Linking genomic and phenotypic traits to interaction outcomes in a synthetic phyllosphere community

Batarseh, Tiffany 1 ; Collado, Jose O.1; Wang, Chuyu1; Mehlferber, Elijah C.2; Alvarez-Navarrete, Rocelia M.1; Wagner, Fiona J.1; Koskella, Britt1

Research facility: University of California, Berkeley

Published Sep 22, 2026 on Dryad. https://doi.org/10.5061/dryad.v6wwpzh9z

Data files

Sep 22, 2026 version files 208.28 KB

Abstract

Microbial community function is difficult to predict from the properties of individual members, in part because the same species can behave differently in isolation and in a community. Using a 16-member synthetic bacterial community isolated from the tomato phyllosphere, with the pathogen Pseudomonas syringae pv. tomato, we combined whole-genome sequencing, phenotypic profiling, and reciprocal spent-media assays spanning over 600 single-species and community-level combinations to ask what predicts the outcome of environmentally mediated interactions. Genomic annotations, metabolic breadth, growth strategy, and phylogenetic relatedness were all poor predictors of interaction outcomes in nutrient rich media, and cross-validated models did not perform better than a mean-only baseline. Community-conditioned media reduced growth substantially more than the average of their member strains predicted, but never more than the most inhibitory member, so that community effects fell consistently between the average and the strongest single-member effect; the number of member strains did not predict suppression, indicating that composition rather than inoculum size drives these effects. Although this community suppresses P. syringae on plants, the pathogen was indistinguishable from commensal members in cell-free assays, both in its sensitivity to community filtrates and in the effect of its own filtrate, indicating that protection is not reducible to the chemical environment the community generates. On tomato seedlings, pathogen challenge raised disease scores on bare plants but not on plants carrying a community, whereas no individual strain altered disease outcome. Across scales, interaction outcomes were governed by the composition of the resident community rather than by the identity, traits, or ecological role of any single member, indicating that the community, together with its host context, should be considered for predicting microbiome function.

Dataset overview

Spent-media (conditioned filtrate) growth assays for a 16-member synthetic bacterial community isolated from the tomato phyllosphere (PhylloStart), together with the model foliar pathogen Pseudomonas syringae pv. tomato DC3000. Each focal strain was grown in cell-free filtrate generated by every other strain individually and by defined community assemblages, and growth was quantified as area under the growth curve (AUC).

Two experiments are represented:

Primary experiment — 40 spent media (17 single-strain, 23 community) against all 17 focal strains. Each medium was generated from a single conditioning culture, so replication within this experiment is technical: single-strain filtrates appear on two assay plates (six wells per focal strain × medium) and community filtrates on one plate (three wells). Files: Raw_Aucs.xlsx, SpentMedia_Mean_AUC.xlsx.
Reproducibility experiment — 22 community compositions, each conditioned in three independently grown biological replicates (66 filtrates), assayed against four focal strains. File: Replicate_AUC_Inhibition.xlsx.

Files

`Raw_Aucs.xlsx`

Single sheet. One row per assay well from the primary experiment (n = 3,264). Well-level AUC values, before any aggregation.

`SpentMedia_Mean_AUC.xlsx`

Three sheets, all derived from Raw_Aucs.xlsx by averaging wells within each spent medium × focal strain combination.

Sheet	Contents
`mean_AUC_by_medium_strain`	Mean AUC and relative growth change for every spent medium × focal strain combination (n = 731).
`mean_AUC_with_media_class`	The same 731 rows, with an added column classifying each medium as single-strain, community, or control.
`mean_dAUC_per_medium`	One row per spent medium (n = 43): the relative growth change averaged across all 17 focal strains.

`Replicate_AUC_Inhibition.xlsx`

Single sheet. Well-level AUC and relative growth change from the reproducibility experiment (n = 287).

Variables

Each unique variable is described once; several recur across files.

Variable	Description
`Bacteria`, `focal_bacteria`	Focal strain whose growth was measured in the spent medium. Isolate codes are defined below.
`Plate`	Assay plate identity within a focal strain (`Plate 1`, `Plate 2`). Plate labels are nested within focal strain: DC3000's Plate 1 is a different physical plate from SC01's Plate 1.
`Well`	96-well plate coordinate (row letter A–H, column 01–12).
`Spent_Media`	Identity of the filtrate in which the focal strain was grown. Naming conventions below.
`Media_Type`	Medium category. `Spent` = conditioned filtrate; `Fresh` = uninoculated King's B medium; `Neg` = negative control (10 mM MgCl₂ buffer inoculated with bacteria, simulating no growth); `Blank` = uninoculated contamination control.
`Spent_Media_Origin`	Whether the filtrate was conditioned by one strain (`Single_Species`), by a defined assemblage (`Community`), or is a control (`Control`, `Fresh_media`, `Neg`).
`auc`, `AUC`, `Mean_auc`	Area under the growth curve, from OD₆₀₀ readings taken every 12 h over 72 h at 28 °C. Dimensionless (OD₆₀₀ × hours). `auc`/`AUC` are single-well values; `Mean_auc` is the mean across wells for a given medium × focal strain combination. Occasional small negative values arise from blank subtraction on non-growing wells.
`Inhibition_Factor`, `Inhibition_Factor_Value`, `Average_InhibitionFactor`	Relative growth change (dAUC), calculated as (AUC in spent medium − AUC in fresh medium) / AUC in fresh medium, using the focal strain's own growth in fresh King's B as the reference. Dimensionless. Zero indicates growth matching fresh medium, negative values indicate suppression, positive values indicate enhanced growth. `Average_InhibitionFactor` is this quantity averaged across all focal strains for a given medium.

Naming conventions

Isolate codes

Code	Species
SC01	Frigoribacterium endophyticum
SC02	Pantoea agglomerans
SC03	Curtobacterium herbarum
SC04	Erwinia tasmaniensis
SC06	Brevibacterium frigoritolerans (syn. Peribacillus)
SC09	Massilia aurea (syn. Telluria)
SC10	Pseudomonas asturiensis
SC13	Exiguobacterium sibiricum
SC14	Pseudomonas rhizosphaerae
SC16	Pseudomonas moraviensis
SC21	Rathayibacter festucae
SC23	Curtobacterium pusillum
SC25	Pseudomonas rhodesiae
SC26	Pantoea allii
SC33	Microbacterium oleivorans
SC35	Bacillus thuringiensis
DC3000	Pseudomonas syringae pv. tomato DC3000 (pathogen)

Spent medium codes — primary experiment

Pattern	Meaning
`SC01` … `SC35`, `DC3000`	Single-strain filtrate conditioned by that isolate.
`SS_` prefix (in `Raw_Aucs.xlsx`)	Single-strain filtrate; `SS_SC10` is the same medium as `SC10` elsewhere.
`FSC_SM`	Intact 16-member PhylloStart community.
`FSC_DC3000`	PhylloStart plus P. syringae DC3000 (17 members).
`DO_SC01` … `DO_SC35`	Single-strain drop-out: PhylloStart with that one member removed (15 members).
`DO_GN`	All Gram-negative members removed.
`DO_GP`	All Gram-positive members removed.
`DO_Pa`	Both Pantoea members removed (SC02, SC26).
`DO_Ps`	All Pseudomonas members removed (SC10, SC14, SC16, SC25).
`DO_EE`	P. moraviensis SC16 and P. rhodesiae SC25 removed.
`Fresh_media`	Uninoculated King's B medium.
`MgCl2_Blank`	Uninoculated 10 mM MgCl₂ buffer.
`Negative_Control`	MgCl₂ buffer inoculated with bacteria.

Note that drop-out codes name the member that was removed, not the members
present.

Spent medium codes — reproducibility experiment

Codes take the form <community>_<replicate>, where replicate is A, B or C and
denotes an independently conditioned biological replicate.

Pattern	Meaning
`DO_01_A` … `DO_35_C`	Single-strain drop-out; the two-digit number is the isolate code (`DO_01` = PhylloStart minus SC01).
`DO_GN_`, `DO_GP_`, `DO_Pa_`, `DO_Ps_`	Group drop-outs, as above.
`FSC_*`	Intact PhylloStart community.
`FSC_DC_*`	PhylloStart plus DC3000.
`KB_A` … `KB_F`	Uninoculated King's B medium controls (six replicates).

Focal strains in this experiment are DC3000, SC14, SC16 and SC21 only.

Methods summary

Filtrates were prepared by growing each strain or assemblage in 40 mL King's B medium at 28 °C for 48 h to stationary phase, then removing cells by centrifugation and 0.22 µm filtration. Focal strains were normalised to OD₆₀₀ = 0.002 and 5 µL inoculated into each filtrate well. Growth was tracked by OD₆₀₀ every 12 h for 72 h at 28 °C. Full methods are in the associated manuscript.