vol. 171, no.

4 the american naturalist april 2008

Natural History Note

Parasite-Induced Fruit Mimicry in a Tropical Canopy Ant

S. P. Yanoviak,1,* M. Kaspari,2,3,† R. Dudley,3,4,‡ and G. Poinar Jr.5,§

1. Florida Medical Entomology Laboratory, Vero Beach, Florida Parasites can modify characteristics of arthropod hosts in
32962; diverse ways to facilitate their transmission to subsequent
2. Department of Zoology, University of Oklahoma, Norman,
hosts (Holmes and Bethel 1972; Hurd 1990; Schmid-
Oklahoma 73019;
3. Smithsonian Tropical Research Institute, Apartado 2072, Balboa, Hempel 1998; Moore 2002). These modifications often are
Panama; behavioral, with infected individuals presenting no other
4. Department of Integrative Biology, University of California, obvious phenotypic differences from uninfected conspe-
Berkeley, California 94720; cifics (Sheiman et al. 2006; reviewed by Moore [2002]).
5. Department of Zoology, Oregon State University, Corvallis, However, frequently the altered behaviors are accompa-
Oregon 97331
nied by physical changes that make the host more con-
Submitted July 16, 2007; Accepted October 23, 2007; spicuous, such as an increase in contrast or size. In such
Electronically published February 15, 2008 cases, the modifications may not be directly involved in
facilitation of transmission (Moore 2002) but will be fa-
vored by selection when the terminal host is a visual pred-
ator prone to consume rare morphs (i.e., oddity selection,
novelty selection, or antiapostatic foraging; Curio 1976;
abstract: Some parasites modify characteristics of intermediate hosts
to facilitate their consumption by subsequent hosts, but examples of Lindström et al. 2001). Many suggestive examples of this
parasite-mediated mimicry are rare. Here we report dramatic changes increased conspicuousness strategy exist, but most are not
in the appearance and behavior of nematode-parasitized ants such that supported by field data (Holmes and Bethel 1972; Hurd
they resemble ripe fruits in the tropical rain forest canopy. Unlike 1990; Moore 2002).
healthy ants, which are completely black, infected ants have bright red, Whereas increasing the conspicuousness of familiar pal-
berry-like gasters full of parasite eggs. The infected gasters are held in
atable prey may be a common and effective parasite trans-
a conspicuous elevated position as the ants are walking, and they are
easily detached from living ants, which also exhibit reduced defensive mission strategy, there are few cases of parasite-induced
responses. This combination of changes presumably makes the infected mimicry, whereby the appearance of an intermediate host
ants attractive to frugivorous birds, which ingest the red gasters and is transformed to the extent that it resembles a completely
pass the parasite eggs in their feces. The feces are collected by ants and different organism. The only terrestrial invertebrate ex-
fed to the developing brood, thus completing the cycle. This is the first ample is the trematode Leucochloridium, which causes the
documentation of parasites causing apparent fruit mimicry in an an-
antennae of infected snails to resemble caterpillars (Moore
imal host to complete their life cycle.
2002). When the intermediate host (i.e., the mimic) has
Keywords: behavior, color, frugivory, interaction, nematode, parasit- a sting or otherwise is distasteful to the subsequent host,
ism, Cephalotes, Myrmeconema. such a transformation should be accompanied by a re-
duction in its defenses. Nematode parasites can have this
effect on Lasius ants, causing them to lack normal defen-
sive responses and venom production (Gösswald 1930),
* Corresponding author. Present address: Department of Biology, University
of Arkansas at Little Rock, 2801 South University Avenue, Little Rock, Ar-
although mimicry is not involved in this example. To our
kansas 72204; e-mail: [email protected]. knowledge, there are no reports of parasites causing ar-
†
E-mail: [email protected]. thropods to resemble plant parts or of extreme mimicry
‡
E-mail: [email protected]. combined with a change in host defenses. Here we describe
§
E-mail: [email protected].
what appears to be such a case in the tropical arboreal ant
Am. Nat. 2008. Vol. 171, pp. 536–544. 䉷 2008 by The University of Chicago.
Cephalotes atratus (Formicidae: Myrmicinae).
0003-0147/2008/17104-42731$15.00. All rights reserved. Parasite-induced physical changes (e.g., distended gas-
DOI: 10.1086/528968 ter, intercaste morphology) are fairly common in ants, and
 Parasite-Induced Fruit Mimicry 537

some parasites modify ant colors as well as anatomy Methods

(Schmid-Hempel 1998; Moore 2002). Classic examples in-
clude the pale, distended gasters of nematode- and trem- The majority of the fieldwork was conducted on Barro
atode-infected ants (Wheeler 1907; Gösswald 1930; Carney Colorado Island (BCI), Panama (9.16⬚N, 79.85⬚W), during
1969) and the altered color of ants infected with gregarines a series of short field trips in May and June of 2005–2007
(Crosland 1988) or cestodes (Heinze et al. 1998; Trabalon and November 2005 and 2006 (see Leigh et al. 1996 for
et al. 2000). These changes generally involve an overall a description of the site). Observations from other forests
lightening in color, often resulting from cuticular thinning in the Panama Canal Zone (e.g., Gamboa, Pipeline Road,
(Crosland 1988; Trabalon et al. 2000; but see Muir 1954). Camino de Cruces) were obtained from a September 2007
Ant-associated parasites may cause other generalized mor- survey conducted by S. Price (University of California, Los
phological effects, such as a reduction or increase in overall Angeles). In addition, Cephalotes atratus colonies were ex-
body size (Passera 1976; Trabalon et al. 2000). amined at four lowland locations in the Peruvian Amazon:
Physical effects of parasites on ants may act in concert (1) in and around the city of Iquitos (3.75⬚S, 73.25⬚W);
(2) along the Iquitos-Nauta highway (4.33⬚S, 73.53⬚W;
with behavioral changes (e.g., sluggishness, lack of escape
Mäki et al. 2001); (3) at the Amazon Conservatory for
response) to facilitate consumption of infected workers by
Tropical Studies (ACTS) and Explornapo field stations
birds or other vertebrates (Carney 1969, 1970; Moore
(3.25⬚S, 72.91⬚W; Vásquez Martı́nez 1997; Madigosky and
1995, 2002; Heinze et al. 1998). Generalized alterations of Vatnick 2000); and (4) at the Amazon Conservation As-
body size are an exception; size polymorphism is common sociation Los Amigos field station (CICRA) 100 km west
in ants, and the typical range of these changes (Ⳳ10%– of Puerto Maldonado (12.57⬚S, 70.10⬚W).
20%) is probably irrelevant to enhanced parasite trans- We estimated infection rates in parasitized colonies by
mission in the absence of altered color and other modi- counting the number of red-gastered and normal workers
fications. The magnitude of physical effects often is visiting baits and returning to the nest via vines (see Poinar
independent of parasite loads (Carney 1969; Trabalon et and Yanoviak 2008). To ensure the accuracy of this
al. 2000), but this may depend on the type of parasite method, 300 completely black foragers were collected from
(Crosland 1988) or the type of response measured (e.g., a parasitized colony on BCI and dissected; none was in-
cuticular hydrocarbons; Trabalon et al. 2000). Also, the fected with nematodes. Counts used to determine infection
type or magnitude of phenotypic alterations may differ rates were conducted on one of the first two days of a
among closely related host or parasite species (Heinze et field trip, were initiated immediately on arrival at the focal
al. 1998). The importance of individual or interspecific nests, and were separated by 13 weeks. Conspicuous be-
host variation to parasite transmission is unknown. haviors, in particular gaster flagging, also were noted dur-
In May 2005, we discovered a C. atratus colony in Pan- ing these counts (i.e., we recorded the gaster position of
ama that included several foragers with conspicuous red ants passing a focal point). The total number of ants ob-
gasters (fig. 1). Like other investigators, we initially as- served per colony in a given count ranged from 60 to 620.
sumed that the red coloration represented a new taxon In addition to counting foragers, we opened an infected
(e.g., a social parasite on C. atratus) or a developmental colony occupying a large Jacaranda copaia (Bignoniaceae)
aberration (de Andrade and Baroni Urbani 1999). Dis- tree at CICRA in Peru and collected 1,602 third- and
section of the gaster revealed that it contained hundreds fourth-instar larvae, 1,693 worker pupae, 40 male pupae,
40 callow workers, and 131 adult workers. Subsets of the
of transparent eggs, each housing a small coiled worm,
collected larvae and pupae were haphazardly selected for
subsequently determined to represent a new genus of tet-
dissection to determine the life cycle of the parasite and
radonematid nematode (Poinar and Yanoviak 2008). Nem-
the infection rate in the colony (Poinar and Yanoviak
atode infections are not commonly observed in Cephalotes
2008). We kept the number of adult infected ants killed
(de Andrade and Baroni Urbani 1999), and only one other in the project to a minimum to avoid disrupting the par-
tetradonematid is known from ants (Nickle and Jouvenaz asite cycle. When possible, live ants were returned to their
1987). Tetradonematids are more typically associated with respective colonies after experiments.
wood-boring beetles and flies (Zervos 1980; Peloquin and Cephalotes atratus workers commonly accumulate
Platzer 1993; Tang et al. 1997). They do not greatly modify around patchy resources, such as fresh bird feces, where
the appearance of these insects, and sparse collection rec- they are conspicuous and relatively easy prey. We hypoth-
ords suggest that they are widespread but undersampled esized that the abundance of infected ants would be dis-
(Zervos 1980). Here we describe the dramatic effects of a proportionately high at such stationary food items. To test
new tetradonematid parasite on C. atratus, which appar- this, we compared the relative abundance of infected ants
ently facilitate its transmission to a frugivorous host. at tuna baits with that of ants transiting along vines with
538 The American Naturalist

a Wilcoxon two-sample test (SAS Institute 2002) using the serting one insect pin on either side of the postpetiole.
count data described above. We similarly tested the relative The pins crossed to create an opening too small for the
frequency of gaster flagging in infected ants versus normal gaster to pass, but large enough to avoid contact with the
ants that were observed during counts. Proportional data postpetiole. The thorax of each ant was grasped in the
were arcsine square-root transformed before analysis (So- terminal clip of a 300-g spring scale, and tension was
kal and Rohlf 1995). Other statistical tests are identified gradually increased along the longitudinal axis of the ant
in the results. until the gaster broke free. Ants were paired a priori by
body size based on HW. We used a similar technique to
measure leg removal forces from seven infected and seven
Effects on Ants
normal ants paired by HW. Ants were secured to the card-
To quantify general physical differences between infected board sheet with pins crossing over the thorax. One midleg
and normal ants, we weighed freshly collected workers to was grasped midfemur in the spring scale clip, and the
the nearest 0.1 mg on an analytical balance and estimated force at breakage of the trochanter-femur joint was re-
body size by measuring head width (HW) behind the eyes corded as described above.
(between the vertex spines) to the nearest 0.01 mm with To put the results of these experiments into a more
digital calipers. We used HW as a surrogate for body size natural context (i.e., to determine whether the red gaster
because it is often correlated with other measures (Corn could easily be plucked from an infected foraging ant), we
1980; Trabalon et al. 2000; but see Kaspari and Weiser measured the force required to dislodge nine normal ants
1999). We dissected the gasters of 10 infected workers to from their typical foraging substrate. A thread was secured
assess general changes in internal anatomy caused by the to the dorsal surface of each ant’s gaster with cyanoacrylate
nematodes. We also collected nematode eggs from three adhesive. A spring scale was connected to the loose end
of the ants and mounted them on microscope slides with of the thread, and the ant was allowed to walk freely on
Euparal following dehydration in an ethanol series. a twig ca. 50 cm long and 1 cm in diameter. The thread
To quantify the change in color of parasitized ants, we was then pulled taut and the tension gradually increased
measured the percentage reflectance of the gasters of 11 until the ant lost its grip on the substrate. We recorded
infected and 11 normal C. atratus workers from 380 to the force at the time of release and measured the HW of
700 nm at 10-nm intervals with a spectroradiometer the ant as described above.
(GretagMacbeth Eye-One, Grandville, MI). We also mea-
sured the reflectance of 11 ripe Hyeronima alchorneoides
Predation by Birds
(Euphorbiaceae) fruits present in the general area of in-
fected colonies for comparison with the red gasters (fig. We conducted three experiments to estimate relative pre-
2). Each gaster or fruit was measured more than five times dation rates on red gasters and to implicate the potential
and repositioned between recordings. To condense the predators. In the first experiment, two similar-sized C.
large amount of information thus obtained, we used the atratus workers, one infected and one normal, were se-
median reflectance value at each wavelength across the cured to a twig ca. 1 cm in diameter and 30 cm long. The
spectrum to generate a representative profile for each in- ants were spaced 110 cm apart, and each was held in place
dividual ant or fruit. The averages of these individual pro- by a loop of fishing line passed around the twig and trans-
files were plotted for comparison. versely over the thorax between the pronotal and pro-
We observed that infected foragers tend to hold their podeal spines. Thirteen replicates of experimental twigs
gasters in an elevated position (i.e., gaster flag) almost were tied to vines in tree fall gaps or to branches in the
continuously (see “Results”). We hypothesized that this forest canopy on BCI, and a green plastic colander was
behavioral change is simply biomechanical compensation suspended beneath each replicate to collect any fallen ant
for the mass added by the parasite load. To test this, we parts. Each colander was preloaded with a gaster removed
monitored the gait of seven normal workers after increas- from a normal C. atratus to ensure that fallen ant parts
ing the weight of their gasters by adhering fine sand to were not being scavenged by other ants. The experiment
fingernail polish freshly painted on the gaster dorsum. The was checked at least once a day until an ant was missing
ants were weighed before and after this treatment as de- its gaster or had died. Six of the 13 replicate pairs ulti-
scribed above. mately were excluded from the analysis: gasters were miss-
When handled and manipulated in the lab, the gasters ing from both experimental ants in two replicates, both
of infected individuals appeared to readily detach from the ants died before being depredated in three replicates, and
body at the postpetiole junction. To quantify the breaking the red gaster was found in the subtending colander in
strength of this junction, we secured seven normal and one replicate.
seven infected ants to a small sheet of cardboard by in- In the second experiment, we placed various colors of
 Parasite-Induced Fruit Mimicry 539

clay balls (red, black, pink, yellow, orange, gray, white, Effects on Ants
blue, green, and brown; Polyform Products, Elk Grove
Village, IL) on the ends of pins (fig. 4). The clay balls were The parasite has multiple physical and behavioral effects
ca. 5 mm in diameter (similar to the gaster diameter of a on the ants (table 1). The most conspicuous of these is
medium-sized C. atratus worker) and were serially ar- the change in color from black to red (figs. 1, 2). Although
ranged along vines or tree branches at the BCI, ACTS, and the greatest degree of color change is restricted to the
CICRA field sites (n p 27 replicate series, all sites com- gaster, many individuals also show some reddening of the
bined). The pins in a given linear series were separated by femora (fig. 1). Dissection of normal and infected workers
ca. 10 cm and inserted at an acute angle such that each revealed that the mechanism for the color change does
clay ball was elevated 1–2 cm above the bark surface. The not involve production of red pigmentation. Instead, the
order of the 10 colors in each series was haphazard. The exoskeletal structure or pigment density of the gaster is
clay balls were examined for damage once per day after altered in infected ants such that it becomes translucent
placement. amber, which, in combination with the yellowish nema-
In the final experiment, we fed one infected and one tode eggs held inside, causes a bright red appearance. Gas-
normal C. atratus worker to a free-range chicken housed ter reddening is not uniform among infected workers, and
at CICRA for use in an unrelated research project. We the extent of discoloration is correlated with certain be-
collected feces deposited by the chicken immediately havioral changes, as described below. Reflectance profiles
before and 1.5, 2.5, and 13 h after feeding it the ants. for the gasters of infected ants were similar in shape and
The feces were preserved in 70% ethanol and examined intensity to those for mature Hyeronima alchorneoides
for the presence of ant parts and nematode eggs under fruits (fig. 2).
40# magnification. Apart from the change in color, parasitized ants con-
To quantify the proportion of bird feces gathered by the sistently exhibit two other external anatomical differences
ants, we collected 1300 large particulate food items from from normal ants. First, infected workers are 10% smaller
C. atratus foragers returning to the nest of an infected but 40% heavier than similar-sized uninfected ants (Poinar
colony on BCI in May 2007. Collections were made be- and Yanoviak 2008; fig. 3). This incongruous weight gain
tween 0900 and 1730 hours for 0.5–2 h per day over 5 is due to the parasite load. The rate of mass increase was
days. All bird feces encountered in these samples were similar between infected and normal ants (t-test for dif-
dissected in 1% saline solution under 40# magnification. ference in slopes; t p 1.41, df p 1, P p .16; fig. 3). Sec-
ond, the parasite causes significant weakening of the post-
petiole-gaster junction. The average (ⳲSE) tensile force
required to remove the gaster from an infected ant is one-
fourteenth (18 Ⳳ 3.9 g) that for a healthy ant (254 Ⳳ
Results 28.2 g; paired t p 9.66, df p 6, P ! .0001). The ants ad-
here to twig surfaces with 15 Ⳳ 1.2 g of force; thus, the
Four of the 101 Cephalotes atratus colonies examined (all parasite-laden gaster may be plucked from a forager with-
sites pooled) were infected: three in Panama (two on BCI out removing the ant from its natural substrate (t p
and one near Camino de Cruces) and one at CICRA in 0.81, df p 14, P p .43). This is not true for healthy ants.
Peru. Infection rates (based on red-gastered workers ob-
served outside the nests) averaged 3.5% for the two par- Table 1: Principal effects of tetradonematid infection
asitized colonies on BCI (range p 1.5%–10.0% of workers on Cephalotes atratus workers
per count). In contrast, only three red-gastered foragers Target characteristic Appearance
were found out of several thousand workers encountered
Physical effect:
in the CICRA colony, but dissection of that colony’s brood Gaster Red, full of eggs
showed the actual infection rate to be ca. 6% (Poinar and Body size (HW) 10% smaller
Yanoviak 2008). The infection rate was not determined for Mass 40% heavier
the colony at Camino de Cruces. Gaster-postpetiole junction 93% weaker
Infected ants did not accumulate disproportionately at Internal organs Nerve cord damage
stationary patches of food resources, as we had hypoth- Behavioral effect:
esized. The average relative abundance of infected workers Gaster flagging Virtually constant
foraging at tuna baits (3.0%, n p 10 counts, range p Gait Erect and unstable
1.2%–5.0%) was similar to their abundance among work- Alarm/defensive pheromone
ers running along vines (3.3%, range p 1.3%–5.0%, production Little or none
General activity Sluggish
z p 0.49, P p .62).
540 The American Naturalist

Figure 1: Normal (A) and infected (B) Cephalotes atratus workers. Scale bar ≈ 2 mm.

This exoskeletal weakening is restricted to the postpetiole- bite when handled) and produce no detectable alarm
gaster junction; there was no difference in the force re- pheromones when intentionally disturbed. (Like many
quired to remove legs from normal (159 Ⳳ 10.6 g) and Cephalotes species, C. atratus produces alarm chemicals
infected (143 Ⳳ 13.0 g) workers (paired t p 1.44, df p that are distasteful and have a strong odor readily detected
6, P p .23). by humans.) They also are the most sluggish and the most
Dissection of infected and uninfected ants revealed that unstable among infected ants. In contrast, ants with in-
the physical effects of the parasite extend to their internal complete reddening of the gaster occasionally release alarm
anatomy. The gastrointestinal tract and Malpighian tubules pheromones when disturbed, show typical levels of ag-
of infected ants were completely intact, but the portion of gression and defense, and have a normal gait apart from
the ventral nerve cord located in the gasters of infected continuous gaster flagging. The extent and intensity of
ants consistently appeared atrophied and dark in color. In redness increases as the nematode embryos develop into
most cases, it was clearly visible as a dark semicircle or first-instar worms (see Poinar and Yanoviak 2008).
coil under the translucent gastral sternites. Further dis-
section showed that the parasites are not found in other
Predation by Birds
regions of the ant apart from the gaster.
Nematode infection has several behavioral effects on the Gaster removal rates were higher for infected than for
ants in addition to the physical effects described above normal ants when they were presented in tethered pairs
(table 1). Most conspicuous among these is nearly constant in the field on BCI (binomial test, P p .055). The missing
gaster flagging (frequency among infected workers p gasters in these replicates were not found in the respective
94.5%, range p 86%–100%, n p 10 counts) relative to colanders, nor were the “seed” gasters removed from the
normal workers (3.9%, range p 2.7%–5.8%; Wilcoxon colanders. In the second predation experiment, red and
two-sample test, z p 3.77, P p .0002). Adding sand to pink clay balls (combined) either were attacked by birds
the gasters of normal workers increased their mass by an or were missing more often than the other colors com-
average of 47% (range p 7%–120%) and caused minor bined (G-test; G p 4.61, df p 1, P p .036). Bird attacks
instability when walking, but did not stimulate continuous were distinguishable from damage caused by other animals
gaster flagging or otherwise alter their behaviors in ways because of the V-shaped impressions left in the clay (fig.
resembling infected workers. 4). Finally, hundreds of intact tetradonematid eggs ap-
Changes in the appearance and behavior of the ants are peared in the chicken feces between 2.5 and 13 h after the
associated with each other and linked to development of chicken was fed an infected ant.
the nematodes. Specifically, individuals with the greatest The majority of the solid material carried by foraging
amount of red coloration are less aggressive (they do not C. atratus consisted of bird feces (68%), followed by dead
 Parasite-Induced Fruit Mimicry 541

Figure 2: A, Reflectance profiles of gasters of normal and infected Cephalotes atratus workers and of ripe Hyeronima alchorneoides fruits. For each
series, n p 11 ants or fruits, and each point is the average (Ⳳ95% confidence interval) of the median values obtained from multiple recordings
per ant or fruit. B, An infected ant is shown on an infructescence of H. alchorneoides for comparison. Scale bar ≈ 1 cm.

insects and insect parts (24%), brood (eggs and larvae; worms migrate to the pupal gaster and mate within the
8%), and unidentifiable material (!1%). We found one callow worker (Poinar and Yanoviak 2008). Shortly after
nematode egg containing a live worm (although not con- eclosion, the parasitized adult ant has a black gaster and
firmed to be a tetradonematid) and several other living mainly tends brood, rarely leaving the nest (Corn 1980).
microscopic worms and parasite eggs in the bird feces Hundreds of developing nematode embryos within the
carried by the ants. gravid female worms cause increased reddening and mod-
ified behavior, probably by sequestering food and exo-
skeletal compounds from the adult ant. Just as a fruit
Discussion reaches peak color when its seeds are ready for dispersal,
the gaster reaches peak redness as the nematode eggs are
Here we show that a parasitic tetradonematid nematode
has multiple physical and behavioral effects on its terminal
host, the common Neotropical arboreal ant Cephalotes
atratus, that likely facilitate its transmission to the same
colony and to new colonies via frugivorous birds. Most
conspicuous among these effects is the striking change in
gaster color from black to red, which earned infected work-
ers separate taxonomic status in the late nineteenth cen-
tury (var. rufiventris Emery; de Andrade and Baroni Urbani
1999; Poinar and Yanoviak 2008).
Our results suggest the following hypothesized and pos-
sibly unique life cycle for this new nematode species. A
frugivorous bird mistakenly consumes an infected gaster
and deposits nematode eggs in its feces while perched in
a tree crown. The feces are collected by a foraging C. atratus
(Corn 1980; this study), but the filter-like proventriculus
of the ant’s gastrointestinal tract prevents passage of nem-
atodes to its midgut (Baroni Urbani and de Andrade 1997;
Roche and Wheeler 1997). Instead, transmission to new
Figure 3: Head width (HW) versus mass regressions for normal (n p
ants occurs when the nematode-laden feces are fed to lar- 84) and infected (n p 49) Cephalotes atratus workers. The slopes of the
vae by workers inside the nest. The nematodes develop two regressions do not differ. Infected equation: mass p 2.19HW2.59; nor-
within the ant larvae, partly stunting their growth. Adult mal equation: mass p 1.51HW2.68.
542 The American Naturalist

general sluggishness caused by the parasite should facilitate

consumption of an infected gaster by a duped frugivore.
Two alternatives to the fruit mimicry scenario must be
considered. First, it is possible that all of the described
changes are irrelevant to transmission (i.e., no specific
paratenic hosts are involved and reinfection occurs within
the nest), as might occur if carcasses of parasitized ants
are fed to larvae. However, this pathway provides no mech-
anism for infection of new C. atratus colonies unless (a)
foundress queens carry the parasite or (b) infected workers
invade, or are enslaved by, other colonies (Breed et al.
1999). We did not find nematodes in either male or queen
pupae. An infected foundress likely would have very low
survival and fecundity, and queen replacement is not
known for C. atratus. Colonies of this species often are
very large (Weber 1957; Corn 1980) and may persist for
many years, if not decades. Thus, if transmission occurs
via enslavement, we would expect infected C. atratus col-
onies to be clumped in space. We believe that the absence
Figure 4: Clay ball perched on an insect pin used in predation experi- of a paratenic host is a very unlikely scenario for this
ments. Bird attacks are indicated by the V-shaped impressions. Scale parasite, but more data are needed to adequately assess
bar ≈ 5 mm. this possibility.
The second alternative explanation to fruit mimicry is
that the parasite causes increased conspicuousness of in-
most infective. These changes coincide with temporal poly- fected ants to generalist insectivores. This hypothesis re-
ethism in C. atratus; the worker ant switches from brood quires a priori that healthy hosts are regularly consumed
care inside the nest to foraging as it ages (Corn 1980). The by predators. Despite being very conspicuous (large, black,
gaster is consumed by a frugivorous bird while the ant is diurnal ants that forage in large numbers, often against
outside the nest, and the parasitic life cycle repeats. light-colored tree surfaces; see Yanoviak and Dudley 2006),
Despite hundreds of hours of observing and recording records of insectivorous vertebrates feeding on C. atratus
bird and ant behaviors in tree crowns, we lack direct ob- are very few (de Andrade and Baroni Urbani 1999). Rel-
servation of bird predation on either infected or healthy ative to their abundance and conspicuousness, ants in gen-
C. atratus. However, all evidence gathered to date indicates eral do not represent a large fraction of the prey consumed
that transmission by duped frugivorous birds is the most by tropical insectivorous birds (Chapman and Rosenberg
parsimonious explanation for the changes induced in C. 1991), and there are no published records of bird predation
atratus by the parasite. Among tropical ants, C. atratus on C. atratus. Our field observations indicate that C. atra-
workers are particularly well suited to this mimicry system tus workers are generally ignored by insectivorous birds,
because of their large size, conspicuous spherical gasters, probably because (like many other Cephalotes species) they
diurnal activity, arboreal habits, and extensive foraging on are well defended with spines, hard exoskeletons, and dis-
bird feces (Corn 1980; this study). Although they will oc- tasteful pheromones (see de Andrade and Baroni Urbani
casionally forage on or near the ground (especially in dis- 1999). A significant reduction in these defenses is not re-
turbed sites; Weber 1957; Corn 1980), C. atratus are pri- quired by the increased conspicuousness hypothesis but
marily arboreal, and the redness of the gasters is most must accompany fruit mimicry for it to be effective. Our
intense (to the human observer) in direct sunlight. Thus, results show that this is the case in parasitized C. atratus.
involvement of an arboreal paratenic host (i.e., an inter- Whereas birds do not regularly feed on C. atratus, other
mediate host required for completion of the parasite’s life predators, such as anteaters and lizards, could support the
cycle but in which no development of the parasite occurs) increased conspicuousness scenario (de Andrade and Bar-
is the simplest scenario. There are many logical candidates oni Urbani 1999). Arboreal ants are a major component
for paratenic hosts among frugivorous birds, including in the diets of tropidurid lizards (Vitt and Zani 1996), and
several species of honeycreepers, tanagers, and small tyran- Uracentron feces collected at the ACTS site contained
nid flycatchers (D. Robinson, personal communication; Cephalotes spp. (S. P. Yanoviak, personal observation).
also see Hosner 2005). Collectively, the reduced alarm de- However, the patchy nature of C. atratus distribution in
fenses, the weakened gaster-postpetiole junction, and the forests requires a reliable mechanism of parasite trans-
 Parasite-Induced Fruit Mimicry 543

mission to new colonies. If this is accomplished by a ver- gest that these behavioral effects are partly caused by dam-
tebrate paratenic host, birds and their feces are the most age to the ventral nerve cord in the gaster, perhaps via
efficient vehicles, given the natural history of C. atratus. mechanical pressure imposed by the parasite mass. An-
It is unlikely that lizards would disperse over great dis- terior progression of this damage as the parasites develop
tances or with sufficient frequency to infect new C. atratus in the adult ant may explain the increased sluggishness
colonies. Moreover, unlike bird feces, feces of anteaters and unstable stance of workers during the latter stages of
and arboreal lizards are relatively dry and fall to the infection.
ground, where they are likely to be consumed by litter Two-host cycles like the one described here are known
ants (Kaspari 1993) long before their discovery by C. atra- for other nematodes, and this strategy is effective for par-
tus foragers. Finally, regardless of the predator, red col- asites of sedentary, central-place foragers such as ants be-
oration (especially in combination with black) is generally cause the nematode can complete its entire life cycle near
aposematic in insects (Chapman 1998); thus, we would the nest (Schmid-Hempel 1998). Other nematode parasites
expect lower predation by generalist insectivores on in- of ants have effects somewhat similar to those described
fected C. atratus. here, including reduced defensive behaviors combined
Physiological mechanisms underlying the coordinated with lethargy (e.g., Gösswald 1930). However, we know of
physical and behavioral effects of the tetradonematid on none that combines a change in defense-based palatability
C. atratus are unknown and deserve further investigation. with fruit mimicry, nor do we know of similarly spectac-
Parasite-induced changes in appearance occur in other ant ular cases in which an insect mimic and its model occupy
taxa, although none suggests fruit mimicry. When such separate biological kingdoms (but see Roy 1993 for a strik-
color changes occur, they generally affect the whole body, ing plant-fungus relationship). Although it is likely that
not specific parts as in C. atratus (Schmid-Hempel 1998). the extensive collection of bird feces by C. atratus played
In the few examples in which pale coloration is restricted a significant role in the evolution of this association, un-
to the gaster, the effect is a by-product of gaster distention, derstanding the interplay of coordinated effects on the ant
not altered pigmentation (e.g., Carney 1969). No color awaits further study of their mechanisms. We hope that
change was observed in other ants infected with tetra- this discovery promotes additional research into other po-
donematids (Nickle and Jouvenaz 1987) or in a Cephalotes tential examples of facilitation of parasite transmission via
minutus infected by a mermithid nematode (de Andrade mimicry.
and Baroni Urbani 1999). The color change in C. atratus
may result in part from cuticular thinning (Crosland Acknowledgments
1988). This is supported by the fact that thinner areas of
exoskeleton on the frontal carinae of C. atratus are trans- We thank N. Ayana, F. Azorsa, M. Knörnschild, S. Price,
lucent orange and that another exoskeletal structure—the M. Reiskind, D. Robinson, A. Roddy, M. U. Valdez, J.
gaster-postpetiole junction—is significantly weakened by Wilson, and N. Zegarra for useful discussions, field and
the infection. Alternatively, the developing nematodes may lab assistance, and/or comments on the manuscript. O.
specifically sequester pigment compounds from the ad- Acevedo, P. Bucur, P. Jensen, L. P. Lounibos, S. Madigosky,
jacent exoskeleton without significantly compromising its and A. Morrison provided logistical support. The Smith-
structural integrity. sonian Tropical Research Institute, the Panamanian Au-
The 10% size reduction we observed in C. atratus is toridad Nacional del Ambiente (ANAM), and the Peruvian
comparable to measured HW changes in other ants that Instituto Nacional de Recursos Naturales (INRENA) pro-
are parasitized by cestodes and nematodes during larval vided permits. The Amazon Conservatory for Tropical
development (Passera 1976; Trabalon et al. 2000). Cepha- Studies, Amazon Explorama Lodges, and numerous pri-
lotes atratus workers show considerable within-colony size vate landowners facilitated access to field sites in Peru.
variation that is essentially continuous (Corn 1980). Given This research was supported in part by the National Geo-
the substantial overlap in the size distribution of infected graphic Society, the Amazon Conservation Association,
and normal workers (fig. 3), the ecological consequences and the BBC Natural History Unit.
of a 10% size reduction in 5% of the workers are prob-
ably irrelevant at the colony level. A number of factors Literature Cited
could cause the smaller body size in infected workers,
Baroni Urbani, C., and M. L. de Andrade. 1997. Pollen eating, storing,
including malnutrition, slower growth rates, or hormonal
and spitting by ants. Naturwissenschaften 84:256–258.
aberrations. Breed, M. D., T. P. McGlynn, E. Stocker, and A. N. Klein. 1999. Thief
The continuous gaster flagging and erect stance of par- workers and variation in nestmate recognition behavior in a
asitized C. atratus workers have not been documented for ponerine ant, Ectatomma ruidum. Insectes Sociaux 46:327–333.
other ants infected with parasites. Our observations sug- Carney, W. P. 1969. Behavioral and morphological changes in car-
544 The American Naturalist

penter ants harboring dicrocoeliid metacercariae. American Mid- Nickle, W. R., and D. P. Jouvenaz. 1987. Tetradonema solenopsis n.
land Naturalist 82:605–611. sp. (Nematoda: Tetradonematidae) parasitic on the red imported
———. 1970. Brachylecithum mosquensis: infections in vertebrate, fire ant Solenopsis invicta Buren from Brazil. Journal of Nematology
molluscan and arthropod hosts. Transactions of the American Mi- 19:311–313.
croscopical Society 89:233–250. Passera, L. 1976. Origine des intercastes dans les sociétés de Pheidole
Chapman, A., and K. V. Rosenberg. 1991. Diets of four sympatric pallidula (Nyl.) (Hymenoptera, Formicidae) parasitées par Mermis
Amazonian woodcreepers (Dendrocolaptidae). Condor 93:904– sp. (Nematoda, Mermithidae). Insectes Sociaux 23:559–575.
915. Peloquin, J. J., and E. G. Platzer. 1993. Control of root gnats (Sciar-
Chapman, R. F. 1998. The insects: structure and function. 4th ed. idae: Diptera) by Tetradonema plicans Hungerford (Tetradone-
Cambridge University Press, Cambridge. matidae: Nematoda) produced by a novel culture method. Journal
Corn, M. L. 1980. Polymorphism and polyethism in the Neotropical of Invertebrate Pathology 62:79–86.
ant Cephalotes atratus (L.). Insectes Sociaux 27:29–42. Poinar, G., Jr., and S. P. Yanoviak. 2008. Myrmeconema neotropicum
Crosland, M. W. J. 1988. Effect of a gregarine parasite on the color n. g., n. sp., a new tetradonematid nematode parasitising South
of Myrmecia pilosula (Hymenoptera: Formicidae). Annals of the American populations of Cephalotes atratus (Hymenoptera: For-
Entomological Society of America 81:481–484. micidae), with the discovery of an apparent parasite-induced host
Curio, E. 1976. The ethology of predation. Springer, Berlin. morph. Systematic Parasitology 69:145–153.
de Andrade, M. L., and C. Baroni Urbani. 1999. Diversity and ad- Roche, R. K., and D. E. Wheeler. 1997. Morphological specializations
aptation in the ant genus Cephalotes, past and present. Stuttgarter of the digestive tract of Zacryptocerus rohweri (Hymenoptera: For-
Beiträge zur Naturkunde, Serie B (Geologie und Paläontologie), micidae). Journal of Morphology 234:253–262.
no. 271. Stuttgart Museum of Natural History, Stuttgart. Roy, B. A. 1993. Floral mimicry by a plant pathogen. Nature 362:
Gösswald, K. 1930. Weitere Beiträge zur Verbreitung der Mermithi- 56–58.
den bei Ameisen. Zoologischer Anzeiger 90:13–27. SAS Institute. 2002. SAS OnlineDoc 9. SAS Institute, Cary, NC.
Heinze, J., O. Rüppell, S. Foitzik, and A. Buschinger. 1998. First Schmid-Hempel, P. 1998. Parasites in social insects. Princeton Uni-
records of Leptothorax rugatulus (Hymenoptera: Formicidae) with versity Press, Princeton, NJ.
cysticercoids of tapeworms (Cestoda: Dilepididae) from the south- Sheiman, I. M., M. F. Shkutin, N. B. Terenina, and M. K. S. Gus-
western United States. Florida Entomologist 81:122–125. tafsson. 2006. A behavioral study of the beetle Tenebrio molitor
Holmes, J. C., and W. M. Bethel. 1972. Modification of intermediate infected with cysticercoids of the rat tapeworm Hymenolepis dim-
host behavior by parasites. Pages 123–149 in E. U. Canning and inuta. Naturwissenschaften 93:305–308.
C. A. Wright, eds. Behavioral aspects of parasite transmission. Sokal, R. R., and F. J. Rohlf. 1995. Biometry. 3rd ed. W. H. Freeman,
Academic Press, London. New York.
Hosner, P. A. 2005. Regurgitated mistletoe seeds in the nest of the Tang, Y., R. Killick-Kendrick, and W. M. Hominick. 1997. Life cycle
yellow-crowned tyrannulet (Tyrannulus elatus). Wilson Bulletin of Didilia ooglypta (Nematoda: Tetradonematidae), a parasite of
117:319–321. phlebotomine sandflies of Afghanistan. Nematologica 43:491–503.
Hurd, H. 1990. Physiological and behavioral interactions between Trabalon, M., L. Plateaux, L. Péru, A.-G. Bagnères, and N. Hartmann.
parasites and invertebrate hosts. Advances in Parasitology 29:271– 2000. Modification of morphological characters and cuticular
318. compounds in worker ants Leptothorax nylanderi induced by en-
Kaspari, M. 1993. Removal of seeds from Neotropical frugivore drop- doparasites Anomotaenia brevis. Journal of Insect Physiology 46:
pings: ant responses to seed number. Oecologia (Berlin) 95:81– 169–178.
88. Vásquez Martı́nez, R. 1997. Flórula de las reservas biológicas de
Kaspari, M., and M. D. Weiser. 1999. The size-grain hypothesis and Iquitos, Perú. Monographs in systematic botany, vol. 63. Missouri
interspecific scaling in ants. Functional Ecology 13:530–538. Botanical Garden, St. Louis.
Leigh, E. G., Jr., A. S. Rand, and D. M. Windsor, eds. 1996. The Vitt, L. J., and P. A. Zani. 1996. Ecology of the elusive tropical lizard
ecology of a tropical forest. 2nd ed. Smithsonian Institution, Wash- Tropidurus [pUracentron] flaviceps (Tropiduridae) in lowland rain
ington, DC. forest of Ecuador. Herpetologica 52:121–132.
Lindström, L., R. V. Alatalo, A. Lyytinen, and J. Mappes. 2001. Strong Weber, N. A. 1957. The nest of an anomalous colony of the arboreal
antiapostatic selection against novel rare aposematic prey. Pro- ant Cephalotes atratus. Psyche 64:60–69.
ceedings of the National Academy of Sciences of the USA 98:9181– Wheeler, W. M. 1907. The polymorphism of ants, with an account
9184. of some singular abnormalities due to parasitism. Bulletin of the
Madigosky, S. R., and I. Vatnick. 2000. Microclimatic characteristics American Museum of Natural History 23:1–93.
of a primary tropical Amazonian rain forest, ACEER, Iquitos, Peru. Yanoviak, S. P., and R. Dudley. 2006. The role of visual cues in directed
Selbyana 21:165–172. aerial descent of Cephalotes atratus workers (Hymenoptera: For-
Mäki, S., R. Kalliola, and K. Vuorinen. 2001. Road construction in micidae). Journal of Experimental Biology 209:1777–1783.
the Peruvian Amazon: causes and consequences. Environmental Zervos, S. 1980. Bispiculum inequale n. gen. & sp. (Nematoda:
Conservation 28:199–214. Tetradonematidae) from New Zealand wood-boring beetles (Cur-
Moore, J. 1995. The behavior of parasitized animals. BioScience 45: culionidae: Platypodinae). New Zealand Journal of Zoology 7:
89–96. 155–164.
———. 2002. Parasites and the behavior of animals. Oxford Uni-
versity Press, Oxford.
Muir, D. A. 1954. Ants Myrmica rubra L. and M. scabrinodis Nyl. as
intermediate hosts of a cestode. Nature 173:688–689. Natural History Editor: Henry M. Wilbur