Dilution of specialist pathogens drives productivity benefits from diversity in plant mixtures

Diverse plant communities tend to be more productive and support key ecosystem functions, with productivity gains commonly attributed to resource partitioning. Yet direct evidence for resource partitioning in biodiversity-manipulation experiments is limited, suggesting other mechanisms may contribute. The pathogen dilution hypothesis posits that diversity buffers host plants from compatible pathogens through the presence of unrelated neighbors, lowering disease impacts and increasing productivity. While widely studied in wildlife and human disease systems and supported by meta-analyses showing biodiversity can reduce disease prevalence, its generality remains debated and direct links to plant productivity are sparse. In plant communities, evaluating pathogen impacts is challenging due to pathogen diversity and measurement difficulties, and many studies focus on symptoms rather than impacts on growth. A complementary approach uses plant-soil feedback (PSF) to capture net effects of host-specific microbial changes on plant fitness. Negative pairwise PSF—where plants perform worse in their own soil than in heterospecific soils—has been shown to promote coexistence and could also reduce monoculture productivity while alleviating deleterious impacts in mixtures via pathogen dilution. The authors test whether host-specific pathogen accumulation drives negative PSF and whether these feedbacks predict complementarity and overyielding in field mixtures, expecting stronger effects in phylogenetically diverse communities and with increasing richness.

Prior work shows plant diversity enhances productivity and ecosystem functions, often attributed to resource partitioning, though direct evidence from biodiversity experiments has been limited. The dilution effect literature indicates biodiversity may reduce disease risk, especially in wildlife and human systems, but its universality is debated. In plant systems, studies of pathogen dilution are fewer, and direct links to productivity are scarce. The PSF framework demonstrates that host-specific soil biota can generate negative feedbacks that stabilize coexistence. Meta-analyses indicate PSF effects are stronger and more negative when pathogens are included in the soil community. Experimental manipulations of soil microbes in mesocosms and fields have implicated pathogens in overyielding and suggested PSF can explain biodiversity–productivity relationships. Theory predicts that host-specific pathogen dynamics can generate overyielding, with stronger effects expected when hosts are phylogenetically dissimilar due to reduced pathogen sharing. The study builds on these foundations to connect host-specific pathogen accumulation, negative PSF, and complementarity in field communities.

The study combined a field biodiversity-manipulation experiment, greenhouse PSF assays, high-throughput sequencing of soil and root microbiomes, and feedback modeling. Field experiment (Experiment 1): - Site: University of Kansas Field Station, eastern Kansas, USA (39°03′09″ N, 95°11′30″ W). Historically tallgrass prairie; tilled until 1970; fallow thereafter with recent native prairie grass establishment. Prior to planting, resident soil was tilled and inoculated with native prairie soil. - Design: 240 plots (1.5 m × 1.5 m) established in May 2018. Eighteen native prairie species from three families (Poaceae, Fabaceae, Asteraceae; six species each). Treatments manipulated species richness (1, 2, 3, 6), phylogenetic dispersion (under-dispersed single-family vs over-dispersed multi-family mixtures), and precipitation (50% or 150% ambient rainfall) using paired rainfall exclusion shelters. Plant compositions were replicated across shelters; precipitation treatments began spring 2019 (both treatments received ambient rainfall in 2018). Precipitation effects on microbiomes were not analyzed here. - Sampling: In September 2018 (4 months after planting), six 20 cm × 1.9 cm soil cores per plot were collected near planted individuals for soil and root microbiomes. Samples were split: a portion was processed for DNA (soil and roots washed and stored at −20 °C) and the remainder stored at 4 °C for PSF assays. - Biomass: In July 2019 (peak biomass), aboveground biomass harvested from 0.1 m² strips, sorted to species, dried, and weighed. Plant cover surveys enabled conversion of cover to biomass via species-specific regressions to estimate plot-level yields. Greenhouse PSF assays (Experiment 2): - Background soil: Field Station soil (pH 5.93; 0.17% total N; 6.7 mg kg⁻¹ Mehlich P; 3.83% OM) sieved, mixed 1:1 with river sand, and steam sterilized twice. - Pot setup: Deep pots (6.4 cm diameter × 25.4 cm height) filled with three layers: 225 ml sterile soil, 50 ml (10%) inoculum from field monoculture plots, and 225 ml sterile soil. - Design: All 18 species were grown in soils inoculated with conspecific monocultures and heterospecific monocultures from same or different families, yielding 81 full-factorial pairwise feedbacks. Replication: conspecific inoculum n=9 per species; heterospecific inocula: three species per family with three replicates each; sterile controls n=3 per species. Total 702 pots in a randomized block design. - Conditions: Greenhouse in Lawrence, KS; natural light 800–1200 μmol m⁻² s⁻¹; 20–30 °C; drip irrigation to ~20% soil moisture. Harvest after 2 months; shoots and roots dried and weighed. - PSF calculation: Pairwise PSF measured as the log response ratio: PSF = ln(α/β) = ln(αx) + ln(βy) − ln(αy) − ln(βx), where αx is species A in A soil, βy is species B in B soil, and αy, βx are reciprocal performances. Negative PSF values indicate conditions favoring coexistence. Molecular methods: - DNA extraction: 0.25 g fresh soil and 0.25 g roots using Qiagen DNeasy PowerSoil. - Amplicon targets and primers: bacteria (16S V4; 515F/806R), fungi (ITS; fITS7/ITS4), AMF (LSU; FLROR/FLR2), oomycetes (ITS; ITS300/ITS4). PCR conditions followed standard protocols; products indexed (Nextera XT), purified (AMPure XP), quantified (Qubit), pooled equimolarly, and sequenced (Illumina MiSeq v3 PE300). - Bioinformatics: QIIME2 pipeline with DADA2 denoising/merging; taxonomy with SILVA 99% (bacteria) and UNITE 99% (fungi). AMF sequences curated via phylogenetic filtering (MAFFT alignment; RAxML GTRGAMMA; 1000 bootstraps) using a custom AMF database. Fungal functional guilds assigned via FungalTraits; plant pathogens defined by “plant_pathogen” lifestyles; saprobes defined accordingly. Oomycete OTUs identified via BLAST against NCBI database and phylogenetic placement. Rarefaction to minimum read depths per group for downstream analyses. Statistical analyses: - Microbiome dissimilarity: Pairwise Bray–Curtis dissimilarities among species’ monoculture soils and roots (vegan package). - PSF estimation: Log-transformed biomass analyzed via linear models with plant species, inoculum species, and their interaction; seedling height as covariate. Marginal means used to compute PSF. Overall PSF tested against zero using random-effects meta-analysis (metafor), with variance of PSF estimates accounting for shared conspecific means. - Linking pathogens to PSF: Regressions of PSF against microbial dissimilarities for bacteria, fungi, AMF, and oomycetes in soil and roots. Model selection compared linear vs random forest; multi-model inference (glmulti) identified key predictors (soil fungal pathogens, soil oomycetes, root fungal pathogens). Best linear predictor of pairwise PSF: PSF_ij = 1.52 × (soil fungal pathogen dissimilarity) − 2.27 × (soil oomycete dissimilarity) − 1.27 × (root fungal pathogen dissimilarity) + 1.589. - Predicted PSF for field plots: Sum over pairs weighted by realized field biomass proportions: Predicted PSF_effect = Σ p_i p_j PSF_ij. - Predicted pathogen dilution: −(Predicted PSF) × (1 − 1/N), where N = species richness in a plot, capturing dilution from heterospecific neighbors. - Biodiversity effects: Complementarity effect (CE) and selection effect (SE) computed from 2019 biomass following Loreau & Hector. Relative yield total (RYT) calculated; RYT > 1 indicates overyielding. Effects of richness and phylogenetic dispersion on biomass, CE, and RYT assessed with ANOVA; CE and RYT tested against 0 and 1 with t-tests. Relationships between CE/RYT and predicted PSF and pathogen dilution analyzed via simple linear regression. Theoretical analysis: - A general frequency-dependent feedback model was parameterized using empirically derived pathogen dissimilarities and PSF predictor (above) to build an interaction matrix σ_ij among up to 11 coexisting species subsets from the pool of 18. Feasibility and local stability of all 262,125 possible communities were evaluated. For feasible, stable communities with negative community-level feedback (2–11 species), predicted PSF effects, pathogen dilution, and complementarity were computed to test theoretical relationships with complementarity across N=17,850 simulated communities.

- Rapid pathogen differentiation: Soil and root pathogen communities diverged among host species and community compositions within four months of field establishment. - Negative PSF predominates: Average pairwise PSF across species was negative (95% CI −0.48 to −0.24; p < 0.0001; N=81), indicating stronger intraspecific than interspecific negative effects. - Family-specific patterns: Within-family PSF was negative for composites (Asteraceae; 95% CI reported as −0.90 to −1.04; p < 0.0001) and grasses (Poaceae; 95% CI −0.71 to 0.09; p = 0.011), marginally negative for composite–legume pairs (95% CI −0.55 to −0.03; p = 0.079), negative for composite–grass pairs (95% CI −1.42 to −0.87; p < 0.0001), neutral for grass–legume pairs (95% CI −0.20 to −0.26; p = 0.81), and positive within legumes (Fabaceae; 95% CI 0.23 to 0.89; p = 0.001). - Pathogen dissimilarity predicts PSF: Only pathogen components of the microbiome predicted PSF. Pairwise PSF became more negative with increasing soil fungal pathogen dissimilarity (R²_adj = 0.036; p = 0.05) and soil oomycete dissimilarity (R²_adj = 0.05; p = 0.025). Multi-model inference identified soil fungal pathogens (p = 0.037), soil oomycetes (p = 0.028), and root fungal pathogens (p = 0.049) as the strongest PSF predictors. The most abundant fungal pathogens differed among most host species, implicating specialist pathogens in feedbacks. - Diversity–productivity linkage via PSF and dilution: In the subsequent field season, complementarity and productivity increased with more negative pairwise PSF measured in the greenhouse, consistent with pathogen dilution driving diversity benefits. Statistical relationships between plant productivity metrics and PSF-based predictors showed significant explanatory power (Fig. 4): models relating productivity to PSF and to predicted PSF effect and pathogen dilution had R² ≈ 0.11–0.12 with highly significant p-values (e.g., R² = 0.11, p = 9.9×10⁻⁸; R² = 0.12, p = 1.8×10⁻⁶; R² = 0.11, p = 9.9×10⁻⁶; R² = 0.11, p = 8.2×10⁻⁶; N=168). - Modeling corroboration: A general feedback model parameterized with empirical interaction strengths reproduced observed patterns: complementarity increased with species richness; complementarity was negatively related to predicted PSF (r = −0.731; p < 1×10⁻⁶) and positively related to predicted pathogen dilution (r = 0.883; p < 1×10⁻⁶) across N = 17,850 simulated communities. The model indicates pathogen dilution can generate linear diversity–productivity relationships over time. - Overall conclusion: Combined empirical (field and greenhouse) and theoretical evidence demonstrate that dilution of host-specific pathogens reduces negative feedbacks in mixtures, driving complementarity and overyielding in diverse plant communities.

The study establishes a causal chain from host-specific pathogen accumulation to negative plant-soil feedback and then to increased complementarity and productivity in diverse plant mixtures via pathogen dilution. Rapid divergence of pathogen communities among hosts generated negative PSF that depressed monoculture yields; mixing species reduced densities of compatible hosts and diluted pathogen impacts, thereby enhancing productivity. Pathogen dissimilarity, rather than broader microbiome components (bacteria, AMF), predicted PSF strength, underscoring specialist pathogens as key drivers of feedbacks and diversity benefits. Field results in the second season linked more negative greenhouse-measured PSF to higher complementarity and overyielding, with relationships strengthened when accounting for expected pathogen dilution from increased richness. Theory parameterized with empirical data reproduced these relationships, suggesting generality and the potential for pathogen-mediated coexistence to sustain productivity gains over time. Context matters: effects may be strongest in systems with coevolved native plants and pathogens and intact soil microbiomes, and could be modulated by global change drivers (warming, precipitation, nutrient enrichment) that influence pathogen dynamics. The findings position pathogen dilution as an alternative and complementary mechanism to resource partitioning for explaining biodiversity–ecosystem functioning relationships.

This work provides strong empirical and theoretical evidence that dilution of host-specific pathogens is a key mechanism generating productivity gains from plant diversity. By linking host-specific pathogen divergence to negative PSF and then to complementarity and overyielding in field mixtures, the study substantiates pathogen dilution as a driver of biodiversity–productivity relationships. The modeling framework further indicates that pathogen dilution can promote coexistence and maintain productivity benefits over time, and predicts stronger benefits with higher species richness and greater phylogenetic dissimilarity. Future research should: (1) quantify the relative contributions of pathogen-mediated PSF versus resource partitioning across ecosystems; (2) identify specific pathogen taxa responsible for negative feedbacks; (3) assess how anthropogenic changes (warming, altered precipitation, nutrient enrichment) modulate pathogen dilution and BEF relationships; and (4) link pathogen dilution-driven productivity gains to other ecosystem functions such as carbon sequestration.

- Attribution to specific pathogens was not possible; multiple pathogens likely contribute to observed PSF, and their individual effects remain unresolved. - Family-level deviations (e.g., positive PSF within legumes) indicate context dependence and potential influences of mutualists (e.g., rhizobia) that were not fully disentangled. - Soil rhizobia data were insufficient for inclusion in predictive models due to limited ASVs and missing values. - Precipitation treatments were established after the first season and precipitation effects on microbial community composition were not analyzed here. - The experiment focused on a set of 18 native prairie species and a specific site with native soil inoculum; generality across other ecosystems, species pools, and disturbance histories requires further testing. - Modeling assumed equal conspecific effect strengths among species to parameterize σ_ij, which may simplify real-world heterogeneity. - Measuring pathogen abundance and diversity remains challenging; conclusions rely on dissimilarity metrics and functional guild assignments that may contain classification uncertainties.