Introduction

Determining the mechanisms and processes that govern the assembly and dynamics of plant communities is of major interest in ecology (Sutherland et al., 2013), and over the last decades, a multitude of coexistence mechanisms have been proposed (Chesson, 2000; Wright, 2002). For example, the classical mechanism proposed by Janzen (1970) and Connell (1971) postulates that coexistence in plant communities may be maintained by specialized natural enemies that cause distance- and density-dependent mortality of early life stages if they are located close to conspecific adults or in areas of high conspecific density, respectively, opening space to be used by heterospecific recruitment. A stabilizing mechanism arises if individuals of a plant species are more frequently attacked as the species becomes more common (a rare-species advantage; Chesson, 2000; Comita & Stump, 2020). Typically, the Janzen-Connell hypothesis is tested by assessing survival of seeds or seedlings located close to adults, or in areas of high conspecific density (Comita et al., 2014), but without explicitly accounting for the organisms mediating these interactions at plant community level. While demographic analyses can reveal if observed patterns are consistent with the Janzen-Connell hypothesis (e.g., Comita et al., 2014), a mechanistic understanding requires the identification of the principal classes of plant-associated organisms driving these patterns (Mangan et al., 2010).
From a plant perspective, interactions with plant-associated organisms can result in a range of effects, from detrimental to beneficial (Bronstein, 2009). Interestingly, the negative effects of antagonists such as herbivorous insects or plant pathogens can be neutralized or reduced by mutualists like leaf epiphytes (Bacon & White, 2016; Pajares-Murgó et al., 2022). The balance of this counteraction varies between plant species, and contributes to plant community dynamics and species coexistence (Agrawal & Maron, 2022; Chomicki et al., 2019; Johnson, 2021). Because antagonists and mutualists operate typically in a distance- and density-dependent manner at the plant neighbourhood scale (Bagchi et al., 2010; Bell et al., 2006; Freckleton & Lewis, 2006), we need to investigate their spatial patterns to better understanding how they may impact plant communities (Perea et al., 2020).
The effect of mutualists and antagonists on plant recruitment is influenced by the seed dispersal mechanism of the species and the emerging “seedscape” (i.e., the environment surrounding a seed that influences its success; Beckman & Rogers, 2013), setting the template for subsequent processes of recruitment and interactions with its neighbours (Nathan & Muller-Landau, 2000; Wiegand et al., 2021). A key aspect of the seedscape is the load of plant antagonists (e.g., herbivorous insects, pathogens) and mutualists (e.g., epiphytes) neighbouring the location of a seed (or seedling). Following the Janzen-Connell hypothesis, neighbourhoods in the seedscape with a high load of antagonists, shared with the seed species, will be harmful for the emerging recruits, while neighbourhoods with a low load of antagonists will be beneficial or neutral. However, little is known about how different antagonist and mutualist organisms distribute among adult and recruit plants and how they shape the seedscape (Beckman & Sullivan, 2023).
Different seed dispersal mechanisms will lead to different densities and spatial patterns in the deposition of seeds in the seedscape, with important consequences for subsequent process of self-thinning (Schupp et al., 2010; Schupp & Fuentes, 1995). For example, zoochorous dispersal generally transports the seeds away from mother plants, placing them into locations where the seeds share less antagonists with neighbouring plants compared to locations close to mother plants. As example, frugivorous birds drop seeds of multiple species under adult plants of other species, leading to multispecies clumps of seeds, seedlings and saplings (Herrera, 1984; Jordano, 2014; Perea et al., 2021). Under these circumstances it is expected that density-responsive antagonists have only moderate effects on the fate of the recruits. In contrast, larger proportions of seeds located close to the parents may suffer strong effects of antagonists (Spiegel & Nathan, 2010). If so, seeds and seedlings will show reduced survival in areas of high share of antagonists, and the remaining recruits that survive into the saplings stage will be located in areas with lower load of shared antagonists.
Techniques of high-throughput sequencing are increasingly effective and affordable, providing new exciting approaches to study plant-microorganism interactions (Põlme et al., 2020). The availability of such data about mutualists and antagonists associated to plant species opens up new avenues for analysing the effect of plant-associated organism on the dynamics and assembly of plant communities (Pajares-Murgó et al., 2022). For example, we showed beneficial effects of arbuscular mycorrhizal fungi in pairwise canopy-recruit interaction networks in Mediterranean forest communities (Garrido et al., 2023). However, limiting the analysis to pairwise plant-plant interaction may produce biased results when recruits tend to be surrounded by multiple plant species with whom they share antagonists or mutualists, because not all of these organisms are host specific (Ali & Agrawal 2012; Gilbert et al., 2015; Gilbert & Webb 2007; Parker et al., 2015).
In this study we consider the entire plant neighbourhood that forms the seedscape. We argue that each plant species hosts typical communities of associated mutualists or antagonists (Gómez et al., 2010), and that the load of mutualists or antagonists shared with its neighbours will determine the fate of recruits in its transition from seed to sapling. Our novel approach involves deriving for each organism type (e.g., herbivorous insects, pathogens or epiphytes) an index\(\delta_{\text{fi}}\) of similarity between two plant species fand i in terms of the number of shared organisms. The conspecific index \(\delta_{\text{ff}}\) is then given by the organisms associated only with the focal species f . Recent techniques of spatial point pattern analysis (Perea et al., 2022; Wiegand et al., 2017) provide the methods to take this novel neighbourhood approach. For example, the summary functionαf ,phy (r ) determines first for each plant the mean number of organisms shared with their neighbours before averaging over all individuals of interest. This allows us to find out if individuals of a given species, life stage or functional group (e.g., fleshy- or dry fruited species) share on average more or less organisms with their neighbours than expected by a null model that locates them randomly in the seedscape.
Our overall objective is to identify non-random spatial patterns in the number of organisms shared by plants and their neighbours, and to compare them to expectations under Janzen-Connell mechanisms. To this end, we analyse data on woody plant species in two fully stem-mapped Mediterranean forest communities with similar species composition, but contrasting species richness and abundances, and different seed dispersal mechanisms (i.e., fleshy fruited species vs. dry fruited species). At each community, we collected and classified the fungi (sequencing) and the insect species inhabiting the phyllosphere of each of the local plant species, including leaf pathogens, sap-sucker insects and chewer insects as antagonists, and leaf epiphytic fungi as mutualists (see supplementary information). This unique dataset allows us to reveal spatial patterns in the similarity of antagonist and mutualist communities between saplings and their saplings and adult neighbours, and adults and their adult neighbours, and whether these patterns are conditioned by the seed dispersal mechanism. Identifying these patterns allow us to better understand plant community assembly during the ontogeny of the species (see Figure 1 for the conceptual framework of this approach).
More specifically, we assess the following questions, i) Is the host-specificity maintained across shared-organisms at species-pairwise levels (analysis 1)? ii) Do the patterns in shared-organisms (between sapling-sapling (analysis 2), sapling-adult (analysis 3) and adult-adult (analysis 4) life stages) meet the contrasting expectations of Janzen-Connell effects for dry- and fleshy-fruited species? These analyses allow us to ask iii) Do adults conserve a spatial signature of Janzen-Connell mechanism? and iv) Are these patterns consistent across communities? (See Table 1 for aims, expectations and technical settings of analyses 1 to 4).