Context-dependent plant responses to arbuscular mycorrhiza mainly reflect biotic experimental settings

Abstract

Mutualistic associations that plant roots form with soil-borne fungi of the phylum Glomeromycota, termed arbuscular mycorrhizas (AM; in the text we often refer to them as mycorrhizas), are among the most ubiquitous symbioses across terrestrial ecosystems (Brundrett & Tedersoo, 2018). The benefits that plants gain from the symbiosis not only depend strongly on environmental parameters, such as light intensity and soil fertility, but also reflect how compatible the Glomeromycotan fungus is with the plant species (the mycorrhizal phenotype: Johnson et al., 1997; Hoeksema et al., 2010). There have been, so far, numerous controlled experiments addressing biomass gains following manipulations of the mycorrhizal status of the plant hosts, and these have been nicely summarized in recent meta-analyses (e.g. Hoeksema et al., 2010; Qiu et al., 2022). Results of meta-analyses, however, are meant to be generalizable under a narrow range of ‘common’ experimental settings (Gurevitch & Hedges, 2001), which in the case of mycorrhizal studies most likely describes short-term experiments on phosphorus-deficient sandy growth substrates, plant hosts growing at low densities and a low diversity of Glomeromycotan propagules inoculated at artificially high densities.

We know much less about experimental conditions that can make mycorrhiza perform exceptionally well or badly. As an example, we know that plant mycorrhizal growth responses can get negative under very low light availability (Konvalinkova & Jansa, 2016), which describes nonetheless an unusual set of growth conditions in the mycorrhizal literature. Plant growth stimulation, by contrast, can be observed frequently across experimental settings that include plant pathogens or root-feeding nematodes (Veresoglou & Rillg, 2012). Traditional meta-analytical approaches focus on average experimental settings that do not capture well those relationships, and it can occasionally be tricky, either because the number of studies is small or because the relationships are not documented sufficiently, to develop dedicated meta-analyses. A way to gain further insights on the topic is through addressing whether pairs of experimental settings that either consolidate or obviate each other in relation to growth effects exist. Documenting these considerations (i.e. interactive effects) could add context towards interpreting past mycorrhizal growth experiments and designing new ones.

There appears to be no standardized way to address such unusual settings. Here, we present results from a systematic literature review aiming to fill this gap (Supporting Information Notes S1). We carried out a quantitative synthesis on subsets of studies from existing meta-analyses showing extreme plant mycorrhizal growth effects. We assessed the degree to which experimental parameters differed between the top-five best- and the bottom-five worse-performing studies per meta-analysis. We compiled a dataset comprising 24 meta-analyses and a total of 240 experimental studies (Notes S1; Fig. S1). Twenty of the studies were duplicates across the meta-analysis, and we repeated our analysis with (see text) and without them (Notes S2), but our results were robust to this consideration. We manually screened those 240 studies for reported experimental settings. We then fitted logistic regression models with our two groups of effect sizes as a response variable: top-performing studies were assigned the value ‘1’ and bottom-performing studies the value ‘0’. With our approach, we assessed evidence in support of a set of four yet unresolved hypotheses (Hypotheses 2–5 in Table 1), addressing whether the results from combing two experimental parameters diverge from additivity (i.e. there are significant interactions and an overarching hypothesis addressing the relative importance with which biotic and abiotic experimental settings contribute to extreme observations) (Hypothesis 1 – Table 1). If a parameter (or pair of parameters) modified sufficiently the growth response of the plant host to mycorrhiza, we could capture this as a change in the frequency with which the parameter (or the pair) was represented in the top and bottom studies and detect it through our statistical models (Notes S1). In our synthesis, we focused only on plant growth responses but we believe that the analysis can be repeated with mycorrhizal root colonization as an alternative response variable.

To our surprise, and subject to the limited pool of abiotic parameters that we considered (Notes S3, S4), we observed small differences in abiotic parameters between top- and bottom-performing studies (Fig. 1). This was the case, even when we visualized the distributions of the abiotic parameters to assess likely dependencies on extreme values or other distribution properties (e.g. Fig. S2). There were comparably stronger differences in relation to biotic variables and single predictor logistic models were only significant for the predictors ‘grass’ (deviance = 12.3, P < 0.001, model S1 as detailed in Notes S2), ‘legume’ (deviance = 3.85, P = 0.049, model S2 as detailed in Notes S2) and C₄ grass (deviance = 12.6, P < 0.001, model S1 as detailed in Notes S2; the coefficient for C₄ grasses was negative. The respective model for C₃ grasses was nonsignificant, and there were no differences between C₃ and C₄ grasses), presenting classifiers describing the host plant. We anticipated (Hypothesis 1) that abiotic parameters would overshadow biotic ones, because in a context-dependent symbiosis, such as arbuscular mycorrhiza (e.g. Hoeksema et al., 2010), unrealistic experimental settings, which in most cases are abiotic, are more likely to induce extreme observations. We are also aware that communities of mycorrhizal fungi under field settings are predominantly shaped by abiotic settings (e.g. van Geel et al., 2018), which we rationalized as evidence that reciprocal benefits from the symbiosis also rely on abiotic conditions. We observed instead that, even under these specific extreme settings, plant mycorrhizal responses mainly depended on some biotic parameters such as the identities of the plant host and mycorrhizal fungus.

We further observed that combining in experiments Graminoid plant hosts with Funneliformis mosseae fungal isolates results in less negative growth effects than either of these experimental parameters alone (evidence in support of Hypothesis 3) and that using F. mosseae can offset some of the positive effects of extending the duration of a study (evidence in support of Hypothesis 2 – Fig. S3; Table S1). Funneliformis  mosseae is a disturbance tolerant mycorrhizal fungus that thrives in agricultural landscapes (Sýkorová et al., 2007) and aggressively colonizes its plant hosts (Jansa et al., 2007). Our understanding was that in short-term experiments, particularly those with fast-growing plants such as grasses, we would observe a high relative frequency of ‘high’ growth responses when the plants had been inoculated with F. mosseae, but this frequency would decline across longer term studies or studies using plants that grow at a slower pace but show stronger responses to mycorrhiza such as legumes (Hoeksema et al., 2010). We further combined these two interaction terms into a single model in which both the positive interaction term for F. mosseae with grasses (deviance 8.40, P = 0.004, model S3 as detailed in Notes S2) and the negative with duration (deviance 9.00, P = 0.003, model S3 as detailed in Notes S2) were significant.

By contrast, we found no evidence that either the interaction term describing sandy soils (but not sandiness of the growth substrate in the growth trials – Notes S1) while experimenting exclusively with Glomeraceae isolates (deviance = 1.73, P = 0.42, Fig. S3; model S4 as detailed in Notes S2), or the respective combination of Glomeraceae with soil pH (deviance = 0.89, P = 0.34, Fig. S3; model S5 as detailed in Notes S2), is of importance for the mycorrhizal growth outcome (i.e. evidence against Hypotheses 4 and 5). The two hypotheses were based on observations that larger spore families such as Gigasporaceae and Acaulosporaceae establish less well in heavy soils (Lekberg et al., 2007), whereas most acid-tolerant arbuscular mycorrhizal fungi do not belong to Glomeraceae (Veresoglou et al., 2013). At the same time, sandy soils tend to have a lower nutrient availability, which is a setting that can promote mycorrhizal growth responses (Hoeksema et al., 2010). The likely difference between Hypotheses 2 and 3 is that these two hypotheses described combinations of biotic and abiotic parameters, and our analysis could be of a lower statistical power than the earlier hypotheses exclusively focused on biotic parameters.

To what level can our findings be generalized? We re-analysed existing experimental studies and, consequently, the range of conditions which we evaluated corresponds to commonly exercised experimental settings. Unlike ordinary meta-analyses, however, our approach allowed us to evaluate pairs of parameters infrequently used together and our results thus reflect a broader range of the possible parameter space. At the same time, it has been impossible to extract information on some experimental parameters such as nutrient availability and microbial biomass in soil. It may well thus be the case that we missed some of the key abiotic settings that determined mycorrhizal growth gains such as P availability, and we worked instead on a set of less important abiotic parameters. A counterargument, however, is that given enough studies (i.e. replication) and assuming that some of our abiotic parameters did modify growth gain, we would have expected to detect changes in the probability with which the studies were classified as top- or bottom-performing.

The take-home message from our analysis was that we were able with our approach to address a set of five hypotheses which remain unresolved in mycorrhizal ecology (Table 1). There is a common perception that it is possible to manipulate controlled experiments addressing mycorrhiza in ways that you can derive any desired growth gain result and the perception is reasonably well supported (Johnson et al., 1997; Johnson & Graham, 2012 but see Smith & Simth, 2013). We are supportive of this perception, but for some reasons, undocumented right now, when synthesizing over the published literature, biotic mycorrhizal parameters contribute to more variance in growth responses than abiotic ones (Fig. 1). Through our synthesis, thereby, we present evidence that it remains possible to robustly synthesize mycorrhizal growth responses in the existing literature across pairs of settings describing host plants and mycorrhizal inoculum type. We feel that this way it should be possible to deliver a new generation of quantitative syntheses in soil ecology that further promotes our understanding on mycorrhiza.

None declared.

SDV conceived the study. MW collected the data with the support of JC. SDV drafted the manuscript with the contributions of MW. MW, JC, TML, JX and SDV contributed comments to the manuscript and approved the final version of it.