Ecology Letters, (2014) 17: 1613–1621
LETTER
Brian L. Anacker,1* John N. Klironomos,2 Hafiz Maherali,3 Kurt O. Reinhart4 and Sharon Y. Strauss1
doi: 10.1111/ele.12378
Phylogenetic conservatism in plant-soil feedback and its implications for plant abundance Abstract We examined whether plant-soil feedback and plant-field abundance were phylogenetically conserved. For 57 co-occurring native and exotic plant species from an old field in Canada, we collected a data set on the effects of three soil biota treatments on plant growth: net whole-soil feedback (combined effects of mutualists and antagonists), feedback with arbuscular mycorrhizal fungi (AMF) collected from soils of conspecific plants, and feedback with Glomus etunicatum, a dominant mycorrhizal fungus. We found phylogenetic signal in both net whole-soil feedback and feedback with AMF of conspecifics; conservatism was especially strong among native plants but absent among exotics. The abundance of plants in the field was also conserved, a pattern underlain by shared plant responses to soil biota. We conclude that soil biota influence the abundance of close plant relatives in nature. Keywords AMF, Arbuscular mycorrhizal fungi, exotic, Glomus etunicatum, introduced, old field, phylogenetic signal, phylogeny, plant community assembly, plant-soil feedbacks. Ecology Letters (2014) 17: 1613–1621
Plant interactions with macro-mutualists and macro-antagonists often exhibit phylogenetic signal, defined as the tendency for relatives to resemble each other more closely in their characteristics than expected by chance (Harvey & Pagel 1991). Moreover, evolutionary history is a useful predictor of the distribution and abundance of species (Donoghue 2008), species associations (Cavender-Bares et al. 2009) and the outcomes of species interactions (Becerra 1997; Agrawal & Fishbein 2006; Weiblen et al. 2006). In contrast, relatively few studies have examined whether reciprocal effects between plant growth and soil community development (plant-soil feedback) are shared above the level of species (i.e. plant clades) (reviewed below), despite the important linkage between soil biota and plant growth and plant abundance in natural communities (van der Putten et al. 1993; Bever et al. 1997; Klironomos 2002; Wardle et al. 2004; Bezemer et al. 2006; Kardol et al. 2006; Reinhart et al. 2012). Conservatism in species responses to soil biota has direct implications for species invasibility and coexistence. For example, if close relatives share positive soil feedbacks, then occupancy by one species will favour co-establishment of close relatives and coexistence. In contrast, if relatives share negative soil feedbacks, then an invader might experience biotic resistance by a community with an already-established close relative. Understanding the degree of phylogenetic conservatism in plant responses to soil biota can therefore shed light on the cause of conflicting reports of whether species are better (e.g. Duncan & Williams 2002) or poorer (e.g. Strauss
et al. 2006) invaders of communities with close relatives. If plant clades share positive soil feedbacks and typically thrive in each others’ soils while others share negative soil feedbacks and therefore cannot stably coexist, then assessing the degree of phylogenetic signal in plant-soil feedbacks can assist in developing strategies for ecological restoration. Several studies have shown that close plant relatives have similar plant-soil feedbacks, but these are all based on a small number of soil microbe or plant species. Reinhart et al. (2012) found significant phylogenetic signal in the response of 95 plant species of tallgrass prairies to two arbuscular mycorrhizal fungi (AMF) species in the genus Glomus. For seven subtropical tree species, close relatives were shown to have more similar negative growth responses to soil fungal pathogens (Liu et al. 2012). Soils trained by two native Potentilla species caused strong negative feedbacks on the growth of a distantly related, non-native Potentilla species (Callaway et al. 2013). Experiments with six grass species showed that close relatives had similar plant-soil feedbacks (Brandt et al. 2009). Plants in the Asteraceae family responded similarly to the application of a systemic fungicide (Bennett & Cahill 2013). Introduced species in New Zealand experienced negative soil feedback in conspecific and congeneric soils, but not in distant relative soils (Diez et al. 2010). These studies suggest that evolutionary processes, like strong developmental constraints or the absence of genetic variation, could interact with ecological processes to determine the nature of plant-soil feedbacks. Phylogenetic signal in plant-soil feedback would, however, need to be quite strong in order to be detected in the face of context-dependent variation in the interaction. Communities
1
3
Department of Integrative Biology, University of Guelph, Ontario, Canada
95616, USA
4
Fort Keogh Livestock and Range Research Laboratory, USDA-Agricultural
2
Research Service, Miles City, MT, 59301, USA
INTRODUCTION
Department of Evolution and Ecology, University of California, Davis, CA, Department of Biology, University of British Columbia, Okanagan campus,
Kelowna, British Columbia, Canada
*Correspondence: E-mail: [email protected]
© 2014 John Wiley & Sons Ltd/CNRS
1614 B. L. Anacker et al.
of soil biota are often highly labile in time and space (Kardol et al. 2006; Reinhart & Clay 2009). Plant–microbe interactions can shift from mutualistic to neutral or parasitic with increasing nutrient conditions (e.g. Hoeksema et al. 2010) and can vary with interspecific plant competition and plant density (e.g. Casper & Castelli 2007). Exotic plant species, in particular, may add additional variation and limit the potential for detecting phylogenetic signal, as many exotic plant species are relatively unresponsive to soil mutualists and soil enemies (Klironomos 2002; Kourtev et al. 2002; Reinhart & Callaway 2006; Seifert et al. 2009; Verhoeven et al. 2009; Bennett & Strauss 2013). Klironomos (2002) suggested that native species exhibited strong negative feedback while dominant exotic species accumulated pathogens at slower rates, reflecting an enemy release phenomenon (Engelkes et al. 2008). Here, we synthesise ecological and evolutionary information to gain greater insights into which components of plant-soil feedbacks predict the relative abundance of plants within communities. We expand on a 2002 study by Klironomos that showed that plant abundance in an old field was correlated with plant-soil feedback (i.e. plant rarity was linked to negative soil feedbacks), as measured by how well a species grew in soils with a history of conspecific vs. heterospecific plants. We incorporate data from two additional experiments on soil feedbacks with mycorrhizal fungi, one of which is newly reported here, that each use the same set of 57 plant species. Importantly, we also test whether the mechanistic linkage between soil feedbacks and plant abundance remains with the consideration of phylogeny. Because we find that plant relative abundance is conserved, it becomes important to test if phylogeny can explain more variation in plant abundance than soil feedback, under the assumption that phylogeny acts as a reliable proxy for additional, unmeasured aspects of plant functional properties and niche space (e.g. seed size, competitive ability, dispersal ability) that might be stronger determinants of field abundance than soil feedback. Our data set represents one of the best available for examining phylogenetic conservatism in plant-soil feedback, as it includes a diversity of native and exotic plant species from a natural community (25 natives; 32 exotics), allows simultaneous examination of plant response to both whole-soil communities and mycorrhizal fractions, and links the effects of soil biota on plant growth to plant abundance in the field.
Letter
Plants were grown in soils with a history of conspecifics and in soils with a history of heterospecifics. The difference in plant growth between the conspecific and heterospecific soil treatments describes the soil feedback, which represents the net effect of positive feedback with soil mutualists and negative feedback with soil antagonists, as well as possible changes in nutrients. Thus, we refer to it as ‘net whole-soil feedback’. The experiment had three stages; two stages to culture the soils followed by a third to measure plant growth. In stage one, each of the 57 plant species were grown in a representative field soil from LTMRS for 10 weeks, after which plants were discarded. In stage two, another seedling of the same species was grown for 10 additional weeks and then discarded. In stage 3, another seedling was grown in 10 pots containing soils conditioned by conspecifics and 10 pots containing soils conditioned by heterospecifics for 10 weeks. At the end of stage 3, plants were harvested, air-dried and weighed. Using stage 3 plants, net whole-soil feedback was calculated as: ln (biomass of plant grown in soil with a history of conspecifics/ biomass of plant grown in soil with a history of heterospecifics). We note that plant growth in these soils may reflect species-specific effects of nutrient depletion, in addition to differences in soil communities (see Discussion). Experiment 2: Feedback with AMF of conspecifics
We assembled data from three experiments, two of which are previously published and one of which is described here for the first time (Experiment 2). The 57 plant species (Table S1) in our study occur at the Long-Term Mycorrhizal Research Site (LTMRS), located within the University of Guelph Arboretum, Guelph, Canada (43° 320 27″ N, 80° 120 54″ W). Each species was present in each experiment.
To isolate the unique effects of feedback with AMF communities, we assessed plant response to AMF species collected from soils that contained conspecific plants in the field vs. controls without AMF additions using the same 57 species from experiment 1. We collected a 300 mL soil core from each of 5–7 haphazardly located plants of each species, bulked the soil samples and extracted the spores using centrifugation (methods in Sikes et al. 2012). Plants were grown in 20-cm diameter, 3 L pots containing a 4 : 1 ratio of sand to Promix BX (Premier Tech, Riviere-du-Loup, QC, USA). The potting soil was sterilised prior to planting by exposing it to 32 kGy ( 10%) of gamma-irradiation over 7 days (McMaster University Nuclear Reactor). For each species, 10 pots were each seeded with ~250 AMF spores; an additional 10 pots were used as controls. During a 2-week seedling establishment period, pots were watered daily, and then watered 2–3 times per week. All plants were fertilised with 300 mL of a ¼ strength Hoagland’s solution every 2 weeks. Aboveground plant parts were harvested after 348–372 days growth, dried and weighed. The experiments were carried out over 3 years (subsets of species were grown in year-long trials that began in May 2003). Feedback with AMF of conspecifics was quantified as: ln(biomass of plant grown in soil with AMF inocula/biomass of plant grown in soil without AMF inocula). While this and the following metrics are not feedback measures in a classical sense (i.e. did not have home vs. away comparisons), we refer to them as soil feedback metrics for simplicity.
Experiment 1: Net whole-soil feedback
Experiment 3: Feedback with Glomus etunicatum
The goal of experiment 1, described in Klironomos (2002), was to describe plant-soil feedback, where plant growth influences soil communities, which in turn influence plant growth.
The third experiment, previously described by Klironomos (2003), assessed feedback with the most common mycorrhizal species at the LTMRS (Maherali & Klironomos 2012). This
MATERIALS AND METHODS
© 2014 John Wiley & Sons Ltd/CNRS
Letter
experiment allowed us to contrast feedback with a widespread, generalist AMF, Glomus etunicatum, the single most commonly observed AMF species across the study site (Maherali & Klironomos 2012), with feedback from host-specific AMF from experiment two. The same 57 plant species described above were grown either with G. etunicatum or a non-mycorrhizal control. Field soil, as described in experiment one, was mixed with silica and sand at a 1 : 1 ratio, autoclaved and placed in 15-cm diameter, 10 L pots. A band of AMF inoculum, composed of Allium porrum roots either precolonised by G. etunicatum collected from the site or nonmycorrhizal, equalling 1 g in weight, was added 2 cm below the soil surface. Each species by G. etunicatum treatment combination had 10 replicates. Plants were grown for 16 weeks and then harvested, dried and weighed. Percentage change was calculated as the percentage difference in biomass between mycorrhizal and non-mycorrhizal treatments (Klironomos 2003). We converted per cent change to a log response ratio: -ln(1per cent change/100) (Allison & Goldberg 2002). Plant abundance
Field abundance was determined by measuring the occurrence of species in 100 randomly located quadrats at LTMRS, measured in both 1998 and 2000 and pooled, as reported in Klironomos (2002). Plant abundance was the number of quadrats each species occupied and ranged from 1 to 88 ( x = 24.43). Plant abundance was log-transformed to meet the assumptions of parametric analysis. Phylogeny
We created a molecular phylogeny for the 57 plants species using maximum likelihood. Details are provided in the supporting information. For two species, sequences of congeners were used as substitutes (Table S2). In these two cases, the species were the lone representatives of their genera in our study. Statistical analysis
All analyses were performed in R version 3.0.2 (R Development Core Team, R. 2014). Effects of soil biota treatments We tested whether the soil biota treatments had effects on plant growth relative to controls (i.e. if they significantly differed from zero). We also tested if the three soil feedback metrics were correlated with one another. To account for any possible phylogenetic non-independence in the correlations among the feedback metrics, we fit phylogenetic generalised least squares (PGLS) regression models using the ‘pgls’ function in the ‘caper’ library. PGLS is described in more detail below. Phylogenetic signal in soil feedback We calculated the K statistic (Blomberg et al. 2003) for the three feedback metrics, using the ‘phylosignal’ function in the ‘picante’ library. K values < 1 indicate that species resemble
Conservatism in plant responses to soil biota 1615
each other less than expected under Brownian motion evolution; K values > 1 indicate that species resemble each other more than expected. The significance of K was assessed by comparison to a random shuffle of trait values at the tips of the phylogeny. We also tested for phylogenetic signal in plant abundance. The abundance of plant taxa may be phylogenetically conserved because plants with similar abundances in any given community may share traits that are linked to abundance and also phylogenetically conserved. For example, ruderal plant species are likely to dominate initially during secondary succession due to shared traits like fast growth rates, small seed size, dispersal ability, etc. Seed size, for example, is phylogenetically conserved (Moles et al. 2005). Thus, a plant’s relative abundance may be an expression of many traits that are conserved and reflect aspects of its biology other than soil feedbacks. Relationship of the three soil feedback metrics to plant abundance in the field We fit all possible models of plant abundance using the three soil feedback metrics. Model fit was compared using AICc and model weights (Burnham & Anderson 2004). We preferred to use PGLS for our regression models, rather than phylogenetic independent contrasts (PICs). It has recently been pointed out that phylogenetic regression is only appropriate when there is phylogenetic signal in model residuals, reflecting the regression assumption of independent errors (Revell 2010). In cases where there is no signal in the residuals, despite relatively strong signal in both the dependent and independent variables, using a model based on PICs can actually lead to inflated errors. The PGLS method we used avoids this problem by estimating phylogenetic signal and regression parameters simultaneously, allowing for a PGLS model to equal an ordinary least squares model when phylogenetic signal in the model residuals are absent (k = 0); likewise, a PGLS model can equal a model fit with PICs when phylogenetic signal in the model residuals are strong (k = 1). A drawback to our model fitting technique is that we cannot statistically compare the models using conventional likelihood ratio tests, as they are fit using different branch length transformations internal to the PGLS function. To test the statistical significance for two model comparisons in particular, however, we refit the constituent models with a single branch length transformation, which was set to be equal to the average k value obtained for the constituent models, and compared the fit of the models using likelihood ratio tests. Distinguishing the effects of phylogeny and soil feedback on plant abundance in the field We used variance partitioning to parse the variation of plant abundance among the soil feedback metrics and phylogenetic components (Desdevises et al. 2003; Peres-Neto et al. 2006). The method required that we create three standard multiple regression models. Our first model (Soil) was a regression of plant abundance on the three soil feedback measures. Our second model (Phylo) was a regression of plant relative abundance on a set of principal coordinates (PCs) that represent the phylogeny. For this, we decomposed the phylogeny into a © 2014 John Wiley & Sons Ltd/CNRS
1616 B. L. Anacker et al.
set of n1 orthogonal PCs, where n equals the number of species, using the ‘PVRdecomp’ function of the ‘PVR’ package. To select the PCs to be included in the multiple regression model, we regressed each PC against plant abundance, retaining those that were significantly related at alpha 0.05. We then fit a multiple regression model using only the retained PCs. We retained two PCs (PC2 and PC4). All retained PCs represented deep divergences in the phylogeny, and thus trait values that show correlations with such PCs could have evolved along the tree and reflect niche conservatism. Criticisms of this approach come from PC axes that contrast paraphyletic groups, and for which it is therefore harder to interpret how a trait might evolve in such a manner (Freckleton et al. 2011). In our case, we only used PC axes that contrasted whole clades. Our third model (SoilPhylo) was a multiple regression that included the terms from the Soil and Phylo models. We derived the four constituent components via subtraction of the R2 values from the three multiple regression models described above as follows (Desdevises et al. 2003): Soil feedback only R2 = SoilPhylo R2Phylo R 2 Phylogeny only R2 = SoilPhylo R2Soil R2 Shared = Soil R2 + Phylo R2SoilPhylo R2 Residuals = 1SoilPhylo R2 The soil biota only R2 represents the relationship of plant abundance and soil feedback after taking into account phylogenetic autocorrelation. Significance testing is possible for the soil feedback only component and the phylogeny only component using redundancy analysis, but not for the shared fraction or the residuals (Peres-Neto et al. 2006). Soil feedbacks in natives and exotics We tested if the three soil feedback metrics differed by plant native status (native/exotic) using PGLS. We repeated the phylogenetic signal K tests using natives only (n = 25) and exotics only (n = 32) and tested for phylogenetic signal in native status itself, using the ‘phylo.d’ function in the ‘caper’ library. We also added plant native status (exotic or native) as explanatory variables to our models of plant abundance. Finally, we repeated the variance partitioning exercise for natives and exotics separately.
RESULTS
Effects of soil biota treatments
We found large differences in the direction of our three feedback measures (Fig. 1). Plant growth was lower in the presence of host-specific soil biota than it was in the presence of soil biota of heterospecific plants (i.e. negative net whole-soil feedback; x = 0.14; P < 0.001). In contrast, plants grew much larger in the presence of conspecific mycorrhizal communities vs. controls (i.e. positive feedback with AMF of conspecifics; x = 0.25, P < 0.001), but not in the presence of the dominant G. etunicatum vs. controls (i.e. neutral feedback with G. etunicatum; x = 0.06, P = 0.10). The three different measures of soil feedback were unrelated to one another (net whole-soil feedback vs. feedback with G. etunicatum, P = 0.56; net whole-soil feedback vs. feedback with AMF of © 2014 John Wiley & Sons Ltd/CNRS
Letter
conspecifics, P = 0.80; feedback with G. etunicatum vs. feedback with AMF of conspecifics, P = 0.79; tests accounted for phylogeny via PGLS). Results were qualitatively similar to results based on simple correlation analysis (P = 0.52, 0.22, 0.63, respectively). Phylogenetic signal in soil feedback
We found significant phylogenetic signal in net whole-soil feedback, even stronger signal in feedback with AMF of conspecifics, and no signal in feedback with G. etunicatum (Table 1; Fig. 1). Plant abundance was weakly conserved (K = 0.19; P = 0.06), which could reflect both shared response to soil biota by relatives, or other traits influencing abundance that are phylogenetically conserved. Some clades contributed greatly to the overall phylogenetic signal. Positive net wholesoil feedback and high plant abundances characterised the predominately native, later-flowering Asteraceae tribe Astereae (Fig. 1) and negative net whole-soil feedback characterised the monocots. On the other hand, feedback with AMF of conspecifics was strong and positive for monocots ( x = 0.68) but not dicots ( x = 0.13). Relationship of three soil feedback metrics to plant abundance in the field
Net whole-soil feedback explained the largest proportion of the variance in plant abundance (51%; model 5 in Table 2) after accounting for phylogenetic non-independence. The multivariate models identified complementarity among the three metrics of soil feedback (Table 2). Specifically, two models were equally good at explaining plant abundance: one incorporated net whole-soil feedback and feedback with G. etunicatum; our likelihood ratio test confirmed that the addition of feedback with G. etunicatum significantly improved the model over the originally published model with only net whole-soil feedback (P = 0.04). A model that included all three metrics, including feedbacks from AMF of conspecifics, was not significantly different from, i.e. equally good as, the best model (P = 0.14; Table 2, model 2). While the very best model of abundance does not include feedback with AMF of conspecifics, we argue that it may be a mistake to dismiss the importance of AMF communities on patterns of field abundance, as the feedback with AMF of conspecifics was very strong, contained the strongest phylogenetic signal and was unique from (i.e. uncorrelated with) the other two feedback metrics (Fig. 1). Distinguishing the effects of phylogeny and soil feedback on plant abundance in the field
The strong relationship between field abundance and soil feedback (adj. R2 = 54.4 for model with the three feedback metrics as explanatory variables) was partitioned into a soil feedback component (adj. R2 = 38.1), a phylogeny component (adj. R2 = 0.0) and a soil feedback + phylogeny component (adj. R2 = 16.3), the latter representing the correlation between the PCs and soil feedback (Fig. 2). The soil feedback component was significant, indicating that the relationship
Letter
Conservatism in plant responses to soil biota 1617
K = 0.27 (< 0.01)
K = 0.42 (< 0.001)
K = 0.19 (0.06)
K = 0.11 (0.62)
Negative feedback Positive feedback n = native n n n n n n n n n n
n n n n n
n n n n n
n n n n n n
–0.4 –0.2 0 0.2 Net whole-soil feedback
1.5 –0.4 0 0.4 0.8 –0.5 0 0.5 1 Feedback with AMF Feedback with Glomus of conspecifics etunicatum
0
1 2 3 4 ln(Plant abundance)
Figure 1 Phylogenetic conservatism in three plant-soil feedback metrics and plant abundance for 25 native and 32 exotic plant species (delineated by ‘n’ on phylogeny tips). Black bars indicate negative plant-soil feedback; hashed bars indicated positive soil feedback.
Table 1 Phylogenetic conservatism in species attributes according to Blomberg’s K, where 0 = random and 1 = Brownian
All species Variable Plant abundance Net whole-soil feedback Feedback with AMF of conspecifics Feedback with G. etunicatum
Natives
Exotics
K
P
K
P
K
P
0.19 0.27 0.42
0.06
LETTER
Brian L. Anacker,1* John N. Klironomos,2 Hafiz Maherali,3 Kurt O. Reinhart4 and Sharon Y. Strauss1
doi: 10.1111/ele.12378
Phylogenetic conservatism in plant-soil feedback and its implications for plant abundance Abstract We examined whether plant-soil feedback and plant-field abundance were phylogenetically conserved. For 57 co-occurring native and exotic plant species from an old field in Canada, we collected a data set on the effects of three soil biota treatments on plant growth: net whole-soil feedback (combined effects of mutualists and antagonists), feedback with arbuscular mycorrhizal fungi (AMF) collected from soils of conspecific plants, and feedback with Glomus etunicatum, a dominant mycorrhizal fungus. We found phylogenetic signal in both net whole-soil feedback and feedback with AMF of conspecifics; conservatism was especially strong among native plants but absent among exotics. The abundance of plants in the field was also conserved, a pattern underlain by shared plant responses to soil biota. We conclude that soil biota influence the abundance of close plant relatives in nature. Keywords AMF, Arbuscular mycorrhizal fungi, exotic, Glomus etunicatum, introduced, old field, phylogenetic signal, phylogeny, plant community assembly, plant-soil feedbacks. Ecology Letters (2014) 17: 1613–1621
Plant interactions with macro-mutualists and macro-antagonists often exhibit phylogenetic signal, defined as the tendency for relatives to resemble each other more closely in their characteristics than expected by chance (Harvey & Pagel 1991). Moreover, evolutionary history is a useful predictor of the distribution and abundance of species (Donoghue 2008), species associations (Cavender-Bares et al. 2009) and the outcomes of species interactions (Becerra 1997; Agrawal & Fishbein 2006; Weiblen et al. 2006). In contrast, relatively few studies have examined whether reciprocal effects between plant growth and soil community development (plant-soil feedback) are shared above the level of species (i.e. plant clades) (reviewed below), despite the important linkage between soil biota and plant growth and plant abundance in natural communities (van der Putten et al. 1993; Bever et al. 1997; Klironomos 2002; Wardle et al. 2004; Bezemer et al. 2006; Kardol et al. 2006; Reinhart et al. 2012). Conservatism in species responses to soil biota has direct implications for species invasibility and coexistence. For example, if close relatives share positive soil feedbacks, then occupancy by one species will favour co-establishment of close relatives and coexistence. In contrast, if relatives share negative soil feedbacks, then an invader might experience biotic resistance by a community with an already-established close relative. Understanding the degree of phylogenetic conservatism in plant responses to soil biota can therefore shed light on the cause of conflicting reports of whether species are better (e.g. Duncan & Williams 2002) or poorer (e.g. Strauss
et al. 2006) invaders of communities with close relatives. If plant clades share positive soil feedbacks and typically thrive in each others’ soils while others share negative soil feedbacks and therefore cannot stably coexist, then assessing the degree of phylogenetic signal in plant-soil feedbacks can assist in developing strategies for ecological restoration. Several studies have shown that close plant relatives have similar plant-soil feedbacks, but these are all based on a small number of soil microbe or plant species. Reinhart et al. (2012) found significant phylogenetic signal in the response of 95 plant species of tallgrass prairies to two arbuscular mycorrhizal fungi (AMF) species in the genus Glomus. For seven subtropical tree species, close relatives were shown to have more similar negative growth responses to soil fungal pathogens (Liu et al. 2012). Soils trained by two native Potentilla species caused strong negative feedbacks on the growth of a distantly related, non-native Potentilla species (Callaway et al. 2013). Experiments with six grass species showed that close relatives had similar plant-soil feedbacks (Brandt et al. 2009). Plants in the Asteraceae family responded similarly to the application of a systemic fungicide (Bennett & Cahill 2013). Introduced species in New Zealand experienced negative soil feedback in conspecific and congeneric soils, but not in distant relative soils (Diez et al. 2010). These studies suggest that evolutionary processes, like strong developmental constraints or the absence of genetic variation, could interact with ecological processes to determine the nature of plant-soil feedbacks. Phylogenetic signal in plant-soil feedback would, however, need to be quite strong in order to be detected in the face of context-dependent variation in the interaction. Communities
1
3
Department of Integrative Biology, University of Guelph, Ontario, Canada
95616, USA
4
Fort Keogh Livestock and Range Research Laboratory, USDA-Agricultural
2
Research Service, Miles City, MT, 59301, USA
INTRODUCTION
Department of Evolution and Ecology, University of California, Davis, CA, Department of Biology, University of British Columbia, Okanagan campus,
Kelowna, British Columbia, Canada
*Correspondence: E-mail: [email protected]
© 2014 John Wiley & Sons Ltd/CNRS
1614 B. L. Anacker et al.
of soil biota are often highly labile in time and space (Kardol et al. 2006; Reinhart & Clay 2009). Plant–microbe interactions can shift from mutualistic to neutral or parasitic with increasing nutrient conditions (e.g. Hoeksema et al. 2010) and can vary with interspecific plant competition and plant density (e.g. Casper & Castelli 2007). Exotic plant species, in particular, may add additional variation and limit the potential for detecting phylogenetic signal, as many exotic plant species are relatively unresponsive to soil mutualists and soil enemies (Klironomos 2002; Kourtev et al. 2002; Reinhart & Callaway 2006; Seifert et al. 2009; Verhoeven et al. 2009; Bennett & Strauss 2013). Klironomos (2002) suggested that native species exhibited strong negative feedback while dominant exotic species accumulated pathogens at slower rates, reflecting an enemy release phenomenon (Engelkes et al. 2008). Here, we synthesise ecological and evolutionary information to gain greater insights into which components of plant-soil feedbacks predict the relative abundance of plants within communities. We expand on a 2002 study by Klironomos that showed that plant abundance in an old field was correlated with plant-soil feedback (i.e. plant rarity was linked to negative soil feedbacks), as measured by how well a species grew in soils with a history of conspecific vs. heterospecific plants. We incorporate data from two additional experiments on soil feedbacks with mycorrhizal fungi, one of which is newly reported here, that each use the same set of 57 plant species. Importantly, we also test whether the mechanistic linkage between soil feedbacks and plant abundance remains with the consideration of phylogeny. Because we find that plant relative abundance is conserved, it becomes important to test if phylogeny can explain more variation in plant abundance than soil feedback, under the assumption that phylogeny acts as a reliable proxy for additional, unmeasured aspects of plant functional properties and niche space (e.g. seed size, competitive ability, dispersal ability) that might be stronger determinants of field abundance than soil feedback. Our data set represents one of the best available for examining phylogenetic conservatism in plant-soil feedback, as it includes a diversity of native and exotic plant species from a natural community (25 natives; 32 exotics), allows simultaneous examination of plant response to both whole-soil communities and mycorrhizal fractions, and links the effects of soil biota on plant growth to plant abundance in the field.
Letter
Plants were grown in soils with a history of conspecifics and in soils with a history of heterospecifics. The difference in plant growth between the conspecific and heterospecific soil treatments describes the soil feedback, which represents the net effect of positive feedback with soil mutualists and negative feedback with soil antagonists, as well as possible changes in nutrients. Thus, we refer to it as ‘net whole-soil feedback’. The experiment had three stages; two stages to culture the soils followed by a third to measure plant growth. In stage one, each of the 57 plant species were grown in a representative field soil from LTMRS for 10 weeks, after which plants were discarded. In stage two, another seedling of the same species was grown for 10 additional weeks and then discarded. In stage 3, another seedling was grown in 10 pots containing soils conditioned by conspecifics and 10 pots containing soils conditioned by heterospecifics for 10 weeks. At the end of stage 3, plants were harvested, air-dried and weighed. Using stage 3 plants, net whole-soil feedback was calculated as: ln (biomass of plant grown in soil with a history of conspecifics/ biomass of plant grown in soil with a history of heterospecifics). We note that plant growth in these soils may reflect species-specific effects of nutrient depletion, in addition to differences in soil communities (see Discussion). Experiment 2: Feedback with AMF of conspecifics
We assembled data from three experiments, two of which are previously published and one of which is described here for the first time (Experiment 2). The 57 plant species (Table S1) in our study occur at the Long-Term Mycorrhizal Research Site (LTMRS), located within the University of Guelph Arboretum, Guelph, Canada (43° 320 27″ N, 80° 120 54″ W). Each species was present in each experiment.
To isolate the unique effects of feedback with AMF communities, we assessed plant response to AMF species collected from soils that contained conspecific plants in the field vs. controls without AMF additions using the same 57 species from experiment 1. We collected a 300 mL soil core from each of 5–7 haphazardly located plants of each species, bulked the soil samples and extracted the spores using centrifugation (methods in Sikes et al. 2012). Plants were grown in 20-cm diameter, 3 L pots containing a 4 : 1 ratio of sand to Promix BX (Premier Tech, Riviere-du-Loup, QC, USA). The potting soil was sterilised prior to planting by exposing it to 32 kGy ( 10%) of gamma-irradiation over 7 days (McMaster University Nuclear Reactor). For each species, 10 pots were each seeded with ~250 AMF spores; an additional 10 pots were used as controls. During a 2-week seedling establishment period, pots were watered daily, and then watered 2–3 times per week. All plants were fertilised with 300 mL of a ¼ strength Hoagland’s solution every 2 weeks. Aboveground plant parts were harvested after 348–372 days growth, dried and weighed. The experiments were carried out over 3 years (subsets of species were grown in year-long trials that began in May 2003). Feedback with AMF of conspecifics was quantified as: ln(biomass of plant grown in soil with AMF inocula/biomass of plant grown in soil without AMF inocula). While this and the following metrics are not feedback measures in a classical sense (i.e. did not have home vs. away comparisons), we refer to them as soil feedback metrics for simplicity.
Experiment 1: Net whole-soil feedback
Experiment 3: Feedback with Glomus etunicatum
The goal of experiment 1, described in Klironomos (2002), was to describe plant-soil feedback, where plant growth influences soil communities, which in turn influence plant growth.
The third experiment, previously described by Klironomos (2003), assessed feedback with the most common mycorrhizal species at the LTMRS (Maherali & Klironomos 2012). This
MATERIALS AND METHODS
© 2014 John Wiley & Sons Ltd/CNRS
Letter
experiment allowed us to contrast feedback with a widespread, generalist AMF, Glomus etunicatum, the single most commonly observed AMF species across the study site (Maherali & Klironomos 2012), with feedback from host-specific AMF from experiment two. The same 57 plant species described above were grown either with G. etunicatum or a non-mycorrhizal control. Field soil, as described in experiment one, was mixed with silica and sand at a 1 : 1 ratio, autoclaved and placed in 15-cm diameter, 10 L pots. A band of AMF inoculum, composed of Allium porrum roots either precolonised by G. etunicatum collected from the site or nonmycorrhizal, equalling 1 g in weight, was added 2 cm below the soil surface. Each species by G. etunicatum treatment combination had 10 replicates. Plants were grown for 16 weeks and then harvested, dried and weighed. Percentage change was calculated as the percentage difference in biomass between mycorrhizal and non-mycorrhizal treatments (Klironomos 2003). We converted per cent change to a log response ratio: -ln(1per cent change/100) (Allison & Goldberg 2002). Plant abundance
Field abundance was determined by measuring the occurrence of species in 100 randomly located quadrats at LTMRS, measured in both 1998 and 2000 and pooled, as reported in Klironomos (2002). Plant abundance was the number of quadrats each species occupied and ranged from 1 to 88 ( x = 24.43). Plant abundance was log-transformed to meet the assumptions of parametric analysis. Phylogeny
We created a molecular phylogeny for the 57 plants species using maximum likelihood. Details are provided in the supporting information. For two species, sequences of congeners were used as substitutes (Table S2). In these two cases, the species were the lone representatives of their genera in our study. Statistical analysis
All analyses were performed in R version 3.0.2 (R Development Core Team, R. 2014). Effects of soil biota treatments We tested whether the soil biota treatments had effects on plant growth relative to controls (i.e. if they significantly differed from zero). We also tested if the three soil feedback metrics were correlated with one another. To account for any possible phylogenetic non-independence in the correlations among the feedback metrics, we fit phylogenetic generalised least squares (PGLS) regression models using the ‘pgls’ function in the ‘caper’ library. PGLS is described in more detail below. Phylogenetic signal in soil feedback We calculated the K statistic (Blomberg et al. 2003) for the three feedback metrics, using the ‘phylosignal’ function in the ‘picante’ library. K values < 1 indicate that species resemble
Conservatism in plant responses to soil biota 1615
each other less than expected under Brownian motion evolution; K values > 1 indicate that species resemble each other more than expected. The significance of K was assessed by comparison to a random shuffle of trait values at the tips of the phylogeny. We also tested for phylogenetic signal in plant abundance. The abundance of plant taxa may be phylogenetically conserved because plants with similar abundances in any given community may share traits that are linked to abundance and also phylogenetically conserved. For example, ruderal plant species are likely to dominate initially during secondary succession due to shared traits like fast growth rates, small seed size, dispersal ability, etc. Seed size, for example, is phylogenetically conserved (Moles et al. 2005). Thus, a plant’s relative abundance may be an expression of many traits that are conserved and reflect aspects of its biology other than soil feedbacks. Relationship of the three soil feedback metrics to plant abundance in the field We fit all possible models of plant abundance using the three soil feedback metrics. Model fit was compared using AICc and model weights (Burnham & Anderson 2004). We preferred to use PGLS for our regression models, rather than phylogenetic independent contrasts (PICs). It has recently been pointed out that phylogenetic regression is only appropriate when there is phylogenetic signal in model residuals, reflecting the regression assumption of independent errors (Revell 2010). In cases where there is no signal in the residuals, despite relatively strong signal in both the dependent and independent variables, using a model based on PICs can actually lead to inflated errors. The PGLS method we used avoids this problem by estimating phylogenetic signal and regression parameters simultaneously, allowing for a PGLS model to equal an ordinary least squares model when phylogenetic signal in the model residuals are absent (k = 0); likewise, a PGLS model can equal a model fit with PICs when phylogenetic signal in the model residuals are strong (k = 1). A drawback to our model fitting technique is that we cannot statistically compare the models using conventional likelihood ratio tests, as they are fit using different branch length transformations internal to the PGLS function. To test the statistical significance for two model comparisons in particular, however, we refit the constituent models with a single branch length transformation, which was set to be equal to the average k value obtained for the constituent models, and compared the fit of the models using likelihood ratio tests. Distinguishing the effects of phylogeny and soil feedback on plant abundance in the field We used variance partitioning to parse the variation of plant abundance among the soil feedback metrics and phylogenetic components (Desdevises et al. 2003; Peres-Neto et al. 2006). The method required that we create three standard multiple regression models. Our first model (Soil) was a regression of plant abundance on the three soil feedback measures. Our second model (Phylo) was a regression of plant relative abundance on a set of principal coordinates (PCs) that represent the phylogeny. For this, we decomposed the phylogeny into a © 2014 John Wiley & Sons Ltd/CNRS
1616 B. L. Anacker et al.
set of n1 orthogonal PCs, where n equals the number of species, using the ‘PVRdecomp’ function of the ‘PVR’ package. To select the PCs to be included in the multiple regression model, we regressed each PC against plant abundance, retaining those that were significantly related at alpha 0.05. We then fit a multiple regression model using only the retained PCs. We retained two PCs (PC2 and PC4). All retained PCs represented deep divergences in the phylogeny, and thus trait values that show correlations with such PCs could have evolved along the tree and reflect niche conservatism. Criticisms of this approach come from PC axes that contrast paraphyletic groups, and for which it is therefore harder to interpret how a trait might evolve in such a manner (Freckleton et al. 2011). In our case, we only used PC axes that contrasted whole clades. Our third model (SoilPhylo) was a multiple regression that included the terms from the Soil and Phylo models. We derived the four constituent components via subtraction of the R2 values from the three multiple regression models described above as follows (Desdevises et al. 2003): Soil feedback only R2 = SoilPhylo R2Phylo R 2 Phylogeny only R2 = SoilPhylo R2Soil R2 Shared = Soil R2 + Phylo R2SoilPhylo R2 Residuals = 1SoilPhylo R2 The soil biota only R2 represents the relationship of plant abundance and soil feedback after taking into account phylogenetic autocorrelation. Significance testing is possible for the soil feedback only component and the phylogeny only component using redundancy analysis, but not for the shared fraction or the residuals (Peres-Neto et al. 2006). Soil feedbacks in natives and exotics We tested if the three soil feedback metrics differed by plant native status (native/exotic) using PGLS. We repeated the phylogenetic signal K tests using natives only (n = 25) and exotics only (n = 32) and tested for phylogenetic signal in native status itself, using the ‘phylo.d’ function in the ‘caper’ library. We also added plant native status (exotic or native) as explanatory variables to our models of plant abundance. Finally, we repeated the variance partitioning exercise for natives and exotics separately.
RESULTS
Effects of soil biota treatments
We found large differences in the direction of our three feedback measures (Fig. 1). Plant growth was lower in the presence of host-specific soil biota than it was in the presence of soil biota of heterospecific plants (i.e. negative net whole-soil feedback; x = 0.14; P < 0.001). In contrast, plants grew much larger in the presence of conspecific mycorrhizal communities vs. controls (i.e. positive feedback with AMF of conspecifics; x = 0.25, P < 0.001), but not in the presence of the dominant G. etunicatum vs. controls (i.e. neutral feedback with G. etunicatum; x = 0.06, P = 0.10). The three different measures of soil feedback were unrelated to one another (net whole-soil feedback vs. feedback with G. etunicatum, P = 0.56; net whole-soil feedback vs. feedback with AMF of © 2014 John Wiley & Sons Ltd/CNRS
Letter
conspecifics, P = 0.80; feedback with G. etunicatum vs. feedback with AMF of conspecifics, P = 0.79; tests accounted for phylogeny via PGLS). Results were qualitatively similar to results based on simple correlation analysis (P = 0.52, 0.22, 0.63, respectively). Phylogenetic signal in soil feedback
We found significant phylogenetic signal in net whole-soil feedback, even stronger signal in feedback with AMF of conspecifics, and no signal in feedback with G. etunicatum (Table 1; Fig. 1). Plant abundance was weakly conserved (K = 0.19; P = 0.06), which could reflect both shared response to soil biota by relatives, or other traits influencing abundance that are phylogenetically conserved. Some clades contributed greatly to the overall phylogenetic signal. Positive net wholesoil feedback and high plant abundances characterised the predominately native, later-flowering Asteraceae tribe Astereae (Fig. 1) and negative net whole-soil feedback characterised the monocots. On the other hand, feedback with AMF of conspecifics was strong and positive for monocots ( x = 0.68) but not dicots ( x = 0.13). Relationship of three soil feedback metrics to plant abundance in the field
Net whole-soil feedback explained the largest proportion of the variance in plant abundance (51%; model 5 in Table 2) after accounting for phylogenetic non-independence. The multivariate models identified complementarity among the three metrics of soil feedback (Table 2). Specifically, two models were equally good at explaining plant abundance: one incorporated net whole-soil feedback and feedback with G. etunicatum; our likelihood ratio test confirmed that the addition of feedback with G. etunicatum significantly improved the model over the originally published model with only net whole-soil feedback (P = 0.04). A model that included all three metrics, including feedbacks from AMF of conspecifics, was not significantly different from, i.e. equally good as, the best model (P = 0.14; Table 2, model 2). While the very best model of abundance does not include feedback with AMF of conspecifics, we argue that it may be a mistake to dismiss the importance of AMF communities on patterns of field abundance, as the feedback with AMF of conspecifics was very strong, contained the strongest phylogenetic signal and was unique from (i.e. uncorrelated with) the other two feedback metrics (Fig. 1). Distinguishing the effects of phylogeny and soil feedback on plant abundance in the field
The strong relationship between field abundance and soil feedback (adj. R2 = 54.4 for model with the three feedback metrics as explanatory variables) was partitioned into a soil feedback component (adj. R2 = 38.1), a phylogeny component (adj. R2 = 0.0) and a soil feedback + phylogeny component (adj. R2 = 16.3), the latter representing the correlation between the PCs and soil feedback (Fig. 2). The soil feedback component was significant, indicating that the relationship
Letter
Conservatism in plant responses to soil biota 1617
K = 0.27 (< 0.01)
K = 0.42 (< 0.001)
K = 0.19 (0.06)
K = 0.11 (0.62)
Negative feedback Positive feedback n = native n n n n n n n n n n
n n n n n
n n n n n
n n n n n n
–0.4 –0.2 0 0.2 Net whole-soil feedback
1.5 –0.4 0 0.4 0.8 –0.5 0 0.5 1 Feedback with AMF Feedback with Glomus of conspecifics etunicatum
0
1 2 3 4 ln(Plant abundance)
Figure 1 Phylogenetic conservatism in three plant-soil feedback metrics and plant abundance for 25 native and 32 exotic plant species (delineated by ‘n’ on phylogeny tips). Black bars indicate negative plant-soil feedback; hashed bars indicated positive soil feedback.
Table 1 Phylogenetic conservatism in species attributes according to Blomberg’s K, where 0 = random and 1 = Brownian
All species Variable Plant abundance Net whole-soil feedback Feedback with AMF of conspecifics Feedback with G. etunicatum
Natives
Exotics
K
P
K
P
K
P
0.19 0.27 0.42
0.06
