four independent gene phylogenies confirm the ancient
TRANSCRIPT
Four independent gene phylogenies confirm the ancient relationshipsof Madagascar endemic species in the Papilio demoleus group(Lepidoptera, Papilionidae).Zakharov, E. V., Smith, C., Lees, D. C., Cameron, A., Vane-Wright, R. I., & Sperling, F. A. H. (2004). Fourindependent gene phylogenies confirm the ancient relationships of Madagascar endemic species in the Papiliodemoleus group (Lepidoptera, Papilionidae). Evolution, 58, 2763-2782. https://doi.org/10.1554/04-293
Published in:Evolution
Queen's University Belfast - Research Portal:Link to publication record in Queen's University Belfast Research Portal
General rightsCopyright for the publications made accessible via the Queen's University Belfast Research Portal is retained by the author(s) and / or othercopyright owners and it is a condition of accessing these publications that users recognise and abide by the legal requirements associatedwith these rights.
Take down policyThe Research Portal is Queen's institutional repository that provides access to Queen's research output. Every effort has been made toensure that content in the Research Portal does not infringe any person's rights, or applicable UK laws. If you discover content in theResearch Portal that you believe breaches copyright or violates any law, please contact [email protected].
Download date:02. Oct. 2021
2763
q 2004 The Society for the Study of Evolution. All rights reserved.
Evolution, 58(12), 2004, pp. 2763–2782
INDEPENDENT GENE PHYLOGENIES AND MORPHOLOGY DEMONSTRATE AMALAGASY ORIGIN FOR A WIDE-RANGING GROUP OF
SWALLOWTAIL BUTTERFLIES
EVGUENI V. ZAKHAROV,1,2,3 CAMPBELL R. SMITH,4,5 DAVID C. LEES,4,6 ALISON CAMERON,7,8
RICHARD I. VANE-WRIGHT,4,9 AND FELIX A. H. SPERLING1,10
1Department of Biological Sciences, University of Alberta, Edmonton, Alberta T6G 2E9, Canada2E-mail: [email protected]
4Biogeography and Conservation Laboratory, Department of Entomology, The Natural History Museum, Cromwell Road,London SW7 5BD, United Kingdom
5E-mail: [email protected]: [email protected]
7Faculty of Biological Sciences, School of Biology, University of Leeds, Leeds LS2 9JT, United Kingdom8E-mail: [email protected]
9E-mail: [email protected]: [email protected]
Abstract. Madagascar is home to numerous endemic species and lineages, but the processes that have contributedto its endangered diversity are still poorly understood. Evidence is accumulating to demonstrate the importance ofTertiary dispersal across varying distances of oceanic barriers, supplementing vicariance relationships dating back tothe Cretaceous, but these hypotheses remain tentative in the absence of well-supported phylogenies. In the Papiliodemoleus group of swallowtail butterflies, three of the five recognized species are restricted to Madagascar, whereasthe remaining two species range across the Afrotropical zone and southern Asia plus Australia. We reconstructedphylogenetic relationships for all species in the P. demoleus group, as well as 11 outgroup Papilio species, using 60morphological characters and about 4 kb of nucleotide sequences from two mitochondrial (cytochrome oxidase I andII) and two nuclear (wg and EF-1a) genes. Of the three endemic Malagasy species, the two that are formally listedas endangered or at risk represented the most basal divergences in the group, while the more common third endemicwas clearly related to African P. demodocus. The fifth species, P. demoleus, showed little differentiation across southernAsia, but showed divergence from its subspecies sthenelus in Australia. Dispersal-vicariance analysis using cladogramsderived from morphology and three independent genes indicated a Malagasy diversification of lime swallowtails inthe middle Miocene. Thus, diversification processes on the island of Madagascar may have contributed to the originof common butterflies that now occur throughout much of the Old World tropical and subtemperate regions. Analternative hypothesis, that Madagascar is a refuge for ancient lineages resulting from successive colonizations fromAfrica, is less parsimonious and does not explain the relatively low continental diversity of the group.
Key words. Biogeography, dispersal, island colonization, Madagascar, Papilionidae, speciation, vicariance.
Received May 7, 2004. Accepted September 1, 2004.
Islands can play an important role in both generating andconserving biodiversity. Species radiations on island archi-pelagos are well studied (e.g., Grant and Grant 2002; Gil-lespie and Roderick 2002; Jordan et al. 2003), although thecontribution of such island diversification to surroundingcontinents remains poorly documented. Under the principlesof island biogeography, island size and distance from themainland are the primary determinants of the accumulationof species that disperse from larger land masses (MacArthurand Wilson 1967). The likelihood that island species willcolonize continents seems relatively low, because islandsusually contain both fewer species and smaller total popu-lations than mainlands. Furthermore, island species are likelyto be at a competitive disadvantage when islands are colo-nized by continental species (Gargominy et al. 1996; Whit-taker 1998; Cowie 2001; O’Dowd et al. 2003). However, thewidespread assumption that islands contribute little to con-tinental biotas remains relatively untested, and the devel-opment of more rigorous methods of phylogenetic and bio-geographic analysis, as well as opportunities for the gener-
3 Present address: 107 Galvin Life Sciences Center, Universityof Notre Dame, Notre Dame, Indiana 46556.
ation of large molecular datasets that can contribute to morerobust reconstructions, have greatly improved the opportu-nities to assess such biogeographic hypotheses.
The island of Madagascar is widely known for its uniqueand diverse flora and fauna. It has been isolated from Africasince the late Jurassic and from other Gondwanan elementssuch as India since the late Cretaceous (Rabinowitz et al. 1983;Storey et al. 1995; Briggs 2003). These conditions have ledto the evolution of a very large number of endemic species,although the processes that have contributed to the biotic di-versity within Madagascar are still poorly understood. Manyendemics represent high-ranking, very distinct groups, such aslemurs and tenrecs. As a result, and because much of theisland’s biodiversity is now under threat of extinction, Mad-agascar is considered one of the world’s 10 most importantregions for biodiversity conservation (Myers et al. 2000).
In spite of the proximity of Madagascar to Africa since itsseparation from that continent about 165 million years ago,the Malagasy flora exhibits remarkably high affinity withIndo-Australo-Malesian floras far to the east (Schatz 1996).Such phytogeographic connections are especially prevalentamong eastern humid forest taxa, and in some cases mayrepresent relictual Cretaceous Gondwanan disjunctions, as
2764 EVGUENI V. ZAKHAROV ET AL.
well as repeated long-distance or stepping-stone dispersalevents across the Indian Ocean. Although the fauna of Mad-agascar shows more connections with Africa (Paulian 1972;Warren et al. 2003), some Oriental affinities are still present.For example, several Malagasy birds show clear affinitieswith Oriental groups that are unknown in Africa (Dorst 1972).The presence of some lineages of plants and animals in Mad-agascar may be explained by Cretaceous dispersals to theisland from India or South Africa, and floral and faunal ex-change with the remains of Gondwana via Antarctica duringthe time of the initial radiation of the angiosperms (Koechlin1972; Schatz 1996; Vences et al. 2001). This scenario mayaccount for the evolution of seven endemic plant familiesand three recently extinct ratite species (Cooper et al. 2001).
In addition, recent phylogenetic evidence and fossils sug-gest that part of the modern vertebrate fauna of Madagascarhad an independent origin from that of the Cretaceous faunaof Madagascar (Krause et al. 1999). There is growing evi-dence for nontectonic dispersal to Madagascar (and Africa)from Laurasia and western Malesia via India along Lemurianstepping-stones in the western Indian Ocean during the Eo-cene-Oligocene (McKenzie and Sclater 1973; Schatz 1996;Jansa et al. 1999). Late Tertiary introduction by rafting hasbeen suggested for an endangered endemic tortoise in Mad-agascar (Caccone et al. 1999). Neogene colonization of Mad-agascar by African ancestors has been reported in Malagasyoscine songbirds (Cibois et al. 2001) and sunbirds in thePliocene (Warren et al. 2003). Yoder et al. (1996) found thatthe Malagasy lemuriforms comprise a monophyletic groupthat probably dispersed from Africa in the early Tertiary anda single African Miocene origin is also posited for carnivores(Yoder et al. 2003). Thus, multiple overseas dispersals to andfrom Madagascar have been proposed both for plants andanimals (see also Renvoise 1979; Fisher 1996; Raxworthy etal. 2002; Treutlein and Wink 2002; Nagy et al. 2003; Venceset al. 2003, 2004). Nonetheless, such hypotheses of overseasdispersal remain tentative in the absence of more and bettersupported phylogenetic evidence for the relationships of dif-ferent elements of the Malagasy biota.
Complex patterns of vicariance and dispersal, along withlong isolation of the island, account for much of the ende-mism of the Malagasy biota. At the species level, at least80% of the Malagasy flora and 75% of its fauna is endemic(Schatz 2001), and this is also the level of species endemismfor butterflies (Lees et al. 2003). For the conservation ofinvertebrates, butterflies of the family Papilionidae have beenchampioned as a flagship group (Collins and Morris 1985).Thirteen species of the family are found on Madagascar (Pau-lian and Viette 1968), of which 10 are endemic to mainlandMadagascar and the Comoros (Ackery et al. 1995). Three ofthe endemics are currently classified by the IUCN as threat-ened (IUCN 2003), and one of these belongs to the Papiliodemoleus group, the subject of this study.
The P. demoleus group, or lime swallowtails, comprisesfive butterfly species from the Old World tropics that sharea distinctive black and yellow pattern. Three of them, P.erithonioides Grose-Smith, 1891, P. grosesmithi Rothschild,1926, and P. morondavana Grose-Smith, 1891, are found onlyon Madagascar, while the other two have wide distributions
in the Afrotropical (P. demodocus Esper, 1798) and Indo-Australian (P. demoleus Linnaeus, 1758) regions.
Papilio demodocus occurs throughout sub-Saharan Africa,extending into Saudi Arabia, Yemen, and Oman. Its existenceon Madagascar, Mauritius, Reunion, and the four ComoroIslands is likely to be the result of recent and perhaps evendeliberate introductions (Paulian 1951; Paulian and Viette1968; Turlin 1994), though it is also possible that the speciesarrived naturally in Madagascar as two of us (D. C. Lees andA. Cameron) have observed it occasionally in natural habi-tats. The species is abundant in anthropogenic habitats andis not threatened.
The larvae feed mostly on Rutaceae but with no record onnative species and the species is considered to be a pest ofCitrus crops (Paulian and Viette 1968; Collins and Morris1985; Ackery et al. 1995). Although some populations inSouth Africa have switched to larval feeding on Apiaceae(van Son 1949; Clarke et al. 1963), there is no evidence thatthey have done this in Madagascar. Occasional use of An-acardiaceae, Asteraceae, Ptaeroxylaceae, and Sapindaceae aslarval host plants has been reported (Ackery et al. 1995),though these records may represent oviposition mistakes be-cause larvae of P. demodocus are not known to develop ad-equately on host plants in these families. Two subspecies arerecognized, P. demodocus demodocus Esper, 1798, from vir-tually the entire range, and the distinctive P. demodocus ben-netti Dixey, 1898, originally described as a separate speciesfrom Socotra Island (Yemen).
The distribution of P. demoleus extends from Iran throughIndia, Sri Lanka, Malaysia, southern China, and Japan, andfrom the Lesser Sunda Islands to mainland New Guinea andAustralia. The species was absent from the Philippines, Su-matra, Java, and Borneo until its recent invasion of the islandsafter their large-scale deforestation by man (Hiura 1973; Cor-bet and Pendlebury 1978, 1992). Most recently, the speciesis now established on the northern tip of Sulawesi (Vane-Wright and de Jong 2003). Six subspecies are currently rec-ognized: demoleus Linnaeus, 1758 (China through South Asiaand Pakistan to the Arabian Peninsula); malayanus Wallace,1865 (Malay Peninsula and Sumatra); libanius Fruhstorfer,1908 (Philippines, Talaud, Sula, Taiwan); sthenelus Macleay,1826 (Sumba and Australia); novoguineensis Rothschild,1908 (Papua New Guinea); and sthenelinus Rothschild, 1895(Flores and Alor). In Papua New Guinea and in Australia,the larvae develop on wild leguminous plants of the genusCullen (Fabaceae; Braby 2000), but no host record is avail-able for the Lesser Sunda island populations (Parsons 1999).All Asian subspecies feed on Citrus, which is commonlyplanted as a crop or ornament in towns and smaller settle-ments and probably facilitates range expansion of P. demo-leus in Asia (Matsumoto 2002).
The rarest of the three Malagasy endemics, P. moronda-vana, is currently classified as potentially ‘‘endangered’’while ‘‘data deficient’’ (IUCN 2003). It is confined to drydeciduous forests of west-southwestern and northern Mad-agascar and is threatened by loss of habitat (Collins and Mor-ris 1985; Conservation International 2003). Our field obser-vations and museum collections suggest that P. grosesmithi(lower risk: near threatened) is more commonly encounteredthan P. morondavana. It is found over a very similar geo-
2765ORIGIN OF ISLAND ENDEMICS
graphic range below about 800 m in dry deciduous and galleryforests of western and northern Madagascar, ranging east asfar as Fianarantsoa, and slightly further southwest than P.morondavana, while P. morondavana is known from the farnorth. According to field observations in 2001–2002 (A.Cameron), P. morondavana can be abundant at the beginningof the wet season at some sites, but appears to be more lo-calized than P. grosesmithi.
Papilio erithonioides (not classified as threatened) occursin spiny forests of the far southwest as well as deciduousand gallery forests in the southwestern, western, and northernparts of the island, also penetrating marginally into north-western and southwestern rainforests. Our recent observa-tions show it to be by far the most common endemic. Sig-nificantly, all three species can be found sympatrically atsome sites, notably the unprotected Kirindy Forest near Mo-rondavana. Papilio demodocus is found in anthropic habitatsthroughout the island including the plateau. Further detailsof the nomenclature, distribution, and bionomics will be pro-vided in a companion publication (C. R. Smith and R. I. Vane-Wright, unpubl. ms.).
Although the life history and geographic variability of theP. demoleus species group is relatively well known (at leastoutside Madagascar), their phylogenetic relationships remainunknown and pose questions about their origin and evolution.Carcasson (1964) considered P. demodocus to be the ancestralspecies, from which the three species endemic in Madagascarwere isolated, one after another, as the result of multipleinvasions of a mainland form. This model corresponds to theconcept of duplex species, in which the rare species, with amore restricted range, owes its origin to isolations and re-unions with the broader range of a sister species (Zeuner1943; Corbet 1944). However, in this case the vicariance thatis normally caused by sea-level rise is replaced by dispersalof a relatively mobile species across a fixed sea level gap.In contrast to a sequential colonization scenario, speciationof Malagasy endemics solely on Madagascar, whether in dif-ferent parts of the island or in contact with each other, remainsa possibility for the P. demoleus group. Endemic Malagasyradiations have been reported for a variety of taxa such astortoises, songbirds, butterflies, and colubrid snakes (Cac-cone et al. 1999; Cibois et al. 2001; Torres et al. 2001; Nagyet al. 2003). However, only a handful of recent studies havesuggested that the Malagasy biota has contributed to the sur-rounding continents.
Here, we report results of phylogenetic analyses based onmorphology and nucleotide sequences of two mitochondrialgenes, cytochrome oxidase subunit I (COI), cytochrome oxi-dase subunit II (COII), and two nuclear genes, wingless (wg),and elongation factor 1a (EF-1a). We use inferred phylog-enies combined with estimated divergence times to recon-struct ancestral areas for the P. demoleus species group andto identify patterns of species dispersal that have contributedto the enigmatic biodiversity of Madagascar. Moreover, anyimprovement in our understanding of the systematics of thesebutterflies provides a better foundation for appropriate con-servation action (Vane-Wright 2003).
MATERIALS AND METHODS
Sampling Strategy
Sampled species and GenBank accession numbers are giv-en in Table 1. Ingroup taxa included all five recognized spe-cies of the P. demoleus group, with sampling across speciesranges where possible. Four out of six described subspeciesof P. demoleus were sampled; P. d. novoguinensis and P. d.sthenelinus were not available for this study. Both subspeciesof P. demodocus were sampled, including P. d. demodocusand P. d. bennetti (a taxon restricted to Socotra Island, Yem-en). However, only two legs from two old pinned specimenscollected in 1967 (British Museum of Natural History En-tomology Department, specimen register nos. 220150 and220151) were available for DNA extraction for P. d. bennetti.
Eleven outgroup taxa were chosen, representing majormonophyletic lineages within the genus based on a recentreconstruction of Papilio phylogeny (Zakharov et al. 2004)to test the monophyly of the P. demoleus group and to relateits phylogenetic position within Papilio. Trees were rootedwith a single neotropical species, Papilio thoas. The originaldata and tree files are available from www.treebase.org (ac-cession number SN1870).
Morphology
A total of 60 characters (Fig. 1, Appendix 1) were scoredby C. R. Smith from wing patterns and male and femalegenitalia for the five species of lime swallowtails (includingboth subspecies of P. demodocus) and all outgroup taxa.Where material was available, male and female genitalia wereexamined in several specimens of each species. Dissectionswere made using standard techniques, after abdomens weresoaked in cold 10% KOH solution overnight and subsequent-ly stored in glycerol. Voucher information for specimensscored for morphological characters is available as supple-mentary online materials at http://dx.doi.org/10.1554/04-293.1.s1.
Although all species of the P. demoleus group share thedistinctive wing pattern of the group, some problems in ho-mology assessment of wing pattern elements were encoun-tered in scoring outgroup species. Certain elements of wingpattern in some species of Papilio are fused or lost com-pletely, making the wing pattern very different from a gen-eralized swallowtail groundplan of black and yellow stripes,spots, and patches. To help hypothesize homology, we eval-uated the position of pattern elements on the wing with re-spect to wing venation and examined series of related species.Some genitalic homologies were likewise problematic. Fur-ther details of the wing patterns and male and female genitalanatomy, and the character hypotheses derived from them,will be provided in a companion publication (C. R. Smithand R. I. Vane-Wright, unpubl. ms.).
Molecular Techniques
Selection of genes for molecular phylogenetic reconstruc-tion was based on previous successful studies in Lepidoptera(Sperling 2003) and other insects (Caterino et al. 2000). Thetwo most commonly used mitochondrial DNA protein-codinggenes, COI and COII, which are generally most useful in
2766 EVGUENI V. ZAKHAROV ET AL.
TA
BL
E1.
Lis
tof
spec
ies
and
spec
imen
sex
amin
ed.
Spe
cies
Loc
alit
yD
ate
Vou
cher
ID1
Hap
loty
peC
OI,
CO
IIE
F-1
aW
ingl
ess
1.P
apil
iode
mol
eus
mal
ayan
usM
alay
sia:
Pen
ang
Isla
ndM
ay16
,19
89C
U:
Lot
No.
1204
DL
1A
F04
4000
3A
F04
4825
2,4
AY
5691
142.
Pap
ilio
dem
oleu
sm
alay
anus
Mal
aysi
a:P
enan
gIs
land
May
16,
1989
UA
FS
:a48
DL
1A
Y56
9049
3.P
apil
iode
mol
eus
mal
ayan
usM
alay
sia:
Per
akA
pril
2002
UA
FS
:165
0D
L1
AY
5690
504.
Pap
ilio
dem
oleu
sm
alay
anus
Mal
aysi
a:P
erak
Apr
il20
02U
AF
S:1
651
DL
1A
Y56
9051
5.P
apil
iode
mol
eus
mal
ayan
usM
alay
sia
Sep
tem
ber
30,
2002
SH
Y:1
855
DL
2A
Y56
9058
6.P
apil
iode
mol
eus
mal
ayan
usIn
done
sia:
Bal
iIs
land
June
1998
UA
FS
:a39
8D
L3
AY
5690
487.
Pap
ilio
dem
oleu
sm
alay
anus
Tha
ilan
d:C
hian
gM
aiM
ay10
,20
01N
P:D
L01
-Q67
7D
L4
AY
5690
558.
Pap
ilio
dem
oleu
sm
alay
anus
Tha
ilan
d:C
hian
gM
aiM
ay10
,20
01N
P:D
L01
-Q67
6D
L1
AY
5690
569.
Pap
ilio
dem
oleu
sm
alay
anus
Vie
tnam
:B
ach
Ma
Nat
.P
ark
June
10,
2001
UA
FS
:165
6D
L1
AY
5690
52
10.
Pap
ilio
dem
oleu
sde
mol
eus
Iran
:H
orm
ozga
nP
rov-
ince
Mar
ch11
,20
02U
AF
S:1
871
DL
5A
Y56
9053
11.
Pap
ilio
dem
oleu
sli
bani
usT
aiw
anS
epte
mbe
r20
02S
HY
:185
6D
L1
AY
5690
5912
.P
apil
iode
mol
eus
liba
nius
Tai
wan
Sep
tem
ber
2002
SH
Y:1
857
DL
1A
Y56
9057
13.
Pap
ilio
dem
oleu
sli
bani
usT
aiw
anS
epte
mbe
r20
02S
HY
:185
8D
L1
AY
5690
6014
.P
apil
iode
mol
eus
sthe
nelu
sA
ustr
alia
:N
ewS
outh
Wal
esN
ovem
ber
12,
2000
NP
:RE
02-A
018
DL
6A
Y56
9092
6A
Y56
9104
AY
5691
15
15.
Pap
ilio
dem
oleu
sst
hene
lus
Aus
tral
ia:
New
Sou
thW
ales
Nov
embe
r12
,20
00N
P:R
E01
-H07
0D
L7
AY
5690
67
16.
Pap
ilio
dem
odoc
usde
mod
ocus
Ken
ya(e
xpu
pa)
Mar
ch15
,19
89U
AF
S:a
146
DD
1A
Y45
7588
5A
Y45
7614
5A
Y56
9112
17.
Pap
ilio
dem
odoc
usde
mod
ocus
Ken
ya(e
xpu
pa)
Mar
ch15
,19
89U
AF
S:1
872
DD
2A
Y56
9070
18.
Pap
ilio
dem
odoc
usde
mod
ocus
Cam
eroo
n:B
uea
Mar
ch11
,20
02U
AF
S:1
653
DD
3A
Y56
9071
19.
Pap
ilio
dem
odoc
usde
mod
ocus
Cam
eroo
n:B
uea
Mar
ch11
,20
02U
AF
S:1
654
DD
4A
Y56
9068
20.
Pap
ilio
dem
odoc
usde
mod
ocus
Cam
eroo
n:B
uea
Mar
ch11
,20
02U
AF
S:1
655
DD
5A
Y56
9069
21.
Pap
ilio
dem
odoc
usde
mod
ocus
Cen
tral
Afr
ican
Rep
ubli
c20
01U
AF
S:1
652
DD
6A
Y56
9072
22.
Pap
ilio
dem
odoc
usde
mod
ocus
Sou
thA
fric
a:N
elsp
ruit
Janu
ary
16,
2003
UA
FS
:192
7D
D7
AY
5690
7323
.P
apil
iode
mod
ocus
dem
odoc
usS
outh
Afr
ica:
Nel
spru
itJa
nuar
y16
,20
03U
AF
S:1
928
DD
8A
Y56
9074
24.
Pap
ilio
dem
odoc
usde
mod
ocus
Sou
thA
fric
a:N
elsp
ruit
Janu
ary
16,
2003
UA
FS
:193
0D
D9
AY
5690
7525
.P
apil
iode
mod
ocus
dem
odoc
usS
outh
Afr
ica:
Nel
spru
itJa
nuar
y16
,20
03U
AF
S:1
929
DD
10A
Y56
9091
6A
Y56
9103
AY
5691
1326
.P
apil
iode
mod
ocus
dem
odoc
usM
adag
asca
r:A
ntsa
lova
Feb
ruar
y12
,20
02B
MN
H(E
):69
7423
DD
11A
Y56
9064
27.
Pap
ilio
dem
odoc
usde
mod
ocus
Mad
agas
car
May
16,
1989
UA
FS
:a58
DD
11A
Y56
9063
28.
Pap
ilio
dem
odoc
usde
mod
ocus
Mad
agas
car:
Tsi
mel
ahy
Janu
ary
19,
2002
CA
SE
NT
:800
1338
DD
12A
Y56
9065
29.
Pap
ilio
dem
odoc
usde
mod
ocus
Mad
agas
car:
Mah
avel
oF
ebru
ary
22,
2002
CA
S:8
0014
18D
D13
AY
5690
6630
.P
apil
iode
mod
ocus
dem
odoc
usM
adag
asca
r:M
ahaf
alay
Feb
ruar
y22
,20
02C
AS
:800
1794
DD
11A
Y56
9067
31.
Pap
ilio
dem
odoc
usde
mod
ocus
Mad
agas
car
2001
UA
FS
:180
1D
D14
AY
5690
6132
.P
apil
iode
mod
ocus
dem
odoc
usM
adag
asca
r20
01U
AF
S:1
802
DD
14A
Y56
9062
33.
Pap
ilio
dem
odoc
usbe
nnet
tiY
emen
:S
ocot
raIs
land
1967
BM
NH
(E):
2201
5034
.P
apil
iode
mod
ocus
benn
etti
Yem
en:
Soc
otra
Isla
nd19
67B
MN
H(E
):22
0151
35.
Pap
ilio
erit
honi
oide
sM
adag
asca
r:A
nkit
sang
aF
ebru
ary
6,20
02B
MN
H(E
):69
7422
E1
AY
5690
956
AY
5690
99A
Y56
9108
36.
Pap
ilio
erit
honi
oide
sM
adag
asca
r:M
ahav
elo
Janu
ary
31,
2001
CA
S:8
0014
77E
1A
Y56
9096
6A
Y56
9100
AY
5691
0937
.P
apil
ioer
itho
nioi
des
Mad
agas
car:
Tsi
mel
ahy
Janu
ary
17,
2002
CA
S:8
0012
25E
1A
Y56
9076
38.
Pap
ilio
erit
honi
oide
sM
adag
asca
r:B
ealo
kaF
ebru
ary
4,20
02C
AS
:800
1520
E1
AY
5690
7739
.P
apil
ioer
itho
nioi
des
Mad
agas
car:
Bem
anat
eza
Mar
ch24
,20
02C
AS
:800
2601
E1
AY
5690
7840
.P
apil
ioer
itho
nioi
des
Mad
agas
car:
Tsi
fota
Mar
ch9,
2002
CA
S:8
0022
21E
1A
Y56
9079
41.
Pap
ilio
erit
honi
oide
sM
adag
asca
r:M
itoh
oM
arch
19,
2002
CA
S:8
0024
17E
1A
Y56
9080
42.
Pap
ilio
erit
honi
oide
sM
adag
asca
r:B
erob
oka
Mar
13,
2002
CA
S:8
0022
74E
1A
Y56
9081
43.
Pap
ilio
gros
esm
ithi
Mad
agas
car:
Kir
indy
Nor
dJa
nuar
y20
,20
02B
MN
H(E
):69
7420
G1
AY
5690
896
AY
5691
01A
Y56
9110
44.
Pap
ilio
gros
esm
ithi
Mad
agas
car:
Am
bahi
beJa
nuar
y31
,20
02B
MN
H(E
):69
7421
G2
AY
5690
906
AY
5691
02A
Y56
9111
45.
Pap
ilio
gros
esm
ithi
Mad
agas
car:
Sep
tL
acM
arch
1,20
02C
AS
:800
1969
G3
AY
5690
8246
.P
apil
iogr
oses
mit
hiM
adag
asca
r:T
sim
elah
yJa
nuar
y18
,20
02C
AS
:800
1315
G2
AY
5690
8347
.P
apil
iogr
oses
mit
hiM
adag
asca
r:Z
ombi
tse
Feb
ruar
y6,
2003
CA
S:8
0343
00G
3A
Y56
9084
2767ORIGIN OF ISLAND ENDEMICS
TA
BL
E1.
Con
tinu
ed.
Spe
cies
Loc
alit
yD
ate
Vou
cher
ID1
Hap
loty
peC
OI,
CO
IIE
F-1
aW
ingl
ess
48.
Pap
ilio
gros
esm
ithi
Mad
agas
car:
Zom
bits
eF
ebru
ary
6,20
03C
AS
:803
4301
G4
AY
5690
8549
.P
apil
iom
oron
dava
naM
adag
asca
r:K
irin
dyN
ovem
ber
30,
2001
CA
S:8
0008
01M
1A
Y56
9093
6A
Y56
9097
AY
5691
06
50.
Pap
ilio
mor
onda
vana
Mad
agas
car:
Am
bovo
n-ao
mby
Nov
embe
r6,
2002
CA
S:8
0029
96M
1A
Y56
9094
6A
Y56
9098
AY
5691
07
51.
Pap
ilio
mor
onda
vana
Mad
agas
car:
And
rong
oni-
beN
ovem
ber
20,
2002
CA
S:8
0330
22M
1A
Y56
9086
52.
Pap
ilio
mor
onda
vana
Mad
agas
car:
And
rong
oni-
beN
ovem
ber
20,
2002
CA
S:8
0330
48M
1A
Y56
9087
Out
grou
pta
xa53
.P
apil
iohe
lenu
sJa
pan
(ex
pupa
)M
arch
29,
1990
UA
FS
:a90
AY
4575
755
AY
4576
195
AY
5691
2154
.P
apil
iopa
ris
Chi
na:
Gua
ndon
gP
rov-
ince
June
20,
2001
UA
FS
:157
2A
Y45
7574
5A
Y45
7605
5A
Y56
9116
55.
Pap
ilio
cons
tant
inus
Ken
ya(e
xpu
pa)
1989
CU
:L
otN
o.12
04A
F04
4002
3A
F04
4836
2,
4A
Y56
9118
56.
Pap
ilio
dela
land
eiM
adag
asca
r:T
alat
akel
yA
pril
–M
ay19
98U
AF
S:9
39A
Y45
7585
5A
Y45
7613
5A
Y56
9117
57.
Pap
ilio
lorm
ieri
Cen
tral
Afr
ican
Rep
ubli
cJu
ly22
,20
01U
AF
S:1
875
AY
5690
886
AY
5691
05A
Y56
9122
58.
Pap
ilio
orib
azus
Mad
agas
car:
Tal
atak
ely
Apr
il–
May
,19
98U
AF
S:9
41A
Y45
7591
5A
Y45
7626
5A
Y56
9119
59.
Pap
ilio
xuth
usJa
pan:
Tok
yoA
ugus
t30
,19
90C
U:
Lot
No.
1204
AF
0439
993
AF
0448
382
,4
AY
5691
2360
.P
apil
iom
acha
onF
ranc
e:C
oudo
ux(e
xpu
pa)
Feb
ruar
y18
,19
88C
U:
Lot
No.
1204
AF
0440
063
AF
0448
282
,4
AY
5691
24
61.
Pap
ilio
anac
tus
Aus
tral
ia:
Que
ensl
and
1997
UA
FS
:942
AY
4575
925
AY
4576
085
AY
5691
2062
.P
apil
ioca
nade
nsis
US
A:
New
Yor
k:R
ich-
ford
June
5,19
97C
U:
Lot
No.
1204
AF
0440
143
AF
0448
162
,4
AY
5691
25
63.
Pap
ilio
thoa
sF
renc
hG
uian
aM
ay30
,19
90U
AF
S:3
02A
Y45
7601
5A
Y45
7632
5A
Y56
9126
1V
ouch
erlo
cati
ons
are
indi
cate
das
:C
U,
inse
ctco
llec
tion
(see
Cat
erin
oan
dS
perl
ing
1999
)of
the
Dep
artm
ent
ofE
ntom
olog
yat
Cor
nell
Uni
vers
ity,
Itha
ca,
NY
;U
AF
S,
DN
Aan
dti
ssue
coll
ecti
onin
F.A
.H.S
.L
abor
ator
yat
the
Uni
vers
ity
ofA
lber
ta,
Edm
onto
n,A
B;
BM
NH
(E),
Dep
artm
ent
ofE
ntom
olog
yat
Nat
ural
His
tory
Mus
eum
,L
ondo
n,U
.K.;
CA
SE
NT,
Dep
artm
ent
ofE
ntom
olog
yat
Cal
ifor
nia
Aca
dem
yof
Sci
ence
s,G
olde
nG
ate
Par
k,S
anF
ranc
isco
,C
A;
SH
Y,
spec
imen
sfr
ompr
ivat
eco
llec
tion
ofS
hen-
Hor
nY
en;
NP,
DN
Aan
dti
ssue
coll
ecti
onin
Nao
mi
Pie
rce
lab
atH
arva
rdU
nive
rsit
y,C
ambr
idge
,M
A.
2A
245-
bpfr
agm
ent
was
sequ
ence
dto
exte
nd39
end
ofpr
evio
usly
publ
ishe
dse
quen
ceof
EF
-1a
(Ree
dan
dS
perl
ing
1999
).3
(Cat
erin
oan
dS
perl
ing
1999
).4
(Ree
dan
dS
perl
ing
1999
).5
(Zak
haro
vet
al.
2004
).6
Spe
cim
ens
wer
ese
quen
ced
for
full
2.3
kbof
CO
I1
CO
IIin
addi
tion
to82
5-bp
frag
men
tat
39en
dof
CO
I.
2768 EVGUENI V. ZAKHAROV ET AL.
FIG. 1. Edited digital images showing morphological characters, with circled numbers referring to numbered character states (Appendix1). (A) Upper and undersides of Papilio demoleus demoleus. Ellipses on wings indicate the position of character traits. (B) Male genitaliaof P. demodocus bennetti. (C) Female genitalia of P. erithonioides. All character states shown are coded as (1) unless indicated otherwise.
delineating taxa at the species level (Sperling 2003; but seeWahlberg et al. 2003), also represent the maternal history ofspeciation in the P. demoleus species-group. Two other loci,wg and EF-1a, are the most commonly used protein-codingnuclear genes in phylogenetic studies in Lepidoptera and areusually considered useful at higher taxonomic levels due totheir lower substitution rates (Sperling 2003). Both nucleargenes appear to have only a single copy in lepidopteran ge-nome and have either no introns (EF-1a) or a 450-bp exonbetween introns (wg).
Mitochondrial genes COI, tRNA-leu, COII and nucleargenes wg and EF-1a were sequenced new by E. V. Zakharov
or retrieved from GenBank for all 11 outgroups and twospecimens per species (10 total) in the ingroup. For EF-1a,an additional fragment of about 245 bp was sequenced fromthe 39-end to extend previously published sequences that wereincomplete for the gene (Reed and Sperling 1999). An 825-bp fragment comprising the 39-end of the COI gene was se-quenced for an additional 40 ingroup individuals, except thetwo specimens of P. demodocus bennetti, which producedonly highly degraded DNA.
Material of varying quality was used for DNA extraction,ranging from alcohol-preserved freshly caught samples todried nonrelaxed specimens and legs of museum pinned but-
2769ORIGIN OF ISLAND ENDEMICS
terflies. Total genomic DNA was extracted either by standardphenol-chloroform procedure as in Sperling and Harrison(1994) or using a Qiagen (Valencia, CA) DNeasy tissue kit.Polymerase chain reactions (PCR) were performed in Biom-etra (Goettingen, Germany) TGradient or TPersonal thermalcyclers using reaction and cycling conditions described pre-viously (Brower and DeSalle 1998; Caterino et al. 2001).PCR products were cleaned using a Qiagen QIAquick PCRpurification kit when only a single DNA band was visible ina gel or, when more than one band was observed, a combi-nation of gel-separation and subsequent purification with aQiagen QIAEX II gel extraction kit. Purified PCR productswere directly sequenced using DYEnamicTM ET terminatorcycle sequencing (Amersham Pharmacia Biotech, Cleveland,OH) or Applied Biosystems (ABI, Foster City, CA) Big Dyeterminator cycle sequencing, under manufacturer’s recom-mendations. Fluorescently labeled sequencing products werefiltered through Sephadex-packed columns, dried, resus-pended, and fractionated on an ABI 377 automated sequencer.All fragments were sequenced in both directions using thesame primers that were used for PCR (Appendix 2, availableonline at http://dx.doi.org/10.1554/04-293.1.s2). Sequenceswere assembled into contiguous arrays using Sequencher,version 4.1 (GeneCode Corp., Ann Arbor, MI).
Phylogenetic Analysis
For morphological data, most parsimonious cladogramswere inferred from the equally weighted and unordered datamatrix using a branch-and-bound search in PAUP* version4.0b10 (Swofford 1998). Bootstrap analysis was done inPAUP using 1000 replicates and used heuristic searches withtree bisection-reconnection (TBR) branch swapping with noadditional random replicates. The results of maximum par-simony analysis were checked using a heuristic search al-gorithm (with 10 random additional replicates and TBRbranch swapping) of NONA 2.0 (Goloboff 1999) spawnedwith the aid of WinClada (Nixon 2002). Bayesian analysisof morphological dataset was performed in MrBayes version3.04b (Huelsenbeck and Ronquist 2001) with the matrix treat-ed as standard binary characters (DATATYPE 5 STAN-DARD), we ran four chains with 5.0 3 105 generations sam-pling every 100th tree. The first 2500 sampled trees werediscarded before computing a consensus tree.
Alignment of nucleotide sequences for all genes was trivialdue to lack of indels and was done by eye with the aid ofBioEdit version 5.0.9 (Hall 1999). MEGA version 2.1 (Kumaret al. 2001) was used to calculate gene statistics such as tran-sition/transversion ratios. PAUP* was used for all parsimony,bootstrap, and decay analyses of molecular and combined data.All parsimony analyses of nucleotide data used heuristicsearches (exhaustive search was applied to morphologicaldata) with 100 random addition sequence replicates and TBRbranch swapping. All base positions were treated as unorderedequally weighted characters. Gaps were treated as missingdata, and multistate characters as polymorphisms. Nonpara-metric bootstrap probabilities (BP) were based on 500 repe-titions of heuristic searches with only 10 random additionreplicates and TBR branch swapping. To determine the numberof additional steps needed to accommodate alternative phy-
logenetic hypotheses, constraint searches were carried out foreach dataset (Bremer 1988). Decay indices were extracted us-ing the program TreeRot (Sorenson 1999).
Bayesian phylogenetic analyses were conducted for com-bined and partitioned datasets in MrBayes version 3.04b(Huelsenbeck and Ronquist 2001) with the general-time re-versible model with gamma distributed rates and proportionof invariable sites applied to the molecular portion of data.Values of specific parameters for the nucleotide substitutionmodel were estimated as part of the analysis and were allowedto vary for individual genes in the combined analyses. Weran four chains simultaneously, three heated and one cold.Each Markov chain was started from a random tree and runfor 1.0 3 106 generations, sampling the chains every 100thcycle. The log-likelihood scores of sample points were plot-ted against generation time to determine when the chain be-came stationary. The first half of the sampled trees was dis-carded as burn-in samples. We ran each analysis three times,each beginning with different random starting trees, and com-pared their apparent stationarity levels for convergence(Huelsenbeck and Bollback 2001). Data remaining after dis-carding burn-in samples were used to generate a majority-rule consensus tree, where percentage of samples recoveringany particular clade represented the clade’s posterior prob-ability (Huelsenbeck and Ronquist 2001). Probabilities of95% or higher were considered as significant support. Themean, variance, and 95% credibility interval were calculatedfrom the set of substitution parameters.
Dispersal-Vicariance Analysis
To reconstruct the distribution history of the P. demoleusgroup we used the dispersal-vicariance approach implementedin program DIVA (Ronquist 1996). In DIVA a fully bifurcatedphylogeny is used to optimize the distribution of ancestralspecies with parsimony. The method is based on optimizationof a three-dimensional cost matrix derived from a simple bio-geographic model. Distributions are described in terms of aset of unit areas, and speciation is assumed to divide ancestraldistributions allopatrically into mutually exclusive sets of unitareas. DIVA finds the optimal distributions of ancestral speciesby minimizing the number of dispersal and extinction events.Unlike other methods in historical biogeography, areas are notrequired to be hierarchically related, and DIVA does not makeany assumptions about the shape or existence of general pat-terns of area relationships.
We used the trees obtained from bootstrap analysis of ourcombined and partitioned datasets. The distribution of eachspecies was classified as present/absent in four different areas.In the optimization we allowed a maximum of two geograph-ical areas for ancestral species, under the assumption that an-cestral populations had limited geographical distributions.
RESULTS
Data Description
Among 60 morphological characters scored for the P. de-moleus group, 16 characters in one or more outgroup specieswere scored missing or inapplicable being dependent on acharacter itself scored as absent (0; Table 2). Two characters
2770 EVGUENI V. ZAKHAROV ET AL.
TABLE 2. Morphological data matrix, for Papilio species. Characters that could not be scored with confidence were coded as doubtful(?), missing or inappropriate (2), or polymorphic (*).
Species
Characters
123456789111111111122222222223333333333444444444455555555556012345678901234567890123456789012345678901234567890
P. thoas 0--010110010100001111000310000100001101101100001110110001110P. canadensis 0--0?011?0101011111111002000001000000000011000011111101010--P. machaon 10-01111?0101111111111002001000---0100110110010111010----100P. anactus 0--01111?010101110-111000-000011010100100110000111010----100P. xuthus 1100111100101011111111003001000---01000101100001110110101100P. paris 10-00000000*101100-10-003100000---01001001100001110010011100P. helenus 110000000000101000-10-003110000---000110011000010-1111101100P. constantinus 10-0111100100-0000-010103111000---01000001100101101110000100P. lormieri 110001010010100001011111311000101010000100100001110110101100P. delalandei 10-000010010101110-011113110000---11000101100001100110101100P. d. demodocus 1111111110111000010110000-0110100000001101110001110010001100P. d. bennetti 111111111011100001011000100110100001001101110001110010001100P. oribazus 0--0110100000-1110-011002100000---010000011000010-0010111100P. erithonioides 111111111111110001111100100010100000001101110001110010001100P. morondavana 10-111111111101011111100200010100001001111110001111111100110P. grosesmithi 111111111011110001111100200111110111001101110001111011100100P. demoleus 111111*011*11000*11111000-01101110001000011011110-1010110101
were scored as doubtful, and four characters were polymor-phic. Across all species, two characters were nonvariable andeight variable characters were parsimony uninformative,leaving 50 parsimony informative characters for the wholedataset. For the ingroup alone, the number of nonvariablecharacters increased to 11 and number of informative char-acters decreased to 31.
The final alignment of DNA sequences numbered 3929nucleotides in the combined dataset, including 1532 bp ofCOI, 68 bp of tRNA-leu, 685 bp of COII, 404 bp of wg, and1240 bp of EF-1a. GenBank accession numbers for sequencesare given in Table 1. Statistics for the ingroup and for alltaxa are given in Table 3. Within the ingroup, COI and COIIhad the highest proportion of informative characters (10.9%and 10.3%, respectively), while wg had 5.9% and EF-1a only3.9% parsimony informative characters. The tRNA-leu se-quences had only one variable nucleotide in the ingroup, andfor that reason this fragment was excluded from all furtheranalyses except the combined dataset with all outgroup taxa.The number of informative characters varied over genes andnucleotide positions and, as is common in such data (Reedand Sperling 1999), most substitutions were observed in thirdcodon positions and second codon positions were the mostconservative in all genes. Very few or no substitutions wereregistered in first and second codon positions in nuclear genesin the ingroup. Relatively low levels of homoplasy were reg-istered in all genes for the ingroup, while the inclusion ofoutgroups increased homoplasy in nucleotide data as seenfrom CI and RI values (Table 3). The same was true forsaturation levels, with transition/transversion ratios decliningin the total dataset compared to the ingroup.
Plotting p-distances versus maximum-likelihood correcteddistances under GTR 1 G 1 I revealed higher levels ofsaturation in COII and COI and lower amounts in nucleardatasets (data not shown). The average percent sequence di-vergence (p-distance) for the ingroup ranged from 1.7% inEF-1a to 2.4% in wg and 5.1% both in COI and COII, while
adding outgroups increased the amount of divergence to 4.4%in EF-1a, 7.7% in wg, 8.0% in COII, and 8.4% in COI.
Phylogenetic Analyses
A branch-and-bound search in PAUP retained six mostparsimonious trees for morphological data (140 steps, CI 50.429, RI 5 0.529). Alternative topologies for both ingroupand outgroup relationships were recovered and no resolutionwas obtained in the strict consensus tree for any outgroupsexcept paris 1 helenus and machaon 1 xuthus (Fig. 2). Thestrict consensus also had an unresolved trichotomy betweenP. erithonioides, P. demodocus and ((P. demoleus, P. groses-mithi), P. morondavana). Levels of support on the morpho-logical tree were very low, and did not allow confidence aboutphylogenetic relationships between species in the P. demoleusspecies-group. However, the bootstrap consensus tree gavea basal position for P. morondavana. Also, when the mostdistant outgroups (based on highest pairwise sequence di-vergences from the ingroup), P. thoas and P. canadensis, wereremoved from the maximum parsimony analysis and P. an-actus was used as an outgroup as the next most distant relativeof lime swallowtails (Zakharov et al. 2004), the strict con-sensus of four most parsimonious trees also showed a basalposition for P. morondavana. NONA heuristic searches ofthe morphological data recovered the same six trees that werefound by branch-and-bound algorithm in PAUP.
Heuristic searches of molecular data with the full set ofoutgroup species resulted in a single tree (1157 steps) forCOI data, five trees for COII (469 steps), one tree for COI1 COII (1641 steps), one tree for wg (267 steps), one treefor EF-1a (476), four trees for wg 1 EF-1a (757 steps), andone tree (2415 steps) for all genes combined (Fig. 3, outgrouptaxa are not shown). Two most parsimonious trees (2571steps) were found for all data including molecular and mor-phological characters. When more than one most parsimo-nious tree was found for a particular partition, the ingroup
2771ORIGIN OF ISLAND ENDEMICS
topologies were identical among trees found for all data ex-cept for the wg 1 EF-1a dataset. Each gene partition pro-duced alternative groupings for the P. demoleus group. How-ever, only COI, COI 1 COII, and all combined data producedrelatively robust trees.
Mitochondrial DNA data suggest that P. morondavana isthe most basal species in the group, while nuclear genes in-dicate P. grosesmithi. All partitions confirm the monophyly ofthe P. demoleus species-group with 91–100% bootstrap pro-portions and decay indices of 6–12, except for morphologicaldata that had a 76% bootstrap proportion and decay index of4. Combining data increased support both for monophyly ofthe P. demoleus species-group and intragroup relationships.
Bayesian analyses of mitochondrial DNA genes and par-titioned wg and EF-1a revealed the same patterns of rela-tionship among ingroup taxa that were obtained in corre-sponding maximum parsimony heuristic searches. Bayesiananalysis of combined nuclear genes converged on the treetopology recovered for the wg dataset. As in maximum par-simony searches of molecular data, the same tree topologywas found for COI, COI 1 COII and all four genes combined.Combining all four genes and morphological data in Markovchain Monte Carlo analyses with all substitution model pa-rameters being estimated individually for gene partitions re-sulted in the highest levels of ingroup node support.
The trees shown in Figure 4 illustrate the cladograms ob-tained for all data combined, for both the P. demoleus groupand representative Papilio outgroups. Maximum parsimonyand Bayesian analyses recovered trees with an almost iden-tical pattern of relationships, with the only difference in theposition of P. anactus. Both maximum parsimony and Bayes-ian analyses showed P. helenus as the sister taxon to the P.demoleus group; however, support for this relationship wasrather low. Partitioned Bremer support (Baker and DeSalle1997) was calculated for each node of the combined maxi-mum parsimony tree to provide a measure of how individualgene data contribute to the total decay indices. Most rela-tionships among outgroups had low bootstrap proportions andconflicting phylogenetic signals from independent gene par-titions. However, our main concern is the relationship withinthe ingroup, where all genes gave high support to monophylyof the P. demoleus group. The gene wg demonstrates thelowest support for a sister relationship between P. demodocusand P. erithonioides and is in greatest conflict with groupingP. demoleus with these two species. This partition also hasnegative support for allying P. grosesmithi with ((P. demo-docus, P. erithonioides), P. demoleus). Data from COII alsodemonstrate visible conflict with the grouping of P. demoleuswith P. demodocus 1 P. erithonioides.
The COI gene appears to have the most stable and reliablephylogenetic signal among closely related species and so the39 half of COI was selected for investigation of intraspecificvariability within the P. demoleus species-group. A total of27 unique haplotypes were identified from 50 ingroup indi-viduals (Fig. 5). We found seven haplotypes for P. demoleus,with the most common haplotype (DL1) shared between pop-ulations of P. d. malayanus and P. d. libanius. None of thesepopulations contained haplotypes of P. d. demoleus (DL5) orP. d. sthenelus (DL6 and DL7). Papilio demodocus revealed14 haplotypes with no haplotypes shared between different
localities in Africa. Also, none of the haplotypes of P. de-modocus from Africa were found in Madagascar. Papilio gro-sesmithi revealed only four haplotypes, two of those (G2 andG3) were found in more than one collecting site. All fourspecimens of P. morondavana, collected in different yearsfrom three distant localities, had the same COI haplotype.No intraspecific variation was revealed within P. erithonioi-des, with all eight analyzed specimens from a wide range ofsampled sites in southern and western Madagascar producingidentical sequence for the 39 half of COI.
Bayesian estimation of phylogenetic relationships of the27 haplotypes identified for the five species of the P. demoleusgroup and 11 outgroup species of Papilio is shown in Figure6. Very little nucleotide variability was found in P. groses-mithi with zero to two substitutions (p 5 0–0.242%) in the825-bp fragment. The degree of sequence divergence betweentwo subspecies of P. demoleus (ssp. malayanus and ssp. li-banius) was also low, with p ranging from 0% to 0.242%.Average p between these two subspecies and P. demoleusdemoleus was only about 1%, whereas the Australian sub-species P. demoleus sthenelus had a 3.8% sequence diver-gence from other subspecies of P. demoleus. Papillio de-modocus samples from Africa did not form a monophyleticlineage. All Malagasy specimens of P. demodocus differedfrom each other by zero to two substitutions (p 5 0–0.242%)and grouped into a monophyletic assemblage (88% posteriorprobability) characterized by a single synapomorphic sub-stitution. Number of substitutions among African P. demo-docus varied from one to 10 (p 5 0.121–1.212%). The av-erage distance between P. demodocus from Africa and fromMadagascar was 0.78%, ranging from 0.36% to 1.212%. Allspecies of the P. demoleus group are well diverged, as in-dicated by the amount of sequence divergence ranging from4.9% to 6.5%.
Biogeography and Ancestral Areas
Dispersal-vicariance analysis (Ronquist 1996) was appliedto the inferred alternative cladograms (Fig. 7) with the dis-tribution of each species classified as present/absent in fourdifferent areas. There were several alternative optimizationsbut in all cases the most optimal reconstruction required threedispersal events (four if P. demodocus arrived in Madagascarnaturally) with different sequences of vicariance events.Based on this form of analysis, the ancestral area of the groupwas inferred to be Madagascar.
DISCUSSION
Conflicting Phylogenetic Signals in Mitochondrial DNA,Nuclear Genes, and Morphology
Except for the most basal relationships within the P. de-moleus species-group, as well as outgroups, we obtainedlargely congruent evidence from the multiple sources of data,including nucleotide sequences from three independent generegions (COI 1 COII, wg, EF-1a) and morphology. We ob-served more discordances between gene partitions and mor-phological data but here, as for mitochondrial versus nucleargenes, the conflict stemmed primarily from disagreementabout basal relationships. Tests for incongruence between
2772 EVGUENI V. ZAKHAROV ET AL.
TABLE 3. Morphological and nucleotide variability.
Morpho-logical
COI
All 1st 2nd 3rd
tRNA-leu
All
COII
All 1st 2nd 3rd
No. characters 60 1532 511 510 511 68 685 229 228 228
Ingroup only (Papilio demoleus group)No. variable 53 196 29 4 163 1 88 16 4 68No. informative 36 158 24 3 131 1 75 13 1 61Autapomorphies 17 38 5 1 32 0 13 3 3 7CI 0.679 0.831 0.882 1.000 0.824 1.000 0.872 0.800 1.000 0.902RI 0.591 0.840 0.900 1.000 0.831 1.000 0.884 0.833 1.000 0.917Ti/Tv 1.960 7.810 0.706 1.636 — 1.844 6.000 0.657 1.328Overall average p 0.051 0.022 0.003 0.129 0.005 0.051 0.027 0.005 0.119
Ingroup 1 outgroupNo. variable 58 462 92 15 355 5 205 48 9 148No. informative 50 315 56 5 254 2 132 25 3 100Autapomorphies 8 147 36 10 101 3 73 19 6 48CI 0.529 0.561 0.587 0.882 0.510 1.000 0.561 0.610 0.923 0.552RI 0.429 0.569 0.594 0.875 0.489 1.000 0.569 0.709 0.750 0.352Ti/Tv 1.130 5.587 1.244 0.877 1.450 1.311 6.562 0.620 0.923Overall average p 0.084 0.040 0.005 0.205 0.013 0.080 0.051 0.006 0.184
partitions revealed that COI and COII were the most con-gruent genes in our data, as should be expected because mi-tochondrial DNA genes are generally not considered to re-combine due to almost exclusively maternal inheritance ofmitochondrial DNA (with few exceptions, e.g., Kondo et al.1990; Kaneda et al. 1995) and therefore to represent a singlelocus with a single history. The observed disagreement be-tween the position of P. grosesmithi in COI and COII treesis thus likely to be due to nucleotide sampling error and thelower number of parsimony-informative characters in the sec-ond gene. Although some incongruence was observed be-tween mitochondrial DNA genes and nuclear genes, the al-ternative phylogenetic hypotheses inferred from nucleargenes were not very robust. Phylogeny estimation based onwg data (the partition with the lowest number of informativecharacters) had the lowest average bootstrap support for theingroup. In fact, there was strong correlation between numberof informative characters and average bootstrap support onthe tree for each gene partition (R2 5 0.75, data not shown).Increased numbers of characters have been shown to con-tribute to both accuracy (e.g., Cummings et al. 1995; Hillis1998) and support (e.g., Felsenstein 1985; Sanderson 1995;Bremer et al. 1999) of phylogenetic trees. Combined dataprovided robust phylogenetic reconstructions that we con-sidered the departure point for further examination of limeswallowtail evolution.
Phylogenetic Relationships
Hancock (1983) considered lime swallowtails to belong tothe subgenus Princeps (Princeps), with the Papilio menes-theus species-group as their sister-group. Based on experi-mental hybridization of P. demoleus with 13 other Papiliospecies, Ae (1979) considered P. dardanus and P. phorcasto be part of the P. demoleus group. However, there is littlecorrelation between hybridization success and genetic simi-larity of swallowtail species (Zakharov et al. 2004), and hy-bridization data can be misleading. A similar situation wasreported for water-striders in the genus Limnoporus, where
one species that was clearly closely related to others, as wasshown by mitochondrial DNA, allozymes, and morphology,demonstrated much higher hybrid incompatibility (Sperlinget al. 1997). Hauser et al. (2002) place the lime swallowtailswithin the subgenus Papilio (Princeps), a group of 19 speciesthat, with the single exception of P. demoleus, is restrictedto Africa and Madagascar.
The most recent evidence from a more comprehensive mo-lecular phylogeny of Papilio (Zakharov et al. 2004) indicatesthat Princeps itself is not monophyletic and lime swallowtailsoccupy a phylogenetic position between two Papilio subdi-visions, Menelaides and Achillides, which are recognized byHauser et al. (2002) as subgenera, with good support forMenelaides as the sister taxon to the P. demoleus group. Ourdata continue to support this relationship, as P. (Menelaides)helenus was found to be the closest outgroup to the P. de-moleus species-group. Papilio lormieri, from the P. menes-theus group, is one outgroup species that was added to thecurrent study to evaluate Hancock’s (1983) hypothesis thatthis species is closely related to the P. demoleus group. How-ever, this species was shown to have a stronger relationshipwith P. delalandei and P. constantinus, which are relativelydistantly related to the P. demoleus group.
Our results leave no doubt as to the monophyly of the P.demoleus species-group, and give strong resolution of intra-group relationships in the combined analyses. Our data do notagree with traditional assumptions about phylogenetic rela-tionships within the P. demoleus group, in which P. demodocusor P. demoleus are considered the most basal species (Car-casson 1964). None of the analyses supports a basal positionfor P. demodocus and this species instead appears to be rel-atively derived. The strongest evidence indicates a clear sisterrelationship between P. demodocus and P. erithonioides, withP. demoleus likely being their sister taxon. There is good sup-port for P. grosesmithi as the sister taxon of these three species,and a basal position for P. morondavana. Thus, the endemicspecies whose conservation status is of most concern, P. gro-sesmithi and P. morondavana, appear to represent the oldest
2773ORIGIN OF ISLAND ENDEMICS
TABLE 3. Extended.
Wg
All 1st 2nd 3rd
EF-1a
All 1st 2nd 3rd
404 135 134 135 1240 413 413 414
32 1 0 31 56 2 0 5424 1 0 23 48 2 0 54
8 0 0 8 8 0 0 80.943 1.000 — 0.941 0.851 1.000 — 0.8460.946 1.000 — 0.944 0.872 1.000 — 0.8652.930 — — 2.730 3.109 — — 2.9580.024 0.003 0.000 0.071 0.017 0.002 0.000 0.050
125 16 3 106 242 11 4 22775 2 1 72 132 4 2 12650 14 2 34 110 7 2 1010.665 1.000 1.000 0.643 0.625 0.846 1.000 0.6320.716 1.000 1.000 0.709 0.609 0.846 1.000 0.6232.282 0.878 — 2.147 3.477 1.538 0.671 3.4370.077 0.014 0.005 0.228 0.044 0.005 0.002 0.126
FIG. 2. Phylogeny for the Papilio demoleus species group andPapilio outgroups computed by PAUP branch-and-bound searchfrom 60 morphological characters, shown as a strict consensus ofsix most parsimonious trees (TL 5 140, CI 5 0.429, RI 5 0.529)rooted with P. thoas. Refer to Figure 3 for ingroup bootstrap andBremer support values.
lineages among lime swallowtails. These relic species speakfor additional protection being given to the dry deciduousforests of Madagascar as a conservation priority (Vane-Wrightet al. 1991; Stiassny and de Pinna 1994).
Another interesting finding is the degree of genetic vari-
ation among and within species of the P. demoleus group.Although percent sequence divergence is quite variableamong closely related Lepidoptera and is not necessarily areliable indicator of species status, there is an obvious trendwith overall sequence divergence increasing at higher taxo-nomic levels. Almost 98% of species recognized through pri-or morphological studies are reported to fit a species levelthreshold value of 3% divergence in COI proposed by Hebertet al. (2003). However, the amount of sequence divergencein COI at the interspecific level within lepidopteran speciescomplexes ranges from 0.1–0.6% (0.4% average) in erminemoths Yponomeuta (Sperling et al. 1995), 0–3.6% (2.1% av-erage) in Choristoneura (Sperling and Hickey 1994), and 0.1–3.7% (2.2% average) in Feltia (Sperling et al. 1996). Diver-gences of less than 1% and up to to 2.5% divergence in COIwas reported for sister species of torticid moths of the genusArchips (Kruse and Sperling 2001). Higher levels of variationin interspecific mitochondrial DNA divergences was dem-onstrated for the one of the most well studied species groupsof the genus Papilio, the P. machaon complex, with valuesranging from ,1% to 8% (Sperling and Harrison 1994).
In lime swallowtails, intraspecific mitochondrial DNA di-vergences ranged from nothing within P. morondavana andP. erithonioides to less than 0.5% between closely relatedsubspecies of P. demoleus, but up to almost 4% sequencedivergence between more distant subspecies of P. demoleus.At the same time, the species themselves were separated bynucleotide differences of 5–6%, which leaves no doubt thatall lime swallowtail species are very distinct by any com-parison to other Lepidoptera species.
Lack of variation in the COI gene in P. erithonioides andP. morondavana, sampled in different localities and time se-ries, requires further investigation. While P. erithonioidesmay still be too evolutionarily young to have accumulatedmuch variation, the latter species is probably a relic. Lackof mitochondrial DNA variation in P. morondavana mightindicate that the ancestral population has gone through oneor more recent bottlenecks.
2774 EVGUENI V. ZAKHAROV ET AL.
FIG. 3. Phylogenetic relationships for species of the Papilio demoleus group inferred from partitioned and combined molecular andmorphological data using maximum parsimony (MP) and Bayesian analyses. Trees represent ingroup relationships from MP bootstrapconsensus trees that were identical to ingroup topologies of MP trees in each analysis that included full set of outgroups (except F andH) and were identical to all Bayesian reconstructions (except H). For F (wg 1 EF-1a), two MP trees had the given ingroup topology;two other trees showed sister relationships for P. grosesmithi with P. demoleus 1 P. morondavana. For H (morphological data), the MPbootstrap consensus is on the left, and Bayesian reconstruction is on the right. Numbers above nodes show bootstrap proportions andBremer support; numbers under nodes indicate Bayesian posterior probabilities.
Biogeography
The approximate age of divergence of the P. demoleusspecies-group from other Papilio is estimated to be about16.8 6 6.7 million years (Zakharov et al. 2004). This dateexcludes any possibility for older vicariance via Cretaceousplate tectonics. Recent reports of invasion of P. demoleusinto islands in Southeast Asia where the species was origi-nally absent indicate the ability of lime swallowtails to ex-pand their range, whether through active flight or with theaid of man and Citrus cultivation (Hiura 1973; Corbet andPendlebury 1978, 1992; Vane-Wright and de Jong 1993).
Phylogenetic reconstruction clearly indicates a basal po-sition for P. morondavana and P. grosesmithi within the P.demoleus species-group. Based on estimated substitutionrates for COI and COII genes (Zakharov et al. 2004), the
ancestors for P. morondavana and the rest of lime swallow-tails diverged between approximately 14.1 and 10.6 millionyears ago, and the split of the ancestor for P. grosesmithifrom its sister group is dated at between 13.5 and 10.3 millionyears ago Dispersal-vicariance analysis based on these re-lationships suggests their origin and speciation within Mad-agascar (Fig. 7), followed by dispersal of lime swallowtailsinto Australia and Asia (becoming P. demoleus) around 10.1to 7.8 million years ago, and Africa (becoming P. demodocus)around 7.3 to 5.6 million years ago.
Due to weakly supported outgroup relationships, our dataare insufficient to resolve with confidence how the ancestorof lime swallowtails came to Madagascar. Based on a cur-rently available molecular phylogeny for the genus Papilio(Zakharov et al. 2004), the P. demoleus species group had a
2775ORIGIN OF ISLAND ENDEMICS
FIG. 4. One of two equally parsimonious trees (TL 5 2533, CI 5 0.548, RI 5 0.554) based on combined dataset. Most parsimonioustrees were different only in some outgroup rearrangements. Bayesian analysis produced tree with almost identical topology (-lnL 517024.14) with different position of Papilio anactus shown by a circled star. Histograms above nodes show partitioned Bremer supports,numbers under nodes indicate bootstrap proportions, total Bremer support, and clade posterior probabilities from Bayesian analyses.
common ancestor with an Oriental lineage of Papilio, thesubgenus Menelaides. Together they are the sister group toPapilio (Achillides), another Oriental subgenus. This suggeststhat an Asian ancestor of the P. demoleus species-group mayhave first reached Madagascar in a manner similar to thehypothesized origin of the Madagascan mycalesine butterflies(Lees 1997; Torres et al. 2001). This scenario is also sup-ported by a close approach to Madagascar of two lineagesof an Indo-Australian genus Euploea (Lepidoptera: Nym-phalidae) to the Seychelles and Mascarene Islands (Ackeryand Vane-Wright 1984).
Colonization of Madagascar would have been followed byspeciation of Malagasy endemics, succeeded by dispersal ofthe ancestors of each of P. demodocus and P. demoleus toAfrica and back to the Oriental region and Australia. Underthe best-supported parsimony reconstructions, P. moronda-
vana, P. grosesmithi, and P. erithonioides would have di-verged from each other on Madagascar itself, whether ondifferent parts of the island or by a sympatric process wherethey were in contact with each other.
The probability of dispersal across the Indian Ocean (c.4000 km) may seem relatively low compared to a scenarioinvolving multiple colonizations of Madagascar from Africa(separated by just c. 400 km). Nonetheless, a similar hy-pothesis has also been proposed for the origin of other rep-resentatives of the Malagasy fauna. Jansa et al. (1999) sug-gested a single invasion of native rodents (Muridae: Neso-myinae) to Madagascar from Asia followed by secondaryinvasion from Madagascar into Africa (but see Jansa andWeksler 2004). Lees (1997) and Torres et al. (2001) alsosuggested that the Malagasy mycalesine butterfly radiationmay have originated from India rather than Africa, and that
2776 EVGUENI V. ZAKHAROV ET AL.
FIG. 5. Sampled localities and distribution of 27 unique haplotypes identified from 50 ingroup individuals of Papilio demoleus speciesgroup from 825-bp fragment of the COI gene. Haplotypes shared between different localities are shown in regular font. Haplotypesrestricted to a single locality are shown as bold. For haplotype definitions and complete list of specimens, refer to Table 1. Arrowsindicate prevailing wind and currents in the Indian Ocean that are influenced by the Asiatic monsoon and trade winds (Walter 1970),including North Equatorial Current (I), Mozambique Current (II), Alguhas Stream (III), West Wind Drift Current (IV), West AustralianCurrent (V), and South Equatorial Current (VI).
a dispersal event from India was followed by subsequentcolonization of Africa by founders from Madagascar.
Long-distance dispersal has been proposed to explain sim-ilar patterns of relationships in chameleons, where the oldestlineages are distributed in Madagascar, and more recentlyderived forms that are found in Africa, the Seychelles, theComoros, and India are believed to be the result of severaldispersal events across the Indian Ocean (Raxworthy et al.2002). Both the oldest fossil records that are available forchameleons and geological age of the volcanic islands ofComoros are substantially younger than the time of Creta-ceous vicariance (see Raxworthy et al. 2002). The Comorosarchipelago has never been in contact with larger land massessince its formation within the last 5.4 million years (Emerickand Duncan 1982); thus, the only possible way the ancestorof the endemic chameleons endemic to these islands couldhave arrived there is by transmarine dispersal.
Oceanic dispersal of terrestrial animals to and from Mad-agascar and other islands in the western Indian Ocean hasbeen favored by many other studies that hypothesized post-Gondwanan transmarine migration by rafting on tangles ofvegetation (e.g., Fisher 1996; Yoder et al. 1996; Nagy et al.2003; Vences et al. 2003). Dispersal across an uninhabitablespace is even more plausible when a taxon possesses a sig-
nificant degree of vagility, as in actively flying Lepidopteraand lime swallowtails in particular.
It also appears that the Malagasy biota has provided sourc-es for multiple radiations that have contributed to the bio-diversity of the surrounding continents and islands. For ex-ample, reconstruction of island radiations in extinct and ex-tant geckos using ancient and recent DNA by Austin et al.(2004) suggested that at least four different archipelagos inthe Indian Ocean have been colonized independently by dis-persals from Madagascar. According to Warren et al. (2003),Madagascar has given rise to two independent sunbird lin-eages in the Aldabra archipelago.
Complex seasonal systems of oceanic currents and windsin the Indian Ocean provide conditions for long-distancetransmarine dispersals (see Fig. 5). The currents of the north-ern Indian Ocean are influenced by the Asiatic monsoon whilethe currents of the southern Indian Ocean are influenced byanticyclonic circulation in the atmosphere (Walter 1970). Thenorthwest monsoon stimulates the North Equatorial Current,which flows from east to west and upon reaching the eastcoast of Africa the largest portion turns southward and even-tually becomes the Mozambique Current. The MozambiqueCurrent flows south along the east coast of Africa and atabout 358S merges with the Alguhas Stream, which flows
2777ORIGIN OF ISLAND ENDEMICS
FIG. 6. Bayesian estimation of the phylogeny of 27 unique haplotypes (from 50 ingroup samples) and 11 outgroups inferred from 825bp from the 39 end fragment of COI (106 cycles, each 100th sample, contype 5 halfcompat, burnin 5 5000). Numbers above nodesindicate clade posterior probabilities, and under nodes indicate bootstrap proportions from maximum parsimony analyses. For haplotypedefinitions and species list, refer to Table 1. For geographic distribution of haplotypes, see Figure 5.
westward along the southern coast of Madagascar and theeast African coast toward the Cape of Good Hope, where itpartly joins the West Wind Drift Current, which then carriesits waters toward southwest Australia. During the southwestmonsoon in August and September, the North Equatorial Cur-rent reverses direction and flows west to east as the MonsoonCurrent.
Thus, a westward dispersal of lime swallowtails across thenorthern part of the Mozambique Channel is more probablebased on prevailing winds and has been inferred to occur onseveral occasions in Lepidoptera (Pierre 1992; Torres et al.
2001), though on more tenuous phylogenetic grounds. Pre-vailing easterly trade winds and ocean currents in the IndianOcean also increase the probability of long-distance dispersalfrom Asia to the western Indian Ocean (Donque 1972; Ren-voise 1979). At the same time, dispersal from southern Mad-agascar and southern Africa can be facilitated by West WindDrifts, as has been hypothesized for some sea stars (Watersand Roy 2004).
Alternatively, the assumptions of dispersal-vicarianceanalysis may simply be wrong, in that the distribution ofthese butterflies may not be parsimonious. For example, ex-
2778 EVGUENI V. ZAKHAROV ET AL.
FIG. 7. Phylogenies of the Papilio demoleus group inferred from (a) the morphology bootstrap consensus (this topology was alsorecovered from COII data partition); (b) COI-COII; (c) wg; and (d) EF-1a, with areas optimized by dispersal-vicariance analysis (Ronquist1996). Each species is coded as present in a particular area, disregarding human introductions. Alternative equally parsimonious opti-mizations are separated by a semicolon at each node. Composite areas (e.g., AB) indicate that the program has calculated the combinedarea as the optimal solution. Maps illustrate the present-day distribution of lime swallowtails.
tinction of intermediate or ancestral taxa would complicateand obscure reconstruction of dispersal, refuting an indige-nous origin of the group within Madagascar. Under a scenarioof dispersal from Africa, there could have been a number ofsuccessive arrivals to Madagascar from an anageneticallyevolving stem lineage, as suggested by Carcasson (1964).
Additional uncertainty is associated with the origin of P.demodocus in Madagascar. It is reported that the presence ofP. demodocus on the island is the result of introduction byman (Paulian 1951; Paulian and Viette 1968; Turlin 1994).
The monophyly of mitochondrial DNA haplotypes found inMadagascan samples of P. demodocus relative to Africanindividuals may be a consequence of founder effect from asingle introduction. However, the fact that the mitochondriallineage already displays moderate genetic variability (fourhaplotypes of seven samples) suggests that colonization ofMadagascar by this lineage may have occurred thousands ofyears ago or even longer, prior to island colonization by manabout 2000 years ago (Dewar 1997).
The origin of P. demoleus in Saudi Arabia and Iran is also
2779ORIGIN OF ISLAND ENDEMICS
believed to be the result of introduction along with the firstimportation of Citrus in the 10th century (Wiltshire 1945;Larsen 1977). The 1% sequence divergence between our sam-ple of P. demoleus from Iran and our samples from SoutheastAsia and eastern Asia indicates that this introduction wouldprobably have come from closer sources, such as the Indiansubcontinent (from which we were unable to obtain samples),which already had a moderate amount of divergence frompopulations farther to the east. Interestingly, all Asian indi-viduals of P. demoleus had very similar or identical haplotypes,supporting the hypothesis of recent range expansion in thisregion. In contrast, there was a striking difference in mito-chondrial DNA between Oriental subspecies and the Austra-lian subspecies P. d. sthenelus. Taking into account host usedifferences between this and other subspecies of P. demoleus,these haplotypes appear to be well-diverged lineages thatmight be considered separate species (Hebert et al. 2003). Atminimum, it appears likely that ancestral populations of P.demoleus in the Oriental and the Australian regions have beenisolated for a long time, from 3.24 to 4.24 million years agobased on substitution rates estimated for mitochondrial DNAin Papilio (Zakharov et al. 2004). Thus P. d. sthenelus mayrepresent the oldest lineage in P. demoleus. More detailedsampling across its range is required to reconstruct the evolutionary history of this species on both sides of Wallace’s line.
Conclusion
Biogeographic scenarios are strongly dependent on the se-lection of phylogenetic hypotheses for a particular group oforganisms, while phylogenetic reconstructions themselvescan be subject to different biases in data. Thus, selection ofinformative loci for a given taxonomic level remains key tounderstanding the evolution and biogeography of any groupof organisms. In our reconstructions of the phylogeny ofspecies in the P. demoleus group, nucleotide sequences haveprovided a useful source of data in addition to traditionalmorphology. Comparison of four genes representing threeindependent loci supports the utility of the COI gene in re-constructing the phylogeny of closely related species, suchas in the P. demoleus group.
Our results provide strong evidence for the basal relation-ships of Malagasy endemic species in the P. demoleus group.Relationships among lime swallowtails on Madagascar in-clude elements similar to those reported in nymphalid but-terflies and even chameleons, indicating not only that post-Cretaceous dispersal across large oceanic barriers has con-tributed to the diversity of Madagascar, but that the Malagasybiota has speciated actively on the island, serving as an in-cubator for radiations that have contributed to the diversityof surrounding continents on all sides of the Indian Ocean.Our study may be the best documented case of this patternto date, despite the fact that morphology does not yet seemto provide enough resolution both for ingroup lime swallow-tails and outrgoup species of Papilio, as it is supported bymultiple independent genes.
Of more immediate importance, our data support a rec-ommendation that further protection should be given to thedry forests of Madagascar (generally underrepresented withinthe protected area system; Dupuy and Moat 1996) to ensure
survival of the endemic species of lime swallowtails in Mad-agascar. These large and charismatic butterflies give new in-sight into the complex history of the unique and ancient bio-diversity of this island, as well as add weight to the urgencyof arresting the rapid clearance of dry deciduous forests onthe island via creation of new protected areas, for the mutualbenefit of other species restricted to these ecosystems.
ACKNOWLEDGMENTS
We are grateful to the following people for help in pro-viding specimens for this study: N. Pierce, D. Lohman, andM. Cornwall from Harvard University (Cambridge, MA,USA); N. Tatarnik and V. Nazari from the University ofAlberta (Edmonton, AB, Canada); H. F. Wong from DecoEnterprise (Taiping, Malaysia); S. Schoeman from ARC, In-stitute for Tropical and Subtropical Crops (South Africa); M.G. Wright from the University of Hawaii (Honolulu, HI,USA); D. Goh from the Penang Butterfly Farm (Penang, Ma-laysia); S. A. Ae (Gifu, Japan); S. H. Yen from The NaturalHistory Museum (London, United Kingdom); and J. Demayfrom SNC Ornithoptera (Agny, France). This study was fund-ed by an NSERC grant to FAHS. Collections from Mada-gascar were partially supported by grant DEB-0072713 fromthe National Science Foundation to B. L. Fisher and C. E.Griswold. Fieldwork that provided the basis for this workcould not have been completed without the gracious supportof the Malagasy people. We are grateful for comments on anearlier version of this manuscript by A. Yoder, N. Wahlberg,and an anonymous reviewer.
LITERATURE CITED
Ackery, P. R., and R. I. Vane-Wright. 1984. Milkweed butterflies:their cladistics and biology. BMNH, London, U.K. and CornellUniv. Press, Ithaca, NY.
Ackery, P. R., C. Smith, and R. I. Vane-Wright. 1995. Carcasson’sAfrican butterflies: an annotated catalogue of the Papilionoideaand Hesperoidea of the Afrotropical region. CSIRO Publishing,Melbourne.
Ae, S. A. 1979. The phylogeny of some Papilio based on interspe-cific hybridization data. Syst. Entomol. 4:1–16.
Austin, J. J., E. N. Arnold, and C. G. Jones. 2004. Reconstructingan island radiation using ancient and recent DNA: the extinctand living day geckos (Phelsuma) of the Mascarene islands. Mol.Phylogenet. Evol. 31:109–122.
Baker, R. H., and R. DeSalle. 1997. Multiple sources of molecularcharacteristics and the phylogeny of Hawaiian drosophilids.Syst. Biol. 46:654–673.
Braby, M. F. 2000. Butterflies of Australia, their identification,biology and distribution. CSIRO Publishing, Collingwood, Aus-tralia.
Bremer, B., R. K. Jansen, B. Oxelman, M. Backland, H. Lantz, andK.-J. Kim. 1999. More characters or more taxa for a robustphylogeny: case study from the coffee family (Rubiaceae). Syst.Biol. 48:413–435.
Bremer, K. 1988. The limits of amino acid sequence data in an-giosperm phylogenetic reconstruction. Evolution 42:795–803.
Briggs, J. C. 2003. The biogeographic and tectonic history of India.J. Biogeogr. 30:381–388.
Brower, A. V. Z., and R. DeSalle. 1998. Patterns of mitochondrialversus nuclear DNA sequence divergence among nymphalid but-terflies: the utility of wingless as a source of characters for phy-logenetic inference. Insect Mol. Biol. 7:73–82.
Caccone, A., G. Amato, O. C. Gratry, J. Behler, and J. R. Powell.1999. A molecular phylogeny of four endangered Madagascar
2780 EVGUENI V. ZAKHAROV ET AL.
tortoises based on mtDNA sequences. Mol. Phylogenet. Evol.12:1–9.
Carcasson, R. H. 1964. A preliminary survey of the zoogeographyof African butterflies. East Afr. Wildl. J. 2:122–157.
Carroll, S. B., J. Gates, D. N. Keys, S. W. Paddock, G. E. Pan-ganiban, J. E. Selegue, and J. A. Williams. 1994. Pattern for-mation and eyespot determination in butterfly wings. Science265:109–114.
Caterino, M. S., and F. A. H. Sperling. 1999. Papilio phylogenybased on mitochondrial cytochrome oxidase I and II genes. Mol.Phylogenet. Evol. 11:122–137.
Caterino, M. S., S. Cho, and F. A. H. Sperling. 2000. The currentstate of insect molecular systematics: a thriving Tower of Babel.Annu. Rev. Entomol. 45:1–54.
Caterino, M. S., R. D. Reed, M. M. Kuo, and F. A. H. Sperling.2001. A partitioned likelihood analysis of swallowtail butterflyphylogeny (Lepidoptera: Papilionidae). Syst. Biol. 50:106–127.
Cibois, A., B. Slikas, T. S. Schulenberg, and E. Pasquet. 2001. Anendemic radiation of Malagasy songbirds is revealed by mito-chondrial DNA sequence data. Evolution. 55:1198–1206.
Clarke, C. A., C. G. C. Dickson, and P. M. Sheppard. 1963. Larvalcolor pattern in Papilio demodocus. Evolution 17:130–137.
Collins, N. M., and M. G. Morris. 1985. Threatened swallowtailbutterflies of the world. The IUCN Red Data Book. IUCN, Gland,Switzerland.
Conservation International. 2003. Madagascar forest cover andclearance ;1900 and ;2000. Map. Center for Applied Biodi-versity Science, Conservation International, Washington, DC.
Cooper, A., C. Lalueza-Fox, S. Anderson, A. Rambaut, J. Austing,and R. Ward. 2001. Complete mitochondrial genome sequencesof two extinct moas clarify ratite evolution. Nature 409:704–707.
Corbet, A. S. 1944. Paleontology without fossils in the ‘birdwing’butterflies. Nature 153:32.
Corbet, A. S., and H. M. Pendlebury. 1978. The butterflies of theMalay Peninsula. 3rd ed. Rev. by J. N. Eliot. Malayan NatureSociety, Kuala Lumpur.
——— 1992. The butterflies of the Malay Peninsula. 4th ed. Rev.By J. N. Eliot. Malayan Nature Society, Kuala Lumpur.
Cowie, R. H. 2001. Decline and homogenization of Pacific faunas:the land snails of American Samoa. Biol. Conserv. 99:207–222.
Cummings, M. P., S. P. Otto, and J. Wakely. 1995. Sampling prop-erties of DNA sequence data in phylogenetic analysis. Mol. Biol.Evol. 12:814–822.
Dewar, R. E. 1997. Were people responsible for the extinction ofMadagascar’s subfossils, and how will we ever know? Pp. 364–377 in S. M. Goodman and B. D. Patterson, eds. Natural changeand human impact in Madagascar. Smithsonian Institution Press,Washington, DC.
Donque, G. 1972. The climatology of Madagascar. Pp. 87–142 inR. Battistini and G. Richard-Vindard, eds. Biogeography andecology in Madagascar. Dr. W. Junk B. V. Publishers, TheHague, The Netherlands.
Dorst, J. 1972. The evolution and affinities of the birds of Mada-gascar. Pp. 615–627 in R. Battistini and G. Richard-Vindard,eds. Biogeography and ecology in Madagascar. Dr. W. Junk B.V. Publishers, The Hague, The Netherlands.
Dupuy, D. J., and J. Moat. 1996. A refined classification of theprimary vegetation of Madagascar based on the underlying ge-ology: using GIS to map its distribution and to assess its con-servation status. Pp. 205–218 in W. R. Lourenco, ed. Bioge-ography of Madagascar. Editions Orstom, Paris.
Emerick, C. M., and R. A. Duncan. 1982. Age progressive volca-nism in the Comores Archipelago, western Indian Ocean andimplications for Somali plate tectonics. Earth Planet. Sci. Lett.60:415–428.
Felsenstein, J. 1985. Confidence limits on phylogenies: an approachusing the bootstrap. Evolution 39:783–791.
Fisher, B. L. 1996. Origins and affinities of the ant fauna of Mad-agascar. Pp. 457–465 in W. R. Lourenco, ed. Biogeography ofMadagascar. Editions Orstom, Paris.
Gargominy, O., P. Bouchet, M. Pascal, T. Jaffre, and J. C. Tourneur.1996. Consequences for biodiversity of plant and animal speciesintroductions in New Caledonia. Rev. Ecol. Terre Vie 51:375–402.
Gillespie, R. G., and G. K. Roderick. 2002. Arthropods on islands:colonization, speciation, and conservation. Annu. Rev. Entomol.47:595–632.
Goloboff, P. 1999. NONA (NO NAME). Ver. 2. Published by theauthor, Tucuman, Argentina. Available via http://www.cladistics.com/aboutNona.htm.
Grant, P. R., and B. R. Grant. 2002. Adaptive radiation of Darwin’sfinches. Am. Sci. 90:130–139.
Hall, T. A. 1999. BioEdit: a user-friendly biological sequence align-ment editor and analysis program for Windows 95/98/NT. Nu-cleic Acids Symp. Ser. 41:95–98.
Hancock, D. L. 1983. Classification of the Papilionidae (Lepidop-tera): a phylogenetic approach. Smithersia 2:1–48.
Hauser, C. L., R. de Jong, G. Lamas, R. K. Robbins, C. Smith, andR. I. Vane-Wright. 2002. Papilionidae: revised GloBIS/GARTspecies checklist. 2nd draft. http://www.insects-online.de/frames/papilio.htm.
Hebert, P. D. N., S. Ratnasingham, and J. R. deWaard. 2003. Bar-coding animal life: cytochrome c oxidase subunit 1 divergencesamong closely related species. Proc. R. Soc. Lond. B. Biol. Sci.270:S96–S99.
Hillis, D. M. 1998. Taxonomic sampling, phylogenetic accuracy,and investigator bias. Syst. Biol. 47:3–8.
Hiura, I. 1973. Butterflies flying across the sea. Soju Shobo, Tokyo.Huelsenbeck, J. P., and J. P. Bollback. 2001. Empirical and hier-
archical Bayesian estimation of ancestral states. Syst. Biol. 50:351–366.
Huelsenbeck, J. P., and F. Ronquist. 2001. MR BAYES: Bayesianinference of phylogeny. Bioinformatics 17:754–755.
IUCN. 2003. 2003 IUCN Red List of Threatened Species. http://www.redlist.org.
Jansa, S., and M. Weksler. 2004. Phylogeny of muroid rodents:relationships within and among major lineages as determined byIRBP gene sequences. Mol. Phylogenet. Evol. 31:256–276.
Jansa, S., S. M. Goodman, and P. K. Tucker. 1999. Molecular phy-logeny and biogeography of the native rodents of Madagascar(Muridae: Nesomyinae): a test of the single-origin hypothesis.Cladistics 15:253–270.
Jordan, S., C. Simon, and D. Polhemus. 2003. Molecular systematicsand adaptive radiation of Hawaii’s endemic damselfly genusMegalagrion (Odonata: Coenagrionidae). Syst. Biol. 52:89–109.
Kaneda, H., J. Hayashi, S. Takahama, C. Taya, K. F. Lindahl, andH. Yonekawa. 1995. Elimination of paternal mitochondrial DNAin intraspecific crosses during early mouse embryogenesis. Proc.Natl. Acad. Sci. USA 92:4542–4546.
Koechlin, J. 1972. Flora and vegetation of Madagascar. Pp. 145–190 in R. Battistini and G. Richard-Vindard, eds. Biogeographyand ecology in Madagascar. Dr. W. Junk B.V. Publishers, TheHague.
Kondo, R., Y. Satta, E. T. Matsuura, H. Ishiwa, N. Takahata, andS. I. Chigusa. 1990. Incomplete maternal transmission of mi-tochondrial DNA in Drosophila. Genetics 126:657–663.
Krause, D. W., R. R. Rogers, C. A. Forster, J. H. Hartman, G. A.Buckley, and S. D. Sampson. 1999. The Late Cretaceous ver-tebrate fauna of Madagascar: implications for Gondwanan pa-leobiogeography. Geol. Soc. Am. Today 9:2–7.
Kruse, J. J., and F. A. H. Sperling. 2001. Molecular phylogenywithin and between species of the Archips argyrospila complex(Lepidoptera: Torticidae). Ann. Entomol. Soc. Am. 94:166–173.
Kumar, S., K. Tamura, I. B. Jakobsen, and M. Nei. 2001. MEGA2:molecular evolutionary genetics analysis software. Bioinfor-matics 17:1244–1245.
Larsen, T. B. 1977. Extension recente en Iraq de l’aire de Papiliodemoleus. Linn. Entomops 42:37–38.
Lees, D. C. 1997. Systematics and biogeography of Madagascanmycalesine butterflies (Lepidoptera: Satyrinae). Ph.D. diss., Uni-versity of London.
Lees, D. C., C. Kremen, and T. Raharitsimba. 2003. Classification,diversity, and endemism of the butterflies (Papilionoidea andHesperioidea): a revised species checklist. Pp. 762–793 in S. M.Goodman and J. Benstead, eds. The natural history of Mada-gascar. Univ. of Chicago Press, Chicago.
MacArthur, R. H., and E. O. Wilson. 1967. The theory of islandbiogeography. Princeton Univ. Press, Princeton, NJ.
2781ORIGIN OF ISLAND ENDEMICS
Matsumoto, K. 2002. Papilio demoleus (Papilionidae) in Borneoand Bali. J. Lepid. Soc. 56:108–111.
McKenzie, D. P., and J. C. Sclater. 1973. The evolution of theIndian Ocean. Sci. Am. 228:63–72.
Myers, N., R. A. Mittermeier, C. G. Mittermeier, G. A. B. DaFonseca, and J. Kent. 2000. Biodiversity hotspots for conser-vation priorities. Nature 403:853.
Nagy, Z. T., U. Joger, M. Wink, F. Glaw, and M. Vences. 2003.Multiple colonization of Madagascar and Socotra by colubridsnakes: evidence from nuclear and mitochondrial gene phylog-enies. Proc. R. Soc. Lond. B. Biol. Sci. 270:2613–2621.
Nixon, K. C. 2002. Winclada. Ver. 1.00.08. Published by the author.Available via http://www.cladistics.com/aboutwinc.htm.
O’Dowd, D. J., P. T. Green, and P. S. Lake. 2003. Invasional ‘melt-down’ on an oceanic island. Ecol. Lett. 6:812–817.
Parsons, M. 1999. Butterflies of Papua New Guinea. AcademicPress, San Diego, CA.
Paulian, R. 1951. Papillons communs de Madagascar. Publicationsof the Institute Recherche Scientifique, Tananarive, Madagascar.
——— 1972. Some ecological and biogeographical problems of theentomofauna of Madagascar. Pp. 411–426 in R. Battistini and G.Richard-Vindard, eds. Biogeography and ecology in Madagascar.Dr. W. Junk B. V. Publishers, The Hague, The Netherlands.
Paulian, R., and P. Viette. 1968. Insects. Lepidopteres Papilionidae.Faune de Madagascar 27. Museum d’Histoire Naturelle, Paris.
Pierre, J. 1992. Les Acraeidae des Comores et de Madagascar (Lep-idopteres Rhopaloceres). Entomologiste 48:351–364.
Rabinowitz, P. D., M. F. Coffin, and D. Falvey. 1983. The separationof Madagascar and Africa. Science 220:67–69.
Raxworthy, C. J., M. R. J. Forstner, and R. Nussbaum. 2002. Cha-meleon radiation by oceanic dispersal. Nature 415:784–787.
Reed, R. D., and F. A. H. Sperling. 1999. Interaction of processpartitions in phylogenetic analysis: an example from the swal-lowtail butterfly genus Papilio. Mol. Biol. Evol. 16:286–297.
Renvoise, S. A. 1979. The origins of Indian Ocean island floras.Pp. 107–129 in D. Bramwell, ed. Plants and islands. AcademicPress, London.
Ronquist, F. 1996. DIVA (dispersal-vicariance analysis). Ver. 1.1.Computer program and manual available by anonymous FTPfrom Uppsala University (ftp.uu.se or ftp.systbot.uu.se).
Sanderson, M. J. 1995. Objections to bootstrapping phylogenies: acritique. Syst. Biol. 44:299–320.
Schatz, G. E. 1996. Malagasy/Indo-Australo-Malesian phytogeo-graphic connections. Pp. 73–84 in W. R. Lourenco, ed. Bioge-ography of Madagascar. Editions Orstom, Paris.
——— 2001. Generic tree flora of Madagascar. (Flore Generiquedes Arbres de Madagascar). Missouri Botanical Garden and Roy-al Botanic Gardens, Kew, U.K.
Sorenson, M. D. 1999. TreeRot. Ver. 2. Boston University, Boston,MA.
Sperling, F. A. H. 2003. Butterfly species and molecular phylog-enies. Pp. 431–458 in C. Boggs, W. Watt, and P. Ehrlich, eds.Butterflies: evolution and ecology taking flight. Univ. of ChicagoPress, Chicago.
Sperling, F. A. H., and R. G. Harrison. 1994. Mitochondrial DNAvariation within and between species of the Papilio machaongroup of swallowtail butterflies. Evolution 48:408–422.
Sperling, F. A. H., and D. A. Hickey. 1994. Mitochondrial DNAsequence variation in the spruce budworm species complex(Choristoneura: Lepidoptera): Mol. Biol. Evol. 11:656–665.
Sperling, F. A. H., J.-F. Landry, and D. Hickey. 1995. DNA-basedidentification of introduced ermine moth species in North Amer-ica (Lepidoptera: Yponomeutidae) Ann. Entomol. Soc. Am. 88:155–162.
Sperling, F. A. H., R. Byers, and D. Hickey. 1996. MitochondrialDNA sequence variation among pheromotypes of the dingy cut-worm, Feltia jaculifera (Gn) (Lepidoptera:Noctuidae). Can. J.Zool. 74:2109–2117.
Sperling, F. A. H., J. R. Spence, and N. M. Andersen. 1997. Mi-tochondrial DNA, allozymes, morphology, and hybrid compat-ibility in Limnoporus water striders (Heteroptera: Gerridae): Dothey all track species phylogenies? Ann. Entomol. Soc. Am. 90:401–415.
Stiassny, M. L. J., and M. C. C. de Pinna. 1994. Basal taxa and the
role of cladistic patterns in the evaluation of conservation pri-orities: a view from freshwater. Pp. 235–249 in P. L. Forey, C.J. Humphries, and R. I. Vane-Wright, eds. Systematics and con-servation evaluation. Oxford Univ. Press, Oxford, U.K.
Storey, M., J. J. Mahoney, A. D. Saunders, S. Kelley, R. A. Duncan,and M. Coffin. 1995. Timing of hot-spot related magmatism andthe breakup of Madagascar and India. Science 267:852–855.
Swofford, D. 1998. PAUP: phylogenetic analysis using parsimony.Ver. 4.0. Illinois Natural History Survey, Champaign, IL.
Torres, E., D. C. Lees, R. I. Vane-Wright, C. Kremen, J. A. Leonard,and R. K. Wayne. 2001. Examining monophyly in a large ra-diation of Madagascan butterflies (Lepidoptera: Satyrinae: My-calesina) based on mitochondrial DNA data. Mol. Phylogenet.Evol. 20:460–473.
Treutlein, J., and M. Wink. 2002. Molecular phylogeny of cycadsinferred from rbcL sequences. Naturwissenschaften 89:221–225.
Turlin, B. 1994. Faune Lepidopterologique de l’Archipel des Co-mores (Ocean Indien), (Rhopaloceres, Sphingidae, Attacidae).Lambillionea 94:81–94.
Vane-Wright, R. I. 2003. Evidence and identity in butterfly sys-tematics. Pp. 477–513 in C. L. Boggs, W. B. Watt, and P. R.Ehrlich, eds. Butterflies: ecology and evolution taking flight.Univ. of Chicago Press, Chicago.
Vane-Wright, R. I., and R. de Jong. 2003. The butterflies of Su-lawesi: annotated checklist for a critical island fauna. Zoolo-gische Verhandelingen 343:3–267.
Vane-Wright, R. I., C. J. Humphries, and P. H. Williams. 1991.What to protect? Systematics and the agony of choice. Biol.Conserv. 55:235–254.
van Son, G. 1949. The butterflies of southern Africa. Part I. Pap-ilionidae and Pieridae. Transvaal Museum Memoirs 3.
Vences, M., J. Freyhof, R. Sonnenberg, J. Kosuch, and M. Veith.2001. Reconciling fossils and molecules: Cenozoic divergenceof cichlid fishes and the biogeography of Madagascar. J. Bio-geogr. 28:1091–1099.
Vences, M., D. R. Vieites, F. Glaw, H. Brinkmann, J. Kosuch, M.Veith, and A. Meyer. 2003. Multiple overseas dispersal in am-phibians. Proc. R. Soc. Lond. B. Biol. Sci. 270:2435–2442.
Vences, M., J. Kosuch, M. O. Rodel, S. Lotters, A. Channing, F.Glaw, and W. Bohme. 2004. Phylogeography of Ptychadenamascareniensis suggests transoceanic dispersal in a widespreadAfrican-Malagasy frog lineage. J. Biogeogr. 31:593–601.
Wahlberg, N., R. Oliveira, and J. A. Scott. 2003. Phylogenetic re-lationships of Phycoides butterfly species (Lepidoptera:Nym-phalidae): complex mtDNA variation and species delimitations.Syst. Entomol. 28:257–273.
Walter, D. 1970. The moonsoon regime of the currents in the IndianOcean. East-West Center Press, Honolulu, HI.
Warren, B. H., E. Bermingham, R. C. K. Bowie, R. P. Prys-Jones,and C. Thebaud. 2003. Molecular phylogeography reveals islandcolonization history and diversification of western Indian Oceansunbirds (Nectarinia: Nectariniidae). Mol. Phylogenet. Evol. 29:67–85.
Waters, J. M., and M. S. Roy. 2004. Out of Africa: the slow trainto Australia. Syst. Biol. 53:18–24.
Whittaker, R. J. 1998. Island biogeography. Oxford Univ. Press,Oxford, U.K.
Wiltshire, E. P. 1945. Studies in the geography of Lepidoptera. III.The zoogeographical classification of West Palearctic species.Entomologist 78:113–136.
Yoder, A. D., M. Cartmill, M. Ruvolo, K. Smith, and V. Vilgalys.1996. Ancient single origin for Malagasy primates. Proc. Natl.Acad. Sci. 93:5122–5126.
Yoder, A. D., M. M. Burns, S. Zehr, T. Delefosse, G. Veron, S.M. Goodman, and J. F. Flynn. 2003. Single origin of MalagasyCarnivora from an African ancestor. Nature 421:734–737.
Zakharov, E. V., M. S. Caterino, and F. A. H. Sperling. 2004.Molecular phylogeny, historical biogeography, and divergencetime estimates for swallowtail butterflies of the genus Papilio(Lepidoptera, Papilionidae). Syst. Biol. 53:193–215.
Zeuner, F. E. 1943. Studies in the systematics of Troides Hubner(Lepidoptera, Papilionidae) and its allies: distribution and phy-logeny in relation to the geological history of the Australianarchipelago. Trans. Zool. Soc. Lond. 25:107–184.
Corresponding Editor: A. Yoder
2782 EVGUENI V. ZAKHAROV ET AL.
APPENDIX 1
Morphological characters and states used in the cladistic analysis of five species of the Papilio demoleus species group and eleven outgroupspecies of Papilio. Characters are illustrated in Figure 1 and will be fully discussed in C. R. Smith and R. I. Vane-Wright (unpubl. ms.).
Wing patternForewing upperside
1. Discal cell with pale scales: absent 5 0; present 5 1.2. Discal cell scales (patterning): random 5 0; organized 5 1.3. Discal cell scale pattern: longitudinal stripes 5 0; transverse bands 5 1.4. Discal cell with an array of three distal marks: absent 5 0; present 5 1.5. Cell R2 with postdiscal mark: absent 5 0; present 5 1.6. Cell R4 with discal mark: absent 5 0; present 5 1.7. Cell R4 postdiscal mark: absent 5 0; present 5 1.8. Cell M1 with postdiscal mark (in the male): absent 5 0; present 5 1.9. Cell 1A with both a postdiscal and a submarginal mark: absent 5 0; present 5 1.
Forewing underside10. Cell R2 with apical mark: absent 5 0; present 5 1.11. Cell M1 with postdiscal mark: absent 5 0; present 5 1.
Hindwing upperside12. Cell R1 with an eyespot: absent 5 0; present 5 1.13. Cell CuA2 with an eyespot: absent 5 0; present 5 1.14. Cell CuA2 with eyespot with a transverse distal border: absent 5 0; present 5 1.15. Cell M1 with postdiscal spot: absent 5 0; present 5 1.16. Cell M2 with postdiscal spot: absent 5 0; present 5 1.17. Cell M3 postdiscal spot: absent 5 0; present 5 1.
Hindwing underside18. Cells R1 and discal cell with pale basal band: absent 5 0; present 5 1.19. Basal band in cells R1 and discal cell: narrow 5 0; broad 5 1.20. Cells R5 to CuA1 with a pattern of blue and orange bands proximal to the submarginal marks: absent 5 0; present 5 1.21. Cell R5 with discal mark: absent 5 0; present 5 1.22. Cell R5 with discal mark reaching root of M1: absent 5 0; present 5 1.23. Cell M2 with marginal mark subdivided into two large marks: absent 5 0; present 5 1.24. Cell M3 with marginal mark subdivided into two large marks: absent 5 0; present 5 1.25. Hindwing tail (size): absent or rudimentary 5 0; short 5 1; medium 5 2; long 5 3.26. Hindwing tail with club: absent 5 0; present 5 1.27. Upperside surface of male forewing with androconial scales: absent 5 0; present 5 1.
Antennae28. Antennal color: unicolorous 5 0; with pale mark on one surface near tip 5 1.
Male genitalia29. Valve rim with postero-ventral expansion: absent 5 0; present 5 1.30. Valve rim with terminal notch: absent 5 0; present 5 1.31. Harpe with vertical projection: absent 5 0; present 5 1.32. Vertical projection of harpe: free 5 0; fused 5 1.33. Vertical projection of harpe: simple 5 0; double 5 1.34. Vertical projection of harpe: contiguous 5 0; detached 5 1.35. Harpe reaching or surpassing valve rim: absent 5 0; present 5 1.36. Harpe with ventral flange serrate: absent 5 0; present 5 1.37. Pseuduncus width: narrow 5 0; broad 5 1.38. Pseuduncus strongly declivous: absent 5 0; present 5 1.39. Uncus with prominent, terminal ridge or projection: absent 5 0; present 5 1.40. Uncus with prominent sub-terminal swelling: absent 5 0; present 5 1.41. Uncus with prominent serrations: absent 5 0; present 5 1.42. Saccus: absent 5 0; present 5 1.
Female genitalia43. Vestibulum with pocket: absent 5 0; present 5 1.44. External genitalic pocket: shallow 5 0; deep 5 1.45. Pocket with ventral lining of pocket: rugose 5 0; lanose 5 1.46. Ostium bursae position: anterior 5 0; central/posterior 5 1.47. Ostium bursae opening on sagittal ridge: absent 5 0; present 5 1.48. Peripheral vestibular plates: absent 5 0; present 5 1.49. Peripheral vestibular plates extending around ostium: absent 5 0; present 5 1.50. Peripheral vestibular plates joining anteriorly: absent 5 0; present 5 1.51. Peripheral vestibular plates with posterior expansion: absent 5 0; present 5 1.52. Peripheral vestibular plates with processes: absent 5 0; present 5 1.53. Lateral ostial plates: absent 5 0; present 5 1.54. Lateral ostial plates extended posteriorly: absent 5 0; present 5 1.55. Lateral ostial plates with lateral ridges: absent 5 0; present 5 1.56. Lateral ostial plates with peripheral lamella: absent 5 0; present 5 1.57. Lateral ostial plates with joining anteriorly: absent 5 0; present 5 1.58. Tonguelike process: absent 5 0; present 5 1.59. Tonguelike process with posterior transverse ridge: absent 5 0; present 5 1.60. Posterior central cup: absent 5 0; present 5 1.