SEARCH

SEARCH BY CITATION

Keywords:

  • arbuscular mycorrhizal fungi;
  • bacteria;
  • fungi;
  • glucosinolates;
  • invasion history;
  • novel weapons hypothesis;
  • plant–soil feedback;
  • terminal restriction fragment length polymorphism (T-RFLP)

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Invaders can gain ecological advantages because of their evolutionary novelty, but little is known about how these novel advantages will change over time as the invader and invaded community evolve in response to each other. Invasive plants often gain such an advantage through alteration of soil microbial communities.
  • In soil communities sampled from sites along a gradient of invasion history with Alliaria petiolata, microbial richness tended to decline, but the community’s resistance to A. petiolata’s effects generally increased with increasing history of invasion.
  • However, sensitive microbial taxa appeared to recover in the two oldest sites, leading to an increase in richness, but consequent decrease in resistance. This may be because of evolutionary changes in the A. petiolata populations, which tend to reduce their investment to allelopathic compounds over time.
  • These results show that, over time, microbial communities can develop resistance to an invasive plant but at the cost of lower richness. However, over longer time-scales evolution in the invasive species may allow for the recovery of soil microbial communities.

Introduction

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

Invasive species can often transform their new communities, and many theories have been proposed to explain their success in the new range (Catford et al., 2009). Many of these theories posit that invaders gain some ecological advantage because of their evolutionary novelty (Callaway & Ridenour, 2004; Hallett, 2006). For example, a lack of coevolutionary history with the native herbivores or pathogens could lead to enemy release (Keane & Crawley, 2002), or the allelochemicals produced by an invasive plant species may be especially toxic to naive native plants (Callaway & Ridenour, 2004).

For many invasive plants, their evolutionary novelty can lead to altered interactions with diverse soil microbial communities. Soil microbial communities can play an important role in driving plant invasions (Callaway et al., 2004; Wolfe & Klironomos, 2005). In many cases, this occurs because the invasive plant changes soil communities in a way that promotes its own growth while inhibiting the growth of native species (Klironomos, 2002). This can be driven by the lack of soil pathogens in the new range (Reinhart et al., 2003, 2005; Blumenthal et al., 2009), the preferential build up of microbial species pathogenic to native plants (Mangla et al., 2008; Beckstead et al., 2010), or the inhibition of species mutualistic with native plants (Stinson et al., 2006; Vogelsang & Bever, 2009).

Altered soil microbial interactions have been suggested to partly underlie the invasion of Alliaria petiolata (garlic mustard) into forest understories throughout eastern North America (Rodgers et al., 2008a). Like most members of the mustard family, this species does not form connections with mycorrhizal fungi, while most of its native competitors benefit from mycorrhizal associations in which certain soil fungi aid plants in nutrient uptake in exchange for fixed carbon. Alliaria petiolata produces a suite of potential allelochemicals, many of which are unique to the species or novel to its new range. These chemicals have been shown to have toxic effects on mycorrhizal fungi (Roberts & Anderson, 2001; Stinson et al., 2006; Burke, 2008; Wolfe et al., 2008). Invasions of A. petiolata may also affect soil microbial communities through more generalized effects on biogeochemical cycles (Rodgers et al., 2008b) or plant diversity (Stinson et al., 2007). By reducing the diversity or abundance of mycorrhizal fungal species, A. petiolata could gain a competitive advantage against native plant species dependent on the mutualism.

The antimycorrhizal action of A. petiolata may be partly a consequence of the evolutionary naivety of fungi in the introduced range, since secondary chemicals of A. petiolata are more toxic to North American vs European arbuscular mycorrhizal fungal (AMF) species (Callaway et al. 2008). Thus, A. petiolata’s allelochemicals may act as a ‘novel weapon’ because of the mismatch of evolutionary histories between the invader and the native fungi. This also suggests that European AMF have evolved some resistance to these compounds and raises the possibility that, given sufficient time, such resistance could evolve in North American species.

If an invader gains an ecological advantage because of its evolutionary novelty, one would predict that invaders with strong impacts would exert pressure on native communities to shift in composition, and on native populations to adapt, in response to these novel traits (Callaway et al., 2005; Carroll et al., 2005; Lau, 2006; Strauss et al., 2006). Depending on the outcome of these ecological and evolutionary responses, this could lead to a reduction in the invader’s ecological advantage. Such a process could lead to ‘boom-and-bust’ invasion dynamics, as have been observed for some invaders (Simberloff & Gibbons, 2004). For example, while many invasive plants benefit from a release from herbivore pressure, some invaders have been found to gradually accumulate herbivores in their native range over time (Strong et al., 1977; Siemann et al., 2006). While the impact of invasive plants on soil communities has been well documented, it is unclear whether these communities will also develop resistance to invaders as herbivore communities have done. Community-level resistance could arise via ecological changes, as sensitive taxa are replaced by more resistant ones, or via evolutionary ones, as resistance traits within populations rise in frequency.

The novel traits of the invader may themselves evolve during the invasion. This evolution may also reduce the impact of the invader, if invaders evolve reduced investment to the novel traits once they have come to dominate an area. Recent research found that A. petiolata populations may evolve lower allelochemical concentrations over time (Lankau et al., 2009), and genotypes with lower concentrations have weaker impacts on soil communities (Lankau, 2010b) [Correction added after online publication 28 October 2010: in the preceding citation, Lankau, in press was corrected to Lankau, 2010b]. This decline in allelochemical concentration may reduce the impact exerted on the soil community, allowing the re-establishment of sensitive species or strains.

Here I compared the microbial communities in soils from sites across a gradient of invasion history with A. petiolata. I used these communities to answer two questions:

  • 1
    Do the richness and composition of microbial communities vary with invasion history?
  • 2
    Are communities with a longer history with A. petiolata more resistant to the invader’s effects?

I specifically looked for evidence of species sorting (a continuous loss of richness and community similarity over time) and recovery (a rise in richness or similarity with increasing age). I predicted that soil communities with a longer history with A. petiolata should be more resistant to additional exposure to A. petiolata. However, it is also possible that recovery of sensitive taxa in older, less toxic A. petiolata populations could lead to decreased resistance in the oldest sites.

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 and estimation of invasion history

Soils were collected from 15 sites in Illinois and Michigan. These sites all had relatively dense A. petiolata populations, but varied in the estimated age of those populations. Population age was estimated from a previous study, which used 650 dated herbarium records to make a spatially kriged map of date of first report for A. petiolata across the eastern half of the USA (Lankau et al., 2009). The age of the A. petiolata at each of the 15 sites sampled for this study was determined from the kriged prediction at their latitude and longitude. While this may not perfectly reflect the specific history of invasion at a given site, it does provide at least broad estimates of regional invasion history, and has been used to successfully predict the allelochemical concentrations and impact on soil communities of specific A. petiolata populations (Lankau et al., 2009; Lankau, 2010b) [Correction added after online publication 28 October 2010: in the preceding citation, Lankau, in press was corrected to Lankau, 2010b]. Site locations and estimated A. petiolata population ages are listed in the Supporting Information Table S1.

Collection

At each site, soil was collected from six distinct locations, all separated by at least 20 m. Soil was collected up to a depth of 15 cm using a 2.5 cm diameter, 15-cm long soil corer (two cores per location), as this layer of soil tends to have the highest root and fungal hyphal density. Three of the samples were taken from directly underneath actively growing A. petiolata plants and mixed; these will be referred to as ‘near A. petiolata’ throughout this paper. The other three samples were located at least 20 m away from any visible A. petiolata plants. These were also mixed, and will be referred to as ‘far from A. petiolata’ throughout. While I can be confident that no A. petiolata plants had grown directly in these sampled soils for at least 2 yr (because of the biennial nature of A. petiolata), it is very likely that these soils had been exposed to A. petiolata in the past, given their proximity to living A. petiolata plants and the patchy nature of A. petiolata populations in space and time (Nuzzo, 1999). Thus, these samples should not be interpreted as controls for exposure to A. petiolata in the past, but rather serve to separate longer-term legacy effects from short-term impacts of an actively growing plant. Soils were collected in the field and stored on ice until brought to the University of Illinois, where they were stored at 4°C for 8 wk until the start of the experiment.

Experimental design

Field soils were used to fill 30 glasshouse pots, one from each location (near or far from A. petiolata) from each site. Each pot was filled with 50 ml of a sterilized background soil, 450 ml of field soil, and then an additional 50 ml of sterilized background soil. As the pots were filled, a subsample of soil was taken from each location at each site and immediately frozen for later DNA extraction. These will be referred to throughout as the ‘initial’ samples. Another subsample was taken for analysis of 14 abiotic factors including: percentage organic matter, concentrations of nitrate, available phosphorous (using Mehlich III extraction and measured on an inductively coupled plasma atomic emission spectrometer), potassium, calcium, and magnesium, soil pH, buffer pH, cation exchange capacity, percent base saturation of potassium (K) and magnesium (Mg), and the percent sand, silt, and clay (A & L Great Lakes Laboratory, Fort Wayne, IN, USA; for protocols see Brown, 1998). Newly germinated A. petiolata seedlings were collected from a nearby population known to produce high root levels of glucosinolates, one of the suspected allelochemicals of A. petiolata (Lankau et al., 2009). Seedlings were collected from a small area (50 × 50 cm) to increase the likelihood of high genetic relatedness. One seedling was planted into each pot, and allowed to grow for 3 months. Seedlings that died were replaced; the majority of replacements occurred in the first 2 wk of growth.

After 3 months of growth, above-ground biomass of the A. petiolata seedlings was removed, a subsample of roots (c. 10 mg) was taken for allelochemical analysis, and a subsample of soil was taken and frozen for later DNA extraction. These samples will be referred to as the ‘final’ samples. Alliaria petiolata biomass was dried for 48 h and weighed. Root tissue was analysed for glucosinolate and flavonoid concentrations following methods described elsewhere (see Lankau et al., 2009; Lankau, 2010a) [Correction added after online publication 28 October 2010: in the preceding citation, Lankau, 2010 was corrected to Lankau, 2010a]. The A. petiolata individuals were chosen to be as genetically homogeneous as possible, so there was relatively little variation in allelochemical concentrations or plant biomass. Neither variation in biomass or allelochemical concentrations explained a significant amount of the variation in microbial community structure.

Analysis of soil microbial communities

Soil communities were analysed by terminal restriction fragment length polymorphisms, in which a common gene region is amplified from mixed community samples with polymerase chain reactions (PCR), and then individual taxa are distinguished based on differences in the placement of restriction enzyme cut sites within the amplified gene region (Kitts, 2001). A total of 500 mg of frozen soil was taken from each of 60 samples (15 sites × 2 locations/site × 2 time-periods, i.e. initial and final) for DNA extraction. DNA was extracted using the FastDNA SPIN Kit for Soil (MP Biomedicals, Santa Ana, CA, USA), followed by an additional chloroform : isoamyl alcohol purification step. The DNA concentrations were standardized to 10 ng μl−1 (based on spectrophotometry) before use as template for PCR. I amplified segments of the ribosomal RNA region by PCRs using a Biometra t-Gradient (Biometra, Goettingen, Germany) using taxon-specific primers for three taxonomic groups: bacteria, fungi in general and AMF in particular. Bacterial PCR targeted the 16S ribosomal RNA region using primer pair 8F and 1492R (Liu et al., 1997; Kitts, 2001), fungal PCR targeted the internal transcribed spacer (ITS) region of the ribosomal RNA gene segment using primer pair ITSFf1 and ITS4r (Koide & Dickie, 2002; St. Laurent et al., 2008) and AMF PCR targeted the small subunit ribosomal RNA region using primer pairs AML1 and AML2 (Lee et al. 2008). The PCR protocols followed those in Liu et al. (1997) for bacteria and St. Laurent et al. (2008) for fungi. Protocols for AMF followed Lee et al. (2008), except that 0.5 g of T4 gene 32 protein (Roche Diagnostics) was added to the PCR mix and PCR cycles were increased to 35 from 30 to improve yields. All forward primers were 6-FAM fluorescence labeled (Operon Biotechnologies, Inc., Huntsville, AL, USA) for detection of fragments with capillary electrophoresis.

Following PCR, products were digested with one restriction enzyme per reaction (Rsa for bacteria, Hha from fungi, and Mbo for AMF; Promega). The Rsa and Hha enzymes were chosen because they provided the most discrimination for their respective taxonomic group in a previous study that used three enzymes per group (Lankau et al., 2010) [Correction added after online publication 28 October 2010: in the preceding citation, Lankau et al., in revision was corrected to Lankau et al., 2010]. Mbo was chosen because it was the most discriminatory of a set of tested enzymes in an in silico digestion of published sequences from Lee et al. (2008). Fragments were sized by capillary electrophoresis on an ABI Prism 3730xl DNA analyser (Applied Biosystems, Carlsbad, CA, USA) using a fluorescent lane standard (ROX1000 for bacteria and fungi, and ROX500 for AMF). Size-calling of the resulting fluorescence peaks was performed using genemapper v 3.0 (Applied Biosystems, Carlsbad, CA, USA) using two basepair allele bins. All analyses were performed using the peak areas for each fragment. Peak area correlates roughly to the relative abundance of the operational taxonomic unit (OTU) represented by that fragment, although this may be affected by primer annealing biases during the PCR (Blackwood et al., 2003). Results were similar when using only peak presence or absence.

Statistical analysis

A mix of multivariate and univariate statistical approaches were used to address the two questions posed above.

Do the richness and composition of microbial communities vary with invasion history?  Community richness was measured as the number of distinct terminal restriction fragment length polymorphism (T-RFLP) peaks in each sample, operating under the generally accepted assumption that each peak corresponds to a unique OTU (Kitts, 2001). For each microbial group (bacteria, fungi, and AMF), an ANCOVA was performed to test whether OTU numbers differed between locations (near or far from A. petiolata), whether they varied linearly or quadratically with the estimated age of the A. petiolata population at the sampled site, and whether this trend differed between the locations.

Community composition was analysed using nonparametric permutation MANCOVAs for each group, using the adonis function in the vegan package in the r statistical language (Oksanen et al., 2005). Location, age (linear and quadratic effects) and their interactions were included as explanatory variables.

In these and all subsequent analyses, differences in abiotic variables were controlled for by first reducing the variance in the 14 measured variables into three principal components (PCs, which cumulatively explained 73% of the variation), and entering the three PCs as additional covariates in the models. The PC loadings for each abiotic variable are listed in Table S2. Latitude and longitude were also included in all models to control for geographic trends, although they were never significant predictors.

Nonmetric multidimensional scaling (NMDS) was used to visualize variation in community composition. A separate ordination of the 30 initial samples was performed for each microbial group. To further visualize the effects of population age tested with the permutational MANCOVA, the scores for each sample on the first NMDS axis were regressed against the estimated age of the A. petiolata population, separately for each location.

Are communities with a longer history with A. petiolata more resistant to the invader’s effects?  Resistance of microbial communities to A. petiolata effects was measured as the degree of change of the community between the initial and final soil samples. For richness, this was calculated simply as the difference in the number of OTUs of the final minus the initial sample. A positive number would imply that more taxa were detected after 3 months of growth with A. petiolata, while a negative number would imply a reduction in detected taxa. Owing to the nature of PCR-based methods, OTU number may not completely reflect the true richness of a sample, especially for broader and hyperdiverse taxonomic groups such as bacteria, where sampling rarefaction curves rarely reach saturation. Thus, differences in the number of OTUs detected in such diverse communities may also represent changes in relative abundance that translates to differences in detection. Nevertheless, differences among samples or between time-points, analyzed with the same methods, could still be informative of relative differences in the detected community composition. Again, ANCOVA was used to test whether the change in richness depended on location, the age of the A. petiolata population (linear and quadratic effects), and their interactions, separately for each microbial group.

Compositional resistance was analysed by calculating the similarity of each sample between the initial and final time-point using the Bray-Curtis similarity metric (also known as the Czekanowski or quantitative Sorensen’s similarity indices, Bray & Curtis, 1957). For two communities, this metric calculates the proportion of species in common divided by the total number of species across both communities (weighted by their abundances), giving a value between 0 and 1. In this case, a high value would indicate little compositional change between the time-points, and thus high resistance. ANCOVAs were used to test the effects of location, age (linear and quadratic effects), and their interaction on this similarity score.

To further elucidate any trends in resistance, each OTU was categorized as either ‘resistant’ or ‘sensitive’ to A. petiolata by calculating the difference in the average peak area of a particular OTU across all initial samples vs all final samples. If this difference was positive, then that OTU tended to decrease in relative abundance after exposure to A. petiolata, and was therefore considered ‘sensitive’. If this difference was negative, then that taxon increased in relative abundance, and was considered ‘resistant.’ I then used ANCOVAs as before with location, age (linear and quadratic effects), and their interaction as explanatory variables, and the number of ‘sensitive’ or ‘resistant’ taxa per sample as the dependent variable. A similar analysis was done using the difference in frequency of occurrence in initial vs final samples – all results were similar between the two analyses and so only the difference in peak area is discussed.

Visually inspecting scatter plots of microbial community structure metrics against estimated population age revealed that the two oldest sampling sites were clear outliers from the linear pattern formed by the other 13 sites (see Figs 1, 2). Therefore, these and all subsequent analyses were performed with all samples and with the two oldest sites removed.

image

Figure 1.  Regressions of the number of operational taxonomic units (OTUs) in initial samples against the estimated age of the Alliaria petiolata population at the sampling site for (a) bacteria, (b) fungi, and (c) arbuscular mycorrhizal fungi (AMF). The best fit quadratic function is shown separately for samples collected near (solid symbols) and far (open symbols) from actively growing A. petiolata plants. Significance tests are presented in Table 1a.

Download figure to PowerPoint

image

Figure 2.  Nonmetric multidimensional scaling ordinations for (a) bacteria, (b) fungi and (c) arbuscular mycorrhizal fungi (AMF) initial communities. Circles represent samples from collected near, and triangles samples collected far, from actively growing Alliaria petiolata plants. The symbol size corresponds to the estimated age of the A. petiolata population at the sampled site (larger symbols = older populations). To aid in interpretation, the scores for each sample along the first ordination axis are graphed against the estimated age of the A. petiolata population for each taxonomic group.

Download figure to PowerPoint

In addition, because the estimation of population ages from herbarium records and spatial kriging likely entailed substantial error, all analyses were performed with two alternative age metrics. First, populations were ranked according to their herbarium estimated age, and then the ranks were used in models in place of the quantitative estimate. Second, sites were lumped into 5-yr bins, resulting in six age categories. The qualitative results did not change compared with the original analysis. Full results from the alternative analyses are available in Tables S3 and S4.

Results

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

Do the richness and composition of microbial communities vary with invasion history?

The initial OTU richness of all three microbial groups (bacteria, fungi, and AMF) varied quadratically with the history of A. petiolata invasion (Table 1a). These patterns were qualitatively similar for samples collected near and far from actively growing A. petiolata plants. The quadratic trend emerged because richness was highest at the youngest sites and declined with age, but then rose again in the two oldest sites (Fig. 1). Without the oldest two sites in the analysis, AMF and bacterial richness instead decreased linearly with increasing invasion history (Table 1a).

Table 1.   Statistical analysis of initial microbial community structure
 (a) OTU richness(b) Community composition
BacteriaFungiAMFBacteriaFungiAMF
FPFPFPR2PR2PR2P
  1. AMF, arbuscular mycorrhizal fungi.

  2. (a) ANCOVA results for the number of operational taxonomic units (OTUs); (b) non-parametric MANOVA results for community composition based on relativized terminal restriction fragment length polymorphism (T-RFLP) peak areas, based on 1000 permutations. Analyses were performed for all samples and for a subset without the two oldest sites, which are strong outliers from the linear pattern. Location refers to whether samples were collected near or far from actively growing Alliaria petiolata plants. Age and Age2 refer to the linear and quadratic effects of the estimated age of the A. petiolata population at the sampled site. Latitude and longitude refer to the geographic location of a sampled site, and PC 1–3 refer to the first three principal components derived from 18 abiotic soil variables measured on each sample. Bold values are significant at P < 0.05. All tests had 1 numerator and 30 denominator degrees of freedom.

All samples
 Location0.000.9610.400.5351.370.2570.030.4760.030.4780.060.044
 Age4.360.0512.840.1084.560.0470.040.1220.040.1850.050.125
 Age214.700.00111.400.0038.980.0080.080.0040.120.0060.080.007
 Age × Location0.050.8292.370.1410.380.5430.030.6590.020.7610.040.233
 Age2 × Location0.180.6750.150.7010.090.7650.020.8700.020.6480.020.715
 Latitude0.950.3420.490.4941.140.3000.030.3510.050.1450.020.593
 Longitude0.930.3490.030.8573.640.0720.050.0740.050.1560.060.032
 PC116.97< 0.00113.330.0022.020.1720.060.0310.060.1150.030.286
 PC22.110.1640.240.6300.030.8660.040.1570.030.3910.050.139
 PC35.600.0291.490.2370.020.8880.050.0540.030.3400.050.105
Without oldest sites
 Location0.060.8040.390.5442.090.1700.030.6660.040.4630.050.190
 Age6.360.0242.770.1176.040.0280.060.0410.090.0430.080.025
 Age20.080.7781.380.2580.270.6110.070.0250.020.7700.060.102
 Age × Location0.010.9132.220.1570.000.9820.020.9190.050.2920.030.581
 Age2 × Location0.060.8170.650.4320.040.8460.050.1110.020.8040.020.859
 Latitude13.340.0030.120.7300.630.4420.030.0370.300.1470.050.185
 Longitude0.720.4120.000.9922.720.1210.050.0520.070.4470.080.023
 PC12.690.12310.750.0050.690.4190.060.0550.060.0660.050.115
 PC21.950.1841.130.3060.090.7720.040.0390.260.5540.050.146
 PC31.520.2392.620.1270.180.6800.050.0610.040.2280.090.012

Similarly, community composition also varied quadratically with invader population age (Table 1b) for all three microbial groups. For AMF only, composition was significantly different between samples collected near and far from A. petiolata, but the pattern with invasion history did not differ between the locations. These patterns are evident in NMDS ordination for AMF and bacteria, where the position of a sample along the first ordination axis showed a quadratic relationship with population age. Communities grew progressively dissimilar along this axis with increasing age, except for communities from the longest invaded sites which were more similar to the youngest ones (Fig. 2a,c). In the absence of the oldest two sites, the composition of general fungi and AMF communities changed linearly with invasion age, while the quadratic pattern remained significant for bacterial communities.

Quadratic trends were not evident in the fungal ordination, despite a significant quadratic effect of age in the non-parametric MANOVA (Fig. 2b). This pattern may have been obscured in the ordination by variation in abiotic factors among soils. For AMF and bacterial communities, the quadratic trend with age was significant whether or not soil abiotic variables were entered as covariates in the model, while for fungal communities the age effect was only significant when controlling for the abiotic variation.

Are communities with a longer history with A. petiolata more resistant to the invader’s effects?

Resistance to A. petiolata effects was determined as the degree of community change between samples taken directly from the field (‘initial’) and the resulting communities after those samples were grown with a common A. petiolata population for 3 months (‘final’). Community change was measured as the difference in number of OTUs, and the Bray–Curtis similarity between community compositions. The change in OTU richness showed consistent quadratic relationship with invasion history for all three taxonomic groups (Table 2), although this effect was only marginally significant for bacterial communities. The youngest and oldest communities tended to have fewer OTUs in the final vs initial samples, while communities with intermediate invasion histories tended to increase in OTU richness in the final samples (Fig. 3). Without the oldest two samples, resistance in terms of OTU richness increased linearly for general fungal and AMF communities. Resistance of bacterial communities was better explained by abiotic and geographic factors rather than A. petiolata population age.

Table 2.   Statistical analysis of microbial community resistance
 (a) Change in OTU richness(b) Bray–Curtis similarity
BacteriaFungiAMFBacteriaFungiAMF
FPFPFPFPFPFP
  1. OTU, operational taxonomic unit; AMF, arbuscular mycorrhizal fungi.

  2. ANCOVA results (a) for the change in the number of OTUs and (b) the Bray–Curtis similarity between the initial and final community (after 3 months of growth with Alliaria petiolata individuals from a common population) for each sample. Analyses were performed for all samples and for a subset without the two oldest sites, which are strong outliers from the linear pattern. Location refers to whether samples were collected near or far from actively growing A. petiolata plants. Age and Age2 refer to the linear and quadratic effects of the estimated age of the A. petiolata population at the sampled site. Latitude and longitude refer to the geographic location of a sampled site, and PC 1–3 refer to the first three principal components derived from 18 abiotic soil variables measured on each sample. Bold values are significant at P < 0.05. All tests had 1 numerator and 30 denominator degrees of freedom.

All samples
 Location0.040.8350.010.9421.860.1900.600.4503.260.0870.110.741
 Age0.400.5356.330.0211.310.2682.650.1210.220.6420.270.612
 Age24.390.0517.010.0168.860.0085.790.0275.870.0268.030.011
 Age × Loc0.310.5830.040.8490.350.5594.560.0473.060.0960.180.677
 Age2 × Loc0.430.5190.330.5720.070.8000.080.7752.090.1641.380.256
 Latitude1.550.2300.680.4192.060.1690.890.3580.200.6633.870.065
 Longitude5.960.0250.030.8702.100.1640.210.6541.980.1760.010.906
 PC18.110.0114.360.0500.290.5941.750.2032.220.1530.670.424
 PC21.730.2051.580.2240.010.9190.001.0000.010.9390.680.422
 PC311.810.0030.920.3510.000.9990.230.6401.050.3181.430.247
Without oldest sites
 Location0.000.9780.300.5914.810.0460.030.8650.970.3410.500.493
 Age1.320.2704.560.0505.290.0373.570.0800.350.5601.540.235
 Age21.980.1812.240.1550.030.8700.010.9151.840.1950.160.697
 Age × Loc0.180.6750.010.9130.150.7012.840.1143.080.1000.240.629
 Age2 × Loc0.230.6390.220.6460.400.5370.160.6950.040.8411.840.196
 Latitude4.560.0510.050.8351.490.2430.070.8000.430.5242.820.115
 Longitude9.720.0080.000.9690.700.4170.000.9981.290.2730.020.897
 PC14.940.0432.940.1070.580.4572.780.1181.830.1960.940.349
 PC20.660.4300.140.7161.160.2990.000.9840.440.5160.620.445
 PC36.930.0201.850.1940.850.3720.080.7791.770.2041.020.330
image

Figure 3.  Regressions of the change in operational taxonomic unit (OTU) number (a–c) and Bray–Curtis similarity (d–e) between initial and final communities for each sample against the estimated age of the Alliaria petiolata population at the sampled site for bacteria (a,d), fungi (b,e) and arbuscular mycorrhizal fungi (AMF) (c,f). The best fit quadratic function is shown separately for samples collected near (solid symbols) and far (open symbols) from actively growing A. petiolata plants. Significance tests are presented in Table 2.

Download figure to PowerPoint

Compositional resistance also showed significant quadratic relationships with invasion history for all three taxonomic groups. However, the direction of this effect differed for AMF communities vs fungal and bacterial communities. In contrast to the pattern with OTU richness, bacterial and fungal compositional resistance was greatest in the oldest sites. The linear increase in resistance was greater in bacterial communities collected near vs far from living A. petiolata plants (Table 2). For fungal communities, samples collected near A. petiolata tended to increase linearly with population age, while those collected far from A. petiolata tended to show a curvilinear pattern, with resistance initially declining but then increasing in the oldest sites. However, the statistical interaction between location and population age was only marginally significant (Table 2).

By contrast, compositional resistance in AMF communities matched the pattern seen in the change in OTU richness. Here, the youngest and oldest sites had the least resistance in both metrics, while communities with intermediate invasion histories had the highest resistance (Fig. 3b). These patterns were dependent on the two oldest sampled sites; without those sites, compositional resistance did not vary significantly with A. petiolata population age (either linearly or quadratically) for any taxonomic group.

For all three taxonomic groups, the number of ‘sensitive’ taxa again showed quadratic trends with age, tending to decline with age until rising again in the oldest sites (with the oldest sites removed, these trends became linear, Table 3). However, a similar pattern was found for ‘resistant’ bacterial and fungal taxa, indicating that this distinction did not separate taxa into groups that responded differently to gradients of invasion history. For bacteria and fungi, the number of ‘sensitive’ and ‘resistant’ taxa was strongly positively correlated across samples (r = 0.66 and 0.78, respectively, P < 0.0001 for both). Conversely, there was no relationship between invasion history and the number of ‘resistant’ AMF taxa in a community (Fig. 4, Table 3), and the correlation between ‘sensitive’ and ‘resistant’ taxa was weaker (although still positive, r = 0.37, P = 0.05).

Table 3.   Statistical analysis of the number of ‘sensitive’ and ‘resistant’ operational taxonomic units (OTUs) in initial samples
 (a) Number of ‘sensitive’ OTUs(b) Number of ‘resistant’ OTUs
BacteriaFungiAMFBacteriaFungiAMF
FPFPFPFPFPFP
  1. AMF, arbuscular mycorrhizal fungi.

  2. ANCOVA results (a) for the number of ‘sensitive’ OTUs, defined as those taxa that decreased in relative abundance on average between initial and final samples and (a) for the number of ‘resistant’ OTUs, defined as those taxa that increased in relative abundance on average between initial and final samples. Analyses were performed for all samples and for a subset without the two oldest sites, which are strong outliers from the linear pattern. Location refers to whether samples were collected near or far from actively growing A. petiolata plants. Age and Age2 refer to the linear and quadratic effects of the estimated age of the A. petiolata population at the sampled site. PC 1–5 refer to the first five principle components derived from 18 abiotic soil variables measured on each sample. Bold values are significant at P < 0.05. All tests had 1 numerator and 30 denominator degrees of freedom.

All samples
 Location0.510.4850.030.8571.500.2370.090.7660.470.5020.140.710
 Age6.660.0185.980.0245.850.0262.910.1042.100.1630.040.852
 Age232.61< 0.000115.210.0019.270.0077.350.0149.590.0061.430.248
 Age × Location1.750.2024.620.0450.010.9180.060.8121.800.1962.730.116
 Age2 × Location0.130.7180.040.8520.030.8760.150.6980.170.6880.270.608
 Latitude2.760.1131.040.3211.870.1890.560.4640.360.5560.110.748
 Longitude3.080.0960.240.6272.260.1500.660.4280.010.9113.700.070
 PC144.34< 0.00019.610.0060.730.4048.400.00912.690.0024.610.046
 PC21.240.2800.000.9540.010.9222.380.1400.300.5920.070.794
 PC36.850.0170.010.9370.100.7505.400.0311.980.1760.170.681
Without oldest sites
 Location0.000.9460.120.7312.030.1760.160.6980.400.5350.690.420
 Age9.820.0075.590.0327.000.0194.430.0532.150.1630.730.406
 Age20.880.3642.850.1120.040.8400.000.9921.060.3191.160.300
 Age × Location0.320.5787.340.0160.000.9820.150.7051.470.2440.020.900
 Age2 × Location0.010.9141.510.2380.200.6580.110.7490.490.4960.290.600
 Latitude2.980.1050.110.7401.100.3121.620.2230.110.7420.050.827
 Longitude3.990.0640.170.6892.170.1621.140.3020.000.9561.780.204
 PC129.19< 0.000111.200.0040.020.8798.300.0119.640.0074.690.048
 PC20.940.3481.810.1990.030.8540.650.4330.920.3520.190.669
 PC34.930.0421.330.2670.220.6482.320.1482.610.1270.010.909
image

Figure 4.  Regressions of the number of sensitive (a–c) and resistant (d–f) operational taxonomic units (OTUs) against the estimated age of the Alliaria petiolata population at the sampled site for bacteria (a,d), fungi (b,e) and arbuscular mycorrhizal fungi (AMF) (c,f). The best fit quadratic function is shown separately for samples collected near (solid symbols) and far (open symbols) from actively growing A. petiolata plants. Significance tests are presented in Table 3.

Download figure to PowerPoint

High initial richness correlated with larger decreases in richness following growth with A. petiolata for all three taxonomic groups (r < −0.7, P < 0.0001 for all). However, initial richness was positively correlated with compositional resistance for bacteria and fungi (r > 0.5, P < 0.001 for both) but was unrelated to AMF compositional resistance (r = −0.14, P = 0.45). The change in OTU richness tended to be negatively related to compositional resistance for bacterial and fungal communities (r = −0.20, −0.37; P = 0.33, 0.03 respectively), but was positively correlated for AMF communities (r = 0.40, P = 0.03). Finally, bacterial and fungal resistances were positively correlated (r = 0.58, P = 0.003), but were uncorrelated with AMF resistance (tending toward negative correlations, r = −0.17, −0.37; P = 0.37, 0.05). See Table S5 for the full table of correlation coefficients.

Discussion

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

Invasive species often gain ecological advantages because of their novelty in a system, leading to altered interactions with native competitors, herbivores and soil communities (Hallett, 2006). However, relatively little is known about the long-term stability of these novel advantages. The ecological impact of the invader could decrease over time if native communities develop resistance to the novel traits of the invader. In addition, evolution in the invader may act to reduce investment to its novel traits. Previous research found that A. petiolata populations tended to evolve reduced investment into allelochemicals over time (Lankau et al., 2009). The reduction in allelochemical concentrations led to weaker impacts on soil microbial communities (Lankau, 2010b) [Correction added after online publication 28 October 2010: in the preceding citation, Lankau, in press was corrected to Lankau, 2010b]. This evolutionary decline in invader impact could potentially allow native microbial communities to recover over long enough time-scales. In this study bacterial, fungal and arbuscular mycorrhizal fungal communities showed both increasing resistance and recovery over a gradient of history with the invasive A. petiolata.

Soil microbial communities collected along a gradient of invasion history showed a complex pattern in which richness and composition progressively changed with longer associations with the invader up to a point, but then appeared to recover in the oldest sites. This is consistent with an initial phase of species sorting, in which microbial species sensitive to the impacts of A. petiolata are progressively lost from communities over time. However, over long enough time-scales, a sufficient decrease in the soil impacts of the A. petiolata population may allow the return of some of these sensitive taxa. This would explain the high richness of the oldest sites, as well as the high compositional similarity of the oldest and youngest sites. This hypothesis was further supported by the analysis of ‘sensitive’ and ‘resistant’ taxa, which found that ‘sensitive’ taxa tended to be common in the youngest sites, declined in abundance with increasing site age, but then rose again in the oldest sites. These results were surprisingly consistent among bacterial, general fungal, and arbuscular mycorrhizal fungal communities and the different metrics of community structure, despite the wide range of diversities and resolution between the groups.

If an invasive plant exerted a constant selection pressure on native species, one would expect resistance to develop through time, such that native communities with a longer association with the invader should show greater resistance to that invader’s effects. However, if the impact of the invader also varied temporally, then the level of resistance in the native community may track the impact of the invader rather than show a progressive increase. In this study, communities showed increasing resistance over time to a point, but then reduced resistance in the oldest sites for some taxa and resistance metrics. Resistance was measured by the degree of change in either taxa richness or community composition between initial samples and the resulting communities after 3 months of growth with A. petiolata individuals from a common, highly allelopathic population. Youngest and oldest communities tended to have the largest decrease in taxa richness – likely the result of the greater number of sensitive taxa in these communities. Samples with intermediate invasion histories, and lower initial richness, were already depauperate in sensitive taxa, and thus showed less change following additional exposure to a highly allelopathic A. petiolata individual. Again, this pattern was consistent across all three taxonomic groups.

The compositional resistance of a community also varied with the history of association with A. petiolata, but in this case the patterns were not consistent among the taxonomic groups. For bacterial and general fungal communities, compositional resistance was highest in the oldest sites, in contrast to the pattern seen for richness. This contradiction may be result in part from the limitations of the analytic method used. The majority of fungal, and about half of bacterial, communities actually had increased OTU richness in the final samples. As there was little opportunity for new microbial taxa to enter the glasshouse pots (except for general contamination from the glasshouse environment, which should have been relatively equal across pots and thus not a major driver of differences along the invasion history gradient), an increase in observed richness may actually reflect an increase in evenness. Polymerase chain reaction-based community fingerprint methods, especially for highly diverse groups, are likely to miss taxa that are not at relatively high abundance. If exposure to A. petiolata reduced the dominance of some highly abundant taxa, and thus increased the relative abundance of other taxa, this could have lead to an apparent increase in richness. Compositional resistance, as measured by the Bray–Curtis similarity index, is simply a function of the number and abundance of taxa common to both the initial and final community, divided by the total number of taxa in both. Thus, similarity would be decreased both by a large reduction in taxa, but also by a large increase. In fact, a close examination of Fig. 3 shows that the youngest fungal communities tended to show a reduction in OTU richness, intermediately aged communities showed an increase, and the oldest ones had very little change. This led to a generally increasing trend in compositional resistance with age, but for different reasons at different invasion ages.

This pattern is also likely influenced by the responses of ‘sensitive’ and ‘resistant’ taxa. For bacterial and fungal communities, the richness of ‘sensitive’ and ‘resistant’ taxa was highly correlated (r > 0.66 for both), indicating that these groups did not respond differentially to gradients in A. petiolata population age or per capita impact. As most fungal and bacterial communities were dominated by ‘resistant’ taxa, the increase in ‘sensitive’ taxa in older sites may have been overwhelmed by a concomitant increase in ‘resistant’ taxa.

By contrast, AMF communities showed the same pattern for richness and composition, with lowest resistance in the oldest and youngest sites. These differences among taxonomic groups may result partly from the resolution of the methods for each group. The AMF communities in soil have a much lower taxonomic diversity than general bacterial or fungal communities, and thus a larger percentage of available taxa were probably amplified by the PCR. The majority of AMF communities had fewer OTUs detected in the final vs initial samples, implying a general loss of taxa following exposure to A. petiolata, likely because there were fewer taxa ‘missed’ in the initial sample that could appear in the final one. Thus the oldest and youngest communities had the lowest compositional resistance primarily because of the loss of taxa in the final sample, while intermediately aged communities had higher similarity because they had relatively minor changes in OTU richness.

The AMF communities also differed from general bacterial and fungal communities in that the majority of AMF taxa were classified as ‘sensitive’ rather than ‘resistant’, and while there was a significant quadratic relationship between invasion history and number of ‘sensitive’ taxa, there was no significant relationship with ‘resistant’ taxa. Thus, while the oldest sites regained some ‘sensitive’ taxa, this gain was not balanced with an increase in ‘resistant’ types, leading to decreased overall resistance for those communities. However, this pattern may be influenced by the higher resolving power of the AMF primers.

While the patterns in microbial communities across the gradient of invasion history were quite consistent across taxonomic groups and several measures of composition and resistance, ultimately, the number of sampled communities was somewhat limited. This is especially true for the oldest sites, where only two sites at the far end of the sampled spectrum had a strong influence on the observed quadratic trends. When those two sites were removed from the analysis, microbial community structure and resistance generally showed linear patterns with invasion history. The quadratic patterns remained strong when controlling for a host of abiotic variables as well as geography, so the results cannot be explained by unusual conditions at these particular sites that are independent of A. petiolata invasion history. Nevertheless, future studies that include more samples from long-invaded sites and other areas of the invaded range are warranted.

The mechanisms underlying A. petiolata’s effects on soil microbial communities are likely multifaceted, including the well-studied allelopathic effects (Vaughn & Berhow, 1999; Roberts & Anderson, 2001; Wolfe et al., 2008), as well as more general effects on nutrient cycling (Rodgers et al., 2008b) and plant diversity (Stinson et al., 2007; Rodgers et al., 2008a). While the responses of bacterial and general fungal communities were highly correlated, neither of these broader community’s responses correlated with those of AMF communities. The AMF communities were also the only ones to show a compositional difference between samples taken near and far from actively growing A. petiolata plants. It is possible that the responses of the functionally and taxonomically diverse bacterial and general fungal were driven by generic aspects of A. petiolata invasion (changes to nutrient cycling, litter biomass, etc.), while the response of the more functionally and taxonomically restricted AMF guild were also influenced by more specific aspects of the invader, such as its allelochemicals (Stinson et al., 2006; Wolfe et al., 2008).

A secondary goal of this study was to compare microbial communities collected near and far from actively growing A. petiolata plants at each site. The AMF community composition differed significantly between samples taken near and far from A. petiolata individuals, consistent with the results of previous studies (Roberts & Anderson, 2001; Stinson et al., 2006; Burke, 2008). However, similar patterns were not evident for general bacterial or fungal communities, and in general the temporal patterns were similar for near and far samples. As A. petiolata populations can be extremely patchy in space and time, and the samples taken from ‘far’ from A. petiolata were still relatively close to the current population, it is likely that these soils had been exposed to A. petiolata in the past. This suggests that the kinds of community changes seen in this study are the result of many generations of exposure, rather than reflecting the effects of individual A. petiolata plants. The patchiness of A. petiolata populations may also help explain why 20 yr (the estimated age of the youngest site) was not long enough to extirpate all of the sensitive taxa from a community. It likely takes many years, even decades, for an A. petiolata population to completely and consistently fill all of the available understory habitat (Nuzzo, 1999), and during that time the impact of A. petiolata on the soil communities will be limited by temporal and spatial patchiness.

While there was evidence for increasing resistance of soil communities over time, it was complicated by a countervailing pattern of declining per-capita impacts of the invader. These patterns combined to give nonlinear relationships, in which soil communities declined in richness initially, as the invader selected for resistant microbial taxa and extirpated sensitive ones, but then regained some of this richness after invasions reached a certain stage. This may have resulted from a lag between ecological and evolutionary effects. In the initial stages of the invasion at a given site, the per-capita impact of the invader is likely relatively constant and the dominant effect on soil communities results from the cumulative impact of the invader over years (and the increasing density of the invader). However, over longer time-scales evolutionary changes in the invader may become important. In this case, A. petiolata tends to evolve reduced per-capita impact over time, owing to a reduction in allelopathic traits. Once this impact has been reduced far enough, sensitive microbial taxa may be able to re-establish in the invaded site. If these microbes play an important role in the establishment and growth of native plants, than this microbial recovery may also aid in the recovery of native plant communities. Consistent with this prediction, Lankau et al. (2009) found that native woody species tended to increase in abundance over a 5-yr interval in sites with the longest history of A. petiolata invasion, while they declined in abundance over the same interval in more recently invaded sites.

The results have important implications for how invasive species are studied and managed. The primary mechanisms driving invasions may vary through the different stages of invasions, and the impact of particular invasive species may also change over time because of changes both in the invader and the invaded community. However, most studies of species invasions do not explicitly consider the history of the species or population in question; Strayer et al. (2006) found that 40% of the studies they reviewed did not even report the number of years since the focal species was first introduced to a new continent. The current results suggest that researchers could come to very different conclusions about the threat posed by a given invader if their research is limited in spatial and temporal scale. These results also suggest that management of invasive species could benefit from understanding how invasive impacts change over time. For example, the recovery of soil communities in areas with the longest history of A. petiolata invasion suggests that restoration of native plant species may be successful in these areas, even without eradicating the invader. By contrast, in younger sites eradication may be the prime goal, to remove the invader before soil communities are severely degraded.

Acknowledgements

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

I thank Joanne Chee-Sanford and Adam S. Davis for access to equipment, and Emily Wheeler for help with molecular techniques and comments on the manuscript. This research was funded by USDA-NRI grant # 2007-02894 and NSF DEB grant # 0918450.

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

Table S1 Locations and estimated Alliaria petiolata population ages at the sampled sites

Table S2 Loadings of soil abiotic variables on the first three principal components

Table S3 Statistical analysis of initial microbial community structure using alternative metrics of population age

Table S4 Statistical analysis of microbial community resistance using alternative metrics of population age

Table S5 Correlation matrix for measures of microbial community composition and resistance

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

FilenameFormatSizeDescription
NPH_3481_sm_TableS1-5.doc175KSupporting info item