Negative plant-soil feedbacks disproportionally affect dominant plants, facilitating coexistence in plant communities

Plant-soil feedbacks (PSFs) are proposed as key regulators of plant species dynamics in grassland communities. Positive PSFs often arise from mutualistic mycorrhizal fungi, whereas negative PSFs are caused by soil-borne pathogens that reduce growth or recruitment of conspecifics. PSFs have been implicated in shaping species richness–productivity relationships, promoting exotic invasions via positive feedbacks, and maintaining species coexistence through frequency-dependent interactions. The enemy release hypothesis (ERH) suggests non-native plants may become invasive when freed from co-evolved enemies; empirical support is mixed. How PSF influences species dominance remains debated: some studies report dominant species experience more positive or less negative PSFs than rarer species, while others find predominantly negative PSFs with dominant species accumulating more species-specific pathogens that can stabilize diversity via negative frequency dependence. Most evidence stems from monoculture mesocosms, yet PSFs measured in monocultures may not translate to communities where interspecific interactions modify feedbacks. Recent work reports weak correspondence between greenhouse monoculture PSFs and field PSFs, underscoring the need for community experiments that quantify the interplay between competition and PSF. This study tests whether monoculture PSFs predict community PSFs across 10 native grassland species and assesses whether PSF contributes to invasive success of three non-native species. If predictive, monoculture PSFs should relate to species dominance in communities; alternatively, PSFs in communities may depend on species dominance, with larger species accumulating more pathogens and receiving stronger negative PSFs. The study also tests ERH predictions that invasive species achieve greater dominance in unsterilized (PSF-present) communities.

Prior research indicates that mutualistic mycorrhizae can generate positive PSFs, while species-specific pathogens often drive negative PSFs. Meta-analyses and experiments have linked soil microbes to classic diversity–productivity patterns and suggested roles for PSFs in community assembly, invasibility, and coexistence. Evidence for ERH is mixed: some studies find invaders face less enemy pressure and more positive PSFs compared to natives, while others show the opposite or context dependence. The influence of PSF on species dominance is debated; some studies report dominant species experiencing more positive or weaker negative PSFs than rare species, whereas others show negative PSFs are common and can accumulate more strongly on dominant species, promoting negative frequency dependence and diversity maintenance. Multiple works caution that PSFs measured in greenhouse monocultures often do not correlate with field measurements, indicating that biotic interactions and environmental context in communities can alter PSFs. Theoretical models predict negative PSFs can promote coexistence via frequency dependence, but robust experimental support in multispecies communities has been limited.

Study system and species: Ten native grassland species typical of moist mesotrophic European grasslands were used: Achillea millefolium, Centaurea jacea, Leucanthemum vulgare (Asteraceae), Lotus corniculatus, Trifolium pratense (Fabaceae), Plantago lanceolata (Plantaginaceae), Agrostis capillaris, Anthoxanthum odoratum, Holcus lanatus (Poaceae), Rumex acetosa (Polygonaceae). Three exotic invasive species were included: Avena sterilis (Poaceae), Lupinus polyphyllus (Fabaceae), Solidago gigantea (Asteraceae). Seeds were sourced from commercial suppliers and local collections. Experimental design: A two-phase PSF design was implemented comprising a conditioning phase and a response phase. In the conditioning phase, plants were grown to condition soil biota (accumulating mutualists/pathogens). In the response phase, conspecifics were grown in the conditioned soil under different microbial treatments to quantify PSFs. Substrate and inoculum: Plants grew in 0.5 L pots with a 95:5 (vol:vol) quartz sand:field-soil inoculum mix. The field soil inoculum was collected from 0–10 cm depth at a Belgian nature reserve (Doode Bemde; 50.815509, 4.644803), homogenized, stored at 5 °C for three days, and used to establish microbial communities. Sand contained no organic matter and had N and P below detection limits. Nutrients were provided weekly via half-strength Hoagland solution; over eight weeks each plant received totals of 30.0 mg N, 2.0 mg P, 222.0 mg K, 55.3 mg Ca, 31.5 mg Mg, 4.9 mg Fe, and trace elements. Watering was via saucers to avoid leaching biota/nutrients. Soil microbial treatments (response phase): Three treatments were applied: (i) Sterilized: steam sterilized (1 h at 121 °C) conditioned soil; (ii) PSFtot: unsterilized conditioned soil containing full soil biota; (iii) PSFpath: pathogen/saprobe filtrate added to sterilized conditioned soil. For PSFpath, conditioned soil was suspended, sieved (1000 to 63 µm) and then filtered through a 20 µm mesh to exclude AMF spores and larger fauna while retaining smaller pathogenic/saprobic microbes; the filtrate was added back to sterilized soil. Visual checks confirmed absence of rhizobial nodules on legumes during the eight-week response phase. Monoculture experiment: Each of the 13 species was grown alone through conditioning and response phases. In the response phase, for each species × treatment combination, n = 6 replicates were grown (total 234 plants per phase). Biomass was harvested after eight weeks; above- and below-ground biomass were dried (70 °C, 72 h) and weighed. Community experiment: Communities comprised the 10 native species (one individual per species) randomly arranged in 23 cm diameter pots. Conditioning was performed by whole communities; nutrient inputs were scaled tenfold to match per-plant inputs. In the response phase, the same three microbial treatments were applied. An additional factor included the presence of one invasive species (A. sterilis, L. polyphyllus, or S. gigantea) placed centrally; four levels: no invasive, A. sterilis, L. polyphyllus, S. gigantea. Design totals: 10 native species per community × 3 soil treatments × 4 invasive levels × 10 replicates = 1200 native individuals per phase; plus 90 invasive plants in the response phase (3 species × 3 treatments × 10). Above-ground biomass was measured for all 1290 plants; below-ground biomass for half the replicates (n = 645), with careful root separation. PSF quantification: Two standardized PSF effect sizes were calculated using above-ground, below-ground, or total biomass. PSFtot: standardized effect of full soil biota (unsterilized conditioned soil) relative to sterile soil. PSFpath: standardized effect of the pathogen/saprobe fraction (pathogen filtrate added to sterilized soil) relative to sterile soil. Standardization divided the difference by species’ mean biomass in sterile soil: St.PSFspecies = Σi=1..n (Biomasstreatment − Biomasssterile)/Biomasssterile,avg. PSFs were computed per individual. Statistics: To test predictability of community PSF from monoculture PSF, standardized major axis regressions compared monoculture vs community PSF (PSFtot and PSFpath), with assumptions checked (Shapiro-Wilk, visual diagnostics). To test dominance effects, mixed-effects linear models related community PSF to species’ average biomass in sterile soil (proxy for dominance), functional group (forb, grass, legume), and their interaction; community identity and invasive presence were random effects. For model fit, PSF response was shifted by +1 and biomass entered as 1/x. Additional linear regressions tested biomass–PSF relationships in monocultures. Significance of individual PSFs was assessed by one-sample t-tests or Wilcoxon signed-rank tests vs 0. Two-way ANOVA tested effects of soil treatment and invasive presence on native community biomass; one-way ANOVA tested soil treatment effects on invasive biomass proportions. Post hoc comparisons used Tukey HSD. Analyses were conducted in R 4.2.1 with lme4 and smatr; graphics in ggplot and Inkscape.

• PSFs were predominantly negative across species in both monocultures and communities. In 9 of 10 native species, PSFpath (pathogen-only) was less negative in community than in monoculture, indicating reduced pathogen impacts in diverse communities.
• Monoculture PSFs poorly predicted community PSFs. Relationships between monoculture and community PSFs were weak (e.g., Fig. 1: R² ≈ 0.29 for PSFtot and 0.12 for PSFpath; Extended Data Fig. 2: R² ≈ 0.06 and 0.03), indicating low translatability of monoculture PSFs to community contexts.
• Dominance-driven negative feedback: Larger, more dominant species in communities experienced significantly more negative PSFs than smaller species (negative frequency dependence). This held for above-ground biomass (significant for PSFpath; marginal for PSFtot), below-ground biomass, and total biomass (Table 1; multiple χ² tests P < 0.001 for biomass effects on PSFpath). Root biomass showed the strongest relationships, consistent with rhizosphere pathogen associations.
• Functional group effects: In communities, grasses experienced significantly more negative PSFs than forbs or legumes for both PSFtot and PSFpath (P < 0.001), an effect absent in monocultures. The size–PSF negative dependency appeared strongest in legumes (significant interaction term; Table 1).
• Some smaller species (e.g., Lotus corniculatus, Centaurea jacea) exhibited positive PSFpath in communities despite negative or neutral PSFpath in monocultures, suggesting indirect positive effects via reduced competition when dominant species are disproportionately suppressed by pathogens.
• Invasive species and ERH: Competitive effects of invasives on natives were independent of soil treatment (interaction non-significant, Table 2). Only Lupinus polyphyllus increased its biomass proportion under PSF-present conditions and showed positive PSFtot in monoculture; Avena sterilis and Solidago gigantea showed no significant soil-treatment-dependent dominance shifts.
• Community-scale effects: Soil biota (PSF) reduced native community biomass (two-way ANOVA on natives: above-ground biomass F = 153.4, P < 0.001; total biomass F = 54.6, P < 0.001). Presence of an invasive reduced native biomass overall (F = 39.3, P < 0.001), though species-specific effects were limited (significant reduction only for A. sterilis). For invasives’ own responses: L. polyphyllus showed significant soil treatment effects (AGB F = 7.5, P < 0.01; TB F = 13.4, P < 0.01), whereas A. sterilis and S. gigantea did not.
• Biodiversity modulation: Negative PSFs became less negative with higher community diversity due to dilution of species-specific pathogens. Mutualistic effects may mask some frequency dependence in PSFtot, which was weaker than in PSFpath.

The study directly addresses whether PSFs measured in monocultures are predictive of PSFs in multispecies communities and evaluates whether PSFs contribute to invasive success. Results show that monoculture PSFs are weak predictors of community PSFs, indicating that interspecific interactions and community context substantially modify feedbacks. In communities, negative PSFs scale with species dominance, producing negative frequency dependence: dominant species accrue stronger species-specific pathogen loads and experience disproportionately negative feedbacks, while rare or small species can benefit indirectly from reduced competition. This mechanism supports species coexistence and biodiversity maintenance, aligning with theoretical predictions that negative PSFs stabilize communities. Functional group differences further shape community PSFs, with grasses experiencing more negative feedbacks than forbs or legumes in communities, possibly due to higher root:shoot ratios or broader pathogen host ranges. Contrary to broad ERH predictions, invasive species did not generally gain a competitive advantage from PSFs in this biodiverse community; only L. polyphyllus showed treatment-dependent increases consistent with more positive feedbacks. The dilution of negative PSFs with increasing diversity and productivity may erode invaders’ advantage via ERH, offering a community-level explanation for reduced invasibility of diverse systems. Overall, community-level PSFs, particularly pathogen-driven components, can alter competitive hierarchies and facilitate coexistence, whereas monoculture PSFs fail to capture these dynamics.

Monoculture-based PSF measurements do not reliably predict PSFs in multispecies communities. In communities, dominant species experience stronger negative, pathogen-driven PSFs than subordinate species, generating negative frequency dependence that promotes species coexistence. Grasses receive more negative PSFs than forbs and legumes in communities, and community-level PSFs reduce overall native biomass. Invasive species did not consistently benefit from PSFs, with only Lupinus polyphyllus showing increased dominance under PSF-present conditions, suggesting ERH effects may weaken in diverse communities. Future work should prioritize community mesocosm experiments that manipulate biodiversity and functional composition to quantify how PSFs interact with competition and context, and should incorporate both root and shoot biomass given the strong root–PSF relationships. Investigating PSF–invasion dynamics across diversity gradients may clarify when ERH contributes to invasion success.

The pathogen/saprobe filtrate method (20 µm filtering) excludes AMF and larger soil fauna but retains many bacterial and fungal taxa; some saprobic microbes may influence plant growth within the time frame, potentially confounding attribution solely to pathogens. Certain nematode eggs may pass filters and recolonize, and differential recolonization after filtration can bias microbial community composition and PSF estimates. Pathogens added via filtrate to sterilized soil may not fully recolonize within eight weeks, possibly reducing observed negative effects in PSFpath relative to PSFtot. Some filtered pathogens may be more sensitive to processing, further biasing treatment effects. These methodological constraints could affect the magnitude and comparability of PSF components.