Most phyllosphere-inhabiting bacteria are believed to arrive on above-ground plant tissues via aerial transmission, including wind and rain , and much of this transmission is likely to originate from neighboring plants , though it is unclear if endospheric bacteria arrive in this way. Greenhouse-grown plants are expected to be relatively isolated from microbial dispersal via these channels, and indeed have been shown to develop communities distinct from those developing outdoors . Since greenhouse plants are typically grown in commercial potting mixes, it is also unlikely that they have the full breadth of bacteria available for recruitment from the soil reservoir . Given that greenhouse-grown plants are likely depauperate in their microbial associations, they could provide a unique opportunity to understand the importance of phyllosphere bacteria by re-introducing these communities in a controlled environment. One promising avenue for investigating the causative effects of plant-microbiota interactions in plant health is using synthetic bacterial communities. Ideally, these synthetic communities represent the phylogenetic diversity of natural phyllosphere communities, but at a tractable level of complexity, drainage for plants in pots allowing for repeatable experimentation. This approach has been used to investigate a variety of plant-microbial interactions , but these synthetic communities also hold great potential as microbial ‘probiotics’ or biostimulants .

The addition of beneficial bacteria would be especially useful in environments where microbial diversity is otherwise reduced and/or where host-microbiome associations have been disrupted by, for example, pathogen establishment or antimicrobial treatments. We examine this question using a defined set of naturally occurring bacteria to establish a synthetic community that we developed based on observed diversity of bacterial species residing on field-grown tomato plants. To investigate interactions between microbial associations and nutrient status we included a commercially available micronutrient supplement at various concentrations. We applied PhylloStart to aerial plant tissues during the first weeks of growth post-germination to mirror early life microbiome establishment and tested, across multiple trials, bacterial growth on leaves, disease resistance against a common bacterial tomato pathogen, and tomato fruit yield. In line with the idea that phyllosphere recruitment is likely limited in the greenhouse, we find that field-transplanted plants sprayed pretransplantation with PhylloStart did not show altered fruit yield in response to amendment. Overall, our results demonstrate that, unless supplemented, phyllosphere bacterial communities establish poorly under common greenhouse growth conditions and highlight the underappreciated role of the above-ground microbiome in shaping plant fitness.

Seeds of tomato variety ‘Moneymaker’ were surface sterilized by shaking in a solution of sodium hypochlorite and 2% Tween 20 for 20 min, followed by two rinses with filter-sterilized H2O. Seeds were sowed in trays with Sunshine Mix #1 and germinated in the greenhouse. When seedlings were ~4 inches tall, they were transplanted into pots and distributed across the greenhouse in a randomized design, where they were grown for the remainder of their development under controlled conditions with supplemented lights to maintain long days and fans to control temperature fluctuations. Plants were grown in the Greenhouse facility at the USDA Plant Gene Expression Center greenhouse in Albany, CA. Nutrient supplementation consisting of Peters Professional 20/20/20 watersoluble fertilizer was applied once per week. A disease suppression program with Floramite and Decathlon at a rate of 1/ 4 tsp per gallon was applied through a controlled sprayer at the rate of 1 to 2 gal per 100 plants. Plants in the field trial were started in the greenhouse and received the same treatment as the plants in the third trial until 7 weeks, when they were moved outside to harden, and transplanted into the Oxford Tract field at UC Berkeley at 8 weeks. After transplanting into the field, plants were watered once a week on a drip system for approximately 6 hours, with 20-20-20 water-soluble fertilizer at a rate of 0.93 lb/ac. The drip irrigation prevented water splashing onto aerial tissues and thus reduced transmission of bacteria into soil/roots. Powdery mildew, a common nuisance in greenhouse environments was detected during the second trial and treated, as per standard protocols, by pruning infected tissue. The plants were randomly dispersed throughout the greenhouse, and regardless of location or treatment we did not see a noticeable difference in presence of powdery mildew, and plants were pruned equally across treatments.PhylloStart was designed to mimic the composition of a fieldgrown tomato phyllosphere bacterial community. The community was designed based on community sequences from tomato plants in the Student Organic Farm at UC Davis .

Isolates were collected to be representative at the family level of bacterial species found above 0.1%, and were collected directly from the leaves of tomato plants grown at the Student Organic Farm or, in 3 cases , from the endpoint of a greenhouse selection experiment . Leaf wash was initially plated on King’s Broth and LB agar plates, followed by MacConkey, and 1% Tryptic Soy agar plates. Individual colonies were selected, amplified, and sequenced at the 16S rRNA locus. Included isolates represent 97.8% of the total bacterial relative abundance in our field sequencing at the family level, with families Enterobacteriaceae, Oxalobacteraceae, Pseudomonadaceae, Bacillaceae, Microbacteriaceae, and a member of Brevibacteriaceae that was identified at high prevalence from later field samples. In total, 16 species were selected , with several members representing species level variation within the selected families. Information on the identity of the PhylloStart synthetic community is available in Supplementary Table 2.For preparation of the PhylloStart consortia, strains were grown for three days at 28°C on a media shaker in KB broth. Cultures were then spun down for 10min at 2500g, and the supernatant was replaced with fresh KB. The optical density at 600nm of each sample was read and the volume of each sample necessary to yield a concentration of 0.2 was calculated. Samples were mixed to yield this concentration and the suspension was frozen in 50/50 KB/ glycerol at -80°C until inoculation. On the day of inoculation, the community was thawed, pelleted as above, and resuspended at a concentration of OD600 = 0.02 in sterile 10 mM MgCl2 buffer with 0.01% Silwet surfactant. For the third greenhouse trial and the field trial we included two concentrations of PhylloStart, one at OD600 = 0.02 and another diluted 100-fold. Plants were sprayed with either PhylloStart or sterile MgCl2 with Silwet onto both sides of all leaves until runoff . Inoculation timing varied among experiments; in the first trial plants were inoculated at weeks 4, 5 and 6, in the second trial at weeks 3, 6 and 10, and in the third trial at weeks 2, 4, and 6 .Given the reduced nutrient composition of the potting mix used to grow plants, we supplemented soil with a commercial micronutrient product, Azomite® in order to determine whether PhylloStart had a nutrient-dependent impact. Azomite® is a soil additive and fertilizer derived from volcanic ash that has been shown to increase the growth and yield of tomato plants . In the first greenhouse trial, we used three Azomite treatments: 5% wt/wt Azomite Granular during sowing and transplanting , 1g of Azomite Ultrafine applied after transplanting at the base of the plant at 7, 9, and 12 weeks after sowing , or both Azomite Granular and Azomite Ultrafine applied as described . This trial included a control treatment with no Azomite or PhylloStart , as well as a PhylloStart only treatment and a treatment with both PhylloStart and Azomite Granular and Ultrafine . In the second greenhouse trial, we applied Azomite Granular to all Azomite treated plants while modifying the Azomite Ultrafine concentration using 1, 2 and 3 grams , as well as a control treatment that did not receive either Azomite or PhylloStart . In the third greenhouse trial and the field trial, we included treatments with Azomite Granular and Ultrafine, using 1 and 2 grams in the greenhouse, and 1 and 3 grams in the field and either a low, or high dose of PhylloStart , a PhylloStart only treatment at both concentrations , and a control treatment . A second field trial was performed at UC Davis to confirm the results from our initial field trial, growing raspberries in pots using the same methods as previously described, with the following treatments; PhylloStart , Control , PhylloStart with 3 grams of Azomite and Control with 3 grams of Azomite .

For a detailed overview of all treatments and replicates across experiments see Supplementary Figure 1.To determine whether PhylloStart conferred pathogen resistance to plants, tomato seeds were prepared as described above, then germinated onto 1% water-agar plates. After 1 week, seedlings were transferred to individual pots containing autoclaved calcined clay medium . In this experiment we focus on a specific nutrient, Phosphorous, using 960 mg of organic fertilizer which was added to each pot at the transplant stage. Plants were randomized with respect to treatment and maintained in a growth chamber at a 15 h day:9 h night cycle for the duration of the experiment. PhylloStart was applied to leaves at a concentration of OD600 = 0.02 with 0.01% Silwet surfactant on three-week-old plants. One week after spraying, an overnight culture of Pseudomonas syringae pathovar tomato PT23 was pelleted and diluted in 10 mM MgCl2 to a concentration of OD600 = 0.0002 and inoculated into three leaves per plant via blunt-end syringe inoculation. At 24 hours post-infection, three hole punches were taken from each inoculated leaf . Leaf discs were homogenized in 1 mL 10 mM MgCl2 in a FastPrep-24 5G sample disruption instrument at 4.0 m/s for 40 seconds. Pseudomonas syringae population density on leaves was obtained through colony forming unit plating.Leaves were sampled in the first and third trials to assay the composition of the epiphytic phyllosphere community. In each case, 5 leaves were collected into 50ml conical tubes from random locations across each plant, in the first trial leaves were collected a week after the last inoculation, when the plants were 8 weeks old, while in the third trial they were collected three weeks after the last inoculation, when the plants were 9 weeks old to determine how much of the community persisted over time. These leaves were weighed and 40ml of sterile 10mM MgCl2 was added, then sonicated for 10 minutes, followed by five seconds of vortexing. The bacteria were pelleted, the supernatant was removed, and samples were frozen at -80°C until DNA extraction and sequencing.Paired-end reads were filtered and trimmed to 230 and 160 base pairs , using DADA2 with default parameters . Following denoising, merging reads and removing chimeras, DADA2 was used to infer amplicon sequence variants , which are analogous to operational taxonomic units , and taxonomy was assigned using the DADA2-trained SILVA database. Using DNA extraction and PCR negative controls from 16S rRNA sequencing, the decontam package was implemented using default settings to identify and remove potential contamination from the samples . The assigned ASVs, read count data, and sample metadata were combined in a phyloseq object for downstream analyses. The phyloseq package was used to calculate beta diversity , and a permutational analysis was performed on data rarified to 400 reads  using the adonis function in the vegan package . Fruit production was analyzed with two-way ANOVAs, and where appropriate, a Kruskal-Wallis test was used to control for unequal variances. Post-Hoc tests were performed with either Tukey’s HSD test or, when a Kruskal-Wallis was performed, Dunn’s Test. Tests were performed in R using the package rstatix . Tests for normality across residuals were performed for all linear mixed-effects model, and data was checked for outliers, normality, and homogeneity of variance prior to running the ANOVAs .We hypothesized that greenhouse-grown plants would be relatively depauperate in their microbial associations. To test this, we inoculated seedlings with the PhylloStart community and harvested leaves one week after the last application. There was no significant differences in community composition from plants treated with or without Azomite and so for the sake of simplicity Azomite samples are not included in the Figure . Supporting our hypothesis, we found a significantly higher abundance of bacteria on the leaves inoculated with PhylloStart than those treated with buffer . Further, in the treated plants, the vast majority of the relative abundance of bacterial sequences were associated with PhylloStart members . Together, these data indicate that there was a robust initial representation of the PhylloStart community on the plant leaves, and that there is minimal development of leaf associated bacteria in the greenhouse in the absence of amendment.