Local adaptation to biotic factors: reciprocal transplants of four species associated with aromatic Thymus pulegioides and T. serpyllum

Authors


Correspondence author. E-mail: bodil.ehlers@biology.au.dk

Summary

  • 1A plant producing secondary compounds may affect the fitness of other plants in the vicinity, and, likewise, associated plants may evolve adaptation to the presence of their ‘chemical neighbour’. Species of the genus Thymus are aromatic plants, well known for their production of aromatic oils whose constitution is dominated by mono- or sesquiterpenes. A polymorphism for the production of the dominant terpene in the oil exists both within and between thyme species.
  • 2Here we examine the effects of two different terpenes produced by Thymus pulegioides and T. serpyllum on the performance of four associated plant species: Achillea millefolium, Agrostis capillaris, Galium verum and Plantago lanceolata. In a reciprocal transplant experiment we studied how plants naturally occurring together with thyme producing either carvacrol or b-caryophyllene perform on soil treated with these compounds.
  • 3We found evidence of local adaptation to the ‘home’ terpene. Plants originating from sites where they grow together with carvacrol-producing thyme plants also perform better on soil treated with carvacrol. One of the associated species (A. millefolium) also showed evidence of local adaptation to the sesquiterpene b-caryophyllene .
  • 4Seed germination and root biomass showed an adaptive response to soil treatment. Vegetation analysis supported the results of the reciprocal transplant experiment. When the associated species performed best on ‘home’ soil, thyme and the associated species also showed a positive spatial association at natural sites of origin. Moreover, coefficients of variation in plant traits were significantly lower on ‘home’ soil compared to other soils for both A. capillaris and A. millefolium, but higher for G. verum.
  • 5Synthesis. Our results show that plant species can adapt to the presence of neighbour plants that produce specific chemical compounds. This supports the idea that local plant communities may be a lot more co-evolved than was previously thought.

Introduction

Local adaptation of plants to their abiotic environment is both well studied and documented. Plants show local adaptation to their home soil (e.g. reviews by Brady et al. 2005 for adaptation to serpentine soils, and Macnair 1987 to heavy metal-contaminated soils), to climate and climate-related factors (e.g. Waser & Price 1985; Galloway & Fenster 2000; Joshi et al. 2001; Bischoff et al. 2006) and to a combination of both (e.g. Macel et al. 2007). However, few studies have focused on plant species adapting to their biotic environment in the form of the presence of other plant species.

The ecological consequences of plant-plant interactions where one plant leaches specific allelochemicals into the surrounding environment have received a great deal of interest (e.g. Muller 1966; Muller et al. 1969; Bartholomew 1970; Rice 1984; Callaway & Aschehoug 2000; Bais et al. 2003). The effects of an allelochemical on the performance of other plants may vary from inhibitory to facilitative and may also depend on the life-stage of the other plant (e.g. Callaway & Walker 1997). The presence of such plants may therefore have a large impact on the species composition and dynamics of the local plant community. Indeed, recent studies have shown that associated plants can evolve local adaptation to the presence of their ‘chemical neighbour’ (e.g. Mallik & Pellissier 2000; Ehlers & Thompson 2004; Callaway et al. 2005).

How a plant responds and potentially adapts to the chemical environment created by its neighbours is also of interest to understand how native plants may respond and subsequently evolve under the presence of an invasive plant species producing allelopathic chemicals to which the native plant community is naïve (e.g. Callaway & Aschehoug 2000; Ridenour & Callaway 2001; Callaway & Ridenour 2004; Vivanco et al. 2004).

Plants producing essential oils or resins are well known examples of species which biochemically affect the performance of associated species. Both resins and essential oils have terpenes as their dominant component (Langenheim 1994). The production of terpenes has many purposes, including defence against herbivores, parasites and pathogens (e.g. Vokou et al. 1984, 1998; Linhart & Thompson 1999). For aromatic plants, the production of essential oils may also be an adaptation to the climate as they increase protection against summer drought and fires (Blondel & Aronson 1999; Thompson 2005). Thus, aromatic plants may have evolved their allelochemicals for reasons other than obtaining a competitive advantage to other plant species, but these chemicals also affect neighbouring plant species which may co-evolve to better tolerate these.

Aromatic plants dominate in regions with a Mediterranean climate (Thompson 2005) and most studies on the effects of essential oils on other species have been conducted in these regions. The impact of essential oils produced by species of the family Lamiaceae and in particular species of the genus Thymus on biotic interactions has received a great deal of interest (Tarayre et al.1995; Linhart & Thompson 1995, 1999; Rahman & Gul 2003; Ehlers & Thompson 2004; Linhart et al. 2005, see also Thompson 2002, and references therein). Evidence for local adaptation of the grass Bromus erectus to the presence of specific monoterpenes produced by Thymus vulgaris was found by Ehlers & Thompson (2004). Bromus erectus individuals originating from populations where T. vulgaris produced a non-phenolic monoterpene germinated and grew significantly better on their own soil compared to B. erectus individuals originating from nearby populations, where the local thyme species produced a phenolic monoterpene.

While terpene concentrations in leaf and litter of plants producing essential oil and resins are well known (e.g. Vokous & Margaris 1986; White 1991; Wilt et al. 1993; Venskutonis 2002), few studies have reported the terpene concentration in the soil underneath such plants. Monoterpenes in soil can be found both as gas in the soil micro-air, and in an aqueous phase (e.g. White 1991; Pavolainen et al. 1998). The concentration of terpene decreases from fresh leaves to decaying litter (Vokous & Margaris 1986; Wilt et al. 1993) and to soil (e.g. White 1991; Wilt et al. 1993). Wilt et al. (1993) found that monoterpene concentration in fresh pine needles was on average 50 times higher than the concentration found in mineral soils underneath the pine trees. However, White (1991) stresses that mineral soil may actually act as sink for monoterpenes that move in the soil both via gas exchange and by water. White (1991) also stresses that detection of unbound monoterpene in soil is difficult, making measures of the concentration of terpene in soils difficult and often underestimated. Heating of soil reduces the concentration of chemical compounds with a low boiling point e.g. volatile compounds such as terpenoids (Herranz et al. 2006). For Mediterranean plant communities, this heating factor may be important in reducing the amount of terpenes in the soil – both due to heating of soil in the warm summer period, and especially by fires. However in Northern European plant communities fires and summer drought are rare, thus possibly favouring an accumulation of terpenes in the soil.

In Denmark, two naturally occurring thyme species (Thymus pulegioides and T. serpyllum) are found in dry grasslands. Both thyme species are polymorphic with respect to the dominating terpene found in their oil. A survey showed that in Danish populations of T. pulegioides the oil is predominately composed of a single monoterpene, most frequently carvacrol (Grøndahl et al. 2008). In contrast, the constitution of the oils of T. serpyllum is more mixed and comprises between two to four different types of terpenes. Indeed, the sesquiterpene β-caryophyllene is very frequent and is one of the dominant compounds in all Danish T. serpyllum populations surveyed (16 populations, Grøndahl, Keefover-Ring and Ehlers, unpublished data). Thymus pulegioides is found in the region around the Baltic Sea where the climate is warmer and the soil more fertile compared to habitats where T. serpyllum is found. The latter is also found in Western Denmark where soils are sandier. Both species, and in particular T. pulegioides, are rare. However, when present, the clonal growth of both species results in dense formation of mats and tuffs which completely dominate the ground vegetation. Hence, in plant communities where they are present, they have the potential to affect their associated species via releasing terpenes into the soil.

As the various terpenes differentially affect the performance of other species, the distribution of thyme chemotypes within and between local populations creates a spatial structure in selection on species co-occuring with thyme. One interesting question is whether associated plants adapt to the presence of neighbouring thyme plants, and even to the local thyme chemotype producing a specific terpene.

The purpose of this study was to examine the effects of two dominating terpenes (carvacrol and β-caryophyllene) found in the essential oil of two naturally occurring Thymus species (T. pulegiodies and T. serpyllum) on the germination and growth of four different plant species co-occuring with Thymus in Danish grassland communities. In particular, we address the following questions:

  • 1Are carvacrol and β-caryophyllene produced by Thymus allelopathic? In other words, are other plants affected by the presence of these chemicals in the soil?
  • 2If so, do other plant species show evidence of local adaptation to species- and ecotype-specific thyme biochemistry? Do plants collected from sites where carvacrol is the dominant allelochemical in co-occurring thyme plants perform better on soil treated with carvacrol than plants collected from sites where β-caryophyllene is the dominant allelochemical, and vice versa?

Methods

study sites and study species

Four sites, all located around the Baltic Sea (two in the east-northeastern part of Jutland and two on Funen) were chosen for this study. At two of the sites, T. pulegioides dominated the ground vegetation, and T. serpyllum dominated part of the ground vegetation at the other two sites. We named these sites TP-1, TP-2, TS-3 and TS-4 respectively. At sites with T. pulegioides, T. serpyllum was not present, and vice versa for sites with T. serpyllum.

We selected four plant species (Agrostis capillaris L. (Poaceae), Achillea millefolium L. (Asteraceae), Plantago lanceolata L. (Plantaginaceae) and Galium verum L. (Rubiaceae)) that were common at all sites, that is, which co-occurred with both T. pulegioides and T. serpyllum, except for P. lanceolata which was only present at three of the four sites (absent from one T. serpyllum site). All four associated species were frequently found growing in very close proximity to thyme plants – even in the middle of the thyme tuffs. We collected seeds of all species from each site.

vegetation analysis

To estimate the level of spatial association between thyme and its associated species in natural sites we performed the following vegetation analysis. At each of the four study sites we randomly positioned 50 Raunkiær circles (Raunkjær 1909). Within each circle (0.1 m2) we recorded the presence or absence of Thymus and each of the four associated study species.

In addition, to estimate the local species diversity of herbaceous plants we randomly laid out 10 plots (40 × 40 cm) at each site and within each plot we recorded every herbaceous species present. The frequency of each species was then used to obtain estimates of two species diversity indices; Shannon's diversity index D, calculated as: D = 1/∑inline image , and Simpson's diversity index H, calculated as H =–∑ Pi × ln Pi, where Pi is the proportion of the ith species and summed over all species.

experimental set up

Seeds and plant individuals of the four associated species from each of the four study sites were grown on three types of soils: soil treated with the T. pulegioides monoterpene; carvacrol, soil treated with the T. serpyllum sesquiterpene; β-caryophyllene and control soil where no terpene was added. We termed these soil types: TP-soil, TS-soil, and Control soil, respectively.

This kind of reciprocal transplant experiment allowed us to examine if each species in general was affected by the presence of terpenes in the soil (by comparing performance on carvacrol and β-caryophyllene treated soil with control soil), and if one type of terpene affected the performance more than the other (compare performance on carvacrol soil with β-caryophyllene soil). Finally we searched for local adaptation by testing if plants originating from a site with carvacrol-thyme oil perform better on carvacrol treated soil than plants originating from a site with β-caryophyllene thyme oil, and vice versa.

soil preparation

Soils (Pindstrup mixture number 1) were prepared to contain 33 µL of terpene per 100 g soil (dry weight, hereafter dry wt.), which is equivalent to 322 and 299 µg g−1 soil of carvacrol and β-caryophyllene respectively. The two types of terpenes were added to the soil in the following way. Liquid carvacrol and β-caryophyllene (Sigma Aldrich) were mixed in separate petri dishes with filter paper, cut into small pieces of approximately 1 cm2 (Filtrak paper sheet, grade 17.95 g m−2) and sealed with plastic film for 24 h after which all liquid had soaked into the filter paper. Filter papers were then mixed carefully into the soil using one single container of soil per treatment. Soil with filter paper was left for another 24 h to homogenize the concentration of terpene before transferring soil to individual germination trays or pots.

The concentration of carvacrol and β-caryophyllene added to the soil was similar to the lower concentrations of essential oils used by Vokou et al. (1984) in their experiment on effects of essential oil on soil respiration. Vokou & Margaris (1986) report that fresh leaves of Thymus contains between 4.5–5 mL volatile oil/100 g. In the carvacrol chemotypes of both T. vulgaris and T. pulegioides, carvacrol makes up between 60% and 80% of the constituent of the oil (Senatore 1996; Thompson et al. 2003; Keefover-Ring, Grøndahl and Ehlers unpubl. result). Assuming carvacrol makes up to 65% of total oil content, one gram of fresh thyme leave from a carvacrol plant contains about 30 mg of carvacrol (density of carvacrol is 0.976 g mL−1). Further, assuming that the soil contains on average 50 times less terpene than that found in fresh leaves (Wilt et al. 1993), we can estimate the concentration of carvacrol in the soil underneath T. pulegioides plants to be on average 600 µg g−1 dry wt. Note that the dilution factor of 50 assumed here is reported to vary from 20 to 130 (Wilt et al. 1993) and we can thus expect the carvacrol concentration to be within the range 230–1500 µg g−1 dry wt. For β-caryophyllene in T. serpyllum a similar assumption can be made but here the concentration is likely smaller because β-caryophyllene contributes less to the total oil composition as compared to carvacrol in T. pulegioides (Stahl-Biskop 2002, Keefover-Ring, Grøndahl and Ehlers unpubl. results). In this study we used a concentration of terpene in the soil of 322 and 299 µg g−1 dry wt. soil for carvacrol and β-caryophyllene respectively. The above calculations demonstrate that the concentrations we used are within the range of what associated plant species experience in the field.

experiment 1: effect of soil type on germination and juvenile biomass

We collected a bulk sample of seeds from at least 20 plants of each of the four associated species at each of the four study sites. All seeds collected were examined and seeds which were determined to be non-viable (empty) were discarded. For each species and from each of the four sites, 450 seeds were selected and divided equally among treatments (TP-soil, TS-soil and Control). Each treatment was replicated three times, with 50 seeds in each replicate.

From February, germination of seeds was scored twice a week for a period of 3 months until germination had reached a plateau. After germination had ceased, all seedlings from each replicate were harvested. For each species and each replicate (within treatment), three seedlings were selected to continue the experiments of adult biomass (see Experiment 2). We carefully chose seedlings which had similar size within each species and across replicates. The remaining seedlings were rinsed, stored in a paper bag and later oven dried at 35–40 °C. The dried seedlings were subsequently weighted using a microgram precision balance and their biomass was divided into total, above-ground and below-ground (root) biomass.

experiment 2: effect of soil type on adult biomass and reproduction

Nine seedlings of each species from each treatment and each site were transferred to pots (size 10 cm diameter) containing a soil type similar to their previous treatment. Each plant was transplanted into its own pot. Hence, for each species 108 plants divided between three treatments were used in the experiments, except for Plantago where only one population of TS origin was sampled yielding 81 individual plants.

Plants were kept in an unheated glasshouse and pots were randomized twice a week. No plants died during the experiment. All plants of Achillea and Plantago flowered whereas only 20 of Agrostis and even fewer of Galium flowered. After flowering at the end of the year, all plants were harvested.

Each harvested adult plant was rinsed, kept in a paper bag and later oven dried as described above for seedlings. After drying, the biomass of plants was assessed using a milligram precision balance.

Due to damage to the roots when removing the soil, we chose not to include root biomass of adult plants in our analysis for all species except Agrostis where careful cleaning was possible without breaking and damaging roots. Above-ground biomass was divided into vegetative and reproductive material. For Achillea, the umbel was removed directly beneath its base. For Plantago, flowering spikes were separated from the rest of the plant. Due to the low number of flowering Galium and Agrostis plants, we decided not to include the data of their reproductive biomass in the analysis.

Leaves of Plantago adults were subject to severe herbivore damage by slugs at the end of the experiment and we therefore only had data for number of spikes of the adult plants.

statistical analysis

The vegetation data consisted in 50 Raunkiær circles at each study site. Within each circle we scored the presence or absence of Thymus and each of the four associated study species. At each site, and for each associated species, we built a 2 × 2 contingency table with presence and absence of thyme and the associated species as rows and column, respectively. To examine if the co-occurrence of an associated species with Thymus differed from a random association we performed a G-test of independence on the 2 × 2 table.

Germination rate, biomass of juvenile (above-ground and roots), adult (above-ground for all species, and for Agrostis also roots), and biomass of flowers (Achillea) and number of spikes (Plantago) were analysed using a mixed model anova with origin (fixed), site nested within origin (random), treatment nested within origin (fixed) and their interaction as predictor variables.

When we detected a significant or nearly significant (P < 0.1) treatment effect, we further used anova contrasts to test for two specific biological hypotheses. First, we tested for an allelopathic effect of terpenes by contrasting performance on soils treated with terpenes (‘home’ and ‘away’) vs. control. Second we tested if performances on the two types of terpenes are indicative of local adaptation by contrasting performances on ‘home’ vs. ‘away’ soils. Due to the many contrasts performed, there was a risk that some may appear significant due to chance alone. To control for the proportion of significant tests that are in fact coming from the null hypothesis (i.e. false discoveries), we applied the Benjamini & Hochberg (1995) false discovery rate (FDR) method. Controlling for an FDR of δ = 5% and ranking the P-values in ascending order, an individual threshold for each P-value can be obtained as: p(i) ≤ δ/m × i, where p(i) is the P-value of rank order i, and m is the number of individual tests (see also Verhoeven et al. 2005 for detailed description on FDR methods).

For clarity we present the data averaged over sites, and report in the text if significant differences between sites within origins are detected. All analyses were performed using the computer program jmp (sas Institute Inc., Cary, NC1999).

Testing for overall biological effects – combining P-values

The several tests of performance on ‘home’ vs. ‘away’ soil can be combined into an overall test for local adaptation across species by combining P-values from each individual test (whether significant or not) using the technique developed by Fisher (See Sokal & Rohlf 1995, page 794–). We test the null hypothesis that roots biomass on ‘home’ soil is no different from ‘away’ soil. Under the null-hypothesis, the test-statistic, calculated as: –2 ∑ln p, (summing the ln P-values over k tests), is χ2-distributed with 2k degrees of freedom, k being number of separate tests.

Testing for differences in coefficient of variation among treatments

If a local environmental condition exerts a selective pressure on a phenotypic trait, and if this trait evolves towards a more adaptive phenotype through natural selection, then one would expect a lower phenotypic variation in the trait under the treatment representing the home–soil environment and a larger variation when plants are grown on a different soil type. In other words, if the differences in phenotype we observe between ‘home’ and ‘away’ treatments are due to local adaptation to the ‘home’ environment, we would expect the lowest phenotypic variation in this trait on ‘home’ soil. One way to test this prediction is to compare the coefficient of variation of the plant traits in the ‘home’ and ‘away’ environment. Under the scenario of local adaptation one would expect the lowest CV in the ‘home’ environment due to selection reducing variation in that trait. We test this by a binomial exact test under the null hypothesis of equal CV in ‘home’ vs. ‘other’ (away and control) environment. We do this for each species and each origin separately.

Results

vegetation analysis

Similar number of herbaceous species and estimates of diversity were found between the four study sites (Table 1). We detected a significant positive spatial association between thyme and Plantago at sites TP-1 and TS-4, and between thyme and Galium, and thyme and Achillea at site TP-2, and for thyme and Achillea also at site TS-3 (Table 2). At all the other sites, the spatial association of thyme with the associated species did not differ significantly from that of a random association based on the marginal frequency of each individual species. Agrostis was present in all the Raunkiær circles hence leaving no power to test for any deviation from random association with thyme.

Table 1.  Number of herbaceous species and diversity indices at each of the four study sites. Data are based on 10 random quadrats (40 × 40 cm) at each site
LocalityTP-1TP-2TS-3TS-4
Number of species26272618
Shannon diversity index D6.236.876.155.05
Simpson diversity index H0.120.170.160.24
Table 2.  Spatial association between thyme and each of four associated species at four natural sites
Site Species:TP-1TP-2TS-3TS-4
Obs.Exp.Obs.Exp.Obs.Exp.Obs.Exp.
  1. Sites TP-1 and TP-2 are sites where the associated species co-occurs with Thymus pulegioides, sites TS-3 and TS-4 with T. serpyllum. Observed (Obs.) and expected (Exp.) frequency of co-occurrence between thyme and associated species is based on 50 Raunkiær circles at each site, and deviation from random association was tested using a G-test of independence. Figures in bold and marked with a *refer to sites where thyme and associated species differ significantly from a random association (G > 3.84, P < 0.05).

Agrostis capillaris0.520.520.60.590.320.300.260.31
Achillea millefolium0.50.480.320.26*0.360.30*0.080.1
Plantago lanceolata0.440.38*0.260.220.140.100.540.48*
Galium verum0.240.200.320.24*0.320.270.180.21

germination, juvenile biomass, adult biomass and reproduction

A global analysis of variance including all four associated species showed a significant effect of species for germination rate, juvenile and adult biomass (Table 3). In addition, we detected significant effects of soil treatment (within species) for germination and adult biomass, and of origin (TP or TS) and the interaction with treatment for germination and juvenile biomass. Due to the strong species effect (Table 3) we present the results of the effects of origin and treatment for each species separately.

Table 3. anova for germination rate, juvenile and adult biomass of the four associated species combined
Trait/EffectsGerminationJuvenile biomassAdult biomass Aboved.f.
TotalAbove/total
  1. Note, for adult biomass, analysis was only performed with three of the four species (data on Plantago not available, see Methods for details). d.f.: degrees of freedom for germination and juvenile biomass and d.f. for adults in parentheses.
    Model R2 is given with degrees of freedom in parenthesis. Trt: Soil treatment. Origin refers to TP or TS origin. F-ratios and their significance are indicated by:P < 0.1; *P < 0.05; **P < 0.01; ***P < 0.001.

Model R20.65 (23, 245)***0.23 (23, 110)***0.46 (23, 110)***0.73 (17, 306)*** 
Trt (species)18.3***0.971.783.56**8 (6)
Origin (species)42.5***3.10*1.712.434 (3)
Trt × origin (species)2.4*0.100.680.368 (6)
Species34.6***13.75***37.0***432.8***3 (2)

In the following, ‘home’ soil refers to plants of TP-origin growing on TP-soil and to plants of TS-origin growing on TS-soil, ‘away’ soil refers to plants of TP-origin growing on TS-soil and to plants of TS-origin growing on TP-soil, and ‘control’ soil refers to plants of both TP- and TS-origin growing on soil where no terpenes were added.

overall effect of terpenes – are carvacrol and β-caryophyllene allelopathic?

All species showed at least one trait in which the performance on soil treated with terpene was significantly different from the performance on control soil.

For germination, Agrostis seeds originating from a TP (carvacrol) site tended to have a higher germination rate on soil treated with a terpene (either carvacrol or β-caryophyllene) compared to control soil (Fig. 1a), whereas the reverse was true for germination of Plantago and Galium seeds (Figs 3a and 4a).

Figure 1.

Performance of Agrostis capillaris plants on three types of soils. White bars represent sites where the species co-occur with T. pulegioides (TP-sites) and grey bars represent sites where the species co-occur with T. serpyllum (TS-sites). Plants from these sites were grown on soil onto which the dominating terpene of T. pulegiodes was added (TP soil), soil onto which the dominating terpene of T. serpyllum was added (TS soil), and soil where no terpene was added (Control). ‘Home’ refers to plants from TP-sites growing on TP soil, and to plants from TS sites growing on TS soil. ‘Away’ refers to plants from TP sites growing on TS soil and to plants from TS sites growing on TP soil. When relevant, contrasts testing within each origin (TP or TS) for the allelopathic effect of terpenes (indicated as terpene vs. control) or the presence of local adaptation (home vs. away) are added above each panel together with their corresponding P-values (see Methods for details).

Figure 3.

Performance of Plantago lanceolata plants on three types of soils. White bars represent sites where the species co-occur with T. pulegioides (TP-sites) and grey bars represent sites where the species co-occur with T. serpyllum (TS-sites). Plants from these sites were grown on soil onto which the dominating terpene of T. pulegiodes was added (TP soil), soil onto which the dominating terpene of T. serpyllum was added (TS soil), and soil where no terpene was added (Control). ‘Home’ refers to plants from TP-sites growing on TP soil, and to plants from TS sites growing on TS soil. ‘Away’ refers to plants from TP sites growing on TS soil and to plants from TS sites growing on TP soil. When relevant, contrasts testing within each origin (TP or TS) for the allelopathic effect of terpenes (indicated as terpene vs. control) or the presence of local adaptation (home vs. away) are added above each panel together with their corresponding P-values (see Methods for details).

Figure 4.

Performance of Galium verum plants on three types of soils. White bars represent sites where the species co-occur with T. pulegioides (TP-sites) and grey bars represent sites where the species co-occur with T. serpyllum (TS-sites). Plants from these sites were grown on soil onto which the dominating terpene of T. pulegiodes was added (TP soil), soil onto which the dominating terpene of T. serpyllum was added (TS soil), and soil where no terpene was added (Control). ‘Home’ refers to plants from TP-sites growing on TP soil, and to plants from TS sites growing on TS soil. ‘Away’ refers to plants from TP sites growing on TS soil and to plants from TS sites growing on TP soil. When relevant, contrasts testing within each origin (TP or TS) for the allelopathic effect of terpenes (indicated as terpene vs. control) or the presence of local adaptation (home vs. away) are added above each panel together with their corresponding P-values (see Methods for details).

The flower biomass of Achillea from TS sites, the root biomass of Plantago juveniles from TP sites, the number of spikes on Plantago adults from TP sites, and the adult above-ground biomass of Galium from both TP and TS sites were all larger on control soil compared to terpene treated soil, indicating a general overall allelopathic effect of terpenes on these traits (Figs 2e, 3b, and 4b respectively). The other traits, which showed a significant effect of treatment, were due to differences among the two terpene treatments (see below) rather than differences among control and terpene soil.

Figure 2.

Performance of Achillea millefolium plants on three types of soils. White bars represent sites where the species co-occur with T. pulegioides (TP-sites) and grey bars represent sites where the species co-occur with T. serpyllum (TS-sites). Plants from these sites were grown on soil onto which the dominating terpene of T. pulegiodes was added (TP soil), soil onto which the dominating terpene of T. serpyllum was added (TS soil), and soil where no terpene was added (Control). ‘Home’ refers to plants from TP-sites growing on TP soil, and to plants from TS sites growing on TS soil. ‘Away’ refers to plants from TP sites growing on TS soil and to plants from TS sites growing on TP soil. When relevant, contrasts testing within each origin (TP or TS) for the allelopathic effect of terpenes (indicated as terpene vs. control) or the presence of local adaptation (home vs. away) are added above each panel together with their corresponding P-values (see Methods for details).

local adaptation to specific thyme chemotype: comparing performance on home vs. away soil

Agrostis capillaris

Juveniles showed a significant treatment effect for total biomass (Ftrt(origin) = 2.9, d.f. = 4, P = 0.02) and root biomass (Ftrt(origin) = 7.1, d.f. = 4, P < 0.001). Juveniles from TP sites had a significantly higher biomass on soil treated with terpene compared to control soil. Moreover, their biomass was highest on soil treated with the ‘home’ terpene (carvacrol) compared to the ‘away’ terpene (β-caryophyllene). This difference was due to the root biomass being highest on the ‘home’ soil (Fig. 1b). No difference in the above-ground biomass among treatment was found. Juveniles from TS-sites also had the highest root biomass on carvacrol (‘away’) soil, but their root biomass did not differ between ‘home’ and ‘control’ soil (Fig. 1b). For adults, plants from TP sites had a significantly higher total biomass on ‘home’ soil compared to ‘away’ soil (total: Ftrt = 3.4, d.f. = 2, P = 0.04, Fig. 1c), and as with juveniles, this difference was due to a higher biomass of roots on ‘home’ soil (root: Ftrt = 3.3, d.f. = 2, P = 0.04, Fig. 1d).

We detected no difference in biomass among soil treatments for adult plants originating from TS-sites.

Achillea millefolium

Germination of Achillea seeds showed a significant interaction between origin and treatment (Forigin × trt = 4.31, d.f. = 4, P = 0.002) due to seeds originating from both TP and TS sites germinating better on ‘home’ soil compared to ‘away’ soil (Fig. 2a).

Juvenile root biomass showed significant effects of both treatment (Ftrt(origin) = 2.4, d.f. = 4, P = 0.04) and site (Fsite(origin)= 5.8, d.f. = 4, P = 0.003). Juveniles from TS sites had significantly higher root biomass on their ‘home’ soil (β-caryophyllene) compared to ‘away’ soil (carvacrol) (Fig. 2b). This home site advantage was detected in plants from both TS sites however the difference in root biomass between ‘home’ and ‘away’ soil was twice as high in plants from site TS-3 compared to plants from site TS-4 (data not shown).

Above-ground biomass of adult plants also showed a significant treatment effect (Ftrt(origin) = 2.6, d.f. = 4, P = 0.04) due to plants originating from TS sites having significantly lower above-ground biomass on ‘home’ soil compared to ‘away’ soil (Fig. 2c). Note that we did not obtain data on root biomass for Achillea adults.

Reproduction in Achillea was estimated as flower biomass. All Achillea plants flowered in the experiment. The mixed model anova showed no effects of origin, site (within origin) or treatment on the variation in flower biomass. However, plants from both TP and TS sites tended to have higher flower biomass on control soil compared to terpene treated soil but this effect was only significant for plants of TS origin (Fig. 2d).

Plantago lanceolata

This species was only represented from three of the four study sites, missing from site TS-3. Germination of all seeds, irrespective of origin, was lower on TP soil compared to both TS- and control soil (Ftrt(origin) = 37.51, d.f. = 4, P < 0.001). Seeds from the TS site germinated better on ‘home’ soil compared to ‘away’ soil (Fig. 3a). For juveniles, the root biomass showed a significant treatment effect (Ftrt(origin) = 6.44, d.f. = 4, P < 0.001). Plants originating from TP sites had the highest root biomass on control soil indicating an overall allelopathic effect of terpenes. Moreover, root biomass was significantly higher on ‘home’ soil compared to ‘away’ soil (Fig. 3b) – a result also found in Agrostis and Achillea, indicating that roots may show an overall general adaptive response to the ‘home’ chemotype. However, a significant site × treatment effect (Fsite × trt(origin) = 16.1, d.f. = 2, P < 0.001) and contrasts among sites showed that the increased root biomass on ‘home’ soil was only found in plants from site TP-1.

For reproduction (here estimated as number of flowering spikes), plants at TP sites produced more spikes on control soil compared to treated soils (Fig. 3c).

Galium verum

Seeds originating from both TP and TS sites germinated better on control soil compared to terpene treated soils (Ftrt(origin) = 8.67, d.f. = 4, P < 0.001, Fig. 4a).

For total biomass of juveniles we detected no overall effects of treatment, but an effect of site (Fsite(origin) = 6.02, d.f. = 2, P = 0.003) and contrasts showed that juveniles from site TP-2 had both heavier roots and above-ground biomass on ‘home’ soil compared to ‘away’ soil (contrast TP2; home vs. away, P = 0.03, data not shown).

For adult above-ground biomass, a significant effect of treatment (Ftrt(origin) = 4.04, d.f. = 4, P = 0.0045) was due to plants originating from both TP and TS sites being heavier on control soil compared to terpene treated soils (Fig. 4b).

controlling for the fdr

Before controlling for the FDR, we detected 10 significant contrasts for performances on ‘home’ vs. ‘away’ soil in total. After applying the Benjamini & Hochberg (1995) FDR method, we found that all ten tests still fall into the category of significant tests where the probability of being a false discovery (i.e. a true type I error) is less than 5%.

overall test of local adaptation – combining p-values

The presence of an overall biological effect on root biomass was confirmed by combining P-values from individual tests on ‘home’ vs. ‘away’ soil. In this overall test we combine P-values from eight contrasts (four species and two origins) and we obtain a test-statistic of 35.7. Under the null hypothesis, this test-statistic is χ2 distributed with 16 degrees of freedom yielding an overall P-value of 0.003. The null hypothesis that root biomass on ‘home’ soil is no different from ‘away’ soil is clearly rejected and we may say that, overall, root biomass is indeed higher on ‘home’ than on ‘away’ soil.

differences in coefficient of variation among treatments

For A. millefolium, we found that plants originating from both TP and TS sites had significantly more traits with a lower CV on ‘home’ soil compared to ‘other’ soils (binomial exact test; P = 0.021 for both TP and TS origin (Figs 5a and b).

Figure 5.

Coefficient of variation (CV) of all plant traits (germination, juvenile and adult biomass) on home versus away soil for Achillea millefolium of TP and TS origin (5a, b), Agrostis capillaris of TP origin (5c) and Galium verum of TP origin (5d). Home soil is TP soil for plants of TP origin (5a, c, and d) and TS soil for TS origin (5b). Away soil is either control soil (indicated by crosses) or soil treated with the ‘away’ terpene (indicated by triangles). The line represents Y = X; excess of points above the line indicate a higher CV on away soil than home soil and vice versa for excess of points below the line.

For A. capillaris, we found that plants from TP sites tended to have more traits with a lower CV on ‘home’ soil compared to control soil (binomial exact test CV-home vs. CV-control, P = 0.08, Fig. 5c). This was mainly due to CV of germination and adult traits individually being nearly significant (germination) or significantly (adult traits) lower on ‘home’ than on control soil (mixed anova, data not shown).

For P. lanceolata originating from both TP and TS sites, and for G. verum plants originating from TS sites, we found no difference in CV among ‘home’ and ‘other’ soils (binomial exact test; P > 0.2). However, in contrast to A. millefolium and A. capillaris, we found that G. verum from TP sites had more traits with a higher CV on its ‘home’ soil compared to ‘other’ soils (binomial exact test, P < 0.01, Fig. 5d).

Discussion

This study shows that the terpenes carvacrol and β-caryophyllene produced by Thymus plants had either inhibitory or stimulatory effects on the germination and growth of four associated plant species. More importantly, the associated plant species responded to the terpenes in ways that suggest local adaptation to the presence of Thymus species. We demonstrate this in three different ways. First, a reciprocal transplant experiment showed that seeds of A. millefolium from both TP- and TS-sites germinated best on ‘home’ soil. A better germination on ‘home’ soil was also found for P. lanceolata seeds from the only TS-site where it was sampled. Moreover, A. capillaris originating from TP sites, and A. millefolium plants from TS sites had a significantly higher biomass on ‘home’ soil, mainly due to increased root biomass. A higher root biomass on ‘home’ soil was also found in P. lanceolata originating from site TP-1, and G. verum from site TP-2. Second, plants originating from sites where larger roots or a better germination rate were detected on the ‘home’ soil were also the sites where vegetation analysis showed a significant positive association in the co-occurrence with thyme plants (Plantago at sites TP-1 and TS-4, Galium at site TP-2, and Achillea at site TS-3, note that Agrostis and thyme was always co-occurring in all quadrats at all sites, hence leaving no power to detect for any deviation in their spatial association). Finally, the two plant species in which the reciprocal transplants showed strongest evidence of local adaptation (A. capillaris and A. millefolium) also had the lowest CV in plant traits on ‘home’ soil, as would be expected if the variation for the traits had been reduced due to selection in the home environment.

Local adaptation is a result of natural selection. It is therefore not surprising that patterns of local adaptation detected in this study could not be demonstrated for all traits and at all sites. Beyond issues of statistical power to detect local adaptation, natural selection could be ineffective in developing local adaptation in some populations either due to lack of relevant genetic variation for a trait, or to ecological factors opposing local adaptation (ecological disturbance, migration from neighbouring populations non-adapted or adapted to a different chemotype). In this context, the fact that we do demonstrate adaptation to specific thyme chemotypes within local populations strongly supports the idea that chemical neighbours, such as thyme, may serve locally as strong selective agents capable of shaping and affecting the whole plant community.

germination is affected by terpenes

Both the monoterpene (carvacrol) and the sesquiterpene (β-caryophyllene) significantly altered the germination of associated species but their effects varied from inhibition to stimulation, depending on the species and origin of plants.

The inhibitory effects of monoterpenes and sesquiterpenes on germination have been documented for a range of different plant species (e.g. Aspelund 1968; Kelsey & Locken 1987; Tarayre et al. 1995; Abrahim et al. 2000; and see Langenheim 1994 for a review). Most studies have examined the effects of terpenes on germination in Petri dishes and using concentrations which may well exceed those occurring in natural soil conditions (see e.g. Inderjit & Weston 2000). Ehlers & Thompson (2004), using soil collected directly underneath the canopy of T. vulgaris plants from natural sites, found that seeds of B. erectus originating from a site where thyme plant produce a non-phenolic monoterpene germinated significantly better on their ‘home’ non-phenolic soil than on soil from sites where thyme produced either a phenolic or another type of non-phenolic monoterpene. The ‘home’-site effect was only detected on the soil collected directly underneath the thyme plants and not on soil collected from the very same sites but away from thyme plants, strongly indicating that B. erectus has adapted to the presence of its local thyme chemotype. Our results corroborate these findings because Achillea seeds germinated better on TS- and TP-soil when the seeds originated from a TS and TP site respectively.

biomass of associated species

Depending on species and origin, an increased root biomass in response to the ‘home’-terpene treatment was detected in juvenile and adult plants. For Agrostis from TP sites, both juvenile and adult plants invested significantly more in root biomass when growing on ‘home’ soil compared to ‘away’ soil. For Achillea, juveniles from TS sites also had significantly heavier roots on ‘home’ soil compared to ‘away’ soil. However, here the above-ground biomass of adults was lower on ‘home’ soil compared to other soils. Unfortunately, we do not have estimates of the root biomass of these adults. We thus cannot say if the decreased above-ground biomass of plants from TS origin growing on ‘home’ soil is due to a compensation caused by increased root growth (as may be indicated from the results of the juveniles) or if total biomass is actually reduced in this treatment. Increased investment in root biomass on the ‘home’ soil was also detected in both Plantago and Galium juveniles from sites TP-1 and TP-2 respectively.

Overall, plants from TP-sites most often showed evidence for local adaptation to ‘home’ soil. The concordance between spatial distribution of thyme plants and associated species at natural TP-sites and the response to our TP treatment suggests that the increased investment in roots may indeed be an adaptation to growing on soil where carvacrol is a common component of the soil. Except for Achillea, we did not find a similar clear response between ‘home’ and ‘away’ soil for plants originating from TS-sites. This may be due to the fact that the oil composition of T. serpyllum plants is typically dominated by 2–4 different terpenes (Stahl-Biskup 2002; Grøndahl, Keefover-Ring and Ehlers unpublished data). Although all thyme plants from the TS-sites had β-caryophyllene as one of the dominant components of the oil, the oil also contained other terpenes. Any response to growing with a T. serpyllym plant may be affected by the action of these other terpenes as well – an effect we did not control for in our experiment. In contrast, the oil content of T. pulegioides plants are clearly dominated by a single monoterpene, carvacrol (Stahl-Biskup 2002; Grøndahl et al. 2008).

increased investment in root biomass and implications for the plant community

Previous studies on effects of both mono- and sesquiterpene on root growth report that terpenes actually decrease the growth of roots and/or damage the root cells (e.g. Muller & Muller 1964; Lorber & Muller 1976; Kelsey & Locken 1987; Abrahim et al. 2000). This is not consistent with our results. However, most studies on the effects of terpenes on root growth were performed using Petri-dishes or plastic containers to observe a plant's response to terpene (either as volatile or liquid). In our study, we measured the effect of terpene on plants by growing them on soils into which terpenes were added. Moreover, we used a lower concentration of terpene than in the above studies. These differences in the experimental set-up may explain why we find a very different effect of terpenes on the growth of plant roots.

We did not find evidence that increased investment in roots caused a significant reduction in the above-ground vegetative biomass. However, we did find some evidence that increased root investment occurred at the expense of reproduction. When root biomass was largest on home soil, the reproductive investment in both Plantago (mean number of spikes per plant) and Achillea (flower biomass) tended to be larger on control soil.

An increased investment in root biomass is a well known response to nitrate limitation (e.g. Mihailiak & Lincoln 1985; Robinson 1994; Ericson 1995). In a review, comprising 129 species, Reynolds & D’Antonio (1996) found that plants had a higher ratio of root weight to total plant weight at low nitrogen availability compared to high nitrogen availability. An increased allocation to root biomass under nitrate limiting conditions may be a general response across plant species. The combined effect of variation in nutrient availability and interaction with other plants may however create a root growth response that differs from the response found when the effects (competition vs. nutrient availability) are studied independently (e.g. Hodge et al. 1999; Schenk 2006). Plants having root neighbours increased the biomass of their roots even though nutrients were not limited, suggesting a response to the root neighbour itself rather than resource depletion. Roots may detect other roots before physical contact via a mechanism that involves allelochemicals (e.g. Bais et al. 2003; and see review by Schenk 2006), possibly sensing the root neighbour as a cue of future competition before direct contact (Murphy & Dudley 2007). In this context, soil organisms and allelochemicals may play a very important role in root interaction between species as the composition and abundance of soil organisms can both increase and decrease nutrient availability to competing roots.

It is well known that terpenes, especially phenolic ones, alter the nitrogen cycle in soils by inhibiting the nitrification process and thereby decreasing the amount of nitrate (White 1986; Pavolainen et al. 1998; Hättenschwiler & Vitousek 2000; Souto et al. 2000; Castells et al. 2003). Associated plant species may adapt to low nitrate availability by increasing their root biomass as indicated in the present study. The inhibition of nitrification by terpenes is due to both direct effects on the physiology of nitrifying bacteria and indirect effects due to immobilization of mineral N (e.g. White 1991; Pavolainen et al. 1998). Soil suspensions exposed to monoterpenes increased production of ammonium and decreased that of nitrate (Pavolainen et al. 1998). A lower availability of nitrate in the neighbourhood of terpene-producing plants may exert negative effects on the establishment and growth of other plant species that prefer nitrate as nitrogen source, and may favour species that prefer ammonium as the major nitrogen source. The terpenes may thus indirectly alter the competitive interactions between plant species in the community. Terpenes also affect the growth and reproduction of soil fungi (e.g. Vokou et al. 1984; Rahman & Gul 2003). Callaway et al. (2003) showed that the abundance level of arbuscular mycorrhizal fungi significantly changed the rank of competing plant species. Similar alterations of competitive interactions may also be at play in plant communities dominated by thyme plants.

As the effects of terpenes are very local, affecting mainly the soil just beneath thyme plants (e.g. Ehlers & Thompson 2004), this may enhance the spatial heterogeneity and could potentially increase biodiversity of the plant community. Such small scale effects of terpenes on plant species richness were demonstrated by Iason et al. (2005) who found that the chemical diversity of terpenes in individual pine trees were correlated with the species richness of woody plants found directly beneath the pine trees.

The present study shows that associated plant species may locally adapt to the presence of another plant species and its specific biochemistry. This may involve adaptation to both direct effects of the thyme allelochemicals on, for example, seed germination, and also indirectly to an altered nutrient availability due to the allelochemicals effect on soil microorgansims. The study supports a recent view (e.g. Lortie et al. 2004; Schenk 2006) that local plant communities may in fact be much more co-evolved than was previously believed.

Acknowledgements

Authors thank J. Christiansen, Fahrudin Zec and all the staff at Påskehøjgård for their help in the glasshouse, and A. Sølling for her help in assessing biomass. Authors are grateful to R. M. Callaway, R. Ejrnæs, K. Keefover-Ring, and J.M. Olesen for discussion and comments on the manuscript, and to T. Bataillon for statistical advice. Two anonymous referees also made valuable comments to the manuscript. The study was financed by a grant from the Danish National Research Foundation (FNU) to BKE.

Ancillary