Annu. Rev. Entomol. 2002. 47:733–71
Copyright c ? 2002 by Annual Reviews. All rights reserved
THE ECOLOGY AND EVOLUTION OF ANT
ASSOCIATION IN THE LYCAENIDAE (LEPIDOPTERA)
Naomi E. Pierce,1Michael F. Braby,1Alan Heath,2
David J. Lohman,1John Mathew,1Douglas B. Rand,1
and Mark A. Travassos1
1Museum of Comparative Zoology, Harvard University, Cambridge, Massachusetts
02138; e-mail: firstname.lastname@example.org; email@example.com;
firstname.lastname@example.org; email@example.com; doug firstname.lastname@example.org;
2Department of Zoology, University of Cape Town, Rondebosch 7700, South Africa;
myrmecophily, parasitism, butterfly, symbiosis, biogeography
of all Papilionoidea. The majority of lycaenids have associations with ants that can be
pae employ complex chemical and acoustical signals to manipulate ants. Cost/benefit
of obligately associated groups. Parasitism typically arises from mutualism with ants,
and entomophagous species are disproportionately common in the Lycaenidae com-
pared with other Lepidoptera. Obligate associations are more common in the Southern
has played in the evolution of the Lycaenidae.
The estimated 6000 species of Lycaenidae account for about one third
INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 734
Variation in Lycaenid-Ant Associations: Definition of Terms . . . . . . . . . . . . . . . . . 734
Entomophagy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 737
MECHANISMS OF INTERACTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 737
Chemical Signaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 738
Acoustic Signaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 743
COSTS AND BENEFITS OF LYCAENID-ANT INTERACTIONS . . . . . . . . . . . . . 745
Effects of Ant Association on Host Plant Choice . . . . . . . . . . . . . . . . . . . . . . . . . . . 746
Ants as a Template for Butterfly Diversification . . . . . . . . . . . . . . . . . . . . . . . . . . . . 747
ENTOMOPHAGY AND ANT ASSOCIATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 748
PIERCE ET AL.
Phylogenetic Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 748
From Mutualism to Parasitism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 753
Evolutionary Constraints on Entomophagy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 753
BIOGEOGRAPHIC DISTRIBUTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 754
Zoogeographical Patterns in Ant Association . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 754
Phylogenetic Patterns in Ant Association . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 755
Origin and Evolution of Ant Association . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 758
CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 759
The estimated 6000 species of Lycaenidae account for about one third of all
Papilionoidea (1,55,208,226). Full or partial life histories have been recorded
for about 20% of these species, and of those whose full life histories are known,
about 75% [(N=665 (Table 1)] associate with ants (95,139,181). These associa-
associations), to complex obligate associations in which larvae are always tended
by ants, often by only a single species (30%) (Table 1). Even when lycaenids
are not myrmecophilous, they may be protected against ant aggression by a suite
of ant-associated adaptations. The Lycaenidae are additionally characterized by
striking life history diversity. Herbivorous species consume an unusually wide
array of different plant families (10,80), and a small number of lycaenids (∼3%
of all associations or 12% of obligate ant associations) are parasitic or predatory
ticularly amenable to comparative studies of life history evolution. Colonel John
N. Eliot laid the groundwork for such research in 1973 by providing what he des-
cribed as a “tentative arrangement” of the higher classification of the Lycaenidae
(83). He later revised this scheme to comprise the five subfamilies we recog-
nize in our treatment here: the Riodininae, Curetinae, Poritiinae, Miletinae, and
Lycaeninae (44,84; Figure 1). A molecular study by Campbell and colleagues
(35) corroborated the broad outlines of Eliot’s hypothesis. The Riodininae form a
monophyletic group and are sometimes considered a separate family (124); they
are the sister taxon to the remaining four subfamilies, which together are mono-
phyletic relative to outgroups from the Nymphalidae (35). For convenience, we
refer to the latter collectively as Lycaenidae sensu stricto (s.s.).
Variation in Lycaenid-Ant Associations: Definition of Terms
Various workers (77,95,104,139,153,154,181) have categorized the degree of
lycaenid-ant associations into three broad types: obligate, facultative, and non-
ant associated. Gray areas between these categories occur, and interpretations
of facultative and obligate have varied. For clarity, we define these interactions
ant-associated species within each lycaenid taxonomic group
(subfamilies and tribes)
The number and proportion (in parentheses) of
Taxonomic groupN Obligate FacultativeNone
Records compiled primarily from (17,21,22,24,39,41,43,44,48,53,70,74,77,79,89,
183,190,195,199,203,205,236,238), and numerous journal articles of single taxa. Doubtful
and hypothetical records such as predictions based on associations of closely related species
are excluded, and information for Riodininae are not included (see text).
as follows and use the term entomophagy to describe the feeding behaviors of
with ants during at least some portion of the life cycle and are dependent on ants
for survival under field conditions. These include both mutualistic and parasitic
only a single species or genus of ant.
Facultative associations are those in which lycaenid larvae are found only in-
termittently associated with ants, either spatially or temporally, and do not require
attendant ants for survival under field conditions. Associations are nonspecific:
Larvae of a particular lycaenid may associate with ants from numerous species,
each partner benefiting from the presence of the other. However, a few species
are facultatively predaceous, occasionally consuming the ants that normally tend
them (e.g., 130).
PIERCE ET AL.
Non-ant-associated, or mymecoxenous, lycaenids are characterized by the ab-
sence of apparent associations with ants. The term myrmecoxenous underscores
the notion that, unlike most lepidopterans, larvae of these lycaenids possess ant-
related adaptations that protect them against aggression, even if they are not ac-
tively tended. Kitching & Luke (153) coined the term to describe species lacking
a specific ant-associated organ, the dorsal nectary organ (see below), but we use
it in a broader sense to describe lycaenids not tended by ants. Chemical defense
from exocrine glands, hairiness, thickness of larval cuticle, and/or construction of
Like other Lepidoptera, the great majority of lycaenid larvae feed exclusively on
living plant tissue. However, some use insect-derived food resources during all or
part of their development. These include (a) ant eggs, larvae and pupae (myrme-
cophagy), (b) ant regurgitations (trophallaxis), (c) Homoptera (homopterophagy),
(d) homopteran honeydew, and (e) other lycaenid larvae (facultative cannibalism
too general when applied to species that supplement an otherwise phytophagous
diet with a nonplant resource or that switch from phytophagy to aphytophagy
between instars. Carnivory is likewise a limited descriptor, as it excludes food
sources such as regurgitations from trophallaxis and homopteran honeydew. We
instead use the term entomophagous for any species that depends on some insect-
derived resource other than plant tissue at some point during its larval phase.
MECHANISMS OF INTERACTION
Ant association has exerted strong selection on lycaenid larval morphology. Pre-
sumably as a defense against ant bites, the cuticle of lycaenid larvae can be up to
20 times thicker than that of larvae from other lepidopteran families, and the head
can be retracted under a sclerotized prothoracic plate (164,165). Lycaenid larvae
also have a much reduced thrashing response in the presence of ants (164,165).
provoke the attention and aggression of ants (240).
DeVries (59) called the strategic deployment of ant-associated organs “entice-
← − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − −
Phylogeny of the Lycaenidae as proposed by Eliot (83), modified in ac-
cordance with Eliot’s subsequent taxonomic revisions (44), and including Harvey’s
treatment of the Riodininae (taxonomic ranks adjusted; dashed lines indicate groups
not thought to be monophyletic) (124) with modification from Penz & DeVries (177).
PIERCE ET AL.
and maintaining ant associations. Most of these adaptations are found in larvae
and pupae, but adults also engage in complex interactions with ants. Lycaenids
manipulate ant behavior in at least three ways: suppression of ant aggression,
maintenance of a “standing guard,” and ant-mediated defensive measures. In ad-
dition to more general adaptations, lycaenids possess two highly specialized sets
of organs used in interactions with ants: those involved in chemical mediation and
those involved in acoustical mediation. For simplicity, we discuss each of these
separately, recognizing that they typically function in concert with each other.
Pore cupola organs
(PCOs) are found on the larvae of every lycaenid species yet examined, except
ficially resemble the much smaller lenticles found on some hesperiid caterpillars
(116), there is little evidence of structural or functional homology. Lycaenid PCOs
may secrete substances to pacify ants that might otherwise attack the soft-bodied
abled ancestral lycaenids to benefit from enemy-free space in the presence of ants
Single-celled epidermal glands called pore cupola organs
the eighth abdominal segment secretes volatile substances that attract and alert
ants if a caterpillar is alarmed (11,38,95,106,132). These organs are also found
in the Curetinae where they are enlarged, occur more centrally on the dorsum, and
may confer mechanical defense (56). In the Miletinae, only species in the genus
Aslauga possess TOs.
In the Lycaeninae, an eversible pair of tentacle organs (TOs) on
organ (DNO) on the seventh abdominal segment produces nutritious secretions
for ants and plays a critical role in the maintenance of ant/lycaenid mutualisms
(11,160,174). It has been called the honeydew gland by analogy with the excre-
ment of homopterans, but it is in fact a specialized exocrine gland (165).
Apparent losses of TOs and the DNO are found throughout the Lycaenidae
and are often correlated with reduced ant association. For example, larvae in the
entomophagous subfamily Miletinae lack a DNO, and all but those in the genus
Aslauga lack TOs as well. They are nonetheless ignored by the ants tending the
Homoptera on which they feed, suggesting that the PCOs may appease ants in this
group (48) and perhaps in the whole family.
Restricted to species in the Lycaeninae, the dorsal nectary
function is unknown. (b) Dish organs: Clark & Dickson (41) described saucer-like
glands located dorsally on the fifth segment of the final instar larvae of Spindasis
et al. (67) noted that the larvae of species of Curetis possess unusual epidermal
organs of unknown function located laterally on the first thoracic segment and the
unusual structures on the larva of the miletine Allotinus major and, in particular,
noted the presence of ovoid flattened organs scattered along the central axis of the
larva and clustered in groups of 20 or more on lateral discs near the posterior end.
Ants showed great interest in these areas, and Kitching proposed that these organs
might be of special significance in the Miletinae.
Riodinine ant organs
been described from three relatively derived ant-associated subtribes [Eurybiiti,
Lemoniiti, and Nymphidiiti (63,124)]. Recent phylogenetic work by Penz &
DeVries (177), however, shows that the Lemoniiti are rooted within the
Nymphidiiti, so the three groups collapse to two, the Eurybiiti and Nymphidiiti.
Nutritious droplets are secreted from paired tentacle nectary organs (TNOs) on
the eighth abdominal segment, while paired anterior tentacle organs (ATOs) on
the third thoracic segment presumably emit volatile compounds (63). Differences
in the location and structure of ant-associated glands between riodinines and their
sister taxa have been cited as evidence for convergence of both form and function
(61). In view of the evolutionary importance of tagmatization in arthropods, how-
ever, and the ease of switching between segment identities, it is possible that these
organs are homologous (34).
Ant associations in the Riodininae have traditionally
SUPPRESSION OF ANT AGGRESSION
some data (3,132,134) suggest that lycaenid larvae suppress ant aggression in
part by mimicking aspects of the pheromones of ant brood. Attendant ants lick
and antennate lycaenid larvae much as they do their own brood. The substance(s)
responsible in both cases is widely dispersed over the cuticle and relatively non-
volatile, persisting for several days after death (27,106,113,182,234). The larvae
of parasitic inquiline lycaenids are often deposited by workers in the brood cham-
ber of the ant nest alongside the immature ants (217,219).
Henning (132,134) showed that the lycaenids Aloeides dentatis and Lepido-
chrysops ignota chemically mimic the brood of their respective attending ants
(3) demonstrated that larvae of the lycaenid Maculinea rebeli produce a profile of
hydrocarbon compounds sufficiently similar to that of Myrmica schencki ant lar-
vae to induce M. schencki workers to carry M. rebeli larvae into their nests, where
the caterpillars eat the brood. They showed that after seven days inside the nest,
became nearly perfect chemical mimics of the ant brood.
Much discussion (48,88,106,164,182) and
PIERCE ET AL.
Some authors nevertheless doubt the existence of brood pheromones in ants
(169,224). Ants can clearly differentiate brood from workers and discriminate
among different types of brood (e.g., worker-biased larvae, queen-biased larvae,
worker pupae) (161). The experimental problem is to distinguish between a feed-
ing response in which an object is carried into the brood chamber from a simi-
lar response elicited by an object coated with brood pheromone. This remains a
weakness of all published studies on lycaenids (3,132,134). Nevertheless, recent
studies (156,221) have demonstrated that cuticular hydrocarbons communicate
colony membership in worker ants, and hydrocarbons may serve a related purpose
in labeling ant brood in some as-yet-unknown way.
Even if cuticular hydrocarbon matching is responsible for suppressing ant ag-
gression toward lycaenid immatures, it remains to be seen whether PCOs are
involved in their production. While PCOs are unique to lycaenid immatures, cu-
ticular hydrocarbons are ubiquitous among insects as they serve essential roles in
waterproofing and osmoregulation (120,172).
MAINTENANCE OF A “STANDING GUARD” OF ANTS
tending caterpillars may depend on the mode of appeasing ant aggression (such
as ant brood chemical mimicry) but may also relate more directly to the quantity
and quality of nutritive rewards offered to ants. Lycaenids producing particularly
valuable secretions would be expected to maintain a larger cadre of dominant,
aggressive attending ants than those producing less valuable secretions, as seen in
The nutritive rewards secreted by immatures of several lycaenid species have
been analyzed chemically. Maschwitz et al. (167) found carbohydrates (13–19%)
and trace amounts of methionine from the DNO secretion of the facultative ant
associate, Lysandra hispana. Pierce (182,188) also found carbohydrates to be an
important component of the DNO secretions of the Australian lycaenid, Jalmenus
evagoras, making up about 10% dry weight. These secretions also contained at
least 14 different free amino acids, particularly serine, in concentrations ranging
from 20 to 40 mM, depending on the time of day (188). DeVries found that
TNO secretions from the riodinine Thisbe irenea contain at least 18 amino acids,
with glutamine and glycine predominating, and small amounts of sugars (<0.5%)
Behavioral assays indicate that host plant quality can affect the secretions from
the field than counterparts feeding on low-quality unfertilized plants, and females
also preferred to lay eggs on these high-quality plants.
Lycaenid larvae can manipulate their attendant ants by strategically varying the
defenses in plants (2). When under a perceived threat (a pinch from forceps),
The persistence of ants in at-
larvae of both Polyommatus icarus (160) and Plebejus acmon (2) secrete more
rewards and/or attractants from the DNO and thereby attract a greater number of
attendant ants. Individual larvae of J. evagoras also regulate secretions depending
each does when alone (12). Beyond a threshold number of ant guards, the benefit
from producing metabolically expensive secretions may have diminishing returns
Wada and colleagues (228) found that the DNO secretions of the parasitic
larvae of Niphanda fusca fed by worker trophallaxis in the nest contain high titres
of glucose and glycine. Recordings from the taste receptors of the attendant ant
glycine when combined with glucose made these solutions much more attractive
to attendant ants.
Eavesdropping on ant trail pheromones has been demonstrated in the behav-
ior of Euliphyra mirifica and Euliphyra leucyana, which are parasites of arbo-
real weaver ants, Oecophylla longinoda (54). Adult females of a close relative,
L. brassolis, typically lay eggs on foliage adjacent to established nests of Oeco-
phylla smaragdina. The first instars probably locate the ant nests by following ant
trail pheromones. Larvae must also move between nests after they have consumed
the brood within a nest or whenever host ants abandon old nests and construct new
ANT-MEDIATED DEFENSE TACTICS
behavior following this signal is similar to that released by ant alarm pheromone.
Mimicry of an ant’s alarm pheromone has been documented in the Australian
spider Habronestes bradleyi, which uses the chemical to disorient its ant prey,
Iridomyrmex purpureus (5).
and often the same chemical provokes an alarm response in species from different
mimic might communicate the need for protection with multiple ant species. Ant
alarm pheromones and TO/ATO pheromones are highly volatile, which makes
pheromone collection for chemical analysis difficult. Henning (132), however,
succeeded in extracting the posterior half of Aloeides dentatis (which contains the
TOs) in dichloromethane, and these extracts elicited an alarm response from at-
tending ants in bioassays. These ants responded similarly when presented with
dichloromethane extracts from conspecific mandibular glands, which produce
alarm pheromones in several ant species (25,142).
Myrmecophilous lycaenid caterpillars can sig-
philous Lycaenidae, or if they do, treat them much as they would any insect
prey. For example, among species of Chrysoritis and Thestor, adults are killed
Ant workers frequently do not interact with the adults of myrmeco-
PIERCE ET AL.
and eaten if they fail to escape from their corral or byre on eclosion. Species
that eclose inside ant shelters are frequently cloaked in “eclosion wool,” a layer
of deciduous scales that slough off in an ant’s mandibles and tarsi, enabling
the teneral adult to slip away. Nevertheless, several observations suggest that
chemical interactions between adults and ants may be more complex than cur-
rently appreciated, and some adults may appease ants that would otherwise attack
of Anoplolepis longipes tending larvae of Curetis regula were keenly interested in
the adults, palpating them with their antennae and appearing to feed near the base
of the butterfly’s extended proboscis. Larvae of the Australian species Ogyris gen-
teneral adults harden their wings in this shelter without harassment (75). Oviposit-
ant, Oecophylla smaragdina, but workers cease attacking once a female has begun
laying eggs perhaps because the eggs release an appeasement pheromone (109).
ants may also be used as mating and oviposition cues. Males of the Australian
species J. evagoras use ants as cues in finding available females (82). The females
of a number of myrmecophilous lycaenids use attendant ants as oviposition cues
(9,108,109,132,185,209,232). The great majority of these are species with obli-
gate ant association (e.g., 185), although some are facultatively associated (232).
Significantly from the point of conservation practices, experiments have failed
to find ant-dependent oviposition not only among females of facultative species
ever, analyses taking into account host plant phenology, intraspecific competition
in egg deposition, and location of ant nests show that ant-dependent oviposition
may occur in at least one species, Maculinea alcon, but simply be difficult to de-
tect experimentally (225). Differences in adoption times depending on ant hosts
have been measured for caterpillars from separate populations of M. alcon (6),
and genetically differentiated populations of this species are suggestive of host ant
Females from populations of J. evagoras show a remarkable degree of speci-
ficity in ant-dependent oviposition, preferring to lay eggs in response to cues
from their natal ant population rather than other populations of ants of the same
infested host plants (82,117,185). A number of ant-associated insects locate ants
through ant trail pheromones (4,33,54), which the chemosensory structures in the
antennae and ovipositors of adult female lycaenids may likewise be capable of
Most sound production in the Lepidoptera has evolved in response to selection on
sexual or defensive traits (90); juvenile sound production in the Lycaenidae also
mediates associations with ants.
PUPAL SOUND PRODUCTION: MECHANISMS AND DISTRIBUTION
produce sound by a mechanism that is widespread among lepidopteran pupae:
stridulation, the act of grating a file lined with teeth against a hardened plate
(71). In lycaenids, the file-and-scraper organ is found in the intersegmental region
between abdominal segments 4–5, 5–6, or 6–7.
Lycaenid pupae can give three distinct signals (73,141). A primary signal,
often produced by stimulation of the pupa, is detectable without amplification. A
secondary signal is of lower amplitude, consisting of a set of clicks produced in
A pupal stridulatory organ has been found in every lycaenid pupa examined
(72,81), including both ant-associated and non-ant-associated taxa. This includes
ten species of riodinine pupa in the tribe Hamearini and five subtribes of the
Riodinini. Each of these riodinine pupae possesses two sets of file-and-scraper
stridulatory organs, located between abdominal segments 4–5 and 5–6.
PUPAL SOUND PRODUCTION: FUNCTION
turbed, and Downey & Allyn (73) concluded that calls act primarily as a deterrent
to predators and parasites. Travassos & Pierce (222) showed that pupal calling
is also involved in ant recruitment. In pairs of pupae of J. evagoras, where one
guard than their silent counterparts. Eastwood & King (78) observed that pupae
of the myrmecophage, Arhopala wildei, emitted a prolonged “burr” lasting sev-
eral seconds upon reintroduction to the ant nest. The pupae also oscillated with a
rapid dorso-ventral movement of the anterior end in a frequency that matched the
frequency with which the host ants tapped the substrate when alarmed.
DeVries (60) surveyed 26 species of pupae in the Riodininae but found no
evidence of pupal sound production. However, experiments conducted by Ross
(197) on Lemonia caliginea (=Anatole rossi) suggest that sound production in
organs, L. caliginea pupae possess a pair of glands on the metathoracic segment
that may produce a chemical attractive to ants (196). Ross found that a fast-drying
lacquer applied to either the paired metathoracic organs or the stridulatory organs,
that the stridulatory organs and metathoracic glands work in concert to attract and
maintain attendant ants.
Calls are induced when pupae are dis-
PIERCE ET AL.
LYCAENID SENSU STRICTO LARVAL SOUND PRODUCTION
(60), although larvae of Hemiargus isola have been recorded to start calling in
the second instar (D. Wagner, personal communication). In general, these calls
resemble a slow drumming compared with the faster chirps of the riodinines.
They travel primarily through the substrate, although they also have an airborne
component (60,201). DeVries (60) first noted that some lycaenid calls have two
signals: a low background sound accompanied by a constant pulse. Travassos &
Pierce (222) found that J. evagoras larvae produce three signals that differ in
acoustic properties and amplitudes: the grunt, drum, and hiss.
Hill (138) localized rapid trembling to abdominal segments five and six of
Arhopala madytus and described a file of teeth found on the posterior margin of
the fifth abdominal segment that rubs against an opposing plate on the anterior
margin of the sixth abdominal segment. A. madytus larvae thus produce calls with
a stridulatory organ similar to that found in lycaenid pupae, although the position
segments 4–7 when producing calls.
Because the organs for sound production in lycaenids remain poorly charac-
terized, we cannot determine how widespread they may be within the family. In
his sound production survey, DeVries (59,60) found that only myrmecophilous
lycaenids produce calls, and several non-ant-associated members of the Eumaeiti
are silent. However, since then, several non-myrmecophilous lycaenids have been
Caleta manovus (100), and Cheritra freja (98). Callophrys rubi, Curetis bulis, and
Curetis santana also produce sound, but accounts of whether ants attend these
species conflict (26 versus 113 and 56 versus 112). Fiedler (97) suggested that
the ability to produce calls may be universal in the Lycaenidae: Whereas non-
ant-tended lycaenids produce simple calls in response to a disturbance, ant-tended
lycaenids have calls of greater complexity that are produced more often.
DeVries (58), Fiedler (98), and D. Wagner (personal communication) observed
that lycaenid larvae produce sounds when disturbed, suggesting a defensive func-
tion. DeVries (60) argued that lycaenid calls, like riodinine calls, attract ants.
Travassos & Pierce (222) found that two calls produced by larvae of the lycaenid,
Moreover, the calls of Maculinea larvae and the stridulations of the Myrmica ants
they parasitize share the same pulse length and dominant frequency, suggesting
a convergence of caterpillar calls on those produced by host ants (66). Lycaenids
that form obligate associations with different ant species may have evolved dif-
ferent acoustic signals. Attendant ants are widely distributed phylogenetically,
including representatives from the subfamilies Myrmicinae, Dolichoderinae, and
Formicinae (77). Of these, only myrmicines are known to produce calls via stridu-
lation, whereas some dolichoderine and formicine ants produce vibratory signals
by drumming body parts on the substrate (142,166).
Lycaenids (s.s.) typically
LARVAL SOUND PRODUCTION IN RIODININAE
different mechanisms than from those of the Lycaenidae (s.s). Some riodinine
larvae have a pair of vibratory papillae on the distal edge of the prothorax, each
one bearing concentric grooves along its length (62). When a caterpillar oscillates
its head, epicranial granulations on the head slide across these grooves, producing
low-amplitude calls that travel solely through the substrate, unlike the airborne
caterpillars of T. irenea with those that had their papillae removed and found that
calling caterpillars were tended by more ants.
Riodinine larvae produce sound by
COSTS AND BENEFITS OF LYCAENID-ANT INTERACTIONS
Early studies of costs and benefits of lycaenid-ant interactions focused on the ben-
efits of the lycaenids, generating two nonexclusive hypotheses. The appeasement
hypothesis is simply that ant-associated lycaenid larvae (excluding those species
parasitic on host ants) benefit from not being attacked by ants. The food rewards
they provide attendant ants can be regarded as a kind of bribery (162,164,165).
The protection hypothesis argues that attendant ants guard lycaenid larvae against
predators and parasites, and in turn they are rewarded with nutritious secretions
in this way, although the importance of ant protection varies among species
(61,107,111,184,186,187,197,198,206,229–231; but see 178). This does not
preclude the possibility that ant-tended lycaenids also appease otherwise aggres-
sive ants and thereby inhabit enemy-free space (10).
Attendant ants can affect lycaenid development and/or reproductive success
(20,16,49,82,105,110,186,192,229,233). In mutualistic interactions, the cost
for this net benefit is sometimes meted out in terms of adult weight and size.
Larvae and pupae of J. evagoras tended by ants are considerably (25%) lighter
and smaller than experimental counterparts not tended by ants (186). Because size
and weight are determinants of female fecundity and male reproductive success in
J. evagoras (82), this size reduction represents a significant cost. Additional costs
can come in the form of increased apparency: Some parasitoids use chemical cues
from host ants to find their lycaenid prey (140,213).
Other lycaenids, such as the facultative Hemiargus isola (229) and the obli-
gate Paralucia aurifera (49), may compensate for the secretions they give up to
attendant ants. H. isola eclose at a heavier weight when tended by one species of
ant, and P. aurifera develop more quickly and gain more weight when tended by
ants perhaps because they spend more time feeding when tended. Some species
exhibit sex-specific effects during development, wherein one sex supports more of
the cost of the mutualism than the other. Thus, Polyommatus icarus males develop
relatively more quickly than their female counterparts when tended by ants (105),
and males of J. evagoras experience relatively less reduction in size than their
female counterparts when tended by ants (20).
PIERCE ET AL.
portion of their lifetimes have unusually variable development (183). Some have
prolonged developmental times, often overwintering as larvae in the nest (133),
some are variable with respect to overwintering (200), and others vary consider-
ably in final adult size (18,78). Elmes et al. (87) showed that phytopredaceous
lycaenids such as species of Maculinea and other parasites of ants also have un-
usual growth patterns between instars. The small, phytophagous early instars have
regular growth, but the entomophagous final instar has >10 times the growth pre-
patterns are similar to those exhibited by their Myrmica ant hosts.
studied; it has been assumed that ants receive a net nutritional benefit from har-
vesting secretions despite the metabolic cost of protecting caterpillars. However,
sometimes the ants do not benefit, as exhibited most dramatically by species that
parasitize host ants. More subtle forms of manipulation have been demonstrated
by the strategic, or inducible, nature of the rewards offered to ants by different
lycaenid species (2,12,160). Natural selection should favor strategies whereby it
is less expensive metabolically to fool ants into attendance while still receiving a
As described above, attendant ants are rewarded with nutritious secretions
access to lycaenid secretions (49,110,111). Queenright colonies of the attendant
ant Iridomyrmex rufoniger showed net gains in growth rate when their food was
supplemented with secretions from larvae of J. evagoras (171,188).
Effects of Ant Association on Host Plant Choice
Experimental cost/benefit studies of lycaenid-ant interactions have elucidated se-
suggested potential evolutionary repercussions that might be expected among ant-
Pierce (180) investigated whether ant-associated lycaenids that reward ants
with protein-rich secretions are also more likely to feed on legumes and other
protein-rich–host plant species. Similarly, Pierce & Elgar (185) assessed whether
of host plant taxa than those that do not. Fiedler subsequently re-examined these
and other patterns using many additional life history records (95,99,101–103).
With a number of conditional caveats, these two patterns of host plant use appear
to be supported when analyzed with larger data sets (101,99). Leguminous–host
obligately associate with ants use a wider range of host plants than lycaenids that
do not associate with ants.
However, none of these comparative studies has taken into account phylo-
genetic effects, in part because of our limited knowledge of the phylogeny of
LYCAENID-ANT ASSOCIATIONS Download full-text
the Lycaenidae, making the validity and/or functional significance of these pat-
terns impossible to evaluate. Additional information about the phylogeny of the
Lycaenidae will be necessary to identify appropriate independent contrasts (92) to
test hypotheses about the evolution of ant association and host plant use.
Ants as a Template for Butterfly Diversification
Attendant ants influence many aspects of lycaenid physiology, behavior, and
ecology. In obligately mutualistic lycaenids, overlapping requirements of suit-
able host plants and attendant ants can lead to population fragmentation and
small population sizes, thus promoting genetic divergence among populations
(10,45,46,114,149,179,206,209). The history of ant association may therefore
be reflected in the cladogenesis of ant-associated butterflies, and at least two lines
of evidence suggest that this is the case.
First, phylogenetic studies of four independent lineages of ant-associated ly-
caenids have shown that sister taxa typically share closely related ant associates:
Taxa within the Aphnaeini, Ogyriti, Zesiiti, and Luciiti all show phylogenetic
conservatism with respect to ant association. The genus Chrysoritis (Aphnaeini)
associates largely with Crematogaster ants (Myrmicinae). Other aphnaeines, in-
cluding species of Phasis and Tylopaedia, also associate with myrmicines. In con-
in the subfamily Formicinae, thereby forming distinctive clades within a largely
myrmicine-associated group (191,211).
Most strikingly, the species of the Australian genus Ogyris form several dis-
tinctive clades that correspond well with their ant affiliations. For example,Ogyris
ing Ogyris ianthis, Ogyris iphis, and Ogyris aenone associates with Dolichoderi-
nae; and the clade containing Ogyris oroetes, Ogyris olane, and Ogyris barnardi
has lost or greatly reduced association (N.E. Pierce, A.A. Mignault, G.S. Adelson,
R. Eastwood, D.J. Lohman, M. Blair & T. Itino, manuscript in preparation).
Within the Australian genus Jalmenus, a small group of closely related taxa
with the highly distinctive Iridomyrmex purpureus–species group and its close
A. myrmecophila and A. brisbanensis, associate with ants in the Dolichoderinae.
A speciation event associated with a host-ant shift to Myrmicinae occurred in the
ancestor of the clade containing the sister taxa Acrodipsas cuprea and Acrodipsas
aurata, which further gave rise to the five members of the illidgei species group.
Although the host ants of four species in this group are still unknown, the larvae
of Acrodipsas illidgei itself also feed on ants in the Mymicinae (76).
Just as shifts onto novel host plants may serve as a key adaptation that permits
diversification, shifts by lycaenids to chemically and behaviorally novel clades