Tropical leaf-cutter ants cultivate the fungus Attamyces bromatificus in a many-to-one, diffuse coevolutionary relationship where ant and fungal partners re-associate frequently over time. To evaluate whether ant–Attamyces coevolution is more specific (tighter) in peripheral populations, we characterized the host-specificities of Attamyces genotypes at their northern, subtropical range limits (southern USA, Mexico and Cuba). Population-genetic patterns of northern Attamyces reveal features that have so far not been observed in the diffusely coevolving, tropical ant–Attamyces associations. These unique features include (i) cases of one-to-one ant–Attamyces specialization that tighten coevolution at the northern frontier; (ii) distributions of genetically identical Attamyces clones over large areas (up to 81 000 km2, approx. the area of Ireland, Austria or Panama); (iii) admixture rates between Attamyces lineages that appear lower in northern than in tropical populations; and (iv) long-distance gene flow of Attamyces across a dispersal barrier for leaf-cutter ants (ocean between mainland North America and Cuba). The latter suggests that Attamyces fungi may occasionally disperse independently of the ants, contrary to the traditional assumption that Attamyces fungi depend entirely on leaf-cutter queens for dispersal. Peripheral populations in Argentina or at mid-elevation sites in the Andes may reveal additional regional variants in ant–Attamyces coevolution. Studies of such populations are most likely to inform models of coextinctions of obligate mutualistic partners that are doubly stressed by habitat marginality and by environmental change.
Evolution in peripheral populations can differ markedly from evolution in more central populations within a species range [1–3]. Peripheral populations often exist at lower densities, produce fewer offspring per individual, are more fragmented and are more prone to local extinction [2,3]. These demographic properties leave distinct population-genetic footprints, such as reduced genetic diversity, increased homozygosity and increased levels of unique (private) alleles in peripheral populations compared with central populations [4,5]. A large body of work has confirmed these general population-genetic predictions, although there are exceptions [5–7]. Much less is known about the population-genetic patterns of host–symbiont associations in marginal habitat, yet studies on symbioses near their distributional range limits are needed to inform models of coextinctions of symbiotic partners that are doubly stressed by habitat marginality and by accelerated environmental change [8,9]. To inform such models of host–symbiont coextinction, we elucidate here the genetic diversities and host-specificities of Attamyces fungi cultivated by leaf-cutter ants at their northern range limit.
The obligate mutualism between leaf-cutter ants (genera Atta and Acromyrmex) and their Attamyces fungi originated 8–12 Myr ago in the South American tropics but extends today into temperate regions [10–13]. In the Southern Hemisphere, leaf-cutter ants reach 44° S latitude . In the Northern Hemisphere, Atta texana reaches 33° N latitude in north-central Texas and northeastern Louisiana [15; this study] and Acromyrmex versicolor reaches almost 36° N latitude in northwestern Arizona . It is unknown whether leaf-cutter ants dispersed into the North American continent before or after the closing of the Isthmus of Panama approximately 1–3 Myr ago . However, leafcutter ants must have reached their current latitudinal limit in the southern USA sometime during the past 10 000 years following the most recent Pleistocene glaciation. The leaf-cutter ant At. texana was abundant in central and east Texas at the time when European settlers arrived [18,19], whereas a historical presence of Ac. versicolor in Northern Arizona is unknown, but likely. Because tropical Attamyces fungi are cold-intolerant and grow best at the warm, stable temperatures of tropical soils (between 20°C and 30°C) [20–23], the cold-sensitivity of Attamyces symbionts is thought to have been one factor that constrained the expansion of the leaf-cutter ant–Attamyces symbiosis from tropical into temperate habitats [14,24].
All leaf-cutter ants depend on symbiotic fungi for food, which grow in gardens tended by the ants in excavated subterranean cavities or in thatched shelters [10,12,19,25]. The fungi cultivated by leaf-cutter ants belong to the agaric tribe Leucocoprineae (fungal anamorph Attamyces bromatificus, teleomorph Leucocoprinus gongylophorus, Agaricales, Basidiomycota; [26,27]). All Attamyces strains genotyped so far showed polyploid-like allele patterns with more than two alleles per locus [24,28,29], consistent with cytological studies indicating that Attamyces cells are multi-nucleate . The polyploid-like allele patterns could derive from a complex heterokaryon (coexistence of genetically differentiated haploid nuclei in the same cell, which is the typical growth form of basidiomycete mycelia), from duplicated genomes within single nuclei or from a combination of both . Because Attamyces fungi have not been found so far growing independently of the ants [31,32], Attamyces appears to be on obligate symbiont. However, at least some Attamyces fungi are fruiting-competent and can produce spore-bearing mushrooms (table 3 in Mueller  summarizes records of fruiting leaf-cutter fungi). Attamyces fruiting bodies in the field are so far known mostly from Acromyrmex leaf-cutter ants that build thatched gardens at ground level [33,34].
Attamyces fungi are clonally propagated by the ants within and between nests [35,36], suggesting the possibility of strict Attamyces asexuality imposed by the ant farmers; however, incongruence between the phylogenetic topologies of segregating genes indicates that clonality of Attamyces is occasionally punctuated by recombination events [37,38]. In laboratory experiments, co-cultivated cultivar genotypes may recombine in artificially created chimaeric gardens [28,39], possibly through the exchange of nuclei between cultivar mycelia. A separate study testing for chimaeric gardens in natural leaf-cutter nests revealed no evidence for Attamyces polyculture in the leaf-cutter ants At. texana and Atta cephalotes, suggesting that Attamyces is grown by the ants in single-strain monoculture throughout the hundreds of gardens within a single mature leaf-cutter nest .
Across the leaf-cutter ant range from Argentina to the USA, the approximately 50 described leaf-cutter species are thought to associate with a single cultivar species (Attamyces bromatificus Kreisel) in a many-to-one coevolutionary relationship [37,40]. Whereas the leaf-cutter ant clade is estimated to be about 8–12 Myr old, the corresponding clade of Attamyces cultivars is significantly younger (about 2–4 Myr old) . Attamyces lineages of recent origin, therefore, may have spread by means of horizontal transfer between leaf-cutter ant lineages (i.e. Attamyces sweeps through the range of leaf-cutter ants). Comparison of fast-evolving genes of Attamyces strains from geographically distant leaf-cutter species across South America  and population-genetic patterns of Attamyces cultivated by sympatric leaf-cutter species in Panama  confirmed the expected sharing of Attamyces lineages between tropical leaf-cutter ant species. Because a local community of tropical leaf-cutter ants shares a corresponding local community of cultivar lineages [38,40,41], Attamyces cultivars in the tropics are thought to evolve within a continually shifting landscape of the microhabitats occupied by the diverse leafcutter ants between which cultivars are exchanged.
The most detailed population-genetic study of Attamyces symbionts to date was conducted by Mikheyev et al.  for sympatric populations of three Atta species and two Acromyrmex species from Panama. Despite vertical transmission of Attamyces strains from mother to daughter nests, different leaf-cutter ant species and genera sometimes share identical Attamyces genotypes (clones), indicating widespread cultivar exchange across Panama. Attamyces genotypes group into six distinct Attamyces genotype-clusters in Panama, but these clusters do not correspond to the five ant hosts (only about 10% of the structure in genetic variance of Attamyces is attributable to species and generic boundaries among ant hosts; ). Frequent exchange of Attamyces clones between heterospecific ant nests or other forms of Attamyces gene flow between nests, therefore, prevents the long-term persistence of specific ant–Attamyces combinations, leading to an overall pattern of diffuse coevolution between leaf-cutter ant hosts and Attamyces symbionts in this tropical population.
At the northern range limit of leaf-cutter ants (southern USA), Attamyces evolution progresses differently than in the tropics, because here leaf-cutter ants do not exist sympatrically with other leaf-cutter species. For example, the Texas leaf-cutter ant At. texana is the only leaf-cutter species within its range, except for a contact zone with Atta mexicana just south of the USA–Mexico border [19,42]. Likewise, the desert leaf-cutter Ac. versicolor overlaps with At. mexicana south of the USA–Mexico border, but Ac. versicolor is the only leaf-cutter ant in its northern range in the USA. Ant–fungus coevolution in these northernmost leaf-cutter populations, therefore, is expected to be tighter (i.e. involving a single ant species and its Attamyces cultivars) compared with the multi-species, diffuse coevolution resulting from sharing of Attamyces strains between sympatric leaf-cutter ant species in the tropics. The expectation of possibly tighter ant–fungus coevolution in the northernmost leaf-cutter populations stimulated our investigations into the population genetics of Attamyces at the northern range limit of the leaf-cutter ant distribution.
2. Material and methods
Attamyces fungi were collected from leaf-cutter gardens excavated from nests throughout the US ranges of the northernmost leaf-cutter ants At. texana (n = 165 Attamyces accessions from an equal number of nests) and Ac. versicolor (n = 35 Attamyces). To place these Attamyces collections in a larger population-genetic context of North American Attamyces, we also collected garden material from five Atta insularis nests from Cuba; from seven At. mexicana nests from northeastern and southern Mexico; and from eight At. cephalotes nests from southeastern Mexico. The electronic supplementary material, table S1 lists information collected for all 220 Attamyces accessions collected from the five leaf-cutter ant species from the USA, Mexico and Cuba.
Attamyces accessions were genotyped with a panel of 12 polymorphic microsatellite loci . Population-genetic patterns were analysed in Structure v. 2.2 [43,44] by clustering individuals into populations on the basis of the multi-locus genotype information. Because Attamyces fungi are multi-nucleate and exhibit genotype profiles of unknown ploidy , we did not calculate standard population-genetic parameters (e.g. heterozygosities; F-statistics, etc.), but treated all alleles as dominant markers for genotype clustering and inference of population-genetic structure, as recommended for polyploid organisms by Falush et al. . Structure requires that individuals differ by at least one marker, and we therefore included in our population-genetic analyses only one representative per microsatellite genotype. Because Attamyces is clonally propagated within and between leaf-cutter nests , 55.7 per cent of the 220 Attamyces accessions possessed a genotype profile that was identical to the profile of at least one other accession (see the electronic supplementary material, table S1). After eliminating duplicates to retain only one representative per genotype, the final dataset analysed in Structure included 93 unique genotypes (each profiled for 91 variable microsatellite markers). A Markov chain Monte Carlo algorithm implemented in Structure identified genetically differentiated populations and inferred contributions of ancestry for each allele (likelihood that an allele derived from one of the inferred populations). Admixture (mixed ancestry) was inferred for a particular fungal accession if different alleles of this genotype were assigned with high likelihood to different populations. Such admixture may result from (i) recombination of alleles (as in sexual organisms); or (ii) exchange of nuclei between differentiated Attamyces strains, generating re-assortment and novel combinations of nuclei that coexist in the multi-nucleate mycelium of Attamyces (see the electronic supplementary material Admixture for a detailed description of the diverse mechanism of genetic exchange in multi-nucleate fungi). Separately for the groups of Attamyces genotypes from At. texana (n = 165) and Ac. versicolor (n = 35), we also estimated the probability of sexual recombination within each group by applying the methods of Bengtsson  and De Fine Licht et al.  in a Bayesian framework implemented in WinBUGS . To visualize the genetic diversity and to validate results obtained from Structure, we calculated a two-dimensional non-metric multi-dimensional scaling solution of the binary distances between cultivar genotypes (figure 1), which was computed in the programs R and GenAlEx (www.anu.edu.au/BoZo/GenAlEx/) . Collection, genotyping and statistical methods are explained in detail in the electronic supplementary material.
(a) Specificity and sharing of Attamyces symbionts at the northern range limit
Structure analyses inferred four populations of Attamyces symbionts cultivated at the northern range limit of the leaf-cutter ant distribution (figure 1). These solutions were convergent between three repeat runs. Modelling more than four populations (K > 4) did not significantly improve likelihood scores (see the electronic supplementary material, figure S2). Two of the four inferred populations represent two dominant cultivar types of At. texana (‘T-group’ and ‘M-group’ Attamyces); one of these two Attamyces populations from At. texana (dark blue, figure 1) also includes Attamyces cultivated by other leafcutter species in Mexico (hence M-group Attamyces) and by Ac. versicolor in Arizona. The second Attamyces population from At. texana (light blue, figure 1) is clearly distinct from the three other populations and was found only in Louisiana and Texas (hence T-group Attamyces), but not in Mexico or the western USA. A third population was found only in Arizona and California (cultivated there by Ac. versicolor, therefore, called V-group Attamyces; red, figure 1). A fourth population includes a more diverse assemblage of Attamyces cultivated by At. insularis in Cuba (hence C-group Attamyces; green, figure 1), by At. mexicana and At. cephalotes in Mexico and by atypical and rare nests of At. texana in Texas (these two rare Attamyces genotypes are shown in figure 1 far right centre, and they are listed as genotypes nos. 74 and 75 in the electronic supplementary material, table S1).
Among the 220 collections, therefore, two Attamyces populations appear ant–species-specific: V-group cultivars from Ac. versicolor and T-group cultivars from At. texana. Attamyces belonging to the remaining two cultivar populations are shared between several leaf-cutter ant species across North America, paralleling the Attamyces sharing observed between sympatric leaf-cutter species in Panama . Additional collections of Attamyces from At. mexicana in northern Mexico may reveal that V-group and T-group Attamyces are less ant–species specific than suggested by the present analysis. This is more likely for the V-group Attamyces of Ac. versicolor that occur in close proximity to known populations of At. mexicana (figure 2a), but less likely for T-group Attamyces that have so far not been found in south Texas and thus not near the range of At. mexicana south of the USA–Mexico border (figure 2b).
(b) Biogeography of Attamyces symbionts cultivated by At. texana
The two dominant Attamyces symbiont populations cultivated by At. texana together comprise 97.6 per cent of the known Attamyces diversity associated with At. texana (56.4% T-group Attamyces accessions, n = 93 of 165 accessions total; 41.2% M-group Attamyces accessions, n = 68; see the electronic supplementary material, table S1). The remaining 3.4 per cent of the Attamyces accessions (n = 4) from At. texana includes: (i) one accession (0.6%) that was significantly admixed under the K = 4 model (more than 30% of an individual's markers assigned to at least two populations; this single accession is listed as genotype no. 73 in the electronic supplementary material, table S1 and is shown in the centre of figure 1); and (ii) three accessions (1.8%) assigned under the K = 4 model to the fourth population of Attamyces that is also associated with several Atta species in Mexico and Cuba (the two genotypes of these three unusual accessions are listed as genotypes nos 74 and 75 in the electronic supplementary material, table S1 and are shown at the far right of figure 1). Assignment of accessions to either T-group or M-group Attamyces does not change when modelling a range of K = 4 to K = 12 populations (see the electronic supplementary material), and T-group and M-group Attamyces populations therefore represent differentiated symbiont types (figure 1) that together dominate the symbiont pool of the leaf-cutter host At. texana.
The populations of T-group and M-group Attamyces largely overlap across the range of the host At. texana, but with two interesting biogeographic differences. T-group Attamyces have so far not been collected in southern Texas (figure 2b and electronic supplementary material, table S1; all 11 accessions from At. texana south of latitude 28.06° N were M-group Attamyces). Second, M-group Attamyces have so far not been collected in Louisiana and east Texas (figure 2b and electronic supplementary material, table S1; all 19 Attamyces accessions collected from far east Texas and Louisiana were T-group Attamyces). Although our Attamyces collections are somewhat limited in the southern range (n = 11 Attamyces accessions) and the eastern range (n = 19 accessions) of At. texana, it appears that T-group Attamyces are restricted to more northern latitudes and are prevalent in the eastern range of At. texana for unknown ecological or historical reasons (e.g. T-group Attamyces are the only types that so far expanded with their host into east Texas and Louisiana).
(c) Biogeography of Attamyces symbionts cultivated by Ac. versicolor
Within Ac. versicolor (35 collections total), most Attamyces can be assigned to one of two populations; 69.6 per cent of these (n = 24) belonged to the V-group Attamyces and 11.1 per cent (n = 6) to the M-group Attamyces population (which are Attamyces lineages shared with the leaf-cutter hosts At. mexicana and At. texana; figure 1 and electronic supplementary material, table S1). Three Attamyces accessions from Ac. versicolor were inferred to be significantly admixed (defining admixture, as above, as more than 30% of the markers assigned to at least two Attamyces populations). M-group Attamyces have been found only in southeast Arizona, and Ac. versicolor appears to cultivate only V-group Attamyces throughout the rest of its US range (figure 2a and electronic supplementary material, table S1). Because of the fewer Attamyces collections from Ac. versicolor (n = 35) compared with At. texana (n = 165), and because Ac. versicolor ranges into Mexico where it is sympatric with At. mexicana, the apparent specificity of V-group Attamyces on Ac. versicolor is provisional and will need to be tested through further collection of Attamyces in northwest Mexico.
(d) Biogeography of Attamyces symbionts cultivated by At. insularis in Cuba and by At. mexicana in Mexico
Despite the marine dispersal barrier between mainland North America and Cuba, and despite the significant evolutionary divergence between the ant At. mexicana and the sister-species pair At. insularis and At. texana , we find no evidence for significant genetic differentiation between Attamyces from At. insularis and some Attamyces from At. mexicana (i.e. Structure assigned Cuban and some mainland Attamyces to the same population (C-group); figure 1 and electronic supplementary material, table S1).
(e) Geographical structure of Attamyces clones
Among the 165 accessions from At. texana, 12.7 per cent (n = 21) represent unique genotypes (collected in only a single At. texana nest), and the remaining 87.3 per cent accessions (n = 144) belong to genotypes that were collected in at least two different ant nests. Some of these Attamyces cultivars with genetically identical marker profiles were collected from nests at surprisingly distant locations. In At. texana, the average distance and the average maximum-distance between genotypically identical Attamyces accessions measured, respectively, 68.0 km (±10.64 s.e.) and 125.0 km (±11.36 s.e.; see the electronic supplementary material, table S1). In the most extreme case, two genetically identical Attamyces genotypes were collected from At. texana nests 490 km apart (southern to central Texas). The most wide-ranging M-group Attamyces genotype ranged over an area of at least 80 948 km2 (across south and central Texas), and the most wide-ranging T-group Attamyces ranged over an area of at least 22 188 km2 (across central and east Texas; see the electronic supplementary material, table S1). On average, M-group genotypes ranged over significantly larger areas (average = 23 973 km2, s.e. = 12 121, n = 7) compared with T-group genotypes (average = 2636 km2, s.e. = 1707, n = 13; t = 2.372, d.f. = 18, two-tailed p = 0.029; see the electronic supplementary material, table S1). Despite the enormous ranges of some Attamyces genotypes, the geographical clustering of Attamyces by genotype (see the electronic supplementary material, table S1) indicates that gene flow of Attamyces is viscous. This population viscosity across the range of the host At. texana is consistent with the observation that female At. texana disperse less than 20 km per mating flight . Overall, the genotype distributions across space confirm the expectation [35,36] that Attamyces cultivars of At. texana are primarily clonally propagated by dispersing females over many ant generations (long enough for Attamyces to spread as clones across areas of thousands of square kilometres).
Among the 35 accessions from Ac. versicolor, 57.1 per cent (n = 20) represented unique genotypes. The average distance and average maximum-distance between Attamyces accessions belonging to the same genotype measured, respectively, 8.9 km (±7.13 s.e.) and 17.49 km (±15.58 s.e.; see the electronic supplementary material, table S1). In the most extreme case, two genetically identical Attamyces genotypes were collected from Ac. versicolor nests 57.1 km apart (in southeast California). The Attamyces genotype with the largest and second-largest geographical distribution ranged, respectively, over only 2.61 km2 (in central Arizona) and 0.56 km2 (in southeast California; see the electronic supplementary material, table S1). As in Attamyces from At. texana, these patterns are likewise consistent with clonal propagation of Attamyces between Ac. versicolor nests; however, the comparatively smaller average range over which genotypically identical Attamyces were found in Ac. versicolor could indicate that, compared with clonal propagation in At. texana, strict clonal propagation of Attamyces occurs over fewer ant generations in Ac. versicolor. For example, clonal propagation in Ac. versicolor may be more frequently punctuated by genetic changes because the polygynous nest-founding of Ac. versicolor  creates frequent opportunities for co-cultivation of genetically differentiated Attamyces strains in incipient gardens (i.e. frequent opportunities for genetic recombination). However, because only 35 Attamyces were collected from Ac. versicolor (compared with 165 Attamyces from At. texana), Attamyces from Ac. versicolor may have been collected at insufficient geographical density relative to the dispersal distances of queens. A larger number of Ac. versicolor nests therefore will need to be collected across Arizona and California to rule out sampling biases.
Under all of the models examined in Structure, a significant fraction of Attamyces accessions was inferred to be admixed (i.e. these accessions combined markers assigned by Structure to at least two different Attamyces populations; see the electronic supplementary material, table S1). In the K = 4 model and under a relatively stringent definition of admixture (at least 30% of a genotype's markers were assigned to different populations), admixed Attamyces genotypes were found at lower frequency in the host At. texana (6.1%, n = 3 admixed genotypes of 49 total) compared with the host Ac. versicolor (11.5%, n = 3 admixed genotypes of 26 total) and compared with the tropical leaf-cutter hosts in Mexico and Cuba (average of 22.2%, n = 4 admixed genotypes of 18 total; see the electronic supplementary material, table S1). These proportions are not significantly different (all χ2 comparisons p > 0.1). Under a less stringent definition of admixture (at least 10% of a genotype's markers were assigned to different populations), admixed Attamyces occurred at comparable levels in At. texana (24.5%, n = 12 admixed genotypes) and Ac. versicolor (26.9%, n = 7 admixed genotypes), but more frequently in the tropical leaf-cutter hosts (average of 66.7%, n = 12 admixed genotypes; see the electronic supplementary material, table S1); this proportion of admixed genotypes among the tropical leaf-cutter hosts is significantly higher (χ12 4.749, p = 0.0293) compared with the two northernmost leaf-cutter hosts. A larger collection of Attamyces from tropical leaf-cutter ants will need to be genotyped to corroborate this pattern and rule out sampling artefacts as an explanation for the observed difference in admixture between temperate and tropical Attamyces. For the Attamyces from At. texana (n = 165) and Ac. versicolor (n = 35), fitting equation 8 of Bengtsson  within a Bayesian framework to the observed genotype distributions implicated very low frequencies of sexual reproduction per generation (0.3% with a 95% highest posterior density confidence interval (HPD-CI) of 0.01–1% for At. texana; 0.6% with a 95% HPD-CI of 0.02–3% for Ac. versicolor).
Population-genetic patterns of Attamyces cultivated near the northern range limit reveal some unique features of the ant–fungus mutualism that have so far not been observed for Attamyces in tropical habitats [37,38,41], but also some parallels between northern and tropical Attamyces.
(a) Parallels between northern and tropical Attamyces populations
Attamyces is largely clonally propagated between parent and offspring nests in both northern and tropical habitats. Clonality is implicated in our study most strongly by the high frequency (87.3%) of Attamyces accessions that were collected as genetically identical genotypes in at least two different At. texana nests. The most abundant Attamyces genotypes were collected across vast ranges (15 000–80 000 km2, see the electronic supplementary material, table S1). Because our population-genetic markers mutate at rates typical for microsatellite loci , the large ranges of some Attamyces clones indicate that the spread of Attamyces clones across hundreds of kilometres can occur rapidly relative to the rates of endogenous (mutational) change. A second parallel between northern and tropical Attamyces populations is that Attamyces lineages are shared between leaf-cutter ant species, exemplified most prominently by the sharing of M-group Attamyces between four leaf-cutter species in North America. A third parallel is that Attamyces lineages occasionally recombine (implicated by the observation of admixed Attamyces genotypes); Attamyces propagation is, therefore, not strictly clonal across an endless series of host generations as proposed by initial phylogenetic studies of Attamyces . These three basic features of Attamyces biology had already emerged in previous population-genetic and phylogenetic analyses of much smaller Attamyces datasets from Central and South America ([37,38,41]; see also discussions in [31,40]).
(b) Differences between northern and tropical Attamyces populations
Several features emerged in the northernmost Attamyces populations that have so far not been documented for Central and South American populations: (i) cultivation of specific Attamyces lineages by only a single ant host (i.e. ant–fungus specializations that tighten ant–fungus coevolution); (ii) regional or ant-specific differences in admixture between Attamyces lineages; and (iii) dispersal by Attamyces over significant distances and across ocean barriers, indicated most prominently by the surprisingly close genotypic proximity between Attamyces representatives from Cuba and Mexico (figure 1). These three differences are discussed in detail in the following sections.
(i) Host–symbiont specializations tighten coevolution at the northern frontier
As expected for peripheral populations, genotypically differentiated cultivar lineages were indeed found in the northernmost leaf-cutter populations, giving rise to specialized host–symbiont combinations. The evidence for host–symbiont specialization is strongest for T-group Attamyces, which emerged as the most distinct symbiont type in the present study (figure 1). This implies that, whereas ant–Attamyces coevolution is rather diffuse in the tropics (several leaf-cutter species interact with several Attamyces lineages in a complex interaction network [38,40,41]), coevolution is tighter for some ant–fungus combinations at the northern range limit. The most prominent cases of coevolutionary tightening were found, first, in Louisiana where At. texana associates with only one narrow subgroup of T-group Attamyces (figure 2b), and second, in California and northern Arizona where Ac. versicolor associates with only a narrow subset of V-group Attamyces (figure 2a). Moreover, among the Attamyces cultivated by At. texana, T-group Attamyces appear to be more tightly coevolving with the host At. texana (because of the exclusive association of this symbiont type only with At. texana) compared with M-group Attamyces (which have close genetic links to Attamyces cultivated by other leaf-cutter species (figure 1). A trendin declining genetic diversity of the cultivated fungi towards the northern range limit and a corresponding tightening of coevolutionary interactions had also been observed in the northernmost fungus-growing ant Trachymyrmex septentrionalis (a non-leaf-cutter ant), which spread after the Pleistocene northwards through the eastern and central USA .
(ii) Admixture rates appear lower in temperate than in tropical Attamyces
Given the biology of United States leaf-cutter ants and their Attamyces fungi, it is surprising that our survey revealed few significantly admixed Attamyces accessions (three in At. texana and three in Ac. versicolor; see the electronic supplementary material, table S1) and that Bayesian modelling implicated very low frequencies of sexual reproduction per Attamyces generation (0.3% for At. texana and 0.6% for Ac. versicolor). Higher frequencies could be expected because of (i) the known capacity of Attamyces for recombination ( and this study; see also the electronic supplementary material Admixture for a detailed description of the diverse mechanism of genetic exchange in multi-nucleate fungi); (ii) the extensive overlap between T-group and M-group Attamyces across the range of At. texana (figure 2b) and the overlap of V-group and M-group Attamyces in southeast Arizona (figure 2a); (iii) the frequent proximity of neighbouring nests with Attamyces from distinct Attamyces-groups (frequently within 100 m of each other; see the electronic supplementary material, table S1); and (iv) polygynous nest-founding and thus possible cultivar mixing in both ant species (5% polygynous nest-founding in At. texana ; between 0 and 100% in Ac. versicolor, depending on the population ).
The low levels of observed recombinants between differentiated Attamyces groups in the northern range may have several, non-exclusive explanations. First, when Attamyces strains are co-propagated by the ants in a common garden or when a novel strain enters the monoculture of an established nest, recombination between differentiated Attamyces strains may be prevented by mycelium–mycelium antagonism (i.e. somatic incompatibility sensu , preventing anastomosis and migration of nuclei between the two mycelia). Second, Attamyces types may be maintained as genotypically differentiated, sympatric lineages because of nucleus–nucleus incompatibilities if nuclei are ever exchanged between two Attamyces strains (i.e. the two nuclei do not function together when residing in the same cell, leading to cell death; or nuclei may compete with each other, leading to elimination of competitively inferior nuclei). Third, Attamyces recombinants may exhibit hybrid inferiority relative to the unrecombined parental types (i.e. nests with hybrid Attamyces therefore may have reduced survivorship, reduced growth rates or reduced fecundity). These three explanations are not mutually exclusive, and they can be tested in laboratory experiments with experimental ant–fungus combinations or with artificially created chimaeric gardens .
The possible lower level of recombination (admixture) of Attamyces from At. texana and Ac. versicolor compared with Attamyces of tropical leafcutter hosts is intriguing and hopefully will stimulate follow-up work to test for this difference with a larger sample of tropical Attamyces. Both Ac. versicolor and At. texana can found their nests polygynously [16,50,53], whereas At. mexicana and At. cephalotes are strictly monogynous ([42,55]; U. G. Mueller 1991–2010, personal observations). If co-cultivation of Attamyces strains during polygynous nest-founding facilitates genetic exchanges, one would expect that the Attamyces of Ac. versicolor and At. texana actually show higher levels of admixture, whereas we found the opposite pattern. This implicates processes other than co-cultivation as primary factors facilitating genetic exchanges between Attamyces lineages.
(iii) Attamyces fungi can disperse independently of the ant hosts across marine barriers
Although the link between mainland and Cuban Attamyces had already emerged in the phylogenetic analysis of Mikheyev et al. , our analysis documents the population-genetic proximity between mainland and Cuban Attamyces populations with greater resolution and for a more comprehensive sample (220 Attamyces strains from North America). Cuba, as part of the proto-Greater Antilles, moved through the gap between North and South America at the end of the Cretaceous (over 60 Myr ago) and may have temporarily abutted what is now Southern Mexico [17,56]. Since the early Tertiary, however, Cuba is thought to have remained separate from the mainland and without temporary land-bridges during changes in sea levels [17,56]. In the absence of gene flow across the ocean between mainland and Cuba, the Attamyces populations in Cuba, therefore, should have become genetically differentiated from the mainland populations, a prediction that is not supported by the observed population-genetic patterns (figure 1). Because the host At. insularis occurs only in Cuba but not on mainland North America, and because At. insularis is significantly derived at both the molecular and morphological levels from the mainland Atta lineages (, suggesting significant evolutionary time to permit divergence of At. insularis), it seems unlikely that the current genetic similarity between mainland and Cuban Attamyces was maintained during the divergence process of At. insularis without gene flow in Attamyces. Consequently, recent or ongoing gene flow in Attamyces between the mainland and Cuba is a more likely explanation for the absence of genetic differentiation between mainland and Cuban Attamyces. Such gene flow could occur via three possible avenues.
Leaf-cutter queens occasionally disperse across the ocean barrier from the mainland to Cuba (e.g. across the 250 km wide Yucatan Channel between the Yucatan Peninsula in Mexico and the Guanahacabibes Peninsula in Cuba); these queens vector Attamyces to Cuba, establish nests long enough to permit Attamyces to be transferred to At. insularis, then these temporary colonizers become extinct. Although other poorly dispersing ant species managed to colonize many islands throughout the Caribbean , we consider this scenario unlikely, primarily because of the rarity of observed leafcutter dispersal across Caribbean islands .
Similar to (i) but assuming human-mediated dispersal, Attamyces fungi were recently introduced to Cuba with leaf-cutter ants brought accidentally by humans from the mainland (e.g. in soil of potted plants); once in Cuba, introduced Attamyces replaced pre-existing cultivars in At. insularis nests. This scenario would receive support if the only other leaf-cutter ant species in Cuba (Acromyrmex octospinosus) can be shown to have colonized Cuba recently, which is not supported by available genetic information (i.e. the Cuban At. octospinosus populations appear genetically differentiated from mainland populations ).
Attamyces is able to disperse independently of the leaf-cutter hosts, either by airborne spore dispersal (spore-bearing mushrooms of Attamyces have been observed on rare occasions; table 3 in Mueller  summarizes the relevant literature), or by dispersal with the help of unknown vectors (e.g. by one of the many arthropods frequenting leaf-cutter nests, such as mites, beetles or moths; [10,15,59]).
Irrespective of these unresolved dispersal mechanisms, the close population-genetic links between mainland and Cuban Attamyces support the view that tropical Attamyces cultivars may be able to survive independently of leaf-cutter ants at least for some time (e.g. as spores, or when dispersed by vectors other than ants). Consequently, Attamyces fungi may be more than enslaved domesticates tied inescapably to the fate of their particular leaf-cutter host [26,37,38,40]. If so, long-distance dispersal and genetic mixing of Attamyces, including oceanic dispersal independent of leaf-cutter ants, may occur throughout the range of Attamyces. Such long-distance dispersal ability of Attamyces would contrast with the dispersal ability of Termitomyces fungi of the termite–Termitomyces symbiosis, in which effective long-distance dispersal of Termitomyces across oceanic barriers appears to be rare (e.g. between mainland Africa and Madagascar ). However, long-distance dispersal and genetic mixing of Attamyces could help explain the much younger coalescence and origin of Attamyces fungi (about 2–4 Myr old) compared with the corresponding origin of leaf-cutter ants (about 8–12 Myr old) .
Peripheral populations of the leaf-cutter ant–Attamyces symbiosis emerge as particularly interesting test cases of coevolution between ant farmers and their cultivated fungi, for two reasons. First, the documented local reduction of ant–species diversity and Attamyces diversity at the northern range limit increases local ant–Attamyces specificity; increased specificity, in turn, should lead to tighter, local coevolutionary interactions. Second, because novel Attamyces variants can sweep through the range of leaf-cutter ants , peripheral populations are least likely to be reached by novel Attamyces variants. These demographic processes should affect peripheral and island populations alike; however, because of dispersal across oceanic barriers, marine island populations are no longer the most promising candidates for finding unusual Attamyces variants. Specifically, if the Cuban populations of Attamyces analysed here are representative, we predict that Caribbean island populations of Attamyces known from Trinidad , Guadeloupe  and Curacao  will likewise retain strong population-genetic links to continental populations. Instead, populations at the periphery of the leaf-cutter distribution (in the USA, in Argentina, perhaps at mid-elevation sites in the Andes) are most likely to reveal unique and ant–species-specific leaf-cutter ant–Attamyces associations. These peripheral populations therefore contribute the most distinct geographical variants in coevolutionary interactions between leaf-cutter ants and their cultivated fungi. Studies of such marginal populations will inform models of coextinctions of obligate symbiotic partners that are doubly stressed by habitat marginality and by environmental change [8,9].
We thank numerous landowners for permission to collect on their properties; Kevin Anderson and John Crutchfield for permission to work at the Hornsby Bend Environmental Research Center and at the Brackenridge Field Laboratory; Kevin Anderson, Don Bass, Family Blackstock, Jason Byrd, Joy Camp, Craig Clark, Rebecca Clark, James Cokendolpher, Kenneth Conaway, Jerry Cook, Stefen Cover, Shawn Dash, Lloyd Davis, Mary Dixson, Diane Dudney, Evan Economo, Tim Fennell, Brush Freeman, Patricia Garlough, Nicole Gerado, Larry Gilbert, Paula Gilleland, Dale Groom, Don Grosman, David Hall, Ray Hampton, Ken Helms, Anna Himler, Rick Knipe, Jean-Nick Jasmin, Bob Johnson, Donald Johnson, Carol Jones, Charles Jones, Joey Jordan, Steve Lanoux, David Long, Malcolm MaCallum, Callen McWilliams, Marcy Marsden, Jack Martin, Mike Merchant, Alex Mintzer, John Moser, Scott Myres, Christian Rabeling, Ed Riley, Justin Rhodes, Andre Rodrigues, Jerod Romine, Ron Schmidt, Jarrod Scott, Carrol Simanek, Gordon Snelling, Roy Snelling, Jeff Sparks, Tracy Villareal, Roy Vogtsberger, Gerald Walls, Debbie Walker, Gordon Walton, Phil Ward, Stuart West, Alex Wild and Royce Wright for collection information; Damian Broglie and Christian Rabeling for drawing electronic supplementary material, figure S1; and Jon Seal, Duur Aanen and two anonymous reviewers for comments on the manuscript. The work was supported by the National Science Foundation (DEB-0920138, DEB-0639879 and DEB-0110073 to U.G.M.; DEB-0949689 to T. R. Schultz, N. J. Mehdiabadi and U.G.M.), the EEB Fellowship Endowment (to A.S.M. and S.E.S.) and the W. M. Wheeler Endowment from The University of Texas at Austin.
- Received January 19, 2011.
- Accepted February 14, 2011.
- This journal is © 2011 The Royal Society