Plant microbiomes are dynamic and undergo successional changes with plant development, possibly with new introductions occurring throughout the plant life cycle. Several bacterial reservoirs for the phyllosphere microbiome have been reported, including the air, insect pollinators, seed, other nearby plants, and meteorological conditions. Te impact of the latter on fresh produce crop microbiomes is of particular interest due to the highly variable nature of weather-related events, variation due to geography, and anticipated changes in precipitation patterns in the coming years due to climate change. Increased precipitation and humidity often favor the development of plant disease. Similarly, the prevalence of several foodborne pathogens including pathogenic Escherichia coli, Campylobacter jejuni, Salmonella enterica and Bacillus cereus has been correlated with elevated environmental temperature and humidity. In trials assessing the fate of E coli, fecal coliforms and enterococci applied to the lettuce phyllosphere, bacterial decline rates were slower under moderate and regular rain patterns. At the community level, rainfall events may coincide with drastic changes in the leaf surface microbiomes of canola plants, although changes due to plant development could be difficult to detangle. Below ground, some soil microbial communities are influenced by drying and wetting frequencies, especially those not normally exposed to large fuctuations in soil moisture. Rain may shift the microbial profile of phyllosphere communities through direct seeding of microbes present in rainwater, growing blueberries in pots splash from surrounding soil, increasing water availability for existing microbes, or by washing of loosely adhered epiphytes, creating opportunities for others to fill their former niche.
Airborne biological particles, including bacteria and fungi, may act as ice or cloud nuclei, particles around which rain droplets form. Levels of bioaerosols are elevated during rain events, and in fact, plants have been suggested as “cloud seeders”. Airborne microbes, classified as bioaerosols, may be transferred to plant surfaces directly via rainfall or indirectlyfrom standing water after rainfall. In fact, Salmonella enterica serovar Typhimurium is capable of aerosolizing from puddles and colonizing tomato plants following simulated rain events. Rain splash dispersal can facilitate the transfer of human enteric bacteria from bulk soil to leaf and fruit surfaces even with the use of plastic mulch as a barrier. Other microbes including plant pathogens are similarly capable of aerosolizing and retaining viability, sometimes incorporating aerosolization as part of their lifecycle. In the face of a changing and more variable climate, managing crop protection to ensure crop diversity, food security and food safety will necessitate a deep understanding of crop systems including their association with microorganisms and how they respond to external conditions and stresses. To garner a more comprehensive understanding of the impact of rain on the phytobiome of fresh produce crops that are vulnerable not only to plant disease but also colonization by human pathogens, we characterized the epiphytic bacterial communities dwelling on two commercially important fresh produce crops. A temporal assessment of the epiphytic bacterial communities of commercially cultivated cucumber fruit following a rain event, and tomato carpoplane and leaves surrounding two rain events was conducted.Sterile deionized water was added to sample bags .
Submerged samples were hand massaged through the bag for 30 s then sonicated in a Branson Ultrasonic Bath 8510 for 3 min at a frequency of 40 kHz to dislodge bacterial cells from the carpoplane and phylloplane. Samples were hand massaged again and sonicated for an additional 3 min before filtration. Carpoplane and phylloplane washes were filtered through sterile 0.22 µm nitro-cellulose filters , and filters were frozen at −80 °C until further processing. Total community DNA was extracted from filters using the MoBio PowerWater kit . Te V1-V3 region of the 16 S rRNA gene was chosen for use in bacterial community profiling using 8F-533R primers. Sequencing was carried out using 300-bp paired-end sequencing on the Illumina MiSeq . Illumina’s protocol for 16 S Metagenomic Sequencing Library Preparation was followed for all samples as previously described.Quality filtering and sequence analysis were carried out using QIIME v. 1.8, Mothur v. 1.34, and Phyloseq v. 1.24.0 in R v. 3.5.0. Prior to alignment, sequences went through several quality filtering steps to remove chimeras, non-target sequences , and sequences less than 100 bp in length. Sequences were aligned to the Greengenes Core Set using PyNAST, and taxonomy assignment utilized the RDP Classifier 2.2. Reads that failed to match the reference database were clustered de novo using UCLUST v. 1.2.22. To ensure comparability between samples, the dataset was subsampled to the lowest common sequencing depth, 8,200 sequences per sample. Sample types were analyzed separately to assess the influence of rainfall events on bacterial diversity for each of these niches. Beta diversity was assessed using both unweighted distance matrices and matrices weighted by relative taxon abundance. Phylogenetic distance was incorporated into both distance matrices using UniFrac.
Bray-Curtis dissimilarity, which includes abundance but not relatedness in calculation of dissimilarity, was also assessed. Adonis , a non-parametric MANOVA from R’s Vegan package, was implemented to assess significance of treatment influence on bacterial community structure. Using Principal Coordinates Analysis generated through R’s Vegan and Phyloseq packages, plots were created to visualize β-diversity. Alpha diversity was assessed on the rarefed dataset in Phyloseq using both Observed OTUs and Shannon Index metrics. Observed OTUs represent the number of unique Operational Taxonomic Units at 97% sequence similarity, while the Shannon Index takes into account the proportional abundances of the observed OTUs. To compare α-diversity between groups of samples, ANOVA was employed followed by Tukey’s HSD tests for pairwise comparisons, in R’s stats package v. 3.5.0. More than 95% of sequences were identified to the family-level and differential abundance analysis was continued at this taxonomic rank for the unrarefed dataset using DESEQ. 2. Sample information, sequence data and contingency matrices are available with Qiita study id 12262 . Sequence data have been deposited in the European Nucleotide Archive at the European Bio-informatic Institute under accession number ERP 118277.Daily precipitation measurements were obtained from the National Oceanic and Atmospheric Administration website , using climatological data collected from a weather station located 9km from the sample site. Only limited weather data was available from the local station, so additional weather measurements, including temperature, barometric pressure, and wind speed, were acquired an Automated Weather Observing Station located 18 km away. Te weather station is operated by the Federal Aviation Administration and administered by NOAA , and data was accessed through Weather Underground .Sequencing metrics. Approximately 4 million raw reads from 94 multiplexed samples with an average length of 463 bases, and an average Q score of 35 were further filtered for quality. Reads 1 and 2 were merged at an average efficiency of 83%. High quality unmerged read 1 was also included in downstream analysis. All samples had a Good’s Coverage value exceeding 0.95, indicating that samples were sequenced to a level nearing saturation. After rarefaction, 770,810 sequences were retained for the final analysis. Weather. A dry period had been recorded prior to the commencement of sampling, with the most recent precipitation dating back to a 21mm rain event on 8/20, 3 weeks prior to the first pre-rain sampling. The first rain event on 9/12 recorded 9.14mm of precipitation and the second event on 9/21 reached 9.65mm . The highest daily temperature was recorded on 9/9 and a wind gust occurred on 9/13 around sampling time. Barometric pressure was low on 9/9 and 9/13 compared to the other sampling dates. Rain 1 was accompanied by thunder and lightning. Cucumber carpoplane. Following the 9/12 rain event , a significant change in cucumber fruit surface bacterial community structure was observed for both unweighted UniFrac and weighted UniFrac analyses . Cucumber fruit samples collected 3 days prior to Rain 1 supported bacterial communities that clustered separately from those collected 1 day after Rain 1, drainage gutter and the largest average unweighted UniFrac distances were obtained between samples collected on 9/9–9/13 and 9/9–9/17 . Although some samples collected on 9/17 generated bacterial community profiles resembling the pre-rain profile, some resembled the immediate post-rain profile, a trend evident across 3 different distance metrics . Therefore, within 4 days a partial return to the pre-rain profile was detected. On the cucumber carpoplane, α-diversity increased significantly following Rain 1, as measured by both Observed OTUs and Shannon Index , escalating from an average of 232 to 310 OTUs per sample, a 33.6% increase . Five days after Rain 1, α-diversity remained elevated compared to pre-rain levels , with an average of 300 OTUs per cucumber.
Many of these OTUs were introduced across all replicates, indicating a common source. A core microbiome analysis was conducted to identify taxa shared by 100% of samples collected on each date and across multiple dates. Thirty-eight OTUs were ubiquitous among all dates. Seventy-four OTUs not observed in pre-rain cucumber carpoplane samples were identified in samples collected 1 day post-rain . Of these, 35 OTUs were retained across all replicates 4 days later. By contrast, only 7 OTUs observed in pre-rain samples were not detected 1 day post-rain, with 3 of these being observed again 5 days post-rain. In addition to the introduction of new taxa, changes in the relative abundance of established taxa on the cucumber surface were observed following Rain 1 . Of 809 total OTUs in the cucumber dataset, 112 were differentially abundant between samples collected on Sept 9 and 13 . At the family level, 16 bacterial families differed between the pre- and post-rain time points. Notably, the family Xanthomonadaceae increased from an average of 1.2% to 9.6% relative abundance following rain , dropping to 2.4% 5 days after rain. The Oxalobacteriaceae exhibited a similar increase, from 0.6% to 7.0% , but in this case average relative abundance remained high after 4 days, at 7.1%. Similarly, the Sphingobacteriaceae and Comamonadaceae, initially detected at less than 0.2% average relative abundance, increased at least an order of magnitude in relative abundance following rainfall, remaining elevated 4 days later . Relative abundance for several of the most dominant bacterial families on the cucumber surface declined or increased following rainfall. Te Sphingomonadaceae decreased from an average of 9.6% to 5.1% relative abundance following rainfall but increased 4 days later to 7.6%. Similarly, the average relative abundance of the family Microbacteriaceae diminished following rainfall , increasing to an even higher average relative abundance later . The Enterobacteriaceae demonstrated an opposite shift that was not significant , increasing 1 day after rain from 18.4% to 21.9%, later returning to 17.2% average relative abundance . Although these changes in relative abundance are indicative of community shifts, they do not necessarily translate to increases or decreases in the absolute abundance of certain taxa. This may be the reason why β-diversity of the tomato carpoplane did not strictly parallel the same trend observed on cucumber fruit. There was an overall effect of sampling date on bacterial communities when analyzed using unweighted UniFrac distance and Bray-Curtis dissimilarity but not weighted UniFrac distance . Throughout the sampling period, unweighted UniFrac distance increased steadily in comparison to the first pre-rain sampling date, with the greatest distance measured between samples collected on 9/9 and 9/25 . On the other hand, following Rain 1, α-diversity on the tomato carpoplane resembled the dynamics seen on cucumbers. Observed OTU count increased from 185 to 251 OTUs per sample from 9/9 to 9/13 . Five days after Rain 1, α-diversity by both measures was indistinguishable from pre-rain and 1 day post-rain levels , at 231 OTUs per sample. Following Rain 2, observed OTU count remained elevated compared to pre-Rain 1 but did not significantly increase beyond 9/13 levels. Analysis of the same dates by Shannon Index suggests that across the sampling period, α-diversity was different only between 9/9 and 9/13 . Following Rain 1, similar trends in the core microbiome as those on the cucumber carpoplane were observed on tomato fruit. For the 3 dates surrounding Rain 1, tomato fruit collected on each day supported OTUs common to all samples collected on that date but not observed on any other date, with 9/17 hosting the most unique core OTUs . Moreover, 42 additional OTUs were present across all post-Rain 1 samples, of which 18 persisted in all samples 4 days later. Following Rain 2, only 17 new OTUs were detected on all tomato fruit samples, 7 of which remained present on all samples on the final sampling day.