Oecologia (2002) 133:492–500 DOI 10.1007/s00442-002-1069-3
POPULATION ECOLOGY
Krista K. Ingram
Flexibility in nest density and social structure in invasive populations of the Argentine ant, Linepithema humile Received: 25 March 2002 / Accepted: 28 August 2002 / Published online: 2 October 2002 Springer-Verlag 2002
Abstract The extraordinary success of unicolonial insect invaders has renewed interest in the origins and maintenance of unicoloniality, a social system characterized by the absence of aggressive behavior between individuals of neighboring conspecific nests. In this study, I explore how traits associated with unicoloniality, particularly nest budding and the local exchange of workers between nests, vary across environments. Comparisons of nest characteristics in three introduced populations of Argentine ants, Linepithema humile, reveal considerable variation in nest density and social structure across populations. The population with the highest nest density has smaller, less productive nests and fewer queens per nest than the two populations with low nest densities. Nestmate relatedness is low in all populations, but multi-locus genotype analyses reveal that levels of connectivity vary among populations. In particular, high nest density is associated with higher levels of genotypic similarity between nests. Assignment distances in the two populations with low nest density are similar to a native population, suggesting that the amount of mixing between neighboring nests is shared among some native and introduced populations. Because the study populations are similar in age and genetic diversity, these results suggest that the social structure of L. humile populations depend on the ecological context. Understanding the patterns of expression of unicolonial traits in different environments helps to shed light on the origins of unicoloniality and to explain the success of Argentine ants as invaders. Keywords Ecological constraints · Introduced species · Multiple queens · Social insects · Unicoloniality K.K. Ingram ()) Department of Organismic and Evolutionary Biology, Harvard University, Cambridge, MA 02138, USA Present address: K.K. Ingram, 371 Serra Mall, Department of Biological Sciences, Stanford University, Stanford, CA 94305, USA, e-mail:
[email protected], Tel.: +1-650-7256791, Fax: +1-650-724
Introduction In unicolonial social insect species, large networks of polygynous (multiple queen) nests are formed by the budding of queens and workers from existing nests (Hlldobler and Wilson 1977). The relatedness among nestmates is low due to the production of brood by multiple queens and the frequent movement of workers and queens between nests. Unicoloniality thus challenges kin selection arguments for evolution of cooperation between non-reproductive workers because cooperators receive only weak inclusive fitness benefits (Hamilton 1964, 1972; Crozier 1979; Nonacs 1988; Keller 1995). Despite the patchy taxonomic distribution of unicoloniality (Bourke and Franks 1995; p. 284), many unicolonial ants enjoy great ecological success as invaders, particularly species exploiting ‘tramp ant’ (Passera 1994) strategies (e.g. living in close association with humans). Unicolonial populations are defined by an absence of aggressive behavior between individuals of neighboring nests (Wilson 1971, p. 457; Passera 1994). In introduced populations of Argentine ants, Linepithema humile, the lack of clear colony boundaries often leads to the formation of expansive, polydomous colonies of interconnected nests (Tsutsui et al. 2000; Giraud et al. 2002). Two key features associated with unicoloniality, the dispersal of nests via budding and the lack of clear colony boundaries mediated by extensive intermixing of workers between nests, suggest that flexibility in social structure may play an important role in the success of invasive unicolonial ants. Such flexibility can evolve under natural ecological conditions in response to frequently disturbed habitats (Bourke and Franks 1995; pp. 284–286) or as a strategy for colonizing gaps in environments with limited nest sites (Hlldobler and Wilson 1977). Many unicolonial species typically inhabit ephemeral nest sites and can frequently change nest locations in response to environmental variation (Hlldobler and Wilson 1990; p. 215). In fact, unicoloniality tends to evolve in species that inhabit frequently disturbed areas, such as sandy slopes or river
493
banks prone to flooding (Bourke and Franks 1995; p. 284). The advantage of maintaining fluidity of brood and workers between nests is that environmental fluctuations are unlikely to influence all nest sites equally (Rosengren and Pamilo 1983). In favorable environments, colonies can expand rapidly via budding. Alternatively, in highly constrained environments, flexibility in social structure allows colonies to persist despite reductions in critical resources. These arguments are similar to the habitat saturation hypothesis for dispersal in bird species with helpers at the nest (Emlen 1984; Brown 1987) and to hypotheses explaining predator/competition avoidance in colonial marine organisms (Jackson and Palumbi 1979; Buss 1980; McKinney and Jackson 1989). Under fluctuating environmental conditions or high disturbance regimes, flexibility in social structure may be a successful strategy in the native habitats of unicolonial species (Chapman and Bourke 2001). Species adapted to naturally disturbed environments might also possess traits that enhance colonization abilities in human-dominated landscapes, such as the ability to produce new nests rapidly (Passera 1994). Thus, flexibility in social structure allows nests to respond to different environments by adopting strategies that vary in the frequency of budding and movement between nests. Depending on the environment, the variation in social structure within native populations or introduced populations can be greater than between native and introduced populations. This paper explores the possibility that variation in the expression of ‘unicolonial traits’ exists in invasive populations and that the sociogenetic structure of unicolonial species depends on the ecological context in which the population is found. If ecological factors can influence the social structure of nests in unicolonial ants, variation in the frequency of budding and movement between nests is expected in invasive populations that experience different environments. In particular, these traits may vary when the environments differ in the level of ecological constraints, i.e. habitat saturation, nest site limitation or competition. Introductions into habitats with relaxed constraints generally lead to high nest densities (Porter et al. 1997). Due to risky dispersal, nests in high constraint environments are expected to bud less frequently, leading to low nest densities. Populations with low nest density are predicted to have larger nests, higher queen numbers per nest, and lower within-nest relatedness than populations with high nest densities. In addition, the mixing of individuals between nests is expected to be less common in lowdensity populations where nests are separated by greater distances. This study examines the social and genetic structure of nests in three introduced Argentine ant populations. The Argentine ant, a highly polygynous, unicolonial species, was originally described from Buenos Aires, Argentina, in 1866 and has since successfully invaded a diversity of habitats worldwide (Newell and Barber 1913; Haskins and Haskins 1965; Huddleston and Fluker 1968; Markin
1970; Ward 1987; Giraud et al. 2002). Previous work has shown that queen number and social structure in Argentine ants change across an ecological gradient of habitat saturation during an invasion in Hawaii (Ingram 2002). The aims of the present study are to determine whether populations invading different environments vary in nest density and to investigate how the social structure and genetic connectivity of nests varies across these populations.
Materials and methods Study populations Three introduced populations of Argentine ants were investigated in this study. The populations were located in Haleakala National Park on the island of Maui in Hawaii, Auwahi dryland forest on the southwestern slope of Haleakala in East Maui, and Aminga, a small town in the northwestern province of La Rioja, Argentina. All three of the populations were introduced approximately 10–30 years ago, an important criteria for studies comparing the genetic structure of nests in introduced populations. The three study populations differ in both habitat type and faunal community. Because competitive constraints strongly influence the frequency of successful dispersal of nests via budding, nest densities within a population are likely to be affected by interspecific competitors. For aggressive ants such as L. humile, competition comes mainly in the form of other ant species (Hlldobler and Wilson 1990). In order to obtain a rough estimate of the number of potential interspecific ant competitors, ant species richness was estimated for each Argentine ant population with bait transects and spot collecting. Each potential competitor was identified to genus and species (if possible) by Stefan Cover (Museum of Comparative Zoology, Harvard University). Vouchers of all specimens were deposited in the Museum of Comparative Zoology at Harvard. The population of Argentine ants in Haleakala National Park was initially discovered in 1967 (Huddleston and Fluker 1968) and is described in detail by Ingram (2002). The data for the present study include a section located in the southwest corner of the expanding population between the elevations of 2,070 and 2,260 m. Due to the unique ecology of the Hawaiian Islands, particularly the lack of eusocial insects native to the archipelago (Wilson and Taylor 1967), there are no known predators or ant competitors of Argentine ants in this area. Only one other ant species, Hypoponera opaciceps, has been collected in the lower elevation site at Haleakala, but this species was found only in non-Argentine ant areas (Cole et al. 1992) and was never collected during the course of this study. Twenty nests were collected from Haleakala in February and March 1997 and 1998. This population has the lowest competitive constraints of the study populations. The Auwahi population is located approximately 10–20 miles from the Haleakala population on the southern slope of the same mountain. In this area, there are no ant predators, but the Argentine ant colonies are in direct competition with an invasive population of the big-headed ant, Pheidole megacephala. These two species are notorious competitors in many parts of the world where they have been introduced (Haskins and Haskins 1965, 1988; Huddleston and Fluker, 1968) and their interactions on the Hawaiian Islands have been aggressive, with Argentine ants displacing bigheaded ants primarily in the higher elevations (Huddleston and Fluker 1968). Since the Argentine ants invaded Auwahi in the 1970s from a higher elevation, the two species have formed an unstable front that oscillates near 1,200 m elevation, but has not moved appreciably over a 2-year study period (unpublished data; Arthur Medeiros, USGS, personal communication). In March 1999, 13 nests were collected along this front, where Argentine ants were in direct, aggressive competition with the big-headed ant colonies.
494 The third study population was introduced into Aminga, Argentina approximately 10–20 years ago. Outside of the native range of L. humile, nests collected from this population represent colonies experiencing constraints imposed by a community of predators, prey, and competitors. Collections of ten nests at this site occurred in February 1999. Using bait transects and spot collecting, a number of ant species were found coexisting with Argentine ants in this area. The collection included two species of Forelius, two species of Dorymyrmex, and single species of Camponotus, Ephebomyrmex, Pogonomyrmex, Pheidole, Acromyrmex, and Solenopsis (S. Cover, personal communication). This list includes some notoriously aggressive species, such as Pheidole, Solenopsis, and Forelius species, as well as species that may not interact extensively with L. humile. For comparison with the introduced populations, data from a native population were collected from Reserva Otamendi, a wetland natural preserve north of Buenos Aires, Argentina. Unfortunately, it was not possible to excavate entire nests at this site or estimate nest size, productivity, or queen number. Workers and brood were collected from six nests at this location in February for genetic analysis. Microsatellite data from these nests were used in the analyses of genotypic connectivity. Nest census L. humile nests (n=49) were collected from each population at midday when the ants aggregated under rocks or at the base of shrubs or trees, just below the soil surface. Entire nests were excavated with shovels (approximately 5–25 l of soil) and placed in a large sampling tray to collect individuals. All nests that had multiple brood chambers at the same nest site were collected as a single nest. The density of individual nests in each population was estimated from the distance to the nearest conspecific nest. Due to unequal variances, significant differences across populations were tested with a Kruskal-Wallis test and differences between populations were tested with nonparametric multiple comparisons with unequal sample sizes (Zar 1999). The contents of each nest were examined for 30 min and all queens, males, and brood (including eggs, larvae, and pupae) found during this period were counted and collected in 70–95% alcohol. Preliminary collections showed that all queens, males, and brood could be collected within 30 min of search time, with the majority actually collected within 15 min. Nest size was estimated by counting the number of workers per nest. For large nests, the number of workers was estimated by subsampling from the excavated nest. The amount of brood production in nests was measured by counting the number of brood collected from the excavated nest. Differences across populations in nest size and brood production were tested for significance with chi-square analyses. Average queen number per nest was calculated as an arithmetic mean, QA. Differences in QA across populations were tested with a Kruskal-Wallis test and differences between populations were tested with nonparametric multiple comparisons with unequal sample sizes (Zar 1999). Because genetic measures of relatedness are based on the average number of matrilines into which progeny are divided (i.e. the reciprocal of queen number), a harmonic mean, QH, was calculated for comparisons with relatedness estimates and, in particular, estimates of effective queen number (Wade 1985; Queller 1993). Queenless nests (nests with brood present but no queen collected) were found only in Haleakala. For this population, QH was calculated as the harmonic mean of all nests in the population (n=64) assuming queenless nests were monogynous. Effective queen numbers (per nest) were calculated from regression relatedness values of nestmate brood (see below).
ized, and boiled in 100 l of a 10% Chelex solution for 15 min. To avoid sperm-derived DNA, queens were dissected in half and DNA was extracted from the head and half of the thorax. After boiling, the extraction solutions were centrifuged for 1 min and supernatant was removed. Samples were stored at –20C. One l of each sample was used in a 12.5-l PCR reaction. Seven primer sets were used to genotype each individual at microsatellite loci: Lihu-H, Lihu-O, Lhum-13*, Lhum-35*, Lhum19*, Lhum-11*, Lhum-52*. Primer sequences and detailed methods of PCR amplification are described in Ingram and Palumbi (2002) and Ingram (2002). PCR products were run on 5% Sequagel acrylamide gels using ABI fluorescent dyes and the gels were subsequently analyzed using GENESCAN software. Average regression relatedness values, r, were measured for nestmate workers, brood and queens, and between workers and brood within a nest. Regression relatedness is the genotypic correlation among individuals in a group or between individuals from different groups (Hamilton 1972; Pamilo 1985). I used the method described by Queller and Goodnight (1989) and the program Relatedness 5.0 to obtain relatedness estimates. For all calculations, nests were weighted equally and standard errors were obtained by jackknifing over loci. Relatedness values were tested for significant departures from zero with one-tailed t-tests. All queens collected from nests were dealate queens and were assumed to be inseminated and potentially reproductive (Krieger and Keller 2000). The number of queens actively reproducing in a nest was estimated as the effective queen number by using the formula, QEFF ¼
ð1Þ
where rBB is the regression relatedness of the brood in a nest (Queller 1993; Ross 1993; Pedersen and Boomsma 1999). Assuming coexisting queens in this population are unrelated and generally mate only once (Keller and Passera 1992; Krieger and Keller 2000), QEFF is the number of queens contributing equally to the observed genetic diversity of female brood in a randomly mating population. QEFF was corrected for inbreeding using Pamilo’s equation (1985, Eq. 5b): 1 3 1F 2F QEFF ¼ þ ð2Þ rBB 4 1þF 1þF where QEFF* is the genetically effective queen number in the absence of inbreeding, F is the inbreeding coefficient, and rBB is the genetic relatedness of brood when there are QEFF* egg-laying queens contributing equally in a nest. Population structure and connectivity Differences in number of alleles between populations were tested with G-tests (Sokal and Rohlf 1995). As microsatellite loci can be considered a random sample of loci from the genome, I used Friedman’s method for randomized blocks to test for heterozygosity between populations (Sokal and Rohlf 1995). Two factors can affect estimates of nestmate relatedness and, in particular, the estimate of effective queen number: the partial genetic isolation by distance due to population substructure (Pamilo 1984) and assortative mating between relatives (Pamilo 1985). F-statistics were calculated in FSTAT (version 2.9.1; Goudet 1995) to measure the degree of population structure (FST) and inbreeding (FIS) in the three populations. R-statistics (measures of population substructure based on allele sizes) were calculated for comparison with Fstatistics in GENEPOP. Due to the significant inbreeding coefficient in Aminga, brood relatedness for this population was corrected using Pamilo’s equation (1985; Eq. 1a):
Sample extraction and microsatellite analysis For the genetic analyses, 6–10 adult workers, 12–18 brood, and all queens were typed from each nest (n=49), representing 885 individuals in total. Alcohol preserved workers and brood (larvae and pupae) were soaked in distilled water for 15–30 min, pulver-
3 ; 4rBB
rBB ¼
rBB 2F ð1 þ F Þ 1 2F ð1 þ F Þ
ð3Þ
where rBB is the regression relatedness and F is the fixation index. The effective queen number for Aminga was then estimated using rBB* and Eq. 2) above.
495 Two methods were used to measure connectivity between nests. First, I calculated the relative differences in nestmate relatedness within and between nests across the three populations in Relatedness 5.0 (Queller and Goodnight 1989). Similarities in relatedness values within- and between-nest suggest a high degree of mixing between nests. Differences between these two values were tested for significance in each population with t-tests. In addition, I used assignment tests to determine the probabilities of multi-locus worker genotypes in nests of each population (Paetkau et al. 1995). Genetic distances between nests within a population were calculated from the proportion of misassigned individuals (http:// www.biology.ualberta.ca/jbrzusto/Doh.php). Small assignment distances represent high rates of genotypic connectivity and mixing between nests. Average pairwise genetic distances of nests were compared between populations with two sample t-tests.
Fig. 1 Differences in nest density in the three study populations of Linepithema humile shown as boxplots of distance to the nearest neighboring nest (H=29.1, P
