Mutualist and pathogen traits interact to affect plant community structure in a spatially explicit model

Empirical studies show that plant-soil feedbacks (PSF) can generate negative density dependent (NDD) recruitment capable of maintaining plant community diversity at landscape scales. However, the observation that common plants often exhibit relatively weaker NDD than rare plants at local scales is difficult to reconcile with the maintenance of overall plant diversity. We develop a spatially explicit simulation model that tracks the community dynamics of microbial mutualists, pathogens, and their plant hosts. We find that net PSF effects vary as a function of both host abundance and key microbial traits (e.g., host affinity) in ways that are compatible with both common plants exhibiting relatively weaker local NDD, while promoting overall species diversity. The model generates a series of testable predictions linking key microbial traits and the relative abundance of host species, to the strength and scale of PSF and overall plant community diversity.


Parameter Definition
sf Host specificity of mutualists (f = m) and pathogens (f = p). When sm = 1, mutualists are complete generalists, and a microbe has an equal affinity for all hosts. When sm = 0.3, the affinity of mutualists for non-preferred hosts is 30% of the affinity of mutualists for preferred hosts. Host affinities of mutualists 1,2 and pathogens 3,4 have been independently investigated in multiple studies. Such studies have shown that mutualists and pathogens can often associate with many host species, but usually have host-specific effects.
gf Fecundity of microbes, represented as a proportion of microbes in any given cell that are dispersed as propagules.
bf Exponent of power law distribution. This indicates the dispersal limitation of both microbial guilds (f = m, f = p), or plants (f = t). Dispersal kernels have been estimated for trees 5 and fungal spores 6 . Power law functions have been shown to fit both plant and fungal dispersal 5,6 . g The impact of microbes on host survival. Higher values of g correspond to simulations in which a given change in the abundance of pathogens or mutualists has a greater effect on seedling recruitment.
h Relative contribution of the mutualist community to seedling recruitment probability.
z Relative fitness of least fit plant species. The fitness values of plant species (i.e. zj where j is the plant species), are evenly distributed between z and 1.
cf Exponent that scales the competitive ability of a microbe according to its effect on host survival. As cf increases, the competitive effect of a microbe decreases when it is associating with a non-preferred host. We assume that cf is greater than zero, meaning that the competitive ability of a microbe associating with a given host decreases as its effect on host survival decreases.
qf Exponent that scales intrinsic growth rate of microbes with host affinity. When the value is negative, microbial populations rapidly crash after their preferred host dies. High values of this parameter allow for a microbe to persist at a site after its preferred host dies. We interpret high values of qp as a representation of taxon's ability to escape competition by occupying alternative life history strategies (e.g. plant pathogens that can persist as saprotrophs or endophytes when they cannot exploit a living host).
rf Intrinsic growth rate of mutualists and pathogens. . Each of these values can be measured empirically to calculate " using potted, or field experiments 9 . Many studies measure PSF using a home-away approach, ( ' ) − ( 1 ), where ( 1 ) is the average performance of species A beneath heterospecifics 10 . Although this approach captures one component of PSF, it cannot predict whether any two species will coexist, because it does not account for the effect of species A on heterospecifics. In the figure above, seedling survival probabilities yield a negative interaction coefficient (i.e. negative PSF). Negative feedback occurs when plants alter their local soil microbiomes to have a relatively more detrimental effect on their own seedlings than the seedlings of other species 11 . Empirical studies show that a local accumulation of relatively species-specific pathogens beneath adult trees creates negative feedback 10,12,13 .

Supplementary Figure 2. Conditional feature contributions -predictors of coexistence.
Point colour represents the proportion of decision trees that predict that all plant species will coexist for the duration of the simulation (i.e. no extinctions). Figure 3. Conditional feature contributions -equilibrium. Specifically, conditional feature contributions from a random forest model using all predictors to determine whether the abundance of the fittest plant species will reach a stable equilibrium (Pe > 0.1) and coexistence (i.e. 5 species coexist until the end of the simulation). Pe represents the equilibrium metric generated from a two-sided t-test, with n = 60 time intervals. No correction for multiple testing was performed, because the P value was intended as a metric. Point colour represents the proportion of decision trees predicting whether all species coexist under equilibrium at the end of the simulation (i.e. the equilibrium metric Pe > 0.1). Figure 4. Conditional feature contributions -negative feedback. Specifically, conditional feature contributions of all predictors of whether strong feedback will develop (random forest 2). Point colour represents the proportion of decision trees that predict that Imax will be negative. Figure 5. Important predictors -PSF-abundance correlation. Specifically, results from a random forest classifier using each variable to predict whether a strong positive correlation (r > 0.8) will develop between host abundance and feedback. a, Importance of each variable (mean decrease in accuracy). b, Conditional feature contribution of variables predicting the relationship between host abundance and feedback strength. In addition to the relative effect of mutualists on seedling survival, h, and the relative host affinity of mutualists and pathogens, sf, the most important variable in determining whether a strong positive correlation between host abundance and feedback developed was qp (the exponent scaling pathogens' intrinsic growth rate with their effect on a non-target host). Effectively, this variable decreases the rate at which pathogens decay in abundance at a site after their preferred host is replaced by a non-target host. As qp increases, pathogens are more able to maintain higher abundances after their host dies. See Supplementary Figure 4 for the conditional feature contributions of every variable.