The strength of negative plant–soil feedback increases from the intraspecific to the interspecific and the functional group level

Abstract One of the processes that may play a key role in plant species coexistence and ecosystem functioning is plant–soil feedback, the effect of plants on associated soil communities and the resulting feedback on plant performance. Plant–soil feedback at the interspecific level (comparing growth on own soil with growth on soil from different species) has been studied extensively, while plant–soil feedback at the intraspecific level (comparing growth on own soil with growth on soil from different accessions within a species) has only recently gained attention. Very few studies have investigated the direction and strength of feedback among different taxonomic levels, and initial results have been inconclusive, discussing phylogeny, and morphology as possible determinants. To test our hypotheses that the strength of negative feedback on plant performance increases with increasing taxonomic level and that this relationship is explained by morphological similarities, we conducted a greenhouse experiment using species assigned to three taxonomic levels (intraspecific, interspecific, and functional group level). We measured certain fitness‐related aboveground traits and used them along literature‐derived traits to determine the influence of morphological similarities on the strength and direction of the feedback. We found that the average strength of negative feedback increased from the intraspecific over the interspecific to the functional group level. However, individual accessions and species differed in the direction and strength of the feedback. None of our results could be explained by morphological dissimilarities or individual traits. Synthesis. Our results indicate that negative plant–soil feedback is stronger if the involved plants belong to more distantly related species. We conclude that the taxonomic level is an important factor in the maintenance of plant coexistence with plant–soil feedback as a potential stabilizing mechanism and should be addressed explicitly in coexistence research, while the traits considered here seem to play a minor role.


| INTRODUCTION
The exact mechanisms maintaining species coexistence remain largely unresolved. With regard to individual plant species, not only abiotic factors (environmental conditions), but also a number of biotic factors such as intraspecific competition (Stoll & Prati, 2001), interspecific competition (Goldberg & Barton, 1992), a species' associated soil community as well as the associated soil communities of other plant species (van de Voorde, van der Putten, & Bezemer, 2011) plays important roles for plant-plant interactions and thus for coexistence between them. Plant-soil feedback as a process potentially maintaining plant species coexistence when acting as a stabilizing mechanism (Chesson, 2000;HilleRisLambers, Adler, Harpole, Levine, & Mayfield, 2012) has received considerable attention (Bever, 2003;Bever, Platt, & Morton, 2012;Bever, Westover, & Antonovics, 1997;Ehrenfeld, Ravit, & Elgersma, 2005;Klironomos, 2002;Kulmatiski, Beard, Stevens, & Cobbold, 2008;van der Putten et al., 2013). This idea is based on the fact that a plant community influences its associated soil community, and the soil organisms have specific feedback effects on their host plants in turn (Bever et al., 1997). This soil community can contain mutualists and/or pathogens (Adewale, Aremu, & Amazue, 2012;Bever, 2003;Bever et al., 1997Bever et al., , 2012. In addition, abiotic mechanisms can lead to feedback effects, such as the release of allelochemical compounds by the plants or specific nutrient depletion (van der Putten et al., 2013). Negative feedback which has been reported to be more common than positive feedback, at least in experimental systems  may enhance species coexistence via increasing negative density dependence (i.e., as a stabilizing mechanisms sensu ChessonChesson (2000)) and thus, support the maintenance of species diversity when it is strong enough to balance out fitness differences between species (Petermann, Fergus, Turnbull, & Schmid, 2008) In contrast, positive feedback might lead to a loss of species diversity (Bever et al., 2012). The prevailing direction of plant-soil feedback may depend on a number of parameters, for example, plant functional group identity (Heinze, Bergmann, Rillig, & Joshi, 2015), plant life form (van de Voorde et al., 2011), plant abundance (Heinze et al., 2015), size-related traits of a plant species (Heinze et al., 2015), the composition of the plant community (Kulmatiski & Kardol, 2008), and whether a plant species is native or invasive (Klironomos, 2002;Reinhart, Packer, Van der Putten, & Clay, 2003).
It has been stated that trait variation at the intraspecific as well as at the interspecific level has an influence on species coexistence (Albert, Grassein, Schurr, Vieilledent, & Violle, 2011;Bolnick et al., 2011). However, most soil feedback studies refer to the interspecific level only, comparing growth of plants on own soil with growth on soil from other species. Recently, it has been shown that plant-soil feedback may operate at the intraspecific level, that is, that there are differences in plant growth on own soil compared with growth on soil from different accessions or genotypes within the same species (Bukowski & Petermann, 2014;Liu, Etienne, Liang, Wang, & Yu, 2015).
Despite indications that certain pathogens may have similar effects on closely related species (Gilbert, Briggs, & Magarey, 2015;Parker et al., 2015), it is still unclear whether there is a difference in feedback strength between the intraspecific and the interspecific levels (van der Putten et al., 2013). Indeed, there is considerable debate on whether the strength of plant-soil feedback experienced by each plant individual in a community is predictable from information on species relatedness. It has been shown that plant-soil feedback may be mediated by plant traits (Heinze et al., 2015) and that, in most cases, closely related species have similar traits (Blomberg, Garland, & Ives, 2003;Burns & Strauss, 2011;Gilbert & Webb, 2007;Webb, Gilbert, & Donoghue, 2006). Anacker, Klironomos, Maherali, Reinhart, and Strauss (2014) have related plant species relatedness to the strength of soil feedback between them, suggesting phylogeny as a major determinant of plant-soil feedback (Brandt, Seabloom, & Hosseini, 2009). On the other hand, this relationship could not be confirmed by a recent metaanalysis (Mehrabi & Tuck, 2015).
To experimentally test whether there is indeed a relationship between the taxonomic relatedness of plants and the strength of the feedback they experience, we conducted a plant-soil feedback experiment at different taxonomic levels. The defined taxonomic levels were as follows: the intraspecific level (different accessions within a species), the interspecific level (closely related species of the same plant family), and the functional group level (species of different functional groups that are very distantly related). We measured plant-soil feedback as relative plant performance on home soil (trained by the same accession/species) versus away soil (trained by another accession/species), whereby positive feedback implies a better plant performance on home soil compared to away soil, and vice versa for negative feedback.
Additionally, we compared accessions and species based on measured morphological traits in order to investigate whether morphological similarities between our accessions/species or individual traits of the accessions/species might explain the feedback effects. We hypothesized that: 1. Plant individuals experience plant-soil feedback, predominantly negative feedback, at all taxonomic levels.

2.
Plant-soil feedback between accessions (i.e., at the intraspecific level) is weaker than between species (interspecific level) and functional groups (functional group level).

3.
The strength of plant-soil feedback can be explained by morphological similarities between accessions/species or by individual traits.

| Experimental species
For the plant-soil feedback experiment, a total of eleven accessions/ species were used as follows: four accessions of one species for the intraspecific level, five species for the interspecific level, and four spe- Tsu-0 (Tsu, Japan), Bur-0 (Burren, Ireland), and Na-1 (Nantes, France).
As Col-0 is the most explored accession being used as wild type or reference accession of A. thaliana in most studies (Fahlgren et al., 2006;Frenkel et al., 2009;Xiao et al., 2001), we decided to use this accession in all three parts of the plant-soil feedback experiment.
To investigate plant-soil feedback at the interspecific level, we  Koch, Bishop, & Mitchell-Olds, 1999;Price, Palmer, & Al-Shehbaz, 1994). According to the latest classification, all species used here rate among a monophyletic group which consists of eleven tribes (Kiefer et al., 2014). Our approach of classifying species and accessions as either closely or distantly related is based on their taxonomic relatedness, disregarding phylogenetic relationships. Phylogeny may have the potential to determine species' relatedness even more precisely; however, we were not able to construct a phylogenetic tree for our experimental accessions/species because precise phylogenetic data are not available for all A. thaliana accessions and its closely related species (but see Lee, Guo, Wang, Kim, and Paterson (2014) for phylogenies of some of our accessions).

| Plant-soil feedback experiment
Following the common approach of conducting plant-soil feedback experiments (Aguilera, 2011;Hendriks et al., 2013;Kulmatiski, Beard, & Heavilin, 2012;Kulmatiski & Kardol, 2008;Petermann et al., 2008;Reinhart, 2012;van de Voorde et al., 2011), the experiment consisted of two phases. This was a training phase consisting of monocultures growing in neutral soil whereby changing its biotic and abiotic conditions in a specific way and a subsequent experiment phase using the trained soil to test its effect on a new generation of plants.
F I G U R E 1 Scheme depicting the design of the experiment phase of the plant-soil feedback experiment. The experiment consisted of an intraspecific, interspecific, and functional group levels. Each accession/species was growing in home soil that had been trained by the same accession/species (indicated by the semicircular arrow) as well as in away soil that had been trained by another accession/species. Within each taxonomic level, every accession/species was growing in all possible away soil types (indicated by the connecting lines). The focal Arabidopsis thaliana accession Col-0 (center) was used in all three parts of the experiment

Intraspecific level Functional group level
Interspecific level

| Training phase
For the first phase, we used a substrate consisting of 70% premixed soil (standard soil and perlite, see below) and 30% inoculum which was field soil taken from an old field site near Freie Universität Berlin.
Using field soil enabled us to investigate the effects of the biotic components of the soil. The field soil had been sieved (mesh size: 2 mm) in order to remove stones, roots, and other large objects. The surface-sterilized seeds were stratified in dry condition in the refrigerator at 5°C for 4 days to ensure synchronous germination. Seeds were sown in pots (height 10 cm, diameter 11 cm) according to the following design. During the entire experiment, plants were growing in monocultures in groups of ten. For this, seeds of one accession/ species were placed in a circle of ten in a pot.

| Experiment phase
Prior to the experiment phase, the soil was prepared as follows. As the soil trained by L. perenne, T. pratense, and P. lanceolata contained thick roots compared to the soil trained by the other plant species, the majority of those roots was removed by sieving (

| Plant material, greenhouse conditions, and soil composition
Seeds of the A. thaliana accessions Col-0, Tsu-0, Bur-0, and Na-1 as  The pairs of replicate plants on home and away soil, respectively, which were used for calculating the log-transformed feedback ratios were selected randomly for each calculated ratio. As the numbers of home and away pots were not equal (six home pots for each accession/species, nine or twelve away pots for each accession/species), some data points of home pots were randomly used twice. To test for any effect that the original pairing might have on the results, we performed an additional bootstrap procedure for the ratio calculation by sampling with replacement for 1,000 iterations.

| Statistical analyses
Specifically, we calculated the log-transformed biomass ratio for each accession/species 1,000 times and calculated 95% bootstrap confidence intervals to determine whether this ratio was significantly different from zero. We used the R package boot (Canty & Ripley, 2013;Davison & Hinkley, 1997) for the bootstrap procedure.  For determining trait similarities between accessions/species, we used the three traits "biomass," "height," and "fitness" measured in the training phase of our experiment as well as trait values from the literature of five categorical traits. The trait "fitness" describes the proportion of individuals per pot that produced seeds. As categorical traits from the literature we used "rosette formation," "life form," "association with arbuscular mycorrhizal fungi (AMF)," "nitrogen fixation (NF)," and "life span" for a further characterization of each accession/species (Table S1)

| RESULTS
On average, plants at all three taxonomic levels experienced nega- The bootstrapping showed that the pairing of the individuals on home and away soils for calculating the feedback ratios did not affect the results ( Figure S1). Trait dissimilarities between accessions/species were not related to the strength of the soil feedback (Fig. 3, Mantel test with 999 permutations, Mantel r statistic: −.1174, p-value: .68). In addition, we did not find single traits to be related to the strength of the soil feedback (Table 2).

| Increasing feedback strength from low to high taxonomic levels
On average, negative plant-soil feedback operated in our experiment, partly confirming our first hypothesis. The average negative feedback was stronger between functional groups (functional group level) than between species (interspecific level) and between accessions (intraspecific level), confirming our second hypothesis. In line with this, T A B L E 2 Linear model results for the effect of single traits on mean feedback values for each accession/species. "Biomass (g)," "height (cm)," and "fitness" were measured during the training phase of the experiment, whereas the categorical traits "rosette," "life form," "association with arbuscular mycorrhizal fungi (AMF)," "nitrogen fixation (NF)," and "life span" were extracted for each accession/species from the literature. The trait "fitness" describes the proportion of individuals per pot that produced seeds F I G U R E 3 Pairwise trait dissimilarities between accessions/ species show no relationship with the average plant-soil feedback of the respective accession/species pair. For determining trait dissimilarities, we used data from eight traits and calculated the Gower distance for each pair of accessions/species. The dissimilarity of accession/species pairs ranged from 0 (low dissimilarity) to 1 (high dissimilarity). To calculate the corresponding plant-soil feedback value, we used biomass data on home soils for each accession/ species divided by the biomass of this accession/species on away soil that was conditioned by the corresponding accession/species from that species pair. See the Results section for the statistical analysis using a Mantel test

Dissimilarity of accessions/species Log(biomass home/biomass away)
the focal A. thaliana accession Col-0 experienced stronger negative feedback at the functional group level than at the interspecific level or the intraspecific level. These results indicate that the taxonomic relatedness of the involved plant species may be an important factor determining plant-soil feedback. This relationship could be driven by pathogens as agents of negative feedback which have been shown to be shared by closely related species . We also observed some (rarer) cases of positive feedback, potentially caused by potentially less specific mutualists such as mycorrhizae (Klironomos, 2002). The relationship between the relatedness of plant species and the strength of plant-soil feedback is controversially debated in the literature. A meta-analysis which was based on results of plant-soil feedback experiments that were performed in the last 20 years could not explain the strength and direction of feedback by phylogeny (Mehrabi & Tuck, 2015). In line with this, a recent study showed that plants did not perform better when growing in soil being trained by a more distinctly related species compared to soil from a closely related species (Mehrabi, Bell, & Lewis, 2015). On the other hand, some studies have found evidence for the relationship between relatedness of plant species and strength as well as direction of plant-soil feedback (Brandt et al., 2009;Liu et al., 2012) which are confirmed by our results for the focal a A. thaliana accession Col-0 as well as for the average feedback across taxonomic levels. These contrasts between studies may indicate that neither the strength nor the direction of plant-soil feedback can be explained by plant phylogeny in all cases, but that we may have to consider a number of different factors and their interplay when ascertaining the cause of observed plant-soil feedback, and relatedness between species is certainly one of them.

| Direction and strength of plant-soil feedback vary among accessions and species
Many of the accessions and species experienced plant-soil feedback, partly confirming our first hypothesis. However, the individual accessions and species differed in the direction and strength of the feedback. With regard to intraspecific plant-soil feedback, we had formerly shown that A. thaliana accessions differ in the strength and direction of experienced feedback (Bukowski & Petermann, 2014 Regarding the functional group level, our results are partly consistent with other studies which found P. lanceolata, L. perenne, and T. pratense to suffer from negative feedback (Bever, 2002;Harrison & Bardgett, 2010;Hendriks et al., 2015;Klironomos, 2002;Petermann et al., 2008). Furthermore, in contrast to our results, it has been shown that plants perform better in soil having been trained by species from the functional group of herbs than in soil trained by grass species due to a depletion of potassium in grass-trained soil .
However, comparing absolute results of different studies is generally difficult due to differences in experimental procedures such as the proportion of inoculum used for the experiment phase which has been shown to indeed alter plant performance (Nicot & Rouse, 1987; Pernilla Brinkman, Van der Putten, Bakker, & Verhoeven, 2010), especially as biotic and abiotic conditions may be strong determinants of the strength and direction of feedback (Ehrenfeld et al., 2005;Heinze et al., 2015;Ke, Miki, & Ding, 2015;Mazzoleni et al., 2015;Nicot & Rouse, 1987;Pernilla Brinkman et al., 2010). Further, we used only one species of each functional group for this experiment, making it impossible to distinguish functional group-specific and species-specific effects, an issue that could be resolved by a more complex experimental design in future studies.

| Can feedback strength be explained by trait dissimilarities?
In our experiment, we did not find a relationship between trait dissimilarities between accession/species or individual traits and the strength of plant-soil feedback, which is in contrast to our third hypothesis that morphology might contribute to feedback effects. It has long been known that soil organisms act as main agents of plant-soil feedback (Bever et al., 1997). In addition, there is a greater awareness that processes such as plant-mediated nutrient cycling may have impacts on those plant-microbial interactions, thereby contributing to the resulting feedback effects (Ehrenfeld et al., 2005;Teste et al., 2017). In this context, plant functional traits have been shown to have an effect on the abiotic (here: chemical) properties of the soil they are growing in (Binkley & Giardina, 1998). The importance of these plant traits as determinants of plant-soil feedback has recently been shown to depend on the composition of the soil communities (Ke et al., 2015;Mazzoleni et al., 2015). Closely related species may still accumulate distinct soil communities leading to differences in plant performance when growing in different soil types and therefore to different feedback effects (Pendergast, Burke, & Carson, 2013). In contrast to that, other studies have shown that more closely related plant species may be more similar to each other than to distinctly related species with regard to the composition of specific soil communities (Gilbert & Webb, 2007;Webb et al., 2006) as well as in terms of ecological traits such as germination rate, seedling survival (Burns & Strauss, 2011), and responses to infestation by pathogens Parker et al., 2015). Contrary to our expectations, we did not find such a relationship between feedback strength and speciesspecific morphological or functional traits of the involved accessions/ species. However, in our study, we only included a limited number of traits. For example, we did not measure possibly relevant belowground traits such as root morphological traits or the ability of plants to influence the composition of soil communities via root exudates, which could clearly have influenced the strength of feedbacks. We also did not consider flowering times or successional stages of the plants, the latter of which has been shown to be connected to the strength of soil feedback (Kardol, Martijn Bezemer, & van der Putten, 2006) This limitation might be part of the reason why we did not find a relationship between traits or trait similarities and feedback strength.
A recent study by Cortois, Schröder-Georgi, Weigelt, van der Putten, and De Deyn (2016) investigated the relationship between feedback strength and several aboveground traits (relative growth rate, specific leaf area) and belowground traits (specific root length, percent arbuscular mycorrhizal fungi colonization) simultaneously and found that the direction and strength of feedback were best explained by the examined belowground traits. Species with high specific root length and low arbuscular mycorrhizal fungi colonization experienced the most negative soil feedback (see also Bennett et al. (2017)). We conclude that future studies that link plant-soil feedback, phylogenetic relatedness, and morphological trait similarities should focus on traits and especially on trait differences in the belowground compartment to better understand the relationship between feedback and phylogeny.

| CONCLUSION
We found that on average, negative plant-soil feedback operated in our experiment, with increasing strength of the negative feedback with increasing taxonomic distance between the involved players. This result has important implications for the assembly and the resulting structure of plant communities. If the taxonomic differences in feedback that we found similarly occur under field conditions, this might, for example, explain why plant communities are more readily invaded by more distantly related species, and why diverse communities are more resistant to invasion (Fargione, Brown, & Tilman, 2003;Petermann et al., 2010). Generally, plant-soil feedback could be an important mechanism maintaining diversity at all taxonomic levels, with stronger structuring effects at high taxonomic levels favoring the highest possible diversity within the limits of the environmental filter.
However, negative soil feedback even emerged at the intraspecific level in our experiment, indicating a contribution to the maintenance of diversity even below the species level, where plant-soil feedback is rarely considered as a coexistence mechanism. While we did not find a relationship between plant-soil feedback and absolute traits or trait similarities between accessions/species our results highlight the importance of explicitly considering relatedness when examining plant-soil feedback as a coexistence mechanism.

ACKNOWLEDGMENTS
We would like to thank Julia Thüringer for ordering seeds and other materials. Special thanks to Viktoria Goncarova for her extraordinary support in the greenhouse. Further greenhouse assistance was kindly provided by Reinhard Broz, Andreas Bukowski, Angelika Bukowski, Paolo Fischer, Janine Hardtke, Mouhammad Shadi Khudr, Johannes Meka, Bernd Richter, and Inga Venjakob. Finally, we thank Susanne Wurst (Freie Universität Berlin) and her research group for providing their drying cabinet, fine scales, and autoclaving bags.