Scientia Terrae Research Institute, B-2860 Sint-Katelijne-Waver, Belgium
Laboratory for Process Microbial Ecology and Bioinspirational Management, Lessius University College, Campus De Nayer, Consortium for Industrial Microbiology and Biotechnology (CIMB), Department of Microbial and Molecular Systems, KULeuven Association, B-2860 Sint-Katelijne-Waver, Belgium
1. Understanding the many factors that affect community composition and coexistence of species in natural environments is one of the central goals in ecology. As germination and establishment of seedlings in orchids are crucially dependent on mycorrhizal availability, the diversity and spatial distribution of orchid mycorrhizal fungi are likely factors that contribute to orchid coexistence.
2. In this study, we combined molecular identification techniques with seed germination experiments and spatial point pattern analyses to investigate to what extent differences in mycorrhizal association patterns affected spatial variation in seed germination and the above-ground distribution of three co-occurring terrestrial orchid species (Anacamptis morio, Gymnadenia conopsea and Orchis mascula).
3. The three species associated with a different set of mycorrhizal fungi, except A. morio and G. conopsea, which shared one fungal associate. The number of fungal lineages detected associating with each species also differed between species, being highest in G. conopsea, which associated with five lineages, and restricted to a single lineage in O. mascula.
4. Seed germination experiments showed that below-ground seed germination was restricted to locations where orchid individuals were most abundant, and quickly declined with increasing distance from the nearest above-ground congener. Spatial point pattern analyses indicated significant fine-scale spatial clustering that was highest in O. mascula and lowest in G. conopsea. Moreover, the spatial distributions of the three species were independent from each other, except for A. morio and G. conopsea.
5.Synthesis. Our results support previous findings that co-occurring orchid species tend to use different mycorrhizal partners. Although the possibility that variation in local environmental conditions affected seed germination could not be completely ruled out, our results suggest that the presence of specific mycorrhizal fungi contributed, at least partly, to the spatial distribution and coexistence of the investigated orchid species.
Understanding the many factors that affect community composition and coexistence of species in natural environments is one of the central goals in ecology. In orchids, germination and establishment of seedlings are crucially dependent on orchid mycorrhiza (Bidartondo & Read 2008; Smith & Read 2008; Rasmussen & Rasmussen 2009; Dearnaley, Martos & Selosse 2012); therefore, the availability of suitable mycorrhizal fungi may explain the above-ground abundance and spatial distribution of terrestrial orchid species (Diez 2007; Jacquemyn et al. 2007; McCormick et al. 2009, 2012). Additionally, direct interspecific encounters, and therefore competition among orchid species, may be reduced if different orchid species associate with different mycorrhizal fungi, which in turn can contribute to the coexistence of the orchid community.
It is generally assumed that the distribution of orchid mycorrhizal fungi is independent from orchid distribution, because orchid fungi are not obligately associated with orchids (Dearnaley, Martos & Selosse 2012; McCormick et al. 2012). However, seed germination experiments in the field, using seed packages, have shown that in several orchid species, seed germination often significantly declines with increasing distance from existing individuals (McKendrick, Leake & Read 2000; Batty et al. 2001; McKendrick et al. 2002; Diez 2007), which suggests that the local availability and abundance of suitable mycorrhizal fungi decline with increasing distance from parent plants. In cases where several orchid species co-occur at a single site, patchy distributions of mycorrhizal fungi combined with high fungal diversity and different preferences for mycorrhizal fungi are the most likely factors contributing to orchid coexistence, as they may reduce competition for nutrients and promote seed germination and seedling establishment (Waterman et al. 2011). Although compelling evidence to support this hypothesis is still limited, recent analyses have shown that co-occurring species of South-African Coryciinae orchids were less likely to share mycorrhizal fungi than would be expected under random community assembly conditions (Waterman et al. 2011), which the authors interpreted as evidence for mycorrhizal fungi being involved in determining coexistence of orchid communities.
In this study, we investigated below-ground seed germination and above-ground distribution patterns of three co-occurring orchid species (Anacamptis morio, Gymnadenia conopsea and Orchis mascula) that show strong differences in mycorrhizal specificity. Gymnadenia conopsea is a generalist towards its mycorrhizal fungi. Sequencing of the internal transcribed spacer (ITS) region revealed a high taxonomic and ecological diversity of fungal associates, including typical orchid mycorrhizas of the Tulasnellaceae and Ceratobasidiaceae as well as several ectomycorrhizal taxa of the Pezizales (Stark, Babik & Durka 2009). Within populations, a high number of fungal associates was also observed, varying between 4 up to 15 different fungal operational taxonomic units (OTUs), often related to different fungal families. Orchis mascula, on the other hand, was shown to be a specialist towards mycorrhizal fungi. Investigating over 30 individuals distributed across a number of populations in Europe, Jacquemyn et al. (2010, 2011) recently concluded that this species almost exclusively associated with a single fungal symbiont related to the Tulasnellaceae. Finally, preliminary research on the fungal associates of A. morio showed that this species associates with a large number of different mycorrhizal fungi, but that within populations, the number of associates is restricted to one or a few mycorrhizal OTUs (H. Jacquemyn, unpubl. results).
Given this strong interspecific variation in mycorrhizal associations, we hypothesized that fungal partners are involved in mediating the coexistence of these three species. More specifically, we first investigated mycorrhizal association patterns in the three study species using molecular detection and identification techniques. We predict that the three co-occurring species share no or few mycorrhizal fungi. We further predict that the probability of successful seed germination is highest near parent plants. To test this hypothesis, a seed germination experiment was used to investigate spatial variation in seed germination and to define suitable areas for seed germination within the study site. These habitat suitability maps were finally used to test the hypothesis that the spatial distribution pattern of each of the three species was independent from one another.
Material and methods
Study Site and Mapping of Orchids
The study was conducted in a site where the three species were found growing together (50°6′23.78″N; 4°43′6.02″E) (Fig. S1 in Supporting Information). This site consists of a calcareous grassland with scattered birch (Betula pendula) and oak (Quercus robur) trees. The grassland contains a large diversity of grass and forb species, including Anthyllis vulneraria, Carlina vulgaris, Cirsium acaule, Gentiana germanica, Helianthemum nummulariuum, Linum catharticum, Polygala comosa, Primula veris, Sanguisorba minor and others. The most abundant grass species was Brachypodium pinnatum, but its abundance was never very high. Unlike many other calcareous grasslands, this grassland had no particularly strong slope. Overall, the soil is very shallow (<5 cm), with no apparent topographical variation within the study site.
In Spring 2009, roots of eight randomly selected plants per species were collected within the study plot to determine patterns of mycorrhizal associations. Roots were surface sterilized (30-s submergence in 1% sodiumhypochlorite, followed by three 30-s rinse steps in sterile distilled water) and microscopically checked for mycorrhizal colonization. Subsequently, DNA was extracted from 0.5 g mycorrhizal root fragments using the UltraClean Plant DNA Isolation Kit as described by the manufacturer (Mo Bio Laboratories Inc., Solana Beach, CA, USA) and 10 times diluted afterwards.
The mycorrhizal community associated with the roots was assessed using a previously developed ITS-based DNA array, enabling the simultaneous detection and identification of 23 OTUs associated with multiple species of the genera Orchis, Anacamptis and Gymnadenia, including O. mascula, A. morio and G. conopsea (Jacquemyn et al. 2011). Amongst some others, the majority of fungi that could be detected belonged to the family of Tulasnellaceae (12 OTUs). OTUs were defined based on an ITS sequence similarity of at least 97% (Lievens et al. 2010; Jacquemyn et al. 2011). Additional cloning and sequencing of fungal ITS sequences of three additional individuals per species sampled at the study site did not result in new fungi, demonstrating the robustness of the array. For DNA array analysis, fungal ITS regions were amplified using ITS1-OF and ITS4-OF (Taylor & McCormick, 2008; Lievens et al. 2010; Jacquemyn et al. 2011) and simultaneously labelled with alkaline-labile digoxigenin (0.15 mmol L−1 digoxigenin-11-dUTP mix; Roche Diagnostics GmbH, Mannheim, Germany). These primers are broad-spectrum basidiomycete-specific primers that are increasingly used to characterize fungal symbionts of orchids (Taylor and McCormick 2008; Shefferson, Kull & Tali, 2008). DNA samples were amplified according to the following thermal cycling profile: initial denaturation at 94 °C for 2 min, followed by 35 cycles of 45 s at 94 °C, 45 s at 58 °C and 45 s at 72 C, with a final elongation step at 72 °C for 10 min. The resulting labelled amplicons were subsequently hybridized to the DNA array as previously described (Lievens et al. 2010; Jacquemyn et al. 2011). All hybridizations were performed twice to check for consistency of the results.
To investigate spatial variation in seed germination, fruits were harvested when seeds were ripe (mid-August 2009), and seeds were then buried using the modified seed package method of Rasmussen & Whigham (1993). Per seed package, approximately 250 seeds were placed within a square of 53-μm mesh phytoplankton netting, enclosed within a Polaroid slide mount. Packages were buried into the soil along six transects within the 40 × 70 m study plot. Transects were 10 m apart and consisted of six sample points per transect that were 5 m apart. At each point, three seed packets (one for each species) were placed vertically in the ground, leading to a total of 142 seed packages that were left in the ground for about 2 years. In April 2011, seed packages were retrieved, gently washed and maintained moist in paper towel for 1 day until examination. Packages were first rinsed with tap water, opened with a mini-knife and rigorously checked under a dissecting microscope for germination. As orchid seed germination stages can be variable (Ramsay, Sivasithamparam & Dixon 1986), germination was considered to have occurred when clear signs of mycorrhiza formation were present and the leaf primordia had developed (stage 3 sensuRamsay, Sivasithamparam & Dixon 1986) (see Fig. S2). For each package, the successful development of protocorms was determined by counting the number of protocorms.
Spatial variation in below-ground seed germination
To investigate whether the probability of germination was affected by the distance to the nearest congener, for each seed package, the distance to the nearest individual was calculated based on the geographic positions of both plants and seed packages. Additionally, we also calculated the distance to the nearest individual for each seed package, regardless of species. Presence–absence of protocorms in seed packages was then related to these distances using generalized linear mixed models using the glmer() function from the LME4 package in R (Bates, Maechler & Dai 2009). In the case of overdispersion, a categorical variable with a different level for each location as a random effect was added to the model (Warton & Hui 2011). Both the distance to the nearest conspecific individual and the distance to the nearest individual were included in these models. However, because inclusion of the distance to the nearest individual did not reveal any significant effects, this variable was not investigated further. To investigate whether seed germination differed between species, species was included in the model as a fixed, categorical variable. We also investigated whether the number of protocorms was affected by the distance to the nearest congener. In this case, a Poisson regression with log link was used. Again, species was included as a fixed categorical factor in the analyses.
Above-ground spatial distribution
Our hypotheses require two groups of spatial analyses. In a first group, we study the univariate spatial pattern of the three orchid species, and in the second, group we study the bivariate association pattern of pairs of orchid species. We used spatial point pattern analysis techniques (Wiegand & Moloney 2004; Perry, Miller & Enright 2006; Law et al. 2009). Assessment of the relative degree of clustering of the univariate spatial patterns of the three orchid species is not straightforward because it may be influenced by the distribution of the mycorrhizal fungi. We, therefore, used mark connection functions (Getzin et al. 2008; Illian et al. 2008) that allow us to quantify for a given distance r the contribution of each species to the overall clustering of all three orchids. We expect that the species with the highest mycorrhizal specificity should contribute most to the overall clustering, whereas the species with the lowest mycorrhizal specificity should contribute least.
The plot of the species distribution pattern (Fig. S1) suggests that the three species are strongly clustered and that their small-scale distributions show, except for the G. conopsea–A. morio pair, little overlap. We can thus hypothesize that different mycorrhizal associations or different environmental conditions favour spatial segregation of the three orchid species and avoid direct interspecific competition. An interesting question is, therefore, what happens if two individuals of different species are close to each other. Do they show on average repulsion or attraction, or are their patterns independently superposed, conditionally on the template set by the mycorrhizal fungi and/or prevailing environmental conditions? To explore this question, we tested for independence of pairs of species, conditioning on both, the univariate spatial structure of each species and a coarse approximation of the effect of the mycorrhizal fungi on the univariate distribution patterns. To this end, we used a specific null model that generates null distribution patterns of the second species of a given pair (the locations of the individuals of the first species were maintained fixed) that could potentially occur, given the germination probability of the second species and its observed characteristics of smaller-scale spatial structure. This approach assumes that the germination probability is responsible for the larger-scale distribution pattern, whereas other processes of population dynamics shape the smaller-scale spatial structures of the distribution pattern.
To quantify the univariate spatial patterns, we used the neighbourhood density function O(r) (also called O-ring statistic; Wiegand & Moloney 2004) that describes the mean density of plants of a given species at distance r from the plants of the species. The neighbourhood density function O(r) is closely related to the pair correlation function g(r), that is, O(r) = λg(r), where λ is the density of plants in the study area (=the number of plants divided by the area of the study plot). We used mark connection functions (Getzin et al. 2008; Illian et al. 2008) to quantify the contribution of each species to the clustering of all three orchid species at a given distance r. Mark connection functions pij(r) represent the conditional probability that for a pair of orchids picked at random, which are separated by a distance r, the first orchid belongs to species i and the second orchid to species j (Getzin et al. 2008; Illian et al. 2008). If species i shows the same degree of clustering than the joined pattern of all species, we find pii(r) = pipi, where pi is the proportion of individuals of species i among all individuals. Note that the mark connection functions are related to the (bivariate) neighbourhood density functions Oij(r) and (bivariate) pair correlation functions gij(r):
where O(r) is the neighbourhood density function of the joined pattern of all three species, and the bivariate Oij(r) is the density of plants of species j at distance r of plants of species i. Equation 1 shows that the mark connection functions pii(r) describe the relative clustering of species i within the joined pattern of all species and tell us how much a given species i contributes at a given distance r to the clustering of the joined distribution pattern of all three species. If species i is at distance r more clustered than the joined pattern, we find that pii(r) > pipi. Conversely, if species i is less clustered than the joined pattern, we find that pii(r) < pipi. Because the mark connection functions are conditional probabilities, they add up to one, that is, p11 + p12 + p13 + p21 + p22 + p23 + p31 + p32 + p33 = 1. This allows us also to make statements about the overlap of different species. If, for example, all species are individually clumped and segregated from each other, we will find at smaller distance r only small values of pij(r) for i ≠ j (because the probability that plants of species i and j are close to each other is low), but the pii(r)’s will sum up to almost one.
For the second analysis, we used bivariate pair correlation functions and nearest neighbour distribution functions. The bivariate pair correlation function gij(r) (Stoyan & Stoyan 1994; Wiegand & Moloney 2004; Illian et al. 2008) is related to the bivariate neighbourhood density function [i.e. Oij(r) = λjgij(r)]. This yields the density of type j points at distance r from type i points where λj is the density of type j points in the study area (= the number of plants of pattern j divided by the area of the study plot). Important additional information is provided by the distribution function (r) of the distances r from species i plants to their kth species j neighbour (Illian et al. 2008). Nearest neighbour statistics are ‘short-sighted’ and sense only the immediate neighbourhood of the points, which makes them especially sensitive to local cluster structures.
Null model for independence
In the second analysis, we tested for potential interactions among pairs of orchid species caused by direct interspecific interactions such as competition or facilitation (i.e. second-order effects) by contrasting the observed bivariate patterns to the null model of independence. A test of independence must be conditional on the spatial structure of the two univariate component patterns, but any relationship between the component patterns needs to be removed (Dixon 2002; Goreaud & Pélissier 2003; Jacquemyn et al. 2011). However, this is complicated in our case by the potential dependence of the distribution of orchids on their associated mycorrhizal fungi. This dependence is a so-called first-order effect because it determines the probability of presence of an orchid at a given location. To reveal the ‘pure’ second-order (interaction) effect, we, therefore, needed to condition additionally on the first-order effect. We used for this purpose the probability of germination of a given species (see section ‘Spatial variation in below-ground seed germination’).
Our null model kept the plants of the first (focal species) fixed and randomized the locations of the second species in a way that it showed approximately the same spatial structure as the observed univariate patterns and approximated the larger-scale distribution pattern imprinted by the mycorrhizal fungi. We accomplished this by extending current methods of pattern reconstruction (Tscheschel & Stoyan 2006; Illian et al. 2008; p407ff; Jacquemyn et al. 2012) to allow conditioning on an intensity function (Wiegand, He & Hubbell in press), which is given in our case by the probability of germination. Basically, an annealing algorithm is used to minimize the differences in several summary statistics between the observed and the reconstructed pattern (in our case up to distances of 16 m). The extended algorithm starts with a pattern generated by a heterogeneous Poisson process (Wiegand & Moloney 2004) that has the same number of points as the original pattern. In this point process, tentative points are randomly and independently distributed but only accepted with the probability of germination. In each simulation step of the annealing algorithm, a randomly selected point is tentatively removed and a new point following the above rules of the heterogeneous Poisson process is proposed instead (Wiegand, He & Hubbell in press). This new point is accepted if the mean deviance between the observed and the simulated summary characteristics decreased (Tscheschel & Stoyan 2006; Illian et al. 2008); otherwise, it is removed and another new point is selected.
We used the pair correlation function, the K-function (Ripley 1981), the distribution functions of the distances to the kth neighbour (Diggle 2003; k = 1, 2, 4, 6, 8, 12, 16, 20, 25, 30) and the spherical contact distribution (Illian et al. 2008) for pattern reconstruction. The spherical contact distribution Hs(r) measures basically the gaps in the pattern (Diggle 2003; Illian et al. 2008). Because these summary characteristics capture different aspects of spatial structure (i.e. neighbourhood density, nearest neighbours and gaps), the reconstructed patterns approximate the small to intermediate scale (i.e. up to 16 m) spatial structure of the observed patterns well (see Fig. S3).
To assess fit of the independence null model, we generated 199 simulated data sets and used the 5th lowest and highest values of our test statistics [i.e. gij(r) or Dij(r)] at distance r as simulation envelopes to depict the range of possible values under the point process model. The simulation envelopes provide approximately 5% intervals but are prone to type I error (Diggle 2003; Loosmore & Ford 2006; Illian et al. 2008). To assess the overall fit of the independence null model, we, therefore, used a goodness-of-fit (GoF) test proposed by Diggle (2003) and Loosmore & Ford (2006). This test reduces the distance-dependent information of the summary statistics for the observed data (k = 0) and the simulated data (k = 1, …, 199) into one single test statistics uk and calculates the rank of the observed uk (k = 0) within all uk. If the rank of u0 is larger than 190, the data show a departure from the null model with a 5% error rate.
By DNA array analysis, the detected fungi associated with the 22 sampled individuals could be grouped into eight OTUs. Six OTUs were affiliated to members of the Tulasnellaceae family (OTU4, OTU7, OTU10, OTU12, OTU15 and OTU18), one was associated with members of the Ceratobasidiaceae (OTU11), and one with members of the Cortinariaceae (OTU9) (Table 1). All OTUs exhibited low levels of sequence similarity between each other (e.g. for the Tulasnellaceae varying between 72% and 87%) (Jacquemyn et al. 2011), indicating that the three orchid species associated with distinct mycorrhizal fungi.
Table 1. List of observed fungal operational taxonomic units (OTUs; defined based on a 97% internal transcribed spacer (ITS) sequence similarity cut-off value) and their frequency in eight, six and eight individuals of Anacamptis morio, Gymnadenia conopsea and Orchis mascula, respectively, that were sampled at the study site
There were marked differences in mycorrhizal association patterns between the three orchid species at the study site (Table 1). Whereas O. mascula exclusively associated with a single fungal OTU (OTU10), five different OTUs were found in G. conopsea (OTU4, OTU9, OTU12, OTU15 and OTU18). In A. morio, three different fungal lineages were observed at the study site (OTU7, OTU11 and OTU18). Interestingly, each species associated with a unique set of fungal OTUs, except for OTU18 that was observed in A. morio and G. conopsea.
Spatial Variation in below-Ground Seed Germination
Germination success (i.e. the proportion of seed packages that contained at least one protocorm) differed markedly between the three species. In G. conopsea, 56% of seed packages contained a protocorm, whereas in A. morio and O. mascula, protocorms were observed in 23% of the seed packages. In seed packages in which seed germination had occurred, the number of protocorms varied between 1 and 11 (mean: 2.86) in G. conopsea, whereas in A. morio and O. mascula, the number of protocorms found in seed packages varied between 1 and 12 (mean: 3.27) and 1 and 5 (mean: 2.00), respectively.
The probability of seed germination declined significantly with increasing distance from the nearest congener in all three species (Fig. 1), but because of the overall higher germination success in G. conopsea, the intercept of the slope between distance and logit(Seed germination) was significantly higher in this species compared with that in A. morio and O. mascula. At a distance of 5 m from the nearest congener, seed packages of A. morio and O. mascula showed a 25% and 20% probability of containing at least one protocorm, respectively, but this value decreased to 0.08% and 0.03% when located at a distance of 10 m from the nearest congener. In G. conopsea, on the other hand, the chance of finding at least one protocorm within a seed package when buried at distances of 5 and 10 m from the nearest congener was 50% and 11%. Similarly, for all three, species protocorm density (i.e. the number of protocorms found within a seed package) significantly declined with increasing distance from the nearest congener (Fig. 2).
Univariate above-Ground Spatial Distribution
Orchis mascula contributed at small distances (i.e. r <120 cm) most to the overall clustering of all three orchid species. When randomly selecting two neighbouring plants (i.e. distance r < 120 cm), the probability that both will be O. mascula plants was larger than 40% (Fig. 3 top). However, when picking any two plants, the probability to obtain two O. mascula plants yields only pOmpOm = 0.15. Gymnadenia conopsea showed at small distances the second most important contribution to overall clustering (pGcpGc = 0.12), whereas A. morio was only slightly more clustered than the overall pattern (pAmpAm = 0.072). Interestingly, the mark connection function of A. morio showed relatively little response with distance, whereas that of O. mascula and G. conopsea showed strong distance responses (Fig. 3 top).
Orchis mascula also contributed most to the overall clustering at intermediate distances (120 cm < r <380 cm), with A. morio ranked second, and G. conopsea contributing least. At distances >380 cm, the contribution to clustering of all three species was similar. The ranking at small and intermediate distances is also evident from the neighbourhood density functions Oij(r). Orchis mascula had the highest neighbourhood density among all three species, whereas G. conopsea showed the lowest neighbourhood density at intermediate scales (Fig. 3 bottom).
Figure S1 suggests that the distribution ranges of the three orchid species showed little overlap, except for G. conopsea and A. morio in the central-right part of the study plot. Indeed, at small distances r, approximately 5% of all pairs were between the two species G. conopsea and A. morio (Fig. 3 top; open triangles). However, most pairs of individuals at small distances were conspecifics, as indicated by the large values of the sum of the three univariate mark connection functions for distances below 1.2 m (i.e. pOmOm + pGcGc + pAmAm; Fig. 3 top, open squares). Summing up all four mark connection functions (i.e. pOmOm + pGcGc + pAmAm + 2pGcAm) yields for all distances below 5m values larger than 0.9, which indicates that the overlap of the pairs O. mascula–G. conopsea and O. mascula–A. morio was very low (Fig. 3 top, closed squares).
Bivariate above-ground spatial distribution
Figure S3D–F show the spatial pattern of the probability of germination used to constrain the reconstructed patterns. The reconstructed patterns show the same characteristic as the observed patterns, but the characteristic small-scale structures are randomly displaced within the template set by the probability of germination (Fig. S3). The GoF test for the 0–5 m distance interval conducted for both the pair correlation function and the nearest neighbour distribution function indicates that all pairs of species, except the pair G. conopsea–A. morio, showed independent spatial patterns (i.e. rank ≤ 190). If plants of G. conopsea were the focal species and that of A. morio the second species, the pair correlation function indicated a weak attraction at distances below 0.5 m (Fig. 4), but the GoF test with the distribution function of the distances to the nearest neighbour was not significant although the function was slightly outside the simulation envelope (Fig. 4). Thus, besides showing some overlap (indicated by the expectation of the pair correlation function under the null model clearly larger than one), plants of G. conopsea and A. morio showed some additional attraction at distances up to 0.5 m.
Significant associations between species may be common in plants and range from significant overlap in distribution area to complete spatial segregation (Wiegand, Gunatilleke & Gunatilleke 2007). These patterns may arise from a range of processes, including localized seed dispersal, intraspecific and interspecific interactions, and local heterogeneity in growth conditions (e.g. topography, resource supply or background vegetation). Because mycorrhizal fungi are required for orchid seed germination and/or protocorm development, differences in mycorrhizal association patterns can, therefore, be expected to have profound effects on below-ground seed germination and above-ground distribution patterns and may lead to strong spatial clustering and spatial segregation of orchids.
Consistent with previous research (Stark, Babik & Durka 2009), we found that G. conopsea associated with the highest number of mycorrhizal fungi. In our study, five different fungal OTUs, belonging to the Tulasnellaceae (OTU4, OTU12, OTU15 and OTU18) and Cortinariaceae (OTU9), were recovered from G. conopsea. These fungal families were also the most common fungal associates of G. conopsea in German populations (Stark, Babik & Durka 2009). Our results were also consistent with previous findings of Jacquemyn et al. (2010, 2011), who found only one fungal OTU in O. mascula. Anacamptis morio, on the other hand, associated with three different fungal OTUs, two belonging to the Tulasnellaceae (OTU7 and OTU18) and one to the Ceratobasidiaceae (OTU11). These results confirm earlier findings of Girlanda et al. (2011), who showed broad specificity towards mycorrhizal fungi in the closely related Anacamptis laxiflora. Investigating mycorrhizal associations in Southern Italy, they found that A.laxiflora associated with four different OTUs, all of them belonging to the Tulasnellaceae and Ceratobasidiaceae. Interestingly, the three species did not share fungal OTUs, except in the case of A. morio and G. conopsea, which shared one fungal OTU, confirming previous findings that co-occurring orchids tend to use different fungal partners (Waterman et al. 2011).
Below-Ground Seed Germination
Given the different mycorrhizal partners and differences in mycorrhizal specificity between the three study species, we expected to find differences in seed germination patterns. Consistent with our hypotheses, the probability of successful germination was highest in the species that associated with the most fungal partners (56% of the seed packages contained at least one protocorm), whereas the species with high mycorrhizal specificity (O. mascula and A. morio) showed the lowest seed germination success (23% of the seed packages contained at least one protocorm). However, regardless of the number of fungal associates, seed germination was in all cases significantly related to the distance to the nearest congener. These results fit to previous seed germination experiments that showed that seed germination probability is inversely related to the distance from the nearest congener (McKendrick, Leake & Read 2000; Batty et al. 2001; McKendrick et al. 2002; Diez 2007), suggesting patchy distribution patterns of mycorrhizal fungi in the soil. However, germination experiments using seed packages do not allow separating effects of mycorrhizal fungi and environmental conditions (McCormick et al. 2012), and it might as well be that environmental conditions further away from adults are not sufficient to promote seed germination. Using a hierarchical Bayesian statistical framework, Diez (2007), for example, showed that at distances larger than 1 m from adult plants, germination success increased with higher levels of soil moisture, higher organic content and lower pH. However, similar to the results presented here, he also found highest germination success in seed packages in places where adult densities were highest, suggesting that mycorrhizal abundance is highest where adults are present. Similarly, McCormick et al. (2012) showed that locations with high abundance of mycorrhizal fungi were more likely to support seed germination of three woodland orchids. They also showed that the fungi that associated with the different orchids were differently affected by environmental conditions, suggesting that germination patterns of orchids may be indirectly affected by environmental conditions by influencing the distribution and abundance of the fungi. Although no apparent topographical variation was observed at our study site, it might be that small differences in edaphic conditions have contributed to the observed patterns of seed germination. Whereas O. mascula and G. conopsea generally occurred in relatively dry conditions, plants of A. morio were found in slightly moister growth conditions. Future research should, therefore, isolate fungal DNA directly from the soil, which will allow creating fungal distribution maps, and relate orchid abundance to the distribution of mycorrhizal fungi.
Above-Ground Spatial Distribution
If seed germination is spatially restricted to areas where the abundance of suitable mycorrhizal taxa is highest, and if different orchid species associate with different sets of mycorrhizal fungi, significant spatial structure is expected. Previous studies have already indicated that patchy distributions are common in orchids. Pierce et al. (2006), for example, compared the spatial distribution of a large set of orchid species in the Monte Barro natural park (Lecco, Lombardy, Italy). Using an index of dispersion (I), they found that most species showed strong and highly significant aggregation patterns. However, few studies have investigated spatial distribution of one species relative to that of another one.
Using advanced spatial point pattern analyses, we have demonstrated that all three orchid species showed strong small-scale clustering, and as a result, picking two nearby orchids with a given small distance (i.e. r < 120 cm) at random results in most cases (i.e. ≈ 95%) in conspecific pairs. For the heterospecific species pair G. conopsea–A. morio, which showed the largest interspecific overlap in the study plot, there was only a 5% probability to find such heterospecific pair within the immediate plant neighbourhood (distances r <120 cm). Thus, the three orchid species managed to avoid each other almost completely. When two individuals came accidentally close to each other, we could not detect a significant signal of species interactions except for a weak positive association of the G. conopsea–A. morio pair at distances below 0.5 m. At intermediate distances outside the direct plant neighbourhood (i.e. between 1.2 and 3.8 m), we found that the contribution of the three species to the overall clustering of the joined pattern showed the same ranking as the strength of the mycorrhizal association. The specialist species O. mascula showed by far the strongest relative clustering and the highest neighbourhood density, the generalist species G. conopsea showed the smallest degree of relative clustering and neighbourhood density, whereas A. morio ranged in between. These results confirm earlier analyses that showed that O. mascula shows pronounced small-scale spatial clustering (Möller 1987; Jacquemyn et al. 2009a,b). However, more research is needed to unequivocally show whether the observed differences in spatial distribution patterns are because of differences in mycorrhizal specificity or different availability of suitable environmental conditions (see McCormick et al. 2012).
Overall, our results support previous findings that co-occurring orchid species tend to use different mycorrhizal partners. Furthermore, our results indicate that above-ground distribution patterns of orchids are directly related to below-ground seed germination. Although the possibility that variation in local environmental conditions affected seed germination could not be completely ruled out, the high diversity of fungal partners and different preferences among species suggests that mycorrhizal fungi contributed, at least partly, to the spatial distribution and coexistence of the investigated orchid species. Different association patterns and patchy distributions of mycorrhizal fungi may lead to patchy distribution patterns of orchids and possibly also constrain the abundance of orchid species. On an ecological time scale, this will have implications for the conservation of orchids, as orchids can only have population sizes that are determined by the spatial distribution of suitable mycorrhizal fungi and the necessary environmental conditions to promote seed germination and because patchy distributions also affect pollen flow patterns and fruit set. Recent research has also shown that patchy distribution patterns of orchid species that can cross to form hybrids affect the spatial extent of hybridization (Jacquemyn et al. 2012). On an evolutionary time scale, patchy distribution of orchids and small population sizes may have contributed to high diversification of orchid species, as they limit gene flow among populations and lead to small effective population sizes, thereby creating ideal conditions for the drift-selection model of orchid speciation proposed by Tremblay et al. (2005).
This research was funded by the European Research Council (ERC starting grant 260601 – MYCASOR and ERC advanced grant 233066 – SPATIODIVERSITY). Marc-André Selosse and one anonymous reviewer provided very useful comments that improved the quality of this manuscript.