Material and Methods

Study sites

This study was conducted in two Mediterranean mixed pine-oak forest communities of the Southeastern of Spain, Mixed Forest of Jaen (MFJ) and Mixed Forest of Segura (MFS, hereafter). The climate is typically Mediterranean, with mean temperatures varying between 6º C in January and 27ºC in July, and scattered rainfall concentrated during autumn and spring (535.4 mm to 611.7 mm). Both forests belong to semi-deciduous and sclerophyllous Mediterranean ecoregion.
Both communities are dominated by Quercus and Pinus spp.Quercus spp. are more abundant in MFJ than in MFS, with Q. faginea being highly abundant in MFJ, Q. ilex common to both sites, and Q. pyrenaica occurring only at MFS. Pinus halepensis also shows high densities in MFJ, but they are lower thanPinus nigra subsp. salzmannii in MFS, which dominates all the canopy layer. The highest species diversity of both communities is located in the sub-canopy and shrub layers. MFJ contains Crataegus monogyna, Juniperus oxycedrus, Phillyrea angustifolia, Phillyrea latifolia, Pistacia Lentiscus and Rhamnus lycioides , among other less abundant species. In turn, Acer granatensis , Crataegus laciniata , Juniperus communis , Prunus spinosa are present in MFS. Although both communities are located relatively close each other, MFJ has almost twice the species richness of woody plants than MFS, being 29 vs 17 respectively (for further information, see Perea et al., 2021a).

Sampling design

At each community we installed a permanent plot of 100×100 m, where the coordinates of all adult tree and shrub individuals were taken. Concentrically to this plot, we located a 50×50 m plot, where we took the coordinates of all saplings. Saplings were individuals with basal diameter > 1cm, without symptoms of being reproductive (i.e. flowers and fruits), and with a size lower than 25% of the typical size of an adult of the species (Alcántara et al., 2019; Verdú et al., 2022). Coordinates for adult plants in the large plots were taken to the closest meter (i.e. 1m), but in the smaller plots they were taken to the closest 0.33m to detect spatial signatures at smaller scales.
At both communities we collect data to characterise for each plant species the organisms of the phyllosphere that are potentially involved in plant community dynamics, both depressing the recruitment (i.e. antagonists such as fungal pathogens, sap sucking insects and chewing insects) or fostering it (i.e. mutualists like fungal epiphytes and endophytes) (Bastida et al., under review , Pajares-Murgó et al., 2022). For further details about data collection and sequencing see supplementary information.

Data analyses

Correlation among plant-associated organisms

We conducted a Mantel test for each plant community to find out if the guilds of plant-associated organisms showed collinear responses. We used the Kendall approach (τ) to add more robustness to this test (not assuming linearity, outlier’s absence or homoscedasticity). This analysis allowed us to assess whether two plant species that share e.g., more pathogens have also a higher probability to share also more epiphytes, sucker insects and/or chewers insects.

Spatial analysis

Our objective is to conduct a spatial analysis of how different guilds of plant-associated organisms are shared between focal plants and their neighbours. To this end we used multivariate techniques of spatial point pattern analysis (Wiegand et al., 2017; Wiegand & Moloney 2014) that allow for such analyses at the level of the entire plant community, and for groups of focal species only (e.g., only dry-fruited and fleshy-fruited species). For the community-level analysis, we used a summary function that is the expected number of plant-associated organisms common to two randomly selected plants that are closer than distance r of each other. This summary function is called “community-level cumulative phylogenetic Simpson index” αphy (r ) (Wiegand et al., 2017) and can be estimated by using standard summary functions of spatial point pattern analysis as:
\(\ \alpha_{\text{phy}}\left(r\right)=\ \sum_{f=1}^{S}{\sum_{i=1}^{S}\delta_{\text{fi}}\frac{\lambda_{f}}{\text{λ\ }}\frac{\lambda_{i}K_{\text{fi}}(r)}{\lambda\ K(r)}}\). (eq.1)
where the similarity index δfi gives the number of OTUs of fungi or insect species shared between the focal speciesf and a second species i , and \(\delta_{\text{ff}}\ \)is the number of unique fungi or insect species hosted by species f . The ratio λi /λ is the relative abundance of species i in the community,Kfi (r ) the bivariate K -function of species i with respect to the focal species f , andK (r ) the univariate K -function of the entire community (Wiegand & Moloney, 2014). Note that the quantity\(\frac{\lambda_{f}}{\text{λ\ }}\frac{\lambda_{i}K_{\text{fi}}(r)}{\lambda\ K(r)}\)in equation 1 is the fraction of pairs of individuals with distance < r where the first plant is of the focal speciesf and the second of species i .
Slight reordering of equation 1 leads to the definition of the corresponding species-level summary function
\(\alpha_{f,phy}\left(r\right)=\ \sum_{i=1}^{S}\delta_{\text{fi}}\frac{\lambda_{i}K_{\text{fi}}(r)}{\lambda\ K(r)}\), (eq.2)
which is the expected number of plant-associated organisms common to two randomly selected plant that are closer than distance r of each other, but where the first plant is of the focal species f . Equations 1 and 2 allows us to define “partial” summary functions where the first plant is e.g., a fleshy-fruited plant. This is accomplished by abundance-weighed averaging (as in equation 1), but only over fleshy-fruited focal species f (Wiegand & Moloney, 2014).
To better interpret our results, we also estimated the partial cumulative spatial Simpson index \(\alpha_{f,S}\left(r\right)\) that gives the proportion of heterospecifics located within distance rof individuals of species (or type) f (Wiegand et al., 2017). It follows from equations 2 by using δfi = 1 forfi and δfi = 0 for f =i .
We want to assess, for example, if saplings of a focal species fshare on average fewer pathogen species with their neighbours than expected. To this end, we compared the actual neighbourhoods of the saplings with that of randomly selected neighbourhoods, testing for independent placement of saplings of species f with respect to their heterospecific neighbours (i.e., sapling-sapling analysis) or with respect to the adult community (i.e., sapling-adult analysis). Because our summary functions include conspecifics, we implemented the independent placement hypothesis as a toroidal shift null model (Lotwick & Silverman, 1982; Wiegand & Moloney, 2014) that conserves the spatial pattern (e.g., aggregation) of conspecifics. To test for significant departures from spatial independence, we conducted 999 simulations of the null models and used standardised effect size (SES) of the normalised summary functions. In case that the standardised effect size is lower than −1.96, or higher than 1.96, we obtain significant departures (p < 0.05) from the null model at the particular neighbourhood distance r (i.e. a point-wise test). IfSES (r ) < −1.96, the mean similarity is smaller than expected, for SES (r ) > 1.96 the similarity is larger than expected, and otherwise they were similar (p > 0.05). To assess the overall significance of departures from the null model for the distance interval of 1 to 10m, we used a global envelopes test based on the standardized effect sizes (Wiegand et al., 2016).