Phylogeographic patterns of Hawaiian Megalagrion damselflies (Odonata: Coenagrionidae) correlate with Pleistocene island boundaries.

Abstract

The Pleistocene geological history of the Hawaiian Islands is becoming well understood. Numerous predictions about the influence of this history on the genetic diversity of Hawaiian organisms have been made, including the idea that changing sea levels would lead to the genetic differentiation of populations isolated on individual volcanoes during high sea stands. Here, we analyse DNA sequence data from two closely related, endemic Hawaiian damselfly species in order to test these predictions, and generate novel insights into the effects of Pleistocene glaciation and climate change on island organisms. Megalagrion xanthomelas and Megalagrion pacificum are currently restricted to five islands, including three islands of the Maui Nui super-island complex (Molokai, Lanai, and Maui) that were connected during periods of Pleistocene glaciation, and Hawaii island, which has never been subdivided. Maui Nui and Hawaii are effectively a controlled, natural experiment on the genetic effects of Pleistocene sea level change. We confirm well-defined morphological species boundaries using data from the nuclear EF-1alpha gene and show that the species are reciprocally monophyletic. We perform phylogeographic analyses of 663 base pairs (bp) of cytochrome oxidase subunit II (COII) gene sequence data from 157 individuals representing 25 populations. Our results point to the importance of Pleistocene land bridges and historical island habitat availability in maintaining inter-island gene flow. We also propose that repeated bottlenecks on Maui Nui caused by sea level change and restricted habitat availability are likely responsible for low genetic diversity there. An island analogue to northern genetic purity and southern diversity is proposed, whereby islands with little suitable habitat exhibit genetic purity while islands with more exhibit genetic diversity.

Full text

Page 1

Molecular Ecology (2005)
14
, 3457–3470doi: 10.1111/j.1365-294X.2005.02669.x
© 2005 Blackwell Publishing Ltd
Blackwell Publishing, Ltd.
Phylogeographic patterns of Hawaiian
damselflies (Odonata: Coenagrionidae) correlate with
Pleistocene island boundaries
Megalagrion

STEVE JORDAN,
*
Department of Ecology and Evolutionary Biology, Box U-3043, University of Connecticut, Storrs, CT 06269-3043,
Station, Pacific Island Ecosystems Research Center, US Geological Survey, PO Box 44, Hawaii National Park, HI 96718,
Biological Survey, Bishop Museum, 1525 Bernice Street, Honolulu, HI 96817
* CHRIS SIMON,* DAVID FOOTE† and RONALD A. ENGLUND‡
†
Kilauea Field
‡
Hawaii
Abstract
The Pleistocene geological history of the Hawaiian Islands is becoming well understood.
Numerous predictions about the influence of this history on the genetic diversity of Hawaiian
organisms have been made, including the idea that changing sea levels would lead to the
genetic differentiation of populations isolated on individual volcanoes during high sea
stands. Here, we analyse DNA sequence data from two closely related, endemic Hawaiian
damselfly species in order to test these predictions, and generate novel insights into the
effects of Pleistocene glaciation and climate change on island organisms.
xanthomelas
and
Megalagrion pacificum
are currently restricted to five islands, including
three islands of the Maui Nui super-island complex (Molokai, Lanai, and Maui) that were
connected during periods of Pleistocene glaciation, and Hawaii island, which has never
been subdivided. Maui Nui and Hawaii are effectively a controlled, natural experiment on
the genetic effects of Pleistocene sea level change. We confirm well-defined morphological
species boundaries using data from the nuclear EF-1α α α α
reciprocally monophyletic. We perform phylogeographic analyses of 663 base pairs (bp) of
cytochrome oxidase subunit II (COII) gene sequence data from 157 individuals representing
25 populations. Our results point to the importance of Pleistocene land bridges and historical
island habitat availability in maintaining inter-island gene flow. We also propose that
repeated bottlenecks on Maui Nui caused by sea level change and restricted habitat
availability are likely responsible for low genetic diversity there. An island analogue to
northern genetic purity and southern diversity is proposed, whereby islands with little
suitable habitat exhibit genetic purity while islands with more exhibit genetic diversity.
Megalagrion
gene and show that the species are
Keywords
: phylogeography, Pleistocene, sea level change
Received 27 February 2005; revision accepted 9 June 2005
Introduction
Enormous advances have been made in deciphering the
evolutionary and demographic processes that shaped
continental biotas during the Pleistocene (e.g. Taberlet
1998; Hewitt 2000; Good & Sullivan 2001; Johnson 2002).
In contrast, the effects of Pleistocene climate change on
et al
.
island biogeography have received much less attention
(but see Chown 1990; Ruedi
et al
Arensburger
et al
. 2004). This may be due to the lack of
direct glacial effects on many of the world’s biotically rich
islands. This subject merits more attention, however,
because glacially mediated sea level variation during the
Pleistocene was on the order of 120 m (Matthews 1990) and
declines in sea level dramatically altered the size and shape
of islands. In the Indonesian archipelago, for example, over
1.53 million km
of new land was available to terrestrial
organisms at maximum glaciation [
and details of the topography and river systems in this
. 1998; Buckley
et al
. 2001;
2
∼
17 000 years ago (ka)]
Correspondence:
Biology, Bucknell University, Lewisburg, PA 17837. Fax: +1 (570)
577–3537; E-mail: sdjordan@bucknell.edu
S
teve
J
ordan, Present address: Department of

Page 2

3458
S. JORDAN
ET AL.
© 2005 Blackwell Publishing Ltd,
Molecular Ecology
, 14, 3457–3470
now-flooded land are known (Voris 2000). Often, changes
in island size and shape have had biological consequences
(see Davies & Bermingham 2002 for an example involving
Caribbean butterflies). These consequences are amplified
by several properties of islands: their degree of isolation,
their relatively small size compared to continents, and
their preponderance of endemic species.
Episodes of warming and cooling during the Pleistocene
occurred at regular intervals (Hewitt 2000) and led to peri-
odic and at times rapid fluctuations in global sea level and
thus island sizes. The genetic consequences of such changes
on islands should differ from those proposed for northern
continental systems (Hewitt 2000), where a pattern of
‘southern (genetic) richness to northern purity’ has been
proposed (Hewitt 1996). Glaciation reduced available
habitat for most studied northern continental taxa. When
the glaciers disappeared, vast areas of open habitat became
available (Hewitt 2001; Taberlet & Cheddadi 2002), although
exceptions may exist (e.g. species that inhabit mountain
tops or sky islands, see Masta 2000; Knowles 2001). On oce-
anic islands the situation may have been quite different,
with glaciation leading to lower sea levels and sometimes
dramatic
increases
in total island area. In interglacial periods,
as Northern Hemisphere continental organisms were
increasing their ranges northward from refugia, many
island taxa were losing available habitat and becoming
more isolated on higher ground.
The islands of Hawaii present an excellent system for
studying the effects of Pleistocene climate change on trop-
ical oceanic islands, in large part because their geomorpho-
logical history is becoming known in some detail (Carson
& Clague 1995; Clague 1996; Hotchkiss
Clague 2002). The Hawaiian Islands are the product of a
mantle plume hot spot that has existed for at least 80 million
years (Myr). This hot spot has remained stationary as the
Pacific tectonic plate has drifted to the north and northwest
above it (Carson & Clague 1995), and lava has broken
through Earth’s crust regularly to form islands. The
Hawaiian Islands are thus arranged linearly, and increase
in age as one moves to the northwest (Fig. 1). Kauai is
the oldest of the high Hawaiian Islands, at 5.1 Myr,
while Hawaii is the youngest, at roughly 0.5 Myr. As
et al
. 2000; Price &
Fig. 1 Map of the high Hawaiian Islands, showing current and historical collection sites for Megalagrion xanthomelas and Megalagrion
pacificum, and the current extent of the submarine Penguin Bank. Island ages based on K-Ar dating are shown (Clague & Dalrymple
1987). In periods of low sea stands, the islands of Molokai, Maui, Lanai, and Kahoolawe were connected into a super-island referred to as
Maui Nui. Approximate boundaries of this island as it existed 20 ka are shown with a dashed line (after Price & Elliott-Fisk 2004). Note
that at this time Kahoolawe stood alone.

Page 3

PHYLOGEOGRAPHY OF HAWAIIAN DAMSELFLIES
3459
© 2005 Blackwell Publishing Ltd,
Molecular Ecology
, 14, 3457–3470
islands move further from the hot spot, subsidence and
erosion decrease their height until they become atolls
and ultimately underwater sea mounts. The depths of
sea channels between Hawaiian Islands are variable,
and some islands that are currently separate have been
connected in the past.
In particular, the islands of Maui Nui (the ensemble of
Molokai, Lanai, Maui, and Kahoolawe) rise from a large,
shallow platform. These islands have been connected
during more than 75% of the past 1.2 Myr and, excluding
Kahoolawe, as recently as 15–20 ka (Price & Elliott-Fisk
2004; see Fig. 1). Molokai and Oahu were connected much
earlier, before approximately 2 million years ago (Ma), via
a high connection through the current Penguin Bank (Fig. 1)
and a broad plain northwest of Molokai. Together these
islands formed the ‘Oahu Nui’ super-island (Price & Elliott-
Fisk 2004). Several authors have emphasized that under-
standing Pleistocene island connections such as these is
fundamental to our understanding of the evolution of the
Hawaiian biota (Funk & Wagner 1995; Craddock 2000). The
island of Hawaii is currently separated from Maui by the
1890-m-deep Alenuihaha Channel and has never been
physically connected to any other Hawaiian high island,
even during low sea stands (Carson & Clague 1995).
Organisms on Maui Nui and Hawaii have effectively
undergone a controlled natural experiment on the genetic
effects of Pleistocene sea level change. At glacial maxima 20 ka,
Maui Nui was roughly 5900 km
current size (Price & Elliott-Fisk 2004). During previous low
sea stands, Maui Nui was even larger, and at its maximum
extent 1.2 Ma, its
∼
14 000 km
of the Big Island of Hawaii (10 458 km
Maui Nui included expansive tracts of lowland habitat (Price
2004), extensive lowland rivers (whose courses can be seen
in sonar-generated seafloor maps), and large coastal pools
(J. Price, personal communication). At glacial minima, such
as the current high sea stand, Maui Nui is fragmented into
four unconnected islands, while Hawaii remains contiguous.
Current patterns of genetic diversity on these islands are
the product of at least three possible evolutionary scenarios.
Under scenario 1, ocean channels between islands represented
significant barriers to dispersal, and the land connections
that existed between islands during periods of low sea
stands did not facilitate gene flow. This idea has received
much support in the literature by authors who have pro-
posed that the formation of sea channels between the
individual sub-islands of the Maui Nui complex may
have led to greater isolation of single volcano populations
(Funk & Wagner 1995; Roderick & Gillespie 1998; Craddock
2000). Under scenario 2, ocean channels between islands
represented significant barriers to dispersal, but organisms
freely colonized and moved across land connections that
existed during low sea stands (Piano
scenario 3, ocean channels were not significant barriers to
2
in size, or nearly twice its
2
exceeded the current size
2
). The contiguous
et al
. 1997). Under
dispersal and organisms moved freely between all land
areas.
Predicted genetic patterns allow us to test these scenarios.
Under scenario 1, we would expect each volcano to harbour
populations with significant divergence due to interrupted
gene flow. Under scenario 2, we would expect to see little
to no genetic differentiation within land areas that are
contiguous at low sea stands: Maui Nui and the island of
Hawaii. Under scenario 3, we would expect to see signs
of gene flow even between Maui Nui and Hawaii, in spite
of the permanent ocean barrier between them.
The endemic Hawaiian damselflies
and
Megalagrion pacificum
are widespread, freshwater insects
that are part of a spectacular radiation of 23 species that
occupy a wide variety of habitats (Jordan
species offer an excellent opportunity to explore the
phylogeographic impacts of Pleistocene climate change for
several reasons. First, these taxa are sister species that are
well-defined morphologically (Polhemus 1997). Second,
they are geographically widespread, with current popu-
lations spanning five Hawaiian islands. Their distribution
includes the islands of Maui Nui and Hawaii, making them
ideal for testing the scenarios outlined above. Intensive
surveying and sampling in the 1990s delineated the
distribution of these species and generated a valuable
collection of alcohol-preserved specimens (Polhemus 1996).
Third, although they can exist in sympatry, these species
prefer slightly different habitats, and so may react differently
to sea level change.
Megalagrion xanthomelas
abundant at sea level and can develop large populations in
spring-fed and even brackish coastal wetlands.
pacificum
is found in shady side channels of streams at low
and intermediate altitudes (Polhemus & Asquith 1996).
Empirical studies examining each of the three scenarios
outlined earlier are needed, and will complement the large
number of evolutionary studies correlating Hawaiian biotic
patterns with plate tectonics, volcanism, and ecology (see
Wagner & Funk 1995 and references within). In this study,
we apply phylogenetic and phylogeographic methods to these
two Hawaiian damselfly species to examine the effects of
Pleistocene climate change on them. We examine the geo-
graphical distribution of mitochondrial genetic variability
across the five islands that make up the current range of these
species to test the three possible scenarios outlined above.
Finally, we explore the implications of our results for the
Pleistocene phylogeography of island organisms in general.
Megalagrion xanthomelas
et al
. 2003). These
is generally most
Megalagrion
Materials and methods
Sampling and specimens examined
Megalagrion xanthomelas
shore of Molokai, common on Lanai, and relictual on Maui
and Oahu (Fig. 1). Although Perkins (1913) noted that
is abundant on Hawaii and the north

Page 4

3460
S. JORDAN
ET AL.
© 2005 Blackwell Publishing Ltd,
Molecular Ecology
, 14, 3457–3470
Table 1
population bearing each of 31 unique COII haplotypes are listed. For nested clade analysis, haplotypes were aggregated into the populations marked with a cross (+)
Populations of
Megalagrion xanthomelas
and
Megalagrion pacificum
sampled for this study. Both species occur at sites shown in bold. The number of individuals from each
Haplotypes:
M. xanthomelasM. pacificum
N
Oahu/Maui NuiIsland of Hawaii
Namexan pacabcdefghijklmnopqrstuvwabcdefgh
Oahu
+Tripler Army Medical Center 2020
North Molokai
+Kalaupapa Peninsula
+Waikolu Valley
+
Pelekunu Valley
patch near Keawenui Stream
+Wailau
46
3
3
2116
2
3
1
and taro 145621
29116111
South Molokai
+Kaluaapuhi Fishpond,
Palaau Wetland22
Lanai
+Pipeline seepage, Maunalei Gulch
+Koele Lodge Ponds
6
4
4
2
2
2
Maui
+Cape Hanamanioa Region
+Haleakala NP, Kipahulu Valley
+Haipua’ena Stream
+Ukumehame Valley, W. Maui
413
1
4
1
31
11
+Hawaii, North Hilo and Hamakua
Maili Stream and small tributary
Waikaumalo Co. Park
Onomea Stream
412
3
1
111
12
2
31
1
311
+Hawaii, South Hilo
Keaukaha
Leleiwi
9
4
3
2
3111
11
+Hawaii, Kau District
Ninole Springs
Kawa Springs
52
7
21
6 1511
Hawaii, Kona District
+Kananone Water Hole
+Kaloko
64
6
11
16112411
Totals130 27 36145312712 11531111111111111 174111111

Page 5

PHYLOGEOGRAPHY OF HAWAIIAN DAMSELFLIES
3461
© 2005 Blackwell Publishing Ltd,
Molecular Ecology
, 14, 3457–3470
M. xanthomelas
is now represented on Oahu by a small population found
in a hundred metres of stream at the Tripler Army Medical
Center (TAMC). A single specimen was collected in 1947
on Niihau, but the species has never been documented on
Kauai.
Megalagrion pacificum
is most abundant on Molokai,
where it often occurs near
M. xanthomelas
Maui and a single individual was recently collected from
Hawaii (Englund 1999), although no significant populations
have been found there.
Megalagrion pacificum
on Kauai, Oahu, and Lanai, but is now extirpated from
these islands (Polhemus & Asquith 1996; Polhemus 1997).
We examined a total of 130
M. pacificum
individuals (Table 1). The disparity in these
numbers reflects the relative abundance of
compared to the more localized
tions of these species were from 25 populations and five
islands, effectively spanning their current known range,
including rare
M. xanthomelas
populations on Maui and Oahu. Efforts were made to ana-
lyse 20 individuals from each population or geographically
close group of populations. However, in many cases this
was not feasible due to variation in population sizes and
the logistics of visiting many different localities. Wings,
abdomens, heads, and terminalia of each specimen have
been preserved in individual vials. These voucher specimens
will be deposited at the Bernice P. Bishop Museum in Honolulu,
and the Kilauea Field Station, Pacific Island Ecosystems
Research Center, US Geological Survey, Hawaii Volcano
National Park. Extracted DNA will remain in the care of S.J.
was once common in Honolulu gardens, it
. It persists on
was once found
M. xanthomelas
and 27
M. xanthomelas
. Our collec-
M. pacificum
individuals from remnant
DNA extraction and sequencing
DNA was extracted from damselfly thoracic muscles
using either a standard phenol–chloroform method
(Sambrook
et al
. 1989) or a QIAGEN DNeasy Tissue Kit.
Mitochondrial DNA (mtDNA) and nuclear primer sequ-
ences and conditions for polymerase chain reaction (PCR)
and cycle sequencing are reported elsewhere (Jordan
2003). We generated 663 bp of DNA sequence data from
the mitochondrial COII gene, which we expected to show
informative variation within species (Simon
using two primers: C2-J-3102 and TK-N-3773. In order to
explore the possibility of introgressive hybridization between
these species, we augmented a pre-existing nuclear DNA
data set by sequencing approximately 1000 bp of the EF-1
gene from two individuals using two primer pairs: EF1-F-
2361 and EF1-R-2765 for the 5
EF1-R-3093 for the 3
′
end (Jordan
autapomorphies and nonsynonymous mutations were
verified by rechecking the original chromatograms.
In order to identify possible contamination, extra-
ordinary results were double checked by gathering the data
again. The Hawaii island
M. pacificum
et al
.
et al
. 1994)
α
′
end and EF1-F-2652 and
et al
. 2003). All sequence
mtDNA was extracted,
amplified, and sequenced twice, in separate months. The
mtDNA from two
M. xanthomelas
Hawaii island, that were found to bear haplotypes from
another clade was amplified and sequenced twice, again in
different months.
individuals from Kaloko,
mtDNA sequence analysis
DNA sequences were aligned by eye and
version 3.0 (GeneCodes). Basic sequence attributes, including
raw (
p
) haplotype divergences, codon site-specific mutation
rates, base composition bias, and amino acid substitutions
were calculated using
mega
2 (Kumar
version 4.0b6–10 (Swofford 1999).
We used three separate classes of data analysis on the
mtDNA data, explained in detail below, (i) phylogenetic
analysis, (ii)
tcs
(Templeton, Crandall, Sing) parsimony and
nested clade analysis (NCA: Templeton
1998) and (iii) analysis of molecular variance (
Excoffier
et al
. 1992).
sequencher
et al
. 2001) and
paup
*
et al
. 1992; Templeton
amova
:
Phylogenetic analysis
Maximum parsimony (MP) and maximum likelihood (ML)
were implemented using
paup
M
.
xanthomelas
and
M
.
pacificum
distances from all other
Megalagrion
data exploration suggested that these distant outgroups
could lead to poorly resolved trees. We therefore carried
out all MP, ML, and Bayesian analyses using no outgroups,
and rooted the resulting trees a posteriori (see below).
We performed heuristic, equally weighted MP searches
using 100 random-addition-sequence replicates and TBR
(tree-bisection–reconnection) branch swapping. MP boot-
straps were performed using a full heuristic search
(10 random-addition-sequence replicates, TBR branch
swapping) and 1000 pseudoreplicates.
Because appropriate ML models are important in molecular
systematics (Sullivan & Swofford 1997), we selected the
best model for our data using the method of Frati
We considered 16 different likelihood models and used a
likelihood-ratio test to select the simplest model that was not
significantly different from the best fitting model (GTR + I +
This simplest model (HKY + I) was used in all ML analyses
in order to minimize computing time and reduce the variance
of the estimated parameters. Parameter values for this model
were estimated on one of the MP trees, and fixed for the first
ML heuristic search (10 random-addition-sequence replicates,
TBR branch swapping, all best trees retained). After a full
heuristic ML search was completed, we re-estimated model
parameters on the resulting tree and used those parameters
to search again, repeating this process until the new search
recovered the same topology as the previous search. In
ML bootstrap searching, model parameter values were
* 4.0b6–10 (Swofford 1999).
display high genetic
species and initial
et al
. (1997).
Γ
).