Is there an association between root architecture and mycorrhizal growth response?

Authors


Summary

  • The symbiosis between arbuscular mycorrhizal (AM) fungi and plants is evolutionarily widespread. The response of plant growth to inoculation by these fungi (mycorrhizal growth response; MGR) is highly variable, ranging from positive to negative. Some of this variation is hypothesized to be associated with root structure and function. Specifically, species with a coarse root architecture, and thus a limited intrinsic capacity to absorb soil nutrients, are expected to derive the greatest growth benefit from inoculation with AM fungi.
  • To test this hypothesis, previously published literature and phylogenetic information were combined in a meta-analysis to examine the magnitude and direction of relationships among several root architectural traits and MGR.
  • Published studies differed in the magnitude and direction of relationships between root architecture and MGR. However, when combined, the overall relationship between MGR and allocation to roots, root diameter, root hair length and root hair density did not differ significantly from zero.
  • These findings indicate that possessing coarse roots is not necessarily a predictor of plant growth response to AM fungal colonization. Root architecture is therefore unlikely to limit the evolution of variation in MGR.

Introduction

The symbiosis between plants and arbuscular mycorrhizal (AM) fungi is one of the most widespread trophic interactions in nature (Brundrett, 1991; Kiers & van der Heijden, 2006). AM fungi are an ancient lineage of obligate biotrophs in the phylum Glomeromycota that form associations with plants in order to obtain energy for growth and reproduction. In return, AM fungi can provide plants with better access to soil nutrients and improve resistance to abiotic and biotic stressors (Newsham et al., 1995; Smith & Read, 2008). Although the symbiosis is facultative for many terrestrial plants, fossil evidence indicates that the relationship is > 400 million yr old and nearly 75% of plant species are estimated to form associations with AM fungi (Pirozynski & Malloch, 1975; Remy et al., 1994; Wang & Qiu, 2006; Brundrett, 2009). Despite the ubiquity of the AM symbiosis, plant growth responses to AM fungi, defined as the biomass difference between a mycorrhizal plant and a non-inoculated plant of the same species (hereafter referred to as the mycorrhizal growth response (MGR); Janos, 2007), vary widely along a continuum from positive to negative (Johnson et al., 1997; Klironomos, 2003).

The structure and function of the root system are expected to influence how plants respond to colonization by AM fungi (Smith & Gianinazzi-Pearson, 1988; Newsham et al., 1995; Smith & Read, 2008). A frequently cited hypothesis is that species with coarse root architecture, characterized by relatively large diameter roots, low root hair density and short root hairs, derive the greatest growth benefit from colonization by AM fungi (Baylis, 1970, 1975; Smith & Gianinazzi-Pearson, 1988; Hetrick, 1991; Newsham et al., 1995; Brundrett, 2002; Fitter, 2004; Smith & Read, 2008). This is because plants with a coarse root architecture have limited intrinsic ability to absorb nutrients (Bates & Lynch, 2001), and are therefore predicted to benefit from the presence of finely structured AM fungal hyphae, which increase the surface area available for the absorption of nutrients, particularly phosphorus (Raven & Edwards, 2001). As originally formulated (Baylis, 1970, 1975; Fitter, 2004), this hypothesis proposed that a fine root architecture (specifically greater root hair length and density) is an alternative to mycotrophy in phosphorus-limited soils. Therefore, root fineness is expected to be negatively correlated with mycorrhizal dependence, defined as ‘the inability of plants without mycorrhizas to grow unless provided with supplemental phosphorus’ (Janos, 2007, p. 88).

Although the original Baylis hypothesis was formulated for mycorrhizal dependence, it has been tacitly extended to include MGR, because the latter property is a general approximation of the degree to which plants can respond to inoculation by AM fungi (Menge et al., 1978; Hetrick, 1991; Klironomos, 2003; Smith & Read, 2008; Hoeksema et al., 2010). The extension of the Baylis hypothesis to MGR was also facilitated by the interchangeable use of mycorrhizal dependence and MGR in the literature, even though they are functionally distinct concepts (Janos, 2007). In addition, MGR is much more frequently reported than mycorrhizal dependence (Hoeksema et al., 2010; Veresoglou et al., 2012; Treseder, 2013), probably because its estimation requires a simpler experimental design (Fitter, 2004; Janos, 2007).

Regardless of its origin, the frequency with which relationships between root traits and MGR are evaluated in the literature (Menge et al., 1978; Graham & Syvertsen, 1985; Hetrick et al., 1988; Hetrick, 1991; Smith & Read, 2008; Smith & Smith, 2011) has established the expectation that coarse roots should be positively correlated with plant growth response to mycorrhizal inoculation (e.g. table 4.6 in Smith & Read, 2008). Nonetheless, there is no consensus on the direction and magnitude of the relationship between these two characteristics. Several comparative studies with wild and agricultural species have reported associations between coarser roots, shorter root hairs and lower root hair density and increased MGR (Hetrick et al., 1988; Manjunath & Habte, 1991; Baon et al., 1994; Declerck et al., 1995; Schweiger et al., 1995; Jakobsen et al., 2005). By contrast, other comparative studies have indicated that root fineness, root hair density and root hair length are either not associated or positively associated with MGR (Graham & Syvertsen, 1985; Peterson & Farquhar, 1996; Duponnois et al., 2001; Siqueira & Saggin-Júnior, 2001; Zangaro et al., 2005, 2007).

To determine the magnitude and direction of the overall relationship between root traits and MGR, a meta-analysis of previous studies was used to evaluate the correlation between root traits and MGR. Based on the tacit extension of the Baylis hypothesis to MGR (Menge et al., 1978; Hetrick, 1991), species for which non-inoculated individuals have lower allocation to roots and coarser root systems were expected to have higher MGR. Specifically, decreased root : shoot ratio, low specific root length (SRL, or the ratio of root length to root mass), thicker root diameter, shorter root hairs and lower root hair frequency were predicted to associate with increased MGR across seed plants. Because of shared ancestry, species are not statistically independent, and this lack of independence could either produce spurious correlations or obscure legitimate correlations (Felsenstein, 1985). To account for this effect, interspecific correlations between root traits and MGR in previously published studies were re-calculated by formally taking into account the phylogenetic relationships among species in the sample. These phylogenetically corrected correlations and their associated error variances were used in the meta-analysis.

Materials and Methods

Data collection

To examine the relationships between root traits and the growth response of plants to AM fungi, Web of Science and Google Scholar were searched using arbuscular, mycorrhizal, mycorrhiza, mycorrhizae, fungal, fungi, AM fungal, AM fungi, AMF and each root trait (specific root length, fine root diameter, root hair length and root hair density) as terms. The results of these searches were screened to determine whether they contained information on quantitative measurements of both a root trait under non-mycorrhizal conditions, an assessment of MGR via direct inoculation with AM fungi, and whether the number of species in the sample was ≥ 4. This level of replication within a study was required in order to calculate the correlation coefficient and its associated error variance for meta-analysis.

From each study, data on plant biomass were obtained for both the AM fungal treatment and the non-inoculated treatment or the ratio of biomass from the two treatments. In addition, information on at least one of SRL, root diameter, root hair length and root hair density was recorded from plants in the non-inoculated treatment. In cases in which SRL was not reported directly, it was calculated as the ratio of root length and root mass if these values were reported. High SRL values typically indicate thinner roots with greater total surface area for nutrient absorption, whereas low SRL values indicate thicker roots with lower total surface area (Craine et al., 2001). Root : shoot ratios were recorded when reported, or calculated from biomass data. Because not all studies reported all root traits, sample sizes for each relationship differed (Table 1).

Table 1. A summary of the growth form, biome and sample sizes of 12 studies that report relationships between at least one of five root architectural traits and the mycorrhizal growth response (MGR)
StudyAbbreviationExtractable soil PGrowth formBiome n λ
Root : shoot ratioSRLRoot diameterRoot hair lengthRoot hair density
  1. For each relationship, the maximum likelihood estimate of λ, or the degree to which residuals were associated with the phylogenetic variance–covariance matrix, are shown. When λ = 0, there is no relationship between variation in the regression residuals and phylogeny and, when λ = 1, there is complete dependence between residual variation and a Brownian model of evolution along the phylogeny. Study abbreviations match those on the x axes of Figs 1 and 2. Soil extractable phosphorus (P), method of P extraction, growth form and biome information are also listed.

  2. a

    Included data from a companion paper: Zangaro et al. (2003).

  3. b

    = 77 for root diameter and root hair density; = 72 for root hair length.

Declerck et al. (1995)DEC7 μg g−1 (Olsen)HerbTropical70.50.50.5
Duponnois et al. (2001)DUP4.8 ppm (Olsen)Herb, treeTropical170.860.92
Graham & Syvertsen (1985)GRA3 mg kg−1 (Mehlich I)TreeSubtropical5110
Hetrick et al. (1988)HET10 ppm (Bray I)Herb, grassTemperate130
Hill et al. (2010)HILFertilizedHerb, grassTemperate410
Manjunath & Habte (1991)MAN0.02 mg l−1 (method unknown)TreeTropical800000
Osonubi et al. (1991)OSO4.82 mg kg−1 (Bray I)TreeTropical410
Pope et al. (1983)POP20 ppm (Bray I)TreeTemperate401
Schweiger et al. (1995)SCHFertilizedHerb, grassTemperate50.9100
Siqueira & Saggin-Júnior (2001)SIQFertilizedTreeTropical290.360.560.44
Stanescu (2012)STA2 mg kg−1 (Olsen)Herb, grassTemperate22111
Zangaro et al. (2005)aZAN1.8 mg kg−1 (Mehlich I)TreeTropical78b00.3400

Quantifying MGR

To quantify MGR, the effect size of the AM fungal treatment on dry plant biomass (either shoot biomass or total biomass, depending on the study) was calculated as the log response ratio, MGR = loge[Xi/Xn], where Xi is the inoculated biomass and Xn is the non-inoculated biomass (Hoeksema et al., 2010). Positive MGR values indicate that plant biomass increased in response to inoculation, whereas negative values indicate that plant biomass decreased in response to inoculation. This metric is preferred over previous formulations of MGR (reviewed in Janos, 2007) because values are not bounded by zero, and its statistical properties are appropriate for meta-analyses (Hedges et al., 1999). In situations in which plants were grown with multiple AM fungal inocula treatments, the response ratio was calculated by averaging Xi across treatments, as described by van der Heijden et al. (1998). In situations in which other calculations of MGR were reported, they were converted to log response ratios using the formulae described by Allison & Goldberg (2002). In situations in which plants were grown across a gradient of soil phosphorus, maximum MGR values were recorded, and plant traits were recorded from the corresponding non-mycorrhizal treatment.

Meta-analysis of relationships between root traits and MGR

A random effects meta-analysis (Borenstein et al., 2009) was used to examine the overall magnitude and direction of associations between the root architectural traits of non-inoculated plants and MGR. A random effects analysis was used because the overall effect size is calculated assuming that variation among studies can occur because of differences in experimental conditions. This is applicable to the present meta-analysis because studies differed in terms of plant species identity, growth conditions and sources of mycorrhizal inoculum. By assuming that these factors vary among studies, a random effects meta-analysis is generalizable to the overall hypothesis (Borenstein et al., 2009).

Before conducting the meta-analysis, the within-study Pearson correlations between root architectural traits and MGR, and their associated variances, were calculated. The correlation coefficient was used because it represents a standardized effect size, which allows the effect sizes among studies to be expressed on the same scale. To account for the effect of shared ancestry in the estimate of correlation coefficients from the original studies, correlation coefficients within each study were calculated from a phylogenetic generalized least-squares (PGLS) regression, where the phylogenetic variance–covariance matrix is incorporated directly into the calculation of effect size (Martins & Hansen, 1997; Pagel, 1999). To facilitate PGLS regression, a phylogenetic tree for all taxa in the meta-analysis was derived from the Angiosperm Phylogeny Group seed plant phylogeny reference tree (APG III; Stevens, 2001) which is contained within Phylomatic (http://phylodiversity.net/phylomatic/; Webb et al., 2008). The resulting tree (Supporting Information Fig. S1) was assigned branch lengths using node ages from Wikstrom et al. (2001) and the ‘bladj’ function in Phylomatic.

To estimate the correlation coefficient in each study, the phylogenetic tree was pruned to create unique trees for each study, depending on the identity of the species present. To quantify the degree of dependence on phylogeny for each study, a scaling factor λ was estimated using maximum likelihood. λ was unique to each study because it depends on the identity of the species sampled, and ranges from zero, indicating complete independence between variation in the regression residuals and phylogeny, to unity, indicating complete dependence between residual variation and a Brownian model of evolution along the phylogeny (Freckleton et al., 2002). When λ = 0, the PGLS regression is identical to ordinary least-squares (OLS) regression. Estimates of λ were performed in R version 3.02 (R Core Team, 2013) and the ‘pgls’ command in the package caper, version 0.5.2 (Orme et al., 2013).

To carry out the random effects meta-analysis for each root architectural trait, the correlations between traits and MGR, and their variances, were combined using the equations and methodology described in Borenstein et al. (2009). Effect sizes from individual studies were weighted by the inverse of their variances before being combined in the meta-analysis. To determine whether the inclusion of a particular study influenced the overall effect size, a sensitivity analysis was carried out. This entailed the iterative removal of one study and the re-calculation of the overall effect size of the meta-analysis with the remaining studies. A two-tailed z-test was used to determine whether the overall effect size for each meta-analysis differed from zero (Borenstein et al., 2009).

Results

The dataset for analysis was compiled from 12 studies (Table 1) comprising 196 experimental trials (Table S1). Studies were performed on plants ranging from 28 to 262 d old and growing in pots in glasshouses or other controlled environments. The experimental trials contained representatives of 40 seed plant families. A plurality of observations came from the Fabaceae, which represented 73 of the 196 trials. Phylogeny was correlated with regression residuals in a majority of studies (Table 1). Of the 33 reported correlations between root architectural traits and MGR, λ was > 0 in 18 cases. Only one study (Manjunath & Habte, 1991) lacked phylogenetic influence on the calculation of correlations.

Root allocation and aspects of root fineness were not associated with MGR in the meta-analysis. The root : shoot ratio was positively correlated with MGR in the study by Zangaro et al. (2005), but no other study contained a statistically significant relationship (Fig. 1a) and the overall effect size did not differ significantly from zero (= 0.26, = 0.32). The overall effect size was insensitive to the iterative removal of studies with one exception. When the study by Stanescu (2012) was removed, the root : shoot ratio was positively correlated with MGR (Fig. 1d; = 0.44, = 0.047). SRL was negatively correlated with MGR in three of five studies (Pope et al., 1983; Graham & Syvertsen, 1985; Manjunath & Habte, 1991), but the overall effect size did not differ from zero (Fig. 1b; = −0.35, = 0.22). The overall effect size for the relationship between SRL and MGR remained non-significant in the sensitivity analysis (Fig. 1e). Root diameter was positively correlated with MGR in the studies by Hetrick et al. (1988) and Manjunath & Habte (1991), but negatively correlated with MGR in the studies by Schweiger et al. (1995) and Zangaro et al. (2005), but the overall effect size did not differ from zero (Fig. 1c, r = 0.048, = 0.81). Similarly, the relationship between root diameter and MGR remained non-significant in the sensitivity analysis (Fig. 1f).

Figure 1.

Pearson correlation coefficients for each study (closed circles), the overall effect size (open triangles) in the meta-analysis (a–c) and the sensitivity of the overall effect size to the iterative removal of single studies (d–f). Relationships are between root : shoot ratio (a, d), specific root length (SRL) (b, e) and root diameter (c, f) of non-inoculated plants and mycorrhizal growth response (MGR). Study abbreviation codes correspond to those listed in Table 1. Vertical lines around the symbols represent the 95% confidence intervals. Correlations that differed significantly from zero are indicated: *, < 0.05; **, < 0.01; ***, < 0.001.

Root hair dimensions and frequency were also not associated with MGR in the meta-analysis. Root hair length was negatively correlated with MGR in the studies by Declerck et al. (1995) and Manjunath & Habte (1991), but positively correlated with MGR in the study by Siqueira & Saggin-Júnior (2001) (Fig. 2a). Nevertheless, the overall effect size did not differ from zero (= 0.029, = 0.90) and this outcome was consistently observed in the sensitivity analysis (Fig. 2c). Root hair density was negatively correlated with MGR in the studies by Declerck et al. (1995) and Manjunath & Habte (1991), but positively correlated with MGR in the studies by Hill et al. (2010) and Zangaro et al. (2005) (Fig. 2b). However, the overall effect size did not differ from zero (= 0.057, = 0.81), an outcome that was also consistently observed in the sensitivity analysis (Fig. 2d).

Figure 2.

Pearson correlation coefficients for each study (closed circles), the overall effect size (open triangles) in the meta-analysis (a, b) and the sensitivity of the overall effect size to the iterative removal of single studies (c, d). Relationships are between root hair length (a, c) and root hair density (b, d) of non-inoculated plants and mycorrhizal growth response (MGR). Study abbreviations correspond to those listed in Table 1. Vertical lines around the symbols represent the 95% confidence intervals. Correlations that differed significantly from zero are indicated: *, < 0.05; **, < 0.01; ***, < 0.001.

Discussion

Meta-analyses of the relationship between root traits and MGR did not support the prediction that coarse root architecture is associated with greater plant growth benefit from AM fungi (Menge et al., 1978; Hetrick et al., 1988; Hetrick, 1991; Manjunath & Habte, 1991; Brundrett, 2002; Smith & Read, 2008). The overall effect sizes of the relationships between MGR and root : shoot ratio, SRL, root diameter, root hair length and root hair density did not differ from zero (Figs 1a–c, 2a,b). The outcome of the meta-analyses was not generally sensitive to the inclusion of a particular study (Figs 1d–f, 2c,d). Specifically, the outcome was not influenced by the exclusion of studies with large sample sizes, and thus high leverage (Zangaro et al., 2005; Table 1). The only exception to this pattern occurred for the relationship between the root : shoot ratio and MGR when the study by Stanescu (2012) was excluded. However, the overall effect size in this case was the opposite of expectations – increased root : shoot ratio in non-mycorrhizal plants was positively associated with MGR. These results suggest that there is no broad support for the hypothesis that species with a coarse root architecture are more likely to derive the greatest growth benefit from colonization by AM fungi.

Although there were no overall associations between root architectural traits and MGR in the meta-analysis, some of the contributing studies had significant effect sizes. The pattern of variation among contributing studies, or lack thereof, suggested three ways in which a non-significant overall effect size occurred. First, for the root : shoot ratio (Fig. 1a), there were generally weak relationships in the contributing studies, which resulted in a non-significant overall effect size. The single study for which there was a statistically significant relationship (Zangaro et al., 2005) did not have sufficient weight to counteract the influence of the majority of studies which showed no significant relationship. Second, for root diameter, root hair length and root hair density (Figs. 1c, 2a,b), there were approximately equal proportions of significant positive and negative relationships among the contributing studies, and these opposite effects averaged out to produce a non-significant overall effect. The tendency for equal proportions of studies to fall on opposite sides of the weighted mean effect is not unusual in meta-analyses, and suggests that sampling error could have been responsible for differences between studies (Palmer, 2000). Third, for SRL, significant correlations were consistently negative, yet did not produce an overall negative effect size (Fig. 1b). Although the overall weighted correlation was the strongest among the root traits (= −0.35), this meta-analysis had the smallest sample size (Table 1), which led to high analysis-wide variance. Thus, it is likely that the SRL meta-analysis lacked the necessary power to reach conclusions about the nature of the relationship between SRL and MGR.

Variation in the magnitude and direction of effect sizes in contributing studies could potentially be associated with covariates, such as growth form, biome/habitat and soil phosphorus content (Table 1). Although there were an insufficient number of studies within each meta-analysis to statistically test for the influence of covariates, a comparison of the covariates among studies suggests that it is unlikely that many of these factors were responsible for effect size variation. This is because similar effect sizes were often found for studies that differed in growth form and habitat. For example, there were significant positive relationships between root diameter and MGR in Hetrick et al. (1988) and Manjunath & Habte (1991), but the former study was performed with temperate herbs and grasses, whereas the latter was performed with tropical trees (Fig. 1c). Similarly, there were significant negative relationships between root diameter and MGR in Schweiger et al. (1995) and Zangaro et al. (2005), even though the former was performed with temperate herbs and grasses and the latter with tropical trees. These types of patterns were also observed for root hair length and root hair density (Fig. 2a,b). Differences in phosphorus availability could have influenced both the morphology of root systems of non-inoculated plants (Raven & Edwards, 2001) and the magnitude of MGR (Hoeksema et al., 2010), and potentially contributed to variation in the magnitude and direction of correlations. Although extractable phosphorus in soil was reported in a majority of studies, differences in the analysis method among studies meant that values could not be compared on an absolute scale (Wolf & Baker, 1985). Therefore, it is not known whether and by how much phosphorus availability in soil influenced inter-study variability in relationships between root architecture and MGR.

Although a relationship between root architecture and MGR was not found in the present meta-analysis, it should be noted that there are other biologically meaningful aspects of root system architecture and plant growth response to AM fungi that were not captured in the analysis. First, the present study used frequently reported root architectural traits, but other less studied aspects of the root system, such as branching frequency (Berta et al., 1995), could be associated with MGR. Second, it is possible that trade-offs between root traits influenced the outcome. For example, plants may enhance nutrient absorption by increasing SRL, root hair length or root hair density, but possibly not all three simultaneously. If alternative strategies for producing fine root architecture are employed by different species (Siqueira & Saggin-Júnior, 2001), the detection of a correlation between root architecture and MGR could be highly dependent on the sample of species used. Similarly, a relationship between root architecture and MGR could be detected if different root architectural metrics could be combined to reflect whole root system function. Finally, the growth response of plants to AM fungi depends on the identity of both the plant and fungal partner (Klironomos, 2003). Therefore, it is possible that a relationship between root architecture and MGR could exist if studies had matched plant species with the AM fungal partners that most positively influenced plant growth.

One explanation for the independence between root architectural traits and MGR is the ability of plants to respond plastically to fungal colonization. For example, several studies have suggested that inoculation with AM fungi causes plants to produce coarser root systems, which could be a coordinated response that facilitates a shift from root-driven to mycorrhizae-driven nutrient uptake (Hetrick, 1991; Berta et al., 1995; Hooker & Atkinson, 1996; Zangaro et al., 2007). This possibility could not be tested in the present meta-analysis because only a small number of studies reported the root architecture of both non-inoculated and inoculated plants (Osonubi et al., 1991; Zangaro et al., 2005). Although root plasticity to AM fungal colonization is a potential explanation for a lack of relationship between root architecture and MGR, plants do not always produce coarser roots following inoculation by AM fungi (Schroeder & Janos, 2005; Zangaro et al., 2005; Wu et al., 2010, 2011). Variability in the outcome of studies that evaluate the plasticity of root architecture to fungal colonization may arise from a lack of control for the effects of allometric scaling between root architecture and plant size. For example, a common plastic response to AM fungal inoculation is reduced root : shoot ratio (Veresoglou et al., 2012). However, root : shoot ratio is also negatively correlated with size and development (McConnaughay & Coleman, 1999). Thus, mycorrhizal plants may have a low root : shoot ratio because they are larger, and not because of a direct effect of AM fungal inoculation (Veresoglou et al., 2012). The potential for size dependence as a complicating factor in the determination of whether fungal inoculation produces coarse roots has not been explored experimentally (Veresoglou et al., 2012). Therefore, future studies of the effects of AM fungal inoculation on root architecture should be performed by comparing inoculated and non-inoculated plants in an ontologically controlled manner (McConnaughay & Coleman, 1999). Such data are necessary to test whether root plasticity to AM fungi is a common feature of plants and influences the outcome of relationships between root traits and MGR.

Even though expectations for a relationship between root architecture and MGR were derived from theories of mycorrhizal dependence, the lack of a relationship does not preclude the existence of a correlation between root architecture and mycorrhizal dependence (Baylis, 1970, 1975; Fitter, 2004). This is because MGR and mycorrhizal dependence are not quantitatively equivalent (Janos, 2007). For example, species that are dependent on AM fungi will necessarily have a positive MGR, whereas species that are not dependent on AM fungi could have MGRs anywhere along a continuum from negative to positive (Sawers et al., 2008). Thus, if there is a relationship between root architecture and mycorrhizal dependence, it was unlikely to be detected in the present analysis because species with different levels of mycorrhizal dependence could theoretically have the same MGR. As has been stated elsewhere (Fitter, 2004; Janos, 2007), mycorrhizal dependence is less frequently quantified than MGR. Moreover, the lack of overlap in these metrics in published studies not only prevents broad tests of the hypothesis between root architecture and mycorrhizal dependence, but also makes it impossible to partition out dependence and non-dependence components from MGR (Sawers et al., 2008). Thus, one recommendation from the present study is to re-iterate the call for researchers to quantify both MGR and mycorrhizal dependence in the same study using standard methods (Janos, 2007; Sawers et al., 2008).

In conclusion, the findings of this meta-analysis indicate that there is no broad support for the expectation that species with a coarse root architecture have higher MGR. Therefore, the tacit extension of the Baylis hypothesis to MGR is not warranted. More generally, these findings also suggest that root architecture need not limit the evolution of interspecific variability in plant growth responses to AM fungal inoculation (Hoeksema et al., 2010). AM fungi are now known to influence many aspects of plant growth, structure and function beyond the acquisition of phosphorus, including the uptake of other elements and water, protection from pathogens and herbivores, and resistance to toxic substances and environmental stressors (Newsham et al., 1995; Powell et al., 2009; Sikes et al., 2010; Chagnon et al., 2013). Furthermore, plant growth responses to AM fungi can be influenced by cost–benefit ratios of resource exchange between the partners, soil resource availability, climatic factors and competition with other plants and fungi (Smith & Gianinazzi-Pearson, 1988; Klironomos, 2003; Maherali & Klironomos, 2007; Bever et al., 2009; Hoeksema et al., 2010; Johnson, 2010; Kiers et al., 2011; Reinhart et al., 2012). Thus, the magnitude and direction of plant growth responses to AM fungal inoculation could be influenced by many different ecological factors, all of which could act as agents of selection on the symbiosis. Because of this, the causes of variation in the mycorrhizal symbiosis should be tested experimentally by quantifying the strength and direction of natural selection on the symbioses, whilst, at the same time, manipulating putative ecological agents of selection. Such studies have been carried out for other types of biotic interactions (Weis & Gorman, 1990; Juenger & Bergelson, 2000; Heath & Tiffin, 2007) and have great potential to unravel the causes of variation in the interactions between plants and AM fungi (Hoeksema, 2010).

Acknowledgements

This work was supported by a Discovery grant from the Natural Sciences and Engineering Research Council (NSERC), the University of Guelph and sabbatical support from the National Evolutionary Synthesis Center (NESCent), NSF #EF-0423641. I thank C. M. Caruso, A. H. Fitter and anonymous reviewers for comments that improved the manuscript. I also thank the scholarly community and administrative staff at NESCent for providing a superb environment for synthesis research.

Ancillary

Advertisement