
s udied o da e. Human a-de ensins 1 o 4 a e localized in
azu ophilic g anules o neu ophils, which has led o hem
being e e ed o as human neu ophil pep ides (HNP1 o
HPN4). The HNPs a e he mos abundan p o ein in
neu ophils and con ibu e o he oxygen-dependen killing
o phagocy osed mic oo ganisms (Ganz e al. 1985; Singh
e al. 1988; Ba eman e al. 1991). Recen ly, HNP1 o
HPN3 ha e been shown o be he majo componen s o
a soluble ac o sec e ed om CD8 T-lymphocy es ha
supp esses HIV-1 eplica ion (Zhang e al. 2002). Human
a-de ensins (HAD) 5 and 6 a e p ima ily loca ed in he
sec e o y g anules o Pane h cells wi hin he c yp s o
Liebe ku¨hn in he small in es ine (Jones and Be ins 1992,
1993; Po e e al. 1997; Salzman e al. 2003b) whe e hey
ha e been implica ed in mucosal hos de ense. Fu he
e idence o he impo an ole hese pep ides play in
mucosal hos de ense has been shown in a ecen s udy in
which ansgenic mice exp essing HAD5 we e comple ely
esis an o Salmonella yphimu ium (Salzman e al.
2003b). HAD5 has also been de ec ed in emale e-
p oduc i e ac (S ina ich e al. 1997; Quayle e al. 1998)
and in b onchial and nasal epi helia (F ye e al. 2000).
Pep ides homologous o HAD5 and HAD6 ha e been
ound in mouse Pane h cells whe e hey a e e med
c yp dins (Ouelle e e al. 1992, 1994).
a-De ensins a e encoded as p opep ides ha equi e
p o eoly ic p ocessing o become ac i a ed. I has been
sugges ed ha he p o egion is cy op o ec i e because
addi ion o i o in i o assays inhibi ac i i y (Ouelle e e al.
1999; Wu e al. 2003). Pane h cell ypsin has been iden i ied
as he p ocessing enzyme o HAD5, bu he p ocessing
enzyme o HAD6 has ye o be iden i ied (Ghosh e al.
2002). In mice, ac i a ion o c yp dins is media ed by ma ix
me allop o einase-7 (MMP-7) (Wilson e al. 1999; Ayabe
e al. 2002). The signal pep ide and p o egion in a-de ensins
exhibi mo e amino acid conse a ion han does he ma u e
an imic obial pep ide egion. Indeed only he six cys eine
esidues and a glycine esidue a posi ion 18 o he ma u e
pep ide a e conse ed be ween all a-de ensins. These
esidues a e impo an o he s uc u e o he molecule
(Hill e al. 1991), whe eas he o he esidues a e ee o a y.
Mos molecula a ia ion ei he wi hin species o
be ween species is caused by he andom ixa ion o
mu a ions ha a e neu al. Mu a ions ha a e dele e ious
o an o ganism a e emo ed by pu i ying selec ion.
Occasionally, mu a ions con e selec i e ad an age o he
o ganisms ha ing hem, and, he e o e, hese mu a ions a e
ixed in he popula ion by posi i e selec ion (adap i e
e olu ion) a a highe a e han expec ed unde neu al
e olu ion. One o he mos s ingen me hods o de ec ing
adap i e e olu ion is o compa e he a e o nonsynonymous
subs i u ions (d
N
) wi h he a e o synonymous subs i u ion
(d
S
). The a io be ween hese a es (x¼d
N
/d
S
) is hen
a eliable measu e o he selec i e p essu e ac ing in
a p o ein-coding gene. I amino acid changes a e neu al,
hey will be ixed a he same a e as synonymous mu a ions
(x¼1). I mu a ions (d
N
) a e dele e ious, d
N
/d
S
¼0, and he
mu a ions will be emo ed by selec ion. I nonsynonymous
mu a ions a e sligh ly dele e ious xis less han 1, and he
coe icien o selec ion ac ing agains he mu a ion will
depend on he popula ion size. Finally, i he amino acid
changes a e selec i ely ad an ageous, hey will be ixed a
a highe a e (x.1). A numbe o me hods ha e been
de eloped o calcula e d
N
and d
S
among lineages (see o
e iew, Yang [2002]). Maximum-likelihood (ML) me hods
a e mos s a is ically sa is ac o y because hey employ an
explici model o e olu ion, aking in o accoun he e ec s
o unequal ansi ion and ans e sion a es, unequal base
Key wo ds: Adap i e e olu ion, an imic obial pep ide, a-de ensin.
E-mail add ess: [email p o ec ed].
Mol. Biol. E ol. 21(5):819–827. 2004
DOI:10.1093/molbe /msh084
Ad ance Access publica ion Feb ua y 12, 2004
Molecula Biology and E olu ion ol. 21 no. 5 ÓSocie y o Molecula Biology and E olu ion 2004; all igh s ese ed.
and codon equencies, and a iable x alues among
lineages in a phylogeny (Yang 1998; Yang and Nielsen
1998). As mos amino acid si es a e expec ed o be
conse ed o main ain key s uc u al and unc ional
cha ac e is ics o a p o ein, and only a small numbe o
si es in a p o ein a e likely o be a ec ed by adap i e
e olu ion, me hods ha a e age subs i u ion a es o e all
si es in a sequence ha e li le powe in de ec ing cases o
posi i e selec ion. Apa om models ha de ec posi i e
selec ion among lineages, models ha can de ec amino
acids si es unde adap i e e olu ion ha e also been
de eloped (Nielsen and Yang 1998; Yang e al. 2000).
Because a-de ensins a e a c i ical componen o he
inna e immune esponse in an ‘‘a ms ace’’ agains as -
e ol ing mic obes, i is possible ha hese molecules a e
subjec o adap i e e olu ion. I has been p e iously
demons a ed ha xis g ea e han 1 in he an imic obial
pep ide egion o a-de ensins, indica i e o posi i e
selec ion (Hughes and Yeage 1997). The Hughes and
Yeage s udy was, howe e , unable o p edic which
pa icula amino acid si es we e unde posi i e selec ion,
as app op ia e models we e unde eloped a ha ime. In
his i s comp ehensi e s udy o all a ailable mammalian
a-de ensins, we implemen ML models o es o posi i e
selec ion among e olu iona y lineages and among amino
acid si es.
Ma e ials and Me hods
Coding sequences and co esponding p o ein se-
quences o 34 a-de ensins om se en di e en species,
Homo sapiens (human), Macaca mula a ( hesus monkey),
Mus musculus (mouse), Ra us no egicus ( a ), O yc o-
lagus cuniculus ( abbi ), Ca ia po cellus (domes ic guinea
pig), and Ca ia cu le i (wild guinea pig) we e downloaded
om GenBank ( able 1). These sequences ep esen all he
ull-leng h a-de ensins cu en ly a ailable in he da abase.
Molecules anno a ed as a-de ensin– ela ed we e no
included in his s udy, as hey con ain a di e en cys eine
mo i in he ma u e pep ide and do no align well in his
egion wi h o he a-de ensins. Inclusion o hese molecules
could bias he esul s by o e es ima ing d
N
. The p o ein
sequences we e aligned using he T-Co ee p og am ( ig. 1
in Supplemen a y Ma e ial online) (No edame, Higgins,
and He inga 2000). A Neighbo -Joining (NJ) ee was
in e ed om he p o ein alignmen using MEGA e sion
2.1, wi h he gamma dis ibu ion model implemen ed o
accoun o he e ogenei y among si es (Kuma e al. 2001).
The shape pa ame e o he gamma dis ibu ion (a) was
es ima ed using he BASEML p og am (implemen ing he
REV model) om he PAML package e sion 3.12 (Yang
1997). One- housand boo s ap eplica es we e ca ied ou
o es he signi icance o each node in he ee. B anches
leading o nodes wi h a boo s ap alue o less han 500
we e collapsed. The opology o he ee was used as he
inpu ee o CODEML and CODEMLSITES p og ams,
also om he PAML package e sion 3.12 (Yang 1997).
To cons uc an alignmen o he coding sequences, he
p o ein alignmen was used as a empla e, and a copygaps
Pe l sc ip was used o align he DNA, main aining he
gaps ha we e p esen in he p o ein alignmen . Any
columns in he DNA alignmen ha had mo e han h ee
gaps we e emo ed.
Va iable Selec i e P essu es Among Lineages
Models o a iable x a ios among lineages we e
i ed by ML o he alignmen o he 34 a-de ensin
sequences. The one- a io model assumes an equal x a io
o all b anches in he phylogeny. The ee- a ios model
assumes an independen x a io o each b anch. The wo
models can be compa ed by a likelihood a io es (LRT).
Twice he log-likelihood di e ence be ween he wo
models is compa ed wi h a
2
dis ibu ion wi h N-1
deg ees o eedom whe e N is he numbe o b anches.
Pos e io Bayesian p obabili ies we e es ima ed o codon
subs i u ions in each b anch o he phylogene ic ee.
Va iable Selec i e P essu es Among Amino Acid Si es
Models o a iable x a ios among si es we e used o
es o he p esence o si es unde di e si ying selec ion
(x.1) and o iden i y hem. Fi e models o he x
dis ibu ion implemen ed in he CODEMLSITES p og am
o he PAML package we e es ed. Model M1 (neu al)
assumes wo classes o si es in he p o ein: he conse ed
Table 1
Accession Numbe s o he 34 a-De ensin Genes Used in
he Analyses
Gene Species Accession Numbe
HNP1 Homo sapiens NM_004084
HNP3 Homo sapiens NM_005217
HNP4 Homo sapiens NM_001925
HAD5 Homo sapiens NM_021010
HAD6 Homo sapiens NM_001926
MNP1 Macaca mula a AF184159
MNP1A Macaca mula a AF184160
MNP2 Macaca mula a AF184161
Mde 1 Macaca mula a AF188268
Mde 3 Macaca mula a AF188269
Mde 4 Macaca mula a AF188271
Mde 6 Macaca mula a AF188272
Mde 8 Macaca mula a AF188270
CRP1 Mus musculus NM_010031
CRP2 Mus musculus U02996
CRP3 Mus musculus NM_007850
CRP4 Mus musculus NM_010039
CRP5 Mus musculus NM_007851
CRP6 Mus musculus NM_007852
NP1 Ra us no egicus U16686
NP2 Ra us no egicus NM_173329
NP3 Ra us no egicus U50353
NP3B Ra us no egicus U50354
ED Ra us no egicus AF115768
NP4 Ra us no egicus NM_173299
NP3A O yc olagus cuniculus M64599
NP4 O yc olagus cuniculus M64601
NP5 O yc olagus cuniculus M64602
MCP1 O yc olagus cuniculus M28883
MCP2 O yc olagus cuniculus M28072
DEF1A Ca ia po cellus D14119
DEF1B Ca ia po cellus D14118
DEF2 Ca ia po cellus X63676
DEF1A Ca ia cu le i X57705
NOTE.—Whe e Re Seq (h p://www.ncbi.nlm.nih.go /Re Seq/) accession num-
be s ha e been assigned, hese ha e been used, o he wise, GenBank accession
numbe s a e gi en.
820 Lynn e al.
si es (x¼0) and he neu al si es (x¼1). Model M2
(selec ion) adds a hi d class o si e, wi h xas a ee
pa ame e , allowing o si es wi h xg ea e han 1. Unde
he disc e e model M2, he p opo ion o si es unde
pu i ying selec ion (p
0
) and p opo ion o si es unde
neu ali y (p
1
) a e es ima ed om he da a. Model M3
(disc e e) uses a gene al disc e e dis ibu ion wi h h ee
classes o si e, wi h he p opo ions (p
0
,p
1
, and p
2
) and he
x a ios (x
0
,x
1
, and x
2
). Model M7 (be a) uses a be a
dis ibu ion, which, depending on pa ame e s pand q, can
ake di e en shapes in he in e al (0, 1). Model M8 (be a
and x) adds an ex a class o si es o he be a (M7) model,
wi h he p opo ion and he x a io es ima ed om he da a,
hus allowing o si es wi h xg ea e han 1. F om hese
models, h ee LRTs compa e M0 (one a io) wi h M3
(disc e e), M1 (neu al) wi h M2 (selec ion), and M7 (be a)
wi h M8 (be a and x), espec i ely. Models M2, M3, and
M8 a e es s o posi i e selec ion among si es. Pos e io
Bayesian p obabili ies o si e classes we e calcula ed o
each amino acid si e. I he x a ios o some si e classes
a e g ea e han 1, si es wi h high pos e io p obabili ies o
hose classes a e likely o be unde posi i e selec ion.
Sliding-Window Analysis o De ec Selec i e
Cons ain s by Maximum Pa simony
Se e al s udies ha e shown ha maximum-likelihood
me hods a e sensi i e o he iola ion o assump ions made
in models o de ec adap i e e olu ion and ha alse
posi i e esul s could be ob ained unde ce ain condi ions
(Suzuki and Nei 2002). We, he e o e, applied a maximum-
pa simony me hod o es o adap i e e olu ion in ou
sequences. We applied he Kimu a-based model o Li
(1993) using a sliding-window p ocedu e (Fa es e al.
2002). B ie ly, he me hod in e s a s a is ically op imum
codon-window size and slides i along he alignmen . We
hen es in each sliding s ep he signi icance o he
nonsynonymous nucleo ide subs i u ions (d
S
), synonymous
subs i u ions (d
N
), and he nonsynonymous- o-synonymous
a e a io (x). The ma hema ical app oach used is based on
he maximum-pa simony me hod o Suzuki and Gojobo i
(1999). The main di e ence howe e esides in he ac ha
he window size is selec ed unde a s a is ical p ocedu e
(Fa es e al. 2002). Ano he ad an age o using his me hod
is ha he numbe o synonymous subs i u ions is es ed o
signi icance, and, hence, sa u a ed synonymous si es, i
any, can be highligh ed and emo ed om he analysis.
Resul s
An NJ phylogene ic ee was econs uc ed om he
amino acid alignmen using MEGA e sion 2.1 ( ig. 1). The
gamma dis ibu ion model was implemen ed o accoun o
he e ogenei y among si es (Kuma e al. 2001). The gamma
shape pa ame e was es ima ed using he BASEML
p og am o be a¼1.39. This ee opology was used in
he subsequen analyses o de ec adap i e e olu ion. To
ensu e ha he opology o he ee was no dependen on
he si es unde posi i e selec ion o he in a ian si es, he
phylogene ic analysis was epea ed wi hou hese si es. The
opology o his ee was essen ially he same as he ee in
igu e 1, wi h some o he esolu ion los (see Supplemen-
a y Ma e ial online). The CODEMLSITES analysis was
epea ed using his opology as inpu and esul ed in he
de ec ion o he same si es as being subjec o posi i e
selec ion (da a no shown).
Mul igene amilies whose membe s ha e he same
unc ion may e ol e in a conce ed ashion ha homog-
enizes he sequences o he membe genes by in e locus
ecombina ion o gene con e sion, such ha sequences
wi hin a species a e mo e simila o each o he han hose
be ween species (Liao 1999). The physical clus e ing
ypical o a-de ensin amilies wi hin species may imply
conce ed e olu ion. The e a e, howe e , wo excep ions
o phylogene ic clus e ing in ou ee ( ig. 1), which may
mean ha he bi h-and-dea h model o e olu ion is mo e
app op ia e (Nei, Gu, and Si niko a 1997). Nei he he
p ima es no he oden s show species-speci ic clades o a-
de ensins. In ac , in he case o he mouse and a , he
FIG. 1.—Neighbo -Joined ee o mammalian a-de ensins. Con-
s uc ed using MEGA e sion 2.1 (gamma dis ibu ion model, 1,000
boo s ap eplica es). B anches wi h less han 50% boo s ap suppo ha e
been collapsed. Mm ¼Mus musculus (mouse), Rn ¼Ra us no egicus
( a ), Hs ¼Homo sapiens (human), Mma ¼Macaca mula a ( hesus
monkey), Oc ¼O yc olagus cuniculus ( abbi ), Cp ¼Ca ia po cellus
(domes ic guinea pig), and Cc ¼Ca ia cu le i (guinea pig).
Posi i ely Selec ed Si es in Mammalian a-De ensins 821
phylogene ic clus e ing appea s o be caused by a unc-
ional ela ionship be ween he di e en a-de ensins. Mice
do no ha e neu ophil de ensins (Eisenhaue and Leh e
1992) and mos o he a a-de ensins a e exp essed in
neu ophils; howe e , he one a a-de ensin exp essed in
he in es ine (ED_Rn) in a simila manne o he mouse
c yp dins, clus e s wi h he mouse a-de ensins. Fu he -
mo e, he le el o sequence di e gence, e en a he p o ein
le el ( ig. 1 in supplemen a y ma e ial a ailable online), is
inconsis en wi h he homogeniza ion o sequences
expec ed unde he conce ed e olu ion model. In he
bi h-and-dea h model o e olu ion, duplica e genes a e
p oduced in a gene amily. Some o he genes unc ionally
di e ge, o he s may be los om he genome, and s ill
o he s become pseudogenes (Liao 1999). Consis en wi h
his model, we ha e de ec ed mul iple a-de ensin pseudo-
genes in he human and mouse genomes (unpublished
da a).
Examining Sa u a ion o Synonymous Si es
Sa u a ion o synonymous si es is an impo an issue
in he de ec ion o posi i e selec ion by he c i e ion ha
d
N
/d
S
is g ea e han 1, as sa u a ion will lead o he
unde es ima ion o d
S
and an in la ion o he d
N
/d
S
a io.
To ensu e ha sa u a ion was no an issue in ou da a se ,
pai wise d
N
and d
S
alues we e calcula ed using he ML
me hod implemen ed by CODEML (Goldman and Yang
1994). A co ela ion analysis be ween d
N
and d
S
was
pe o med using SPSS e sion 11.0 ( ig. 2). I synony-
mous si es a e sa u a ed, we migh expec a quad a ic
model wi h an inc easing slope o be be e i o he da a
han a linea model. I synonymous si es a e no sa u a ed,
a linea model should i he da a be e han a quad a ic
model. A quad a ic model wi h a dec easing slope implies
ha when a e age alues o nonsynonymous nucleo ide
subs i u ions a e examined, mos o hem show s ong
pu i ying selec ion and no sa u a ion o synonymous si es
exis s in ou da a ha could in la e d
N
/d
S
alues.
Va iable Selec i e P essu es Among Lineages
To es o a iable x a ios among lineages, he one-
a io model (Goldman and Yang 1994), which assumes he
same x a io o all lineages, was compa ed wi h he ee-
a io model (Yang 1998), which assumes an independen
x a io o each b anch, using he LRT. The log-likelihood
alue o he one- a io model is ‘
0
¼23888.76, while he
alue o he ee- a io model is ‘
1
¼23862.20.
Compa ison o 2‘¼2(‘
1
2‘
0
)¼53.12 (d ¼60)
e eals ha he ee- a io model is no signi ican ly be e
han he one- a io model (P.0.5). The ee- a io model
does, howe e , p edic a iable x alues among lineages,
some o which a e g ea e han 1 ( ig. 3). Because he LRT
did no e eal a signi ican di e ence, we canno conclude
ha he e is e idence o posi i e selec ion among he a-
de ensin lineages. The ela i ely low x alues in some o
he mouse lineages may be a esul o he ac ha all he
mouse a-de ensins a e exp essed only in he in es ine and
no in he neu ophils (Eisenhaue and Leh e 1992), and
as a g oup hey may no be subjec o he same selec i e
p essu es as he o he species ha ha e bo h neu ophil and
in es ine a-de ensins.
Va iable Selec i e P essu es Among Amino Acid Si es
To es o posi i e selec ion a indi idual amino acid
si es, LRTs we e ca ied ou be ween model M0 (one a io)
and M3 (disc e e), M1 (neu al) and M2 (selec ion), and
M7 (be a) and M8 (be a and x) ( able 2). All 3 models
(M2, M3 and M8) ha allow o selec ion ( able 3) a e
signi ican ly a o ed o e he o he models (P,0.001) in
all cases ( able 3). The si es p edic ed o be unde posi i e
selec ion wi h pos e io p obabili ies g ea e han 0.95 a e
in ag eemen be ween models M2, M3 and M8, excep o
FIG. 2.—Co ela ion be ween d
N
and d
S
alues as es ima ed by maximum likelihood.
822 Lynn e al.
si es 25 and 62. Nei he o hese si es is p edic ed by M8,
and si e 25 is only p edic ed by M3.
a-De ensins a e encoded as p ep opep ides ha a e
p o eoly ically clea ed o elease he C- e minal an i-
mic obial pep ide. All o he si es p edic ed o be unde
posi i e selec ion unde model M8, he mos s ingen
model, a e loca ed in he ma u e an imic obial pep ide and
no in he p ep opep ide egion ( ig. 4a). Compa ison o
he dis ibu ion o posi i ely selec ed si es o a Poisson
dis ibu ion e ealed ha he clus e ing o hese si es in he
an imic obial pep ide was s a is ically signi ican (P,
0.001). Adap i e e olu ion o he ma u e pep ide is likely
o ha e d i en he conside able amino acid a ia ion in his
egion compa ed wi h he p ep opep ide egion.
The si es p edic ed o be subjec o posi i e selec ion
ha e been displayed on he h ee-dimensional s uc u e o
HNP3 ( ig. 4b). Al hough si es unde posi i e selec ion
occu h oughou he molecule, i is no able ha almos all
o he si es in he coil egion ( esidues 7 o 14) a e subjec
o adap i e e olu ion. The only si e no p edic ed is Glu14,
which is known o o m a sal b idge wi h A g6 and as
such is ela i ely in a iable (Hill e al. 1991). I is likely
ha his egion has an impo an pa o play in he
unc ion and ac i i y o he an imic obial pep ide.
Many o he si es de ec ed o be unde posi i e
selec ion using ML-based models we e also de ec ed using
he sliding-window–based me hod. Some o he s, howe e
(amino acid si es 62, 63, 68, 69, 78, and 82), did no show
signi ican di e ences be ween d
S
and d
N
. Among he
posi i e-selec ed amino acid si es, he a e age x alue was
2.972, being signi ican ly highe han 1 and highe han
he expec a ion unde neu ali y, e en a e co ec ing o
mul iple es s (mul iple sliding-window es s) using
Bon e oni co ec ion (Z ¼11.47; P,0.001). We ha e
no de ec ed sa u a ion o synonymous si es and hence x
alues a e no in la ed by his bias. We ha e o s ess,
howe e , ha maximum-pa simony me hods a e e y
conse ed and may be subjec o he p oblem o possible
con e gences. Despi e his ac , ou esul s using maxi-
mum pa simony a e qui e coinciden wi h hose using ML
me hods and do no a ec he inal conclusions o his
s udy.
FIG. 3.—Phylogeny o mammalian a-de ensins. The abb e ia ions used a e he same as in igu e 1. B anch leng hs we e es ima ed by maximum
likelihood unde he ee- a io model, which assumes an independen x alue o each b anch. B anches wi h no x alues shown had alues ¼‘.x
alues g ea e han 1 a e shown in bold. The opology and numbe o b anches is he same as in igu e 1 (some b anches a e e y sho and a e no
isible on his ee).
Posi i ely Selec ed Si es in Mammalian a-De ensins 823

FIG. 4.—Si es p edic ed o be unde posi i e selec ion in mammalian a-de ensins. (a) Si es p edic ed o be unde posi i e selec ion a e highligh ed
in he mul iple sequence alignmen and (b) he dime s uc u e o he ma u e HNP3 pep ide (PDB en y ¼1DFN). Si es shown in ed a e hose si es
p edic ed o be unde posi i e selec ion (model 8). Pos e io p obabili ies o hese si es a e all g ea e han 0.95. Si es shown in blue a e he si es ha a e
100% conse ed ac oss all OTUs. The ma u e an imic obial pep ide o HNP3 is highligh ed. The ma u e pep ides o o he a-de ensins may di e in
size sligh ly a ei he e minus. The numbe ing o esidues in he ma u e pep ide o HNP3 in (a) co esponds o he numbe ing assigned o he HNP3
s uc u e in he P o ein Da a Bank (PDB) da abase (h p://www. csb.o g/pdb/) (b). No e ha esidue 2 in (a) and (b) co esponds o esidue 63 in able 3,
and so on. The s uc u e o HNP3 was displayed using RasMol e sion 2.7.2.1(h p://www.open asmol.o g/so wa e/ asmol/).
824 Lynn e al.
Discussion
In his s udy, we ha e de ec ed posi i e selec ion a
se e al amino acid si es loca ed in he ac i e an imic obial
pep ide egion o a-de ensins. I is likely ha as mammals
e ol ed o occupy new niches, hey we e aced wi h a new
ange o mic obial pa hogens. E olu ion o an imic obial
pep ides wi h new sensi i i ies capable o a ge ing no el
in ec ious agen s would con e a selec i e ad an age. Ou
esul s a e consis en wi h adap i e e olu ion o he a-
de ensins in esponse o such a challenge and he e olu ion
o speci ici y o di e en mic obes. The e is expe imen al
e idence ha a-de ensin pep ides a e di e se in hei
po ency agains di e en pa hogens. The human neu ophil
de ensins HNP1, HNP2, and HNP3 a e no ac i e agains
he o al g am-nega i e bac e ium Ac inobacillus ac ino-
myce emcomi ans (Miyasaki e al. 1990a), whe eas abbi
NP-1 shows s ong ac i i y (Miyasaki e al. 1990b).
T ophozoi es o Gia dia lamblia a e e y sensi i e o
mouse c yp dins 2 and 3 bu no o c yp dins 1 and 6 (Aley
e al. 1994). No ably, he gia dicidal ac i i y has been
a ibu ed o he p esence o a posi i ely cha ged a ginine
esidue a posi ion 15 in he ac i e pep ide (Ouelle e and
Be ins 2001). In ou s udy, his si e is p edic ed o be
unde adap i e e olu ion. Ra and abbi de ensins ha e
also been shown o ha e a iable ac i i ies agains a panel
o g am-posi i e and g am-nega i e bac e ia (Kohashi
e al. 1992).
Small changes in he p ima y s uc u e o hese
molecules can ha e eno mous e ec on hei po ency.
HNP1 and HNP3 di e only by one esidue a he N-
e minus o he ma u e pep ide, and, ye , HNP1 exhibi s
po en ac i i y agains Candida albicans, whe eas HNP3
has li le e ec . (Leh e e al. 1988; Raj, An ony aj, and
Ka unaka an 2000). This si e was also p edic ed o be
subjec o posi i e selec ion in ou s udy. Simila ly, he
ac i i y o mouse c yp din 4, he mos po en o he mouse
a-de ensins agains a b oad spec um o mic obes
(Ouelle e e al. 1994), is dependen on he p esence o
one o wo esidues a he N- e minus o he ac i e
pep ide. Remo al o hese esidues can o ally elimina e
he an imic obial ac i i y o c yp din 4 (Ouelle e e al.
2000).
I is p obable ha he e olu ion o di e en
mechanisms o in e ac ion wi h mic obial memb anes has
been impo an in he e olu ion o speci ici y o pa icula
a-de ensins o pa icula mic obes. Human HNP2 o ms
s able mul ime ic po es in model memb anes (Wimley,
Sels ed, and Whi e 1994), whe eas abbi NP-1 does no ,
bu pe meabilizes memb anes by c ea ing la ge, sho -
li ed de ec s (H is o a, Sels ed, and Whi e 1996). These
al e na i e mechanisms a e likely o ha e a ying e ec s,
depending on he cons i u ion o he memb ane, which is
a iable among mic obes.
Many mic obes ha e e ol ed mechanisms ha
a emp o e ade and sub e he ac ions o an imic obial
molecules (Ganz 2001). I is likely ha his ongoing ‘‘a ms
ace’’ wi h mic obes has been a signi ican o ce d i ing
he adap i e e olu ion o he a-de ensins. Fo example, he
human pa hogens, Pseudomonas ae uginosa, En e ococ-
cus aecalis, and S ep ococcus pyogenes elease de ma an
sulpha e, a compound ha binds o and neu alizes HNP1
(Schmid chen, F ick, and Bjo ck 2001). Fu he mo e, he
pa hogen Salmonella en e ica can elici a dec ease in he
exp ession o mouse c yp dins (Salzman e al. 2003a).
In his s udy we ha e p o ided e idence ha se e al
amino acid si es in he ac i e pep ide o mammalian a-
de ensins a e unde posi i e Da winian e olu ion. This
wo k will assis in he design o in i o analyses o
unc ionally and s uc u ally ele an si es. By syn he i-
cally changing he esidues a si es p edic ed o be unde
posi i e selec ion, which a e he si es mos likely o be o
unc ional impo ance, i is possible o al e he ac i i y o
hese molecules agains pa icula pa hogens and gain
insigh in o hei mechanisms o ac i i y.
Posi i e selec ion in he ma u e an imic obial egion
o o he an imic obial pep ide amilies has also been
demons a ed. E idence o posi i e selec ion has been
de ec ed in p ima e b-de ensins (Bonio o e al. 2003;
Semple, Rol e, and Do in 2003), mu ine b-de ensins
(Mo ison e al. 2003), he D osophila and opin an ibac-
Table 2
E idence o Adap i e E olu ion Among Si es in Mammalian a-De ensins
Model P Pa ame e s ‘d
N
/d
S
Posi i ely Selec ed Si es
M0: one a io 1 x¼0.9342 23888.76 ¼x
M1: neu al 1 p
0
¼0.108, x
0
¼023837.45 0.8918
p
1
¼0.892, x
1
¼1
M2: selec ion 3 p
0
¼0.107, x
0
¼023751.88 1.6211 62,63,68,69,70,72,73,74,77,78,82,84,86,87,89
p
1
¼0.670, x
1
¼1
p
2
¼0.223, x
2
¼4.263
M3: disc e e 5 p
0
¼0.188, x
0
¼0.075 23740.19 1.264 25,62,63,68,69,70,72,73,74,77,78,82,84,86,87,89
p
1
¼0.585, x
1
¼0.821
p
2
¼0.227, x
2
¼3.386
M7: be a 2 p¼0.479, q¼0.273 23797.38 0.6374
M8: be a and x4p
0
¼0.791, p¼0.452, q¼0.230 –3738.80 1.258 63,68,69,70,72,73,74,77,78,82,84,86,87,89
p
1
¼0.209, x¼3.513
Table 3
Likelihood Ra io Tes (LRT) o De ec Adap i e E olu ion
Models 2‘
2
Value d P- alue
M1 e sus M2 2(–3837.45 23751.88) 171.14 2 ,0.001
M0 e sus M3 2(–3888.76 23740.19) 297.14 4 ,0.001
M7 e sus M8 2(–3797.38 23738.80) 117.16 2 ,0.001
Posi i ely Selec ed Si es in Mammalian a-De ensins 825
e ial pep ide (Da e-I o e al. 2002), and in a numbe o
amphibian an imic obial pep ides (Duda, Vanhoye, and
Nicolas 2002). The de ec ion o posi i e selec ion in hese
impo an modula o s o he inna e immune esponse
p o ides e idence ha despi e he e olu ion o he
adap i e immune esponse in jawed e eb a es mo e han
450 MYA, he inna e immune esponse con inues o
unc ion as a c i ical elemen in hos de ense.
Acknowledgmen s
We would like o hank wo anonymous e iewe s o
help ul commen s. This esea ch was suppo ed by he
Food Ins i u ional Resea ch Measu e (F.I.R.M.) G an 01/
R&D/D/135 om he I ish Depa men o Ag icul u e,
Food and Ru al De elopmen .
Li e a u e Ci ed
Aley, S. B., M. Zimme man, M. He sko, M. E. Sels ed, and F. D.
Gillin. 1994. Killing o Gia dia lamblia by c yp dins and
ca ionic neu ophil pep ides. In ec . Immun. 62:5397–5403.
Ayabe, T., D. P. Sa chell, P. Pesendo e , H. Tanabe, C. L.
Wilson, S. J. Hagen, and A. J. Ouelle e. 2002. Ac i a ion o
Pane h cell alpha-de ensins in mouse small in es ine. J. Biol.
Chem. 277:5219–5228.
Ba eman, A., A. Singh, C. Shus ik, W. M. Ma s, and S. Solomon.
1991. The isola ion and iden i ica ion o mul iple o ms o he
neu ophil g anule pep ides om human leukemic cells. J.
Biol. Chem. 266:7524–7530.
Bonio o, M., A. Tossi, M. DelPe o, S. Sgubin, N. An che a, D.
San on, J. Mas e s, and S. C o ella. 2003. E olu ion o he
be a de ensin 2 gene in p ima es. Genes Immun. 4:251–257.
Da e-I o, A., K. Kasaha a, H. Sawai, and S. I. Chigusa. 2002.
Rapid e olu ion o he male-speci ic an ibac e ial p o ein
and opin gene in D osophila. J. Mol. E ol. 54:665–670.
Duda, T. F. J ., D. Vanhoye, and P. Nicolas. 2002. Roles o
di e si ying selec ion and coo dina ed e olu ion in he
e olu ion o amphibian an imic obial pep ides. Mol. Biol.
E ol. 19:858–864.
Eisenhaue , P. B., and R. I. Leh e . 1992. Mouse neu ophils lack
de ensins. In ec . Immun. 60:3446–3447.
Fa es, M. A., S. F. Elena, J. O iz, E. Ba io, and A. Moya. 2002.
A sliding window-based me hod o de ec selec i e cons ain s
in p o ein-coding genes and i s applica ion o RNA i uses. J.
Mol. E ol. 55:509–521.
F ye, M., J. Ba gon, N. Daule bae , A. Webe , T. O. Wagne ,
and R. G opp. 2000. Exp ession o human alpha-de ensin 5
(HD5) mRNA in nasal and b onchial epi helial cells. J. Clin.
Pa hol. 53:770–773.
Ganz, T. 2001. Fa al a ac ion e aded: how pa hogenic bac e ia
esis ca ionic polypep ides. J. Exp. Med. 193:F31–F34.
Ganz, T., M. E. Sels ed, D. Szkla ek, S. S. Ha wig, K. Dahe , D.
F. Bain on, and R. I. Leh e . 1985. De ensins: na u al pep ide
an ibio ics o human neu ophils. J. Clin. In es . 76:1427–
1435.
Ghosh, D., E. Po e , B. Shen, S. K. Lee, D. Wilk, J. D azba, S.
P. Yada , J. W. C abb, T. Ganz, and C. L. Be ins. 2002.
Pane h cell ypsin is he p ocessing enzyme o human
de ensin-5. Na . Immunol. 3:583–590.
Goldman, N., and Z. Yang. 1994. A codon-based model o
nucleo ide subs i u ion o p o ein-coding DNA sequences.
Mol. Biol. E ol. 11:725–736.
Hill, C. P., J. Yee, M. E. Sels ed, and D. Eisenbe g. 1991. C ys al
s uc u e o de ensin HNP-3, an amphiphilic dime : mecha-
nisms o memb ane pe meabiliza ion. Science 251:1481–
1485.
H is o a, K., M. E. Sels ed, and S. H. Whi e. 1996. In e ac ions
o monome ic abbi neu ophil de ensins wi h bilaye s:
compa ison wi h dime ic human de ensin HNP-2. Biochem-
is y 35:11888–11894.
Hughes, A. L., and M. Yeage . 1997. Coo dina ed amino acid
changes in he e olu ion o mammalian de ensins. J. Mol.
E ol. 44:675–682.
Jones, D. E., and C. L. Be ins. 1992. Pane h cells o he human
small in es ine exp ess an an imic obial pep ide gene. J. Biol.
Chem. 267:23216–23225.
———. 1993. De ensin-6 mRNA in human Pane h cells:
implica ions o an imic obial pep ides in hos de ense o he
human bowel. FEBS Le . 315:187–192.
Kagan, B. L., M. E. Sels ed, T. Ganz, and R. I. Leh e . 1990.
An imic obial de ensin pep ides o m ol age-dependen ion-
pe meable channels in plana lipid bilaye memb anes. P oc.
Na l. Acad. Sci. USA 87:210–214.
Kohashi, O., T. Ono, K. Ohki e al. (8 co-au ho s) 1992.
Bac e icidal ac i i ies o a de ensins and syn he ic abbi
de ensins on S aphylococci, Klebsiella pneumoniae (Chedid,
277, and 8N3), Pseudomonas ae uginosa (mucoid and
nonmucoid s ains), Salmonella yphimu ium (Ra, Rc, Rd,
and Re o LPS mu an s) and Esche ichia coli. Mic obiol.
Immunol. 36:369–380.
Kuma , S., K. Tamu a, I. B. Jakobsen, and M. Nei. 2001.
MEGA2: molecula e olu iona y gene ics analysis so wa e.
Bioin o ma ics 17:1244–1245.
Leh e , R. I., T. Ganz, D. Szkla ek, and M. E. Sels ed. 1988.
Modula ion o he in i o candidacidal ac i i y o human
neu ophil de ensins by a ge cell me abolism and di alen
ca ions. J. Clin. In es . 81:1829–1835.
Li, W. H. 1993. Unbiased es ima ion o he a es o synonymous
and nonsynonymous subs i u ion. J. Mol. E ol 36:96–99.
Liao, D. 1999. Conce ed e olu ion: molecula mechanism and
biological implica ions. Am. J. Hum. Gene . 64:24–30.
Miyasaki, K. T., A. L. Bodeau, T. Ganz, M. E. Sels ed, and R. I.
Leh e . 1990a. In i o sensi i i y o o al, g am-nega i e,
acul a i e bac e ia o he bac e icidal ac i i y o human
neu ophil de ensins. In ec . Immun. 58:3934–3940.
Miyasaki, K. T., A. L. Bodeau, M. E. Sels ed, T. Ganz, and R. I.
Leh e . 1990b. Killing o o al, g am-nega i e, acul a i e
bac e ia by he abbi de ensin, NP-1. O al Mic obiol.
Immunol. 5:315–319.
Mo ison, G. M., C. A. Semple, F. M. Kilanowski, R. E. Hill, and
J. R. Do in. 2003. Signal sequence conse a ion and ma u e
pep ide di e gence wi hin subg oups o he mu ine be a-
de ensin gene amily. Mol. Biol. E ol. 20:460–470.
Nei, M., X. Gu, and T. Si niko a. 1997. E olu ion by he bi h-
and-dea h p ocess in mul igene amilies o he e eb a e
immune sys em. P oc. Na l. Acad. Sci. USA 94:7799–7806.
Nielsen, R., and Z. Yang. 1998. Likelihood models o de ec ing
posi i ely selec ed amino acid si es and applica ions o he
HIV-1 en elope gene. Gene ics 148:929–936.
No edame, C., D. G. Higgins, and J. He inga. 2000. T-Co ee: A
no el me hod o as and accu a e mul iple sequence
alignmen . J. Mol. Biol. 302:205–217.
Ouelle e, A. J., and C. L. Be ins. 2001. Pane h cell de ensins
and inna e immuni y o he small bowel. In lamm. Bowel Dis.
7:43–50.
Ouelle e, A. J., D. Da moul, D. T an, K. M. Hu ne , J. Yuan,
and M. E. Sels ed. 1999. Pep ide localiza ion and gene
s uc u e o c yp din 4, a di e en ially exp essed mouse
Pane h cell alpha-de ensin. In ec . Immun. 67:6643–6651.
Ouelle e, A. J., M. M. Hsieh, M. T. Nosek, D. F. Cano-Gauci,
K. M. Hu ne , R. N. Buick, and M. E. Sels ed. 1994. Mouse
826 Lynn e al.
Pane h cell de ensins: p ima y s uc u es and an ibac e ial
ac i i ies o nume ous c yp din iso o ms. In ec . Immun.
62:5040–5047.
Ouelle e, A. J., S. I. Mille , A. H. Henschen, and M. E. Sels ed.
1992. Pu i ica ion and p ima y s uc u e o mu ine c yp din-1,
a Pane h cell de ensin. FEBS Le . 304:146–148.
Ouelle e, A. J., D. P. Sa chell, M. M. Hsieh, S. J. Hagen, and M.
E. Sels ed. 2000. Cha ac e iza ion o luminal Pane h cell
alpha-de ensins in mouse small in es ine: a enua ed an i-
mic obial ac i i ies o pep ides wi h unca ed amino e mini.
J. Biol. Chem. 275:33969–33973.
Po e , E. M., L. Liu, A. O en, P. A. An on, and T. Ganz. 1997.
Localiza ion o human in es inal de ensin 5 in Pane h cell
g anules. In ec . Immun. 65:2389–2395.
Quayle, A. J., E. M. Po e , A. A. Nussbaum, Y. M. Wang, C.
B abec, K. P. Yip, and S. C. Mok. 1998. Gene exp ession,
immunolocaliza ion, and sec e ion o human de ensin-5 in
human emale ep oduc i e ac . Am. J. Pa hol. 152:1247–
1258.
Raj, P. A., K. J. An ony aj, and T. Ka unaka an. 2000. La ge-
scale syn hesis and unc ional elemen s o he an imic obial
ac i i y o de ensins. Biochem. J. 347(p 3):633–641.
Raj, P. A., and A. R. Den ino. 2002. Cu en s a us o de ensins
and hei ole in inna e and adap i e immuni y. FEMS
Mic obiol. Le . 206:9–18.
Salzman, N. H., M. M. Chou, H. de Jong, L. Liu, E. M. Po e ,
and Y. Pa e son. 2003a. En e ic salmonella in ec ion inhibi s
Pane h cell an imic obial pep ide exp ession. In ec . Immun.
71:1109–1115.
Salzman, N. H., D. Ghosh, K. M. Hu ne , Y. Pa e son, and C. L.
Be ins. 2003b. P o ec ion agains en e ic salmonellosis in
ansgenic mice exp essing a human in es inal de ensin.
Na u e 422:522–526.
Sa chell, D. P., T. Sheynis, Y. Shi a uji, S. Kolushe a, A. J.
Ouelle e, and R. Jelinek. 2003. In e ac ions o mouse Pane h
cell alpha-de ensins and alpha-de ensin p ecu so s wi h
memb anes. p osegmen inhibi ion o pep ide associa ion
wi h biomime ic memb anes. J. Biol. Chem. 278:13838–
13846.
Schmid chen, A., I. M. F ick, and L. Bjo ck. 2001. De ma an
sulpha e is eleased by p o einases o common pa hogenic
bac e ia and inac i a es an ibac e ial alpha-de ensin. Mol.
Mic obiol. 39:708–713.
Sels ed, M. E., D. Szkla ek, and R. I. Leh e . 1984. Pu i ica ion
and an ibac e ial ac i i y o an imic obial pep ides o abbi
g anulocy es. In ec . Immun. 45:150–154.
Semple, C. A., M. Rol e, and J. R. Do in. 2003. Duplica ion and
selec ion in he e olu ion o p ima e be a-de ensin genes.
Genome Biol. 4:R31.
Singh, A., A. Ba eman, Q. Z. Zhu, S. Shimasaki, F. Esch, and S.
Solomon. 1988. S uc u e o a no el human g anulocy e
pep ide wi h an i-ACTH ac i i y. Biochem. Biophys. Res.
Commun. 155:524–529.
Suzuki, Y., and T. Gojobo i. 1999. A me hod o de ec ing
posi i e selec ion a single amino acid si es. Mol. Biol. E ol.
16:1315–1328.
Suzuki, Y., and M. Nei. 2002. Simula ion s udy o he eliabili y
and obus ness o he s a is ical me hods o de ec ing posi i e
selec ion a single amino acid si es. Mol. Biol. E ol. 19:1865–
1869.
S ina ich, D. M., N. A. Wol , R. Gomez, B. Gonik, and R.
Rome o. 1997. De ec ion o human de ensin 5 in ep oduc i e
issues. Am. J. Obs e . Gynecol. 176:470–475.
Wilson, C. L., A. J. Ouelle e, D. P. Sa chell, T. Ayabe, Y. S.
Lopez-Boado, J. L. S a man, S. J. Hul g en, L. M. Ma isian,
and W. C. Pa ks. 1999. Regula ion o in es inal alpha-de ensin
ac i a ion by he me allop o einase ma ilysin in inna e hos
de ense. Science 286:113–117.
Wimley, W. C., M. E. Sels ed, and S. H. Whi e. 1994.
In e ac ions be ween human de ensins and lipid bilaye s:
e idence o o ma ion o mul ime ic po es. P o ein Sci.
3:1362–1373.
Wu, Z., A. P ahl, R. Powell, B. E icksen, J. Lubkowski, and W.
Lu. 2003. F om p o de ensins o de ensins: syn hesis and
cha ac e iza ion o human neu ophil p o alpha-de ensin-1 and
i s ma u e domain. J. Pep . Res. 62:53–62.
Yang, Z. 1997. PAML: a p og am package o phylogene ic
analysis by maximum likelihood. Compu . Appl. Biosci.
13:555–556.
———. 1998. Likelihood a io es s o de ec ing posi i e
selec ion and applica ion o p ima e lysozyme e olu ion.
Mol. Biol. E ol. 15:568–573.
———. 2002. In e ence o selec ion om mul iple species
alignmen s. Cu . Opin. Gene . De . 12:688–694.
Yang, Z., and R. Nielsen. 1998. Synonymous and nonsynon-
ymous a e a ia ion in nuclea genes o mammals. J. Mol.
E ol. 46:409–418.
Yang, Z., R. Nielsen, N. Goldman, and A. M. Pede sen. 2000.
Codon-subs i u ion models o he e ogeneous selec ion p es-
su e a amino acid si es. Gene ics 155:431–449.
Zhang, L., W. Yu, T. He e al. (13 co-au ho s). 2002.
Con ibu ion o human alpha-de ensin 1, 2, and 3 o he
an i-HIV-1 ac i i y o CD8 an i i al ac o . Science 298:995–
1000.
Na ya Sai ou, Associa e Edi o
Accep ed Decembe 10, 2003
Posi i ely Selec ed Si es in Mammalian a-De ensins 827