Genetic determination of the queen caste in an ant

hybrid zone
Glennis E. Julian*†, Jennifer H. Fewell*, Jürgen Gadau‡, Robert A. Johnson*, and Debbie Larrabee*
*Department of Biology, Arizona State University, Tempe, AZ 85287; and ‡Department of Behavioral Physiology and Sociobiology, Biozentrum,
University of Würzburg, Würzburg D-97074, Germany

Communicated by Bert Hölldobler, University of Würzburg, Würzburg, Germany, April 12, 2002 (received for review February 24, 2002)

The question of how reproductives and sterile workers differen- ants, Pogonomyrmex rugosus and Pogonomyrmex barbatus, which
tiate within eusocial groups has long been a core issue in socio- seems to be associated with hybridization. These two sister
biology because it requires the loss of individual direct fitness in species are widespread in deserts and grasslands of the south-
favor of indirect or group-level fitness gains. The evolution of central and southwestern United States and northern Mexico
social behavior requires that differentiation between workers and (15). Of the two species, P. barbatus has a more eastern
female reproductives be environmentally determined, because distribution, occurring from western Louisiana to central Ari-
genetically determined sterility would be quickly eliminated. Nev- zona, whereas P. rugosus occurs from central Texas to California.
ertheless, we report clear evidence of genetic caste determination The two species have broadly overlapping geographic ranges
in populations of two seed harvester ant species common to the from central and southeastern Arizona to western Texas, but
southwestern USA, Pogonomyrmex rugosus and Pogonomyrmex both species also inhabit large areas of allopatry (15).
barbatus. The genetic differentiation between workers and queens We used randomly amplified polymorphic DNA (RAPD)
is found only in areas of sympatry of the two species, and thus genetic markers to examine genotypic patterns among sibling
appears to arisen from hybridization. Our data suggest that this workers, alate queens (alates), and males from colonies of P.
hybridization has had a profound historical effect on the caste rugosus and P. barbatus in areas of sympatry and allopatry.
determination systems and mating patterns of each of these Morphological data suggest that these two ant species hybridize
species. in areas of overlap (16), and both species often possess mtDNA
of the sister species in these areas (17). Mating aggregations of
both species within contact zones have been observed to contain
A n overriding principle of social insect biology is that the
determination of reproductive (queens) versus sterile
(workers) individuals within a social group is primarily environ-
low numbers of congeneric reproductives, supporting the pos-
sibility of hybridization (18).
mentally governed (1–3). A sterile caste can evolve under kin Methods
selection only if the genes for sterility are expressed condition-
We collected workers, alates, and males from colonies of P.
ally, because any allele that invariably caused sterility could
rugosus and P. barbatus during the mating flight season (July
not be passed on and would be quickly eliminated from the
through September) in areas of sympatry and allopatry. Ants
population (4, 5).
were immediately frozen at ⫺80°C or placed in 100% ethanol.
Numerous studies have accumulated evidence that differences
Both P. barbatus and P. rugosus have one reproductive queen per
in larval environment, particularly larval nutrition, determine
colony, and queens of both species mate with multiple males
whether a female egg will develop into a fully capable repro- (18). Thus, all individuals collected from a given colony were

EVOLUTION
ductive or a small worker female (6, 7). When a nutritional or offspring of the same queen.
other physiological threshold is reached during larval develop- Our sympatic site was in southeastern Arizona (Cochise
ment, neurosecretory changes result in elevated juvenile hor- County) and southwestern New Mexico (Hidalgo County),
mone levels, triggering development of reproductive potential where we collected five colonies of P. rugosus and nine colonies
(8). Recent research has focused on identifying genes involved in of P. barbatus. Our sites for allopatric samples were in Texas and
the developmental cascade underlying this polyphenism (9). Arizona, where we collected 17 colonies of P. rugosus and seven
However, the hypothesis that the underlying mechanism for colonies of P. barbatus. Colonies of P. rugosus were collected in
queen-worker caste differentiation could itself become geneti- Maricopa County (six colonies), Pinal County (five colonies),
cally based lacks support. and Navajo County (six colonies), Arizona, whereas samples of
There have been rare reported exceptions of genetic influence P. barbatus were collected in Yavapai County (three colonies),
on caste determination; however, these have involved the gen- Arizona, and Tarrant County (three colonies) and Wichita
eration of different queen types in ants with polymorphic County (one colony), Texas.
queens, rather than the fundamental differentiation of queens DNA was extracted from each individual according to Landry
and workers. Allelic differences between regular, winged queens et al. (19) or Gadau (20); abdomens of workers and alates were
and an intermediate, wingless queen have been described in the removed before extraction. The RAPD PCR was performed
slave maker ant, Harpegoxenus sublaevis (10), and a genetically according to Williams et al. (21). We screened an initial colony
based queen polymorphism also has been demonstrated in a of workers, alates, and males of each species with 45 different
species of an Australian ant, Monomorium (11). A third case, 10-base random primers. We then chose a 10-bp primer OPC9
using allozymes as genetic markers, has demonstrated that a (Operon Technologies, Alameda, CA) showing a fragment-
second queen morph in Acanthomyops is actually a genotype length polymorphism that was useful for identifying heterozy-
resulting from hybridization (12). In Melipona bees, it has been gotes (22, 23). All colonies were then screened with the C9
suggested that queen determination is in part genetically con- primer. Heterozygous individuals displayed three different
trolled by double heterozygosity at two independent loci (13). bands, one at 510 kb, one at 550 kb, and a heteroduplex band at
However, this example is highly controversial (14), and the 590 kb. We verified that the marker was a codominant fragment-
environmental influence on caste remains strong because all
genotypes become workers if given insufficient food.
This article presents clear evidence for genetic queen-worker Abbreviation: RAPD, randomly amplified polymorphic DNA.
caste determination in populations of two species of harvester †To whom reprint requests should be addressed. E-mail: [email protected].

www.pnas.org兾cgi兾doi兾10.1073兾pnas.112222099 PNAS 兩 June 11, 2002 兩 vol. 99 兩 no. 12 兩 8157– 8160

Table 1. Results of presence兾absence RAPD markers among sibling alate queens, workers, and males from a single colony of P.
barbatus and P. rugosus
Alate queens Workers Males

P. barbatus (Colony 5)
Primer (size) 1 2 3 4 5 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5
C9 (600) 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0
R11 (300) 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0
R11 (400) 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0
D18 (490) 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0
S19 (1000) 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0
G4 (650) 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0
Q9 (850) 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0
R11 (600) 0 0 0 1 0 0 1 1 0 0 0 0 1 1 0 1 0 0 0 1
Q9 (450) 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 0 1
P. rugosus (Colony D)
Primer (size) 1 2 3 4 5 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5
C9 (600) 1 0 1 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0
P13 (630) 1 0 1 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0
S8 (1300) 1 0 1 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0
T20 (750) 1 0 1 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0
N11 (650) 1 0 1 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0
K4 (300) 1 0 1 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0
D18 (495) 0 0 0 0 1 0 1 0 0 1 1 0 0 0 0 0 1 0 0 1
A1 (800) — 0 0 1 0 0 0 1 0 — 1 0 1 0 0 1 0 0 1 1

1 ⫽ band present, 0 ⫽ band absent.

length polymorphism by mixing PCR-amplified DNA products locus, in which genotypic frequencies were similar for alates and
from males of each alternative allele, heating the samples, and workers (Table 2). As a result, the proportion of heterozygotes
running them on a gel, which allows reannealing of the hetero- differed significantly between the sympatric and allopatric
duplex. All three bands appeared, indistinguishable from a groups (Fig. 1). In P. barbatus within the contact zone, heterozy-
heterozygous individual. gosity was completely associated with caste, whereas in the
allopatric population heterozygosity was not different between
Results alates and workers (Fig. 1). Heterozygosity levels in allopatric P.
We found distinct intra-colonial differences between the geno- rugosus were low for both alates and workers, because the 550
types of workers and their reproductive siblings in colonies allele was rare. However, heterozygosity was not associated with
within the contact zone. In a sample colony of each species, 13 caste (Table 2), and some colonies contained both a heterozy-
primers revealed 16 variable markers; 12 displayed complete gous queen and heterozygous daughter alates.
linkage with the worker caste (Table 1). Specifically, workers
displayed bands that were absent in their alate queen sisters or
in the males. The most parsimonious explanation for these
genotype differences is that the workers and queens come from Table 2. Genotypes revealed by RAPD genetic marker C9 for
different patrilines. Because Hymenoptera are haplo-diploid, alate queens, workers, and males from sympatric and allopatric
workers from the same father share all markers specific to that populations of P. rugosus and P. barbatus
patriline. Males are produced parthenogenetically, and their
Homozygotes Heterozygotes Homozygotes
genotypes reveal the queen’s genotype. Therefore, markers
Caste 510兾510 510兾550 550兾550 Total
absent in males but present in workers are inherited patrilinially.
Not all markers showed this distinct caste difference; four P. rugosus (sympatric) (n ⫽ 5 colonies)
markers showed variation independent of caste and sex (Table Queens 15 4 5 21
1). Most likely the queens are heterozygous (present兾absent) for Workers 0 33 0 33
these markers and the genotypes of their mates are absent, Males 17 — 12 29
because males—representing the queens genotype—also segre- P. rugosus (allopatric) (n ⫽ 17 colonies)
gated for all of these markers. Queens 41 2 0 43
In addition, we found distinct differences between alates and Workers 56 5 0 61
workers by using a fragment-length polymorphism (C9–550兾 Males 46 — 1 46
510) for colonies within the contact zone. All workers in the P. barbatus (sympatric) (n ⫽ 9 colonies)
sympatric populations of both P. rugosus and P. barbatus were Queens 29 0 38 67
heterozygous for the C9 marker (Table 2, Fig. 1). In contrast, Workers 0 60 0 60
their alate sisters were significantly more likely to be homozy- Males 37 — 30 67
gous for one of the two alleles. All P. barbatus alates (n ⫽ 67) and P. barbatus (allopatric) (n ⫽ 7 colonies)
80% of P. rugosus alates (n ⫽ 21) were homozygous. All males Queens 4 6 5 15
within a given colony shared the same allele, indicating that the Workers 12 11 13 36
queens of colonies in the sympatric zone are all homozygous. Males 4 — 5 9
This allele matched the one present in their sibling alates. Genotypes were assigned based on the fragment-length polymorphic
The allopatric populations of both species showed a more marker that revealed two alleles of different lengths, one at 510 bp, the other
typical pattern of within-colony genetic variation for the C9 at 550 bp. Males are hemizygous and therefore possess only one allele.

8158 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.112222099 Julian et al.

 across successive generations. Because any successful colony
must produce both workers and alate queens, this system has
become a case of obligate polyandry. Note that queens who
mated only with males of the opposite type can still successfully
found colonies (produce workers) but should have a reduced
fitness because they can produce only male sexuals.
We hypothesize that the two multilocus genotypes, which are
currently within each species, arose from a past hybridization
event. This hybridization between the two species generated
genetic incompatibilities in diploid females, which were then
apparently only able to develop into sterile workers. However,
because queens of these species mate with multiple males, they
could still raise female sexuals, too, as long as they had also
mated with at least one compatible male. Thus, because of the
already present polyandry, the hybridization was not selected
against. Once hybridization occurred, a second evolutionary
step had to occur, the elimination of homozygotic (or nonhy-
Fig. 1. Proportion heterozygous individuals comparing alates and workers
of P. barbatus and P. rugosus (identified morphologically) in areas of sympatry
brid) workers. Although hybridization would theoretically
versus allopatry. There is a significant difference between sympatric and provide an excess of heterozygote workers, there is no a priori
allopatric populations in the frequency of heterozygotes in queens and expectation that it would also prevent homozygotes from
workers. (P. rugosus, ␹2 ⫽ 26.1, P ⬍ 0.000; P. barbatus, ␹2 ⫽ 40, P ⬍ 0.000.) The becoming workers. This second step was presumably under
low amount of heterozygosity in allopatric P. rugosus is a result of one allele, strong selection because of kin conf lict over caste determina-
550, being rare in that population. However, allopatric populations show no tion. In a polyandrous system (with associated reduction in
difference in the amount of heterozygosity between castes. The asymmetry in within-colony relatedness) and with a hybrid worker caste in
heterozygosity between alates and workers in sympatric populations reveals
place, it becomes more advantageous for a nonhybrid female
a genetic caste determination.
to become a queen. Selection on these females also may have
been enhanced by selection on males to father reproductives
Discussion rather than workers. An alternative, more mechanistic expla-
nation for the genotype-phenotype correlation in hybrid col-
The evolution of caste determination is central to the evolution
onies would be that in colonies where most individuals (all
of complex sociality. Kin selection as an explanation for euso-
heterozygotes) cannot develop into queens because of hybri-
ciality is based on the assumption that sterile workers forgo
dogenesis, the remaining homozygous larvae may monopolize
direct fitness to help their closely related siblings reproduce (1).
This argument for social evolution also depends on conditional the attention of workers seeking to turn female larvae into
expression by any genes underlying caste determination (4, 5). In queens. This mechanism could also explain the otherwise
contrast to this expectation, our data show clear evidence of puzzling fact that same-species gamete combinations do not
genetically based caste determination between reproductive and produce viable workers in mixed colonies but do in allopatric
sterile females. Within areas of sympatry, genetic differentiation populations. However, under this scenario we would expect
between alates and workers was essentially absolute. All workers that in founding colonies of mixed colonies we should find
were heterozygous for the C9 locus, whereas almost all alates homozygous workers, but this remains to be tested.
This pattern of worker heterozygosity and queen homozygos-

EVOLUTION
were homozygous. An additional 14 RAPD primers also indi-
cated genotypic differentiation between alates and workers ity may not be unique to Pogonomyrmex. Hung and Vinson (27)
(Table 1). These data collectively suggest that the C-9–550兾510 presented allozyme evidence consistent with genetic caste de-
locus is a marker for a large genomic linkage group affecting termination in fire ants, Solenopsis geminata and Solenopsis
caste determination. xyloni. Because they did not have DNA-based markers, they
How can a system of genetic caste determination arise? We interpreted their results to suggest differential enzyme expres-
suggest that caste determination in P. rugosus and P. barbatus is sion between reproductive queens and workers.
related (at least historically) to hybridization. Our results show The ability to shunt nonconspecific offspring into sterile castes
an extreme genetic effect on caste within areas of sympatry, but offers eusocial species a unique mechanism for countering the
we found no similar pattern in allopatric populations of either negative fitness consequences of hybridization (24). Because
species. We hypothesize that ants in the sympatric zone have workers are valuable to colony success, a differential caste
responded to introgression with a unique form of colony-level or trajectory for conspecific versus nonconspecific sperm would
social hybridogenesis (24). This hybridogenesis is similar on a offset negative fitness consequences of interspecies mating.
colony level to that found on an organismal level in certain Shunting heterospecific sperm into workers would also limit the
species of guppies (Poeciliopsis) (25). These Poeciliopsis females movement of hybrid genomes into F2 generations, because
hybridize with males of another species, but during meiosis of the reproductives are almost universally the product of conspecific
offspring only the maternal genes remain in viable gametes (26). matings. If so, eusociality may paradoxically contribute to hybrid
Thus the male’s genome is not transferred beyond the F1 colony success, but in turn limit the extent of hybridization
generation, which is the same in our system. beyond the F1 generation.
How could hybridogenesis in Pogonomyrmex work? Our data
indicate two separate multilocus genotypes present within both We thank T. Polony for assistance in collection of samples; S. Poloner
species, which are revealed by the 510 and 550 markers and the for assistance with PCR; and S. Cahan, T. Dowling, J. Harrison, J.
Parker, D. Wheeler, and members of the Social Insect Research Group
additional RAPD markers (Tables 1 and 2). To survive and
for discussion. We also thank three anonymous referees whose com-
reproduce, homozygous queens must mate with a male of ments helped to improve the manuscript. D.L. was supported by a
matching type to produce reproductively capable daughters. To Howard Hughes grant for undergraduate research. Research was funded
produce sterile workers they must also mate with a male of the by National Institute of Mental Health Grant R29 MH51329 to J.H.F.
opposite type. However, because workers do not reproduce, the This work was also funded by grants from the Deutsche Forschungsge-
genetic information of nonmatching males is not transferred meinschaft to J.G. (Ga-661兾1 and SFB 554兾TPC5).

Julian et al. PNAS 兩 June 11, 2002 兩 vol. 99 兩 no. 12 兩 8159

 1. Hamilton, W. D. (1964) J. Theor. Biol. 7, 1–52. 15. Johnson, R. A. (2000) Sociobiology 36, 89–102.
2. Wilson, E. O. (1971) The Insect Societies (Belknap of Harvard Univ., Cam- 16. Cole, A. C. (1968) Pogonomyrmex Harvester Ants (Univ. of Tennessee, Knox-
bridge, MA). ville).
3. Holldobler, B. & Wilson, E. O. (1990) The Ants (Belknap of Harvard Univ., 17. Parker, J. (1999) Dissertation (Arizona State Univ., Tempe).
Cambridge, MA). 18. Holldobler, B. (1976) Behav. Ecol. Sociobiol. 1, 405–423.
4. Seger, J. (1981) J. Theor. Biol. 91, 191–213. 19. Landry, B. S., Dextraze, L. & Boivin, G. (1993) Genome 36, 580–587.
5. Queller, D. C. & Strassman, J. E. (1998) BioScience 48, 165–175. 20. Gadau, J. (1996) Mol. Ecol. 6, 785–792.
6. Brian, M. V. (1957) Annu. Rev. Entomol. 2, 107–120. 21. Williams, J. G. K., Kubellik, A. R., Livak, K. J., Rafalski, J. A. & Tingey, S. V.
7. Wheeler, D. E. (1986) Am. Nat. 128, 13–34.
(1990) Nucleic Acids Res. 18, 6531–6535.
8. Nijhout, H. F. (1999) BioScience 49, 181–192.
22. Hunt, G. J. & Page, R. E. (1992) Theor. Appl. Genet. 85, 15–20.
9. Evans, J. & Wheeler, D. E. (1999) Proc. Natl. Acad. Sci. USA 96, 5575–5580.
23. Shoemaker, D. D., Ross, K. G. & Arnold, M. L. (1994) Mol. Ecol. 3, 531–539.
10. Winter, U. & Buschinger, A. (1986) Entomol. Generalis 11, 125–137.
11. Fersch, R., Buschinger, A. & Heinze, J. (2000) Insects Soc. 47, 280–284. 24. Seifert, B. (1999) Insects Soc. 46, 45–52.
12. Umphreys, G. J. & Danzmann, R. G. (1998) Biochem. Syst. Ecol. 26, 431–440. 25. Beukeboom, L. W. & Vrijenhoek, R. C. (1998) J. Evol. Biol. 11, 755–782.
13. Kerr, W. E. (1950) Genetics 35, 143–152. 26. Vrijenhoek, R. C. (1978) Science 199, 549–552.
14. Ratnieks, F. L. W. (2001) Behav. Ecol. Sociobiol. 50, 467–473. 27. Hung, A. C. F. & Vinson, S. B. (1977) Science 196, 1458–1459.

8160 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.112222099 Julian et al.