Skip to content

Advertisement

  • Research article
  • Open Access

Fagus sylvatica seedlings show provenance differentiation rather than adaptation to soil in a transplant experiment

BMC Ecology201818:42

https://doi.org/10.1186/s12898-018-0197-5

  • Received: 30 October 2017
  • Accepted: 26 September 2018
  • Published:

Abstract

Background

Understanding and predicting the response of tree populations to climate change requires understanding the pattern and scale of their adaptation. Climate is often considered the major driver of local adaptation but, although biotic factors such as soil pathogens or mutualists could be as important, their role has typically been neglected. Biotic drivers might also interact with climate to affect performance and mycorrhizae, in particular, are likely to play a key role in determining drought resistance, which is important in the context of adaptation to future environmental change. To address these questions, we performed a fully reciprocal soil–plant transplant experiment using Fagus sylvatica seedlings and soils from three regions in Germany. To separate the biotic and abiotic effects of inoculation, half of the plants were inoculated with natural soil from the different origins, while the rest were grown on sterilized substrate. We also imposed a drought stress treatment to test for interactions between soil biota and climate. After 1 year of growth, we measured aboveground biomass of all seedlings, and quantified mycorrhizal colonization for a subset of the seedlings, which included all soil–plant combinations, to disentangle the effect of mycorrhiza from other agents.

Results

We found that plant origin had the strongest effect on plant performance, but this interacted with soil origin. In general, trees showed a slight tendency to produce less aboveground biomass on local soils, suggesting soil antagonists could be causing trees to be maladapted to their local soils. Consistently, we found lower mycorrhizal colonization rate under local soil conditions. Across all soils, seedlings from low elevations produced more annual biomass than middle (+ 290%) and high (+ 97%) elevations. Interestingly, mycorrhizal colonization increased with drought in the two provenances that showed higher drought tolerance, which supports previous results showing that mycorrhizae can increase drought resistance.

Conclusions

Our findings suggest that soil communities play a role in affecting early performance of temperate trees, although this role may be smaller than that of seed origin. Also, other effects, such as the positive response to generalists or negative interactions with soil biota may be as important as the highly specialized mycorrhizal associations.

Keywords

  • Ectomycorrhizae
  • Plant-fungal interactions
  • Drought resistance
  • Genetic adaptation
  • Reciprocal transplant
  • Soil

Background

Adapting to climate change will be a great challenge for many ecosystems as climate patterns alter and extreme events, such as droughts, become more frequent [1]. One important factor that will likely influence the environmental response of natural populations is the scale and drivers of adaptation (e.g. [2]). While the existing literature shows that a majority of populations exhibit patterns of local adaptation [3, 4], the drivers of these patterns are less clear. Climatic conditions are the most frequently invoked factor to explain adaptive divergence, but climate may explain only a small part of the local advantage [4, 5]. Adaptation to local biotic factors, such as soil biota, may also be important but has been less frequently considered.

Soil biota play a key role in the functioning of ecosystems, influencing plant performance, community structure and ecosystem dynamics [6, 7]. Soil communities are also highly diverse and variable at small spatial scales across the landscape [8]. It is therefore likely that plants can gain an important advantage by genetically adapting to their local soil community [9]. For instance, Pickles et al. [10] showed that most populations of a widespread tree species, Pseudotsuga menziesii, are locally adapted across the species range and suggested that soil fungi were the main drivers of this pattern. Because of their longevity and large geographic range, tree species encounter a range of different soil communities and there is therefore a large potential for them to specialize to their local soil biota. In some cases, single individuals have even been shown to perform better with their very local soil community [11]. In contrast, there have been reports of partial [1214] or general maladaptation to soil [15], suggesting that other factors and interactions determine whether trees locally adapt to their soils. These cases of maladaptation can be caused by the faster generation time of antagonistic soil microbes, which may mean they are able to rapidly overcome the hosts’ defences [16]. As stressed by Rúa et al. [9], in a recent meta-analysis on the extent of plant local adaptation to soil, while there are strong reasons to suspect that soil biota play a key role in plant local adaptation, it is important to investigate how strongly different populations are adapted to their local soil community, and whether this interaction is dominated by mutualistic or parasitic associations.

The effects of soil biota on trees, and their role in generating local adaptation, can also determine how plants respond to global change, particularly to increasing drought [17]. An increase in drought frequency and severity with climate change [1] is likely to cause increased plant stress and may even result in episodes of mass tree mortality [18, 19]. This could be offset by beneficial soil microbes, which may confer drought resistance: for instance, substantial reductions in soil moisture (more than 7%) have been shown to differentially affect shoot biomass, photosynthetic rate and leaf nitrogen of mycorrhizal and non-mycorrhizal plants [20]. Specific combinations of mycorrhizal and host genotypes have also been shown to completely alter the host’s drought tolerance and, potentially, its response to future environmental change [21]. It has even been suggested that increasing temperatures may increase the frequency of positive mycorrhizae-plant interactions [22], which might in turn enhance differences between populations if mycorrhizae are drivers of local adaptation. An increase in aridity can therefore indirectly affect trees by influencing their interactions with soil biota [23]. However, more experiments are needed to confirm these effects and to increase our understanding of the consequences of climate change for plant-soil pathogen interactions (discussed in [24, 25]). The lack of consistent patterns in the interactions between mycorrhizal fungi, plants and global change drivers ([17, 26, 27]), calls for more controlled experiments that test the interacting effects of environmental stress and specialized soil biota [23].

To predict and manage climate change impacts, it is imperative to understand how economically and ecologically important species will respond to future changes. In Europe, where a few tree species dominate the landscape, their ability to adapt to new climatic conditions greatly influences the functioning of large natural ecosystems. European Beech (Fagus sylvatica, hereafter ‘Fagus’) is one of the most common tree species in Europe. It is also one of the most extensively studied species, due to its wide distribution and economic importance [28]. Fagus populations are almost continuous in the centre of their range, in central Europe. As a consequence, populations are connected by high gene flow (reviewed in [29]) suggesting that between-population genetic differentiation may be low. At the same time, the populations encounter a large range of environmental conditions, including gradients in climate, topography, and potentially in biotic interactions. This, and the reported, high within-population genetic variability [30], suggests that Fagus populations have a large potential for small-scale genetic differentiation [31]. In fact, adaptation to very local conditions has already been reported for Fagus in marginal and central parts of the range [32], as well as along elevational gradients [33]. Although there is a large body of research on the adaptation of Fagus to climate, the extent to which it might be adapted to biotic factors and how these might interact with changing climatic conditions is still largely unexplored.

Here, we examined the interactive effects of local mycorrhizae, tree provenance and drought on early tree growth in a reciprocal soil–plant transplant experiment. This experiment was carried out under common garden conditions with soil and seedlings of Fagus sylvatica from three regions in Germany that are located at three distinct elevations. To assess the interacting effects of soil communities (especially ectomycorrhizal fungi) and drought on tree performance, we applied fully factorial drought and soil inoculation treatments to our seedlings. Specifically, we addressed the following questions: (i) do Fagus seedlings produce more aboveground biomass when growing in soil inoculated with biota from their local origin? (ii) Are mycorrhizal associations more frequent in local soil conditions? (iii) Does drought modify plant–soil biota interactions? And (iv) does mycorrhizal colonisation affect plant tolerance to drought?

Methods

Biodiversity Exploratories and the tree provenances

Our experiment was conducted within the context of the German Biodiversity Exploratories. They are an interdisciplinary and long-term scientific program that studies the effects of land-use intensity and biodiversity on ecosystem functioning in German forests and grasslands. Plots are located in three regions: in the North-East, Center, and South-West of Germany: Schorfheide-Chorin (hereafter ‘Low’), Hainich-Dün (hereafter ‘Middle’), and Schwäbische Alb (hereafter ‘High’). From North to South, the elevations of our study regions are: ELow = 3–140 m, EMiddle = 285–550 m, and EHigh = 460–860 m a.s.l. [34]. As a consequence, temperature and precipitation between regions do not vary as would be expected solely from their latitude. We therefore refer to the soil and plant origins by their elevations, rather than by their latitudinal positions, as this more clearly indicates the climatic differences between them. Populations at higher elevations (southern) actually experience higher precipitation (both annually and during summer) and lower temperatures than those located at middle and low elevations (central and northern latitudes respectively), and thus experience higher water availability at several soil depths (see Table 1). The most drought stressed region is therefore the low elevation one, followed by the middle and then high elevation regions.
Table 1

Mean climatic parameters for each Biodiversity Exploratory region, as measured directly on the monocultural Fagus sylvatica plots from 2009 to 2016

Exploratory site

Code

Summer P (mm)

Summer T (°C)

Annual P (mm)

Annual T (°C)

Soil moisture at 10 cm (%)

Soil moisture at 20 cm (%)

Schorfheide-Chorin

Low

177

16.96

623

8.57

13.67

10.49

Hainich-Dün

Middle

169

15.40

512

7.67

26.19

25.62

Schwäbische Alb

High

273

15.27

808

7.45

32.49

28.21

Bern

CG

358

17.00

986

8.60

Annual values for precipitation and temperature (Annual P, Annual T) express whole-year average across plots, while seasonal values (Summer P and Summer T) are June–August averages. Soil moisture 10 cm and Soil moisture 20 cm are June to August mean percentages of volumetric water content at the specified depth. The temperature and precipitation conditions for the location of the common garden experiment (Bern, CG) are also shown

In our experiment we used the three Exploratory regions as tree provenances (i.e. distinctive seed origins). Between February and April 2014, we collected 2–3 year old naturally regenerated Fagus seedlings, together with surrounding soil samples, from each of the three Biodiversity Exploratory regions. We chose to work with naturally regenerated seedlings, so that we could study effects of soil and drought on older seedlings, which have already passed the environmental filters imposed during the germination and establishment stages. This would not have been possible if we had established the seedlings from seeds germinated under controlled conditions. We took the soil and seedlings from mature monocultural Fagus stands, located near to the site of seedling collection in each region (see [34]). We collected approx. 4 L of top soil, which was then sieved to remove big particles, roots and other organic material and mixed thoroughly before being used for the inoculations. We also added to this the soil obtained from gently shaking the roots of the seedlings, as this rhizosphere soil may be more likely to contain mycorrhizal spores. In total, we collected 450 seedlings from the three regions. We brought them to the Botanical Garden of the University of Bern (Switzerland) to conduct the experiment. The seedlings from the low and high elevation regions were transported by car immediately after collection while those from middle elevations were sent by mail.

Experimental design

All phases of the experiment took place in a location independent of the three provenance origins, in the Botanical Garden of Bern, Switzerland (46°57′07″ N–7°26′43″E, 501 m a.s.l.). Bern has a temperate climate, with mean annual temperatures of 8.6 °C and mean annual precipitation of 986 mm (see Table 1). This means Bern is warmer and wetter than any of the source regions, meaning no provenance should be best adapted to the local climate. This allowed us to focus on the effects of the soil community and the experimentally imposed drought. The experiment followed a complete factorial design with 36 different treatment combinations (3 soil origins × 3 plant origins × 2 inoculation treatments × 2 drought treatments). We randomly assigned the seedlings to one of 25 blocks, each one including all possible combinations of soil, plant, and inoculation. Plants were randomly positioned within each block. Before planting, we washed the root systems of the seedlings to remove all adhering soil and fungal material. We did not cut off the potentially mycorrhized root tips of the seedlings, as this could have seriously damaged their root system and compromised their survival. We used a standard seedling substrate, which was a blend of white and black sphagnum peat buffered with coco fibre (Klasmann-Deilmann GmbH, Geeste, Germany) that had previously been sterilized by autoclaving. This soil mix is commonly used in greenhouses and forest nurseries and it has previously been used in studies looking at mycorrhizae, which have shown that it is a suitable substrate for ectomycorrhizae and good colonisation has been observed on seedlings grown in this soil [35]. All seedlings were planted in 13 × 13 × 13 cm square pots. We inoculated the pots with 30 mL of natural soil, thoroughly mixed with the standard substrate before planting (see [36]). We preferred this inoculation method over a complete soil reciprocal transplantation to minimize the effect of abiotic differences between soils. Half of the seedlings were inoculated with natural soil from the three origins and, to further separate biotic from abiotic effects of soil addition, the seedlings in the ‘not inoculated’ treatment received 30 mL of sterilized soil from the corresponding soil origin. The sterilization of the inoculum liberated the organic nitrogen in the soil (see Additional file 1: Fig. S1, calculated as total soil nitrogen). It is difficult to say whether this is a permanent release, but we consider this effect to be negligible as the amount of inoculum was small relative to the volume of standard substrate and the standard substrate was fairly rich in nitrogen (Additional file 1: Fig. S1). In any case, in order to reduce the potential effect of nutrient liberation due to sterilization, 2 months after establishing the experiment, we fertilized each pot with 3 g of a slow-release fertilizer rich in nitrogen, phosphorus and potassium (Tardit Top, Hauert Günther Düngerwerke GmbH, Nuremberg, Germany). It possible, although unlikely, that this low level of fertilization affected mycorrhizal colonization (see e.g. [37]) and community structure. However, as the levels of added nutrient were low, we expect this effect to be small. To prevent cross-contamination between pots via watering, we placed them on a raised wooden structure.

During establishment all plants were watered for 1 min per block every other day, which was sufficient to maintain soil humidity. Following this, on July 27th 2015, we moved all plants to an open-sided greenhouse, adjacent to the common garden. This was done in order to control water input, since 12 of the blocks were randomly assigned to a 6 week drought treatment, lasting until 6th September. Other than preventing any precipitation reaching the seedlings, the conditions inside the greenhouse are expected to be similar to those outside. The drought treatment simulated a 50% reduction of water during a short-to-medium length drought, which, according to most climate change scenarios, is likely to become much more frequent in the future [1]. We chose a slight drought treatment over a moderate or extreme one as we aimed to study the consequences of a realistic reduction in water availability for the seedlings in the near future [1]. Consequently, the response of plant-soil interactions to severe or extreme drought conditions, though an interesting question, lay outside of the scope of this paper and should be addressed in future work. During the treatment, ‘Drought’ blocks were watered every 4 days for 1 min per block, while ‘control’ blocks were watered every other day. The drought plants showed signs of stress, with strong leaf wilting and slight discoloration, but no leaf loss was reported. Due to very high temperatures inside the greenhouse from August 5th to August 15th, we decided to reduce the intensity of drought by watering control seedlings daily and drought seedlings every other day. Afterwards, the drought treatment was resumed as before. At the end of the treatment the plants were moved back to their original location, outside of the greenhouse, until the end of the experiment.

Measurements

At the beginning of the experiment, we measured plant height to account for initial size differences and to control for potential maternal or storage effects [38]. In November and December 2015, after the seedlings had shed their leaves and entered winter dormancy, we collected the biomass of the seedlings, above the root collar. The samples were air dried in the oven at 70 °C for 3 days and then quickly weighed with a precision balance before they were rehydrated. We removed occasional dry leaves still attached to the seedlings before weighing. Final aboveground biomass values were corrected for initial height so that we measured aboveground biomass growth in the experiment.

We also quantified ectomycorrhizal fungi on the root system of a subset of the seedlings. We selected 6 blocks of 18 seedlings each (3 from the drought and 3 from the control treatment, 108 individuals in total) to quantify and identify their mycorrhizal associations. We used the staining protocol for mycorrhizal quantification from Vierheilig et al. [39] after adjusting the heating times and KOH concentrations to maximize mycorrhizal visibility in our samples. We took 10 segments of thin root tips per tree, cleaned them with water and submerged them at 90 °C in a 10% KOH solution for 10 min. After rinsing with distilled water we stained them with a 10% ink-vinegar solution for 2 min at boiling temperature, and again cleaned them with distilled water. We kept the samples in slightly acidified water over night to remove any excess stain. The stained root fragments were inspected under a 16 × − 64 × stereoscopic magnifier (Leica AG, Wetzlar, Germany) as the colonized roots are clearly visible following staining. We measured the number of total and mycorrhizal tips for all root fragments and averaged them to produce a single estimate of mycorrhizal colonization per seedling. Although the staining method is more often used in arbuscular mycorrhizal associations, it is also sometimes used to characterize ectomycorrhizae (e.g. [40]), because it allows researchers to visualize the Hartig net under the stereoscope, and thus provide a better estimate of the abundance of active mycorrhizae [41]. However, we acknowledge that this may come at a cost of underestimating the colonization rate of very superficial fungi species, which may have been damaged by the clearing method. Similarly, since clearing and staining may affect mycorrhizal morphotype characteristics, we did not assess the diversity and species composition of the mycorrhizal communities in our samples.

Unfortunately, we have no measurements of root biomass, which would have complemented the measures of aboveground biomass and could have clarified the observed patterns. Two reasons played a role in our decision not to measure belowground biomass. First, it is not possible to measure belowground biomass non-destructively and root length-biomass relationships are not reliable for tree seedlings [42]. We therefore felt that we could not measure root biomass at the end of the experiment because such a measure would have been too strongly affected by any difference in initial mass and, unlike for aboveground biomass, we could not correct for these initial differences in root biomass. Second, the few studies that explored the relationship between above- and below- ground biomass through the lifecycle of trees have reported a consistent and strong correlation between them (e.g. [43]). Therefore, although we only measured aboveground biomass growth, we hope that our results capture the general patterns of seedling performance and that the measured change in aboveground biomass should correlate with the change in belowground biomass. However, it is certainly possible that we would have found clearer responses if we had measured root biomass production during the experiment.

Statistical analysis

We used linear mixed models (lme) to test the effects of plant and soil origin on aboveground biomass (AGB). We included block as a random factor to account for potential spatial variability within the experimental site. The full model for AGB included all possible interactions between variables:
$$\begin{aligned} lme \, (log\left( {AGB} \right)\,\sim \,soil.origin \, * \, plant.origin \, \quad * \, drought \, * \, inoculum\, + \,height.0, \, random \hfill \\ \, = \,\sim 1|block, \, method = \hbox{''}ML\hbox{''}) \hfill \\ \end{aligned}$$

Aboveground biomass was log-transformed to meet normality and heteroscedasticity model assumptions. Initial plant height was included as a covariate. Consequently, biomass estimates presented here are standardized by initial height, so that they represent changes in biomass during the experiment (see [38]).

To analyse mycorrhizal colonization (mycoperc) we used generalized linear models (glm). We treated block as a fixed effect, which allowed us to fit glms instead of mixed models. The full model was then defined as:
$$\begin{aligned} glm \, (mycoperc\,\sim \,soil.origin \, * \, plant.origin \, \quad * \, drought \, * \, inoculum\, + \,block, \, family \\ \, & = \,quasibinomial) \\ \end{aligned}$$

We used a quasibinomial distribution to reduce the overdispersion problems we encountered when using a simple binomial [44]. In all cases, we did model selection by backward deletion, based on likelihood ratio tests and the differences in Akaike’s Information Criterion (ΔAIC of more than 2 AIC units was considered threshold for variable selection) between models to find the most plausible model [44]. All statistical analyses were performed with the R software [45] using the ‘nlme’ to fit linear mixed models [46] and ‘MuMIn’ package to calculate R2 values for the mixed models [47].

Results

Seedling performance was affected both by plant and soil origin (Fig. 1). However, the comparison between provenances showed that, after correcting for initial conditions, seedlings from the low elevation provenances produced most biomass and this was consistent in all soils. Low elevation seedlings produced on average 4.47 g more biomass (+290%) than those from middle elevations; and 2.96 g more (+97%) than the high elevation provenances (Fig. 1, Additional file 1: Table S3). This strong hierarchy suggests clear genetic differentiation between provenances. However, maternal effects and storage could also play a role in driving such patterns and, despite our efforts to minimize them by standardizing by initial size, we cannot completely rule out this possibility.
Fig. 1
Fig. 1

Dried aboveground biomass by soil origin and plant origin in the common garden experiment. Open symbols are model predictions after standardizing by initial size. Mean values are displayed as filled symbols with letter code per region. Means of local combinations are displayed as darker colour and with double symbol border. Regions names coded as in Table 1. F-value and significance of the interaction are also displayed

Seedlings from both high and low elevation provenances performed worse in local soil than they did in foreign soil, while seedlings from middle elevations performed slightly better in their own soil (plant origin by soil origin interaction, Table 2, Fig. 1). In the high and low provenances, this may suggest either local maladaptation to mutualists such as mycorrhizae, or stronger negative soil feedback from local pathogens. However, the lack of a significant interaction between inoculation (live vs. sterile) and soil origin (p = 0.10, Table 2), suggests that this response may have been at least partially caused by abiotic differences. The chemical characterization of the soils showed that sterilization strongly modified the concentration of nitrate, phosphate and ammonia (Additional file 1: Fig. S1). Soil pH, on the other hand, remained largely unaffected by sterilization, as expected (Additional file 1: Table S2). Middle elevation soils had neutral pH, while the high and low ones were slightly acidic. This was also true for the standard substrate, which may contribute to explain the poorer performance of middle elevation seedlings in our experiment.
Table 2

Model selection and levels of significance in the linear mixed model analysing seedling aboveground biomass

Aboveground biomass

Explanatory variable

numDF

F-value

p-value

 

ΔAIC

Intercept

 

1668.74

< 0.0001

***

 

Height.0

1

165.20

< 0.0001

***

 

Drought

1

3.44

0.08

n.s.

 

Soil.origin

2

2.84

0.06

n.s.

 

Plant.origin

2

254.28

< 0.0001

***

 

Drought : plant.origin

2

3.78

0.02

*

 

Soil.origin : plant.origin

4

4.81

0.0009

**

 

Inoculum

1

 

0.24

n.s.

− 0.61

Soil.origin : inoculum

2

 

0.10

n.s.

+ 0.60

Drought : soil.origin

2

 

0.36

n.s.

− 1.96

Drought : inoculum

1

 

0.65

n.s.

− 1.79

Drought : soil.origin : inoculum

2

 

0.45

n.s.

− 2.42

Drought : soil.origin : plant.origin

4

 

0.80

n.s.

− 6.35

Plant.origin : inoculum

2

 

0.91

n.s.

− 3.81

Drought : plant.origin : inoculum

2

 

0.80

n.s.

− 3.54

Soil.origin : plant.origin : inoculum

4

 

0.80

n.s.

− 6.38

Drought : soil.origin : plant.origin : inoculum

4

 

0.86

n.s.

− 6.69

Variables are shown in reverse order of the backward variable selection (i.e. variables at the bottom of the table were deleted first from the full model). Variables retained in the minimal model are shown in bolditalics on top. numDF specifies the number of degrees of freedom for each variable. The difference in Akaike’s Information Criterion (ΔAIC) is shown for the variables that were deleted from the model. Both the p value from the likelihood ratio test (p-value) and the ΔAIC values were considered and variables were only retained in the model if their deletion resulted in a significant p-value and a ΔAIC > 2. In the minimal model, fixed factors alone explained 66% of the deviance (R2m = 0.66), while both fixed and random increased the deviance explained by 4% (R2c = 0.70)

Drought reduced the growth of each provenance differently (drought by plant origin interaction, Table 2, Fig. 2). Provenances from high elevations were most affected by the stressful conditions, while seedlings from middle and low elevation regions barely responded to the treatment, independent of the soil in which they were growing. This may suggest pre-adaptation to increasingly stressful conditions in our middle to low elevations, while populations at higher altitudes seemed less adapted to drought as they normally experience colder and wetter climate conditions (Table 1).
Fig. 2
Fig. 2

Effect of drought on aboveground biomass for each provenance. Open symbols are model predictions after standardizing by initial size. Mean values are displayed as filled symbols with letter code per region as in Fig. 1. F-value and significance of the interaction are displayed

Two interactions significantly affected the rate of mycorrhizal colonization: those between drought and plant origin, and between soil origin and inoculum (Table 3). Regarding the first interaction, for two of three provenances, we observed an increase in the number of mycorrhizal root tips in the seedlings subjected to drought (Fig. 3), supporting a role for mycorrhizal interactions in plant resistance to extreme events. Consistent with this idea, seedlings from higher elevations, which were the most affected by drought (Fig. 2), also had reduced mycorrhizal colonisation under stressful conditions (Fig. 3). In contrast, inoculation with active soil significantly increased the amount of mycorrhizal root tips for the higher and lower elevation provenances, most strongly for the higher elevation seedlings (Additional file 1: Fig. S2).
Table 3

Model selection and levels of significance in the generalized linear model analysing the percentage of mycorrhizal root tips

Mycorrhizal root tips (%)

Explanatory variable

numDF

F-value

p-value

 

Soil.origin

2

0.18

0.84

Plant.origin

2

2.26

0.11

Drought

1

0.28

0.60

Inoculum

1

1.82

0.18

Block

1

1.47

0.23

Drought: plant.origin

2

3.15

0.04

*

Soil.origin: inoculum

2

3.54

0.03

*

Drought: inoculum

1

 

0.09

n.s.

Plant.origin: inoculum

2

 

0.21

n.s.

Drought: soil.origin

2

 

0.30

n.s.

Soil.origin: plant origin

4

 

0.73

n.s.

Drought: plant.origin : inoculum

2

 

0.05

n.s.

Soil.origin: plant.origin : inoculum

4

 

0.11

n.s.

Drought: soil.origin : plant.origin

4

 

0.11

n.s.

Drought: soil origin : inoculum

2

 

0.49

n.s.

Drought: soil.origin : plant.origin : inoculum

4

 

0.09

n.s.

Variables are shown in reverse order of the backward variable selection (i.e. variables at the bottom of the table were deleted first from the full model). Variables retained in the minimal model are shown in bolditalics on top. numDF specifies the number of degrees of freedom of each variable

Fig. 3
Fig. 3

Effects of drought on the percentage of mycorrhizal root tips. Small symbols denote raw measurements while mean values per provenance are displayed as big symbols with standard error bars. F-value and significance of the interaction are also displayed

Discussion

No evidence for local adaptation to soil

Our results show that plant and soil origins interact to affect the performance of transplanted seedlings. However, we did not observe an advantage for seedlings grown with home soil, as would be expected in the case of tree local adaptation. Instead, two of the three studied provenances produced less aboveground biomass when transplanted into pots inoculated with local soil, which might suggest maladaptation of Fagus to local soil biota, such as specialized antagonists [48]. It is also possible for ectomycorrhizae to behave antagonistically [10], although it is unclear what determines whether the fungi act as mutualists or parasites. In the case of parasites and pathogens, host maladaptation could be caused by a higher ability of the antagonists to adapt to their hosts than of the hosts to adapt to the antagonists, thanks to their bigger population size and shorter generation time [49, 50]. In these ‘Red Queen dynamics’, the parasites are ahead in the co-evolutionary arms race [16] and, as a consequence, plant performance can be increased by physically escaping from highly adapted antagonists, through dispersal, somewhat analogous to the ‘enemy release’ hypothesis that leads to greater performance of invasive species in their introduced range [51]. Although our results are consistent with antagonist driven host maladaptation, we did not find a difference between inoculated and non-inoculated treatments in the degree of local disadvantage, as would be expected if maladaptation was biotically driven. This is not because the inoculation effects had disappeared because even at the end of the experiment inoculated soil had significantly more mycorrhizae than non-inoculated soils. It is therefore unclear what caused the signal of soil maladaptation even in the non-inoculated pots. Maladaptation to abiotic soil conditions would be more surprising but could explain our results. Alternatively, the different abiotic soil conditions may have selected for different soil communities as biota recolonised the pots. To determine with certainty which of these hypotheses is more likely, characterisation of the soil communities of the different provenances would be needed. Nevertheless our results do show that variation in soils is an important driver of tree seedling performance and that effects depend on the genetic origin of the seedlings.

It should be noted, however, that due to our experimental design, there are some caveats that should be taken into consideration when interpreting our results. First, despite washing their roots, seedlings might have started the experiment with different rhizosphere communities and therefore with their own mycorrhizae and pathogens present. Given that the soil had different effects it is unlikely that this could explain all the results, however, initial differences between seedlings could have affected the soil biota that established in each pot. Similarly, there was no opportunity for us to ascertain the degree of contamination by soil biota from the surrounding of the experiment. This is a common problem for all soil experiments done under open environmental conditions (in contrast to closely controlled environments, e.g. an ECOTRON). Finally, since we could not reliably measure the mycorrhizal community composition for each population, we cannot ensure that the inoculum was effective in creating different mycorrhizal communities in our treatments. Previous literature (discussed below in each section) supports our methods, but these shortcomings should be taken into account.

Provenance differentiation

Contrary to our expectations, we observed clear and consistent differences between our tree provenances. In widespread and highly connected populations, genetic differences between populations are commonly assumed to be low [52], while within-population variability remains high [30]. Therefore, for widespread species, only populations located at the edges of their natural range are expected to show strong genetic differentiation [28]. However, our results indicate that seedlings transplanted from populations in the central part of a species range can still be strongly differentiated despite presumably high levels of gene flow between the populations. It is possible that the provenances differed in their response to the growing conditions in Bern. These were most similar to the conditions in the origin of the higher elevation population, in terms of precipitation, but were warmer than conditions experienced by any population in their origin, and actually closer to the temperatures commonly experienced in the lower elevation region. It is therefore not clear that climatic conditions would have favoured the growth of any particular provenance. It remains unclear what factors might be driving this population differentiation and further work with these provenances using both reciprocal transplant experiments and common garden experiments are needed to identify the drivers of divergent selection.

Our results also show significant differences between provenances in their ability to tolerate drought. Climate change models currently do not consider local adaptation to drought or extreme climate events, but if adaptation to local, stressful conditions is widespread, this could modify the responses of forests to the changing climate. It would mean that the effect of drought on forests is likely to be highly variable [18]. Intraspecific differences in drought tolerance is a key component of future local adaptation that needs to be considered both in climate change models [53], and in the development of local adaptation theory [54].

An important factor to consider in experimental tests of local adaptation are maternal and storage effects, which can confound the observed adaptation patterns [55]. Despite their relevance, few studies explicitly control for them [4]. Usually, plants are grown for 1- 2 generations under common conditions to control for maternal and storage effects, and only the subsequent generations are used in the transplants (e.g. [55]). However, this methodology cannot be applied to long-lived organisms, such as trees, for obvious practical reasons. Therefore, alternative methodologies have been proposed. We used one of the most common ones, which accounts for maternal and storage effects by standardizing growth by initial plant size (e.g. [38]). This aims to control for differences in levels of stored reserves and in pre-experimental conditions between the different provenances. Using naturally regenerated plants allowed us to work with older seedlings and to explore the effects of our treatments on seedlings which have already passed some early environmental filters. It would interesting, however, to compare the results from seedlings transplanted from naturally regenerated trees, as in our case, with plants grown from seeds or under sterile conditions, to compare and confirm the strength of the observed patterns, and observe any potential confounding effects of transplanting established field seedlings into greenhouse pots. Unlike plants germinated under experimental conditions in isolated pots, naturally regenerated plants can profit or be harmed by the extensive mycorrhizal network that connects tree roots in natural forest soils. This is likely to affect the ability of fungi to uptake and transport water and thus, potentially the plant’s drought resistance. We cannot explore this question using our experimental design, but future work that combines plants germinated under experimental conditions and naturally regenerated ones would be useful to test the importance of these factors.

Mycorrhizal abundance

Contrary to what was has been observed for other species [10, 36], we did not find an overall increase in mycorrhizal colonisation when tree seedlings were grown on their local soil (soil origin by plant origin interaction, Table 3). We did, however, observe an effect of inoculation for two of the provenances, which suggests that inoculation did affect the mycorrhizal community. Overall, our results rather suggest that mycorrhizal communities were unresponsive to plant origin, soil origin, and stressful conditions; and indicate that these associations may be more flexible than usually considered. However, even without changes in abundance, mycorrhizal species composition could have differed between provenances. It is also possible that the seedlings maintained part of the associations they already had despite our efforts to carefully clear the roots before transplanting, or that particular associations with symbionts present in the experimental site happened to favour a particular tree phenotype [21]. We also cannot exclude the possibility that a provenance or particular community of inoculum is, by chance, better adapted to grow under our experimental conditions. This adaptation to laboratory conditions is a shortcoming of all common garden approaches (discussed in [54]). We have no expectation for any of our provenances or soil communities to be better adapted to growing on our potting soil mix, since none of the soils in the origin of the provenances are located in peatlands or flooded areas, which would be closer to the potting soil. However, a better assessment of the fungal community composition and parallel reciprocal transplant experiments would be necessary to clarify this in the future.

The small differences in mycorrhizal colonization between local and foreign soils might also suggest that mycorrhizal communities do not differ between regions as much as expected, likely due to the high connectivity between Fagus populations in central Europe [29], which would tend to homogenise the populations of associated biota. This is in agreement with recent studies that have shown very low levels of endemism in arbuscular mycorrhizae, although these may differ in their response compared to ectomycorrhizae [56]. If ectomycorrhizae also show low endemism then trees would be unlikely to adapt to local mycorrhizal communities. On the other hand, Pena et al. [57] recently showed that mycorrhizal communities occurring in different forest types in the Biodiversity Exploratories were highly differentiated. Since our seedlings were sampled from monocultural Fagus stands in all three regions, it is possible that we would have found larger differences in mycorrhizal communities if we had compared between forest types rather than geographical locations.

The small differences we observe in mycorrhizal colonization between provenances (see e.g. [58]) likely indicate that some community homogenization have happened during the common garden experiment, despite our attempts to minimize cross-contamination between plots. This colonization via greenhouse borne spores or between pots is difficult to quantify and prevent, and it can potentially confound the measurements of mycorrhizal diversity and composition. And it is possible that a larger ratio of inoculum/substrate would have provided better resistance to this homogenization or potential external colonization by common mycorrhiza (but see [36]). Previous experiments that used the same substrate as we employed, suggested it is relatively resistant to external mycorrhizal colonization, as non-inoculated plants remained free of root colonization (in their case of arbuscular mycorrhizal fungi) throughout the experiment [59]. However, the role of specialized mycorrhizae might be underestimated in our experimental settings compared with that of natural conditions. While pre-grown sterile seedlings would have reduced the problem of potential pre-existing spores or fungi over the roots, they would have been similarly sensitive to cross-contamination between pots. However, by using naturally established plants we are able to work with older, more mature seedlings, and therefore to test for local adaptation at a later life history stage than is common in tree seedling experiments.

Effects of drought on the plant-mycorrhizae association

We found a small increase in mycorrhizal colonization under more stressful environments, consistent with the hypothesis that plant resistance to extreme events is affected by mycorrhizal associations [60, 61]. This could be a result of mycorrhizae increasing host nutrient uptake [62] which may in turn facilitate drought resistance. These small differences appeared even under the light drought treatment which we imposed here and under more extreme droughts, a stronger response of mycorrhizae could be expected.

Drought resistance is likely to be of key importance for seedling survival under future climatic conditions [53] and its dependence on specialized, small-scale biotic interactions may imply that tree responses to future conditions can vary substantially depending on whether climate change decouples interactions with their soil biota [23, 63]. An increasing number of studies support this idea: for example, Liang et al. [64] showed that altitudinal migration of trees in response to climate change can be strongly modified by the biotic interactions that plants encounter in the new area. Similarly, forest stands experiencing recent high mortality had lower colonization rates by ectomycorrhizae, suggesting that the reduced mycorrhizal associations had hampered tree performance and increased the likelihood of tree mortality in the face of extreme conditions [65]. While models predicting species responses to climate change have increased in complexity and accuracy by incorporating some biotic and abiotic interactions [62, 66], and intra-specific variability [53], the interactions between plant and soil biota are yet to be included.

Conclusion

Soil effects on seedling performance suggested a general dominance of antagonistic interactions between soil biota and Fagus sylvatica provenances in Germany. However, the effect of soil origin was smaller than that of plant origin. We found a consistent hierarchy in population performance, with lower elevation provenances showing increased growth compared with medium and higher elevation ones. This differentiation in seedling performance was not related to our experimental variables. We did, however, find evidence for a positive effect of mycorrhizal associations on seedling resistance to stressful conditions. Overall, our results suggest that the genetic differentiation between populations of Fagus sylvatica and their interactions with generalist soil biota are most important for seedling responses to climate change.

Abbreviations

AGB: 

aboveground biomass

AIC: 

Akaike’s information criterion

Fagus

Fagus sylvatica

glm: 

generalized linear models

lme: 

linear mixed models

Mycorperc: 

percentage of mycorrhizal colonization

Declarations

Authors’ contributions

RDM conceived and led the project. RDM, MF and EA designed and set the experiment. FS quantified and identified mycorrhizal colonization while RDM measured the other variables. RDM and EA performed the statistical analyses. All authors contributed to data interpretation and writing in consecutive versions of the manuscript.

Acknowledgements

We thank the managers of the three Exploratories: Kirsten Reichel-Jung, Katrin Lorenzen, Martin Gorke, and all former managers for their work in maintaining the plot and project infrastructure, Christiane Fischer for giving support through the central office, Michael Owonibi for managing the central data base, and Markus Fischer, Eduard Linsenmair, Dominik Hessenmöller, Jens Nieschulze, Daniel Prati, Ingo Schöning, François Buscot, Ernst-Detlef Schulze, Wolfgang W. Weisser and the late Elisabeth Kalko for their role in setting up the Biodiversity Exploratories project. Falk Haensel, Stephan Sylvio Wöellauer, and Thomas Nauss provided the weather data summarized in Table 1. We thank Martin Schwarz, Joerg Hailer, Judith Hinderling, Maximilian Frei, Juliet Blum, Floris Schenk, and Wayne Dawson for their valuable help with the field collection and experimental set up. We thank Stefan Blaser for his assistance in the mycorrhizal identification, and 2 anonymous reviewers who provided valuable feedback on the manuscript.

Competing interests

The authors declare that they have no competing interests.

Availability of data and materials

The datasets used and analysed during the current study are available from the corresponding author on reasonable request.

Consent to publish

Not applicable.

Ethics approval and consent to participate

Fieldwork permits were issued by the responsible state environmental offices of Baden-Württemberg, Thüringen, and Brandenburg (according to § 72 BbgNatSchG).

Funding

Not applicable.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Authors’ Affiliations

(1)
Institute of Plant Sciences, University of Bern, Altenbergrain 21, 3013 Bern, Switzerland

References

  1. Orlowsky B, Seneviratne SI. Global changes in extreme events: regional and seasonal dimension. Clim Change. 2012;110:669–96.View ArticleGoogle Scholar
  2. Valladares F, Matesanz S, Guilhaumon F, Araújo MB, Balaguer L, Benito-Garzón M, Cornwell W, Gianoli E, van Kleunen M, Naya DE, et al. The effects of phenotypic plasticity and local adaptation on forecasts of species range shifts under climate change. Ecol Lett. 2014;17:1351–64.View ArticleGoogle Scholar
  3. Leimu R, Fischer M. A meta-analysis of local adaptation in plants. PLoS ONE. 2008;3:e4010.View ArticleGoogle Scholar
  4. Hereford J. A quantitative survey of local adaptation and fitness trade-offs. Am Nat. 2009;173:579–88.View ArticleGoogle Scholar
  5. Joshi J, Schmid B, Caldeira MC, Dimitrakopoulos PG, Good J, Harris R, Hector A, Huss-Danell K, Jumpponen A, Minns A, et al. Local adaptation enhances performance of common plant species. Ecol Lett. 2001;4:536–44.View ArticleGoogle Scholar
  6. Reynolds HL, Packer A, Bever JD, Clay K. Grassroots ecology: plant-microbe-soil interactions as drivers of plant community structure and dynamics. Ecology. 2003;84:2281–91.View ArticleGoogle Scholar
  7. Kulmatiski A, Beard KH, Cobbold SM. Plant-soil feedbacks: a meta-analytical review. Ecol Lett. 2008;11:980–92.View ArticleGoogle Scholar
  8. Lang C, Seven J, Polle A. Host preferences and differential contributions of deciduous tree species shape mycorrhizal species richness in a mixed Central European forest. Mycorrhiza. 2011;21:297–308.View ArticleGoogle Scholar
  9. Rúa MA, Antoninka A, Antunes PM, Chaudhary VB, Gehring C, Lamit LJ, Piculell BJ, Bever JD, Zabinski C, Meadow JF, et al. Home-field advantage? Evidence of local adaptation among plants, soil, and arbuscular mycorrhizal fungi through meta-analysis. BMC Evol Biol. 2016;16:1–15.View ArticleGoogle Scholar
  10. Pickles BJ, Twieg BD, O’Neill GA, Mohn WW, Simard SW. Local adaptation in migrated interior Douglas-fir seedlings is mediated by ectomycorrhizas and other soil factors. New Phytol. 2015;207:858–71.View ArticleGoogle Scholar
  11. Smith DS, Schweitzer JA, Turk P, Bailey JK, Hart SC, Shuster SM, Whitham TG. Soil-mediated local adaptation alters seedling survival and performance. Plant Soil. 2012;352:243–51.View ArticleGoogle Scholar
  12. Macel M, Lawson CS, Mortimer SR, Šmilauerová M, Bischoff A, Crémieux L, Doležal J, Edwards AR, Lanta V, Bezemer TM, et al. Climate vs. soil factors in local adaptation of two common plant species. Ecology. 2007;88:424–33.View ArticleGoogle Scholar
  13. Hoeksema JD, Chaudhary VB, Gehring CA, Johnson NC, Karst J, Koide RT, Pringle A, Zabinski C, Bever JD, Moore JC, et al. A meta-analysis of context-dependency in plant response to inoculation with mycorrhizal fungi. Ecol Lett. 2010;13:394–407.View ArticleGoogle Scholar
  14. Grassein F, Lavorel S, Till-Bottraud I. The importance of biotic interactions and local adaptation for plant response to environmental changes: field evidence along an elevational gradient. Glob Change Biology. 2014;20:1452–60.View ArticleGoogle Scholar
  15. Wagner V, Antunes PM, Ristow M, Lechner U, Hensen I. Prevailing negative soil biota effect and no evidence for local adaptation in a widespread Eurasian grass. PLoS ONE. 2011;6(3):e17580.View ArticleGoogle Scholar
  16. Van Valen LM. A new evolutionary law. Evol Theory. 1973;1:1–30.Google Scholar
  17. Kivlin SN, Emery SM, Rudgers JA. Fungal symbionts alter plant responses to global change. Am J Bot. 2013;100:1445–57.View ArticleGoogle Scholar
  18. Allen CD, Macalady AK, Chenchouni H, Bachelet D, McDowell N, Vennetier M, Kitzberger T, Rigling A, Breshears DD, Hogg EH, et al. A global overview of drought and heat-induced tree mortality reveals emerging climate change risks for forests. For Ecol Manag. 2010;259:660–84.View ArticleGoogle Scholar
  19. McDowell NG, Williams AP, Xu C, Pockman WT, Dickman LT, Sevanto S, Pangle R, Limousin J, Plaut J, Mackay DS, et al. Multi-scale predictions of massive conifer mortality due to chronic temperature rise. Nat Clim Change. 2016;6:295–300.View ArticleGoogle Scholar
  20. Kennedy PG, Peay KG. Different soil moisture conditions change the outcome of the ectomycorrhizal symbiosis between Rhizopogon species and Pinus muricata. Plant Soil. 2007;291:155–65.View ArticleGoogle Scholar
  21. Gehring CA, Sthultz CM, Flores-Rentería L, Whipple AV, Whitham TG. Tree genetics defines fungal partner communities that may confer drought tolerance. Proc Natl Acad Sci USA. 2017;114:11169–74.View ArticleGoogle Scholar
  22. Rillig MC, Wright SF, Eviner VT. The role of arbuscular mycorrhizal fungi and glomalin in soil aggregation: comparing effects of five plant species. Plant Soil. 2002;238:325–33.View ArticleGoogle Scholar
  23. Bardgett R, Manning P, Morriën E, De Vries FT. Hierarchical responses of plant-soil interactions to climate change: consequences for the global carbon cycle. J Ecol. 2013;101:334–43.View ArticleGoogle Scholar
  24. Van der Putten WH, Macel M, Visser ME. Predicting species distribution and abundance responses to climate change: why it is essential to include biotic interactions across trophic levels. Philos T R Soc B. 2010;365:2025–34.View ArticleGoogle Scholar
  25. Letho T, Zwiazek JJ. Ectomycorrhizas and water relations of trees: a review. Mycorrhiza. 2011;21:71–90.View ArticleGoogle Scholar
  26. Mohan JE, Cowden CC, Baas P, Dawadi A, Frankson PT, Helmick K, Hughes E, Khan S, Lang A, Machmuller M, et al. Mycorrhizal fungi mediation of terrestrial ecosystem responses to global change: mini-review. Fungal Ecol. 2014;10:3–19.View ArticleGoogle Scholar
  27. Karst J, Randall MJ, Gehring CA. Consequences for ectomycorrhizal fungi of the selective loss or gain of pine across landscapes. Botany. 2014;92:855–65.View ArticleGoogle Scholar
  28. Magri D, Vendramin GG, Comps B, Dupanloup I, Geburek T, Gömöry D, Latalowa M, Litt T, Paule L, Roure JM, et al. A new scenario for the Quaternary history of European beech populations: palaeobotanical evidence and genetic consequences. New Phytol. 2006;171:199–221.View ArticleGoogle Scholar
  29. Wang KS. Gene flow in European beech (Fagus sylvatica L.). Genetica. 2004;122:105–13.View ArticleGoogle Scholar
  30. Porth I, El-Kassaby YA. Assessment of the genetic diversity in forest tree populations using molecular markers. Diversity. 2014;6:283–95.View ArticleGoogle Scholar
  31. Steiner KC, Berrang PC. Microgeographic adaptation to temperature in pitch pine progenies. Am Midl Nat. 1990;123:292–300.View ArticleGoogle Scholar
  32. Kreyling J, Buhk C, Backhaus S, Hallinger M, Huber G, Huber L, Jentsch A, Konnert M, Thiel D, Wilmking M, et al. Local adaptations to frost in marginal and central populations of the dominant forest tree Fagus sylvatica L. as affected by temperature and extreme drought in common garden experiments. Ecol Evol. 2014;4:594–605.View ArticleGoogle Scholar
  33. Csilléry K, Lalagüe H, Vendramin G, González-Martínez SC, Fady B, Oddou-Muratorio S. Detecting short spatial scale local adaptation and epistatic selection in climate-related candidate genes in European beech (Fagus sylvatica) populations. Mol Ecol. 2014;23:4696–708.View ArticleGoogle Scholar
  34. Fischer M, Bossdorf O, Gockel S, Hänsel F, Hemp A, Hessenmöller D, Korte G, Nieschulze J, Pfeiffer S, Prati D, et al. Implementing large-scale and long-term functional biodiversity research: the Biodiversity Exploratories. Basic Appl Ecol. 2010;11:473–85.View ArticleGoogle Scholar
  35. Lu N, Yu M, Cui M, Luo Z, Feng Y, Cao S, Sun Y, Li Y. Effects of different ectomycorrhizal fungal inoculates on the growth of Pinus tabulaeformis seedlings under greenhouse conditions. Forests. 2016;7:1–14.Google Scholar
  36. Johnson NC, Wilson GWT, Bowker MA, Wilson JA, Miller RM. Resource limitation is a driver of local adaptation in mycorrhizal symbioses. Proc Natl Acad Sci USA. 2010;107:2093–8.View ArticleGoogle Scholar
  37. Rincon A, Alvarez IF, Pera J. Inoculation of containerized Pinus pinea L. seedlings with seven ectomycorrhizal fungi. Mycorrhiza. 2001;11:265–71.View ArticleGoogle Scholar
  38. Raabová J, Münzbergová Z, Fischer M. The role of spatial scale and soil for local adaptation in Inula hirta. Basic Appl Ecol. 2011;12:152–60.View ArticleGoogle Scholar
  39. Vierheilig H, Schweiger P, Brundrett M. An overview of methods for the detection and observation of arbuscular mycorrhizal fungi in roots. Physiol Plant. 2005;125:393–404.Google Scholar
  40. Gómez-Aparicio L, Domínguez-Begines J, Kardol P, Ávila JM, Ibáñez B, García LV. Plant-soil feedbacks in declining forests: implications for species coexistence. Ecology. 2017;98:1908–21.View ArticleGoogle Scholar
  41. Brundett M, Bougher N, Dell B, Grove T, Malajczuk N (eds). 1995. Working with mycorrhizas in forestry and agriculture. ACIAR Monograph 32. Camberra, Australia.Google Scholar
  42. Vadeboncoeur M, Hamburg S, Yanai RD. Validation and refinement of allometric equations for roots of northern hardwoods. Can J For Res. 2007;37:1777–83.View ArticleGoogle Scholar
  43. Peichl M, Arain MA. Allometry and partitioning of above- and belowground tree biomass in an age-sequence of white pine forests. For Ecol Manag. 2007;253:68–80.View ArticleGoogle Scholar
  44. Crawley MJ. The R book –. 2nd ed. Imperial College London at Silwood Park: John Wiley & Sons Ltd; 2014.Google Scholar
  45. R Core Team. 2015. R: A language and environment for statistical computing: R foundation for statistical computing, Vienna, Austria.Google Scholar
  46. Pinheiro J, Bates D, DebRoy S, Sarkar D, R Core Team. 2018. nlme: Linear and Nonlinear Mixed Effects Models. R package version 3.1-137. https://CRAN.R-project.org/package=nlme.
  47. Barton K, Barton MK. 2018. Package ‘MuMIn’. R package version 3.1, ftp://155.232.191.229/cran/web/packages/MuMIn/MuMIn.pdf.
  48. Crémieux L, Bischoff A, Šmilauerová M, Lawson CS, Mortimer SR, Doležal J, Lanta V, Edwards AR, Brook AJ, Tscheulin T, et al. Potential contribution of natural enemies to patterns of local adaptation in plants. New Phytol. 2008;180:524–33.View ArticleGoogle Scholar
  49. Kaltz O, Shykoff JA. Local adaptation in host-parasite systems. Heredity. 1998;81:361–70.View ArticleGoogle Scholar
  50. Gandon S. Local adaptation and the geometry of host-parasite coevolution. Ecol Lett. 2002;5:246–56.View ArticleGoogle Scholar
  51. Keane RM, Crawley MJ. Exotic plant invasions and the enemy release hypothesis. Trends Ecol Evol. 2002;14:164–70.View ArticleGoogle Scholar
  52. Gonzalo-Turpin H, Hazard L. Local adaptation occurs along altitudinal gradient despite the existence of gene flow in the alpine plant species Festuca eskia. J Ecol. 2009;97:742–51.View ArticleGoogle Scholar
  53. Benito-Garzón M, Alía R, Robson TM, Zavala MA. Intra-specific variability and plasticity influence potential tree species distributions under climate change. Glob Chan Biogeogr. 2011;20:766–78.View ArticleGoogle Scholar
  54. Gutschick VP, BassiriRad H. Extreme events as shaping physiology, ecology, and evolution of plants: toward a unified definition and evaluation of their consequences. New Phytol. 2003;160:21–42.View ArticleGoogle Scholar
  55. Kawecki TJ, Ebert D. Conceptual issues in local adaptation. Ecol Lett. 2004;7:1225–41.View ArticleGoogle Scholar
  56. Davison J, Moora M, Öpik M, Adholeya A, Ainsaar L, Bâ A, Burla S, Diedhiou AG, Hiiesalu I, Jairus T, et al. Global assessment of arbuscular mycorrhizal fungus diversity reveals very low endemism. Science. 2015;349:970–3.View ArticleGoogle Scholar
  57. Pena R, Lang C, Lohaus G, Boch S, Schall P, Schöning I, Ammer C, Fischer M, Polle A. Phylogenetic and functional traits of ectomycorrhizal assemblages in top soil from different biogeographic regions and forest types. Mycorrhiza. 2016. https://doi.org/10.1007/s00572-01-0742-z.View ArticlePubMedGoogle Scholar
  58. Baxter JW, Dighton J. Ectomycorrhizal diversity alters growth and nutrient acquisition of grey birch (Betula populifolia) seedlings in host-symbiont culture conditions. New Phytol. 2001;152:139–49.View ArticleGoogle Scholar
  59. Perner H, Schwarz D, George E. Effect of mycorrhizal inoculation and compost supply on growth and nutrient uptake of young leek plants grown on peat-based substrates. Hort Sci. 2006;41:628–32.Google Scholar
  60. Shi L, Guttenberger M, Kottke I, Hampp R. The effect of drought on mycorrhizas of beech (Fagus sylvatica L.): changes in community structure, and the content of carbohydrates and nitrogen storage bodies of the fungi. Mycorrhiza. 2002;12:303–11.View ArticleGoogle Scholar
  61. Kernaghan G. Mycorrhizal diversity: cause and effect? Pedobiologia. 2005;49:511–20.View ArticleGoogle Scholar
  62. Smith SE, Read DJ. Mycorrhizal symbiosis. New York: Academic Press; 2008.Google Scholar
  63. Wisz MS, Pottier J, Kissling WD, Pellisier L, Lenoir J, Damgaard CF, Dormann CF, Forchhammer MC, Grytnes JA, Guisan A, et al. The role of biotic interactions in shaping distribution and realised assemblages of species: implications for species distribution modelling. Biol Rev. 2013;88:15–30.View ArticleGoogle Scholar
  64. Liang E, Wang Y, Piao S, Lu X, Camarero JJ, Zhu H, Zhu L, Ellison AM, Ciais P, Peñuelas J. Species interactions slow warming-induced upward shift of treelines on the Tibetan Plateau. Proc Natl Acad Sci USA. 2016;113:4380–5.View ArticleGoogle Scholar
  65. Swaty RL, Deckert RJ, Whitham TG, Gehring CA. Ectomycorrhizal abundance and community composition shifts with drought: predictions from tree rings. Ecology. 2004;85:1072–84.View ArticleGoogle Scholar
  66. Guisan A, Thuiller W. Predicting species distribution: offering more than simple habitat models. Ecol Lett. 2005;8:993–1009.View ArticleGoogle Scholar

Copyright

© The Author(s) 2018

Advertisement