Geographic structure in a widespread plant–mycorrhizal interaction: pines and false truffles


Jason D. Hoeksema, National Evolutionary Synthesis Center, 2024 W. Main Street, Suite A200, Durham, NC 27705, USA.
Tel.: +1 919 668 4582; fax: +1 919 668 9198;


Mutualistic interactions are likely to exhibit a strong geographic mosaic in their coevolutionary dynamics, but the structure of geographic variation in these interactions is much more poorly characterized than in host–parasite interactions. We used a cross-inoculation experiment to characterize the scales and patterns at which geographic structure has evolved in an interaction between three pine species and one ectomycorrhizal fungus species along the west coast of North America. We found substantial and contrasting patterns of geographic interaction structure for the plants and fungi. The fungi exhibited a clinal pattern of local adaptation to their host plants across the geographic range of three coastal pines. In contrast, plant growth parameters were unaffected by fungal variation, but varied among plant populations and species. Both plant and fungal performance measures varied strongly with latitude. This set of results indicates that in such widespread species interactions, interacting species may evolve asymmetrically in a geographic mosaic because of differing evolutionary responses to clinally varying biotic and abiotic factors.


Putatively mutualistic interactions between species are common in all biological communities, and recent studies show that they are highly variable in their ecological outcomes and potential coevolutionary dynamics. Ecological outcomes between pairs of species can range from mutualism to parasitism (e.g. Johnson et al., 1997; Thompson & Cunningham, 2002), and even simple models of mutualistic interactions predict evolutionary outcomes ranging from fixation of single mutualistic genotypes through positive frequency-dependent selection (e.g. Law, 1985; Parker, 1999) to fluctuating polymorphisms driven by negative frequency-dependence (Bever, 1999). Temporal and spatial variation in abiotic and biotic conditions can add further complexity to the structure of mutualism (Thompson, 1988; Bronstein, 1994a,b; Johnson et al., 1997; Nuismer et al., 2003a). As predicted by the geographic mosaic theory of coevolution (Thompson, 1994, 2005), this complexity leads to varying coevolutionary selection on the species involved. Consequently, mutualistic interactions are likely to exhibit a strong spatial structure in their coevolutionary dynamics, resulting in variable patterns of local adaptation and maladaptation.

Geographic mosaics in coevolving interactions have now been shown both in mathematical models (e.g. Nuismer et al., 2003b) and in an increasing range of empirical studies (e.g. Brodie et al., 2002; Thompson & Cunningham, 2002; Benkman, 2003; Laine, 2005). A geographic mosaic in a coevolving species interaction results from a combination of three factors: (1) geographic variation in the structure of selection, e.g. ranging from mutualism to parasitism, (2) geographic variation in the intensity of coevolutionary selection, resulting in coevolutionary hotspots and coldspots and (3) trait remixing because of gene flow, random genetic drift and metapopulation dynamics (Thompson, 1994, 2005). As a result of this combination of factors, species are expected to exhibit significant geographic structure with respect to traits important for the coevolving interaction.

Through this process, local adaptation can evolve at multiple spatial scales: parasites can be adapted to individual host genotypes (Van Zandt & Mopper, 1998), to specific host populations (Lively et al., 2004), or to regional collections of host populations (e.g. Thrall et al., 2002). The geographic mosaic of coevolution can result in maladaptation between some populations of interacting species (e.g. Kaltz et al., 1999) or complex geographic interaction structures with no evidence for local adaptation (e.g. Burdon et al., 1999). Furthermore, spatial patterns of local adaptation can range from discrete to clinal. In the case of discrete local adaptation, performance of one or both species is distinctly higher in local, sympatric pairings of populations compared with most or all nonlocal or allopatric pairings (e.g. Lively et al., 2004). Other species interactions exhibit a clinal pattern of local adaptation, in which performance gradually declines in pairings between more and more distant populations (e.g. Ebert, 1994), or in which traits of an interaction vary along a latitudinal cline (e.g. Toju & Sota, 2006). Indeed, mathematical models of the geographic mosaic of coevolution have suggested that such clinal patterns should be common in widespread interactions (e.g. Nuismer et al., 2000), and that both local adaptation and maladaptation should be common (Nuismer et al., 1999); nevertheless, we have few datasets collected across large spatial scales with which to test these predictions and inform efforts to refine these models, especially for interactions that vary along the mutualism–parasitism continuum. Furthermore, most published analyses of cross-inoculation experiments only test for discrete patterns of local adaptation and not for clinal patterns.

When species interactions involve multiple species on one or both sides of the interaction, complex networks of coevolutionary dynamics are possible and local adaptation or maladaptation can develop at multiple phylogenetic scales. For example, a symbiont species interacting with multiple host species could evolve host-specific races, could adapt to individual host populations regardless of host species, or could adapt to multiple specific host lineages within populations. Investigations of local adaptation at multiple phylogenetic scales have been rare, and are needed in order to understand coevolutionary dynamics in diverse networks of species interactions.

The interactions between coniferous plants and ectomycorrhizal fungi are particularly amenable to analysis of the spatial scale, taxonomic scale, and spatial pattern of coevolving geographic structure in putatively mutualistic species interactions. Conifers typically interact with diverse suites of ectomycorrhizal fungi, and most ectomycorrhizal fungi interact with multiple host plant species (Horton & Bruns, 2001). Classically, the fungi act as mutualists, extending from host plant root systems into the soil, providing enhanced access to soil nutrients in exchange for host photosynthate (Smith & Read, 1997); however, the net ecological outcome of the interaction for the plant can range widely from mutualism to parasitism, depending on the taxa involved and on environmental conditions (e.g. Jones et al., 1990; Gorissen & Kuyper, 2000). Furthermore, many of these interactions are widely distributed, and thus are likely to vary in ecological outcome and consequent coevolutionary dynamics along significant environmental gradients.

Here, we report results from an experimental investigation of geographic structure in the interaction between pines and false truffles along the west coast of North America. Three coastal pine species occur in a partially overlapping pattern along a narrow, north-to-south linear gradient from Alaska to Mexico, interacting with many of the same ectomycorrhizal fungi throughout much of this range, including members of the genus Rhizopogon (false truffles). This spatial structure provides an opportunity to investigate the spatial scale of evolving geographic interaction structure in a widespread putative mutualism.

Previously, ectomycorrhizal fungi in the genus Rhizopogon have been deemed to have evolved host-specificity at or above the level of host plant genus. Molecular data suggest that the genus Rhizopogon has diversified into five major sub-genera (Grubisha et al., 2002), and three of these sub-genera, Rhizopogon, Roseoli and Versicolores, are considered to be specialists on the plant genus Pinus (Molina et al., 1999). We hypothesized that, because these ectomycorrhizal interactions occur across a broad geographic range and a wide variety of ecological conditions, they are likely to have evolved in a geographic mosaic, with significant variation in compatibility between the fungi and their Pinus host plants at multiple spatial and phylogenetic scales and in complex spatial patterns. Our goal was to evaluate how large-scale geographic patterns of adaptation may develop in a symbiont species that interacts with three host species that are mostly allopatric but collectively form an almost continuous latitudinal gradient of host availability.

We performed a cross-inoculation experiment with multiple populations of the three coastal pine species and multiple populations of the false truffle R. occidentalis. The experiment was designed to characterize the geographic structure of compatibility between the plants and the fungus, especially potential clinal geographic patterns of interaction structure. We measured performance of both the host and the symbionts, to facilitate testing whether the host and symbionts have evolved differently in response to the interaction.


Experimental organisms

Shore pine (Pinus contorta var. contorta Dougl. ex Loud.), bishop pine (P. muricata D. Don) and Monterey pine (P. radiata D. Don) all occur in monodominant to near-monodominant stands along the West Coast of North America. Shore pine has the broadest native range of the three, inhabiting a largely continuous stretch of coastal muskeg and dune habitats from south-east Alaska to northern California. Bishop pine and Monterey pine, which are close sister species within the genus Pinus (see, e.g. Grotkopp et al., 2004), are locally dominant but geographically restricted in their native range, occurring in only a few isolated coastal mainland and island populations in California and northern Mexico.

Ectomycorrhizal fungi in the genus Rhizopogon Fries (Basidiomycota, Rhizopogonaceae) form mycorrhizae with host conifers throughout all stages of forest succession; however, in coastal pine forests many Rhizopogon species seem to be especially common at the edges of these habitats, or after disturbances such as fire, during seedling establishment and early growth. Rhizopogon species produce meiotic spores in rounded, enclosed sporocarps commonly called ‘false truffles’, which develop at or below the soil surface (Fig. 1) and have evolved to be eaten and dispersed by mammals (Molina et al., 1999; Trappe & Claridge, 2005). Many Rhizopogon spores seem to persist in a soil ‘spore-bank’ akin to seed banks in plants (e.g. Baar et al., 1999; Taylor & Bruns, 1999). We focused our study on Rhizopogon occidentalis Zeller & Dodge, which is one of the most common constituents of the spore-bank guild of ectomycorrhizal fungi in West Coast pine forests (Taylor & Bruns, 1999; Horton, 2002). Rhizopogon species are unusual among ectomycorrhizal fungi in being particularly amenable to experimentation; their spores can be used to directly inoculate numerous host plants with no prior fungal culturing necessary (Castellano & Molina, 1989; Molina et al., 1999).

Figure 1.

 Seeds were collected from nine different pine populations of three different pine species, and sporocarps were collected from four different populations of the ectomycorrhizal fungus, Rhizopogon occidentalis, on the west coast of North America. From two of the sites, the Oregon Dunes shore pine site (ORDU) and the Ano Nuevo, California Monterey pine site (ANCA), both seeds and fungi were collected.

Collection of pine seeds and fungal spores

Maternal families of seeds from two different haphazardly chosen trees were collected from each of three shore pine populations, four Monterey pine populations and two bishop pine populations (Table 1, Fig. 1). From two of these populations – shore pine forest in the Oregon dunes and Monterey pine forest at Ano Nuevo, California – two different R. occidentalis sporocarps were collected (Table 1, Fig. 1). In addition, two R. occidentalis sporocarps were obtained from two different bishop pine populations (Pt. Reyes and Santa Cruz Island, California; Table 1, Fig. 1) from which pine seeds were not collected. Pairs of sporocarps within each population were collected at least 30 m apart, so that each sporocarp was most likely produced by a different fungal genet. Fungi from the first three populations were collected between December 2002 and January 2003. From each of these sporocarps, a subset of the tissue was saved for deposition as a voucher collection in the Natural History Museum at the University of California, Santa Cruz, and the rest of the material was homogenized in de-ionized water and refrigerated until the time of inoculation of the experiment (c. 6 months). The fungi from Santa Cruz Island (provided by L. Grubisha) were collected and dried in January 2002, and then were re-hydrated for use in the experiment, which was initiated in April 2003. Voucher specimens of the latter collections are deposited on the Mycological Herbarium at Oregon State University.

Table 1.   Sites of collection of pine seeds and Rhizopogon occidentalis sporocarps. Accession codes are included for the fungi.
SiteDominant pine speciesLatitude and longitudePine seeds collected?Fungal sporocarps collected?
Mitkof Island, southeast Alaska (SEAK)Shore pine56°48′63′′N
Oregon Dunes National Recreation Area (ORDU)Shore pine43°47′92′′N
XX (JDH 27 and JDH 28)
Mendocino County, California (MOCA)Shore pine38°57′61′′N
Salt Point State Park, California (SNCA)Bishop pine38°34′80′′N
Pt. Reyes National Seashore, California (PRCA)Bishop pine38°03′46′′N
 X (JDH 44 and JDH 45)
Ano Nuevo, California (ANCA)Monterey pine37°03′99′′N
XX (JDH 61 and JDH 62)
Pt. Lobos State Park, California (MTCA)Monterey pine36°30′97′′N
Montana de Oro State Park, California (SLCA)Bishop pine35°16′04′′N
Santa Cruz Island, California (SICA)Bishop pine34°04′00′′N
 X (LG 1095 and LG 1099)
Guadalupe Island, Mexico (GUMX)Monterey pine29°8′87′′N
Cedros Island, Mexico (CEMX)Monterey pine28°07′00′′N

Experimental set-up

Cross-inoculation experiments, in which the performances of hosts and symbionts from a large number of widely distributed populations are tested in sympatric and allopatric combinations in a common environment, represent a useful approach to quantifying geographic structure in species interactions (Gandon & Van Zandt, 1998; Kawecki & Ebert, 2004). We paired plants from nine pine populations with fungi from four fungal populations to investigate the spatial and taxonomic scale of geographic structure in their compatibility. All factorial combinations of plant and fungal populations were included in the experiment, except for the pairings of fungi from the bishop pine forest at Pt. Reyes, California, with plants from two of the other bishop pine populations (Salt Point and Montana de Oro State Park, California). Thus, 34 different plant population × fungus population pairings were created, including two that paired sympatric plants and fungi (Oregon Dunes shore pine forest and Ano Nuevo, CA Monterey pine forest). Allopatric pairings ranged in geographic distance from c. 73 km (fungi from Ano Nuevo Monterey pine forest paired with Monterey pines from Monterey, CA) to c. 2723 km (fungi from Santa Cruz Island bishop pine forest paired with shore pines from SE Alaska). For each plant population by fungus population combination, each of the two maternal seed families was randomly paired with one of the two fungal sporocarps, and each of these pairings was replicated three times (i.e. with three different inoculated seedlings) (Fig. 2).

Figure 2.

 Pairings between plant and fungal populations were created by inoculating each of two different maternal seed families from a plant population with spores from one of two different fungal sporocarps from a fungal population. Each pairing between a sporocarp and a maternal seed family was replicated three times.

An important reason for choosing R. occidentalis for this study is that it is one of the few mycorrhizal fungi for which data exist on population genetic structure (L. Grubisha, unpublished data). Analyses of microsatellite markers from some of the same populations we studied have shown that for many of these fungal populations, much more neutral genetic variation exists between populations compared to within populations (L. Grubisha, personal communication). Thus, we chose to allocate more of our sampling effort and experimental design to capturing potential between-population variation in ecological traits of the plant–fungus interaction. Within each population, we sampled seeds from two trees and spores from two sporocarps, providing us with genetically variable, sexually produced offspring for both species. This design results in a greater sampling of genetic diversity through paternal contribution than maternal contribution, and represents a compromise between sampling more populations and more individuals per population.

Seeds were surface-sterilized in 1% sodium hypochlorite solution, thoroughly rinsed and cold-stratified for 4 weeks. Seeds were germinated in a thin layer of autoclaved peat/perlite (c. 70%/30%) potting medium (Promix HP; Premier Horticulture, Inc., Dorval, Quebec, Canada) and grown for 6–8 weeks. The manufacturer reports Promix HP as having the following characteristics: pH 5.2–6.2, electrical conductivity 1.3–2.0 m\mho cm−1, air porosity 20–25% by volume, water-holding capacity 50–70% by volume, bulk density 0.13–0.16 g cm−3, 70–130 ppm nitrate, 25–80 ppm phosphate, 70–150 ppm potassium, 130–210 ppm calcium and 20–45 ppm magnesium. Seedlings were selected for uniformity of size, carefully removed from the germinating flats, and their roots dipped in an aqueous slurry of spores (containing c. 2.2 × 107 spores mL−1) made from the appropriate fungal sporocarp. Seedlings were then planted in cylindrical pots (5 cm diameter × 18 cm deep, D16 Deepot Cells; Stuewe & Sons, Inc., Corvallis, Oregon, U.S.A.) and randomized in a grid, with each pot set inside of another to protect from potential contamination by water from below. Growth chamber conditions were set with a 10-h night-time period at 10 °C and a 14-h day at 23 °C, with c. 225 μmol m−2 s−1 of light at plant height. To maximize opportunities for fungal colonization in each pot, after 6 weeks each pot was re-inoculated with the appropriate fungus by adding c. 107 spores diluted in 5 mL of de-ionized water. Approximately 5 weeks later, each pot was fertilized by adding 13.5 mL of 1/2 strength modified Hoagland's solution (0.815 g L−1 Hoaglands modified basal salt mixture; Phytotechnology Laboratories, Shawnee Mission, Kansas, U.S.A.). Every 6 weeks, the entire experiment was completely re-randomized within the growth chamber to minimize the influence of any heterogeneity of conditions within the chamber.

Data collection

At the time of initial mycorrhizal inoculation, we measured the length of the needle-bearing stem on each plant, a measurement we have found previously to be predictive of total dry mass. We then used previously established regression equations of total dry mass (g) on needle-bearing stem length (green length, mm) for each pine species to estimate total dry mass of each plant at the time of inoculation (shore pine: ln(mass) = ln(green length)*1.396 − 5.646, n = 93, R2 = 0.65; bishop pine: ln(mass) = ln(green length)*1.97 − 7.823, n = 14, R2 =0.80; Monterey pine: ln(mass) = ln(green length)*1.64–6.659, n = 24, R2 = 0.60). Approximately 24 weeks after seedlings were initially transplanted into pots and inoculated, the experiment was harvested: plants were removed from pots, and the substrate was carefully washed from the roots. Mycorrhizal fungus performance was estimated by counting the number of pine root tips colonized by the fungus, and plant root length was estimated using a grid-intersect method (Newman, 1966). A root tip was scored as colonized if a fungal mantle was visible, and a subset of these root tips was checked to confirm the presence of a Hartig net. Plant leaf, stem and root biomass were separated, dried at 60 °C for 48 h, and weighed. Plant relative growth rate (RGR) from inoculation to the end of the experiment was estimated as [ln(m2) − ln(m1)]/(no. of days), where m1 is estimated total dry mass at time of inoculation and m2 is measured total dry mass at the end of the experiment.

Final leaf mass, final root length and total estimated RGR were used as measures of plant performance. RGR is thought to be an important measure of plant performance, as it integrates a variety of plant physiological components and is independent of plant size (Hunt, 1990; Hunt & Cornelissen, 1997). Root length may be an indicator of competitive ability for soil resources (Craine et al., 2005), and leaf mass should be positively associated with photosynthetic potential. Early performance of seedlings has been found to be an important predictor of later success in field studies of demography in pine populations (Landis et al., 2005) and a variety of other species (reviewed by Harper, 1977; e.g. Young, 1985). We also calculated plant root : shoot biomass ratio because it is often predicted to vary significantly among plants depending on the relative importance of limitation by aboveground and belowground resources (Bloom et al., 1985; Tilman, 1988; Aikio & Markkola, 2002).

As Rhizopogon species develop on the roots of pine hosts, they form clusters of dichotomously branched, colonized root tips, and each colonized root tip is a site of nutrient exchange between the plant and the fungus. As an estimate of fungal performance we used the number of pine root tips colonized (irrespective of clustering because the number of colonized tips represents the number of sites for potential nutrient exchange), expressed on a per unit root length basis in order to control for effects of available root system size. In these pine seedlings, new root tips susceptible to colonization can and do emerge from all parts of the root system, including from the thickest part of the root near the root collar itself (J.D. Hoeksema, personal observation); therefore, total root length is a reasonable proxy for the potential number of colonization sites for the fungi.

It is inherently problematic to quantify the fitness of clonal soil microbes such as mycorrhizal fungi (see Pringle & Taylor, 2002 for a discussion of the problem). Nevertheless, ectomycorrhizal fungi like Rhizopogon are obligate symbionts on their host plants, and as a consequence, the extent to which they colonize their hosts is expected to positively correlate with their ability to obtain fixed carbon for growth and sexual reproduction. Thus, the number of root tips colonized by an ectomycorrhizal fungus is likely to be positively correlated with fungal fitness (although not necessarily with plant fitness; Nara, 2006). When controlling for the size of the host plant's root system, the extent of fungal colonization is expected to reflect overall compatibility between the host plant and the fungus. One consequence of using blended fungal sporocarp material to inoculate each replicate pine seedling is that the compatibility expressed between plant and fungus in each experimental unit is necessarily a product of an interaction between the seedling and the most successful spore genotypes present in a pot (although the level of genetic diversity among spores within a sporocarps is unknown). Furthermore, the plant and fungal growth responses we measured are all thought to be quantitative traits, which decreases the likelihood of obtaining an extremely nonrepresentative sample of a trait from a population. Thus, observing low compatibility in a particular plant × fungus population combination is significant, indicating an overall lack of compatibility between numerous combinations of plant and fungal genotypes from those populations. Although we could not directly measure fitness of the plants and fungi in these experiments, our measures of success are likely to be correlated with fitness, and thus likely to inform arguments about adaptation (Pringle & Taylor, 2002).

In addition to the inoculated mycorrhizal fungus, R. occidentalis, some plants were colonized by an additional nontarget ectomycorrhizal fungus (exhibiting a smooth brown mantle morphology – photographs available upon request), which most likely entered the growth chamber via airborne spores. Approximately 75% of the replicates were scored for the abundance of the contaminant, on a scale of 0–3 (0 = none; 1 = sparse, i.e. fewer than 10 root tips colonized; 2 = moderate, i.e. 10–30 root tips colonized; 3 = common, i.e. more than 30 root tips colonized). The large majority (80%) of scored replicates were either not colonized by the contaminant or were sparsely colonized. Prevalence of the contaminant did not vary across plant or fungal populations, but did differ significantly among the three plant species, with shore pine experiencing almost no colonization, and significantly less than the other two plant species. Across all scored pots, there was no significant correlation between contamination score and colonization by the target fungus, R. occidentalis, and noncolonized root tips were present in all replicates, suggesting the potential for additional colonization by the target fungus despite the presence of the contaminant. Thus, contamination by other microbial species is unlikely to have impacted colonization by the target fungus, R. occidentalis.

Data analysis: discrete geographic interaction structure

We examined geographic structure in plant and fungal performance at multiple scales, using a nested, hierarchical statistical model. Specifically, we used a mixed, cross-nested model with plant species, plant population nested within plant species, fungal population, and the two-way interactions between fungal population and the other two factors as fixed factors. Plant maternal seed family and fungal sporocarp were treated as random factors nested within plant and fungal population, respectively. We used restricted maximum likelihood estimation of parameters for all analyses (MIXED procedure in SAS v.8; SAS Institute, Inc., Cary, North Carolina, U.S.A.). When significant interactions were found, for example between plant population and fungal population, specific hypotheses of local adaptation were tested using a series of contrasts for the whole dataset and for each plant population, plant species, or fungal population separately (as in Thrall et al., 2002). For example, we tested for local adaptation of fungal populations to their sympatric plant populations by performing three different types of contrasts in which local adaptation is indicated by significantly higher performance in sympatric pairings compared with allopatric pairings. First, fungal colonization across the two sympatric pairings was compared with fungal colonization across all of the allopatric pairings in the study. Secondly, colonization by each fungal population with their sympatric plant population was compared with pairings of those fungi with allopatric plant populations. Thirdly, fungal colonization by each fungal population with their sympatric plant population was compared with colonization by the other fungal populations with that same plant population.

Pair-wise separations of means were made by comparing least-squares means with t-tests, and P-values were adjusted for multiple comparisons using the Tukey HSD procedure. Plant final leaf mass, root length and root :shoot ratio were log-transformed to achieve normality. Because mycorrhizal root tip colonization data could not be transformed to achieve normality, all fixed effect P-values reported for these data are from randomization tests (with 10 000 iterations) performed within SAS PROC MIXED by using a macro wrapper (Cassell, 2002). For analyses of plant growth, mean seed mass of maternal seed families was tested as a covariate, as were plant RGR and plant latitude in analyses of fungal colonization. All analyses of discrete geographic interaction structure were repeated for two different datasets: (1) the full dataset, including all three plant species and all four fungal populations and (2) a reduced, but fully factorial, dataset, excluding bishop pines and the two fungal populations found in bishop pine forest. A fully factorial design can potentially provide a more accurate test of geographic structure involving interactive effects of plants and fungi, such as local adaptation, because all possible combinations of plant and fungal populations are included. Because the full dataset provides information about more plant and fungal populations, however, we report only results for the full dataset when results do not differ qualitatively between these two analyses.

Data analysis: clinal geographic interaction structure

We explored potential clinal patterns of geographic interaction structure by analysing plant and fungal performance as a function of the geographic distance between the populations of plant and fungal origin in each experimental pairing (Ebert, 1994; Thrall et al., 2002; Lively et al., 2004). For analysis of plant growth, plant species and plant population (nested within plant species) as well as their interactions with geographic distance were included in the model. Similarly, fungal population and its interaction with geographic distance were included in analyses of fungal colonization. Analyses of clinal geographic interaction structure were performed with two different approaches, using either raw data for performance or mean values for each population-by-population pairing as the response variable. In the former approach, plant maternal seed family and fungal sporocarp were included as random factors nested within plant and fungal population, respectively. Results for the two approaches were always qualitatively similar, so we only present results using mean performance values. When significant main or interactive effects of geographic distance were found in these analyses, the relationship between performance and distance was explored across all or particular plant or fungal populations using simple linear regression. We were not attempting to detect interactive effects of plants and fungi in analyses of clinal geographic interaction structure, and so all analyses of clinal structure used the full dataset, including all three plant species and all four fungal populations.

Because our seed collection sites spanned a north–south transect across more than 28° of latitude, and previous research has found strong trends in plant growth parameters with latitude (e.g. Li et al., 1998), we also explored the relationship between mean performance measures and latitude of plant and fungal populations, using simple linear regression. Because the three host plant species are largely aligned on a north-to-south gradient, analyses of latitudinal patterns in this system may be somewhat redundant to simple comparisons of plant and fungal performance among the three plant species; however, not only do bishop and Monterey pine overlap in their distributions, but each pine species exhibits significant latitudinal spread, allowing for the possibility of latitude explaining variation in plant and fungal performance not captured by simply comparing means among the three plant species. Thus, to explore the effects of latitude beyond simple host plant species effects, we compared two different statistical models: a simple model with host plant species as the only explanatory variable vs. a model with both host plant species and latitude as explanatory variables. If both of these models were significant overall, we then compared the fit of the two models using a likelihood ratio test, concluding that latitude explains additional variation beyond plant species effects when the more complex model exhibited a significantly better fit than the simple model. In the latter cases, we then explored the impact of latitude on performance using simple linear regression. In the case of fungal performance, we could not fit models containing host plant species as the only explanatory variable for mean population performance, as host plant species is not replicated among the four fungal populations. In this case, we simply regressed fungal performance against latitude to explore the impact of latitude, with the knowledge that host plant species is one of the factors that varies with latitude.


Discrete geographic interaction structure

For the full dataset, fungal colonization averaged 0.98 (± 0.05 SE, n = 204) tips cm−1 of root, and only varied significantly across the three plant species (Table 2). Specifically, fungal colonization was significantly higher on bishop pine than on the other two species (bishop pine: 1.52 ± 0.17 SE, n = 36; Monterey pine: 0.91 ± 0.080 SE, n = 96; shore pine: 0.80 ± 0.068 SE, n = 72). Although fungal colonization was marginally influenced by interactive effects of fungus population with plant species and plant population (Table 2), contrasts testing for discrete local adaptation of fungi to their sympatric plant populations and for host-specificity on their sympatric plant species, both overall and for each fungal population separately, did not support any hypotheses of discrete local adaptation or host-specificity.

Table 2.   Results from statistical analysis of fungal colonization (root tips colonized per cm root length) in the full dataset, using restricted maximum likelihood estimation of parameters (SAS Proc MIXED).
Fixed effectsNumDFDenDFFP
Plant population (plant species)
Plant species28.29.350.004
Fungus population33.70.830.461
Fungus population*plant population (plant species)17158.21.610.076
Fungus population*plant species5158.22.070.078
Random effectsEstimateSEZP
  1. P-values for fixed effects are from randomization tests (using 10 000 iterations) performed within SAS PROC MIXED using a macro wrapper (Cassell, 2002). Type III tests of fixed effects and Wald tests of covariance parameter estimates for random effects are shown.

Plant maternal seed family (plant population*plant species)0.0480.0391.240.107
Fungus sporocarp (fungus population)0.2220.1731.290.099

In the analysis of the fully factorial dataset (which excluded bishop pines and the bishop pine populations of R. occidentalis) fungal colonization did not vary between the two plant species, shore pine and Monterey pine. We found, however, evidence for substantial geographic interaction structure, with fungal colonization varying significantly because of an interaction of fungus population and plant population (Table 3). Once again, contrasts provided no support for hypotheses of discrete local adaptation. Furthermore, the lack of a significant interaction between plant species and fungal population suggests a lack of host-specificity, as in the analysis of the full dataset. Rather, the Oregon Dunes fungi exhibited significant variation in colonization when paired with the seven different plant populations, and the Ano Nuevo fungi did not (Fig. 3). In contrast, at a smaller scale, in separate analyses for the two fungus populations the Ano Nuevo fungi exhibited significant variation in colonization between plant–fungus combinations within each plant population by fungus population pairing (Wald Z = 1.73, P = 0.0422) and the Oregon Dunes fungi did not (Wald Z = 0.97, P = 0.165).

Table 3.   Results from statistical analysis of fungal colonization (root tips colonized per cm root length) in the fully factorial dataset (lacking bishop pine and bishop-pine associated fungi), using restricted maximum likelihood estimation of parameters (SAS Proc MIXED).
Fixed effectsNumDFDenDFFP
Plant population (plant species)56.02.460.098
Plant species16.00.170.662
Fungus population11.90.780.395
Fungus population*plant population (plant species)562.03.150.015
Fungus population*plant species162.01.330.264
Random effectsEstimateSEZP
  1. P-values for fixed effects are from randomization tests (using 10 000 iterations) performed within SAS PROC MIXED using a macro wrapper (Cassell, 2002). Type III tests of fixed effects, and Wald tests of covariance parameter estimates for random effects are shown.

Plant maternal seed family (plant population*plant species)0.0430.0381.110.134
Fungus sporocarp (fungus population)0.2480.2610.950.171
Figure 3.

 Colonization (no. of root tips per cm root length) by two different populations of the ectomycorrhizal fungus, Rhizopogon occidentalis, across seven shore pine and Monterey pine populations. For the Ano Nuevo, CA fungal population, colonization was not significantly different across the seven plant populations. For the Oregon Dunes fungal population, means with different letters are significantly different (α = 0.05). Grey bars indicate shore pine populations, and white bars indicate Monterey pine populations. The two sympatric pairings are indicated with an ‘S’.

For the full dataset, all four measures of plant performance (RGR, final leaf mass, final root length and final root : shoot ratio) differed significantly across plant species and among populations nested within plant species, and effects involving fungi were never significant (Table 4). Specifically, RGR (overall mean =0.011 ± 0.005 SE) was significantly lower in shore pine than the other two species (Fig. 4a), final leaf mass (g) (overall mean = 0.31 ± 0.019 SE) was significantly higher in Monterey pine than in the other two species (Fig. 4b), and final root length (cm) (overall mean =568 ± 26 SE) was significantly higher in Monterey pine than bishop pine with shore pine intermediate between and not significantly different from the other two (Fig. 4c). Root : shoot ratio (overall mean = 1.03 ± 0.02 SE) was significantly higher in shore pine than in the other two species, intermediate in bishop pine, and significantly lower in Monterey pine than in the other two species (Fig. 4d). For the fully factorial dataset, plant growth again only varied across plant species and plant population; thus, the analysis of the fully factorial dataset was redundant with the analysis of the full dataset, and we do not report the results of that analysis. When plant maternal family mean seed mass was used as a covariate in these analyses, it did not explain a significant proportion of variation in plant growth in any analysis, and did not qualitatively change the importance of other factors in the analyses.

Table 4.   Results from statistical analysis of plant growth in the full dataset, using restricted maximum likelihood estimation of parameters (SAS Proc MIXED). Type III tests of fixed effects, and Wald tests of covariance parameter estimates for random effects are shown.
Fixed effectsNumDFDenDFFP
 Plant population (plant species)69.16.980.005
 Plant species29.639.11< 0.0001
 Fungus population3161.20.9230.431
 Fungus population*plant species5161.20.3100.906
 Fungus population*plant population (plant species)17161.20.7350.764
Random effectsEstimateSEZP
Plant maternal seed family (plant population*plant species)0.000001891.25E-061.520.064
Fungus sporocarp (fungus population)0nanana
Fixed effectsNumDFDenDFFP
Final leaf mass
 Plant population (plant species)65.911.340.005
 Plant species26.888.41< 0.0001
 Fungus population33.10.3340.804
 Fungus population*plant species5159.00.7030.622
 Fungus population*plant population (plant species)17159.02520.7750.719
Random effectsEstimateSEZP
Plant maternal seed family (plant population*plant species)0.001670.003650.4560.324
Fungus sporocarp (fungus population)0.007850.007980.9840.162
Fixed effectsNumDFDenDFFP
Final root length
 Plant population (plant species)67.36.0690.015
 Plant species28.07.3490.015
 Fungus population33.40.5000.705
 Fungus population*plant species5158.50.5630.728
 Fungus population*plant population (plant species)17158.51.2000.270
Random effectsEstimateSEZP
Plant maternal seed family (plant population*plant species)0.005590.005391.0380.150
Fungus sporocarp (fungus population)0.009250.008771.0540.146
Fixed effectsNumDFDenDFFP
Root : shoot ratio
 Plant population (plant species)68.43.620.0456
 Plant species29.441.7< 0.0001
 Fungus population33.91.420.361
 Fungus population*plant species5158.30.270.928
 Fungus population*plant population (plant species)17158.31.110.342
Random effectsEstimateSEZP
Plant maternal seed family (plant population*plant species)0.002920.003250.8970.185
Fungus sporocarp (fungus population)0.000110.001320.08200.467
Figure 4.

 Overall growth of three different pine species (a: relative growth rate, b: final leaf mass, c: final root length and d: final root : shoot ratio). Means with different letters are significantly different (α = 0.05).

Clinal geographic interaction structure

Mean fungal colonization significantly declined with increasing distance between plant and fungal populations of origin (main effect of distance F1,26 = 6.16, P = 0.0198; Fig. 5), suggesting the possibility of clinal local adaptation of fungal populations to plant populations. Fungal colonization did not vary across fungal populations (F3,26 = 1.5, P = 0.237) and there was no interaction between fungal population and distance (F3,26 = 0.12, P = 0.948). Plant RGR and plant latitude, when used as covariates in these models, did not explain significant variation in fungal colonization and did not impact the importance of other factors. Plant growth did not vary with distance between plant and fungal populations of origin. Rather, as in analyses of discrete structure, plant growth varied only among plant species and plant populations. Thus, these analyses are redundant to those reported above for the discrete model and are not reported here.

Figure 5.

 Decline in mean colonization of pines by the ectomycorrhizal fungus, Rhizopogon occidentalis, with increasing distance between plant and fungal populations in pairings. To illustrate the effect of distance, plotted data are residuals from a model lacking distance effects. Black markers indicate the Ano Nuevo, California (ANCA) population of fungi (Monterey pine), grey triangles indicate the Pt. Reyes National Seashore, California (PRCA) population of fungi (bishop pine), grey squares indicate the Santa Cruz Island, California (SICA) population of fungi (bishop pine) and white markers indicate the Oregon Dunes (ORDU) population of fungi (shore pine).

Some measures of mean plant population performance, as well as mean fungal population colonization levels, varied significantly with the latitude of the population. Specifically, fungal colonization increased significantly with fungal population latitude (Fig. 6a). Variation in both RGR and root : shoot ratio was significantly better explained by a model containing both plant species and latitude as explanatory variables, compared with a model containing only plant species (RGR: Likelihood ratio = 12.3, d.f. = 1, P = 0.0005; root : shoot ratio: Likelihood ratio = 6.3, d.f. = 1, P = 0.0121). Variation in final plant leaf mass was not better explained by a model containing latitude as an explanatory variable, compared to a model containing only host plant species (Likelihood ratio = 2.8, d.f. = 1, P = 0.0943), and neither model explained significant variation in final root length. In contrast to the pattern for fungal colonization per unit root length, plant RGR significantly declined with increasing plant population latitude (Fig. 6b). Root : shoot ratio increased significantly with latitude (Fig. 6c).

Figure 6.

 (a) Increase in mean fungal population root colonization (no. of root tips per cm root length), (b) decline in mean plant population relative growth rate and (c) increase in mean plant population root : shoot ratio with increasing latitude. Black markers indicate Monterey pine, grey markers indicate bishop pine, and white markers indicate shore pine.


The geographic mosaic of ectomycorrhizal interactions

Our results suggest the presence of significant evolved geographic structure at multiple scales in the interaction between false truffles and the three pine species that dominate coastal environments along the west coast of North America. The false truffle R. occidentalis exhibits substantial variation in colonization of host roots across host plant species, among host plant populations within species, and between different host–symbiont combinations within populations (Tables 2 and 3, Fig. 3); a portion of this variation is consistent with a clinal pattern of local adaptation to hosts (Fig. 5). In contrast, variation in plant performance was substantial, but appears to have evolved in response to biotic or abiotic factors that vary with latitude.

Previous discussions of variation in performance of Rhizopogon ectomycorrhizal fungi across different host plants have focused on patterns of specialization of Rhizopogon species and subspecies on particular plant genera such as Pinus and Pseudotsuga (Molina et al., 1992; Grubisha et al., 2002). Such patterns may hold true as aggregate results across particular fungal species, and indeed we found no evidence for host-specificity of R. occidentalis populations on their sympatric Pinus host species; however, our data reveal substantial variation in response of R. occidentalis to host plant variation, at multiple scales of host plant phylogeny and spatial distribution below the level of host plant genus – among Pinus species, among particular populations of pines and even among half-sib families within plant populations. For example, overall fungal colonization was higher on bishop pine than on the other two species (Table 2), and one fungus population differed in its response to different plant populations whereas another fungus population did not (Fig. 3). These particular results may be unique to the abiotic conditions used in our experiment (e.g. temperature, photoperiod and substrate), but they demonstrate that at least under some conditions, not all Pinus genotypes are equivalent partners for R. occidentalis.

In contrast, plant growth did not vary significantly among pairings with the four different fungal populations. Rather, significant differences in plant growth were primarily among Pinus populations and species. The Pinus species differed significantly in RGR, final leaf mass, final root length and root : shoot ratio (Table 4, Fig. 4), with Monterey pine exhibiting the highest values for the first three and the lowest value for the latter parameter. The result for RGR is consistent with a previous study of variation in growth rates among Pinus species (Grotkopp et al., 2002), which found that Monterey pine had the highest RGR among 29 pine species. In our study, variation in plant growth parameters was significant, independent of the identity of the ectomycorrhizal fungus symbionts.

Pattern and scale of local adaptation

In a widely distributed interaction between two species, adaptation by one species to local populations of the other species, and to local abiotic conditions, is predicted to generate substantial geographic structure at multiple scales and in a variety of patterns (Gandon, 1998; Kaltz & Shykoff, 1998; Kawecki & Ebert, 2004; Thompson, 2005). For example, individual demes of one species can be adapted to individual genotypes of another in a discrete pattern (Van Zandt & Mopper, 1998), populations of one species can be locally adapted to populations of another in a discrete pattern (Lively et al., 2004), or populations of one species can be adapted to those of another in a clinal pattern across large geographic distances (Ebert, 1994). We found no evidence that populations of R. occidentalis are locally adapted to their natal host plant species in a discrete pattern (i.e. host-specificity), no evidence for a discrete pattern of local adaptation of fungal populations to particular plant populations, and no evidence of any pattern of local adaptation (discrete or clinal) by pines to populations of R. occidentalis. Rather, the decline in average fungal colonization when fungi were paired with plant populations at increasing distances from their natal population (Fig. 5) suggests a degree of local adaptation by fungi to nearby plant populations in a clinal pattern, and supports the hypothesis that selection for local adaptation to host populations is indeed one force affecting the evolution of host-colonization traits in R. occidentalis. The almost continuous distribution of fungal colonization levels we observed, along with the clinal pattern of local adaptation, are consistent with recent reports that colonization levels in ectomycorrhizal interactions are quantitative traits under polygenic control (Tagu et al., 2005). Because our experimental design included only two sympatric pairings of plant and fungal populations, our power to detect discrete local adaptation was necessarily less than our power to detect clinal local adaptation, as the power of the latter analysis is largely a function of the distribution of allopatric pairings across distance. Thus, although our results do not suggest any pattern of discrete local adaptation, it would be desirable in future studies to test the robustness of this result by including additional sympatric pairings.

Plant and fungal performance varied significantly with latitude, and these effects of latitude were not simply due to the north–south distribution of the three pine host species. The significant and opposite changes with latitude for plant RGR vs. plant root : shoot ratio and fungal colonization (Fig. 6) suggest that environmental gradients associated with latitude have had different evolutionary impacts on the host and symbionts. At more northern latitudes, selection may have favoured increased allocation of plants belowground to root mass, perhaps at the expense of overall RGR, whereas selection acted to increase average colonization intensity in more northern fungal populations. Again, these results may be particular to the conditions used in our experiment; however, the patterns are striking enough, particularly in the case of fungal colonization levels, to warrant further investigations in multiple environments including field conditions. Trends of increasing or decreasing average growth rate with latitude have been found for a variety of taxa (e.g. Boehlert & Kappenman, 1980; Li et al., 1998; Jonassen et al., 2000), and hypothesized explanations for such patterns vary widely. One common explanation for increasing growth characteristics with latitude, such as we observed in the ectomycorrhizal fungus R. occidentalis (Fig. 6), is that faster growth rates allow organisms to make the most efficient use of the short growing season at higher latitudes. This could be true for more northerly populations of ectomycorrhizal fungi in coastal pine forests, especially as these more northerly fungal populations may tend to be associated with slower-growing plants than southern populations (Fig. 6). An alternative explanation is that lower colonization rates of more southerly populations of R. occidentalis indicate adaptation to lower productivity habitats, where mean annual rainfall is lower. Inherently slower growth may be an advantage in lower-productivity habitats, if faster-growing genotypes are inherently more nutrient-demanding and therefore suffer disproportionately greater reductions in growth under low-nutrient conditions (Hunt & Cornelissen, 1997).

In contrast to fungal colonization, plant RGR declined with increasing latitude (Fig. 6). Other studies within and across plant species have found a variety of trends of plant growth rates with latitude, ranging from strongly negative to strongly positive (Li et al., 1998; Ryser & Aeschlimann, 1999; Miyazawa & Lechowicz, 2004). Li et al. (1998), for example, also found a decrease in RGR with latitude across populations of Arabidopsis thaliana. They suggested that lower daily irradiation and average air temperatures may select for higher leaf area ratios (LAR) or specific leaf areas (SLA) in higher-latitude plant populations, and that a negative relationship between these variables and RGR produces a negative relationship between latitude and RGR in A. thaliana. We did not measure SLA or LAR in our study, but Grotkopp et al. (2002) found a positive relationship between RGR and both SLA and LAR across 29 different pine species, including the three we studied, suggesting that across pine species, selection for higher SLA or LAR would result in higher RGR. Thus, unlike in the Arabidopsis system (Li et al., 1998), indirect selection for lower RGR at higher latitudes would likely be driven by direct selection for lower SLA or LAR at higher latitudes. Perhaps in the coastal plant populations we studied, which are subject to high winds and salt deposition, survival may depend on thicker, well-protected leaves (with lower SLA and/or LAR) and slower growth rates, especially at higher latitudes. It would be ideal to test hypotheses about adaptation to the latitudinal gradient by directly measuring selection on plant and fungal traits in multiple populations along the gradient (as in, e.g. Toju & Sota, 2006 for a plant–insect interaction).

The few other comparable investigations of large-scale geographic interaction structure in putative nutritional mutualisms between plants and soil microbes, in which both plant and microbial populations were varied in a cross-inoculation experiment, have obtained results that are qualitatively similar to those we found here. Those studies reported significant geographic interaction structure and polymorphism in host–symbiont compatibility at multiple spatial and taxonomic scales, but little or no evidence for a discrete pattern of local adaptation (Acacia-rhizobia: Burdon et al., 1999; Thrall et al., 2000; plant-arbuscular mycorrhizal fungus: Klironomos, 2003; Sylvia et al., 2003). For example, Klironomos (2003) found substantial variation within and between co-occurring plant species in response to different arbuscular mycorrhizal fungi, ranging from mutualism to parasitism, and found a more extreme range of responses in sympatric combinations of plants and fungi than in allopatric combinations. Similarly, Acacia-rhizobia interactions in Australia seem to be characterized by significant polymorphism at multiple spatial and taxonomic scales, with no evidence for local adaptation (Burdon et al., 1999; Thrall et al., 2000). Such results are not consistent with simple models of positive frequency-dependent mutualistic selection within populations (e.g. Parker, 1999), which predict fixation of single mutualistic genotypes within local populations. Parker's (1999) model predicts a mosaic of symbiotic phenotypes among populations, but may be more appropriate for the examples of less diverse plant–rhizobia interactions that he discusses. Rather, such geographic structure in compatibility between species interacting in a putative nutritional mutualism is likely to be generated by a combination of spatially and temporally variable selection on the interaction, perhaps ranging from mutualism to parasitism, and gene flow among populations, as predicted by models of the geographic mosaic of coevolution (e.g. Nuismer et al., 2003b).

Asymmetries in local adaptation and geographic interaction structure

We found very different patterns of geographic variation in performance in R. occidentalis compared with its pine host plants. Specifically, we observed substantial geographic variation in both plant and fungal populations that affected fungal colonization (Tables 2 and 3, Figs 3 and 5); however, plant growth was influenced only by geographic variation within and between plant species (Table 4, Figs 4 and 6), and not by geographic variation in R. occidentalis. Even though it is possible that under different abiotic conditions, plant growth could be significantly affected by geographic variation in ectomycorrhizal fungi, this pattern illustrates how different selective forces are likely to influence the evolution of two interacting species due to asymmetries between them in evolutionary history or life history traits.

For instance, traits of Pinus host plants that could affect the performance of their obligate mycorrhizal symbionts, such as root proliferation rate or overall growth rate, are likely to have evolved independently towards polymorphism within and between pine populations, resulting in current diversifying selection on fungi (as discussed by Burdon et al., 1999 for the interaction between legumes and rhizobia). Grotkopp et al. (2004) found evidence that genome size is positively associated with RGR across Pinus species, and effects of phylogeny on RGR have been found in a variety of plant taxa including pines (e.g. Antunez et al., 2001; Grotkopp et al., 2002). In habitats with slow-growing host plant species, selection on fungi might be for higher rates of root colonization per unit root length, which could account for the opposite trends we observed of plant and fungal performance with latitude. Similarly, chemical diversity and neutral genetic diversity within and among pine populations and species, which has been found to be substantial (Moran et al., 1988; Burdon, 1992; Hong et al., 1993; Aitken & Libby, 1994; Wu et al., 1998, 1999), may not be the result of coevolution with ectomycorrhizal fungi; however, if such diversity influences the ability of ectomycorrhizal fungus genotypes to colonize and interact with pine host plants, it could select for a diversity of fungal genotypes within and between populations with respect to compatibility with chemically different hosts.

A significant asymmetry in the natural history of this interaction is that ectomycorrhizal fungi are obligate symbionts of plants, and in coastal pine forests, R. occidentalis typically has few other potential host species besides the locally dominant pine. The pine host, in contrast, typically associates with a diversity of ectomycorrhizal symbionts. As a result, selection on Pinus host plants may be for a core set of traits related to its interaction with the entire guild of ectomycorrhizal fungi (Thompson, 2005), whereas selection on R. occidentalis may be for more specific traits related to particular sets of plant genotypes present in local populations. This scenario leads to the prediction that local adaptation should be a more prevalent pattern in the ectomycorrhizal fungi than in the plant hosts, as we found in our study.

Asymmetries in gene flow rates and/or generation times may also contribute to asymmetric patterns of evolved geographic structure in species interactions. For example, higher rates of among-population gene flow in parasites compared with hosts may facilitate local adaptation of parasites to host populations (Gandon & Michalakis, 2002). In the interaction between coastal pines and R. occidentalis, however, if an asymmetry in rates of gene flow exists it is probably in the opposite direction. Rhizopogon species produce belowground sporocarps that are dispersed locally by small mammals or decompose in place (Molina et al., 1999), with dispersal distances on the order of hundreds of meters (Ashkannejhad & Horton, 2006). As a consequence, their scale of dispersal and rates of between-population gene flow may be much lower than those of their pine hosts, whose airborne pollen can probably routinely travel distances on the order of kilometres to tens of kilometres (e.g. Schuster & Mitton, 2000). Recent studies of genetic structure in R. occidentalis found strong differentiation between nearby populations (L. Grubisha, personal communication), and studies of other mammal-dispersed ectomycorrhizal fungi have found significant genetic differentiation between populations at relatively small geographic scales (Murat et al., 2004; Kretzer et al., 2005). Genetic differentiation among populations of pines on the west coast of North America is significant (Aitken & Libby, 1994; Wu et al., 1998, 1999), but almost certainly does not exceed that found among populations of R. occidentalis. Recent models of the evolution of mutualistic interactions suggest that the more slowly evolving partner in an interaction may have an evolutionary advantage (Doebeli & Knowlton, 1998; Bergstrom & Lachmann, 2003), which could lead to an asymmetric pattern of local adaptation. Minimum generation times of the three Pinus species range from 4 to 5 years (Grotkopp et al., 2002), whereas that of R. occidentalis is probably 1 year or less (J.D. Hoeksema, personal observation), so once again, the asymmetry in local adaptation we observed is not in the direction predicted.

Overall, these results indicate that substantial evolved geographic structure exists in interactions between pines and false truffles, especially clinal geographic patterns that differ between the host and the symbiont. These results represent a step forward in our understanding of the spatial and taxonomic scales of evolved geographic structure in diverse, widespread interactions that may fluctuate along the mutualism–parasitism continuum. In such interactions, asymmetries between the two species may drive divergent evolutionary responses to geographic variation in biotic and abiotic variables. As we continue to expand our understanding of how evolved geographic structure in these interactions is influenced by environmental variation, we will develop a more comprehensive understanding of the coevolutionary dynamics of diverse species interactions and will thus be better able to manage and conserve populations involved in these interactions.


The authors are grateful to T. Bensen, M. Cuautle, S. Dwiggins, C. Fernandez, S. Forde, C. Hays, J. Vargas Herandez, K. Horjus, A.-L. Laine, B. Piculell, J. Springer, J. Trappe, M. Whitlock and two anonymous reviewers for comments on earlier drafts of the manuscript, and to T. Bruns, L. Grubisha, R. Molina and J. Trappe for discussion and expertise throughout the course of the project. J.D. Hoeksema was supported by a National Science Foundation Postdoctoral Fellowship in Microbial Biology (DBI-0200129) and an award from the Monterey Pine Forest Ecology Research Cooperative. J.N. Thompson was supported by grants from the National Science Foundation (DEB-344147 and DEB-0073911).