(Anisoptera: Libellulidae)

The morphological ‘forms’ of

Palpopleura lucia

(Drury)

areseparate species

as evidenced

by

DNA

sequencing

(Anisoptera: Libellulidae)

A. Mitchell¹ and M.J . Samways²

ReceivedApril2,2004 / Revised andAcceptedOctober22,2004

INTRODUCTION

Intense debateoverthe taxonomicstatusof

Palpopleura

lucia

(Drury)

has con-tinued since the

description

of P. lucia and P.

portia

asseparate species in 1773

(P1NHEY,

1962).

PINHEY

(1951,

1985)

treated “lucia

“portia,”

and the West African

“graffei”

_as “forms” of P. lucia. Current_taxonomy

posits

two

subspecies,

P. I. lucia and P. I.

portia.

While females of thetwo

subspecies

aredifficultto

distin-1School of Molecularand CellularBiosciences, UniversityofNatal,PrivateBag X01, Scottsville, 3209,SouthAfrica;— e-mail:[email protected]

2

DepartmentofEntomologyandNematology, FacultyofAgriculturalandForestry Sciences, UniversityofStellenbosch,PrivateBag XI,7602Matieland,SouthAfrica;

-e-mail;[email protected]

P. lucia isawidespreadAfrican sp. withacheckered taxonomichistory.Currently2 ‘forms’orsubspecies,P. l. lucia and P. l.portiaarerecognized, althoughdebateoverthe

taxonomicstatusof these taxahashardlylet_upoverthe last 230years.The 2 ‘forms’ showdistinctive _wingpatterndifferences_althoughother_aspectsof their_morphology

areverysimilar.Theycan occurhighly sympatricallyatsomelocalities in southern

Af-rica,aswell aselsewhere,thusraisingthequestionof whethertheyaretwospeciesor oneperhapswith balancedpolymorphism.DNA_sequencedata from the ITS2 and COI geneswerecollected fromspecimensof both these ‘forms’toassess morerigorouslythe taxonomicstatusof thesetaxa.Thecloselyrelated P.deceptor (Calv.)andP.jucunda (Ramb.)wereincluded in the dataset toprovideabaseline forcomparisons. Specimens from all 4taxawerefrompoolsof the floodplainof the SabieR., KrugerNational Park,SouthAfrica,andwerepotentiallyabletointerbreed. Bothphylogenetic

analy-sesandcomparisonsof_sequencedivergencelevels_{strongly support}thehypothesisthat the 2 ‘forms’ ofP. luciaarereproductivelyisolatedand should be accorded fullspecies

guish,

the males show distinc-tive

patterning

_on the

wings

(Fig.

1).

The

subspecies

are

sympatric

overmuch of their

geographical

ranges, from the Eastern

Cape

toEast and West

Africa, suggesting

that

they

are not

biological

sub-species. Although

atsome lo-calities both_{occur as} adults

atthesame

pool, although

P. l.lucia overallseemsto

prefer

warmerconditions

(PINHEY,

1985).

The

question

of their taxonomicstatus is also

eco-logically significant,

because if

they

wereindeed

nonspecific,

this would

possibly

be balanced

polymorphism,

unknowntodate in theOdonata,

raising

the

question

of what selection pressures

might

be

responsible

for

maintaining

such

polymorphism.

We

investigate

here the taxonomicstatusof thesetaxa

using

DNA

sequencesof the second internal transcribed

spacer

region (ITS2)

of the nuclear ribosomal DNA repeat

region,

and the mitochondrial

cytochrome

oxidase I

(COI)

_gene.

MATERIAL AND METHODS

WesequencedDNAfrom21malespecimens, includingtenspecimensofP.I. lucia

,fivespecimens of P. I.portia,three specimensofP.jucunda (Ramb.)and threespecimensofP.deceptor (Calv.). Im-portantis that allspecimenswerecollected onthesameday (14March 2002 withinakm of eachother, and thereforehighly sympatric, alongthe floodzoneof the Sabie_{River, Kruger}National Park,South Africa(24° 59’S,31°28’E.,320ma.s.1.),andpreserved directlyinto 100% ethanol.

TotalgenomicDNAwasextracted fromapproximately20 mgofflightmusclesbymeansof the QiagenDNeasy™Tissue Kit.

For the ITS2_gene,initial PCRamplificationsusedtheprimersITS5 (5’-GGAAGTAAAAGTCG-TAACAAGG-3’)and 28S-25R(5’-TATGCTTAAAYTCAGCGGGT-3’),yieldingPCRfragmentsof

ap-proximately900bpinlength.EachPCRfragmentwasfractionatedon a1%agarosegel,2-5piofgel

con-tainingthe PCRfragmentwasremoved withasterilemicropipette tipand used inareamplification

reac-tionusinganinternalupstreamprimer,ITS2-1F_{(5’-CATGAACATCGACATYTTGAACGC-3’),}and theoriginaldownstreamprimer (28S-25R) yieldingaPCRfragmentofapproximately520bp.Theprimers

used for the COI_genewereJerry/Cl-J-2183(5’-CAACATTTAliltGATTTTTTGG-3’)and_Georgina/ Cl-N-2786_{(5’-GGATAATCTGAATAWCGWCG-3’), yielding}PCR_fragments647_bpin_length.

PCRwasperformedon aPerkin ElmerGeneAmpPCRSystem2400 under thefollowingconditions:

94°C for 3minutes,32cycles (or24cycles forreamplifications)of(94°Cfor 30_sec.,55°C for 30_sec,72°C for 60-90sec),72°C for 7min,4°C hold. Reaction volumes were30-50pl,and the reaction mixture

con-tained:IXPCRbuffer,2mMMgCl,, 200pMof eachdNTP,10pmolof each PCRprimer,0.7 units of RocheTaqDNApolymerase,andapproximately 250ngofgenomicDNA/RNA.PCR reactionswere

cleanedusing QiagenPCRpurificationcolumns and directly sequencedin both directions usingthe ABIBigDye™Terminatorv3.0Cycle SequencingKit, followingthe manufacturer’srecommended Fig. 1. Wingpatterningof_Palpopleural.lucia(above)and

conditions.Sequenceswerevisualized on anABI 3100 GeneticAnalyser.

DNA_{sequencechromatograms}wereedited andcontigsassembled_usingthe Staden_package (STA-DEN, 1996).Consensus sequences wereautomatically aligned usingClustalX(THOMPSONetah, 1997)and then manuallycorrectedusingSe-Al v2.0a7(RAMBAULT, 2001). Phylogenetic analyses

wereperformedunder both the maximumparsimony(MP)and maximum likelihood(ML)criteria

us-ingPAUP—4.0M0(SWOFFORD, 1998).A10,000iterationpartition homogeneitytestwasfirst per-formed. Each datasetwasanalysed separately,andfinally,the combineddatasetwasanalysed.Branch and bound searcheswereperformedunderMP,while200-replicaterandom addition sequence heuristic searcheswereperformedunder ML. Theextentofsupportfor internal nodeswasestimatedbymeans

of bothparsimony-basedanddistance-based (ML-model) bootstrap analyses,eachemploying 1,000 iterations.FollowingFRATIetal.(1997),16 different models of sequence evolutionwerecomparedfor their fit_toeach dataset(ITS2,COI and combineddata)bymeansof likelihood ratio tests.The model chosenforMLanalysiswastheonewith the least number of freeparameters,and whichwasnot sig-nificantlydifferent from themostcomplexmodel,thegeneraltime-reversiblemodel(GTR +1+G).For all three data sets, the best modelprovedtobe theHasegawa,Kishino and Yano model(HASEGAWA etal., 1985),withgamma-distributedrates(HKY85+G).Model_parameterswereestimated from the

data for each datasetseparately.Alternativephylogenetic hypotheseswereassessedby performingML constraint searchesandcomparingtheresultingtreeswith the ML tree,usingKishino-Hasegawa (K-H)andShimodaira-Hasegawa(S-H)tests,asimplementedin PAUP*4.

RESULTS AND DISCUSSION

All DNA sequences

reported

hereare

deposited

in GenBank

(accession

numbers ITS2: AY582759-AY582777;COI;

AY582778-AY582796).

The datasetconsisted of 474 nucleotides of ITS2 sequence and 595 nucleotides of COI sequence, fora total of 1069 sites. For ITS2 there_were59 variable sites

(12.4%

ofthe

total)

while for COI therewere 143 variable sites

(24%).

Of the 21

specimens

sampled,

two lacked ITS2 sequence and anothertwo _{lacked COI sequence, due}to

sequencing

difficulties.

Analyses

of the combined datasettherefore used

only

the 17

complete

Comparison ITS 2 COI

mean s min max n1 mean s min max n

Within_deceptor - - ... 0.78 0.09 0.67 0.84 3 Within_jucundu 3.40 2.43 0.68 5.34 3 Withinportia 0.30 0.13 0.28 0.46 3 2.72 1.29 0.67 4.37 10 Within lucia 0.86 0.46 0.22 1.59 15 0.74 0.36 0.17 1.51 28 luciavsportia 3.66 0.18 3.28 3.91 18 10.82 0.32 10.25 11.61 40 jucundavslucia 4.35 0.16 4.14 4.60 6 10.17 0.47 9.16 11.24 24 jucundavsportia 3.75 0.00 3.75 3.75 3 10.37 0.45 9.67 11.11 15 deceptorvsjucunda 9.76 0 9.76 9.76 1 10.71 0.43 10.09 11.27 9 deceptorvslucia 9,17 0.36 8.83 9.72 6 13.46 0.26 12.94 14.02 24 deceptorvsportia 9.48 0.17 9.40 9.61 3 12.07 0.50 11.09 12.77 15

1For ITS2anumber ofspecimensshowed identical sequences (e.g,withinP.deceptorandP.jucunda) and allduplicatesequenceswereeliminated beforecomparisonofdivergencelevels.

TableI

Uncorrectedpairwise divergencevalues(%)

Comparison

mean s

ITS 2

min max n1 mean s

COI min max n Within_deceplor _ _ _ _ - 0.78 0.09 0.67 0.84 3 Withinjucunda - - - 3.40 2.43 0.68 5.34 3 Withinportia 0.30 0.13 0.28 0.46 3 2.72 1.29 0.67 4.37 10 Within lucia 0.86 0.46 0.22 1.59 15 0.74 0.36 0.17 1.51 28 luciavsportia 3.66 0.18 3.28 3.91 18 10.82 0.32 10.25 11.61 40 jucundavslucia 4.35 0.16 4,14 4.60 6 10.17 0.47 9.16 11.24 24

jucundavsportia 3.75 0.00 3.75 3.75 3 10.37 0.45 9.67 II.11 15

deceptorvsjucunda 9.76 0 9.76 9.76 1 10.71 0.43 10.09 11.27 9 deceptorvslucia 9.17 0.36 8.83 9.72 6 13.46 0.26 12.94 14.02 24 deceptorvsportia 9.48 0.17 9.40 9.61 3 12.07 0.50 11.09 12.77 15

sequences.The

partition homogeneity

testwas

non-significant (p

=

0.095),

there-fore the datawere

analysed

both

separately

and in combination.

Table I summarizes the uncorrected

divergence

values observed in

pairwise

com-parisons

among all sequences. For ITS2 sequences, the average

pairwise divergence

levels between P. l. lucia and P. l.

portia

arenot

statistically

differenttothose be-tween these

subspecies

and P.

jucunda.

In

fact,

the maximum

divergence

observed between

specimens

of P I. lucia and P. I.

portia (3.91%)

is_greater than the maxi-mum

divergence

observed between P.

jucunda

and P. I.

portia (3.75%).

The_more

distantly

related P.

deceptor

shows 8.8- 9.7%

divergence

from all othertaxa. A

similar_patternisseenfor COI sequences, with themeanand maximum

divergenc-es between P. I. lucia and P. I.

portia (10.8%

and 11.6%

respectively) being

greater than the

corresponding

values for

comparisons

between either

subspecies

and P.

jucunda (10.3%

and 11.1%

respectively). Again, comparisons involving

P

deceptor

give larger

values

(up

to

14.2%).

For COI data

alone, parsimony analysis

ofthe 17taxa dataset

produced

30 short-est treesof

length

=202_steps,Cl

(excluding

uninformative

characters)

=_0.77,and RI=

0.91.

For ITS2 data

alone, parsimony analysis

of the 17taxadataset

produced

78 shortesttreesof

length

=72_steps,Cl

(excluding

uninformative

characters)

=

0.96,

and RI =0.98. Both

analyses

recovered the four nominatetaxawith >90%

bootstrap

supportin everycase. The branches that

collapsed

in the strict consen-sus treeswerethose

indicating relationships

within the fourtaxa.

Analyses

of the

complete

19taxadatasetsfor each_genewereinagreementbutgavenoadditional

information,

therefore

they

_arenotconsidered further here. Maximum likelihood

analyses

gavealmost identical resultstothe MP

analyses,

the

only

differences

be-ing

in the

weakly supported relationships

within the fourtaxa.

Figure

2A summarizes the

relationships

_{recovered among the four}taxafor the COI data

alone,

under both MP and ML

criteria,

while

Figure

2B shows thesame

forthe ITS2 data alone. Thesetreesdiffer in their

placement

of P. I.

portia.

The COI data

places

P. I. lucia and P. I.

portia

assister groups, with

bootstrap

support of 82% under ML and 69% underMP,while the ITS2 data

places

P I._portiaand Fig.2.Summaryofrelationshipsamongthe foursympatric taxarecovered under both MLand MP criteria._{Bootstrap support}levels underMLandMPcriteriaareshown above and

be-lowbranches, respectively: (A)COI dataonly;

-(B)ITS2 dataonly. Palpopteura

P.

jucunda

as sister groups, with

bootstrap

supportof 66% under ML and72%under MR

The combined dataset recov-ered the_same

relationships

among taxa as the COI

analysis,

with

Figure

3

clearly showing

the

un-ambiguous

and broad

separation

of the

subspecies

of P. lucia. Vis-ual

inspection

of internalbranch

lengths

confirms the initial

find-ings suggested by

examinationof

pairwise divergence levels,

thatthe distance between P. I. lucia and P. l.

portia

isatleast_asgreatasthat between either taxonand P.

ju-cunda. The

monophyly

of each

subspecies

also is

strongly

sup-ported.

Totestthe

significance

of dif-ferences in MLscorebetween the

competing hypotheses,

heuristic searcheswere conductedon the COI

(and combined)

datasets while

constraining

the

topology

tothat shown in

Figure

2B

(ob-tained with the ITS2

data),

and vice versa.

Resulting

trees were

compared

tothe MLtreefor each dataset

by

meansof K-H and

S--F1tests

(Tab. II).

All of theteststatistics_were

non-significant. Thus, despite

seem-ingly

reasonable

bootstrap

_supportvalues for both of the

competing topologies,

noneof the datasetshas the

resolving

_powerto

distinguish

_{among them.} Further-more,reverse constraint searches in which either P. l. luciaorP. l.

portia

were con-strainedtobenot

monophyletic yielded

treeswhichwere

significantly

different from the MLtrees

(K-H

tests, p=

0.000-0.003;

S-H tests, p=

0.001-0.018).

These results_suggestthat thetwo

subspecies

of P. lucia each should be accorded full

species

status.

Specimens

ofthesetaxawerecollectedfrom

sympatric

popula-tions, apparently

with full

opportunity

for

interbreeding. Indeed,

_some

specimens

of P. I. lucia and P. I.

portia

were evencollected withinafew metres ofeachother from thesame

pool. Sampling

of such

specimens provides

avery conservativetestof the taxonomicstatusof thesetaxa. That

is,

while limited

geneflow between thetaxa Fig. 3. MLtreeforthe combined data set, -In L=

2,949.27958.HKY85 modelparameterswere asfollows: A=0.25575,C= 0.22220,G=0.22755,T=0.29450; Ts/Tv ratio=3.54019(k=_{7.1506344); shape parameter}

atthis

locality

wouldnot

necessarily

indicate that

they

were

conspecific,

the_apparentlack of_geneflow between thetaxa

provides

very strongevidence that

they

are

reproductively

isolated,

and therefore full

species.

Instead of

sampling specimens

fromacross the

geographical

_{range of the}

species,

we choseto concentrateon

specimens

thathad been collected in_sympatryin ordertoavoid the

confounding

effectsof

geographical

variation. Inessence, we have examined the “worst case scenario” of

sympatric populations

and_yetwestill obtained very clear results that

Palpopleura

l. lucia and P. l.

portia

are

reproductively

isolatedfrom oneanother,and should be

regarded

asseparate

species, Palpopleura

lucia

(Drury,

1773)

and P.

portia (Drury, 1773).

ACKNOWLEDGEMENTS

We thank KLAAS-DOUWE DIJKSTRA and VIOLA CLAUSN1TZER fordiscussion,and STUART TAYLOR forscanningin thefigureof thewingpatterning.Financial supportwasfrom the National ScienceFoundation,South Africa.

REFERENCES

FRAT1, E,C.SIMON,J, SULLIVAN & D.L.SWOFFORD, 1997. Evolution of the mitochondrial cytochromeoxidaseII genein Collembola. J. mol Evol. 44; 145-158.

HASEGAWA, M.,H. KISHINO & T.YANO, 1985.Datingof thehuman-ape splitting bya molecu-lar clock of mitochondrial DNA. J. mol. Evol.21: 160-174.

PINHEY, E.C.G.,1951. Thedragonfliesof Southern Africa. Transv. Mus. Mem. 5; 1-335.

P1NHEY, E.C.G.,1962. Adescriptive catalogueof the Odonataof the African continent(upto De-cember 1959). Publfoescult. Co. Diam. Angola59: 1-323.

PINHEY, E.C.G., 1985. Asurveyof thedragonflies (Odonata)of South Africa,2:Anisoptera,J.ent. Soc. sth.Afr.48: 1-48

RAMBAULT, A., 2001. Se-Al.SequenceAlignmentEditor. hnp.llevolve.zoo.ox.ac.uklSoftwarelSe-AUmain.html

STADEN, R„1996. The StadenSequence Analysis Package.Mol. Biotech. 5;233-241.

SWOFFORD, D.L., 1998. PAUP*_{Phylogeneticanalysis using parsimony *(aml}othermethods).

Sin-auerAssociates,Sunderland/MA.

THOMPSON, J.D.,T.J.GIBSON,F.PLEWNIAK,F.JEANMOUGIN &D.G.HIGGINS,1997. The ClustalX windows interface:flexiblestrategiesformultiplesequencealignmentaidedby quality analysistools. Nucl. Acids Res. 24: 4876-4882.

TableII

Summaryoftestsfor the_significanceof like-lihood differences between trees with the

topologiesshown inFigures2A and 2B

K-Htest S-H test

Combined data p=0.770 p=0.803 ITS2only p=0.624 p=0.778 COIonly p=0.849 p=0.824