During the peak flowering seasons of 2016 and 2017, standing crop floral nectar was sampled from two sites in northern California: Stebbins Cold Canyon Reserve (Stebbins) in Winters, CA and at flowering plots maintained at the Laidlaw Honey Bee Facility (Bee Biology) in Davis, CA. The sites are approximately 18 miles apart, so are unlikely to be linked by pollinator dispersal, and differ in pollinator species composition and anthropogenic influence, but share a subset of plant species.

Between late February and early July, flowers were sampled every two-four weeks from available plant species, with approximately 10 flowers of each plant collected per week (collections were limited by floral availability and reserve collection restrictions to protect plant populations). When possible, flowers were sampled from multiple individuals and sub-populations or plots (Supplementary Table S1). Care was taken to sample flowers that had been open at least one day if possible, to allow the opportunity for floral visitation and microbial immigration to flowers. Individual inflorescences were collected, placed upright in humidified boxes and kept cool until extraction and plating, no more than 5 hours later. In the lab, flowers were destructively sampled. Nectar was collected using 10 µl microcapillary tubes, and volume quantified. Nectar was diluted in 30 µl of sterile water (D0), then diluted 10 and 100 fold (D1 and D2 respectively) in sterile phosphate-buffered saline. To assess fungal and bacterial abundance, 50 µl of the 10 fold dilution (containing 5 µl of D0) and 50 µl of the 100 fold dilution (containing 0.5 µl of D0) were plated on yeast media agar (YMA) containing chloramphenicol and Reasoner’s agar containing cycloheximide (R2A Oxoid formula with 20% sucrose), respectively. All samples were plated the day of collection. For convenience, we refer to the total number of colonies on YMA as “fungi” and colonies on R2A as “bacteria” throughout the manuscript although some colonies on each media type may be comprised by microbes resistant to the antimicrobial compounds used here (e.g. bacteria resistant to chloramphenicol (Dhami et al. 2018)). The threshold for detection was approximately 8 live cells for YMA media and 60 cells for R2A media in the original nectar sample. Negative controls were included and plated to detect potential contamination and samples discarded if contamination was detected (detected on 1 date; these samples were removed from the analysis). Agar plates were incubated at 28^oC and colony-forming units (CFUs) counted after 48-72 hours. The total number of CFUs and CFU density for each nectar sample was calculated based on dilutions and original sample volume. Over the course of the study, 1 825 nectar samples were collected and plated on two media types. In 2016, representative colonies were picked haphazardly from plates collected over the entire season including all sites and plant species and frozen in glycerol.

A subset of microbial strains from glycerol stocks were identified using MALDI-TOF and spectra were compared to Bruker Bacteria and Eukaryote libraries and a custom in-house database curated from previously identified microbial isolates from nectar (Supplementary Methods 1).

Insect-based microbial dispersal assay

To empirically test whether nectar microbes differ in their dispersal ability, we developed an assay of microbial dispersal resulting from the feeding activity of western flower thrips (Frankliniella occidentalis). The assay consisted of 5 sterilized adult female thrips freely foraging within a 96-well plate in which 1/3 of wells contained a microbial suspension (200uL 10,000 cells/uL) in sweetened tryptic soy broth (TSB, adding 15% sucrose and 15% fructose) and the remaining 2/3 of wells were filled with sterile TSB (200uL). Thrips were sourced from a colony held by Diane Ullman (see supplementary methods S4 for assay methods details). Microbial isolates from field sampling in this study (see below), and other nectar-associated microbe collections were used (6 bacteria, 5 fungi). After thrips foraged for 24 hours at 30°C under 16L:8D light, thrips were killed by adding 320uL of ethyl acetate to the plate bottom and the media was incubated for 5 additional days. Control replicates with thrips and no added microbes were included in all trials to ensure adequate sterilization and trial consistency. Following incubation, we assessed microbial occupancy of initially sterile wells by calculating the deviation of optical density (OD 600nm) from a blank control plate (no thrips, no microbes) with an occupancy threshold equal to the mean OD of control wells +6 standard deviations.

PacBio sequences are available at NCBI SRA under Bioproject #727253 http://www.ncbi.nlm.nih.gov/bioproject/727353

Thrips analyses are contained in: thripsPlatesBulked.csv , thrips_plate_reader_fileGat...

Field microbial incidence and abundance data are analyzed in: DispersalMsCode_4.29.21.R which uses data in FieldData5.19.20.csv

PacBio (full-length 16S) data are analyzed using : Stebbins_PacBio_DADA2_16S_marshall.R and Stebbins_PacBio_DADA2_ITS_marshall.R

Output files from raw analysis are contained in: ITS_DADA2output.RDS and 16S_DADA2output.RDS and analyzed using: Stebbins_PacBio_Analysis.R which also incorporates metadata found in sampleSubmit_RECORD_2.