Seasonal assembly of nectar microbial communities across angiosperm plant species: Assessing contributions of climate and plant traits

Plant-microbe associations are ubiquitous, but parsing contributions of dispersal, host filtering, competition, and temperature on microbial community composition is challenging. Floral nectar-inhabiting microbes, which can influence flowering plant health and pollination, offer a tractable system to disentangle community assembly processes. We inoculated a synthetic community of yeasts and bacteria into nectars of 31 plant species while excluding pollinators. We monitored weather and, after 24 hours, collected and cultured communities. We found a strong signature of plant species on resulting microbial abundance and community composition, in part explained by plant phylogeny and nectar peroxide content, but not floral morphology. Increasing temperature reduced microbial diversity, while higher minimum temperatures increased growth, suggesting complex ecological effects of temperature. Consistent nectar microbial communities within plant species could enable plant or pollinator adaptation. Our work supports the roles of host identity, traits, and temperature in microbial community assembly, and indicates diversity-productivity relationships within host-associated microbiomes.

Methods
Creating nectar microbiome inoculum

We selected five microbe species (Table S1) that are common, widely distributed representatives of nectar microbiomes in various plant species, including those in Northern California (Vannette 2020; Vannette et al. 2021): the yeasts Metschnikowia reukaufii and Aureobasidium pullulans, and the bacteria Neokomagataea thailandica, Acinetobacter pollinis, and Apilactobacillus micheneri. We created our microbial inoculum (Fig. 1A; Supporting Information: Inoculum preparation) as described in Cecala & Vannette (2024). The inoculum contained ~104 cells µL-1 of each species (5 x 104 total cells µL-1).

Floral bagging and inoculation

We conducted 11 rounds of floral inoculation on the University of California, Davis campus (38.540 °N, 121.756 °W) (USA: California: Yolo County) from 22 March to 29 June 2023. In the morning the day before inoculations, we bagged ~10 unopened flower buds on each of 5 to 8 species of flowering plants (Fig. S1) to prevent the transfer of microbes by pollinators. We secured organza bags (7x8.5 cm, 10x13 cm, or 13x18 cm) around flowers, removing all open flowers prior to sealing the bag. Each time we handled flowers, we inspected for any breaches by ants or thrips.

Flowers that opened within bags were inoculated between 0900 to 1100 h. To inoculate a flower, 1 µL inoculum, carried into the field on ice, was delivered onto the nectary using a micropipette and autoclaved tips (Fig. 1B), then flowers were tagged with a unique identifier, and re-bagged. Each week, we inoculated roughly 5 to 8 flowers per plant species (~40 flowers per week). Over the course of the study, we recorded temperature extrema (afternoon highs and overnight lows) for all inoculation days from a local weather station (Fig. 1C, Fig. S2).

Nectar extraction and plating

Roughly 24 hours after inoculation, we excised flowers from plants, sealed them in containers and transported them to the laboratory. Inside a laminar flow hood, we used glass microcapillary tubes (VWR, Drummond) to extract and measure the volume of total nectar in each flower (Fig. 1D). We quantified microbes in nectar as in Cecala and Vannette (2024). Briefly, we diluted pure nectar in DPBS, plated aliquots on each of three agar media types, and incubated for 6 days, after which CFUs were identified and tallied (Supporting Information: Quantifying microbes in nectar). Microbial growth from four bagged flowers breached by crawling insects did not differ markedly from that of other flowers, and remained in analyses.

For each flower, we calculated: (1) the density of CFUs per µL nectar, by dividing the number of CFUs per plate by the actual volume of pure nectar in the aliquot; and (2) the estimated total abundance of CFUs per flower, by multiplying our calculated density (1) by the total volume of nectar originally extracted from that flower. The above values (1 and 2) were calculated for each inoculated microbe individually and for all five species collectively.

For comparison with real nectars and to test our inoculum in artificial solutions, we also added 1 µL inoculum to 10 µL of 30% m/m sucrose and an artificial nectar containing sugars and peptone (“experimental controls”; n=6 replicates each; Supporting Information: Media recipes) in strip tubes. Tubes were sealed and incubated at 25 °C for 24 hours, then processed identically to actual nectar samples.

Determination of floral traits

We estimated concentrations of hydrogen peroxide (H2O2), a known antimicrobial reactive oxygen species found in some nectars (Carter & Thornburg 2004; Mueller et al. 2023), in the nectar of separate, non-inoculated flowers of most sampled plant species (Supporting Information: Peroxide quantification, Table S2). Peroxide values from non-inoculated nectar represent initial conditions which would be experienced by microbes arriving in flowers. To assess the contribution of  floral morphology, we scored floral phenotypes of all plant species on the basis of 28 binary traits used in past studies (Faegri & Pijl 1979; Ollerton et al. 2009) to represent pollination syndromes in multivariate space using Bray-Curtis dissimilarity. We determined trait states through a combination of observation and reference with the Jepson eFlora (ucjeps.berkeley.edu/eflora). We also encoded other traits of particular interest such as inflorescence density and corolla fusion.

Scope of collected data

We excluded from analysis five plant species for which we had few, low quality samples (Table S2). In total, we inoculated 398 flowers across 31 species of plants, 372 of which (93.5%) contained nectar after 24 hours (range: 7 to 16 flowers per species; mean=12 flowers per species). The absence of nectar in flowers did not coincide with any recorded variables. These species comprised 29 genera in 21 families. From the 372 nectar samples, we tallied 1,016,048 CFUs on agar media, of which 99.94% were our inoculated species: 72,242 Metschnikowia; 16,640 Aureobasidium; 20,121 Neokomagataea; 795,332 Acinetobacter; 111,149 Apilactobacillus. We classified 564 CFUs as non-inoculated bacteria or fungi (0.056% of all CFUs), likely originating from other plant tissues or the environment, and excluded these from analyses.

Statistical analysis

We conducted analyses in R (R Core Team 2024). Using package lme4 (Bates et al. 2015), we constructed linear mixed effect models with nectar volume, total and by-species CFU density, and CFU Shannon-Wiener diversity index as dependent variables. As independent variables, we included nectar volume and temperature extrema, and plant species as a random intercept effect. We obtained type III sums of squares, F- and P-values, and Kenward-Roger degrees of freedom using function ‘Anova’ in package car (Fox & Weisberg 2019). We inspected model residuals for normality and variance inflation factors to assess multicollinearity. We also created separate linear models with either plant species or nectar peroxide concentration as a fixed effect, as peroxide data was not collected for three species (Table S2). For linear models in which we included a quadratic predictor, we conducted a likelihood ratio test comparing the goodness of fit of the models with and without the quadratic term.

To test if microbial community composition (as Bray–Curtis dissimilarity) differed by plant species and temperature extrema, we used function ‘adonis’ in package vegan (Oksanen et al. 2020) to perform a permutational multivariate analysis of variance. We used function ‘betadisper’ to examine multivariate homogeneity of dispersions across plant species. Community composition was visualized using non-metric multidimensional scaling (NMDS) ordination, and we tested for significant microbe species vectors using function ‘envfit’. As above, a separate analysis was conducted with peroxide concentration as a predictor variable. To test for co-occurrence between microbe species flowers, we generated Pearson correlation matrices on CFU densities, for both our entire dataset and for each plant species individually, and visualized matrices using package ‘corrplot’ (Wei & Simko 2021).

To estimate plant phylogenetic relationships among sampled plant species, we used the function ‘phylo.maker’ in package V.PhyloMaker2 (Jin & Qian 2022) using the reference plant phylogeny GBOTB.extended.TPL. Using this tree, we tested for a phylogenetic signal of nectar volume, CFU densities, and Shannon diversity using function ‘multiPhylosignal’ in package picante (Kembel et al. 2010) with 10,000 simulations.

To test for relationships between plant phylogenetic relatedness and multivariate microbe community composition, we created a pairwise distance matrix of plant phylogenetic relatedness using function ‘cophenetic.phylo’ in package ape (Paradis & Schliep 2019). We compared this distance matrix to a Bray-Curtis dissimilarity matrix of the mean CFU densities of each microbe by plant species using a Mantel test via function ‘mantel’ in package vegan, calculating Spearman’s ρ with 10,000 permutations. We also created a Bray-Curtis dissimilarity matrix of plant species based on floral trait data and compared this to the two aforementioned matrices. We controlled for the effect of plant phylogenetic distance on pollination syndrome using a partial Mantel test via function ‘mantel.partial’. We generated correlograms for all Mantel tests using the function ‘mgram’ in package ecodist (Goslee & Urban 2007). Figures were created using package ggplot2 (Wickham 2016) and tree plots using ggtree (Yu et al. 2017) and custom function ‘ggtreeplot’ (Hackl 2018).