Nectar contains antimicrobial constituents including hydrogen peroxide, yet it is unclear how widespread nectar hydrogen peroxide might be among plant species or how effective against common nectar microbes. Here, we surveyed 45 flowering plant species across 23 families and reviewed the literature to assess the field-realistic range of nectar hydrogen peroxide (Aim 1). We experimentally explored whether plant defense hormones increase nectar hydrogen peroxide (Aim 2). Further, we tested the hypotheses that variation in microbial tolerance to peroxide is predicted by microbe isolation environment (Aim 3); increasing hydrogen peroxide in flowers alters microbial abundance and community assembly (Aim 4), and that microbial community context affects microbial tolerance to peroxide (Aim 5). Peroxide in sampled plants ranged from undetectable to ~3000 µM, with 50% of species containing less than 100 µM. Plant defensive hormones did not affect hydrogen peroxide in floral nectar, but enzymatically upregulated hydrogen peroxide significantly reduced microbial growth. Together, our results suggest that nectar peroxide is a common but not pervasive antimicrobial defense among nectar-producing plants. Microbes vary in tolerance and detoxification ability, and co-growth can facilitate the survival and growth of less tolerant species, suggesting a key role for community dynamics in microbial colonization of nectar.

Aim 1: Assessing natural levels of hydrogen peroxide in floral nectar

We sampled the floral nectar of 45 flowering plant species (Plantae: Angiospermae) (in addition to four cultivars of Epilobium canum) growing on the campus of University of California, Davis (38.540 °N, 121.756 °W, USA: California: Yolo County) and in the Sierra Nevada Mountain Range (Incline Lake, USA: Nevada: Washoe County) between March-July 2023 & 2024. We selected plant species that were in flower in spring/summer (2023 & 2024), produced nectar in volumes high enough to be extracted using glass microcapillary tubes (VWR International, Drummond Scientific Company, 50 μL) and have been previously studied with interest in microbial community inoculation ] (J. Cecala et al., 2024). Nectar was always extracted, protected from UV light under aluminum foil, kept at room temperature and processed with 3 hours of flower collection to reduce potential peroxide degradation.

To determine the concentration of nectar peroxide, we used one of two lab assays (Table 1, *for Amplex Red). For the first method, 5 μL of floral nectar was assayed using the colorometric Amplex^TM Red Hydrogen Peroxide/Peroxidase Assay Kit (ThermoFisher, Waltham, MA), and measured at an absorbance of 560 nm on a spectrophotometer microplate reader to determine hydrogen peroxide concentration (SYNERGY HTX, Agilent, Santa Clara, CA) (See supplemental methods).  A hydrogen peroxide calibration curve was run for each Amplex Red assay performed in addition to blanks and peroxidase-free negative controls. For the Amplex Red assay the limit of detection (LOD) is 50 nM, while our limit of quantification (LOQ) is 1 μM. Any absorbance values below our LOQ we reported as < 1 μM. Alternatively, we used colorometric mid range peroxide test strips with 0.5-2 μL of nectar (WaterWorks^TM Mid Range Peroxide Check, Industrial Test Systems Inc., Rock Hill, SC) to determine peroxide concentration for some samples. We used a standard hydrogen peroxide curve to assess the performance of the peroxide test strips, which have discrete color changes for <0.5, 2, 5, 10, 25, 50, and 100 ppm. For the test strip assay the LOD and LOQ are 0.5 ppm (15 μM). Any color change that matched <0.5 ppm we reported as 0.5 ppm, and for those that were fainter than <0.5 ppm we reported as <15 μM.  Despite having lower precision, test strips exhibited a linear response to peroxide which the Amplex^TM Red kit did not (due to oversaturation and the reversal of the colorometric reaction), so the test strips gave us greater confidence in the range of measurement without the need to measure multiple dilution points (See supplemental methods) . We note, however, that the peroxide strip test method may detect peroxides that include but are not limited to hydrogen peroxide, therefore, any values obtained from the test strips we refer to as peroxide, rather than hydrogen peroxide concentrations. We detected no significant degradation of prepared H₂O₂ solutions (10 or 35 μM) after 3 hours (Supplemental Figure S1).

In addition, we assessed the existing body of literature on hydrogen peroxide concentration in nectar, recording concentrations and detection methods used (Table 2, color coded by study). We used search terms including nectar, peroxide, hydrogen peroxide, nectar defense, and nectar antimicrobial compounds.

Aim 2: Do plant defense hormones induce upregulation of floral nectar hydrogen peroxide? 

To explore whether the production of nectar hydrogen peroxide may be induced by common plant hormones in living flowers, we tested the effects of methyl jasmonate and methyl salicylate. Jasmonic acid and salicylic acid upregulate defensive plant pathways associated with herbivory and microbial invasion (Bruinsma et al., 2008; Hernandez-Cumplido et al., 2016; Hoffmeister & Junker, 2017; Radhika et al., 2010). We were curious whether the methyl volatile conjugates of these plant hormones (the forms used in plant-plant signaling) could alter concentrations of hydrogen peroxide in the nectar of long-blooming Peritoma arborea (bladderpod). We performed these manipulative field experiments on the campus of the University of California, Davis.

We selected two spatially separated (by 0.25 mi) patches of bladderpod. We treated recently opened flowers by spraying them with approximately 650 μL (350 mL spray bottle) of either a methyl jasmonate (1 mM in 2% EtOH), methyl salicylate (1 mM in 2% EtOH), or 2% ethanol (control) spray solution (Thaler et al., 2001). We held a container under the sprayed flower to catch any excess liquid and prevent the treatment from impacting other parts of the plant. Care was also taken to avoid flooding the flowers with liquid (and diluting the nectar solution) by positioning flowers so they faced downward when spray was applied, such that the lower parts of the petals, as well as the tips of the stigma and anthers received treatment. We then placed mesh bags over the flowers to prevent pollinator visitation and removal of floral nectar.

After 24 hours, we collected the individual treated flowers. Floral nectar was extracted using glass capillary tubes (VWR International, Drummond Scientific Company, 50 μL) and the nectar column’s length combined with a conversion factor based on the tube size was used to calculate the total volume of nectar extracted per flower. Peroxide concentrations were assessed using the Amplex^TM Red Kit (Thermo Fisher Scientific) using 5 μL of nectar, as previously described in Experiment 1 (also, see supplemental methods).

Aim 3: How does microbial identity and isolation source influence tolerance to hydrogen peroxide?

To test how microbial identity and isolation source impacts tolerance to hydrogen peroxide, we selected fungi isolated from floral petal tissue, queen bumble bee regurgitant, bumble bee honey pot stores, and various floral nectars. For each assay, each strain was freshly streaked from freezer stocks (Table 2) onto yeast media (YM) agar containing 0.1% v/v of 100 mg/mL chloramphenicol (antibiotic) in methanol solution to inhibit bacterial growth. After at least 72 h of growth on agar at 25 °C, we created pure culture suspensions at 0.1 optical density at 600 nm (OD₆₀₀) in sterile 20% sucrose solution (0.2 micron syringe filter). We prepared artificial nectar growth media (see Aim 3 recipe in supplemental methods) and sterilized it with a 0.2 micron sterile filter. We prepared hydrogen peroxide stock solutions through a set of serial dilutions into artificial nectar using stabilized 3% hydrogen peroxide (Thermo Scientific Chemicals) resulting in final in-plate concentrations of 0 μM, 100 μM, and 2000 μM.

Microbial growth was assayed by monitoring optical density at 600 nm (OD₆₀₀) within 96 well growth plates. Each well contained 120 μL of the artificial nectar growth media, 10 μL of appropriate hydrogen peroxide stock solution, and 20 μL of microbial culture suspension. Each treatment combination was replicated in triplicate, and negative controls lacking microbe inoculum were run at each hydrogen peroxide concentration to ensure lack of microbial contamination. We incubated the microplates at room temperature in the dark and used a spectrophotometric microplate reader (SYNERGY HTX, BioTek Gen5 Software) to measure optical density at 600 nm every 2 hours for 24 hours. Before each read the plates were shaken linearly for 30 seconds. We scaled the OD₆₀₀ measurements for each microbial taxon relative to their control (0 μM H₂O₂) because microbial growth varied among species depending on the media type. This allowed us to better visualize the impact of hydrogen peroxide treatments on each microbe type relative to one another.

Aim 4:  How does hydrogen peroxide concentration shape microbial assembly in the nectar of living flowers?

To assess whether artificially elevated hydrogen peroxide in nectar affects the composition and growth of an inoculated microbial community, we performed manipulative field experiments with Epilobium canum (Greene) P.H.Raven (Myrtales: Onagraceae) on The University of California, Davis campus from June to August 2023. Epilobium canum was selected because of its harvestable nectar volume and floral phenology and longevity, typically 3-5 days.

We prepared synthetic microbial communities and enzyme solutions for floral inoculation experiments. We selected microbial species isolated from flowers or pollinators that were morphologically distinguishable when grown on media (Table 2: recipe in supplemental methods) (J. M. Cecala & Vannette, 2024). Each microbe was plated from freezer stock onto agar and incubated for at least 72 h of growth at 25 ℃.  Then suspensions of each of the five species were adjusted to approximately 10⁶ cells using a hemocytometer and aliquots stored at -80 ℃ (Cecala & Vannette, 2024). To create inocula the day of the field experiment, aliquots from each of the five species were thawed, combined, and brought up to ~100µL with the freezer stock solution to yield an inoculum containing roughly 10⁴ cells/µL of each of the five species, or equivalently 5 x 10⁴ total cells/µL. The microbial inoculum was kept on ice. Each of the species was proven to be successfully culturable from the individually aliquoted inocula on media plates prior to use in experiments. To manipulate the concentration of hydrogen peroxide in living flowers, we prepared a solution of the hydrogen peroxide-generating enzyme, glucose oxidase (GOX) (isolated from Aspergillus niger, Type VII, ≥1000,000 units/g, Sigma-Aldrich) at a concentration of 0.32 units of enzyme/μL in 40% sugar solution (80% glucose: 20% sucrose), immediately prior to beginning the experiment. This solution, as well as a boiled (inactive, boiled for 2 hr) glucose oxidase (B-GOX) solution at an equivalent concentration to act as a negative control, were kept on ice. This strategy was advantageous because (1) the nectar redox cycle nectarin protein NEC5 identified in Nicotiana is a glucose oxidase and produces hydrogen peroxide via the same mechanism (C. Carter & Thornburg, 2004), (2) hydrogen peroxide can rapidly degrade, so this enzymatic method allows for a renewal of hydrogen peroxide through the activity of the GOX, and (3) a previous study showed success of this method and enzyme concentration in living Nicotiana nectar (Bezzi et al., 2010).

The day prior to our field experiment, we covered closed flowers of Epilobium canum in two discrete patches with mesh bags to exclude pollinator visitation and microbial deposition (Francis et a,l 2023). Once all microbial inocula and enzyme solutions were prepared fresh each day, we identified bagged flowers that had opened and could be used in our experiment. For the first experimental treatment, we inoculated flowers with 10 μL (3.2 units) of active glucose oxidase enzyme and 1 μL of microbial inocula (n = 10, split between two clumps). Control flowers were inoculated with 10 μL of 40% sucrose solution and 1 μL of microbial inocula while additional flowers were inoculated with 10 μL boiled glucose oxidase solution and 1 μL of microbial inocula. Solutions were delivered into flowers using a micropipette and sterilized 10 μL tip. The inoculated flowers were marked with permanent marker, tagged, and mesh bags replaced. Each treatment had 10 replicates per trial, and two separate trials were conducted. The prepared microbial inoculum, the active glucose oxidase, boiled GOX, and control solution (sugar substrate) were all kept on ice while out in the field.

After 24 hours, flowers were harvested and brought back to the laboratory where microbial growth and composition were assessed. Occasionally, we would find that thrips and ants had made it inside of some bagged flowers and their presence noted. Nectar was extracted in a hood using glass capillary tubes, total nectar volume of each flower quantified (VWR International, Drummond Scientific Company, 50 μL), and diluted for plating. We also noted any bleached corolla or pistil tissue, likely caused by oxidative tissue damage from reactive oxygen derivatives of hydrogen peroxide. For floral nectar samples with greater than 5 μL extracted volume, we added 5 μL of nectar into 45 μL of DPBS buffer. For extracted nectar samples of less than 5 μL, 50 μL DPBS buffer was added to the full amount of nectar sample. We vortexed and centrifuged the diluted samples, then plated 15 μL of each sample on YM, TSA, and MRS plates and spread using sterile glass beads. Plates were incubated at 28 ℃ for 7 days. CFUs were identified by microbial taxa based on morphological features and counted for each plate (Cecala & Vannette, 2024). For plates where the CFUs were too numerous/dense to count with confidence, we overlaid a transparent grid and tallied CFU subsamples in at least six 0.1 or 1.0 cm² squares, which we used to estimate the total CFUs per plate for a given microbe. If CFU’s were too dense for even this method, having formed a lawn, we assigned it the highest countable density from another plate of the same microbe type.

Using 5 μL undiluted aliquots of extracted nectar above, we quantified peroxide concentrations with mid (previously used) and high range (0, 100, 200, 400 ppm) peroxide test strips (WaterWorks^TM Mid Range Peroxide Check, Industrial Test Systems Inc., Rock Hill, SC), (Bartovation Peroxide Test Strips 0-400 ppm, Bartovation, Queens, NY) to validate enzyme effects on peroxide concentration.

For some of the samples we also used a handheld refractometer (Bellingham + Stanley, Kent, UK) to measure Brix (Supplemental Figure S2).

Aim 5: How does community vs individual growth context impact microbial growth?

To explore whether co-growth of microbes influences microbial tolerance to hydrogen peroxide, we compared the growth of microbe isolates individually and in communities at different concentrations of hydrogen peroxide, in sterile artificial nectar solution to mimic floral nectar conditions (see Aim 5 recipe in supplemental methods). Using the same microbial isolates previously used in the field experiment of Aim 4 (Table 2), we created a community suspension of all five taxa as well as suspensions for each individual taxa. To create a community inoculum the day of the in vitro lab experiment, we combined aliquots from each of the five species as described previously. To make microbial isolate suspensions, an aliquot from a given species containing roughly 10⁴ cells/µL was brought up to a volume of 100 μL with freezer stock solution. We prepared hydrogen peroxide stock solutions through a set of serial dilutions into the artificial nectar media with final concentrations of 0, 10, 50, 100, and 2000 μM.

For each replicate, we added 39 μL of artificial nectar, 10 μL of hydrogen peroxide treatment and 1 μL of microbial suspension for a total volume of 50 μL in PCR tubes, which we selected based on nectar volumes often extracted from Epilobium canum during prior field trials. Each treatment combination--six different microbial inocula (community of all five microbes and each microbe individually) x five hydrogen peroxide treatment concentrations--was performed in triplicate. We vortexed and centrifuged the tubes and then incubated at 28 ℃ for 24 hours, protected from UV light with aluminum foil. After the 24 hours, we assessed microbial growth as described in Aim 4 above. We also determined the concentration of hydrogen peroxide using 5 μL of each solution applied to mid-range peroxide colorimetric test strips. By comparing control, individual, and community samples we assessed how peroxide degradation (Supplemental Figure S3) and if microbial growth and community membership altered hydrogen peroxide concentration.

Statistical Analysis:

We conducted all statistical analyses in R (R Core Team, 2021) and used ggplot (Wickham, 2016). In Aim 1, to estimate plant phylogenetic relationships among the sampled plant species, we used the function ‘phylo.maker’ in package V.PhyloMaker2 (Jin & Qian, 2022) using the reference vascular plant mega-phylogeny GBOTB.extended.TPL. Using this tree and the function ‘multiPhylosignal’ in the package picante (Kembel et al., 2010) we tested for a phylogenetic signal of nectar peroxide concentration.

To test whether defensive growth hormones altered concentrations of hydrogen peroxide in Peritoma arborea nectar or nectar volume (Aim 2), we used one-way ANOVA in the package car. We obtained F- and P values using Kenward-Roger degrees of freedom. For Aim 3, we tested whether microbial growth (measured by OD₆₀₀) differed among concentrations of hydrogen peroxide according to microbial identity or isolation source and their interaction using type III ANOVA and the drop1 function. We performed post-hoc analysis using the ‘TukeyHSD’ function to assess the source of variation in the predictor variable, microbial isolation source. For Aim 4, we compared how enzyme treatments influenced floral nectar peroxide concentration using an ordinal model (because test strip data are ordinal) implemented using the function ‘clm2’ in the ordinal R package, as well as effects on total microbial community composition, and growth of individual species using separate ANOVAs. We used the peroxide test strip designations per product instructions to assign different peroxide levels used in the analysis and visualization. To evaluate which enzyme treatments contributed to differences observed in total microbial community composition and individual microbe growth, we performed a post-hoc Tukey analysis.

In Aim 5, we examined the impact of community vs isolate growth context, hydrogen peroxide concentration and their interaction on microbial growth using III ANOVAs, with separate models for each tested microbe. Additionally, we examined if measured peroxide concentration varied with microbial growth context, starting hydrogen peroxide treatment concentration and their interaction using ordinal models (clm2) as above, with significance assessed using c² tests. To evaluate which microbe taxa contributed to differences observed in the final environmental hydrogen peroxide concentration, we performed a post-hoc analysis using ‘TukeyHSD’.  Finally, to test if microbial community composition (as Bray-Curtis dissimilarity) that formed after 24 hours differed among hydrogen peroxide treatments, we used function ‘adonis’ in package vegan (Oksanen et al., 2007) to perform a multivariate analysis of variance (PERMANOVA). Microbial community composition was visualized using non-metric multidimensional scaling (NMDS) ordination (Supplemental Figure S4).