pseudogenization dynamics in bats reveal ancestral vomeronasal signaling, then pervasive loss

13 Pages·2017·0.34 MB·English

ORIGINAL ARTICLE
doi:10.1111/evo.13187
Trpc2 pseudogenization dynamics in bats
reveal ancestral vomeronasal signaling,
then pervasive loss
LaurelR.Yohe,1,2 RamatuAbubakar,1 ChristinaGiordano,1 ElizabethDumont,3 KarenE.Sears,4,5
StephenJ.Rossiter,6 andLilianaM.Da´valos1,7
1DepartmentofEcologyandEvolution,StonyBrookUniversity,StonyBrook,NewYork11794
2E-mail: [email protected]
3DepartmentofBiology,UniversityofMassachusetts,Amherst,Massachusetts01003
4CarlR.WoeseInstituteforGenomicBiology,UniversityofIllinois,Urbana,Illinois61801
5SchoolofIntegrativeBiology,InstituteforGenomeBiology,UniversityofIllinois,Urbana,Illinois61801
6SchoolofBiologicalandChemicalSciences,QueenMaryUniversityofLondon,LondonE14NS,UnitedKingdom
7ConsortiumforInter-DisciplinaryEnvironmentalResearch,StonyBrookUniversity,StonyBrook,NewYork11794
ReceivedSeptember25,2016
AcceptedDecember30,2016
Comparative methods are often used to infer loss or gain of complex phenotypes, but few studies take advantage of genes
tightly linked with complex traits to test for shifts in the strength of selection. In mammals, vomerolfaction detects chemical
cues mediating many social and reproductive behaviors and is highly conserved, but all bats exhibit degraded vomeronasal
structureswiththeexceptionoftwofamilies(PhyllostomidaeandMiniopteridae).Thesefamilieseitherregainedvomerolfaction
afterancestralloss,orthereweremanyindependentlossesafterdiversificationfromanancestorwithfunctionalvomerolfaction.
Inthisstudy,weusetheTransientreceptorpotentialcationchannel2(Trpc2)asamolecularmarkerfortestingtheevolutionary
mechanismsoflossandgainofthemammalianvomeronasalsystem.WesequencedTrpc2exon2inover100batspeciesacross
17of20chiropteranfamilies.MostfamiliesshowedindependentpseudogenizingmutationsinTrpc2,butthereadingframewas
highly conserved in phyllostomids and miniopterids. Phylogeny-based simulations suggest loss of function occurred after bat
familiesdiverged,andpurifyingselectionintwofamilieshaspersistedsincebatssharedacommonancestor.Asmostbatsstill
displaypheromone-mediatedbehavior,theymightdetectpheromonesthroughthemainolfactorysystemwithoutusingtheTrpc2
signalingmechanism.
KEY WORDS: Pheromone,purifyingselection,relaxedselection,Trpc2,vomeronasalsystem.
Discoveringtheevolutionarytrajectoryandselectiveforcesshap- Lynch and Wagner 2010; Wiens 2011). These studies infer the
ing complex traits across the Tree of Life is a central task of re-evolution of traits after ancestral loss, sometimes multiple
evolutionary biology. The intricate mechanisms underlying or- times. Those analyses remain controversial, however, as unsuit-
ganismal complexity inspired Louis Dollo’s evolutionary “law” ableinferencemethodshavebeenused,andthetruecomplexity
in1893(Dollo1893).Dollodescribesevolutionas“irreversible” oftraitsisoftenignored(GoldbergandIgic´ 2008).Fewstudies
and“discontinuous,”andtraitcomplexitypreventsevolutionfrom have actually turned to the molecular mechanisms determining
returningorganismstotheirpreviousstate.Butphylogeniesand traitfunction(Sassietal.2007;ChristinandBesnard2009;Se-
comparativemethodstoinfercomplextraitevolutionhavechal- her et al. 2012), or the selective forces shaping traits. Here, we
lengedDollo’slaw(CollinandCipriani2003;Domesetal.2007; analyzethemolecularevolutionofamammalianchemosensory
(cid:2)C 2017TheAuthor(s).Evolution(cid:2)C 2017TheSocietyfortheStudyofEvolution.
1 Evolution
L. R. YOHE ET AL.
accessory (Fig. 2, Table S1) and nonfunctional genes for detecting
olfactory bulb
pheromones(CooperandBhatnagar1976;FrahmandBhatnagar
1980;WibleandBhatnagar1996;BhatnagarandMeisami1998;
main olfactory
epithelium BhatnagarandSmith2007;Youngetal.2010).Thislossispuz-
zlingbecausebatsarenocturnal,andoftenliveincoloniesofmany
vomeronasal
organ
individualsandexhibitadiversityofsocialsystemsofthekindre-
quiringpheromonecommunication.Basedontheseobservations,
onehypothesismightlinktheevolutionofflightandecholocation
olfactory bulb
toreducedselectivepressureonawell-developedsocialsmellsys-
tem,generatingapotentialshiftbetweenprimarysensorymodal-
ities in bats. There are, however, two major exceptions to bat
Figure 1. Schematicofasagittalinteriorviewofabatnosede-
vomeronasal loss: New World leaf-nosed bats (Phyllostomidae)
picting the two nasal chemosensory systems in mammals. The
main olfactory system primarily innervates the olfactory bulb, andlong-fingeredbats(Miniopteridae).Speciesinthesefamilies
whereas the vomeronasal organ primarily innervates the acces- have well developed vomeronasal machinery (Wible and Bhat-
soryolfactorybulb(dottedline). nagar1996;BhatnagarandMeisami1998;BhatnagarandSmith
2007; Zhao et al. 2011). Exactly how this pattern has evolved
system,examiningchangesinselectionunderlyingthelossofthis remains poorly understood and the selective forces shaping this
complexsystemthroughoutthediversificationofbats. systemareunclear.
Chemical signaling and perception among conspecifics in Inmammalianvomerolfaction,theTransientreceptorpoten-
mammals occurs via pheromones and other chemical cues de- tialcationchannel,subfamilyC,member2(Trpc2)geneisanin-
tected in the secondary olfactory system by a sense known as dicatorofafunctioningvomeronasalsystem,asitisseeminglyin-
vomerolfaction (Fig. 1). Pheromones play an important role in dispensableforpheromonesignaltransduction(Mastetal.2010).
social communication and mediate many behaviors critical to Trpc2 is a transmembrane ion channel that becomes activated
fitness,includingconspecificrecognition,courtshipandmating, after a pheromone binds to the G-protein-coupled vomeronasal
parental care, territoriality, and even the detection of sick indi- receptors in the neurons of the vomeronasal organ (Liman and
viduals (Fortes-Marco et al. 2013; Boillat et al. 2015; Mart´ın- Dulac2007).Thissignalisprimarilyrelayedtotheaccessoryol-
Sa´nchezetal.2015).Pheromonesareoftenhighlyspecific,and factorybulbinthebrain(Fig.1),whereitisinterpreted(Keverne
arekeyregulatorsofoffspringandgrouprecognitioninmammals 1999).Thedisruptionofnumeroussocialbehaviorsthroughloss
(SmadjaandButlin2009;Pitcheretal.2011,2015;Rizvanovic ofTrpc2signalinghasbeenwellestablished(Yu2015).Inrodents,
etal.2013;Cinkova´andPolicht2014;Pe´ronetal.2014).Witha Trpc2 knockout males mount both males and females, whereas
fewexceptions,vomerolfactionisubiquitousacrossmammals. knockoutfemalesmountasiftheyweremalesandreducetheir
Different environments and resources may select for fine- maternalcareforpupsbecausetheycannotdetectandprocessthe
tuned perception of one particular sense over another. Because pheromonal cues that normally mediate this behavior (Leypold
theneuralpathwaysofsensorysystemsareenergeticallyexpen- et al. 2002; Stowers et al. 2002; Kimchi et al. 2007; Hasen and
sivetomaintain(NivenandLaughlin2008),selectiononsenses Gammie2009;Yu2015).AfunctionalTrpc2istightlylinkedtoa
no longer critical to fitness may relax, and sensory organs may functioningmammalianpheromone-signalingpathway,providing
become vestigial (Fong et al. 1995). Well-known examples in- anidealmolecularmarkerformodelingvomeronasalevolution.
cludeblindnessincavefish(Jefferyetal.2003;Jeffery2005),and Pseudogenes are the genetic signatures of once-functional
loss of sweet taste in carnivores (Jiang et al. 2012). Because of systems on which selection has relaxed. Mammalian lineages
itsimportancetosurvivalandreproduction,mammalianvomerol- lackingvomeronasalstructuresfrequentlyhaveapseudogenized
factionhasbeenlostinonlyafewgroups,andinmostsuchcases Trpc2,indicatingasensenolongerundernegativeselection.Mul-
theselossesareassociatedwithchangesinecology.Forexample, tipleaquaticandamphibiousmammals,OldWorldprimates,and
independent losses in cetaceans, sirenians, and some pinnipeds 12speciesofbatshavepseudogenizedTrpc2genes(Zhangand
probablyrelatetorelaxedselectionfornasalchemosensoryper- Webb2003;Yuetal.2010;Zhaoetal.2011).Apreviousstudy
ception under water (Yu et al. 2010; Kishida et al. 2015). In notedonlythreeofthe15analyzedspeciesofbatswithanintact
contrast,lossesinOldWorldprimateshavebeenexplainedbythe Trpc2, including two phyllostomids and one miniopterid (Zhao
switchtoadiurnalniche,presumablyfollowinggreaterreliance etal.2011).Theratioofratesofnonsynonymoustosynonymous
onvisualcues(Garrettetal.2013). substitutions(ω)canbeusedtocharacterizeselection(oritsab-
The loss of vomerolfaction in bats is poorly understood. sence)onagene,withneutrallyevolvinggeneshavingequalrates
Mostbatsexhibithighlydegradedorabsentvomeronasalorgans ofamino-acid-changingandsilentsubstitutions(ω=1).Theshift
2 EVOLUTION 2017
Trpc2 INFERS LOSS, NOT GAIN, OF BAT VOMEROLFACTION
fromstrongpurifyingselectiononfunctionalgenes(ω<<1)to
relaxedselectivepressurewillgraduallyshiftωcloserto1,sothat
genesonwhichselectionhasrelaxedmorerecentlywilldisplay
lower ω than genes pseudogenized for longer. Although some
mammalianTrpc2studiesquantifyω,notinglowerωinbranches
with lineages with an intact Trpc2, most focus on the presence
and absence of disruptive mutations. Here, we use the molecu-
larevolutionofvomeronasalsignalingmechanismstoexplicitly
modelalternativeevolutionaryhypotheses,anddiscovertheevo-
lutionarydynamicsleadingtothepresent-daypatternoffunction
andlossinthevomeronasalsystem.
We propose two evolutionary scenarios for vomeronasal
loss in bats. First, and more parsimonious based on the distri-
bution of vomeronasal system loss (see Fig. 1 for morphology,
Fig.S1forparsimonyreconstruction),theancestorofallbatslost
vomeronasalfunction,withtwoinstancesofsecondarygainofthe
systemintwofamiliesofbats(WibleandBhatnagar1996).Ifso,
Trpc2 reactivated after ancestral selection had relaxed, perhaps
tothepointofinactivation.Inthesecond,andless-parsimonious
scenario, Trpc2 has remained under strong purifying selection
throughouttheearlydiversificationofbats,andhasmorerecently
lostfunctionatleastsixdifferenttimes(Zhaoetal.2011).Given
sufficientsamplingandappropriatemodels,itispossibletodis-
tinguish between these two scenarios by estimating the rates of
synonymous and amino-acid-changing substitutions of internal
branchesofthephylogeny(Sassietal.2007).
Todistinguishbetweenthesehypotheses,andilluminatethe
evolutionary forces shaping vomeronasal evolution in bats, we
modeled the evolution of Trpc2 with a large taxonomic sample
acrossnearlyallfamiliesofbats.Usingratesofmolecularevolu-
tionestimatedthroughoutthephylogeny,wesimulatedancestral
sequences to model the probability of the two alternative evo-
lutionary histories of the bat vomeronasal system. Our findings
demonstratehowmodelsofmolecularevolutioncaninformthe
Nonsense mutations occur first in the reading frame unless the
mutation is shared, for which the first shared mutation in the
readingframeislisted(Fig.S2forfulldescription).Subordersand
familiesarelabeled.SeeTableS1forreferencesformorphologi-
calcharacters.Arrowstotherespectivenodeorbranchindicate
either environmental shifts or sensory specializations or losses
(TableS3forreferences).Dichromacyistheabilitytodetecttwo
distinctcolorwavelengthsandallspecies,includingMusmusculus
Figure 2. Patterns of pseudogenization in Trpc2 exon 2 in bats aredichromaticunlessotherwisenoted.Imagenumberdesignates
sampled.Redslashesindicateeitherstopcodonorindelmutation
thecorrespondinglineageandillustration.Illustrationcredit:Ernst
that rendered the protein nonfunctional. Some mutations were
Haeckel, public domain (1, 2, 5, 11, 19); Alcide d’Orbigny, public
sharedamonglineages.Blackbrancheswereusedtocalculatethe
domain (3); G. H. Ford, public domain (4); Adria´n Tejedor, under
backgroundrates(ω )forPAMLandRELAXtests.Branches
background license(6,7,8,9,13,14,15,18);GeorgeE.Dobson,publicdomain
markedinredarelineageswithpseudogenizedTrpc2experiencing
(10,12);NicholasHolowka,withpermission(16);BonnieSumner,
relaxedselectionandwereusedasforegroundrates(ω ).
foreground withpermission(17,20).
EVOLUTION 2017 3
L. R. YOHE ET AL.
mechanismsshapingcomplextraitgainorlossinacomparative reading frame of M. musculus and the reading frame presented
framework.Besidesilluminatingtheevolutionarymechanismsof in Zhao et al. (2011) as proxies for the functional exon. Trpc2
loss of a seemingly indispensable sense for perceiving and re- was considered a pseudogene if a stop codon and/or frameshift
sponding to chemically mediated social behaviors, our analyses mutationwaspresentinthesequence.
illuminateasensorysystemwhosefunctioncontinuestopuzzle
biologists.
ANALYSES OF MOLECULAR EVOLUTION
Earlyinpseudogeneevolution,singlebasepairsubstitutionsaccu-
mulate(Marshalletal.1994).Intheabsenceofconstraints,amino-
Materials and Methods
acid-changing substitutions occur at a higher rate than when
Trpc2 AMPLIFICATION the protein was under purifying selection (Marshall et al. 1994;
Tomodelevolutionarychangesinvomeronasalsignaltransduc- Sassietal.2007).Eventually,moredramaticdisruptivemutations,
tioninbats,wegeneratednewTrpc2sequencesfor104species suchasnonsensemutationsorindelsshiftingthecodonreading
(Fig.2,TableS2),representinganorder-of-magnitudegainintax- frame, may fix (Marshall et al. 1994). Pseudogenes have likely
onomicsamplingovertheonlypreviousstudy(Zhaoetal.2011). evolvedneutrallyformanygenerations,accumulatingnonsynony-
WefocusedmostofoursamplingonfamiliesinwhichTrpc2was mouschangesasquicklyassynonymoussubstitutionsbeforestop
hypothesizedtobefunctional: Miniopteridae(n= 7)andPhyl- codonsorindelsbecomefixed.Inthecontinuedabsenceofselec-
lostomidae(n=85).Phyllostomidsareparticularlyspeciesrich, tion,theformerlyfunctionalgenewilleventuallybecomeasetof
with >170 lineages, and we sampled (cid:2)50% of known species. nucleotidesimpossibletorelatetotheproteinitonceencoded.
Usingapreviouslypublishedprimerpairforeachspecies(Zhao If a nonsense mutation or indel is present, Trpc2 may be
etal.2011),weamplifiedthesecondexonoftheTrpc2β-isoform evolving neutrally. In contrast, an intact gene reflects purifying
((cid:2)496 bp), the form of the protein primarily expressed in the selection,althoughinsomecasestheopenreadingframemayhave
vomeronasal organ (Yu et al. 2010). This is the longest exon of evolvedneutrallyintherecentpast.Wetestedforpurifyingandre-
the13-exongene((cid:2)900aminoacidsintheentiretyofTrpc2)for laxedselectionbyestimatingtheratioofratesofnonsynonymous
thisisoform,andencodesforanankyrin-repeatdomainassociated tosynonymousmutations(ω),inwhichωcloseto0indicatespu-
withprotein–proteininteraction(Yuetal.2010).Weobtainedbat rifyingselection,andωnearerto1suggestsrelaxedselection.To
tissuesamplesfromseveralmuseumcollections(TableS2formu- estimatetheserates,weusedtwomaximum-likelihoodmethods
seumaccessionnumbers).WeusedthefollowingPCRprotocol implementedinPAMLandRELAX(Yang2007;Wertheimetal.
for each 25 µL reaction: 12.5 µL Econotaq Plus Green (Luci- 2014). PAML estimates ω for different branch classes and RE-
gen),3µL10µMforwardprimer,3µL10µMreverseprimer, LAXdistinguishesifωestimatescorrespondtorelaxedorstrong
2–4 µL template DNA, and 2.5–4.5 µL dH2O. The following selection.Bothprogramsrequiredaphylogeny.Weinferredthe
protocol was used for amplification: 94˚C for 5 min; 35 cycles Trpc2 gene tree using maximum-likelihood (see Supporting In-
of94˚C30secdenaturation,(cid:2)58˚C30secannealing,and72˚C formation),butthistreewasincongruentwithestablishedspecies
extension; and 72˚C 5 min final extension. Annealing tempera- trees,likelybecauseoftheshortexonreadingframe.Therefore,
tureswereadjustedfordifferentspecies,butallprotocolshadan we used a published phylogeny of phyllostomids (Rojas et al.
annealingtemperaturebetween52and60˚C.Ampliconswerese- 2016),andcombineditwithapublisheddatedphylogenyofall
quencedusingstandardSangersequencingmethodsprovidedby bats(ShiandRabosky2015).Thetreewasprunedtomatchthe
theUniversityofArizonaGeneticsCoreFacility.Allsequences datausingthegeigerpackageinR(Harmonetal.2008).Branch
weremanuallycheckedforqualityusingGeneiousversion8.1.7 lengthswereignored,asPAMLandRELAXestimatethese.Both
(Kearseetal.2012). programsalsorequiredthecodonstructuretobemaintained,with
Forcomparativeanalyses,wecombinednewsequencesfrom nostopcodonmutationsinthereadingframe.Thus,weremoved
103specieswithpreviouslypublishedbatTrpc2exon2sequences nonsense sites by converting stop codons to gaps andremoving
(13species),andtheTrpc2exon2forM.musculus(Zhaoetal. theentirecolumnfrominsertionmutations.Deletionswerekept
2011). Two species (Pteropus vampyrus and Myotis lucifugus) intacttomaintainthereadingframe.Downstreamsequencesafter
presentinapreviousanalysis(Zhaoetal.2011)werenotincluded these mutations were kept in order to maximize the amount of
here,aspublishedsequenceswerefromadifferentexon,unavail- information in the alignment. After removing and editing these
able from published bat genomes. All sequences were aligned sites, 163 codons were available for nearly all lineages. We ex-
using the Geneious aligner. The alignment of previously pub- pectedhigherωinpseudogenizedbranchescomparedtolineages
lishedsequenceswasvisuallyinspectedandmanuallycorrected withanintactgene.
toensureouralignmentsmatchedthosefromZhaoetal.(2011). To model molecular evolution across bats, we first set up
Toidentifynonsenseandindelmutations,weusedboththeopen a null model estimating a single ω across all branches (M0)
4 EVOLUTION 2017
Trpc2 INFERS LOSS, NOT GAIN, OF BAT VOMEROLFACTION
in the PAML version 4.8e codeml routine (Bielawski and Yang Hypothesis A: Hypothesis B:
2005;Yang2007).Wethentestedtwomodelsinwhichbranches Trpc2 was retained Trpc2 was reactivated
were binned into distinct classes. First, we classified the tree
into two distinct classes (M2.1): those with nonsense and in-
del mutations (test/foreground branches) and those retaining an
intactTrpc2exon,includinginternalbranchesonthetree(back-
ground branches). Our second test classified the tree into three
distinctclasses(M2.2):thosewithnonsenseandindelmutations
(same test branches as M2.1), those terminal branches for lin-
eageswithanintactreadingframe,andallinternalbrancheson
thetree.WethenfittedthecodemlM2branchmodeltoestimate
ωforthebackgroundbranches(ω )andtest(ω )
background foreground
branchesforM2.1(BielawskiandYang2004).ForM2.2,wefit
theM2branchmodeltoestimateωforthebackgroundbranches
(ω ),andtwotestbranchclasses(ω forfunctionalfore-
background 1
Trpc2 functional
groundbranches,ω fornonfunctionalforegroundbranches).To Trpc2 nonfunctional
2
determine which model better fit the data, we used a likelihood
Figure 3. HypothesizedevolutionaryscenariosofTrpc2through-
ratiotestofM2againstthenullM0.
outthediversificationofallbatfamilies.
WealsousedtheprogramRELAXimplementedinHyPhy
version2.22totestfortwodistinctratesofωbetweenlineages
withpseudogenizingmutationsversusallotherbranches,andto recovered from the data are consistent with a null process of
distinguish positive from relaxed selection because increased ω evolutionatasingleωthroughoutthetree.
may be indicative of either (Pond et al. 2005; Wertheim et al. Second, to distinguish between alternative hypotheses of
2014). RELAX first estimates ω among three rate classes for vomeronasalevolution,wemodeledtwoscenariosofTrpc2evo-
each branch using a branch site-random effects likelihood (BS- lution. If Trpc2 were functional in the ancestor of bats and was
REL)methodandthenfitsaparameterkindicatingthestrengthof subsequently pseudogenized on multiple independent branches,
selection.ωrateclassesaretransformedbyraisingtothepower thenωestimatedforthebackgroundbranchesofthetreewouldre-
ofk(ωk),suchthatk>1pusheshighandlowclassesawayfrom flectstrongpurifyingselection,andwouldbedistinctlylowerthan
1 (indicative of strong selection) or k < 1 scales the high and estimatesforpseudogenizedforegroundbranches.Hence,values
low rate classes toward 1 (indicative of relaxed selection). We estimated from the data should fit within nonoverlapping dis-
estimated a single k (k = 1) for all branches and then fitted a tributions for background (ωbackground) and foreground branches
modelestimating kforeachofthetwobranchclassesspecified (ωforeground) simulated ω from the M2.1 model (Hypothesis A,
inthePAMLanalyses. Fig.3).
Alternatively, if Trpc2 had been reactivated twice after
evolving neutrally as bats first evolved, then Trpc2 background
SIMULATION EXPERIMENTS brancheswouldreflectthesehighωvalues,untilthetwo“reac-
We used simulations to validate our ω estimates. Using the tivation” events on the ancestral branches of the families Phyl-
evolverNSbranches routine in PAML, we simulated 1000 sets lostomidae and Miniopteridae. We simulated background and
of codon alignments modeled under the same evolutionary pa- foreground branches with increased ω values (ω = 0.5). If the
rameters as the Trpc2 data for all simulated scenarios. To test evolutionaryhistoryofTrpc2wereconsistentwiththisscenario,
therobustnessoftheobservedωestimates,weexpandedonthe then frequency distributions of background and foreground ω
frameworkofapreviousstudythatusedcodonmodelstotestfor shouldoverlap(HypothesisB,Fig.3),andvaluesestimatedfrom
relaxed selection (Sassi et al. 2007). First, to evaluate the false thedatashouldfallwithinthedistributionsofthissimulation.For
positiverate,wesimulatedalignmentsunderanullmodelcentered all simulations, we tested whether resulting distributions were
ontheestimatedM0valueofω.Thesimulatedcodonalignments distinctusingatwo-sampleStudent’st-test.
matchedthenumberofcodonsandnumberofspeciesofthealign-
ment, and the estimated branch lengths, transition/transversion
Results
ratio,andcodonfrequenciesofM0.WethenappliedM2.1branch
modelsincodemltoeachreplicatedatasettoobtainadistribution To estimate the evolutionary history of vomerolfaction in bats,
ofω andω .Thisanalysistestedwhethervalues wesequencedthesecondexonofTrpc2ofmorethan100species
background foreground
EVOLUTION 2017 5
L. R. YOHE ET AL.
Table 1. EstimatedparametersofrelaxedselectiontestsusingbranchmodelsinPAML.
Model ω ω κ TL np lnL LR P
background foreground
M0(null):oneratiomodel 0.216 0.216 6.76 6.72 231 −6054 — —
M2.1:two-branchclasses 0.096 0.541 6.80 6.71 232 −5975 157.2 <<0.001
M2.2:three-branchclasses 0.085 ω :0.541 6.80 6.71 233 −5975 157.2 <<0.001
1
ω :0.101
2
Grayindicatesthebest-fitmodelusedforinterpretationandsimulations.ω,ratioofnonsynonymoustosynonymoussubstitutions;ωbackground forM0is
ωforallbranches,forM2ωforinternalandfunctionalbranchesfortheM2,two-branchclassmodel;ω1,foregroundfornonfunctionalbranches;ω2,
foregroundforfunctionalbranches,transition/transversionrate;TL,treelength;np,numberofparameters;lnL,loglikelihood;LR,likelihoodratio;P,P-value
oflikelihoodratiotestofM2modelsrelativetothenullM0model.
of bats, including every chiropteran super family and three
quartersofthe20familiesintheorder(seeFig.2,TableS2for 400 M2.1background
M0
species sampled). The alignment spanned 163 codon positions, background
M2.1
approximately20%ofthe891aminoacidproteinsdescribedin 300 foreground
the mouse (Yu et al. 2010), and included all sites analyzed by cy M0foreground
n
e
Zhaoetal.(2011).Wedetected13independentpseudogenizing u
eq200
mutations in Trpc2 across bats (Table S4, Fig. 2), twice the Fr
numberfoundinapreviousstudy(Zhaoetal.2011),withseveral
speciescontainingmultiplepseudogenizingmutationswithinthe 100
sequence (Table S4, Fig. S2). New sequencing revealed intact
genesinallsurveyedmembersofthefamilyMiniopteridae(n= 0
7),andinamajorityofsurveyedmembersofPhyllostomidae(n= 0.00 0.25 0.50 0.75 1.00
ω
81).Surprisingly,however,twoinstancesofTrpc2pseudogeniza-
tion were recorded within phyllostomids: a small monophyletic Figure 4. Background and foreground ω estimates of null and
clade of three Caribbean endemic nectar-feeding bats shared a alternativecodemlsimulations.
singlebasepairdeletionatcodonposition88(TableS4,Fig.S2),
demonstratingarecentpseudogenizationintheancestorofthese
threespecies,andaprematurestopcodonatcodonposition154 of negative selection (Table 1). As stop codon and frameshift
inthenectarivorousphyllostomidChoeroniscusminor(TableS4; triplets were removed from the alignment to run the routines,
Figs. 2, S1). Wider taxonomic sampling also revealed pseudo- the estimates of ω severely underestimate ω, and thus
foreground
genization events in six other families, all within the suborder providealowerlimittothetruevalueofωforthesebranches.
Yangochiroptera. The ratio of the rates of nonsynonymous Thevariationinωassociatedwithpseudogenizationwascor-
to synonymous substitutions was used to model molecular roboratedwithRELAX,whichestimatestheωstatisticforevery
evolution in Trpc2. We tested whether separate estimates of branchonthetreeandaparameterktoindicatevaryingstrength
ω for nonfunctional lineages and functional/internal branches ofselection.Thenullmodelestimatesωforeachbranchofthe
(M2) were a better model fit than a single ω for all branches treebutdoesnottransformthebranches,whereasthealternative
in PAML. All models were tested with five different initial model estimates a value of k that transforms ω for two speci-
values for ω, and all parameter estimates converged on the fiedbranchclassesandtestsifthealternativebetterfitsthedata.
same values. M2.1 and M2.2 provided significantly bet- TheRELAXmodelwithakparametertransformingωacrossthe
ter fit than the null model (Table 1; M2.1: χ2(1) = 157.2, tree for two branch classes, fit significantly better than the null
P = 1.0 × 10–6; M2.2: χ2(2) = 157.2, P = 1.0 × 10–6), and (Table 2; χ2(1) = 124.8 P = 1.0 × 10–6). This indicates selec-
we were able to estimate two distinct ω values for two branch tioninlineageswithnonfunctionalTrpc2relaxedrelativetothe
classes.Thetwo-branchclassmodel(M2.1)andthethree-branch background and functional branches, as k is much less than 1
(M2.2) class model had identical likelihoods. Thus, we used (k=0.10).
M2.1forallsubsequentanalyses,asitwasthebest-fitmodelwith TheωestimatesfromthesimulatedM2.1datadidnotoverlap
the fewest parameters. Nonfunctional species had ω five times withanybackground(t=–144.8,degreesoffreedom[df]=1471,
higher (ωforeground = 0.541) than the background ω of species P=2.2×10–16)andforeground(t=132.5,df=1381,P=2.2
withfunctionalTrpc2(ωbackground=0.096),indicatingrelaxation ×10–16)estimatesfromthenullmodelsimulations(Fig.4).This
6 EVOLUTION 2017
Trpc2 INFERS LOSS, NOT GAIN, OF BAT VOMEROLFACTION
Table 2. EstimatedparametersofrelaxedselectionusingRELAX.
Model lnL np AIC TL k LR p
c
ω v.ω
background foreground
Null −5941 254 12,397 6.50 1 — —
Alternative −5878 255 12,273 6.54 0.10 124.8 <<0.001
k=1isthenullhypothesis.k>1indicatespositiveselectionandk<1indicatesrelaxedselection.lnL,log-likelihood;np,numberofparameters;AICc,
samplesizedcorrectedAkaikeInformationCriterion;TL,treelength;k,selectionintensity;LR,likelihoodratio;P,p-valueoflikelihoodratioofalternative
relativetonullforeachtest.
implies low type I error and that our estimates of ω from the calexplanation.Whiletheexplanationforvomerolfactorylossin
best-fitM2.1modelarerobust. batsremainsmysterious,ourfindingsalsohighlightthepowerof
Figure5displaysresultsofthesimulatedevolutionarysce- molecular evolution to inform the evolution of complex pheno-
narios modeled using the M2.1 branch classes (Fig. 5A), in typesacrossphylogenies.
which Trpc2 had experienced purifying selection along internal
branches(Fig.5B: x¯background =0.097;95%confidenceinterval Trpc2 LOSS IN MAMMALS
[CI]:0.096,0.097;x¯foreground=0.479;CI:0.477,0.482)orselec- Innearlyallmammals,pheromone-mediatedbehavioriscritical
tionhadrelaxedthroughoutinternalbranchesofthetree(Fig.5 tosurvivalandreproduction.Frommaternalaggressionwhenpro-
x¯background=0.545;CI:0.544,0.546;x¯foreground=0.546;CI:0.542, tectingyoungfromintruderstosexidentificationduringcourtship
0.550). The values of ωbackground and ωforeground estimated from and mating (Stowers et al. 2002; Kimchi et al. 2007; Liberles
the data (Table 1) fall well within the respective simulated dis- 2014;CoronaandLe´vy2015),andfromterritorialurine-marking
tributionsinHypothesisAandarecompletelyinconsistentwith to reproductive synchronization (Rasmussen and Schulte 1998;
HypothesisB(Fig.3).ThisstronglysuggestsTrpc2functionhas Hurst and Beynon 2004; Stockley et al. 2013), the machinery
been independently lost and was functional during early bat di- to detect and respond to pheromones is conserved and under
versification. strong selection from the platypus to platyrrhine (New World)
primates.SeveredisruptionsinsocialbehavioroccurwhenTrpc2
isrenderednonfunctional,suchasfailingtocareforoffspring,or
Discussion
mistakingmalesfromfemales,andviceversa(LimanandDulac
Pheromone chemosensation mediated by the vomeronasal sys- 2007;HasenandGammie2009;FraserandShah2014;Yu2015).
temhasrarelybeenlostinterrestrialmammals(Yuetal.2010). Unsurprisingly, this gene is highly conserved. In fact, among
Hence,thewidespreadpseudogenizationofTrpc2inbatsissur- nonvolantmammals,widespreadsamplinghasrevealedjustfour
prising,andcountertoexpectationsbasedontrait-basedanalyses independentcasesofpseudogenizationoftheTrpc2gene:inthe
ofvomeronasalfunctionandloss.Wesoughttocharacterizethe ancestor of the fin whale (Balaenoptera physalus) and dolphin
forcesshapingtheevolutionofTrpc2andvomerolfactioninbats (Tursiopstruncatus),intheriverotter(Lutralutra),intheharbor
bymodelingmolecularevolutiontotestwhetherselectionrelaxed seal(Phocavitulina),andintheancestorofcatarrhineprimates
priorto,orfollowingthediversificationofmajorextantbatfam- (ZhangandWebb2003;Yuetal.2010;Ibarra-Soriaetal.2014).
ilies.Understandingthesequenceofshiftsinselectionprovides Inbats,wefoundmorethanthreetimesasmanylossesoffunction
evidenceforexplanationsforrelaxedselection,suchasinvading asinallnonflyingmammals.Whileextantbatshaveevolved13
anewenvironmentorsensoryexchange.Wediscoveredthatthe independentnonsensemutationsinTrpc2sincetheylastshareda
evolutionary history of vomerolfaction in bats is more complex commonancestor(cid:2)60Ma(ShiandRabosky2015),similardis-
than previously appreciated. We outline three key findings: (1) ruptionshaveevolvedonlyfourtimesinthe(cid:2)150Madivergence
complete pseudogenization of Trpc2 has independently evolved amongallothermodernmammals(dosReisetal.2012).
at least 13 times across bats, including twice within one family Mapping the absence of vomeronasal structures across bat
in which most species have functional vomeronasal organs; (2) familiessupportsancestralorganlossfollowedbyasmallnumber
whilerelaxedselectionexplainsthesemultiplelossesoffunction, ofsystemgains(WibleandBhatnagar1996).Trpc2mayhavebeen
atleasttwolargecladeshaveretainedTrpc2understrongpurify- evolvingneutrallybeforeselectionstrengtheneditintheancestors
ingselection;and(3)onceTrpc2functionwaslost,itwasnever offamiliespossessingafunctionalgenetoday.Whilereactivation
regained. Unlike in aquatic mammals or mammals switching to ofapseudogeneisuncommon,itisnotimpossible(Marshalletal.
primarily visual communication, the pattern of vomerolfactory 1994;Takunoetal.2008).Apseudogenewithaframeshiftorstop
lossesinbatsseemstobedecoupledfromanyobviousecologi- codonmutationcouldberepairedbyback-mutation,apreviously
EVOLUTION 2017 7
L. R. YOHE ET AL.
A B C
100
400
Simulated outcome
ω
ωbackground
75 foreground
300 Observed outcome
y ω
enc ωbfoarcekggrroouunndd
u200 50
q
e
r
F
100 25
0 0
0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00
purifying ω relaxed purifying ω relaxed
Figure 5. (A) Background (black) and foreground (pink) branches from the M2.1 used in the simulation experiment. (B) Simulated
andestimatedratiosofnonsynonymoustosynonymoussubstitutions(ω)inTrpc2.IfTrpc2hadbeenunderstrongpurifyingselection,
simulated distributions of functional lineages and internal branches (ω ) should be distinct from simulated distributions of
background
estimates of nonfunctional lineages (ω ) and the null model. (C) If Trpc2 had been reactivated, ω estimates should
foreground background
overlapwithω .Dottedlinesindicatevaluesofωestimatedfromthedata.
foreground
deletedsegmentbyaninsertionfromacloselyrelatedgene,exon McGowen et al. 2014; Kishida et al. 2015). For aquatic mam-
shuffling,orgeneconversion(Doxiadisetal.2006;Takunoetal. mals,ashifttolifeunderwatermadeitimpossibletodisperseand
2008). detect pheromonal cues in a manner similar to terrestrial mam-
Alternatively,strongpurifyingselectiononTrpc2couldhave mals. Cetaceans and sirenians have highly degraded olfactory
persistedduringdiversificationofallmajorbatlineages,andsub- andvomeronasalmorphology;dolphinsandwhalesshowances-
sequentlyrelaxedindependentlyinmultiplebatfamilies.Losses tralpseudogenizationofTrpc2,andcarnivoranharborsealsand
offunctionincriticalvomeronasalsignaltransductionhavebeen sea otters have pseudogenized Trpc2 (Mackay-Sim et al. 1985;
independent, with the first loss occurring (cid:2)50 Ma or before Yu et al. 2010; McGowen et al. 2014). Bats are the only mam-
(Table S4, Fig. S2). Estimates of ω for internal branches and mals capable of powered flight, but this shift seems unrelated
functionallineages(ω )weremuchlowerthanthosefrom to vomeronasal loss. Airborne pheromonal cues and chemical
background
lineages with detectable loss-of-function mutations (ω ), cuestransmittedduringroostinginteractionsarestilleffectivein
foreground
indicatingselectionhasrelaxedinmanybranches.Theωvalues the aerial environment of bats. Bats also possess a diversity of
estimatedfromthedatawerecompletelyinconsistentwithsimu- external glands, many of them sexually dimorphic, that excrete
lationswithconstantlowω(uniformstronglynegativeselection) scented compounds, suggesting chemical cues are widely used
orhigherω(uniformrelaxedselection),andwereconsistentwith (Scully et al. 2000). Finally, the evolution of flight in the an-
simulations of distinct ω background and foreground branches. cestorofextantbatsisdecoupledfromthemultipleindependent
Together,theseresultssupportstrongpurifyingselectiononTrpc2 pseudogenizationeventsdetected(Fig.2).
haspersistedinbothinternalbranchesandextantfunctionallin- A shift from nocturnal to diurnal habits has also been
eages. Additionally, the narrow distribution and low value of suggested to change selective pressure on vomerolfaction.
ω indicates selection did not relax in those branches, Thoughnotalldiurnalprimateslackvomerolfaction,allnoctur-
background
whilethehighervalueandgreatervarianceofω reflects nal primates have functional vomeronasal systems (Evans and
foreground
thevaryingageofpseudogenizationeventsacrossthephylogeny Schilling1995;BhatnagarandMeisami1998;Smithetal.2011;
(Fig.5B):recentpseudogenizationeventshavenotaccumulated Garrettetal.2013).Nocturnalmouselemurs,inparticular,show
asmanymutationsasolderevents.Insteadofreactivation,quan- diversifyingselectiontoexpandvomeronasalreceptorrepertoires
titative analyses of evolutionary rates support the retention of (Youngetal.2010;Yoderetal.2014).Severalstudieshavenoted
functionalTrpc2intwobatfamilies. detecting chemical cues in low-light conditions is under strong
selection among nocturnal mammals, which have refined and
CHANGES IN ENVIRONMENT DID NOT INFLUENCE well-developed nasal chemosensory systems and a more dis-
Trpc2 LOSS IN BATS criminating sense of smell (Barton et al. 1995; Wang et al.
Change in environment is one of two main explanations 2010;Garrett etal.2013).Assomeprimates evolveddiurnality
of vomeronasal loss and pseudogenization of Trpc2 among and became more visual, the selective pressure to maintain a
nonvolant mammals (Zhang and Webb 2003; Liman 2006; well-developed social chemosensory system may have relaxed.
8 EVOLUTION 2017
Trpc2 INFERS LOSS, NOT GAIN, OF BAT VOMEROLFACTION
This explanation for the loss of vomerolfaction is unlikely to Meisami1998).Similarly,apreviousstudyfoundnosupportfor
apply for bats, as nearly all bats are nocturnal and most lack anexchangebetweendichromaticvisionandvomeronasalfunc-
vomerolfactorysystems. tion(Zhaoetal.2011;Jonesetal.2013).Finally,andmostsur-
prisingly,phyllostomidswithwell-developedvomeronasalorgans
SENSORY SYSTEM EXCHANGE DID NOT INFLUENCE alsohavelargernumbersofdistinctolfactoryreceptors,indicat-
Trpc2 LOSS IN BATS ing a lack of exchange between olfaction and vomerolfaction
Sensory exchange is another mechanism through which selec- (Hayden et al. 2014). With increased sampling of Trpc2 among
tioncanchangeovertime,perhapsshapingthecuriouspatterns phyllostomid bats with diverse sensory adaptations, we find no
of vomerolfaction in bats (Zhao et al. 2011; Jones et al. 2013; support for a simple sensory exchange between vomerolfaction
Hayden et al. 2014). Maintaining neural sensory tissue is ener- andothersenses.Ourfindingschallengenovelenvironmentsor
geticallyexpensive(NivenandLaughlin2008).Ifenvironmental sensoryexchangeasexplanationsforthelossofvomerolfaction,
demandsselectforspecializationoradaptationinanothersense, andopenupnewinterpretationsformechanismsofloss.
natural selection on a less constrained sensory system may re-
lax. The relationship between trichromatic vision in primates MAIN OLFACTORY SYSTEM MAY COMPENSATE FOR
and decreased numbers of olfactory receptors is one proposed VOMERONASAL LOSS
example of sensory exchange (Gilad et al. 2004), though this Compensation between the main olfactory and vomerolfactory
exampleremainscontentious(Matsuietal.2010).Mostbatspri- systemisahithertounexploredmechanismfortherelaxationof
marily forage using echolocation and many communicate with selection on vomerolfaction and its eventual loss. Instead of an
ultrasonicsocialcallsandhaveevolvedaremarkablediversityof increaseorexpansioninthemainolfactorysystemcorrelatedwith
adaptationsforacutehearing(Jonesetal.2013),suggestingasen- vomeronasalloss(ahypothesisrejectedbyHaydenetal.2014),
soryexchange.Butthereisnoevidenceofdecreasedrelianceon olfactionandvomerolfactioncouldberedundant,allowingselec-
olfaction among echolocating bats, or of decreased importance tivepressuresonvomerolfactiontorelaxinsomespecies.Recent
of echolocation among bats with vomerolfaction. For example, neurobiological, developmental, and behavioral studies suggest
Tadaridabrasiliensis,anefficientecholocatinginsectivore,hasa there is more interaction between the two nasal chemosensory
disruptedTrpc2,butmothersstillscent-marktheirpupstoiden- systems than previously thought (Sua´rez et al. 2012; Baum and
tify them within the crowded colonies (Gustin and McCracken Cherry2014;OmuraandMombaerts2014;Ma2015).Inmice,
1987).Additionally,allminiopteridshavewell-conservedTrpc2 for example, some pheromone receptors may also be present in
genes, but also rely on vocalizations to locate their insectiv- themainolfactoryepithelium(OmuraandMombaerts2014;Kan-
orous prey (Miller-Butterworth et al. 2007). The evolution of ageswaranetal.2015),possiblyexplainingwhysomeapparently
constant-frequency echolocation with Doppler shift compensa- pheromone-mediated behavior is maintained in species lacking
tion is an extreme sensory adaptation that enables focusing on thevomeronasalorgan.WhileTrpc2hasbeenpseudogenizedin
insectpreydespiteaclutteredbackground(Fentonetal.2012).If theharborsealandriverotter(Yuetal.2010),bothspeciesretain
this adaptation resulted in sensory exchangefor vomerolfactory awell-developedmainolfactorysystemandbehavioralstudiesin-
function, the pseudogenization of Trpc2 should roughly corre- dicatethepresenceofsomeolfactorysocialcommunication(van
spondwithitsevolution.Thispatternisnotevident:mostlossof Valkenburghetal.2011;Stoffeletal.2015).Finally,thedebateon
functionmutationsoccurredamongnon-constant-frequencybats, whetherhumansarecapableofdetectingpheromonesisstillfar
and Pteronotus parnellii —the only New World bat capable of fromresolved,asanatomicaladvancesareongoing(Smithetal.
constant-frequencyecholocation—stillpossessesanintactTrpc2 2014; Wessels et al. 2014; Vasuki et al. 2016), and behavioral
geneandvomeronasalstructures(BhatnagarandMeisami1998; experimentssuggestthatwecanperceiveconspecificchemosen-
Fentonetal.2012). sorycues(RadulescuandMujica-Parodi2013;Lu¨bkeandPause
Some bats have other extreme sensory adaptations besides 2015;Gelsteinetal.2011).
echolocation (Jones et al. 2013). Many of these adaptations Wealsofindsimilarcuriousexceptionsacrossbatslacking
are found within phyllostomids, one of two families with con- vomeronasalorgans.Despitethegregariousnessofmanybats—
served vomerolfaction. For example, the common vampire bat somespeciesresideincaveswithmillionsofindividuals(Kunz
(Desmodus rotundus) is the only mammal capable of infrared 1982; Iskali and Zhang 2015), while others in smaller roosts
thermoperception (Gracheva et al. 2011), and UV vision has maintain stable relationships among individuals (Kunz et al.
evolvedinthenectarivorousGlossophagasoricina(Winteretal. 1994)—most lack the vomeronasal organ that mediates social
2003;Mu¨lleretal.2009).Nevertheless,bothofthesespecieshave interactions in most mammals. Relying only on the main
well-developedvomeronasalorgansandanintactTrpc2(Frahm olfactorysystemmaybesufficientformanyanimalsthatspend
andBhatnagar1980;WibleandBhatnagar1996;Bhatnagarand part of their lives underwater or in the air, as long as some
EVOLUTION 2017 9
L. R. YOHE ET AL.
of the chemical detection mechanisms in the vomeronasal and DATA ARCHIVING
main olfactory systems are redundant. The loss of a sensory Models, code, and input for analyses are deposited on Dryad (DOI:
http://dx.doi.org/10.5061/dryad.gr1pk);newsequencedataaredeposited
system does not preclude processing of the signal in some
toGenbank(KX537508-KX537612).
other way; indeed many “earless” frogs can still detect sound
(Lindquistetal.1998;Boisteletal.2013).Ifthishypothesiswere
correct, the main olfactory system should contain mechanisms LITERATURE CITED
todetectpheromoneligands,andexpressvomeronasalreceptors. Barton,R.A.,A.Purvis,andP.H.Harvey.1995.Evolutionaryradiationof
visualandolfactorybrainsystemsinprimates,batsandinsectivores.
Functional assays testing for chemosensory redundancy offer a
Philos.Trans.R.Soc.LondonSer.BBiol.Sci.348:381–392.
promisingavenueforfuturework.
Baum,M.J.,andJ.A.Cherry.2014.Processingbythemainolfactorysystem
ofchemosignalsthatfacilitatemammalianreproduction.Horm.Behav.
68:53–64.
Conclusion Bhatnagar, K. P., and E. Meisami. 1998. Vomeronasal organ in bats and
primates:extremesofstructuralvariabilityanditsphylogeneticimpli-
Because of its importance to social communication, vomerol-
cations.Microsc.Res.Tech.43:465–475.
faction is rarely lost in mammals. Here, we have identified the Bhatnagar,K.P.,andT.D.Smith.2007.Lightmicroscopicandultrastruc-
greatest number of losses of function in this crucial system in tural observations on the vomeronasal organ of Anoura (Chiroptera:
Phyllostomidae).Anat.Rec.(Hoboken)290:1341–1354.
anymammalianorder,andquantifiedandmodeledthemolecular
Bielawski,J.P.,andZ.Yang.2004.Amaximumlikelihoodmethodforde-
evolutionaryhistoryofthekeyregulatorofvomeronasalsystem
tectingfunctionaldivergenceatindividualcodonsites,withapplication
function,Trpc2.Theseanalysesreveallosseshavebeenindepen- togenefamilyevolution.J.Mol.Evol.59:121–132.
dentandvariedthroughtime,incontrastwithstrongandsustained ———.2005.Maximumlikelihoodmethodsfordetectingadaptiveprotein
purifyingselectionintwobatfamiliesandtheirancestors.Neither evolution.Pp.103–124inStatisticalmethodsinmolecularevolution.
Springer,NewYork.
shiftstonovelenvironmentsnorpreviouslyproposedsensoryex-
Boillat,M.,L.Challet,D.Rossier,C.Kan,A.Carleton,andI.Rodriguez.
changesexplainvomerolfactorylossinbats.Instead,wepropose
2015. The vomeronasal system mediates sick conspecific avoidance.
the olfactory system may compensate for vomeronasal loss, as Curr.Biol.25:251–255.
manybatswithoutTrpc2stilldemonstratepheromone-mediated Boistel,R.,T.Aubin,P.Cloetens,F.Peyrin,T.Scotti,P.Herzog,J.Gerlach,N.
Pollet,andJ.-F.Aubry.2013.Howminutesooglossidfrogshearwithout
behaviors.ThemolecularevolutionofTrpc2inbatssuggeststhat,
amiddleear.Proc.Natl.Acad.Sci.USA110:15360–15364.
inlinewithDollo’slaw,oncevomerolfactionislost,itdoesnot
Christin,P.A.,andG.Besnard.2009.TwoindependentC4originsinAris-
evolveback. tidoideae (Poaceae) revealed by the recruitment of distinct phospho-
enolpyruvatecarboxylasegenes.Am.J.Bot.96:2234–2239.
Cinkova´,I.,andR.Policht.2014.Discriminationoffamiliarityandsexfrom
ACKNOWLEDGMENTS chemicalcuesinthedungbywildsouthernwhiterhinoceros.Anim.
ThisworkwassupportedbytheNationalScienceFoundationGraduate Cogn.18:385–392.
Research Fellowship and NSF-DEB 1442142. The Indiana University Collin,R.,andR.Cipriani.2003.Dollo’slawandthere-evolutionofshell
Mason server funded through NSF-DBI 1458641 provided computing coiling.Proc.Biol.Sci.270:2551–2555.
resourcesforthesimulationsinthisstudy.Fortissuesamples,wethank Cooper, J. G., and K. P. Bhatnagar. 1976. Comparative anatomy of the
thefollowingindividualsandcollections:N.B.Simmonsandstaffofthe vomeronasalorgancomplexinbats.J.Anat.122:571–601.
AmbroseMonellCryo-CollectionattheAmericanMuseumofNatural Corona,R.,andF.Le´vy.2015.Chemicalolfactorysignalsandparenthoodin
History;B.PattersonandJ.D.PhelpsattheFieldMuseum;J.Cook,J. mammals.Horm.Behav.68:77–90.
Dunnum,andM.CampbellattheMuseumofSouthwesternBiology;V. Dollo, L. 1893. The laws of evolution. Bull. Soc. Belge Ge´ol. Pale´ontol.
PachecoattheMuseodeHistoriaNaturalattheUniversidadNacional Hydrol.7:165–166.
MayordeSanMarcos;C.ConroyattheMuseumofVertebrateZoology Domes,K.,R.A.Norton,M.Maraun,andS.Scheu.2007.Reevolutionof
oftheUniversityofCalifornia,Berkeley;K.HelgenattheSmithsonian sexuality breaks Dollo’s law. Proc. Natl. Acad. Sci. USA 104:7139–
Institution;B.LimandJ.EngerattheRoyalOntarioMuseum;andR.J. 7144.
BakerandtheMammalogystaffattheMuseumofTexasTechUniversity. dosReis,M.,J.Inoue,M.Hasegawa,R.J.Asher,P.C.J.Donoghue,andZ.
Wethank E. Leung for assistancewithlaboratory work. Wethank the Yang.2012.Phylogenomicdatasetsprovidebothprecisionandaccuracy
University of Arizona Genetics Core Facility for Sanger sequencing. inestimatingthetimescaleofplacentalmammalphylogeny.Proc.R.
Forhelpfulcommentsandsuggestionsonthemanuscript,wethankN. Soc.BBiol.Sci.279:3491–3500.
Holowka, two anonymous reviewers, and B. Igic´ for suggestions and Doxiadis,G.G.M.,M.K.H.vanderWiel,H.Brok,N.Groot,N.Otting,
stimulatingdiscussion.WethankN.Holowka,B.Sumner,andA.Tejedor B.’tHart,J.vanRood,andR.Bontrop.2006.Reactivationbyexon
forcontributingillustrations. shufflingofaconservedHLA-DR3-likepseudogenesegmentinaNew
Theauthorscontributedasfollows.LRYconceivedtheideas,designed Worldprimatespecies.Proc.Natl.Acad.Sci.USA103:5864–5868.
andexecutedthestudy,collectedandanalyzedthedata,andwrotethe Evans,C.,andA.Schilling.1995.Theaccessory(vomeronasal)chemorecep-
manuscript. RA and CMG assisted in data cxollection and edited the torsysteminsomeprosimians.Pp.393–411inL.Alterman,A.Gerald
manuscript.LMDhelpeddesignthestudy,collectedsamples,andwrote Doyle,andM.K.Izard,eds.Creaturesofthedark.Springer,York,NY.
themanuscript.SJRassistedinsamplecollection.SJR,KES,andERD Fenton,M.B.,P.A.Faure,andJ.M.Ratcliffe.2012.Evolutionofhighduty
wereinvolvedinfundingacquisitionandprovidedfeedback. cycleecholocationinbats.J.Exp.Biol.215:2935–2944.
10 EVOLUTION 2017