Ornament evolution in dragon lizards: multiple gains and widespread losses reveal a complex history of evolutionary change.

Abstract

The expression in females of ornaments thought to be the target of sexual selection in males is a long-standing puzzle. Two main hypotheses are proposed to account for the existence of conspicuous ornaments in both sexes (mutual ornamentation): genetic correlation between the sexes and sexual selection on females as well as males. We examined the pattern of ornament gains and losses in 240 species of dragon lizards (Agamidae) in order to elucidate the relative contribution of these two factors in the evolution of mutual ornamentation. In addition, we tested whether the type of shelter used by lizards to avoid predators predicts the evolutionary loss or constraint of ornament expression. We found evidence that the origin of female ornaments is broadly consistent with the predictions of the genetic correlation hypothesis. Ornaments appear congruently in both sexes with some lineages subsequently evolving male biased sexual dimorphism, apparently through the process of natural selection for reduced ornamentation in females. Nevertheless, ornaments have also frequently evolved in both sexes independently. This suggests that genetic correlations are potentially weak for several lineages and sexual selection on females is responsible for at least some evolutionary change in this group. Unexpectedly, we found that the evolutionary loss of some ornaments is concentrated more in males than females and this trend cannot be fully explained by our measures of natural selection.

Full text

Page 1

eScholarship provides open access, scholarly publishing
services to the University of California and delivers a dynamic
research platform to scholars worldwide.
University of California
Peer Reviewed
Title:
Ornament evolution in dragon lizards: multiple gains and widespread losses reveal a complex
history of evolutionary change
Author:
Ord, T J, University of California, Davis
Stuart-Fox, D
Publication Date:
05-01-2006
Publication Info:
Postprints, UC Davis
Permalink:
http://escholarship.org/uc/item/38r8200z
Additional Info:
This is an electronic version of an article published in the Journal of Evolutionary Biology 19:
797-808, 2006.
Keywords:
agamidae, comparative methods, ecological constraint, genetic correlation, lizard, ornaments,
phylogenetic reconstruction, sexual dimorphism, sexually selected traits
Abstract:
The expression in females of ornaments thought to be the target of sexual selection in males
is a long-standing puzzle. Two main hypotheses are proposed to account for the existence of
conspicuous ornaments in both sexes (mutual ornamentation): genetic correlation between the
sexes and sexual selection on females as well as males. We examined the pattern of ornament
gains and losses in 240 species of dragon lizards (Agamidae) in order to elucidate the relative
contribution of these two factors in the evolution of mutual ornamentation. In addition, we tested
whether the type of shelter used by lizards to avoid predators predicts the evolutionary loss
or constraint of ornament expression. We found evidence that the origin of female ornaments
is broadly consistent with the predictions of the genetic correlation hypothesis. Ornaments
appear congruently in both sexes with some lineages subsequently evolving male biased sexual
dimorphism, apparently through the process of natural selection for reduced ornamentation in
females. Nevertheless, ornaments have also frequently evolved in both sexes independently. This
suggests that genetic correlations are potentially weak for several lineages and sexual selection
on females is responsible for at least some evolutionary change in this group. Unexpectedly, we
found that the evolutionary loss of some ornaments is concentrated more in males than females
and this trend cannot be fully explained by our measures of natural selection.

Page 2

Ornament evolution in dragon lizards: multiple gains and
widespread losses reveal a complex history of evolutionary change
T. J. ORD* & D. STUART-FOX?
*Section of Evolution and Ecology, University of California, Davis, USA
?School of Animal, Plant and Environmental Science, University of the Witwatersrand, Wits, South Africa
Introduction
Males of many animals use exaggerated physical fea-
tures, or ornaments, as sexual advertisements or signals
during aggressive encounters. Such ornaments include
the spectacular plumage of many birds (e.g. Andersson
et al., 2002) and the elongated tail filaments and dark
vertical bars exhibited by many fishes (e.g. Basolo,
1990,1991; Moretz & Morris, 2003). The origin of such
elaborate ornamentation is a topic of long-standing
interest to biologists. Sexual selection (Darwin, 1874) is
widely believed to be the driving force behind the
evolution of most extravagant male traits. Yet, structures
assumed to be exclusively selected for in males are also
frequently observed in females (reviewed in Amundsen,
2000b). The prevailing view is that sexual selection acts
on males to produce male biased sexual dimorphism. The
widespread occurrence of ornamentation in both sexes is
therefore puzzling, but surprisingly few studies have
examined this phenomenon. Furthermore, traditional
emphasis has been on explaining the evolutionary gain
of ornamental traits targeted by sexual selection in males.
However, a growing number of studies are showing that
many sexually selected traits are often lost more times
than they have evolved (reviewed in Wiens, 2001). This
trend is taxonomically widespread (e.g. insects – Emlen
et al., 2005; fish – Basolo, 1996; amphibians – Emerson,
1996; lizards – Wiens, 1999; Quinn & Hews, 2000; birds –
O ¨deen & Bjo ¨rklund, 2003; de Kort & Ten Cate, 2004) but
remains largely unexplained (Wiens, 2001).
Two main hypotheses are proposed to account for the
existence of conspicuous ornaments in both sexes:
genetic correlation between the sexes (e.g. Lande,
1980) and/or sexual selection on females as well as
Correspondence: T.J. Ord, Section of Evolution and Ecology, University of
California, Davis, One Shields Avenue, Davis, CA 95616-8755, USA.
Tel.: +1 530 752 3075; fax: +1 530 752 1449;
e-mail: tjord@ucdavis.edu
ª 2005 THE AUTHORS 19 (2006) 797–808
JOURNAL COMPILATION ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY
797
Keywords:
Agamidae;
comparative methods;
ecological constraint;
genetic correlation;
lizard;
ornaments;
phylogenetic reconstruction;
sexual dimorphism;
sexually selected traits.
Abstract
The expression in females of ornaments thought to be the target of sexual
selection in males is a long-standing puzzle. Two main hypotheses are
proposed to account for the existence of conspicuous ornaments in both sexes
(mutual ornamentation): genetic correlation between the sexes and sexual
selection on females as well as males. We examined the pattern of ornament
gains and losses in 240 species of dragon lizards (Agamidae) in order to
elucidate the relative contribution of these two factors in the evolution of
mutual ornamentation. In addition, we tested whether the type of shelter used
by lizards to avoid predators predicts the evolutionary loss or constraint of
ornament expression. We found evidence that the origin of female ornaments
is broadly consistent with the predictions of the genetic correlation hypothesis.
Ornaments appear congruently in both sexes with some lineages subsequently
evolving male biased sexual dimorphism, apparently through the process of
natural selection for reduced ornamentation in females. Nevertheless, orna-
ments have also frequently evolved in both sexes independently. This suggests
that genetic correlations are potentially weak for several lineages and sexual
selection on females is responsible for at least some evolutionary change in this
group. Unexpectedly, we found that the evolutionary loss of some ornaments
is concentrated more in males than females and this trend cannot be fully
explained by our measures of natural selection.
doi:10.1111/j.1420-9101.2005.01050.x

Page 3

males (e.g. Amundsen, 2000a,b). Darwin realized that
many of the structures he postulated to have evolved as a
consequence of sexual selection on males were also
expressed in females. He explained this phenomenon
through a process of ‘inheritance’ in which traits selected
for in males become ‘transferred’ to females (Darwin,
1874, pp. 236). Lande (1980; Lande & Arnold, 1985)
formalized this idea with a mathematical model that
explained how a genetic correlation between the sexes
might lead to traits selected for only in males to also
appear in females. Depending on the strength of this
correlation, a trait initially expressed in both sexes could
eventually become restricted to one sex through natural
selection. Natural selection will favour reduced orna-
mentation if the production or maintenance of conspic-
uous ornaments is energetically costly (e.g. Basolo &
Alcaraz, 2003) and/or increases the risk of predation
(e.g. Wallace, 1889; Stuart-Fox et al., 2003; see also
Langerhans et al., 2005). While the strength of sexual
selection is enough to override the influence of natural
selection on males, this cost will select against the
expression of ornaments in females and ultimately result
in the evolution of sexual dimorphism. The genetic
correlation hypothesis predicts an evolutionary gain in
ornamentation in both sexes simultaneously, followed by
an eventual loss of ornamentation in females.
Alternatively, female ornamentation may itself be
the product of sexual selection (West-Eberhard, 1983;
Amundsen, 2000a,b). Frequent intrasexual disputes
over territories and other resources among members
of both sexes have been documented in a variety of
taxa (e.g. fish – Forsgren et al., 2004; lizards – Martins,
1993; birds – Langmore, 1998; Heinsohn et al., 2005;
Siefferman & Hill, 2005). Because the sexes share the
vast majority of their genome, intrasexual selection will
presumably favour the evolution of similar ornamental
traits in both males and females, particularly in the
presence of genetic correlations (Amundsen, 2000a).
Male mate choice is another mechanism by which
ornamentation might evolve in females and may even
promote female biased sexual dimorphism (Amundsen,
2000a; Griggio et al., 2005). The mutual sexual selec-
tion hypothesis predicts an evolutionary gain and
subsequent retention of ornamentation in both sexes.
Most importantly, the historical appearance of orna-
ments in males and females does not necessarily have
to occur in concert. For example, sexual selection
might initially produce male-biased sexual dimorphism
followed by mutual ornamentation when homologous
traits are sexually selected in females.
Here, we examine macroevolutionary patterns of
ornament gain and loss within the family Agamidae
(dragon lizards) from across the ‘old’ world (Africa,
Asia and Australo-Papuan region) in order to elucidate
the relative contribution of genetic correlation and
sexual selection on mutual ornamentation. Sexual
ornaments in these and many other lizards include
gular sacs, nape/dorsal crests, fleshy spines/horns and
rostral appendages (Fig. 1) that are extended or exag-
gerated through posturing during territorial disputes
and mayalso providecues
(Darwin, 1874; Carpenter & Ferguson, 1977; Greer,
1989; Shine, 1990; Manthey & Schuster, 1996; Wat-
kins, 1998; Schulte et al., 2002; Townsend & Larson,
2002; Osborne, 2005). In a previous study (Stuart-Fox
& Ord, 2004), we found dragon species that possessed
many ornaments are also highly sexually dimorphic in
head size, the latter characteristic typically reflecting
the intensity of sexual selection experienced by species
(e.g. male-male competition often selects for males
with larger body or head size – Andersson, 1994).
However, dragons living in open habitats, where there
is an expected increase in predation pressure, are
significantly less likely to evolve ornaments, and this
trend seems to be more pronounced in females than
males (Stuart-Fox & Ord, 2004). Natural selection
appears to have constrained the evolution of ornamen-
tation in dragon lizards although the precise mechan-
ism is unknown. We hypothesized that the types of
shelters used by species occupying open vs. closed
habitats may exert selection on dragon ornamentation
(Stuart-Fox & Ord, 2004). Specifically, species occupy-
ing open habitats tend to flee into dense ground
vegetation, rock crevices or burrows as their primary
escape mechanism whereas closed forest species tend to
hide behind tree trunks or branches. Species occupying
open habitats should therefore be less likely to evolve
protruding ornaments. We test this hypothesis by
examiningtheinfluence
evolution of ornamentation. Our goal was to determine
whether natural selection (via shelter type) explains
ornament loss, especially in females as predicted by the
genetic correlation hypothesis.
Few attempts have been made to examine the
macroevolutionary origins of sexual dimorphism in
sexually selected traits (Dunn et al., 2001; Wiens, 2001;
O ¨deen & Bjo ¨rklund, 2003; Emlen et al., 2005), especially
with respect to the relative contribution of genetic
correlation and mutual sexual selection on the presence
of ornaments in both sexes (Amundsen, 2000b; Wiens,
2001). In addition, the influence of natural selection on
the pattern of gains and losses of sexually selected traits
remains largely unexplored within a comparative frame-
work. In this study, we use data on ornamentation in
males and females for 240 species of dragon lizard to
reveal historical patterns of ornament gain and loss. First,
we determine if dragons are derived from sexually
monomorphic or dimorphic ancestors and whether
evolutionary loss in ornamentation differs between the
sexes. Second, we test whether natural selection –
specifically, the type of shelter used by lizards to avoid
predators – predicts the loss, or constrains the evolution,
of exaggerated ornaments and whether the influence of
natural selection differs by sex.
during matechoice
ofsheltertypeonthe
798
T. J. ORD AND D. STUART-FOX
ª 2005 THE AUTHORS 19 (2006) 797–808
JOURNAL COMPILATION ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY

Page 4

Methods
Data collection
We compiled a database on agamid ornamentation and
shelter type by examining museum specimens and
reviewing published literature (see Stuart-Fox & Owens,
2003 for details). Seven different ornament types were
observed for specimens representing 240 species and
65% of all recognized agamid lizards globally (Macey
et al., 2000; Appendix S1). These ornament types were:
rostral appendage, supra-ocular spines/ridges, nape crest/
spines, gular sac, dorsal crest/spines, tail crest/spines and
enlarged cheeks (e.g. Fig. 1). Physical features unlikely to
result from sexual selection were not scored. For exam-
ple, the Australian thorny devil, Moloch horridus, has an
impressive array of fleshy body spines not included in our
analyses because they putatively function in crypsis and
predator avoidance.
The variety of ornaments across the Agamidae family is
quite extraordinary, which complicates attempts to trace
their evolutionary origins. Some physical features can be
reasonably interpreted as convergent or homologous
across species (e.g. gular sacs), while others are structur-
ally quite different making classification more difficult
(e.g. some rostral appendages – Fig. 1b, d). Rather than
attempt to delineate ornaments by subjective interpreta-
tions of appearance (e.g. are a few, large dorsal spines in
one species homologous to many, smaller dorsal spines in
another species?), we instead grouped structures into
‘ornament types’. Doing so will tend to underestimate
the number of gains and losses reconstructed for some
ornaments (e.g. nape crests and nape spines). Our results
should therefore be considered conservative.
Information on shelter type was also obtained for 180
species using published sources (Appendix S1). Shelter
type was defined as the refuge reported to be used by a
species to avoid a potential predation threat. Four general
Fig. 1 Agamid ornaments: (a) Hypsilurus dilophus, Papua New Guinea, possessing gular pouch (G), nape (N) and dorsal (D) crest/spines; (b)
Ceratophora stoddarti, Sri Lanka, possessing rostral (R) ornament and nape crest/spines; (c) Calotes nigrilabris, Sri Lanka, possessing enlarged
cheeks (C), gular pouch, nape and dorsal crest/spines; and (d) Lyriocephalus scutatus, Sri Lanka, showing rostral ornament, gular pouch, and
supra-ocular (S) spines/ridges. Photos by DSF.
Ornament evolution in dragon lizards
799
ª 2005 THE AUTHORS 19 (2006) 797–808
JOURNAL COMPILATION ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY

Page 5

shelter types were found in the literature: rock crevices,
soil burrows or ‘sand-shimmying’ (species that hide by
burying themselves in loose sand), ground vegetation
(species reported to run under or into small bushes or
grass clumps), and trees/shrubs (species that climb onto
or hide behind trunks or branches of trees and large
shrubs). These are not mutually exclusive, with some
species using more than one shelter type (e.g. the
Australian painted dragon, Ctenophorus pictus, uses both
burrows and ground vegetation). For several riparian
lizards, the primary escape response is to leap into water
(e.g. the Asian sailfin lizards, Hydrosaurus spp., and the
Australian eastern water dragon, Physignathus lesuerii).
These species were scored as zero for all four shelter-
types.
Comparative analyses
A composite phylogenetic tree was created from several
sources as no single phylogeny included all species of
interest (see Stuart-Fox & Ord, 2004). The majority of
species were positioned using the MtDNA-based phy-
logenies presented in Schulte et al. (2002, 2003) and
Macey et al. (2000). To resolve relationships within
Phyrnocephalus and Draco, we used Pang et al. (2003)
and McGuire & Heang (2001), respectively. Gonocepha-
lus grandis, Gonocephalus chameleontinus, and Agama
hispida were placed following Moody’s (1980) morpho-
logical hypothesis. Species not represented in any
phylogeny were left as polytomies within their respect-
ive genera and resolved randomly across 1000 alter-
native trees using
MACCLADEMACCLADE
Maddison, 1992, 1999). These randomly resolved trees
werethen used to examine
ornament evolution in both ancestor state reconstruc-
tions (e.g. Wiens, 1999) and regression analyses (e.g.
Losos, 1994; Martins, 1996; Stamps et al., 1997). We
assumed branch lengths to be of equal length because
branch length information was not available for most
of the species investigated. Species synonyms were
checked using the EMBL reptile database (http://
www.embl-heidelberg.de/?uetz/LivingReptiles.html).
We began our investigation by mapping evolutionary
transitionsbetween ornament
dimorphism through ancestor state reconstructs using
linear parsimony (Swofford & Maddison, 1987; as
implemented in MacClade). Species were scored as zero
when an ornament type was absent in both sexes, one if
present in both sexes, two when present only in males
(male biased dimorphism). There were no cases where
ornaments were present only in females (female biased
dimorphism). Although ornaments present in both sexes
may be larger in males, we were primarily interested in
the presence or absence of mutual ornamentation. For
simplicity, we use the term ‘dimorphism’ to refer to cases
in which the ornament is present only in males (not
simply larger in males).
V3.08A V3.08A
(Maddison &
generalpatterns of
monomorphismand
To compare the relative frequency of evolutionary
change in the sexes more closely, we also reconstructed
the presence (1) or absence (0) of ornament types for
males and females separately. All ancestor state analyses
were conducted with data entered into MacClade as
‘unordered’ and mapped onto phylogenies with the
assumption that gains and losses in character states were
equally likely (e.g. Wiens, 1999; de Kort & Ten Cate,
2004).
Next, we tested whether the type of shelter used by an
animal could reliably predict differences in the number of
ornaments exhibited by species. Specifically, we wanted
to know whether the use of rock crevices and/or
burrowing behaviour (i.e. hiding in soil burrows or
sand-shimmying) has constrained the evolution of
ornaments and whether this trend is differentiated by
sex. The total number of ornaments possessed by each
sex was tallied and entered into a Phylogenetic General
Least Squares (PGLS) regression model (Martins &
Hansen, 1997) with shelter type represented by four
dichotomous independent variables (i.e. rock crevices,
burrows/sand-shimmying, ground vegetation and trees/
shrubs). We then followed the technique outlined by
Martins (1994) to construct 95% confidence intervals
around regression slopes to identify those shelter types
correlated with ornament number. As applied here, PGLS
uses maximum likelihood to calculate an a parameter
that measures the extent phenotypic variation across
taxa can be explained by phylogeny and subsequently
controls for this effect in the regression model. If a is
manually set to 0, results are identical to Felsenstein
(1985)’s independent contrasts (FIC), while forcing a to
be very large (e.g. 15.50+) produces results that essen-
tially ignore phylogeny all together (TIPS). When PGLS is
allowed to estimate a, this parameter can be interpreted
as the extent trait data are phylogenetically conservative,
or the product of rapid evolutionary change (Martins &
Hansen, 1997). All regression analyses were conducted in
COMPARE V4.6 (Martins, 2004).
COMPARE V4.6
Results
Ornament diversity
Of the 240 species included in our study, 70% of species
have at least one ornament, with the most common
feature being a nape crest or spines (Table 1). Possessing
a gular sac and dorsal crest or spines was also widespread,
while the rarest ornament type was a rostral appendage.
Highly ornamented species exhibiting four or more
ornaments are taxonomically diverse, distributed across
32 species and 13 genera. The most spectacularly orna-
mented dragons topping our list come from Asia, specif-
icallytheSriLankan snout
thescelorhinos), Indonesian/New Guinea forest dragons
(Hypsilusus dilophus, Hypsilusus nigrigularis; Fig. 1), Sri
Lankan lyre-headed agama (Lyriocephalus scutatus; Fig. 1)
agama (Harpesaurus
800
T. J. ORD AND D. STUART-FOX
ª 2005 THE AUTHORS 19 (2006) 797–808
JOURNAL COMPILATION ª 2005 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY