Plant-systemic neonicotinoid (NN) insecticides can exert non-target impacts on organisms like beneficial insects and soil microbes. NNs can affect plant microbiomes, but we know little about their effects on microbial communities that mediate plant-insect interactions, including nectar-inhabiting microbes (NIMs). Here we employed two approaches to assess impacts of NN exposure on several NIM taxa. First, we assayed in vitro effects of six NN compounds on NIM growth using plate assays. Second, we inoculated a standardized NIM community into nectar of NN-treated canola (Brassica napus) and assessed survival and growth after 24 hours. With few exceptions, in vitro NN exposure tended to decrease bacterial growth metrics. However, the magnitude of decrease and the NN concentrations at which effects were observed varied substantially across bacteria. Yeasts showed no consistent in vitro response to NNs. In nectar, we saw no effects of NN treatment on NIM community metrics. Rather, NIM abundance and diversity responded to inherent plant qualities like nectar volume. In conclusion, we found no evidence NIMs respond to field-relevant NN levels in nectar within 24 h, but our study suggests that context, specifically assay methods, time, and plant traits, is important in assaying effects of NN on microbial communities.

EXPERIMENTAL PROCEDURES 

Overview of experiments

To determine if and how exposure to NN residues affects the growth and abundance of NIMs, we conducted two separate experiments. First, we grew seven microbes as pure (single-species) cultures in artificial broths spiked to contain one of four concentrations of each of six NN compounds. We monitored microbial growth over 72 h and calculated maximum growth rate (r) and carrying capacity (K). Second, we inoculated a standardized community of four microbe taxa (a subset of those from the first experiment) into floral nectar of greenhouse-grown, potted canola plants. Plants had been treated with either low or high doses of two commercial NN formulations. Plants were also irrigated at either a low or high rate to monitor for any effect of water availability on nectar characteristics, NN translocation, or microbe community metrics.

In vitro plate reader experiment

We selected seven microbe species to assay growth in the presence of set concentrations of six major NN compounds. The selected species occur in floral nectar, are relatively well-studied, and in some cases have been found to influence behaviors of flower-visiting animals or plant reproduction (Vannette, 2020) (Table 1). All microbe strains were sourced from suspensions in glycerol and sucrose stocks at -80 °C. Prior to each plate reader run, we streaked stock on agar media and incubated plates at 25 °C for 72 h. Yeasts were streaked on yeast media (YM) agar and bacteria on tryptic soy (TS) agar, except for Apilactobacillus micheneri, which was streaked on de Man, Rogosa and Sharpe (MRS) agar + 2% m/v fructose (Vuong and McFrederick, 2019). We also prepared liquid broth analogs of each media type (omitting agarose) for use in well plates. All agar and broth media contained 0.1% v/v of a solution of either chloramphenicol (antibacterial; in yeast media) or cycloheximide (antifungal; in bacterial media) in methanol (10% m/v).

We acquired PESTANAL^® analytical standards (Millipore Sigma, St. Louis, MO) for each of six major NN compounds (Table 2). We created a separate stock solution for each compound in sterile distilled water by adding 2 mg of the respective compound to 100 mL of sterile distilled water, yielding a concentration of 2 x 10⁴ ppb or µg L^-1. Stock solution bottles were wrapped in foil and kept in a container at 5 °C to prevent photodegradation of NNs (Borsuah et al., 2020). These six stock solutions were used as spikes (as in Meikle et al., 2022) to achieve specific concentrations of each compound in the corresponding broth for each microbe assay (see below; see also Supplementary Material).

To determine if NN type and concentration influence the growth of microbes in vitro, we conducted successive runs using two spectrophotometer microplate readers (models SYNERGY HTX and 800 TS; Agilent, Santa Clara, CA) simultaneously, using a consistent plate layout for all runs (Fig. S1). We grew each microbe strain in pure culture in two 96-well plates, run at the same time in the two readers, with each run comprising one microbe. Prior to a run, we prepared separate solutions of the appropriate broth for the focal microbe spiked to contain either 1000 ppb, 100 ppb, 10 ppb, or a no-NN control of each of the six NN compounds (see Table 2 for the ecological context of these concentrations). This resulted in a total of eight replicate wells for each of these 24 treatments per experimental run. For the inoculum, we prepared a suspension of the focal microbe by scraping a 2-mm bolus from agar into 3.5 mL of the appropriate broth and vortexing. Per treatment, we inoculated six of the eight wells containing 180 µL of sterile broth with 20 µL of inoculum. In the remaining wells, we prepared 200 µL of non-inoculated broth to monitor for contamination across treatments. Immediately after inoculation, plates were sealed with a lid and wax film to minimize evaporation, then loaded into readers and incubated continuously at 25 °C (30 °C for Apilactobacillus; McFrederick et al., 2017) for 72 hours. To estimate changes in cell concentration over time, readers recorded the optical density at λ=600 nm (OD₆₀₀) of all wells after shaking (6-mm diameter at 6 Hz) every 15 min for 72h.

Statistical analysis

We performed all statistical analyses in R (R Core Team, 2023). We analyzed microbial growth using the function ‘SummarizeGrowthByPlate’ in the package growthcurver (Sprouffske, 2020), which fits a logistic growth equation to OD vs. time, and estimates maximum growth rate (r) and maximum OD (K) for each well over the 72-h period. OD values for inoculated wells were blank-corrected by subtracting the mean OD of all non-inoculated wells at each time point. Using the package lme4 (Bates et al., 2015), we performed linear mixed-effect models (LMMs) for each microbe taxon, with either r or K as dependent variables, and NN type, concentration, and their interaction as independent variables, and plate reader as a random intercept effect. We obtained type III sums of squares and Kenward-Roger degrees of freedom using the ‘Anova’ function in the package car and checked all mixed models for multicollinearity (VIF > 2.0) and normality of residuals.

In planta greenhouse experiment

To assess how NN application and plant water availability impact a microbe community in floral nectar, we conducted an experiment in a glass greenhouse on the University of California, Davis campus (USA: California: Yolo County; 38.5361 °N, -121.7475 °W). We obtained seeds of spring canola, Brassica napus L. ‘CP930RR’ (Land O’Lakes, Inc., Arden Hills, MN, USA), not previously treated with any chemicals. We sowed seeds in 60 2.5-gallon pots (25.7 cm x 23.2 cm) of “UC Mix C” soil (1:1 peat and sand) in three cohorts (18 February, 4 March, and 18 March 2022). After germination, we culled plants to six per pot.

We applied commercial NN formulations to pots according to label specifications once the first buds were produced in a cohort, around 14-18 days before inoculations. Pots were treated with either a high dose (25 mg active ingredient per pot, or 4.2 mg per plant), a low dose (2.5 mg AI per pot, 0.42 mg per plant), or a no-dose control of the respective formulation. For reference, a commercially treated seed typically contains from 0.2 to 1.3 mg of AI (Goulson, 2013; Wood and Goulson, 2017). We included two different NNs in the experiment, applied singly: imidacloprid, as Marathon^® 1% Granular (OHP, Bluffton, SC, USA), and dinotefuran, as Safari^® 20 SG (Valent U.S.A. LLC, San Ramon, CA, USA). We selected these two formulations based on multiple factors, including usage in agricultural settings, differences in solubility and leaching potential (Bonmatin et al., 2015), and approved usage in potting media and greenhouse settings.

Each pot was additionally assigned in a crossed fashion to one of two irrigation treatments. Irrigation rate was controlled by inserting one high- or low-flow irrigation spike (Primerus Products, Encinitas, CA, USA) per pot, each connected to one central irrigation line (as in Cecala and Wilson Rankin, 2021). A high-flow spike emitted 2.7 times the water (0.61 L/min) as a low-flow spike (0.23 L/min). All pots were automatically irrigated simultaneously over the soil surface at 07:00 h daily for 60 s, or up to 120 s on hotter days to prevent wilting.

Experimental flowers were selected and inoculated with a standardized microbe community containing 10⁴ cells µL^-1 of each of a subset of four species from the plate reader experiment (Table 1) in a 20% v/v glycerol, 20% m/v sucrose stock (4 x 10⁴ total cells µL^-1). Microbes were stored in pure culture aliquots at -80 °C that were thawed and mixed the morning of each day of inoculations. All four microbes in the inoculum were confirmed to be successfully culturable on agar media prior to inoculations. Using a micropipette, we delivered 0.5 µL of inoculum into each lateral nectary (total 1 µL per flower) of newly opened canola flowers and tagged them.

After 24 h, we excised all inoculated flowers using sterilized forceps and transported them to the laboratory. Inside a laminar flow hood, we used 10-µL microcapillary tubes to remove all nectar from each inoculated flower, estimated nectar volume from length of the fluid column, and expelled it into individual strip tubes. We also measured whole flower mass after nectar removal. We added 50 µL of Dulbecco’s phosphate-buffered saline (DPBS 1x) to each nectar sample, vortexed tubes, then plated a 15 µL aliquot of the solution onto each of YM, TS, and MRS agar in 100 x 15 mm Petri dishes using plating beads. We then incubated plates for 7 days at 25 °C and stored them at 5 °C. We tallied CFUs per plate and classified them into morphotypes. A representative CFU of each morphotype was sequenced by single-gene PCR and barcoded using NCBI Nucleotide BLAST (“blastn”; blast.ncbi.nlm.nih.gov).

To determine the concentration of NN residues in canola nectar, we collected separate samples of nectar from newly opened, uninoculated flowers. To obtain sufficient volumes for residue analysis, we pooled up to 12 flowers from plants in the same pot. Samples were kept at -20 °C until analysis, at which point we diluted them in ultrapure water and analyzed them via electrospray ionization (ESI) LC-MS (Martel et al., 2013) on an Orbitrap machine. For LC-MS analysis, 5µL of samples were injected onto a Thermo C18 Accucore column (2.1 mm x 50 mm). A standard reverse phase gradient (solvent: Optima grade water and acetonitrile (Fisher, MS grade), plus 0.1% formic acid) was run over 12 minutes at a flow rate of 250 µL min^-1 and the eluent was monitored for positive ions by a Thermo Scientific Q-Exactive HF operated in profile mode. Source parameters were 4kV spray voltage, capillary temperature of 275 °C and sheath gas setting of 20. Spectral data were acquired at a resolution setting of 60,000 FWHM with the lockmass feature, which typically results in a mass accuracy < 2 ppm. Analytical standards of imidaclpoprid and dinetofuran (5 ppm in Milli-Q^® ultrapure water) were dissolved in methanol and diluted into mobile phase for quantitation. Standard curves were run for every set of samples, from which we back-calculated sample residue concentrations. Extracted Ion Chromatograms (XICs) utilizing a 10 ppm mass window for each of the compounds were used for quantitation.

Statistical analysis

To determine how irrigation level and NN application impacted our model floral microbe community, we constructed linear mixed models in lme4 with nectar volume, flower mass, CFU abundance (summed across the three media types), CFU density (per µL nectar), and CFU Shannon diversity per flower as dependent variables. As independent variables, we included irrigation rate, NN dose, their interaction, and NN type (nested within dose). To test if differences in microbial community composition (as Bray–Curtis dissimilarity) were related to treatments, we also performed a permutational multivariate analysis of variance (permANOVA) using function ‘adonis’ in the package vegan (Oksanen et al., 2020). Multivariate homogeneity of dispersions within independent variable groups were checked with the PERMDISP2 procedure using function ‘betadisper’ in vegan. We visualized distance between samples with non-metric multidimensional scaling using function ‘metaMDS’ and confirmed ordination stress was sufficiently low in k=2 dimensions using function ‘dimcheckMDS’. All figures were created using package ggplot2 (Wickham, 2016).