SEARCH

SEARCH BY CITATION

Keywords:

  • coevolution;
  • community;
  • ecology;
  • evolution;
  • host;
  • local adaptation;
  • mutualism;
  • symbiont

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

1. Variation in the effectiveness of mutualistic associations is well characterized in plant–soil symbiont interactions, yet there are little empirical data providing insight into how such variation evolves and persists in natural settings.

2. Heterogeneity in the strength and direction of co-evolutionary selection among spatially discrete demes is predicted to be important for the maintenance of genetic variation in species interactions. Here, we experimentally test the potential for local adaptation to generate phenotypic divergence among wild host–symbiont populations using two leguminous host species that differ in their specificity for rhizobial partners.

3. Molecular characterization of host populations and associated rhizobial communities revealed significant among-population genetic differentiation. Reciprocal cross-inoculation experiments testing for variation in the fitness of nine populations of Acacia salicina and A. stenophylla in response to inoculation with rhizobia revealed variation in host response to the mutualism (both host species), and the benefit conferred by different rhizobial populations (A. salicina only). However, there was no indication that host population-by-rhizobial population interactions influence the outcomes of mutualism for the host.

4. We further examined potential correlations between (i) plant response to inoculation and (ii) rhizobial effectiveness, with variation in soil fertility at the sites from which plant and rhizobial samples originated. Data from the cross-inoculation experiments revealed no correlations between soil chemistry, water availability and either host or rhizobial performance. However, analysis of results from an extensive whole-soil inoculation trial including nearly 60 A. salicina and A. stenophylla sites showed a significant negative correlation between levels of soil nitrogen and plant response to inoculation.

5. Overall, these findings suggest that selection for local adaptation may play little role in maintaining phenotypic variation in these interactions. We hypothesize that mutualistic interactions occurring among communities of hosts and symbionts do not favour co-evolutionary divergence among populations.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Mutualisms between plants and soil symbionts (e.g. rhizobial bacteria, mycorrhizal fungi) play key roles in terrestrial ecosystems. In addition to regulating plant productivity via nutrient acquisition and cycling (Bever, Westover & Antonovics 1997; Sprent 2007; van der Heijden, Bardgett & van Straalen 2008), dynamic feedbacks between plants and microbial symbionts play important roles in many ecological and evolutionary processes relevant to the establishment and maintenance of diversity in plant and soil communities (Bever, Westover & Antonovics 1997; Reynolds et al. 2003; Wardle 2006). Importantly, the ecological outcomes of these associations can be highly variable depending on the genetic and environmental context within which the interaction occurs (Heath & Tiffin 2007; Thrall et al. 2007a; Hoeksema et al. 2010). In natural systems, symbiotic partners vary widely in genetic identity (Lafay & Burdon 1998; Helgason et al. 2002), partner specificity (Thrall, Burdon & Woods 2000; Heath 2010; Thrall et al. 2011) and the fitness benefits they provide to partners (Parker 1995; Burdon et al. 1999; Vogelsang, Reynolds & Bever 2006; Heath 2010). Precise outcomes can be further influenced by both the physical (Thrall et al. 2009; Johnson et al. 2010) and biotic (Larimer, Bever & Clay 2010) environments within which such interactions occur. However, despite the importance of plant–soil microbe mutualisms to ecosystem function, the ecological and evolutionary processes that contribute to the maintenance of diversity in symbiotic traits within such interactions are not well understood.

In natural systems, organisms typically interact within local patches connected by dispersal. In such spatially heterogeneous landscapes, organisms can evolve to become adapted to their local environmental conditions (Williams 1966). In mutualisms, local adaptation occurs when the mean fitness of one (or both) partners is higher when interacting with a partner from its own habitat (e.g. Hoeksema & Thompson 2007; Johnson et al. 2010). Much recent work in ecology and evolution has sought to understand how spatial heterogeneity can determine patterns of geographic adaptation in species interactions. This body of work clearly demonstrates that local selection can be a strong force maintaining genetic structure in species interactions, but that local adaptation should not always be the expected outcome within spatially structured environments (for review, see Kawecki & Ebert 2004; Thompson 2005; Greischar & Koskella 2007; Hoeksema & Forde 2008; Laine 2009). Rather, complex patterns of adaptation and maladaptation may be the rule [as envisaged by the Geographic Mosaic Theory of Coevolution (Thompson 2005)], with precise outcomes dependent on variation in the genetic basis of the interaction (Wade 2007), the environmental context within which the interaction takes place (Thrall et al. 2007a), and on life-history characteristics of the interacting partners, including dispersal ability (Garant, Forde & Hendry 2007) and partner specificity (Barrett et al. 2009).

Patterns of adaptive divergence in species interactions partly reflect the tension between the diversifying effects of local selection and the homogenizing effects of gene flow (Garant, Forde & Hendry 2007; Greischar & Koskella 2007; Hoeksema & Forde 2008). In particular, there exists a general expectation that the capacity for natural selection to structure populations will be constrained when rates of gene flow are high, because of the ‘swamping’ of local populations with alleles common throughout the metapopulation (Gandon et al. 1996; Gomulkiewicz et al. 2000; Nuismer, Thompson & Gomulkiewicz 2000). Furthermore, asymmetrical patterns of dispersal and gene flow among interacting species are predicted to have important consequences for co-evolutionary dynamics (Gandon & Michalakis 2002). For example, as long as rates of gene flow are not too high, parasites dispersing at greater rates than their hosts are expected to have greater potential to locally adapt to their hosts (and vice-versa when hosts disperse more than their antagonists), because new alleles are introduced into populations upon which selection can act (Garant, Forde & Hendry 2007).

In natural plant–soil microbe mutualisms, individual hosts and symbionts are generally embedded within communities of potentially interacting partners (Hoeksema 2010; Larimer, Bever & Clay 2010). The degree of specialization for any given partner in an interaction is therefore an important life-history trait governing the breadth and complexity of interactions (i.e. network structure) that may develop locally (Thompson 2005; Barrett et al. 2008; Guimarães , Jordano & Thompson 2011). Variation in partner fidelity and network structure can have dynamic impacts at the ecological level, potentially influencing encounter rates (Power & Mitchell 2004), community stability (Okuyama & Holland 2008) and fitness outcomes of species interactions (Bever, Westover & Antonovics 1997). Consequently, such heterogeneity can also strongly influence patterns of adaptation among populations. For example, investigation of the role of host fidelity in local adaptation suggests that compared to generalists, symbionts with narrow host ranges are more likely to be locally adapted to their hosts (Lajeunesse & Forbes 2002). Such differences are likely due in part to relatively strong reciprocal selection pressures imposed on partners in tightly coupled, specialized interactions (Strauss & Irwin 2004; Barrett et al. 2009). Patterns of local adaptation in multispecies interactions will also depend to some extent on the community context within which the interaction occurs. For example, potential for local adaptation in generalists will likely differ as available partners and the frequency with which they are encountered vary among localities (Lankau & Strauss 2008; Kniskern, Barrett & Bergelson 2011). Furthermore, disruptive selection via processes like negative feedback (Bever, Westover & Antonovics 1997) may limit potential for local adaptation in generalist species (Kawecki 1998). Yet to date, little attention has been paid to how variation in partner fidelity may mediate local adaptation in mutualisms.

In any species association, genetic variation in interacting partner traits is required for co-evolutionary change. In legume–rhizobial interactions, it is well established that symbiotic outcomes can depend on host genotype by rhizobial genotype (GH × GR) interactions. In particular, two features, nodule formation (infectivity) and mutualistic benefit (i.e. nitrogen fixation by rhizobia; resources supplied by host to rhizobia), have been shown to vary depending on both host and rhizobial genotypes. First, establishment of a symbiosis (i.e. infection) requires successful exchange of specific molecular signals between partners (Masson-Boivin et al. 2009). While levels of partner specificity during the colonization phase may vary greatly, it appears generally true that symbioses (whether effective or not) can only be established between a more or less limited set of plant and rhizobial genotypes (Perret, Staehelin & Broughton 2000). Allelic variation in genes determining host and rhizobial specificity has been identified in several interactions (for review, see Devine & Kuykendall 1996; Perret, Staehelin & Broughton 2000; Masson-Boivin et al. 2009; Yang et al. 2010). In addition, variability in mutualistic benefits for both partners post-infection has been shown to be controlled by GH × GR interactions (Heath 2010). In particular, it is well established that fitness benefits afforded to hosts by rhizobia (i.e. nitrogen fixation) can vary depending on the genotypes of both the rhizobia and the host (e.g. Gibson 1964; Gibson & Brockwell 1968; Parker 1995; Mårtensson & Rydberg 1996; Burdon et al. 1999; Heath, Stock & Stinchcombe 2010).

Partners in mutualisms may be expected to adapt not only to each other, but also to their physical environment (Thompson 2005). Because nutrients are currency in plant–soil microbe mutualisms (Vogelsang, Reynolds & Bever 2006; Heath & Tiffin 2007), soil fertility is predicted to be a key environmental variable influencing co-evolutionary trajectories (Thrall et al. 2007a). Specifically, it has been suggested that beneficial mutualisms may be more likely to evolve and be maintained in nutrient-deficient soils (Thrall et al. 2007a; Johnson et al. 2010). Consistent with such a scenario, there is increasing evidence that under conditions of high nutrient availability, symbionts may become essentially parasitic (Johnson, Graham & Smith 1997; Hoeksema & Schwartz 2003; Verbruggen & Kiers 2010), or conversely, hosts may evolve reduced dependence on the mutualism (Johnson, Graham & Smith 1997; Ryan et al. 2005). It is also becoming clear that the evolution of plant–soil symbionts can be influenced by heterogeneity in other environmental factors such as soil pH (Richardson & Simpson 1989; Garau et al. 2005), soil salinity (Thrall, Bever & Slattery 2008) and water availability (Stahl & Smith 1984). Together, these studies suggest that genotype × environment (G × E) and genotype × genotype × environment (G × G × E) interactions have strong potential to generate and maintain local patterns of adaptation in mutualistic interactions (Thompson 2005). However, too few studies exist to make broad generalizations regarding the potential for environmental heterogeneity to mediate co-evolutionary outcomes in mutualisms.

Here, we use reciprocal-inoculation experiments with natural populations of Australian Acacia (A. salicina and A. stenophylla) and associated communities of nitrogen-fixing bacteria (rhizobia) to evaluate the potential for host adaptation to local communities of rhizobial mutualists. Local adaptation of hosts to rhizobia should be detectable as a population-level GH × GR interaction, with rhizobial inoculation benefiting sympatric hosts more than allopatric hosts. To place our experimental results within a broader evolutionary context, we examine the potential for gene flow to mediate or reinforce the effects of local selection by characterizing patterns of population genetic structure in hosts and rhizobia using molecular markers.

We further assess our results in the light of variation in (i) partner fidelity of host species and (ii) soil chemistry. With regard to partner fidelity, experimental and descriptive surveys of natural populations indicate distinct, qualitative differences between host species with regard to the range of bacteria with which they effectively associate (Thrall et al. 2007b; Hoque, Broadhurst & Thrall 2011). In particular, A. stenophylla forms effective symbioses with a smaller range of isolates compared with A. salicina, so that, at least in a comparative sense, A. stenophylla may be considered a specialist, and A. salicina a generalist. We hypothesized that potential for population-level variability in rhizobial effectiveness, and hence local adaptation, would be stronger for the specialist A. stenophylla. Furthermore, rhizobial abundance in soils where A. salicina and A. stenophylla occur is negatively related to levels of soil nitrogen (Thrall et al. 2007b), leading us to hypothesize that rhizobial effectiveness may also be negatively correlated with levels of soil nitrogen in their source localities.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Study Sites, Materials and Experimental Design

We investigated the potential for local adaptation in interactions between root-nodule-forming bacteria and two leguminous host species; Acacia salicina and A. stenophylla. For each species, plant seeds and rhizosphere soils were collected at nine spatially discrete localities distributed within the Murray-Darling Basin in south-eastern Australia (Table 1). For A. salicina, pairwise geographic distances between allopatric populations ranged from c. 38 to 693 km, while for A. stenophylla, these distances ranged from c. 66 to 876 km. These sites represent a subset of those described in Thrall et al. (2007b). For both A. salicina and A. stenophylla, we tested performance of host plants with all combinations of sympatric and allopatric rhizobial communities. In total, 162 different plant-population × bacterial-population combinations were assessed. For uninoculated controls, we included nitrogen plus (N+) and nitrogen minus (N) treatments for each of the 18 host seed sources.

Table 1.   Location, climate and soil characteristics of Acacia salicina and Acacia stenophylla populations used in cross-inoculation studies
SpeciesSite (abbreviation)Latitude and longitudeMean annual precipitation (mm)NO3-N (mg kg−1)Soil Organic Carbon (%)pH
Acacia salicinaBringagee (Br)34°16′ S 145°45′ E3827·31·65·7
Conargo (Con)34°17′ S 145°10′ E374322·15·3
Lake Urana (LU)35°16′ S 146°12′ E4237·32·87·5
Lila Park (LP)33°12′ S 147°19′ E442224·46·3
Combadello (CPG)29°33′ S 149°37′ E5477·736·9
Pretty Pine (PP)35°24′ S 144°50′ E3717·327
Lynwood (Lyn)29°44′ S 150°32′ E674123·36·9
Warren Weir (WW)31°41′ S 147°50′ E465362·77·1
Yang Yang (YY)34°28′ S 144°18′ E3123·43·67·6
Acacia stenophyllaBangate Bridge Weir (BBW)29°41′ S 147°48′ E429171·26·9
Balranald (Bal)34°42′ S 143°35′ E31351·76·8
Lake Cargelligo (LC)33°12′ S 146°27′ E4027·62·86·6
Wallamundry Creek (WC)33°10′ S 147°09′ E426192·96·4
Hay Weir (HW)34°31′ S 144°42′ E3780·51·47·1
Mildura (Mild)34°10′ S 142°10′ E2812·228·3
Muggabah Creek (MC)33°46′ S 144°55′ E322111·17·1
Lake Victoria (LV)34°03′ S 141°17′ E258180·856·9
Willandra National Park (WNP)33°11′ S 145°07′ E333101·96·0

At each site, soil was collected from around the base of 15–20 plants using a 50-mm soil auger. Sampled plants were a minimum of 10 m apart. Soil samples were then bulked and left to air dry. Dried soils were thoroughly mixed, passed through a 2-mm sieve and stored at room temperature for up to 3 months prior to characterization of soil chemistry and rhizobial trapping (Thrall et al. 2007b).

Bacterial strains used in this study represent a subset of those reported in Hoque, Broadhurst & Thrall (2011), and detailed bacterial isolation procedures can be found in that paper. Briefly, bacteria were isolated from soils using resident host species as trap plants. Two-week-old seedlings from each species were inoculated with soil suspensions from each site. After 4–6 weeks growth under laboratory conditions, individual root nodules were collected, surface sterilized and ground to a fine paste. An aliquot of this paste was then streaked onto yeast-mannitol agar plates. After 3–12 days, well-separated colonies were transferred to fresh plates and subcultured until a pure culture was obtained. Using this process, we obtained between 16 and 26 isolated strains per population. For inoculation, individual bacterial strains were grown in 20 ml of yeast-mannitol broth and incubated with shaking for 5–7 days. Following standardization of optical densities, rhizobial communities were reconstituted by mixing isolates from each population in equal proportions. For each treatment in the inoculation study, 5 ml of the resultant suspension was added directly to the base of each 2-week-old seedling.

Seeds were collected in equal amounts from 10 to 20 plants in each locality and then bulked. Prior to germination, seeds were surface sterilized in 100% ethanol for 30s, transferred to 4% (w/v) sodium hypochlorite for 3 min, rinsed twice with sterile distilled water and scarified in concentrated sulphuric acid for 45 min. The seeds were rinsed with sterile water 10 times and left at room temperature in the final wash overnight. The seeds were then spread over sand/vermiculite flats, watered daily and left to germinate. Emerging seedlings were planted into pots 7–10 days after germination, and the soil surface was covered with a layer of polyurethane beads to limit splashing among pots and cross-contamination. Plants were grown under standard glasshouse conditions, watered with N-free 1:50 diluted McKnight’s solution (1949) three-four times weekly. Plants were harvested 13 weeks after inoculation, and above-ground parts were oven dried and weighed. While the focus of these experiments was plant growth, at the time of harvest plant roots were separated from the soil and a range of nodulation characteristics also recorded, including: (i) presence/absence of nodules; (ii) nodule number (<10, 10–50, >50) and (iii) nodule functionality based on colour and size (ranging from 1 to 5: 1 = small-non-N2 fixing with white centres; 5 = large N2 fixing nodules with pink/red centres).

Statistical Analyses of Plant Growth

Plant dry weights were square root-transformed to fit assumptions of normality. We used two-way anovas to test whether host growth varied in response to host or rhizobial origin and whether performance varied with particular geographic combinations of host and rhizobia. With regard to evidence for local adaptation, a significant interaction term would indicate that some combinations of host and symbiont populations resulted in better (or worse) plant growth. Subsequent decomposition of the resultant variance matrix into sympatric vs. allopatric components allows for specific testing of local patterns of adaptation (Kaltz et al. 1999).

For both A. salicina and A. stenophylla, we examined correlations between rhizobial effectiveness, host response to inoculation and soil chemistry across each of the nine populations (Table 1) using multiple linear regression. Specifically, stepwise multiple regressions were carried out using individual soil chemistry variables [inorganic nitrogen (NO3: mg kg−1), soil pHW, phosphorus (mg kg−1) and organic carbon (%) and annual rainfall (mm)] (Table 1) as predictors of plant dry weights (Fig. 1). See Thrall et al. (2007b) and references therein for details regarding soil chemistry methods. In addition, because nine data points confer limited power to reliably detect linear relationships, we extended our analysis to include data describing plant responses to whole-soil inocula collected from 28 A. salicina sites and 30 A. stenophylla sites. These data were generated as part of a broader geographic survey of symbiont populations (Thrall et al. 2007b), but had not been analysed with respect to potential effects of soil chemistry on rhizobial effectiveness. For that study, plants were sourced from a common commercial stock; thus, issues to do with local adaptation could not be addressed. Models were fitted using R, version 2.8.1, which is available as a free download at http://www.r-project.org.

image

Figure 1.  Variation in mean host growth responses [dry weight (g)] for each host-population rhizobial-community combination relative to N controls (i.e. a value of 0 along the Y-axis means no difference from uninoculated controls). Error bars show the standard error of the mean.

Download figure to PowerPoint

Genetic Characterization and Analysis of Plant and Rhizobial Populations

Genomic DNA was extracted from c. 10 mg of freeze-dried phyllode tissue as described by Broadhurst et al. (2006). AFLP amplification largely followed that of Vos et al. (1995) with the following modifications: c. 400 ng of genomic DNA template was digested for each sample using EcoRI/MseI, the EcoI-A-MseI-C pre-amplification was diluted 1:30 prior to selective amplification, and selective amplification primers were fluorescently labelled. Initial screening of 12 primer combinations identified six with polymorphic and repeatable banding patterns: EcoI-AGC/MseI-CTC (Fam), EcoI-AGT/MseI-CTC (Pet), EcoI-ACC/MseI-CTC (Vic), EcoI-AGC/MseI-CTG (Fam), EcoI-ACC/MseI-CTG (Vic) and EcoI-AGA/MseI-CTG (Ned). Amplicons were visualized on an ABI 3130XL sequencer using a LIZ 500 bp internal standard (Applied Biosystems, Foster City, CA, USA), scored for the presence/absence of bands using GeneMapper version 4.0 software (Applied Biosystems) and a binary matrix generated for each host.

For Acacia, we used an analysis of molecular variance (amova) to apportion total genetic variation within and among geographic localities of each host. To summarize genetic relationships among individual plants, ordination of pairwise genetic distances within each host species was conducted using principal co-ordinate analysis (PCO). Genetic distance calculations and amova (based on 999 permutations) were performed using the software GenAlEx. The first three dimensions of the PCO were plotted using Sigmaplot. To more precisely identify patterns of genetic similarity and gene flow among geographical localities, STRUCTURE version 2.3.2 (Pritchard, Stephens & Donnelly 2000), which is an individual-based Bayesian clustering method, was used to estimate the distribution of individuals among K genotypic clusters without a priori population assignment. The admixture model was assumed because seed dispersal along waterways was considered highly likely with frequencies correlated among populations and alpha inferred from the data set. Five runs for each = 2–7 with a 50 000 burn-in followed by 500 000 MCMC replications were run with the optimal K value for each species estimated by Structure Harvester v0.6.1 (Earl 2011) using the posterior likelihood and variance against K via the Evanno, Regnaut & Goudet (2005)ΔK method.

Rhizobial strains were genotyped using PCR–RFLP of the 16s ribosomal RNA gene, following procedures described in detail in Hoque, Broadhurst & Thrall (2011). Restriction profiles, based on the banding patterns of six restriction enzymes, were determined for each isolate with each unique profile being considered as representing a distinct operational taxonomic unit (OTU). Generic affiliations of isolates were inferred through comparison of OTU banding profiles to 16s reference sequences, as described in Hoque, Broadhurst & Thrall (2011). We used analysis of molecular variance (amova) (Excoffier, Smouse & Quattro 1992) to analyse PCR–RFLP data with respect to the geographic partitioning of genetic variation within and among populations. Genetic relatedness among all pairs of rhizobial populations was evaluated using PCR–RFLP data and unweighted pair-group mean analysis (UPGMA) based on estimates of Nei’s genetic distance (1972). The UPGMA tree was constructed using MEGA version 5. amova and Nei’s distance estimates were calculated using GENALEX version 6.0 (Peakall & Smouse 2006). To test for correlations between host and rhizobial genetic structure, symbiotic effectiveness and host response to inoculation, we constructed matrices of (i) host and rhizobial genetic dissimilarities and (ii) matrices of the difference in growth response of Acacia populations. We tested for correlation between matrices using Mantel tests (implemented in R).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Geographic Structure in Plant Fitness and Rhizobial Benefit

For both A. salicina and A. stenophylla, the outcomes of rhizobial inoculation were generally positive, although the degree of benefit varied widely for different host–rhizobia combinations (Fig. 1). Overall, A. salicina plants inoculated with rhizobia increased in dry weight (relative to the uninoculated N control) by a factor of 1·9, while inoculation resulted in 2·6-fold increase for A. stenophylla. For A. salicina, variation in plant growth was attributable to differences in the geographic origin of rhizobial strains (Fig. 2; F8,721 = 16·6, P < 0·0001), and to a lesser extent, the geographic origin of plant seeds (F8,721 = 4·81, P < 0·0001). In particular, we observed a general consistency in the relative performance of rhizobial communities that was largely independent of host geographic origin (Fig. 2). For example, rhizobia trapped from Bringagee soils were among the least effective, regardless of the source of A. salicina seeds, while rhizobia from Lila Park soils were highly effective for all host populations (Fig. 2). Specific combinations of host and rhizobia had no effect on host growth, ruling out any possibility of local adaptation (Table 2; Fig. 2). In contrast, for A. stenophylla, rhizobial origin did not significantly impact on host performance (Table 2). Rather, the main source of factorial variation in plant growth was attributable to differences in host geographic origin (Fig. 3; F8,726 = 18·83, P < 0·0001). For example, plants from Wallamundry Creek generally grew relatively poorly, while plants from Balranald grew relatively well, regardless of the rhizobia with which plants from a given population were inoculated (Fig. 3). Like A. salicina, for A. stenophylla, the interaction between host and rhizobial origin had no significant effect on host growth (Table 2; Fig. 3).

image

Figure 2.  Mean weights of Acacia salicina plants from nine populations (Table 1) when inoculated with all combinations of rhizobia collected from the same localities. Dark bars represent sympatric host–rhizobial combinations. Lines within each plot represent the mean dry weight of uninoculated control plants grown in the absence of added nitrogen. Error bars show the standard error of the mean.

Download figure to PowerPoint

Table 2.   Analysis of variance testing the effects of host origin, rhizobial origin and the host–rhizobial origin interaction on plant growth (relative to uninoculated nitrogen-free controls) for Acacia salicina and A. stenophylla
SourceA. salicinaA. stenophylla
d.f.SSFPd.f.SSFP
Host origin812·294·81<0·0001834·479·67<0·0001
Rhizobial origin842·4716·60<0·000183·510·980·45
Host × Rhizobia6410·210·500·9966420·060·700·96
Residual721230·542  726323·6  
image

Figure 3.  Mean weights of Acacia stenophylla plants from nine populations (Table 1) when inoculated with all combinations of rhizobia collected from the same localities. Dark bars represent sympatric host–rhizobial combinations. Lines within each plot represent the mean dry weight of uninoculated control plants grown in the absence of added nitrogen. Error bars show the standard error of the mean.

Download figure to PowerPoint

Environmental Correlates

To investigate the role of environmental heterogeneity in driving variation in host response and rhizobial performance, we examined how variation in the response of plant populations to inoculation (mean plant dry weights for each host source averaged across all inoculum sources), and mean effectiveness of rhizobial communities (mean dry weights for each rhizobial community averaged across seed sources), may depend on a range of environmental variables. Specifically, stepwise multiple regressions were carried out using individual soil chemistry variables (nitrogen, pH, phosphorus, organic carbon) and annual rainfall (Table 1) as predictors of plant dry weights (Fig. 1). For the nine A. salicina populations used in pairwise cross-inoculation trials, we found no significant environmental predictors of either mean rhizobial performance (final model: inline image = 0·375, F2,6 = 4·2, P = 0·08), or mean plant response to inoculation (final model: inline image = 0·06, F3,5 = 1·20, P = 0·40). Likewise, for A. stenophylla, we found no significant predictors of mean plant response to inoculation, and no variables were retained in the final model.

We extended our investigation of environmental influences on the evolution of plant–soil mutualisms to include results from a previous glasshouse inoculation study that used whole-soil inocula collected from 28 A. salicina and 30 A. stenophylla populations (Thrall et al. 2007b), including the sites used in the current local adaptation study. These data describe the response of plants from a common seed source to inoculation with whole soils and thus are interpreted here as the net effectiveness of the microbial community in promoting plant growth. In this extended analysis, for A. salicina, two significant predictor variables were maintained in the final model (overall model: inline image = 0·356, F2,25 = 8·464, P < 0·01), organic carbon (b = 0·80, P < 0·01) and nitrogen (0·05, P = 0·013). For A. stenophylla soils, we found no significant environmental predictors of variation in the plant growth promoting abilities of microbial communities (final model inline image = 0·145, F2,27 = 2·291, P = 0·12).

Genetic and Community Structure

Genetic diversity was higher at the within-population than among-population level for both host species, as indicated by amova (A. salicinaΦ 0·195, < 0·001; A. stenophyllaΦ 0·286, < 0·001). For A. salicina, structure analyses indicated that = 5 was the optimal number of genetic groups (Fig. S1, Supporting information). While genetic differentiation was evident among populations, this was not associated with any clear geographic structure or isolation-by-distance (IBD) effects (Mantel test, P > 0·05). For ordination analyses, the first three principal components for A. salicina accounted for c. 67% of the total variation, with most individuals grouped within their respective population cluster and with some separation among populations across the three axes (Fig. 4). Separation among populations was much stronger in A. stenophylla with Mildura being removed from all other populations (Fig. 4). For ordination analyses, almost 73% of the PCO variation in A. stenophylla was accounted for by the first three axes (Fig. 4) with individuals also mostly grouping with plants from their respective population. While individuals in most populations showed very little evidence of admixture, some Warren Weir samples appeared to be admixed with Lynwood, while several Lake Urana samples had affinities with Bringagee. For A. stenophylla, the optimal number of genetic groups revealed by structure analyses was = 6 (Fig. S2, Supporting information). Mildura and Bangate Bridge formed their own discrete genetic clusters while Lake Cargelligo and Muggabah Creek were also a discrete group. Admixture was more evident in this host. For example, Balranald had individuals with a Lake Cargelligo-Muggabah background, which may be consistent with hydrological flows to the southwest. Lake Victoria was also highly admixed with samples having their own unique genetic profiles as well as backgrounds indicative of Lake Cargelligo-Muggabah and to a lesser extent Wallamundry Creek-Willandra NP. Hay Weir also included several samples with a Wallamundry Creek-Willandra NP background. Geographical structuring of genetic groups was again not strong, and we found no evidence for IBD effects (Mantel test, P > 0·05).

image

Figure 4.  Principal coordinates plots based on genetic distances among Acacia salicina and Acacia stenophylla AFLP genotypes. Each genotype is coloured according to host population of origin.

Download figure to PowerPoint

With respect to rhizobial communities, 186 and 193 root-nodule bacteria were isolated from the nine A. salicina and A. stenophylla soils, respectively. Rhizobial inocula representing reconstituted symbiont communities for each population were made up of equal mixtures ranging from 16 to 24 isolates (average: 21·6 isolates per mixture). For each species, rhizobial communities were only moderately differentiated with regard to restriction fragment length variation among localities, as indicated by UPGMA clustering (Fig. 5) and amovaPT = 0·147, P < 0·01 for A. salicina; ΦPT = 0·126, P < 0·01) for A. stenophylla. PCR–RFLP data further indicated that rhizobial communities were not strongly differentiated by host species, with only 5% of variation explained by differences among host of origin, as compared to 13% among populations and 82% among individuals within populations (amova: ΦRT = 0·046, ΦPR = 0·137, P < 0·01).

image

Figure 5.  Unweighted pair-group mean analysis (UPGMA) phylogram based on pairwise estimates of Nei’s genetic distance among for rhizobial communities. Individual fingerprints for each rhizobial isolate were obtained by restriction digestion of the 16s ribosomal RNA gene.

Download figure to PowerPoint

With regard to comparative population genetic structures of Acacia and rhizobia, we found no significant relationships between pairwise genetic distance among rhizobial populations and corresponding pairwise genetic distance among host populations (Mantel tests; A. salicina, r = −0·02, P = 0·52; A. stenophylla, r = −0·13, P = 0·71). Likewise, we were unable to establish correlative relationships between similarities in rhizobial community structure, rhizobial effectiveness and soil chemistry (Mantel tests; P > 0·1 for all comparisons), nor any significant correlation between host genetic similarity and host response to inoculation (P > 0·1 for both comparisons).

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

We predicted that plant populations would exhibit local adaptation in response to inoculation with sympatric and allopatric rhizobia, detectable as a population-level GH × GR interaction, with rhizobial inoculation benefiting sympatric hosts more than allopatric hosts. In contrast to these expectations, our data indicate that population-level variation in plant fitness is not a function of variation in the effectiveness of different combinations of host and rhizobia. Instead, for A. salicina, we found that host populations vary in response to rhizobial inoculation, and that rhizobial populations also vary significantly in effectiveness. For A. stenophylla, we found that host populations varied in response to rhizobial inoculation, but that rhizobial populations did not vary significantly in performance. In other words, while we found evidence for strong variation in host response to inoculation, and, for A. salicina, in rhizobial effectiveness among localities, we found little potential for among-population genetic interactions to influence outcomes of plant–rhizobial interactions. These results are interesting given that geographically structured variation in traits important to the outcome of species interactions has been shown to be a pervasive feature of many natural host–symbiont associations (Thompson 2005).

The seeming absence of population-level variation in the infectivity and effectiveness of different combinations of host and rhizobia may reflect the influence of various factors. Here, we discuss two possible general explanations for our results – first, that geographically variable GH × GR interactions exist, but we failed to detect them; and secondly, that strong population-level GH × GR effects may be uncommon in plant–rhizobial symbioses.

A general caveat of ecological experiments performed under controlled conditions relates to the extent to which results can be used to draw inferences regarding outcomes expected under natural conditions. In our experiments, we focused on plant growth during seedling and juvenile stages, not on capturing estimates of total lifetime fitness. It is well established that enhanced seedling and juvenile vigour are critical for successful establishment in semi-arid and highly seasonal environments (Thrall et al. 2005; Leck, Parker & Simpson 2008 and references therein). However, while there is no evidence to suggest that rhizobial specificity and effectiveness might vary with host ontogeny, as Palmer et al. (2011) elegantly demonstrate for ant-acacia interactions, a full assessment of the outcomes of species associations involving long-lived perennial plants may require measuring outcomes of species interactions at multiple points in time.

In addition, while we explicitly set out to simulate aspects of the community context within which individual plants and soil symbionts naturally interact, the densities and relative frequencies of rhizobial genotypes in our inocula are only approximations of community structure under natural conditions. In particular, the relative frequencies, densities and distribution of different rhizobial genotypes in the soil are likely to vary throughout the soil matrix, making the precise reconstruction of rhizobial communities extremely challenging. The consequences of such variation for plant growth are largely unexplored, and it is possible that our results would differ given more precise reconstruction of soil microbial communities.

Finally, while our experimental approach (i.e. utilization of bulked seed and multi-isolate rhizobial inocula for different populations) facilitated examination of patterns of adaptation across multiple populations and species, and arguably more closely mirrors natural conditions (where plants are exposed to communities of potentially interacting rhizobial partners), it limits power to draw conclusions regarding pairwise interactions that might underlie patterns in our data. In particular, variation in specificity and effectiveness is widely documented between individual legume and rhizobial genotypes (e.g. Gibson 1964; Gibson & Brockwell 1968). Such polymorphism has also been demonstrated within natural populations (e.g. Burdon et al. 1999; Heath 2010). It is likely the approach used here masked the presence of some GH × GR effects between individual genotype pairs within populations. Therefore, without specifically examining interactions at this scale, we cannot conclude that GH × GR interactions are lacking within populations, nor that identical (co) evolutionary trajectories define all populations – only that genetic interactions are seemingly absent at the among-population scale.

Given these caveats, we argue that the most likely explanation for the lack of significant between-population interactions is that host populations do not vary in response to inoculation with different rhizobial populations. We further consider it unlikely that the absence of GH × GR interactions reflects genetic swamping via extensive gene flow. While evidence for some admixture was evident for both host species, the molecular and phenotypic data indicate clear and significant genetic delineations among most geographic localities, suggesting that gene flow is unlikely to be so high as to completely homogenize populations with respect to potential for variation in GH × GR responses. Similarly, although our data are based on variation in a single gene and hence provide less precision with which to characterize differences among locations in the genetic structure of rhizobial communities, it is clear from this, and other studies (e.g. Bissett et al. 2010), that rhizobial community composition also differs significantly among localities.

Our results are largely consistent with findings from previous studies reporting on geographically structured variation within legume–rhizobial interactions. For example, Burdon et al. (1999), using half-sib host families and individual rhizobial isolates, report a general lack of host-population × rhizobial-origin effects in interactions between locally collected rhizobia and three Acacia spp., although significant variation was detected among some isolates and half-sib families. Heath (2010) also found no evidence for population-level GH × GR effects in interactions between Medicago trunculata and Sinorhizobium meliloti, although significant GH × GR effects were detected within populations. Similarly, Wilkinson, Spoerke & Parker (1996) find no evidence for population-level GH × GR interactions structuring variation within three lineages of Amphicarpaea bracteata and their locally associated rhizobia. A lack of geographically structured GH × GR variation in interactions between A. salicina, A. stenophylla and their rhizobia may therefore represent a common characteristic of plant–rhizobial mutualisms.

The considerable empirical support for the general role of G × G and G × G × E interactions in structuring genetic variation across landscapes in host–symbiont interactions has come mostly from relatively specialized antagonistic interactions (but see Johnson et al. 2010), where strong reciprocal selection for phenotype mismatching can promote spatially divergent selection and rapid genetic change in either partner (e.g. Greischar & Koskella 2007). In contrast, in wild Acacia–rhizobial associations, relatively non-specific mutualistic interactions are formed by assemblages of interacting plant populations and multi-species bacterial communities. Such differences are likely to be a key point of contrast with regard to expectations for locally divergent selection in pairwise species interactions (Lajeunesse & Forbes 2002; Kawecki & Ebert 2004; Barrett et al. 2009). Specifically, it seems likely that exposure to a diverse set of local partners may limit the potential for strongly geographically divergent adaptation in soil mutualist associations (but see Hoeksema 2010). Furthermore, we hypothesize that the selection for general compatibility among partners should be strong (Law & Koptur 1986), even if the association was suboptimal in terms of the relative fitness benefits received. This is because plants and soil symbionts can only associate with physically proximate partners, and some nitrogen (or carbohydrate) is likely better than none. For example, plants highly specialized to sympatric rhizobia may be poor competitors compared with more generalist hosts when dispersing outside of local demes, as encounters with compatible rhizobia will be less likely. There is at least some evidence that legumes that are rarer or have more restricted distributions are more specialized with regard to their rhizobial associations (Thrall, Burdon & Woods 2000).

General theory predicts that beneficial species interactions are more likely to emerge and be maintained in poor-quality environments (Thrall et al. 2007a). In plant–soil mutualisms, this is because freely available nutrients can offset the benefits of entering into mutualistic interactions (Streeter 1988). Thus, geographic variation in soil nitrogen concentrations might be predicted to lead to significant spatial structure in the effectiveness of rhizobial communities (Heath, Stock & Stinchcombe 2010; Johnson et al. 2010). Despite the intuitive appeal of this prediction, we could find no evidence that variation in soil nitrogen was associated with variation in rhizobial effectiveness. Similarly, we found no relationships between rhizobial effectiveness and other key soil variables, including soil phosphorus, soil pH and water availability. We noted, however, that our analysis of results from broader surveys using whole-soil inocula demonstrated a significant negative correlation between soil nitrogen levels and the plant growth promoting abilities of microbial communities. This finding supports the idea that beneficial mutualisms may be more likely to evolve under conditions where the traded resource is limiting.

One of the original objectives of this study was to examine how patterns of adaptation in host–symbiont interactions may be influenced by variation in the partner fidelity of host species. While our ability to address this question is obviously limited by the fact we can only compare among two host species, we argue that A. salicina and A. stenophylla represent reasonable points of comparison, given their similar geographic ranges and ecologies. Comparison of these species in terms of partner fidelity is supported by previous work demonstrating distinct, qualitative differences between A. salicina and A. stenophylla in the breadth of bacteria with which they effectively associate. Molecular studies of natural populations of these species across their geographic ranges found a smaller range of rhizobial phylotypes on A. stenophylla than A. salicina (Hoque, Broadhurst & Thrall 2011). Whole-soil inoculation studies across 58 sites also showed that A. stenophylla preferred its own rhizobial communities while A. salicina grew equally well with soils from either host (Thrall et al. 2007b). Consistent with these findings, inoculation trials with a large set of individual rhizobial strains derived from a diverse set of host species indicated that A. stenophylla has a reduced ability to effectively interact with a broad range of isolates compared with A. salicina (Thrall et al. 2007b; P.H. Thrall unpublished data). Thus, at least in a comparative sense, A. stenophylla may be considered a specialist, and A. salicina a generalist.

Our a priori hypothesis was that potential for population-level variability in rhizobial effectiveness, and hence local adaptation, would be stronger in the specialist A. stenophylla than for the generalist A. salicina. However, as discussed earlier, we found no evidence for variation in the effectiveness of specific combinations of host and rhizobial populations. Rather, in our experiments, rhizobial communities associated with the specialist A. stenophylla varied little in effectiveness among populations. In contrast, the major source of variation in the growth of the generalist A. stenophylla was attributable to variation in rhizobial origin. Such differences may develop among the hosts differing in partner fidelity if the structure of rhizobial communities is influenced by positive feedback. For example, for A. stenophylla, a lack of variation in the effectiveness among rhizobial communities may reflect a higher fidelity of that host for a subset of strains, thus promoting the development of generally beneficial rhizobial communities. In contrast, for more generalist interactions, host species may be more likely to associate with diverse strains exhibiting greater variation in effectiveness, promoting the maintenance of more functionally variable rhizobial communities.

In conclusion, we found no evidence for spatially variable GH × GR interactions between host populations and their associated root-nodule bacterial communities, suggesting that selection for local adaptation may play little role in maintaining phenotypic variation in these interactions. However, we found extensive genetic differentiation among populations, as well as significant variation in host response and rhizobial effectiveness for two host species. Moreover, different host species associate with diverse and different (albeit sometimes overlapping) suites of bacterial partners (Hoque, Broadhurst & Thrall 2011; P.H. Thrall unpublished data). Characterizing the spatial and taxonomic scales relevant to understanding the maintenance of diversity in interactions between legumes and rhizobia, and how evolutionary processes might be influenced by local host diversity and environmental heterogeneity, represents a critical step towards a general understanding of plant–soil mutualisms.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

We thank Shamsul Hoque, Jacqui Mackinnon and Jo Slattery for technical support. Three anonymous reviewers made useful comments on a previous version of this manuscript. LGB was able to participate in this research thanks to the support of the Australian Research Council. The financial support of the NSW Environmental Trust is also gratefully acknowledged.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Lay Summary

Figure S1. Plot of the estimate group membership coefficient (Q) for each A.  salicina individual for each cluster (K) based on the most probable number of genetic populations = 5.

Figure S2. Plot of the estimate group membership coefficient (Q) for each A.  stenophylla individual for each cluster (K) based on the most probable number of genetic populations = 6.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
FEC_1940_sm_LaySummary.pdf41KSupporting info item
FEC_1940_sm_FigS1.pdf41KSupporting info item
FEC_1940_sm_FigS2.pdf46KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.