P o eome-Wide Analysis o Func ional Di e gence in
Bac e ia: Explo ing a Hos o Ecological Adap a ions
B ian E. Ca ey
1.
, Tom A. Williams
1.
, Xiaowei Jiang
1
, Ch is ina To
2
, Ka s en Hokamp
1
,
Ma io A. Fa es
1,3
*
1Depa men o Gene ics, Uni e si y o Dublin, T ini y College, Dublin, I eland, 2Depa men o Molecula E olu ion, E olu iona y Biology Cen e, Uppsala Uni e si y,
Uppsala, Sweden, 3In eg a i e Sys ems Biology G oup, Ins i u o de Biologı
´a Molecula y Celula de Plan as, CSIC-Uni e sidad Poli e
´cnica de Valencia (UPV), Valencia, Spain
Abs ac
Func ional di e gence is he p ocess by which new genes and unc ions o igina e h ough he modi ica ion o exis ing ones.
Bo h gene ic and en i onmen al ac o s in luence he e olu ion o new unc ions, including gene duplica ion o changes in
he ecological equi emen s o an o ganism. No el unc ions eme ge a he expense o ances al ones and a e gene ally
accompanied by changes in he selec i e o ces a cons ained p o ein egions. We p esen so wa e capable o analyzing
whole p o eomes, iden i ying pu a i e amino acid eplacemen s leading o unc ional change in each p o ein and
pe o ming s a is ical es s on all abula ed da a. We apply his me hod o 750 comple e bac e ial p o eomes o iden i y
high-le el pa e ns o unc ional di e gence and link hese pa e ns o ecological adap a ions. P o eome-wide analyses o
unc ional di e gence in bac e ia wi h di e en ecologies e eal a sepa a ion be ween p o eins in ol ed in in o ma ion
p ocessing (Ribosome biogenesis e c.) and hose which a e dependen on he en i onmen (ene gy me abolism, de ense
e c.). We show ha he e olu ion o pa hogenic and symbio ic bac e ia is cons ained by hei associa ion wi h he hos , and
also iden i y unusual e en s o unc ional di e gence e en in well-s udied bac e ia such as Esche ichia coli. We p esen a
desc ip ion o he oles o phylogeny and ecology in unc ional di e gence a he le el o en i e p o eomes in bac e ia.
Ci a ion: Ca ey BE, Williams TA, Jiang X, To C, Hokamp K, e al. (2012) P o eome-Wide Analysis o Func ional Di e gence in Bac e ia: Explo ing a Hos o
Ecological Adap a ions. PLoS ONE 7(4): e35659. doi:10.1371/jou nal.pone.0035659
Edi o : Jona han H. Badge , J. C aig Ven e Ins i u e, Uni ed S a es o Ame ica
Recei ed Janua y 10, 2012; Accep ed Ma ch 21, 2012; Published Ap il 26, 2012
Copy igh : !2012 Ca ey e al. This is an open-access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion License, which pe mi s
un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal au ho and sou ce a e c edi ed.
Funding: This s udy was suppo ed by a g an om he Spanish Minis e io de Ciencia e Ino acio
´n (BFU2009-12022) and a g an o he Resea ch F on ie s
P og am (10/RFP/GEN2685) om Science Founda ion I eland. The unde s had no ole in s udy design, da a collec ion and analysis, decision o publish, o
p epa a ion o he manusc ip .
Compe ing In e es s: The au ho s ha e decla ed ha no compe ing in e es s exis .
* E-mail: [email p o ec ed]
.These au ho s con ibu ed equally o his wo k.
In oduc ion
Mos new genes, unc ions, and ac i i ies o igina e h ough he
modi ica ion o exis ing ones. The e olu iona y p ocess ha gi es
ise o unc ional di e ences be ween ela ed genes is called
unc ional di e gence [1,2]. A he species le el, unc ional
di e si ica ion is p ima ily associa ed wi h adap i e adia ions,
when a single ances o di e en ia es in o mul iple descendan
species, each adap ing by na u al selec ion o one o a new se o
ecological niches (Schlu e 2000) [3]. Following his heo y,
en i onmen al a ia ion igge s di e gen na u al selec ion,
leading o he eme gence o niche specialis s. In many cases,
species unde he same ecological condi ions di e in hei abili y
o adap o new niches, e en when hey s em om he same
ances o [4,5]. The e o e, o he ac o s such as gene ic cons ain s
also play an impo an ole in he p ocess o unc ional di e gence.
The p ocess o unc ional di e gence, o depa u e o a gene
om i s ances al unc ion, is cons ained by he equi emen o
main ain he o iginal unc ion: mu a ions ha con e a new
unc ion a e likely o in e e e wi h he ances al unc ion and
he e o e a e elimina ed by nega i e selec ion. This cons ain can
be elaxed when selec ion o he ances al unc ion is weakened,
ei he h ough gene duplica ion (and he e o e edundancy), o
h ough changes o he en i onmen inhabi ed by he o ganism o
a combina ion o bo h hese ac o s. A e gene duplica ion, one
copy o he gene can be ee o e ol e in a new di ec ion i he
o he con inues o pe o m he ances al unc ion (neo unc iona-
liza ion). Al e na i ely, ances al unc ions can be pa i ioned
be ween he wo gene copies, po en ially leading o la e
specializa ion o sub unc ionaliza ion [1,2,6,7,8]. Majo changes
in he en i onmen o ecological niche can also lead o a elaxa ion
o selec i e cons ain s on ances al unc ions, al hough his p ocess
is less well cha ac e ized. Fo example, endosymbio ic bac e ia
ha e los many o he genes hei ee-li ing ela i es need o
ob ain nu ien s om he en i onmen [9], bu ha e also
expe ienced unc ional di e gence in ce ain genes [10,11].
P oka yo es a e ex ao dina ily ich in biological di e si y,
whe he measu ed in e ms o numbe o species [12,13], habi a
ange [14], o he b ead h o ene gy sou ces and biochemical
pa hways hey can exploi in o de o su i e [15]. E en
pho osyn hesis and oxida i e phospho yla ion – he mains ays o
euka yo ic ene gy me abolism – a e bac e ial in en ions acqui ed
by endosymbiosis du ing ea ly euka yo e e olu ion [16]. How did
his p oka yo ic di e si y e ol e, pa icula ly when he ixa ion o
gene duplica ions appea s o be somewha mo e equen in
euka yo es [17]?
Adap i e e olu ion in p oka yo es is p omo ed by a leas h ee
main ac o s: i s , a high s eng h o selec ion ela i e o
PLoS ONE | www.plosone.o g 1 Ap il 2012 | Volume 7 | Issue 4 | e35659
euka yo es, on accoun o hei gene ally la ge popula ion sizes
[18]; second, hei abili y o ob ain genes by ho izon al gene
ans e (HGT), which enables he sha ing o niche- ele an
unc ions be ween dis an ly- ela ed mic obes li ing in he same
en i onmen [19]; and hi d, hei use o s ess-induced hype mu-
a ion [20], which may inc ease he p oduc ion o adap i e
a ian s as a ‘‘las gasp’’ esponse o a challenging en i onmen .
Al hough we know ha hese p ocesses can d i e ecological
adap a ion in p oka yo es, iden i ying he ac ion o gene ic
a ia ion ha is associa ed wi h hese unc ional changes emains a
challenging p oblem. In he case o bac e ia, whole-genome
analyses mus ake in o accoun widesp ead HGT, which means
ha di e en genes o en disag ee on he o e all species ee [21].
This is a conside able p oblem o analyses o unc ional
di e gence, which equi e a ee in o de o de e mine he b anch
upon which a pa icula ai a ose.
The a ionale o p e ious me hods o iden i y unc ional
di e gence, and indeed he new app oach desc ibed he e, de i es
om he neu al heo y o Kimu a [22] (1983), which p edic s ha
esidues impo an o he unc ion o a p o ein will be unde
s ong unc ional cons ain and he e o e e ol e slowly. These
conside a ions ha e mo i a ed he de elopmen o a numbe o
me hods o iden i ying changes in selec i e cons ain s on p o ein-
coding genes and on single amino acid si es and lineages in a
phylogene ic ee [23,24,25,26,27,28,29,30,31,32,33]. A he
p o ein le el, Gu [34,35,36] de eloped a Bayesian app oach o
iden i y unc ional di e gence, which has become he mos widely
used. Compa isons o amino acid si e-speci ic e olu iona y a e o
esidue conse a ion be ween wo homologous clades can
he e o e be used o iden i y amino acid si es a which selec i e
cons ain s ha e changed, po en ially indica ing unc ional di e -
gence.
Recen ly, we ha e de eloped a new dis ance-based me hod
which explo es a bi u ca ing phylogene ic ee, es ing o
unc ional di e gence a each node by compa ing he wo
downs eam clades o an ou g oup in o de o iden i y si es a
which subs i u ion a es pe amino acid si es ha e shi ed [11,37].
Simila o o he me hods, ou app oach was limi ed o es s o one
gene a a ime, unless he phylogeny o all genes could be ixed in
ad ance.
Because gene al pa e ns o unc ional di e gence and hei link
o ecological changes canno be unde s ood by he analysis o
single genes, in he p esen s udy, we ha e op imised ou me hod
o (i) handle analyses o unc ional di e gence ha include
hund eds o comple e p o eomes, (ii) add ess he ac ha he
phylogenies o indi idual p o eins do no necessa ily ag ee wi h he
ue phylogeny, as is o en he case wi h o ganisms ha acqui e
genes h ough HGT, (iii) p o ide an in ui i e p obabili y
assignmen o each es which akes he unde lying phylogeny
o he sequences in o accoun and (i ) explo e all le els o each
gene ee, es ing o unc ional di e gence a each node.
Using his no el me hod o de ec unc ional di e gence, we
in e pa e ns o adical change o each p o ein indi idually, and
hen clus e species acco ding o he unc ional ca ego ies (de i ed
om COG [38]) in which hey exhibi signi ican e idence o
unc ional di e gence. We p o ide a as , open sou ce implemen-
a ion o ou me hod in he C++ p og am CAFS (Clus e ing
analysis o unc ional shi s). We pe o m an analysis o unc ional
di e gence on 750 bac e ial p o eomes. This se includes bac e ia
om a ious di e en ecological niches and he e o e p o ides a
good da ase o iden i ying ecology- ela ed unc ional di e gence.
Ou app oach (i) e eals s iking pa e ns o con e gen e olu ion
in phylogene ically dis inc bu ecologically ela ed g oups o
bac e ia, including pa hogens, endosymbion s, and he mophiles,
(ii) p o ides addi ional suppo o he iew ha bac e ia ha e a
conse ed se o co e unc ions, wi h a mo e a iable me abolic
laye and (iii) p o ides a de ailed pic u e o how indi idual species
o unusual bac e ia ha e di e ged om hei closes ela i es.
Resul s and Discussion
A conse ed unc ional co e and a iable c us in he
e olu ion o bac e ial p o eomes
An ob ious sign o unc ional di e gence (also unde s ood he e
as changes in subs i u ion a es pe amino acid si e in p o eins)
would be a se o homologs ha spans mul iple COG ca ego ies. In
his s udy we ocus only on hose alignmen s whe e all sequences
ha e he same COG anno a ion. This ep esen s he majo i y o
homologs and is a e lec ion o he ela i ely b oad cha ac e o he
COG ca ego ies.
The kinds o unc ional shi s ha we de ec on he basis o
conse ed, adical amino acid subs i u ions a e he e o e sub le
and no no iceable om simply compa ing he COG classi ica ions
ac oss homologous sequences. We used chi-squa ed es s o
e alua e he di e ences in unc ional di e gence be ween COG
gene ca ego ies in ou da ase (see Figu e 1). We compa ed he
p opo ion o posi i e es s o unc ional di e gence wi hin each o
he 19 COG ca ego ies o he backg ound expec a ion, which was
calcula ed by combining all ca ego ies. I genes in di e en
unc ional ca ego ies ha e simila p opensi ies o unde go
unc ional di e gence, we would expec he p opo ion o posi i e
es s in each ca ego y o be simila o he mean, esul ing in ew
signi ican cases o en ichmen . Howe e , eigh een o he nine een
ca ego ies we e ei he en iched o impo e ished o unc ional
di e gence, while only one ca ego y ailed o de ia e signi ican ly
om he backg ound expec a ion.
To es whe he his pola iza ion o ou da ase was simply due
o an a i ac – o ins ance, he use o a non-conse a i e
en ichmen es – we pe o med simula ions in which he genes in
ou o iginal da ase we e andomly assigned o one o he 19 COG
ca ego ies be o e es ing o en ichmen . In hese simula ions,
e en s o unc ional di e gence we e much mo e e enly dis ibu ed
Figu e 1. Di e en ca ego ies o genes expe ience di e en
le els o unc ional di e gence. P opo ion (FD) is he p opo ion o
es ed b anches wi h a leas one unc ionally di e gen si e ac oss all
gene ees in a pa icula unc ional ca ego y. Ca ego ies a e labeled
acco ding o he COG on ology sys em [38,73]. Eigh een o he
nine een ca ego ies all in o wo g oups: signi ican ly en iched o
impo e ished. Mos in o ma ion p ocessing genes (K, J, L, A) all in o he
la e g oup, while me abolic unc ions (E,F,G,H,P,Q) and genes in ol ed
in de ense (V) o ound on he cell su ace (M) a e en iched o adical
change. Desc ip ion o COG ags can be ound in Table S5.
doi:10.1371/jou nal.pone.0035659.g001
Func ional Di e gence Media es Adap a ions
PLoS ONE | www.plosone.o g 2 Ap il 2012 | Volume 7 | Issue 4 | e35659
among he ca ego ies, so ha 93% o ca ego ies we e nei he
en iched no impo e ished o unc ional di e gence ela i e o he
backg ound le el. This esul indica es ha he p obabili y o
unc ional change is no e enly dis ibu ed among he eal
ca ego ies: he e is a s a k di ision be ween en iched and
impo e ished ca ego ies. This suppo s he idea ha bac e ial
p o eomes comp ise a ela i ely unchanging co e ( ha is, genes in
impo e ished ca ego ies) coupled wi h a se o mo e a iable
unc ions (en iched ca ego ies), as p e iously no iced [39,40,41].
The impo e ished ca ego ies a e almos exclusi ely hose
in ol ed wi h in o ma ion s o age and p ocessing, including
DNA eplica ion, ecombina ion, and epai (L); ansc ip ion
(K), ibosome biogenesis (J); and cell di ision (D). Me abolic genes
we e among hose en iched o unc ional di e gence, including
genes in ol ed in he me abolism o coenzymes (H), seconda y
me aboli es (Q), ca bohyd a es (G), amino acids (E) and nucleo-
ides (F). Along wi h hese me abolic ca ego ies, cell wall and
en elope genes (M) and cellula de ense mechanisms (V) we e
among he mos en iched ca ego ies in ou analysis, highligh ing
he c i ical ole o he en i onmen in di ec ing lineage-speci ic
episodes o unc ional change. Taken oge he , ou esul s ag ee
wi h a numbe o p e ious epo s indica ing ha p o eins in ol ed
in in o ma ion p ocessing a e mo e conse ed ac oss la ge
e olu iona y dis ances han hose in ol ed in me abolism
[39,40,41,42].
An addi ional poin bea s emphasizing he e: since ou me hod
con ols o he le el o conse a ion a each node on he ee, he
signi icance o a pa icula subs i u ion pa e n depends on he
backg ound e olu iona y a e so ha in slow-e ol ing p o eins,
ela i ely conse a i e subs i u ions a e de ec ed as signi ican
e en s o unc ional di e gence, whe eas only e y unusual
subs i u ion pa e ns will a ain signi icance in as -e ol ing
p o eins. The e o e, ou esul s indica e ha in o ma ion p ocess-
ing genes a e no only mo e conse ed han o he s pu ely in e ms
o e olu iona y a e, bu ha hey also expe ience less unc ional
change e en aking his low a e o sequence e olu ion in o
accoun .
Why a e in o ma ional genes unde g ea e unc ional con-
s ain han he es o he p o eome? One possibili y, which
ollows C ick’s concep o he ‘‘ ozen acciden ’’ [43], is ha oo
many o he genes depend on he basic unc ions o ansla ion,
ansc ip ion, and epai : unc ional changes in hese genes would
dis up many o he sys ems in he cell. This hypo hesis is
suppo ed by he obse a ion ha he COG ca ego y con aining
p o ein a icking and chape ones (O) is also impo e ished: he
co e ac i i ies o gene alis chape ones such as G oEL and DnaK
a e equi ed o he p ope olding o many di e en p o eins in
bac e ial cells [44].
Hos in e ac ions cons ain unc ional change in
pa hogenic and symbio ic bac e ia
Does he ecological niche o an o ganism in luence he pa e n
o unc ional change i expe iences? To answe his ques ion, we
e alua ed he en ichmen o unc ional di e gence in each species
ela i e o he o he s in ou da ase . To calcula e he en ichmen
s a us o each species, we used he same s a is ical s a egy as
employed o en ichmen by unc ional ca ego y: we calcula ed a
backg ound p opo ion o success ul es s o unc ional di e gence
o e all species, and hen compa ed his o he p opo ion o each
species indi idually using chi-squa ed es s (a ull able o hese
esul s can be ound in Table S1). We also used chi-squa ed es s o
iden i y associa ions be ween hese h ee en ichmen pa e ns
(en ichmen , impo e ishmen , o nei he ) and o ganism li es yle,
as is summa ized in Table 1. While he e was no s a is ically
signi ican di e ence be ween psych ophiles and mesophiles in
e ms o unc ional di e gence (chi-squa ed = 0.9762, P = 0.6138),
no indeed was he e signi icance when compa ing he mophiles
o ee li ing bac e ia.
The mophiles did howe e p esen a con as ing pa e n o
unc ional di e gence in compa ison o he gene al pa e n, wi h
wo COG ca ego ies being en iched o unc ional di e gence in
he mophiles while being impo e ished in gene al. These
ca ego ies a e di ec ly ela ed wi h he su i al o cells unde hea
s ess: ca ego y K, which comp ises mos ly ansc ip ion ac o s,
and ca ego y L, which is in ol ed in DNA eplica ion,
ecombina ion and epai . Abo e ce ain empe a u e h eshold,
molecula pa hways unde go d ama ic empe a u e induced
al e a ions ha d i e o cy o oxici y, adiosensi iza ion and
he mo ole ance [45,46]. Among all he esponses ha ake place
in he cell unde high empe a u es, inhibi ion o DNA, RNA and
p o ein syn hesis is he esponse ha in ol es a complex and ine-
uning o egula ion mechanisms, mainly o ches a ed by an-
sc ip ion ac o s [46]. One such impo an egula ed mechanism is
he induc ion o hea -shock p o eins, pa icula ly in ol ed in
Table 1. E ec o o ganism li es yle on unc ional di e gence.
Li es yle Compa ison En iched Nei he Impo e ished Signi icance
Psych ophile Mesophile 2/61 6/433 1/66 N.S.
The mophile Mesophile 7/61 22/433 1/66 N.S.
Pa hogen Non-pa hogen 22/77 272/294 47/38 *** (2)
In acellula pa hogen O he pa hogen 0/22 26/246 4/43 N.S.
Symbion Non-symbion 4/95 36/530 13/72 *
*In acellula endosymbion All o he s 4/224 14/360 15/133 *** (2)
All in e ac o s F ee-li ing 44/55 410/156 70/15 *** (2)
Associa ions be ween li es yle and en ichmen o unc ional di e gence: he numbe s o genomes in each ca ego y a e gi en in he o m Li es yle/Compa ison.
Signi icance was assessed wi h Ya es-co ec ed chi-squa ed es s, o Fishe es s when he expec ed coun was lowe han 5 o any one cell in he con ingency able.
Signi icance codes: N.S. =P.0.05;
*=P,0.05,
** = P,0.01,
*** = P,0.001.
I an associa ion was signi ican , ‘‘+’’ o ‘‘2’’ deno e he di ec ion o he shi associa ed wi h he li es yle being es ed. Fo ins ance, in e ac o s a e signi ican ly
impo e ished (2) compa ed o ee-li ing bac e ia.
doi:10.1371/jou nal.pone.0035659. 001
Func ional Di e gence Media es Adap a ions
PLoS ONE | www.plosone.o g 3 Ap il 2012 | Volume 7 | Issue 4 | e35659
mi iga ing he cy o oxic e ec s due o he non-speci ic agg ega ion
o un olded and dena u ed p o eins [47,48].
In e es ingly, we ound ha all bac e ia ha in e ac wi h a hos
as an in eg al pa o hei li es yle (including pa hogens, pa asi es,
symbion s and commensals) we e signi ican ly impo e ished o
unc ional di e gence in compa ison o hei ee-li ing ela i es
(see Table 1). This esul is somewha su p ising because
pa hogens and symbion s gene ally expe ience highe a es o
e olu ion han ee-li ing bac e ia, al hough much o he inc ease
can be a ibu ed o heigh ened gene ic d i [9]. Ou esul s
sugges ha once he o e all conse a ion le el o p o eins is
accoun ed o , hese bac e ia ha e unde gone less unc ional
change han hei ee-li ing ela i es. This esul can be explained
by g ea e ecological cons ain s on hos -associa ed bac e ia, which
mus adap o he highly speci ic en i onmen o hei hos . In
pa icula , pa hogenic and symbio ic bac e ia p e e en ially lose
me abolic genes as hey no longe equi e he capaci y o exploi as
wide a ange o nu ien sou ces as ee-li ing bac e ia [9].
Since hese a e p ecisely he kind o genes ha a e mos
amenable o unc ional change (Figu e 1), hei loss om hos -
associa ed bac e ia explains he ela i e impo e ishmen o
unc ional di e gence in hese p o eomes. The emaining genes
a e also unde s ong cons ain s imposed by he specialized
en i onmen hey a e in, limi ing he e o e any oppo uni y o
unc ional di e gence (To and Fa es 2008, 2009). Howe e ,
a iabili y in genome size is a complica ing ac o in his analysis
because hos -associa ed bac e ia end o ha e smalle genomes
han hei ee-li ing ela i es. Fo ins ance, endosymbio ic
bac e ia o insec s unde wen subs an ial educ ion in he gene
con en , wi h genomes sizes anging be ween 144 kb and 792 kb
depending on he hos (in compa ison, E. coli K12 has a genome
size o 4.639 Mb) (See o example [49,50,51,52,53,54,55]. Since
unc ional di e gence o en ollows gene duplica ion [1], i migh
be expec ed ha la ge genomes would be en iched o new
unc ions in compa ison o smalle ones.
Does genome size alone accoun o he obse ed di e ences
be ween hos -associa ed and ee-li ing bac e ia? To es his
possibili y, we modeled genome en ichmen and impo e ishmen
o unc ional di e gence as a unc ion o li es yle (hos -associa ed
s. ee-li ing) and genome size (in nucleo ides) using a gene alized
linea model, a sa u a ed model was i using he glm unc ion in R
(R De elopmen Co e Team, 2010), wi h en ichmen o
impo e ishmen o unc ional di e gence as he esponse a iable
(binomial e o s), and genome size (con inuous, bp) and li es yle
(ca ego ical, ee-li ing o hos -associa ed) as he explana o y
a iables. This was simpli ied o a minimal adequa e model using
he s ep unc ion. The in e ac ion be ween genome size and
li es yle was non-signi ican and was emo ed du ing model
simpli ica ion (see Table S2). Bo h li es yle and genome size we e
signi ican , wi h hos -associa ed bac e ia signi ican ly mo e likely
o be impo e ished (P = 1.28610
213
) and, pe haps su p isingly, a
modes endency owa ds impo e ishmen in la ge genomes
(P = 0.03). The e o e, a ia ion in genome size does no accoun
o he obse ed di e ences in unc ional di e gence be ween hos -
associa ed and ee-li ing bac e ia.
To be e de ine he e ec o li es yle on unc ional di e gence,
we iden i ied he unc ional ca ego ies wi h he g ea es consis en
di e ences in en ichmen s a us be ween hos -associa ed and ee-
li ing bac e ia. In e es ingly, genes in ol ed in esicula anspo
and sec e ion sys ems (U) we e en iched o unc ional di e gence
in hos -associa ed bac e ia bu nei he en iched no impo e ished
in ee-li ing bac e ia, while signal ansduc ion genes (T) we e
impo e ished in hos -associa ed bac e ia bu en iched in hei ee-
li ing ela i es (Table S3). This pa e n can be eadily unde s ood
in e ms o he li es yles o hos -associa ed bac e ia, as pa hogens
use elabo a e sec e ion sys ems o deli e ing oxins and o he
i ulence ac o s o hei hos [56], while symbion s p o ision hei
hos s wi h nu ien s as pa o hei mu ually bene icial ela ionship
[57,58]. In addi ion, he impo e ishmen in hos -associa ed signal
ansduc ion genes e lec s hei adap a ion o a ela i ely cons an
hos en i onmen , which is conside ably mo e s able han he
luc ua ing condi ions expe ienced by hei ee-li ing ela i es.
Wha is he ela i e impo ance o genome size a ia ion
compa ed o unc ional di e gence in bac e ial adap a ion? This
ques ion is di icul o answe because bo h o hese e olu iona y
phenomena a e a wo k in he p ocess o ecological bac e ial
adap a ions. The e o e, mic o-e olu iona y p ocesses, such as
unc ional di e gence, necessa ily accompany mac o-e olu iona y
p ocesses, such as genome sh inkage o HGT, du ing bac e ia
adap a ion o di e en ecological condi ions. The iming and
ela i e impo ance o each o hese phenomena is, ne e heless,
a ied o e he di e en s ages o adap ing o a new en i onmen .
Taking he example o symbio ic bac e ia o insec s, hese bac e ia
a e cha ac e ized by a d ama ic genome s eamlining, high
mu a ion a es and he unc ional di e gence o genes in ol ed
in an endosymbio ic li es yle. When did hese p ocesses occu ? We
p edic ha HGT was impo an o he ee-li ing ances o o
hese bac e ia o acqui e pa hogenic genes and in ade he
euka yo ic cells o he hos . D ama ic gene loss in hese in ading
bac e ia may ha e become he nex impo an e olu iona y leap,
making bac e ia dependen upon he hos . Finally, unc ional
di e gence may ha e con ibu ed impo an ly o he e inemen o
he adap a ion o hese bac e ia o he no el ecological condi ions.
F om p o eome-wide o esidue-le el unc ional
di e gence
In o de o isualize he esul s o ou unc ional di e gence
analysis, we pe o med wo-dimensional hie a chical clus e ing on
he en ichmen s a us (en iched, impo e ished, o nei he )
associa ed wi h each species and unc ional ca ego y – ha is,
we clus e ed species acco ding o simila i ies in hei en ichmen
s a us ac oss he 19 unc ional ca ego ies, esul ing in he hea map
and dend og am in Figu e 2a (comple e dendog am and hea map
is a ailable in Figu e S1). This is a powe ul and in ui i e way o
ep esen ou esul s because i e eals he o e all pa e ns in he
da a – such as he ex eme conse a ion among in o ma ional
genes, pa icula ly hose in ol ed in ibosome biogenesis (J) - while
also highligh ing indi idual, lineage-speci ic excep ions o he
gene al ends. In his sec ion, we demons a e he u ili y o his
app oach by using he hea map o iden i y species ha ha e
unde gone majo unc ional shi s.
Al hough op-le el bac e ial g oups (such as he di isions o he
p o eobac e ia, he Fi micu es, Ac inobac e ia, and so on) a e no
esol ed in ou dend og am o unc ional di e gence (Figu e 2),
amily and genus-le el ela ionships o en a e, p esumably because
o close phylogene ic ela edness, sha ed gene con en , and
simila i y o ecological niche. This allows us o iden i y indi idual
species wi h a ypical pa e ns o unc ional di e gence. A
pa icula ly s iking case is ha o he Ba onella genus (Figu e
2b), which a e a g oup o in acellula pa asi es ha in ec and
eplica e in e y h ocy es [59]. O he ou Ba onella species in ou
da ase , only one – Ba onella bacilli o mus – is en iched o
unc ional di e gence in cell mo ili y genes (N), wi h he o he s
being impo e ished (2 species) o nei he en iched no impo e -
ished (1 species).
Rema kably, his is he only membe o he genus ha possesses
lagella [60]. Since e y h ocy es lack an ac i e cy oskele on, hey
canno be induced o ake up ex e nal bac e ia by in agina ion
Func ional Di e gence Media es Adap a ions
PLoS ONE | www.plosone.o g 4 Ap il 2012 | Volume 7 | Issue 4 | e35659
[61]. Ins ead, e y h ocy e in asion by Ba onella species is an ac i e
p ocess [62]. The mechanism employed by Ba onella bacilli o mus
in ol es he use o i s lagella [63] and is mo e e icien han ha o
o he Ba onella species, wi h up o 80% o e y h ocy es in ec ed
[62,64]. This appea s o be a clea case whe e ou app oach has
iden i ied an in e es ing, lineage-speci ic case o adap a ion o a
specialized ecological niche.
Figu e 2. Visualizing high-le el pa e ns o unc ional di e gence. We used hie a chical clus e ing o e eal he main pa e ns o unc ional
di e gence in ou da ase o 750 bac e ial p o eomes. (a) The comple e hea map, wi h a dend og am co esponding o ca ego y clus e ing, and
species clus e ing along he le hand side. Visualizing he da a in his way e eals he ex eme impo e ishmen o p o eins in ol ed in ibosome
biogenesis (J), as well as he en ichmen o ca ego ies in ol ed in in e ac ion wi h he en i onmen (E, M, G, H, C, P) ac oss all species. (b) Lineage-
speci ic e en s o unc ional di e gence picked ou om he hea map (dend og am colo s deno e he egions expanded upon – a la ge e sion o
he comple e hea map is a ailable as Figu e S1). Unlike o he Ba onella species, B. bacilli o mus is impo e ished o di e gence in cell mo ili y genes
(N), and is unique among Ba onella species in using a lagellum o in ec e y h ocy es. (c) Two s ains o E. coli – SMS 3–5 and UMN026 – ha e
phylogene ically a ypical pa e ns o unc ional di e gence: he cons ain s on cell mo ili y (N) a e among hose ha ha e elaxed ela i e o he o he
s ains in SMS 3–5, while UMN026 is uniquely en iched o sec e ion sys em (U) genes.
doi:10.1371/jou nal.pone.0035659.g002
Func ional Di e gence Media es Adap a ions
PLoS ONE | www.plosone.o g 5 Ap il 2012 | Volume 7 | Issue 4 | e35659

Ou hea map u ns up su p ises e en among ela i ely well-
cha ac e ized species (Figu e 2c). As expec ed, closely ela ed E.
coli and Shigella s ains clus e oge he a he bo om o he
dend og am (Figu e 2a). E. coli SMS 3–5, a mul id ug- esis an ,
hea y-me al ole an s ain isola ed om a pollu ed indus ial
en i onmen [65] is dis inguished om o he E. coli s ains on he
basis o a elaxed unc ional cons ain in he ca ego y o cell
mo ili y (N); mos o he s a e impo e ished o unc ional
di e gence, while SMS 3–5 is en iched. This p o ile co ela es
well wi h wha is known abou he biology o his s ain, which is
unique among sequenced E. coli genomes in possessing a second,
in ac la e al lagella sys em called Flag-2, in addi ion o he
no mal pe i ichious lagella ound in o he E. coli s ains [65,66].
This sys em was o iginally cha ac e ized in a di e en s ain, 042,
whe e i has been ende ed non unc ional by a ameshi
mu a ion in one o he componen genes [66], al hough i appea s
o be comple e in SMS 3–5 [65].
Ano he E. coli p o eome wi h an unusual pa e n o unc ional
di e gence is O17:K52:H18 (s ain UMN026), a mul id ug-
esis an s ain ha causes u ina y ac in ec ions [67]. Unique
among E. coli and Shigella species, his s ain is en iched o
unc ional di e gence among genes in ol ed in sec e ion (U).
In es iga ion o he genes unde lying his en ichmen e ealed
unc ional di e gence in he Vi B8 and Vi B9 genes, which
encode co e p o eins in a Type IV sec e ion sys em ound only in
wo E. coli s ains – UMN026 and 018 (ED1a), al hough he la e
species is no en iched in his ca ego y. In o he bac e ia, Type IV
sys ems a e in ol ed in he exchange o DNA wi h he
en i onmen , as well as he deli e y o e ec o p o eins o hos
cells [68]. Since hese wo p o eins a e impo an componen s o
he Type IV sec e ion sys ems o o he bac e ia, unc ional
di e gence in hese genes may be in ol ed in adap ing he sys em
o an UMN026-speci ic ole (see Figu e 3).
To gain u he insigh in o he possible implica ions o he
UMN026-speci ic changes in hese p o eins, we mapped he
speci ic esidues unde unc ional di e gence in Vi B8 (also ou pu
by CAFS) on o he Ag obac e ium ume aciens c ys al s uc u e [69]. O
he 14 si es unde unc ional di e gence (see Table S4), 5 could be
mapped on o he c ys allized egion o he p o ein. O hese 5, 4
a e a o close o posi ions p e iously shown o be o unc ional
impo ance. Th -196( esidues numbe ed acco ding o he A.
ume aciens sequence), which CAFS de ec ed as being unde
unc ional di e gence in UMN026, is di ec ly in ol ed in he
s abiliza ion o he Vi B8 homodime [69], as is Leu-211, ano he
unc ionally di e gen si e. Addi ionally wo si es iden i ied by ou
app oach a e a posi ions ha sugges hey may ha e an indi ec
ole in dime iza ion. Val-218 is loca ed be ween wo o he esidues
(Leu-217 and Val-219) ha a e in ol ed in dime o ma ion, while
Phe-127 is adjacen o Se -128, a conse ed esidue ha s abilizes
he in e ac ion su ace on Vi B8. The unc ion o he o he si e
de ec ed unde unc ional di e gence, Val-183, is cu en ly
unknown. Taken oge he , hese esul s indica e ha unc ional
di e gence in E. coli UMN026 Vi B8 has occu ed a esidues
impo an in o ming he homodime , which may ha e impo an
implica ions o he o e all s uc u e and unc ion o he complex.
Wi h no c ys al s uc u e a ailable o Vi B9, i is mo e di icul o
e alua e he unc ional signi icance o he si es de ec ed he e.
Fu he , we de ec ed unc ional di e gence on 19 b anches o he
Vi B9 ee, sugges ing ha his p o ein expe iences a mo e gene al
pa e n o adical change.
Me hods
Design and implemen a ion
Ou analysis o unc ional di e gence, he indi idual s eps o
which a e de ailed below, is summa ized in Figu e 4.
Figu e 3. Amino acid esidues unde unc ional di e gence in
E. coli
UMN026 Vi B8. Righ : he s uc u e o a Type IV sec e ion sys em
ound only in wo s ains o E. coli. CM = cy oplasmic memb ane, OM = ou e memb ane. The complex s uc u e is based on ha o Ba on (2006) [84].
In UMN026, he cen al complex p o eins Vi B8 and Vi B9 a e unde unc ional di e gence. Le : O he i e si es de ec ed by CAFS ha could be
mapped o he Vi B8 c ys al s uc u e [69], he e is e idence ha ou a e in ol ed in o ming he Vi B8 homodime , sugges ing ha unc ional
di e gence a hese posi ions is in ol ed in al e ing he qua e na y s uc u e o he complex.
doi:10.1371/jou nal.pone.0035659.g003
Func ional Di e gence Media es Adap a ions
PLoS ONE | www.plosone.o g 6 Ap il 2012 | Volume 7 | Issue 4 | e35659
Sequences, o hology, and alignmen
The i s s ep in a whole-p o eome analysis o unc ional
di e gence is he g ouping o o hologs wi hin he species o
in e es . We lea e o hology assignmen o which a numbe o
ools a e al eady in use [70], o he use s’ choice acco ding o hei
own needs. Fo he p esen analysis, we e ie ed pai wise
o hology assignmen s o 750 comple ely-sequenced bac e ial
genomes om he OMA da abase [71,72], ep esen ing all
bac e ial da a in he Oc obe 2009 e ision o he da abase.
We chose he OMA p ojec o i s e y b oad phylogene ic
co e age, as well as he a ou able pe o mance o i s algo i hm
agains o he cu en o hology assignmen me hods [70]. In
addi ion o p o iding pai wise o hology calls, he OMA algo i hm
assembles s ic o hologous g oups in which e e y membe is
di ec ly o hologous o e e y o he . The a ionale o his s ic
app oach o g ouping is he exclusion o pa alogs, which is
impo an o a numbe o po en ial applica ions o he OMA
da abase, such as phylogene ic analysis. Un o una ely, hese
g oups a e unsui able o unc ional di e gence analysis ac oss
la ge phylogene ic dis ances because lineage-speci ic gene dupli-
ca ions end o b eak up genuine o hologs in o mul iple,
o e lapping g oups ( ha is, clus e ing p oblems a ise because
pai wise o hologies a e no necessa ily ansi i e). Using hese
g oups in ou analysis would esul in mul iple es ing o he same
clade, each ime wi h o e lapping bu incomple e sampling o
downs eam sequences. The inclusion o bo h o hologs and
lineage-speci ic pa alogs in he same g oup is, howe e , o no
conce n in ou pe -species compa ison o di e gence be ween
di e en unc ional ca ego ies o genes, because ou me hod elies
on indi idual gene ees and no a single ‘‘species ee’’ o de ec
unc ional di e gence (see below).
The e o e, we decided o build ou own g oups om he
pai wise homology assignmen s in OMA, wi h he less s ingen
equi emen ha any sequence in a g oup be connec ed o a leas
one o he sequence by pai wise homology. This s a egy p oduces
Figu e 4.
CAFS
p og am wo k low. A e alignmen s ha e been buil o each gene in he analysis, he alignmen s a e so ed by unc ional
ca ego y. In his case, he COG sys em was used [38], bu any o he on ology can be used as well. T ees a e buil o each gene using BIONJ [76] and
he JTT subs i u ion model [77], and si es a e sco ed o unc ional di e gence on each b anch. Signi icance is assessed by simula ing a dis ibu ion o
es sco es unde a model o neu al e olu ion, aking he eal phylogeny in o accoun and using he False Disco e y Ra e app oach o co ec o
mul iple es ing. Fo each species and unc ional ca ego y, we use chi-squa ed es s o e alua e whe he he species is en iched o impo e ished o
unc ional di e gence in ha ca ego y, and hen clus e species acco ding o simila i ies in hei p o ile ac oss all 19 ca ego ies. This app oach enables
us o accoun o HGT while iden i ying in e es ing and a ypical pa e ns o unc ional change in he da a, as discussed in he main ex .
doi:10.1371/jou nal.pone.0035659.g004
Func ional Di e gence Media es Adap a ions
PLoS ONE | www.plosone.o g 7 Ap il 2012 | Volume 7 | Issue 4 | e35659
g oups con aining all o hologs and pa alogs o a gi en gene, as
app op ia e o analysis o unc ional di e gence. Howe e , he
app oach is ulne able o e oneous homology calls in he o iginal
da abase, because a single alse call will cause wo un ela ed
g oups o sequences o be me ged.
To assess he possible e ec o alse OMA homology
assignmen s on ou da ase , we used he ele an genomic da a
a NCBI o assign COG on ology ags o each sequence [38,73].
We hen calcula ed he equency o he modal COG ag in each
g oup (see Figu e 5). The la ges g oup (4,788 alignmen s)
con ained only one COG ag each, alida ing ou app oach o
g ouping homologs (since COG ca ego ies a e ela i ely b oad,
ela ed sequences a e expec ed o be anno a ed wi h he same ag).
To a oid ambigui y in he clus e ing o unc ional ca ego ies, we
only analyzed hese single- ag alignmen s. We hen il e ed ou
poo ly-cha ac e ized g oups (anno a ed wi h he ambiguous R o
S COG ca ego ies) and any g oup con aining less han 9 sequences
(one ou g oup and 4 sequences downs eam o he inne nodes),
which we chose as he minimum numbe equi ed o analysis
(bo h as a equi emen o s ingency and also compa ison o
simila so wa e Gu 1999). The inal da ase comp ised 3,813
g oups, which we e hen analyzed wi h ou no el app oach
(CAFS: Clus e ing Analysis o Func ional Shi s). O he unc ional
classi ica ions such as Gene On ology (GO) can also be used.
Howe e , cau ion is equi ed because GO con ains o e lapping
ca ego ies and alignmen s wi h mul iple ags can lead o
ambiguous esul s.
Sequence alignmen s we e buil o each g oup wi h MUSCLE
[74], using he de aul pa ame e s. Da a on he ecological niches
occupied by he species included in he analysis was e ie ed om
HAMAP [75] and om he Genome da abase a NCBI.
A ypical alignmen o 78 sequences akes 2 minu es and
40 seconds o analyze o unc ional di e gence using CAFS on a
s anda d desk op compu e , including NJ ee-building. A he
o he ex eme, he la ge-scale analysis epo ed below (44,416 es s
o unc ional di e gence/3,813 alignmen s) ook 92 hou s on a 40-
node clus e .
Building gene ees
When analyzing en i e p o eomes o unc ional di e gence, he
use o a species ee o in e e en s on each b anch is p oblema ic:
ex ensi e ho izon al gene ans e (HGT), pa icula ly among
p oka yo es, means ha genomes may no be ela ed in a ee-like
way [21]. We he e o e calcula ed a ee o each gene (se o
homologous sequences) in he da ase using BIONJ [76] (see Tex
S1 and Table S6 o a jus i ica ion o he use o BIONJ and
compa ison wi h he use o maximum-likelihood ees), unde he
JTT model o p o ein sequence e olu ion [77] along wi h a
gamma dis ibu ion(n = 4, alpha = 1.0) o co ec o among si es
a ia ion o e olu iona y a es(a ixed alpha alue was used
because o he ime cons ain s in ol ed in assessing alpha and
o he pa ame e s o all alignmen s). Calcula ions o ha gene
we e hen made exclusi ely using he esul ing ee.
Sco ing unc ional di e gence
We he e de ine unc ional di e gence as he po en ial depa u e
o he de i ed p o ein unc ion om i s ances al one as a esul o
amino acid changes a impo an unc ional si es. The e o e
unc ional di e gence is de ec ed on he basis o shi s in
subs i u u ion a es pe amino acid si e in p o eins. This analysis
o unc ional di e gence can be used o p o ide a lis o candida e
genes o u he expe imen al es ing. Ou me hod iden i ies
amino acid si es wi hin a p o ein, which ha e adical subs i u ions
be ween clades and a e s a is ically signi ican . This is ca ied ou
in each o he lineages o a ee, wi h each lineage being a clus e o
4 o mo e sequences.
The me hod s eps h ough he phylogene ic ee and calcula es
unc ional di e gence sco es a each o he inne nodes. Fo each
si e o he p o ein, ou app oach compa es he amino acid
composi ion be ween wo clades o ha o an ou g oup. This
compa ison is pe o med using BLOSUM62 amino acid subs i-
u ion ma ix [78], indeed any subs i u ion ma ix can be used.
BLOSUM62 and ela ed ma ices p o ide an empi ical measu e
o he likelihood o he ansi ion o one amino acid o any o he
o he 20 (including i s conse a ion). Sco es o unc ional
di e gence (FD
sco e
) o each column a e gi en by:
FDsco e~
!
XX 1{!
XX 2
SX1{X2
ð1Þ
whe e !
XX1,2 a e he mean subs i u ion sco es o he ansi ion om
clades on ei he side o he bi u ca ion in he phylogene ic ee
ela i e o he ou g oup and SX1{X2, he s anda d e o o
unequal sample sizes wi h unequal a iances, is gi en by:
SX1{X2~ i i i i i i i i i i i i i i i i
s2
1
n1
z
s2
2
n2
sð2Þ
Figu e 5. Numbe o COG assignmen s ( ags) o each g oup o
homologous sequences. We included bo h o hologs and pa alogs
in ou sequence g oups, because we a e in e es ed in unc ional
di e gence. The majo i y o g oups consis ed o sequences ha had all
been assigned o he same COG ca ego y, sugges ing ou g ouping
s a egy did no lump oge he un ela ed sequences due o spu ious
o hology calls. Since COG ca ego ies a e ela i ely b oad, we do no
gene ally expec unc ional di e gence o cause a sequence o shi
om one ca ego y o ano he , an hypo hesis ha is also bo ne ou by
he clus e ing o ela ed sequences wi hin he same ca ego y. In ou
s udy, o a oid ambigui y we only use alignmen s in he g oup wi h a
single COG ag.
doi:10.1371/jou nal.pone.0035659.g005
Func ional Di e gence Media es Adap a ions
PLoS ONE | www.plosone.o g 8 Ap il 2012 | Volume 7 | Issue 4 | e35659
Signi icance es ing
To es he signi icance o unc ional di e gence e en s, we
simula ed mul iple sequence alignmen s o he same size as he eal
alignmen bu in which p o eins we e e ol ed unde a neu al
e olu ion model. Because unc ional di e gence was es ed in
p o ein alignmen s, he seed ances al sequence was p o ein based
and his e ol ed unde he JTT model.
Fo ou simula ions, we used he gene-speci ic ee opology and
b anch leng hs calcula ed abo e. We buil a leas 1000 such
simula ed alignmen s (mo e simula ions a e c ea ed i mean and
s anda d de ia ion ha e no con e ged wi hin a di e ence o
1*10
26
a e 1000), in each o which we sea ched o unc ional
di e gence and calcula ed a sco e acco ding o equa ion (1). To
allow o he gaps in he sequences we simula ed alignmen s wi h
he numbe o columns equal o he leng h o he inpu alignmen
minus he a e age numbe o gaps in each species. This sea ch
esul ed in a null dis ibu ion o he es sco e agains which P-
alues o he eal da a we e calcula ed. These alues we e hen
co ec ed o mul iple es ing by he False Disco e y Ra e me hod
[79] using an alpha alue o 5% as he h eshold o signi icance.
Following his p ocedu e, b anches on he ee ha s ill possess a
leas one signi ican amino acid si e we e conside ed o be unde
unc ional di e gence o he pu poses o en ichmen and
clus e ing.
En ichmen analysis
Once all alignmen s we e analyzed, we pe o med h ee
di e en en ichmen es s o ask h ee di e en biological
ques ions. These a e based on a chi-squa ed es :
x2~X
n
i~1
Oi{Ei
ðÞ
2
Ei
ð3Þ
Whe e O
i
is he obse ed equency o genes/alignmen s unde
unc ional di e gence, E
i
is he expec ed equency and n is he
numbe o possible ou comes o each e en . We used he
en ichmen es s o iden i y (i) species and (ii) ca ego ies o genes
ha expe ienced signi ican ly mo e (en iched: O
i
–E
i
.0) o
signi ican ly less (impo e ished: O
i
–E
i
,0) unc ional di e gence
when compa ed o he backg ound le el ( ha is, P,0.05 in a chi-
squa ed es ). We hen calcula ed (iii) he en ichmen s a us o each
ca ego y wi hin each species, in o de o iden i y lineage-speci ic
shi s in he pa e n o unc ional di e gence. I should be no ed
ha he chi-squa ed es used scales o he size o each o he
g oupings conside ed in ou esul s.
Hie a chical clus e ing
We c ea ed a hea map om he en ichmen s a us o unc ional
ca ego ies wi hin species o help isualize he s uc u e in ou la ge
da ase . To do his we used he hea map.2 unc ion om he gplo s
lib a y in R (R De elopmen Co e Team, 2010). This unc ion
pe o ms wo-dimensional hie a chical clus e ing acco ding o
Euclidean dis ance and ou pu s a hea map oge he wi h a
co esponding dend og am. Visualizing he esul s o he analysis
in his way allows iden i ying unusual pa e ns o unc ional
di e gence in pa icula unc ional ca ego ies o con e gen
unc ional di e gence among phylogene ically un ela ed sequenc-
es.
Implemen a ion
CAFS was implemen ed in C++ and is a ailable unde he
GNU Gene al Public License .3 o Linux, Mac and Windows.
The code was w i en using he GNU Scien i ic Lib a y and he
Bio++ lib a ies [80]. The p og am is accompanied by ull
documen a ion and enables he use o pe o m se e al di e en
kinds o analyses, including he iden i ica ion o lineage-speci ic
unc ional di e gence in a gene-o -in e es (such as ha epo ed
by Williams e al. (2010) [37]) and he kind o mul i-p o eome
in es iga ion epo ed he e. The la es e sion o he code and
documen a ion is a ailable a h p://bioin .gen. cd.ie/, a esm/
so wa e/so wa e.h ml.
Compa ison o o he applica ions and assessmen o
E o
As men ioned p e iously many p og ams ha e been de eloped
o p edic pu a i e si es o unc ional di e gence in a mul iple
sequence alignmen . O hose, he mos commonly used me hod is
DIVERGE [81], he e o e we used his as a benchma k. We ound
ha ou me hod could analyze la ge alignmen s, au oma e
p o eome scale analyses and pe o ms analyses on agging sys ems
which DIVERGE does no (Also see Table S7, Tex S2, Table S8
and Tex S3 o a de ailed compa ison o ou me hod wi h
Di e ge). Ou me hod also pe o ms simula ions, which scales he
cu o alue o each alignmen , his scaling allows us o analyze all
homologs and hence is no limi ed o o hologs. Ou p og am
au oma es he building o ees wi h a mo e de ailed model o
e olu ion and be e ee building algo i hm. Gi en ha he ee
building algo i hm is dis ance based we pe o med a compa ison
o ees buil using RaxML [82], 92% o he si es iden i ied using
he dis ance based ees we e iden ical o hose using he
maximum likelihood me hod. Gi en he di e ences be ween ou
so wa e and ha o DIVERGE i is e y di icul o make
compa isons in e ms o he si es epo ed by each p og am.
DIVERGE compa es wo clades a e a duplica ion e en and ou
so wa e compa es wo clades and an ou g oup. We do so in o de
o assess e en s o unc ional di e gence a any es able node on a
ee and also because we eel i p o ides s onge e idence o
unc ional di e gence. Ano he widely used so wa e o iden i y he
s eng h o selec ion in p o ein-coding genes is PAML. This
package employs he p og ams yn00 and codeml o es ing he
D
n
/D
s
a io. This means PAML wo ks on nucleo ide le el while
CAFS uses p o ein sequences, which gi es i he abili y o assess
di e gences o g ea e magni ude. An addi ional ea u e o ou
p og am is he abili y o easily au oma e and ca y ou la ge
analyses, as shown in his s udy. Ano he philosophical di e ence
be ween bo h app oaches is ha , o de ec posi i e selec ion, using
p og ams like PAML a s ong signal is equi ed, which would
make i di icul o de ec episodic posi i e selec ion. CAFS only
equi es an amino acid subs i u ion ixed by posi i e selec ion in a
unc ionally impo an egion o he p o ein ollowed by s ong
pu i ying selec ion.
Gi en he di icul y o inding de ini i e posi i e con ols o an
analysis o his na u e we el i impo an o demons a e ha
he e would no be a la ge alse posi i e a e. We simula ed 20
alignmen s unde a codon model wi h neu al e olu ion (non-
synonymous, Ka, o synonymous, Ks, a es a io = Ka/Ks = 1)
using he e ol e package in PAML [83]. These alignmen s we e
con e ed o amino acids and analysed wi h CAFS unde he
de aul alpha alue o 0.05, wi h his alue he expec a ion would
be 5% o si es being epo ed as unc ionally di e gen . Ou
so wa e epo ed an a e age o 2.3% o si es as unc ionally
di e gen . Gi en his pe cen age we a e con iden ha ou
me hodology o signi icance es ing and implemen a ion o alse
disco e a e is no suscep ible o a la ge numbe o alse posi i es.
Fu he de ails o his sec ion can be ound in Tex S1 and S2 and
Tables S6 and S7.
Func ional Di e gence Media es Adap a ions
PLoS ONE | www.plosone.o g 9 Ap il 2012 | Volume 7 | Issue 4 | e35659