Genetic royal cheats in leaf-cutting ant societies
William O. H. Hughes*†‡and Jacobus J. Boomsma*
*Centre for Social Evolution, Department of Biology, University of Copenhagen, Universitetsparken 15, 2100 Copenhagen, Denmark; and†Institute of
Integrative and Comparative Biology, University of Leeds, Leeds LS2 9JT, United Kingdom
Edited by Bert Ho ¨lldobler, Arizona State University, Tempe, AZ, and approved January 24, 2008 (received for review October 29, 2007)
Social groups are vulnerable to cheating because the reproduc-
tive interests of group members are rarely identical. All coop-
erative systems are therefore predicted to involve a mix of
cooperative and cheating genotypes, with the frequency of the
latter being constrained by the suppressive abilities of the
former. The most significant potential conflict in social insect
colonies is over which individuals become reproductive queens
rather than sterile workers. This reproductive division of labor
is a defining characteristic of eusocial societies, but individual
larvae will maximize their fitness by becoming queens whereas
their nestmates will generally maximize fitness by forcing larvae
to become workers. However, evolutionary constraints are
thought to prevent cheating by removing genetic variation in
caste propensity. Here, we show that one-fifth of leaf-cutting
ant patrilines cheat their nestmates by biasing their larval
development toward becoming queens rather than workers.
Two distinct mechanisms appear to be involved, one most
probably involving a general tendency to become a larger adult
and the other relating specifically to the queen–worker devel-
opmental switch. Just as evolutionary theory predicts, these
‘‘royal’’ genotypes are rare both in the population and within
individual colonies. The rarity of royal cheats is best explained
as an evolutionary strategy to avoid suppression by cooperative
genotypes, the efficiency of which is frequency-dependent. The
results demonstrate that cheating can be widespread in even the
most cooperative of societies and illustrate that identical
principles govern social evolution in highly diverse systems.
caste ? polymorphism ? social evolution ? polyandry ? social insect
members are clones, their reproductive interests will differ,
and individuals may benefit by exploiting the cooperative
efforts of other group members (1). Genetic polymorphism for
cheating is therefore predicted to arise in all nonclonal social
systems (2, 3) and is particularly well characterized in the social
microbes (4–6). However, such genetic cheating does not
appear to occur in one of the classic examples of cooperation,
the development of social insect larvae into either reproduc-
tive queens or sterile workers. This lack of genetic cheating is
thought to be because queens and workers normally differ
morphologically, the caste destiny of larvae is determined by
environmental cues, and these cues are controlled by adult
workers that are unable to bias larval development nepotisti-
cally (7–10). The only known exception to this lack of genetic
variation for queen–worker developmental propensity is re-
stricted to artificially induced emergency queen production in
honeybees (11–14), queens of which are normally determined
solely by being fed a special diet of ‘‘royal jelly.’’ However,
whether constraints prevent genetic variation for caste destiny
under normal conditions, or whether, as predicted by evolu-
tionary theory, it in fact occurs but is rare and thus hard to
detect, is unknown.
The leaf-cutting ant Acromyrmex echinatior is particularly
suitable for examining whether rare, cheating lineages occur
because queens mate with multiple (haploid) males (polyan-
dry) (15). Their worker and queen offspring (diploid females)
thus consist of a number of full-sister lineages (patrilines) that
lthough social groups most obviously exhibit cooperation,
they are also often the scene of conflict. Unless group
are half-sisters toward each other. Individuals of different
patrilines within colonies share the same rearing conditions
and, when colonies are headed by a single mother queen
(monogyny), the same maternal genotype on average. They
therefore differ only in their paternal genotype, which enables
the ready detection of genetic variation. Patrilines of A.
echinatior are known to differ in their propensities to develop
into the two main worker castes [large workers (LW) or small
workers (SW)] (16, 17), which may improve the efficiency of
colony division of labor (16–20). To establish whether royal
patrilines occur, we sampled workers and daughter queens
(gynes sensu stricto) from five mature, monogynous colonies
of A. echinatior. We then genotyped them at four polymorphic
microsatellite loci and assigned them to the different patrilines
within each colony.
Results and Discussion
Significant differences in patriline representation between
queens and workers were found in three colonies (Fig. 1). The
result was not due solely to very rare patrilines (Fig. 1) or to
false discoveries resulting from multiple tests [supporting
information (SI) Table 1]. It was also not due to temporal
changes in sperm use because this does not seem to occur in
mature colonies of A. echinatior (21) and because the workers
and queens were in any case from the same age cohort. There
is therefore a significant genetic influence on queen–worker
caste determination in A. echinatior. Unlike cases where
hybridization or parthenogenesis affect queen–worker deter-
mination (22, 23), the genetic influence on caste fate in A.
echinatior represents standard genetic variation and, unlike the
only previous example in the honeybee (11–14), one that is
expressed under normal conditions. In addition, the variation
is expressed while adult workers have the opportunity to
influence larval caste fate, unlike the case of stingless bees
where some larvae developing in sealed cells selfishly alter
their caste destiny (7, 24, 25). Importantly, most of the
queen-biased patrilines in A. echinatior also produced at least
some workers (Fig. 1), so the genetic influence appears to
involve royal genotypes being predisposed to become queens,
rather than their destiny being fixed.
The caste system of leaf-cutting ants allows an unusually direct
insight into the proximate mechanisms responsible for genetic
variation for royalty. Queens, LW, and SW represent distinct size
categories (Fig. 1), and, as found previously (16, 17), colonies also
showed genetic variation between patrilines in their propensity to
develop into LW or SW (SI Table 2). A positive, across-patriline
indicate a caste-biasing mechanism based on an intrinsic ability to
develop into a larger adult. A lack of relationship between the two
Author contributions: W.O.H.H. and J.J.B. designed research; W.O.H.H. performed re-
search; W.O.H.H. analyzed data; and W.O.H.H. and J.J.B. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
‡To whom correspondence should be addressed. E-mail: email@example.com.
This article contains supporting information online at www.pnas.org/cgi/content/full/
© 2008 by The National Academy of Sciences of the USA
April 1, 2008 ?
vol. 105 ?
no. 13 www.pnas.org?cgi?doi?10.1073?pnas.0710262105
skews would imply that the genetic influence on royalty is de-
coupled from that on worker caste determination. In fact, the two
? 0.04x ? 0.15, radj
? 0.45; F2,26? 10.6, P ? 0.0004; Fig. 2) rather
than linear function (F1,27? 2.2, P ? 0.149, radj
overrepresented in LW were disproportionately likely to develop
into queens, but the same was also true for patrilines that were
overrepresented in SW.
The concave relationship between queen–worker skew and
LW–SW skew (Fig. 2a) indicates that two separate mecha-
nisms are involved in royalty propensity and that both are
? 0.04). Patrilines
linked to worker caste determination. The excess queen pro-
duction of the patrilines that were also overrepresented in LW
rather than SW can be explained relatively easily by a direct
genetic influence on adult size. Such an effect would be
produced if larvae have a lower response threshold for the
environmental cues, such as nutrition, that act at specific
thresholds to program the developmental trajectories of larvae
(26, 27). Larvae of royal–LW patrilines would then be more
likely to develop into a larger adult, i.e., into queens rather
than workers and into LW rather than SW (Fig. 3a). The
relationship between queen-bias and overrepresentation
among SW is most probably caused by a similar effect but
specifically relating to the queen–worker developmental
switch (Fig. 3a). Larvae that receive environmental cues
appropriate for SW development become SW as normal, but
larvae that receive cues appropriate for LW development tend
to become queens (Fig. 3a). The actual caste proportions
observed for the three types of patrilines fit these proposed
mechanisms (Fig. 3b), the principles of which will also hold for
alternative sequences of caste determination to those modeled
here (e.g., if queen–worker determination occurs early in
development and LW–SW determination occurs later) (27).
s - r e
k r o
s - n
small worker-skewedlarge worker-skewed
Worker caste skew
e l p
q f o e r a
v i t a l e
SEM direct reproductive success of patrilines (share of queen sample relative
to share of worker sample, with 0 indicating that the shares were equal)
grouped into six royal–small worker (SW) patrilines (above and left of the
and right of the regression line x-intercept in a), and the 19 normal unbiased
Relationship between worker caste determination and royalty. (a)
46 51 72 1 3
70 73 12 8
11 110 29 5 8
e l p
s f o n
o i t r o
12 27 104 29 30 54 8 6
5 20 15 52 14 52 127 37
photo shows the three female castes: small worker, large worker, and queen.
Data presented are proportions of individuals sampled per patriline for five
colonies that were workers (clear) or queens (colored). The significance of
Fisher’s (Top bars) and G tests (Bottom bars) are shown testing the null
0.05; ns, P ? 0.05). Sample sizes are given above the columns and are circled
when the individual tests for a patriline showed that its ratio differed signif-
icantly from the overall ratio for the colony. In the figure, the worker pro-
portions have been adjusted for the actual numbers of individuals genotyped
in each colony such that a proportion of 0.5 is expected for each caste in all
Hughes and BoomsmaPNAS ?
April 1, 2008 ?
vol. 105 ?
no. 13 ?
Although both royalty-biasing mechanisms most probably
relate to changes in response to caste-determining environ-
mental cues, the contrast between them is important. The
mechanism for royal–LW patrilines involves a change in
worker caste propensity as well as queen–worker propensity.
The trait may therefore plausibly have evolved because of the
cooperative benefit of its effect on worker caste determination
(16–19), with any selfish benefit from royalty biasing being a
pleiotropic side effect or later evolutionary development.
However, the change for royal–SW patrilines only appears to
affect queen–worker determination and is thus most likely to
have arisen primarily for selfish benefits. This distinction is
bolstered by the relative direct fitness gained by royal–LW and
royal–SW patrilines: the former gain very little direct fitness
compared with normal patrilines, whereas the direct fitness of
the latter is enhanced by almost 500% (Kruskal–Wallis test:
that the genetic disposition shown by royal–SW has evolved for
selfish reasons and represents a genuine case of cheating.
Further work will be needed to resolve whether genetic
variation for cheating traits is discrete or quantitative, as well
as whether effects of dominance and epistasis make the degree
of expression of cheating genes different depending on the
other maternal and paternal genes that are present in a colony.
In addition to predicting that cooperative systems should be
vulnerable to cheating, evolutionary theory also predicts that
suppression by cooperators will constrain the frequency of
royal cheats (2, 3). Our results suggest that this prediction is
true for A. echinatior. At the population level, significant
cheating was limited to 20% of patrilines (6 of 30) in our
sample of five colonies. In addition, royal cheats were also rare
at the intracolony level (Fig. 4a). Most patrilines produced
?20% of the individuals sampled in their colony, consistent
with the average effective mating frequency for A. echinatior
of 5.3 (15). Royal–LW patrilines had similar frequencies (Fig.
4a), but royal–SW patrilines were significantly rarer (Kruskal–
Wallis test: ?2
they are constrained by frequency-dependent selection. As in
some other systems (4, 6), this selection may result from
2? 18.4, P ? 0.0001; Fig. 2b). It therefore seems most likely
2? 6.89, P ? 0.032; Fig. 4a), which suggests that
group-level efficiency costs or the direct suppression of cheat-
ing larvae by cooperative genotypes. Alternatively, the rarity
of cheats may be an evolutionary strategy to avoid suppression.
Ant larvae signal queen potential at a normally distributed
critical size, and adult workers then confirm this potential by
feeding the larvae appropriate diet if they assess their signal
and size to be congruent with the colony norm (26, 27). Unlike
some other social insects (7, 24, 25), workers are therefore
potentially able to ignore or eliminate cheating larvae if they
detect them, and royal–SW patrilines are especially likely to
stand out from the colony norm. Being rare in any particular
colony may make them harder to distinguish from the left tail
of the size distribution of cooperative queen-potential larvae
(Fig. 4b). The within-colony rarity of royal–SW patrilines is
further confirmation that the trait has evolved primarily for
selfish reasons. There is no obvious reason why royal genes that
have arisen for cooperative benefits should also be selected to
be rare, whereas rarity is exactly what evolutionary theory
predicts for royal cheats (2, 3).
Our findings underline that social evolution in very different
taxonomic groups can be understood from the same general
principles. The nonidentical reproductive interests of group
members inevitably result in individual-level selection favoring
cheating and the antagonistic coevolution of cheat suppression
(1). In systems as diverse as selfish genetic elements (28),
worker reproduction in social insect colonies (29) and pun-
ishment in human societies (30), suppression by cooperators
can prevent cheats being common but cannot eliminate them
altogether. Our data confirm this theory for royalty biasing in
the leaf-cutting ant A. echinatior. As our present results show,
units of environmental cue
proportional make-up of sample
determination. Larvae of normal genotypes receiving 1 arbitrary unit of the
caste-determining environmental cue (e.g., food) switch from the small
larvae receiving 2 units switch from the worker to queen developmental path
(solid line). Royal–LW patrilines have lower thresholds for both switches.
(b) Mean ? SEM proportion of individuals sampled that were SW, LW, and
normal unbiased patrilines.
Mechanisms of royalty biasing. (a) Proposed mechanisms of caste
e l p
s s l a
d i v i d
n i f o n
o i t r o
Larval size at queen-worker developmental switch
n i d
e r s l a
d i v i d
n i f o n
o i t r o
colonies of the six royal–small worker (SW) patrilines, the four royal–large
worker (LW) patrilines, and the 19 normal unbiased patrilines. (b) Proposed
suppression mechanism that induces rarity. Larvae in normal patrilines (gray)
signal queen potential at a normally distributed critical size (arbitrary units)
with royal patrilines having slightly (royal–LW, blue) or substantially (royal–
SW, red) lower means. Royal–SW patrilines are 75% rarer than other patrili-
nes, which makes their queen-biased larvae harder for adult workers to
distinguish from the distribution tail for normal patrilines and allows them to
www.pnas.org?cgi?doi?10.1073?pnas.0710262105 Hughes and Boomsma
cheating at low frequencies can occur in even the most tightly Download full-text
integrated social systems and may thus be more common than
Materials and Methods
Workers and virgin queens (gynes sensu stricto) were sampled from five
mature colonies of A. echinatior that had been collected from Gamboa,
Panama. For each colony, workers and queens of similar cuticular colora-
tion were sampled on the same day to minimize age variation. From four
of the colonies (Ae125, Ae129, Ae153, and Ae158), 94 queens and 94
workers (split equally between SW and LW) were sampled. From the fifth
colony (Ae48), 94 queens were collected, and a sample of 200 workers from
the same collection date has been analyzed previously (16). Queens and
workers were genotyped at four polymorphic microsatellite loci: Ech1390,
with internal size markers. Multilocus offspring genotypes were used to
infer the genotypes of colony queens and their multiple mates, and these
genotypes were confirmed by using the program MATESOFT, version 1.0
(32). It was then possible to assign the sampled queens and workers to
patrilines within their colony with negligible detection errors of patrilines.
Individuals whose paternities could not be established because of failed
PCR amplification or being heterozygous with the same alleles as a het-
erozygous mother queen at one or more diagnostic loci were excluded
from the analysis (84 individuals, 7% of total). We determined whether
patrilines differed in their ratios of queens to workers by: (i) Fisher’s exact
tests, in which all patrilines were included (33); and (ii) G tests of hetero-
geneity, in which the rarest patrilines were excluded such that no more
than 20% of cells in the overall analysis for any particular colony had
expected frequencies of less than five (34). Both methods examined
whether patrilines differed from the total number of queens:workers
genotyped for each colony in a uniform direction. Because the analyses
involved multiple tests, we controlled for the false-discovery rate by using
the program QVALUE (35). Additional worker samples were genotyped for
two highly queen-biased patrilines.
ACKNOWLEDGMENTS. We are grateful to the Smithsonian Tropical Re-
search Institute for providing facilities in Gamboa, the Autoridad Nacional
del Ambiente (ANAM) for permission to collect and export the ants to
Denmark; D. R. Nash for providing the photograph in Fig. 1; S. Mathiasen
for technical assistance; and members of the Social Insect Laboratory,
University of Sydney, and Centre for Social Evolution, University of Copen-
hagen, for comments on earlier versions of the manuscript. This work was
supported by grants from the Carlsberg Foundation (to W.O.H.H.), the
Foundation (to J.J.B.).
2. Keller L, ed (1999) Levels of Selection in Evolution (Princeton Univ Press, Princeton).
3. Frank SA (1998) Foundations of Social Evolution (Princeton Univ Press, Princeton).
4. Ross-Gillespie A, Gardner A, West SA, Griffin AS (2007) Frequency dependence and
cooperation: Theory and a test with bacteria. Am Nat 170:331–342.
5. Strassmann JE, Zhu Y, Queller DC (2000) Altruism and social cheating in the social
amoeba Dictyostelium discoideum. Nature 408:965–967.
6. Gilbert OM, Foster KR, Mehdiabadi NJ, Strassmann JE, Queller DC (2007) High relat-
edness maintains multicellular cooperation in a social amoeba by controlling cheater
mutants. Proc Natl Acad Sci USA 104:8913–8917.
7. Bourke AFG, Ratnieks FLW (1999) Kin conflict over caste determination in social
Hymenoptera. Behav Ecol Sociobiol 46:287–297.
Rev Entomol 51:581–608.
9. West SA, Griffin AS, Gardner A (2007) Evolutionary explanations for cooperation. Curr
10. Boomsma JJ, et al. (2003) Informational constraints on optimal sex allocation in ants.
Proc Natl Acad Sci USA 100:8799–8804.
11. Cha ˆline N, Arnold G, Papin C, Ratnieks FLW (2003) Patriline differences in emergency
queen rearing in the honeybee, Apis mellifera. Insect Soc 50:234–236.
12. Osborne KE, Oldroyd BP (1999) Possible causes of reproductive dominance during
emergency queen rearing by honeybees. Anim Behav 58:267–272.
13. Tilley CA, Oldroyd BP (1997) Unequal subfamily proportions among honeybee queen
and worker brood. Anim Behav 54:1483–1490.
14. Moritz RFA, et al. (2005) Rare royal families in honeybees, Apis mellifera. Naturwis-
mating singly. Nature 428:35–36.
16. Hughes WOH, Sumner S, Van Borm S, Boomsma JJ (2003) Worker caste polymorphism
has a genetic basis in Acromyrmex leaf-cutting ants. Proc Natl Acad Sci USA 100:9394–
17. Hughes WOH, Boomsma JJ (2007) Genetic polymorphism in leaf-cutting ants is phe-
notypically plastic. Proc R Soc London B 274:1625–1630.
18. Waibel M, Floreano D, Magnenat S, Keller L (2006) Division of labor and colony
kin structure, and rate of perturbations. Proc R Soc London B 273:1815–1823.
Trends Ecol Evol 22:408–413.
20. Mattila HR, Seeley TD (2007) Genetic diversity in honeybee colonies enhances produc-
tivity and fitness. Science 317:362–364.
21. Bekkevold D, Frydenberg J, Boomsma JJ (1999) Multiple mating and facultative po-
lygyny in the Panamanian leafcutter ant Acromyrmex echinatior. Behav Ecol Sociobiol
22. Nonacs P (2006) The ecology and evolution of hybridization in ants. Ecology 87:2141–
23. Sykes E, West SA (2005) Evolution: Revenge of the clones. Curr Biol 15:R547–R549.
24. Wenseleers T, Ratnieks FLW, Ribeiro MdF, Alves DdA, Imperatriz-Fonseca VL (2005)
Working-class royalty: Bees beat the caste system. Biol Lett 1:125–128.
25. Wenseleers T, Ratnieks FLW (2004) Tragedy of the commons in Melipona bees. Proc R
Soc London B 271:S310–S312.
26. Ho ¨lldobler B, Wilson EO (1990) The Ants (Belknap, Cambridge, MA).
28. Burt A, Trivers RL (2006) Genes in Conflict (Belknap, Cambridge, MA).
30. Fehr E, Gachter S (2002) Altruistic punishment in humans. Nature 415:137.
31. Hughes WOH, Boomsma JJ (2006) Does genetic diversity hinder parasite evolution in
social insect colonies? J Evol Biol 19:132–143.
32. Moilanen A, Sundstrom L, Pedersen JS (2004) MATESOFT: A program for deducing
parental genotypes and estimating mating system statistics in haplodiploid species.
Mol Ecol Notes 4:795–797.
33. Mehta CR, Patel NR (1996) SPSS Exact Tests for Windows (SPSS, Chicago), Version 7.0.
34. Sokal RR, Rohlf FJ (1995) Biometry (Freeman, New York).
35. Storey JD, Tibshirani R (2003) Statistical significance for genomewide studies Proc Natl
Acad Sci USA 100:9440–9445.
Hughes and BoomsmaPNAS ?
April 1, 2008 ?
vol. 105 ?
no. 13 ?