The genetic basis of a plant–insect coevolutionary
Christopher W. Wheat*†‡, Heiko Vogel*, Ute Wittstock*§, Michael F. Braby¶?, Dessie Underwood**,
and Thomas Mitchell-Olds*††
*Max Planck Institute for Chemical Ecology, Beutenberg Campus, Hans Knoell Strasse 8, 07745 Jena, Germany;¶School of Botany and Zoology, Australian
National University, Canberra ACT 0200, Australia; **California State University, 1250 Bellflower Boulevard, Long Beach, CA 90840; and††Department of
Biology, Duke University, Durham, NC 27708
Edited by May R. Berenbaum, University of Illinois at Urbana–Champaign, Urbana, IL, and approved October 23, 2007 (received for review July 5, 2007)
Ehrlich and Raven formally introduced the concept of stepwise co-
evolution using butterfly and angiosperm interactions in an attempt
to account for the impressive biological diversity of these groups.
However, many biologists currently envision butterflies evolving 50
to 30 million years (Myr) after the major angiosperm radiation and
thus reject coevolutionary origins of butterfly biodiversity. The un-
resolved central tenet of Ehrlich and Raven’s theory is that evolution
of plant chemical defenses is followed closely by biochemical adap-
tation in insect herbivores, and that newly evolved detoxification
mechanisms result in adaptive radiation of herbivore lineages. Using
one of their original butterfly-host plant systems, the Pieridae, we
identify a pierid glucosinolate detoxification mechanism, nitrile-spec-
ifier protein (NSP), as a key innovation. Larval NSP activity matches
the distribution of glucosinolate in their host plants. Moreover, by
using five different temporal estimates, NSP seems to have evolved
shortly after the evolution of the host plant group (Brassicales) (?10
Myr). An adaptive radiation of these glucosinolate-feeding Pierinae
followed, resulting in significantly elevated species numbers com-
pared with related clades. Mechanistic understanding in its proper
historical context documents more ancient and dynamic plant–insect
insights provide the tools for detailed molecular studies of coevolu-
tion from both the plant and insect perspectives.
adaptive radiation ? Brassicales ? Pieridae ? diversification ?
Bayesian relaxed molecular clock
major angiosperm radiation occurred ?140 to 100 million years
ago (Mya), fossil data suggest that diversification of ‘‘primitive’’
Lepidoptera occurred before this time and butterflies radiated
long after these host plants (?75 Mya) (1–3). Many espouse this
recent butterfly origin, which necessarily implies a very limited
role, if any, for coevolution in butterfly diversification (1–4).
However, others posit a much older age of butterflies (?100
Mya), with speciation influenced by angiosperm evolution and
the breakup of the supercontinent Gondwana (5, 6). This lack of
consensus on both the timing of butterfly diversification, which
resulted in the ?17,000 extant species today, and the role of
coevolution arises from the notably poor fossil record of Lepi-
scenarios, no studies have tested for the effects of key innova-
tions on butterfly diversification even though coevolution
through key innovations was first introduced to science using
butterflies and their angiosperm host plants as exemplars (8–
12). To explore the potential role of coevolution in butterfly
diversification, we focus on the family Pieridae, composed of the
advances in functional genomics and phylogenetics in this family
provide a unique opportunity to resolve the controversies briefly
reviewed above (see also ref. 5).
Pieridae use three major host plant groups: the Fabales
(Legumes), the Brassicales (glucosinolate-containing plants ex-
he relative timing of adaptive radiations in host plants and
their butterfly herbivores is controversial. Although the
emplified by the cabbages and Arabidopsis), and mistletoes.
Phylogenetic reconstruction of almost 90% of the Pieridae
genera (74 recognized genera plus six subgenera, based on 1,066
indicate that Fabales feeding is the ancestral state of Pieridae
(Fig. 1). The Fabales feeders are the Dismorphiinae and nearly
all Coliadinae, whereas the sister to the Coliadinae, the Pierinae,
primarily feed on Brassicales (Fig. 1) (12). Within Pierinae, there
are two subsequent derived shifts away from glucosinolate
feeding onto mistletoes and other species. Thus, the Pierinae
represent a single origin of glucosinolate feeding (Fig. 1).
The evolutionary appearance of the plant order Brassicales
(Eurosid II, Dicotyledons) presented a radical new chemical chal-
lenge for insect herbivores, known as the glucosinolate-myrosinase
and most widely studied chemical plant defenses (15–17). Its
effectiveness as an anti-herbivore defense becomes apparent upon
tissue damage, such as insect feeding. Tissue damage brings the
formerly compartmentalized myrosinase enzyme into contact with
nontoxic glucosinolates, which it hydrolyzes into breakdown prod-
ucts such as isothiocyanates (18–20). Whereas Homo sapiens may
find these breakdown products enjoyable condiments (e.g., mus-
herbivores (15, 21–23).
We have identified two independent lepidopteran detoxifica-
tion mechanisms for the glucosinolate-myrosinase defense sys-
tem at biochemical and molecular levels by means of functional
genomics approaches, beginning with glucosinolate sulfatase
(GSS) in the diamondback moth Plutella xylostella (Plutellidae)
(24). GSS desulfates glucosinolates, producing metabolites that
no longer act as substrates for myrosinases. The second, called
nitrile-specifier protein (NSP), has recently been identified for
the pierid butterfly Pieris rapae (20). NSP, expressed solely in the
larval midgut, promotes the formation of nitrile breakdown
products instead of toxic isothiocyanates upon myrosinase-
catalyzed glucosinolate hydrolysis (20). The GSS and NSP
detoxification mechanisms are distinctly different from each
other, as well as the other identified host plant detoxification
Author contributions: C.W.W. and T.M.-O. designed research; C.W.W., H.V., and U.W.
performed research; U.W., M.F.B., D.U., and T.M.-O. contributed new reagents/analytic
tools; C.W.W., H.V., and U.W. analyzed data; and C.W.W., H.V., and U.W. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
†Present Address: Pennsylvania State University, Department of Biology, 208 Mueller
Laboratories, University Park, PA 16802.
‡To whom correspondence should be addressed. E-mail: firstname.lastname@example.org.
§Present Address: Institute of Pharmaceutical Biology, Braunschweig University of Tech-
nology, 38106 Braunschweig, Germany.
?Present Address: Biodiversity Conservation Division, Department of Natural Resources,
Environment and the Arts, P.O. Box 496, Palmerston NT 0831, Australia.
This article contains supporting information online at www.pnas.org/cgi/content/full/
© 2007 by The National Academy of Sciences of the USA
December 18, 2007 ?
vol. 104 ?
no. 51 ?
mechanisms of the Lepidoptera: the cytochrome P450 monoox-
ygenase gene superfamily in Papilio butterflies used against
furanocoumarins (25) and the flavin-dependent monooxygenase
system of the arctiid moth Tyria jacobaeae used against pyrroli-
zidine alkaloids (26).
With these issues in focus, we document a key innovation and
its role in butterfly diversification by (i) assessing the evolution-
ary origin of a host plant detoxification mechanism involved in
an ancient host plant shift, (ii) generating a robust estimate of
when this detoxification mechanism evolved relative to the age
n i r e i P
of 85 Mya, shown on the x axis. Yellow branches refer to terminal taxa that are predominately Fabales feeding, whereas green branches are glucosinolate
feeding, and blue are derived non-glucosinolate feeding (gray branches indicate unknown host plant associations). Pseudopontiinae and Dismorphiinae, sister
clades to Coliadinae plus Pierinae, are not shown but can be seen in SI Figs. 4 and 5. Higher taxon groupings are shaded for visualization. Butterfly images are
of subfamily exemplar taxa. Red bar indicates the lower 95% confidence interval from an EF-1? fossil calibrated, Bayesian relaxed molecular clock estimation
of the Pierinae–Coliadinae divergence. Butterfly photos: M. Borsch (A. crataegi), S. Coombes (A. cardamines, C. hyale), M. Rowlings (C. evagore), H. Vogel (P.
www.pnas.org?cgi?doi?10.1073?pnas.0706229104Wheat et al.
of the colonized host plant family, and (iii) testing for an effect
of the host shift on butterfly diversification rate.
We begin by asking a mechanistic question, focusing on whether
all Pierid glucosinolate feeders use NSP, not NSP, or the other
identified glucosinolate detoxification mechanism of the Lepi-
doptera, GSS. Midgut assays of larvae from 13 Pieridae species
from North America, Europe, Africa, and Australia, represent-
ing each of the major subfamilies or groups, reveals complete
concordance between NSP activity and glucosinolate feeding
(Fig. 1; and see supporting information (SI) Fig. 3 and SI Table
1). Consistent with initial studies of these mechanisms on Pieris
rapae, none of the butterfly species assayed displays any detect-
able GSS activity, and P. xylostella and other moths have no NSP
activity. Within the Pieridae, glucosinolate feeding exactly
matches the presence of NSP activity in larvae, suggesting a
single evolutionary origin with a subsequent loss after secondary
host plant shifts to Santalales (Fig. 1, SI Fig. 3, and SI Table 1).
These results support NSP as a key innovation within the
Pieridae enabling glucosinolate detoxification (27).
Fossil and molecular data agree that the Brassicales appeared
by 90 to 85 Mya, which is much earlier than the parallel evolution
of glucosinolates in the Euphorbiaceae (58 Mya) (28). Pierinae
fossils with modern relatives in the Brassicales feeding clade
appear ?34 Mya, ?50 Myr after the first fossil Brassicales (13,
28). Fossils inherently provide only minimum temporal esti-
mates. Thus, the fossil record is compatible with either a long
delay of insect adaptation after the Brassicales radiation, a
contemporaneous radiation of Brassicales and their Pierid her-
bivores, or anything in between. Among the major insect orders,
Lepidoptera have the poorest fossil record, probably as a result
of their wing structure (8, 75). In stark comparison, the fossil
record of beetles is very rich, and beetles provide some of the
best evidence for coevolution (29), although this evidence has
recently been questioned by a recalibrated phylogeny of the leaf
beetles (30). In this article, we directly address the timing of the
appearance of the glucosinolate-feeding Pierinae, using several
independent molecular datasets and various calibration methods
to generate a robust estimate of when Pieridae evolved relative
to their Brassicales host plants.
Our temporal analysis used two standard molecular phyloge-
netic genes, cytochrome oxidase I (COI) and elongation factor
1? (EF-1?), with two and three different temporal estimations,
respectively (Fig. 2). First, we compared COI sequence from
Colias eurytheme (Coliadinae) and Pieris napi (Pierinae) and
estimated the time of their divergence at 89.12 and 96.32 Mya.
These two estimates are respectively derived (i) from an Insecta-
wide average substitution rate of 0.022% per million years for
COI 2nd position data calibrated on ancient, nonlepidopteran
fossils (31), and (ii) from COI 1st and 2nd position data
calibrated on butterfly fossils from within the Nymphalinae
subclade of Nymphalidae (32). Second, we compared the EF-1?
sequence distance between our Coliadinae and Pierinae repre-
sentatives to the mean EF-1? sequence distance between
temporally calibrated representatives in the Nymphalidae and
Papilionidae butterfly families (32, 33). Nymphalidae estimates
were calibrated as stated above, and the Papilionidae samples
were calibrated by using biogeographic vicariance events in-
ferred during the breakup of Gondwana. Molecular evolution
rates of the nuclear EF-1? gene are likely to be similar among
these families, as Nymphalidae and Papilionidae are Pieridae’s
nearest family clades (34). Together these EF-1? distance com-
parisons suggest that Coliadinae diverged from Pierinae ? ?62
and ?82 Mya, respectively (Fig. 2).
Pieridae-specific temporal reconstruction used a Bayesian
relaxed molecular clock method on EF-1? data from across
as minimum node dates (35, 36)(SI Figs. 4 and 5). Temporal
reconstruction can be affected by assumptions used in analyses,
(Mya) whereas the y axis gives the species numbers (Log scale) for each clade. The upper and lower 95% confidence limits are shown for the expected species
numbers of a clade diversifying at a rate equal to that of the sister clade to Pierinae plus Coliadinae (i.e., Dismorphiinae plus Pseudopontiinae) with either no
extinction (dashed lines) or a high extinction rate (90%, solid lines). Current diversity levels for Pierinae as a whole (triangle), the Pierinae glucosinolate feeders
only (square), and Coliadinae (diamond) are plotted with an estimated origin 85 Mya. The horizontal bar within these groupings represents the overlap among
the molecular divergence estimates shown below, using the fossil and molecular estimates of the age of the Brassicales as an upper bound. (b) COI and EF-1?
estimates and the lower 95% confidence limit of Bayesian relaxed molecular clock estimates. Temporal estimates from Brassicales fossils are shown at bottom.
Current and expected diversity, and divergence estimates for the Coliadinae and Pierinae subfamilies of Pieridae. (a) The x axis gives the age of clades
Wheat et al. PNAS ?
December 18, 2007 ?
vol. 104 ?
no. 51 ?
such as tree topology, node calibration, clock-like evolution, and
other prior estimates (37). In our analysis, we explored the
potential effects of alternative topologies, fossil node place-
ments, clock assumptions, and other prior estimates on temporal
inferences. These alternatives had little effects on results except
to increase age estimates (SI Table 2). The most temporally
conservative model predicts the appearance of Pieridae with a
95% confidence interval of 166 to 79 Mya, which agrees with
related clock-based estimates (13). In sum, our five temporal
estimates are consistent and provide a robust assessment. The
evolutionary time to the appearance of the Brassicales, most
likely within 10 million years (Figs. 1 and 2).
Ehrlich and Raven hypothesized, but were unable to test, that
the ‘‘main diversification of [Pieridae] occurred after it became
associated with the [Brassicales]’’ (8). Others have posited a
relationship between host plant use and species abundance, but
this relationship has not been tested (11, 12). We calculate and
compare the standing diversity of the sister clades Coliadinae
and Pierinae (38), which are a priori expected to have similar
levels of diversification because they are sister clades (39). There
are both more genera and species of Pierinae than Coliadinae
(57 vs. 18 genera, 830 vs. 220 species). To assess diversification
rate and extinction effects giving rise to these differences in
extant taxa, we use the diversification rate of Dismorphiinae plus
Pseudopontiinae (sister clade to Coliadinae plus Pierinae) as the
background rate for generating expected species diversity con-
fidence intervals through time. Coliadinae species diversity is at
or below the expected numbers of species. However, there are
significantly more species of Pierinae than expected, even if
non-glucosinolate feeders are excluded, suggesting that the
increased diversification of the Pierinae resulted from the host
plant shift onto Brassicales (Fig. 2).
The evolution of the NSP glucosinolate detoxification gene was
a novel, key innovation facilitating the host plant shift of
ancestral Pierinae from Fabales to Brassicales feeding with
macroevolutionary consequences. Our analysis of major clade
representatives across Pieridae demonstrates that only glucosi-
nolate-feeding Pierinae show NSP activity, indicating a single
mechanistic basis of glucosinolate feeding originating within the
Pierinae. Multiple independent temporal reconstructions pro-
vide a robust estimate placing the Pierinae host shift soon after
the appearance of the Brassicales plant order. This host shift also
resulted in significantly increased speciation compared with
related clades. Together, these results provide strong support for
the central tenet in Ehrlich and Raven’s coevolution theory, that
key biochemical innovations foster increased speciation (8, 40).
Ancestral Pieridae butterflies did not provide the initial
herbivory pressure for Brassicales chemical defense evolution,
nor are they solely exerting sufficient pressure for the continued
evolution of the glucosinolate-myrosinase system of the Brassi-
cales, as lepidopteran host plant coevolution is expected to be
diffuse (8, 40, 41). Yet, comparative functional genomic analysis
of the glucosinolate-myrosinase system in the family Brassi-
caceae reveals ongoing evolution in response to herbivory
pressure (42–44). Multiple enzymes, with allelic variation, are
involved in the biosynthesis of diverse glucosinolate structures
and relative concentrations (45–47). This resulting intra- and
interspecific glucosinolate profile variation results in herbivory
level variation (48–51). In addition, signatures of positive selec-
tion at the molecular level have been identified at these genes,
which are undergoing gene duplication and neofunctionaliza-
tion, documenting ongoing glucosinolate metabolic evolution
escape and radiation from herbivore pressure is continuing to
evolve in derived lineages with herbivory consequences.
Studies of plant–insect coevolution have evolved significantly
since the days of Ehrlich and Raven (e.g., refs. 53–64). Butterfly
support for the parallel coevolutionary process, i.e., cospecia-
tion, has proven difficult to find, as cospeciation-driven diver-
sification seems to be rare among insects and butterflies in
particular, with Ehrlich and Raven’s escape and radiate coevo-
lutionary mechanism likely more common (8, 10, 40, 53, 65, 66).
The only strong case for butterfly coevolution is found in the
genus Papilio (Papilionidae), where ?75% of this genus feeds on
plants containing furanocoumarin-based chemical defenses, syn-
thesized by plants using members of the cytochrome P450
monooxygenases gene superfamily (11, 25, 67). Papilio butterfly
species diversity increases with host plant furanocoumarin di-
versity (11). Interestingly, furanocoumarins both induce and are
detoxified by Papilio species using other members of the cyto-
chrome P450 superfamily (68). Cytochrome P450 participation
on both sides of the plant–insect interaction highlights the
diverse functional roles members of this gene superfamily play
in environmental response (68). In fact, the generalist Noctuid
moth Helicoverpa, which is ?100 Myr divergent from Papilio sp.,
independently uses other P450s to detoxify furanocoumarins,
highlighting the repeatability of this gene superfamily’s role in
detoxification. Cytochrome P450s are also part of the glucosi-
nolate biosynthetic pathway in Brassicales plants. Interestingly,
in Lepidoptera specialized on glucosinolates, only the novel
molecular mechanisms discussed in this article (GSS and NSP)
have been identified. Given the apparently rapid host shift of the
Pierinae onto Brassicales, we are investigating the evolutionary
history of the NSP gene to assess its ancestral, pre-glucosinolate
Over 40 years ago, the authors who introduced us to coevo-
lution using butterflies claimed that their ‘‘predictions cannot be
tested’’ (8). The work presented here documents a key coevo-
lutionary innovation in the Pieridae completely independent of
other known lepidopteran detoxification mechanisms. More-
over, this work indicates a much more dynamic interaction
between butterfly herbivores and their host plants than currently
envisioned. We present our work here in the hope that evolu-
tionary and ecological functional genomics will encourage in-
creased mechanistic study of plant–insect interactions.
Materials and Methods
NSP and GSS Assay. Last instar larvae actively feeding on preferred host plants
(SI Table 1) were used to test for NSP activity using methods described in ref.
20. Briefly, benzylglucosinolate and myrosinase were incubated with larval
midgut extracts. Dichloromethane extracts of these assay mixtures were then
analyzed by GC-MS with internal standard, with total ion current traces
identifying NSP activity (see SI Fig. 3 for representative traces and more
assays in this study to detect potentially low enzyme activity levels.
Phylogenetic Reconstruction. Bayesian phylogenetic reconstruction, using a
GTR ? I ? G nucleotide substitution model for a total of 1 million generations
as implemented in Mr. Bayes v.3.1.2 (69, 70), was performed by using two
representatives Pseudopontia or Dismorphiniae (SI Fig. 4). Tree 2 used all
available EF-1? Pieridae sequences with Pseudopontia paradoxa as an out-
group (SI Fig. 5). Trees 1 and 2 place Dismorphiniae sister to Coliadinae plus
their terminal branching patterns, whereas subfamily relationships were sta-
ble (see SI Materials and Methods for analysis details).
Ancestral Node Dating. COI. Recent analysis of Nymphalidae COI data, using
Vanessa, from distant Nymphalinae clades, for 1st and 2nd position rate
change per million year provides an estimate of 0.0004372 (26.8 changes per
943 bp; 26.8/943 ? 0.02841; 0.02841/65 ? 0.0004372 substitutions per million
www.pnas.org?cgi?doi?10.1073?pnas.0706229104Wheat et al.
years). There are 43 1st plus 2nd position changes between C. eurytheme and Download full-text
P. napi, giving a temporal estimate of the Coliadinae and Pierinae divergence
at 96.32 Mya (43/1021 ? 0.04211; 0.04211/.0004372 ? 96.32).
EF1?. The estimated EF-1? mean genetic distance and standard deviation
between Coliadinae (n ? 15) and Pierinae (n ? 60) taxa is 12.928% (SE ?
0.0095), whereas that between Papilio species (n ? 55) and Pachliopta nep-
tunus (Papilionidae) is nearly 3% lower at 10.029% (SE ? 0.1115). Genetic
distance was calculated by using a Tamura 3-Parameter distance model as
implemented in Mega 3.1, with standard error calculated with 500 bootstrap
shape parameter ? 1.4 [as per the estimated value for the dataset (33)]. The
two Papilionidae groups are calculated to have diverged 89.1 to 82.5 Mya,
suggesting a comparable or even older divergence between Coliadinae and
Pierinae. Similar EF-1? calibrated estimates are available from Nymphalidae.
there is a mean distance of 7.64% (SE ? 0.0078). This distance is 60% of the
divergence found between Coliadinae and Pierinae representatives, suggest-
ing a Coliadinae Pierinae divergence much greater than 65 Mya.
EF1? using a relaxed clock method. Temporal estimation using a Bayesian
relaxed molecular clock method was implemented by using MULTIDIVTIME
(35, 73) on the two Bayesian analysis EF-1?-generated trees (SI Figs. 4 and 5).
Prior estimates (SI Table 2) explored model response across many different
age set to the oldest known fossil Pierinae at 34 Mya. For all combinations of
prior values, the Markov Chain was then sampled every 100 cycles for a total
of 10,000 samples. Posterior distributions were approximated based on these
10,000 samples. Fossil node placement initially followed Braby et al. (13), but
here fossils are assigned as a minimum date estimate and the effect of more
conservative node placement was explored by using the next most ancestral
node (SI Figs. 4 and 5 and SI Table 2). For each run, across two trees, the prior
conditions, resulting estimates, and confidence limits are listed in SI Table 2.
Absolute Diversification Rate Calculations. There are both more genera and
numbers are significantly different from equal number expectations (?2?
10.8, P ? 0.001; ?2? 193.5, P ? 0.001). However, this simple calculation does
not assess diversification rate and extinction effects giving rise to these
differences in extant taxa. To address these effects, we perform a statistical
analysis of a stochastic birth-and-death processes, employing a method-of-
moments estimator following Magallon and Sanderson (38), which takes the
age of the clades in question, together with their extant species diversities,
and allows for them to be compared with their expected levels of diversifica-
for more details). Sister clades Coliadinae and Pieridinae are by definition of
equal age, and, for an unbiased, phylogenetically relevant estimator of their
expected species diversity, we chose the diversification rate of their sister
clade, Dismorphiinae plus Pseudopontiinae. By using this rate, 95% confi-
dence limits were calculated for expected numbers of species across different
clade ages with either no extinction or a high (90%) extinction rate (see SI
Materials and Methods for more details; SI Table 3). Groups with a greater
than expected number of species are considered to be ‘‘excessively species
than under no extinction due to having a higher variance; a high number of
speciation events are needed to offset the high number of extinctions to
maintain the same net diversification rate as when there is no extinction.
ACKNOWLEDGMENTS. We thank K. Bargum, M. Clauss, J. Marden, S. Nylin, R.
Schilder, N. Wahlberg, and W. Watt for comments on various stages of this
manuscript; R. Oyama and J. Thorne for analysis assistance; and A. Bonkewitz
and D. Heckel for specimen collection help. M. Borsch, S. Coombes, and M.
A. cardamines, and C. hyale and C. evagore, respectively). This work was
funded by the Max Planck Society and National Science Foundation Grant
IBN-0412651 (to C.W.W.).
1. Labandeira CC, Dilcher DL, Davis DR, Wagner DL (1994) Proc Natl Acad Sci USA
2. Vane-Wright D (2004) Nature 428:477–478.
3. Magallon SA, Sanderson MJ (2005) Evolution (Lawrence, Kans) 59:1653–1670.
4. de Jong R (2003) Invertebr Syst 17:143–156.
5. Braby MF, Trueman JWH, Eastwood R (2005) Invertebr Syst 19:113–143.
6. Miller JY, Miller LD (2001) in Biogeography of the the West Indies: Patterns and
Perspectives, eds Woods CA, Sergile FE (CRC Press, Boca Raton, FL), pp 127–150.
7. Labandeira CC, Sepkoski JJ (1993) Science 261:310–315.
8. Ehrlich PR, Raven PH (1964) Evolution (Lawrence, Kans) 18:586–608.
9. Wahlberg N, Brower AVZ, Nylin S (2005) Biol J Linn Soc 86:227–251.
10. Janz N, Nylin S (1998) Evolution (Lawrence, Kans) 52:486–502.
11. Berenbaum M (1983) Evolution (Lawrence, Kans) 37:163–179.
12. Braby MF, Trueman JWH (2006) J Evol Biol 19:1677–1690.
13. Braby MF, Vila R, Pierce NE (2006) Zool J Linn Soc 147:238–275.
14. Hall JC, Iltis HH, Sytsma KJ (2004) Syst Bot 29:654–669.
15. Chew F (1988) in Biologically Active Natural Products, ed Culter HG (Am Chem Soc,
Washington, DC, pp 155–181.
16. Mitchell-Olds T, Clauss MJ (2002) Curr Opin Plant Biol 5:74–79.
17. Wittstock U, Halkier BA (2002) Trends Plant Sci 7:263–270.
19. Andreasson E, Jorgensen LB, Hoglund AS, Rask L, Meijer J (2001) Plant Physiol
20. Wittstock U, Agerbirk N, Stauber EJ, Olsen CE, Hippler M, Mitchell-Olds T, Gershenson
J, Vogel H (2004) Proc Natl Acad Sci USA 101:4859–4864.
21. Li Q, Eigenbrode SD, Stringham GR, Thiagarajah MR (2000) J Chem Ecol 26:2401–2419.
22. Lichtenstein EP, Morgan DG, Mu ¨ller CH (1964) J Agric Food Chem 12:158–161.
Phytochemistry: From Ethnobotany to Molecular Ecology, Recent Advances in Phyto-
chemistry, eds Romeo JT (Elsevier, Amsterdam), Vol 37, pp 101–125.
24. Ratzka A, Vogel H, Kliebenstein DJ, Mitchell-Olds T, Kroymann J (2002) Proc Natl Acad
Sci USA 99:11223–11228.
25. Li WM, Schuler MA, Berenbaum MR (2003) Proc Natl Acad Sci USA 100:14593–14598.
26. Naumann C, Hartmann T, Ober D (2002) Proc Natl Acad Sci USA 99:6085–6090.
27. Berenbaum MR, Favret C, Schuler MA (1996) Am Nat 148:S139–S155.
28. Wikstrom N, Savolainen V, Chase MW (2001) Proc R Soc London Ser B 268:2211–2220.
29. Farrell BD (1998) Science 281:555–559.
30. Go ´mez-Zurita J, Hunt T, Kopliku F, Vogler AP (2007) PLoS ONE 1:e360.
31. Gaunt MW, Miles MA (2002) Mol Biol Evol 19:748–761.
32. Wahlberg N (2006) Syst Biol 55:703–714.
33. Zakharov EV, Caterino MS, Sperling FAH (2004) Syst Biol 53:193–215.
34. Wahlberg N, Braby MF, Brower AVZ, de Jong R, Lee MM, Nylin S, Pierce NE, Sperling
FAH, Vila R, Warren AD, Zakharov E (2005) Proc R Soc London Ser B 272:1577–1586.
35. Kishino H, Thorne JL, Bruno WJ (2001) Mol Biol Evol 18:352–361.
36. Thorne JL, Kishino H, Painter IS (1998) Mol Biol Evol 15:1647–1657.
37. Ho SYW, Phillips MJ, Drummond AJ, Cooper A (2005) Mol Biol Evol 22:1355–1363.
38. Magallon S, Sanderson MJ (2001) Evolution (Lawrence, Kans) 55:1762–1780.
39. Mitter C, Farrell B, Wiegmann BM (1988) Am Nat 132:107–128.
40. Futuyma DJ (2000) Plant Species Biol 15:1–9.
41. Hougen-Eitzman D, Rausher M (1994) Am Nat 143:677–697.
42. Kliebenstein DJ, Kroymann J, Mitchell-Olds T (2005) Curr Opin Plant Biol 8:264–271.
43. Raybould AF, Moyes CL (2001) Heredity 87:383–391.
44. Renwick JAA (2002) Entomol Exp Appl 104:35–42.
45. de Quiros HC, Magrath R, McCallum D, Kroymann J, Scnabelrauch D, Mitchell-Olds T,
Mithen R (2000) Theor Appl Genet 101:429–437.
47. Kroymann J, Textor S, Tokuhisa JG, Falk KL, Bartram S, Gershenzon J, Mitchell-Olds T
(2001) Plant Physiol 127:1077–1088.
48. Heidel AJ, Clauss MJ, Kroymann J, Savolainen O, Mitchell-Olds T (2006) Genetics
49. Windsor AJ, Reichelt M, Figuth A, Svatos A, Kroymann J, Kliebenstein DJ, Gershenzon
J, Mitchell-Olds T (2005) Phytochemistry (Amsterdam) 66:1321–1333.
50. Kroymann J, Donnerhacke S, Schnabelrauch D, Mitchell-Olds T (2003) Proc Natl Acad
Sci USA 100:14587–14592.
51. Mauricio R (1998) Am Nat 151:20–28.
52. Benderoth M, Textor S, Windsor AJ, Mitchell-Olds T, Gershenzon J, Kroymann J (2006)
Proc Natl Acad Sci USA 103:9118–9123.
53. Pellmyr O (2003) Ann Mo Bot Gard 90:35–55.
54. Forde SE, Thompson JN, Bohannan BJM (2004) Nature 431:841–844.
55. Cornell HV, Hawkins BA (2003) Am Nat 161:507–522.
56. Thompson JN, Fernandez CC (2006) Ecology 87:103–112.
57. Thompson JN (2005) Curr Biol 15:R992–R994.
58. Thompson JN (2005) Geographic Mosaic of Coevolution (Univ of Chicago Press,
59. Scriber JM (2002) Entomol Exp Appl 104:217–235.
60. Becerra JX (1997) Science 276:253–256.
61. Becerra JX (2003) Proc Natl Acad Sci USA 100:12804–12807.
62. Becerra JX, Venable DL (1999) Proc Natl Acad Sci USA 96:12626–12631.
63. Thompson JN, Cunningham BM (2002) Nature 417:735–738.
64. Nuismer SL, Thompson JN (2006) Evolution (Lawrence, Kans) 60:2207–2217.
65. Janz N, Nylin S, Wahlberg N (2006) BMC Evol Biol 6:4.
66. Edwards SV, Beerli P (2000) Evolution (Lawrence, Kans) 54:1839–1854.
67. Berenbaum MR (2001) Ann Mo Bot Gard 88:45–59.
68. Berenbaum MR (2002) J Chem Ecol 28:873–896.
69. Huelsenbeck JP, Ronquist F (2001) Bioinformatics 17:754–755.
70. Ronquist F, Huelsenbeck JP (2003) Bioinformatics 19:1572–1574.
71. Pollock DD, Watt WB, Rashbrook VK, Iyengar EV (1998) Ann Entomol Soc Am 91:524–
72. Kumar S, Tamura K, Nei M (2004) Brief Bioinform 5:150–163.
73. Thorne JL, Kishino H (2002) Syst Biol 51:689–702.
74. Wiegmann BM, Yeates DK, Thorne JL, Kishino H (2003) Syst Biol 52:745–756.
75. Kristensen NP, Scoble MJ, Karsholt O (2007) Zootaxa, in press.
Wheat et al. PNAS ?
December 18, 2007 ?
vol. 104 ?
no. 51 ?