Caffeine Consumption Helps Honey Bees Fight a Bacterial Pathogen

ABSTRACT Caffeine has long been used as a stimulant by humans. Although this secondary metabolite is produced by some plants as a mechanism of defense against herbivores, beneficial or detrimental effects of such consumption are usually associated with dose. The Western honey bee, Apis mellifera, can also be exposed to caffeine when foraging at Coffea and Citrus plants, and low doses as are found in the nectar of these plants seem to boost memory learning and ameliorate parasite infection in bees. In this study, we investigated the effects of caffeine consumption on the gut microbiota of honey bees and on susceptibility to bacterial infection. We performed in vivo experiments in which honey bees, deprived of or colonized with their native microbiota, were exposed to nectar-relevant concentrations of caffeine for a week, then challenged with the bacterial pathogen Serratia marcescens. We found that caffeine consumption did not impact the gut microbiota or survival rates of honey bees. Moreover, microbiota-colonized bees exposed to caffeine were more resistant to infection and exhibited increased survival rates compared to microbiota-colonized or microbiota-deprived bees only exposed to the pathogen. Our findings point to an additional benefit of caffeine consumption in honey bee health by protecting against bacterial infections. IMPORTANCE The consumption of caffeine is a remarkable feature of the human diet. Common drinks, such as coffee and tea, contain caffeine as a stimulant. Interestingly, honey bees also seem to like caffeine. They are usually attracted to the low concentrations of caffeine found in nectar and pollen of Coffea plants, and consumption improves learning and memory retention, as well as protects against viruses and fungal parasites. In this study, we expanded these findings by demonstrating that caffeine can improve survival rates of honey bees infected with Serratia marcescens, a bacterial pathogen known to cause sepsis in animals. However, this beneficial effect was only observed when bees were colonized with their native gut microbiota, and caffeine seemed not to directly affect the gut microbiota or survival rates of bees. Our findings suggest a potential synergism between caffeine and gut microbial communities in protection against bacterial pathogens.

herbivore defense (2,11), depending on plant tissue location and concentration. For example, caffeine has been detected in the nectar and pollen of Citrus and Coffea species at concentrations up to 1.1 mM (13) and 6.7 mM (9), respectively. These concentrations are much lower than the ones found in leaves and seeds of Coffea, which have been reported to be as high as 124 mM (8). Caffeine is bitter tasting to several animals, but bees are not usually deterred by caffeine in nectar, and actually they prefer sucrose solutions containing naturally occurring concentrations of caffeine (14)(15)(16)(17). However, when caffeine is present in doses higher than 1 mM, bees can be deterred by the bitter taste (5).
Plant secondary metabolites in nectar and pollen are also associated with parasite tolerance (31)(32)(33). Caffeine, for example, can diminish the severity of fungal disease by reducing spore density of the microsporidian Vairimorpha spp. (previously known as Nosema spp. [34]) and increasing longevity in both honey bees (31,35) and bumble bees (36), although this is not observed for the trypanosomatid parasite Crithidia bombi (37). Low doses of caffeine also impact virus infection (38,39), by increasing the expression of immunity genes and reducing the levels of Deformed wing virus in honey bees (38) and increasing survival rates of honey bees infected with Israeli acute paralysis virus (39).
The effects of caffeine on bacterial growth have also been tested, with several bacteria being able to grow in the presence of 13 mM caffeine (40). However, bacterial inhibition is observed at higher concentrations, such as 50 mM (41). Interestingly, the gut microbiota of a major insect pest can mediate caffeine detoxification as a mechanism to deal with the high concentrations of caffeine found in Coffea plants (42), and there is some evidence that caffeine consumption can initially increase the diversity and abundance of the honey bee gut microbiota, which is followed by a temporal stabilization of the microbiota (43).
In this study, we investigated the effects of naturally relevant concentrations of caffeine on survival rates, gut microbial communities, and bacterial infection susceptibility of the Western honey bee, Apis mellifera. We performed experiments with hive bees and with age-controlled bees, lacking or containing a native microbiota, to investigate how caffeine consumption and the gut microbiota protect bees against an opportunistic bacterial pathogen, Serratia marcescens, known to cause disease and decrease survival rates of honey bees at both larval (44) and adult stages (45,46).

RESULTS
To investigate the effects of caffeine consumption on the gut microbiota and on the survival rates of honey bees, we performed three trials of a two-step in vivo experiment in which bees were first treated with caffeine and then challenged with S. marcescens. In the first trial, we used bees of various ages collected from inside a hive. During the caffeine treatment step, we did not observe changes in survival rates between control and caffeine-treated bees after 7 days (analysis of variance [ANOVA] test, COXME model; x 2 = 8.61, df = 2, P = 0.91, N = 348), and survival rates were higher than 90% for all groups (Fig. 1A). However, we observed significant changes in survival rates between groups during the Serratia challenge step (ANOVA test, COXME model; x 2 = 30.25, df = 5, P = 1.32e-05, N = 279). Control and caffeinetreated bees not challenged with Serratia exhibited higher survival rates than bees challenged with Serratia after 7 days (Fig. 1B, Table 1). Interestingly, 1 mM caffeine-treated bees challenged with Serratia exhibited nonsignificantly higher survival rates (;74%) than 0.1 mM caffeine-treated bees (;57%) or control bees (;48%) challenged with Serratia (Fig. 1B, Table 1).
Also, control and caffeine-treated bees not challenged with Serratia did not exhibit significant changes in survival rates compared to 1 mM caffeine-treated bees challenged with Serratia ( Fig. 1B, Table 1), suggesting a potential benefit of 1 mM caffeine consumption in protection against this opportunistic bacterium.
To further investigate the potential contribution of caffeine treatment in protection against Serratia, we performed two additional trials with age-controlled, newly emerged bees, in order to minimize the effects of variable age in our results (Fig. 2). In these replicate trials, we included two major groups of microbiota-deprived and microbiota-colonized bees, to account for and understand the effects of the microbiota during the caffeine treatment and the Serratia challenge. In these trials, the caffeine treatment step affected survival rates in microbiota-deprived groups (ANOVA test, COXME model; x 2 = 7.76, df = 2, P = 0.02, N = 919) ( Fig. 2A), but not in microbiota-colonized groups (ANOVA test, COXME model;

Caffeine Prevents Bacterial Infection in Honey Bees
Microbiology Spectrum x 2 = 3.69, df = 2, P = 0.16, N = 990) (Fig. 2C). More specifically, microbiota-deprived bees treated with 1 mM caffeine exhibited lower survival rates than control bees ( Fig. 2A, Table 2). However, during the Serratia challenge step, we observed significant treatment effect on survival rates for both microbiota-deprived (ANOVA test, COXME model; x 2 = 136.90, df = 5, P = 2.20e-16, N = 747) ( Fig. 2B) and microbiota-colonized groups (ANOVA test, COXME model; x 2 = 142.33, df = 5, P = 2.20e-16, N = 766) (Fig. 2D). In the microbiota-deprived group, Serratia challenge, but not caffeine treatment, significantly impacted survival rates, with pathogen-challenged bees showing lower survival, as expected ( Fig. 2B, Table 2). In the microbiota-colonized group, on the other hand, both caffeine treatment and Serratia challenge affected survival rates. Bees treated or not with caffeine and not exposed to Serratia exhibited similar survival rates to one another and higher survival than control bees or bees treated with 0.1 mM caffeine and exposed to Serratia (Fig. 2D, Table 3). Interestingly, 1 mM caffeine-treated bees exposed to Serratia exhibited similar survival rates as control bees not exposed to Serratia and significantly higher survival rates than control bees or 0.1 mM caffeine-treated bees exposed to Serratia (Fig. 2D, Table 3). Although not statistically significant, there was a trend in increased survival rates for microbiota-colonized bees treated with caffeine compared to control bees when the pathogen was absent (Fig. 2D, Table 3).

DISCUSSION
Plant-derived metabolites have long been used by humans as prophylactic or therapeutic medication for the treatments of diseases (50,51). This is also true for other animals, including insect pollinators (52,53). Bees, for example, collect metabolites from plant resins, nectar, and pollen as a self-medication mechanism against parasites (6,54). In our study, we were particularly interested in investigating the effects of caffeine on honey bee susceptibility to bacterial infection. We found that caffeine consumption helped honey bees to better survive infection by the opportunistic bacteria S. marcescens. This finding contributes to the understanding of the beneficial effects of low doses of caffeine in animal health and protection against disease. Caffeine, as a neuroactive alkaloid, can affect both the behavior and physiology of bees (5), but it can also provide a beneficial role in protection against parasites, such as Vairimorpha spp. (31,35,36), and viruses, such as Deformed wing virus and Israeli acute paralysis virus (38,39), increasing the survival rates of infected bees. S. marcescens is an opportunistic bacterium of honey bees that has been reported to cause diseases at both larval (44) and adult stages (45,46). This bacterium also increases mortality rates of adult bees exposed to antibiotics (55). Our findings, showing that caffeine consumption improves survival rates of honey bees following exposure to S. marcescens, raise the question of whether caffeine can also provide a similar benefit to other bees or honey bees exposed to other bacterial pathogens, such as Bacillus pulvifaciens, which causes powdery scale, and Spiroplasma spp., involved in May disease (56,57). Moreover, the most discussed bacterial infections in honey bees are associated with brood diseases, primarily involving American and European foulbrood, which are caused by Paenibacillus larvae and Melissococcus plutonius, respectively (58,59), and future studies should continue investigating the roles of specific plant-derived metabolites, such as caffeine, in the treatment of bacterial infections affecting brood development (60).
As important as the finding of protection against pathogens is the finding that exposure to low doses of caffeine has no evident fitness cost to bees. Survival rates of caffeineexposed bees usually increase (22) or remain the same (30), at least under concentrations found in nectar. In our study, caffeine exposure did not affect survival rates of honey bees, although we observed a trend toward increased survival rates. However, caffeine, similar to other alkaloids, has dose-dependent effects on pollinator behavior and activity, and exposure to concentrations similar to what is found in leaves and seeds can be toxic and negatively impact animal health (61). Since bees collect plant resins to make propolis (54), concentrations of caffeine in these resources should be also taken into consideration. Indeed, high concentrations of caffeine may act as a deterrent and may even be lethal to bees (62). In other animals, such as caterpillars, parasitized individuals prefer diets rich in toxic plant metabolites, which improve their survival rates, though this reduces fitness of uninfected individuals (63,64).
We investigated the effects of caffeine on gut microbial communities of honey bees and did not find an impact on the abundance or composition of the main bacterial species in the bee gut, suggesting that caffeine has little or no impact on the honey bee gut microbiota. This finding is important because the gut microbiota plays important roles in   (69,70), and protection against pathogens (71)(72)(73). Furthermore, perturbations of the gut microbiota can lead to immune system dysregulation (74,75) and increased susceptibility to infection (55,74,76). However, a previous study found an impact of caffeine consumption on the diversity and abundance of the honey bee gut microbiota, but this impact only occurred during the first few days of exposure (43). Since we only checked potential microbiota changes after a week of exposure and only used 16S rRNA amplicon sequencing, we may have missed a potential effect in the beginning of treatment or on the fine-scale strain diversity of the honey bee microbiota.
We also performed experiments with microbiota-deprived bees and microbiotacolonized bees, and the protective effect of caffeine against S. marcescens was only observed in microbiota-colonized bees. These results demonstrate that an intact microbiota plays a crucial role in protection against pathogens (71,73), as has been observed in cases of dysbiosis in honey bees (55,74,77). The molecular mechanisms behind the synergistic effect of caffeine and the gut microbiota in protection against S. marcescens deserves further investigation. They may reflect interactions between caffeine and specific members of the gut microbiota. In some organisms, such as the berry borer (or coffee borer beetle), caffeine is metabolized by the microbiota, which allows this pest to feed on and damage coffee beans (42). Further studies are needed to investigate whether members of the bee gut microbiota metabolize caffeine and whether this underlies the beneficial effects described here and in previous studies. Moreover, direct effects of caffeine on S. marcescens cannot be dismissed, as previous studies found that high concentrations of caffeine inhibited this opportunistic pathogen in vitro (78). It is also known that different bacterial species can metabolize caffeine (79), including some strains of S. marcescens (80).
Conclusions. Our findings demonstrate that caffeine consumption does not affect gut microbial communities or survival rates of honey bees. Interestingly, specific concentrations of caffeine can improve survival rates of microbiota-colonized honey bees exposed to the bacterial pathogen S. marcescens compared to bees not exposed to caffeine. Since this effect is not observed for microbiota-deprived bees, the establishment of the native gut microbiota must be crucial for caffeine to exert a protective effect against Serratia in bees. These findings expand previously established evidence that caffeine ingestion increases honey bee protection against natural enemies, such as fungal parasites and viruses, as our findings showed protection against an opportunistic bacterial pathogen.

MATERIALS AND METHODS
Honey bee rearing. Honey bees (A. mellifera) were obtained from outdoor hives kept on the rooftop of J. T. Patterson Laboratories Building at the University of Texas at Austin (latitude, 30.287913; longitude, 97.736183). These hives are self-sufficient.
Treatment preparation. Caffeine standard (99.7% purity, lot number W24A011) was purchased from Alfa Aesar by Thermo Fisher Scientific. Stock solutions of 10 mM caffeine were prepared with heated molecular biology-grade water and used to prepare working solutions with final concentrations of 0.1 mM or 1 mM caffeine in filter-sterilized 0.5 M sucrose syrup.
Caffeine exposure and Serratia challenge experiments. (i) First trial. Approximately 360 honey bee workers (not age controlled) were collected from inside a hive in fall 2017. They were brought into the lab, immobilized at 4°C, and randomly transferred to single-use cages constructed from plastic cups (81). Bees were in groups of 28 to 30 bees in each of 12 cup cages. Cup cages were split into three treatment groups, which were provided (i) sterile sucrose syrup, (ii) 0.1 mM caffeine in sterile sucrose syrup, or (iii) 1 mM caffeine in sterile sucrose syrup. Survival rates were monitored for 7 days, and dead bees were removed in a daily census. After that, treatments were removed, and 16 bees were sampled from each treatment group and preserved at 280°C. The remaining bees were briefly immobilized with carbon dioxide, mixed according to treatment group, and then transferred to new cup cages in groups of 15 to 16 bees for a total of six cup cages per treatment group. Finally, each treatment group was divided into two subgroups: one subgroup was used as a control and fed sterile sucrose syrup, whereas the other group was challenged with the opportunistic pathogen Serratia marcescens strain kz19 in sucrose syrup. Survival rates were monitored for 7 days, and dead bees were removed in a daily census.
(ii) Second trial. Thousands of late-stage pupae, with pigmented eyes but lacking movement, were removed from brood frames from a different hive in fall of 2020. Pupae were transferred to clean plastic bins and placed in an incubator at 35°C and ;60% relative humidity to simulate hive conditions until emerging as adults. Approximately 1,200 healthy newly emerged workers (age controlled) were randomly transferred to cup cages in groups of 30 to 35 bees for a total of 36 cup cages. These cup cages were first split into two major

Caffeine Prevents Bacterial Infection in Honey Bees
Microbiology Spectrum groups. Bees from one major group were allowed to acquire their normal microbiota by addition of a suspension of freshly prepared gut homogenate (see below) from hive bees to sterile pollen, while bees from the other major group were prevented from acquiring microbiota and provided only sterile pollen. Then, cup cages from each major group were divided into three treatment groups and fed sterile sucrose syrup, 0.1 mM caffeine in sterile sucrose syrup, or 1 mM caffeine in sterile sucrose syrup. Survival rates were monitored for 7 days, and dead bees were removed in a daily census. Finally, each treatment group was divided into two subgroups. One subgroup was used as a control and provided only sterile sucrose syrup, whereas the other subgroup was challenged with the opportunistic pathogen S. marcescens kz19 in sucrose syrup. Survival rates were monitored for an additional 7 days, and dead bees were removed in a daily census.
(iii) Third trial. Thousands of late-stage pupae were removed from brood frames from a different hive in fall of 2022 and managed similar to the second trial. Approximately 960 healthy newly emerged workers were randomly transferred to cup cages in groups of 20 bees for a total of 48 cup cages. Cup cages were split into two major groups, microbiota-deprived and microbiota-colonized groups, then split into three treatment groups which were fed sterile sucrose syrup, 0.1 mM caffeine in sterile sucrose syrup, or 1 mM caffeine in sterile sucrose syrup. Survival rates were monitored for 7 days, and dead bees were removed in a daily census. Cup cages in which we observed high mortality rates due to syrup leakage or feeding tube clogging were removed from the following step and analyses. Then, cup cages from each treatment group were divided into two subgroups to be used as controls or challenged with S. marcescens kz19. Survival rates were monitored for an additional 7 days, and dead bees were removed in a daily census.
Preparation of gut homogenates. For the second and third trials, half of the bees were allowed to acquire a normal microbiota by providing gut homogenate solutions mixed with irradiated pollen. Gut homogenate solutions were prepared by aseptically pulling out guts from healthy workers from the same hive and homogenizing them with equal proportions of 1Â phosphate-buffered saline (PBS) and sterile sucrose syrup. For each cup from the microbiota-colonized group, 200 mL of gut homogenate solution was mixed with irradiated pollen and provided in small troughs. For each cup from the microbiota-deprived group, only 200 mL of 1:1 PBS-sterile sucrose syrup solution was mixed with irradiated pollen.
Preparation of Serratia marcescens suspension. For each trial, a S. marcescens kz19 suspension in sucrose syrup with an optical density (OD) of 0.5 was prepared and provided in feeding tubes to half of the cup cages in each treatment group. S. marcescens kz19 was grown in LB broth at 37°C overnight. Then, the OD was measured at 600 nm, and bacterial cells were diluted to an OD of 0.5 in 1:10 PBS-sucrose syrup and provided to the cup cages from the Serratia challenge group. For consistency, cup cages from the control group were provided 1:10 PBS-sucrose syrup.
DNA extraction. Honey bees sampled from the first trial (16 bees from each group, 48 bees in total) were dissected with sterilized forceps under aseptic conditions, and DNA was extracted from individual bee guts using a previously described protocol (82), with some adaptations. Briefly, guts were homogenized with 100 mL of cetyl triethylammonium bromide (CTAB) buffer (0.1 M Tris-HCl [pH 8.0], 1.4 M NaCl, 20 mM EDTA, and 20 mg/mL cetrimonium bromide), resuspended in additional 600 mL of CTAB buffer and 20 mL of proteinase K solution (0.1 M Tris-HCl, 26 mM CaCl 2 , 50% glycerol, and 20 mg/mL proteinase K), and transferred to a capped vial with 0.5 mL of 0.1 mm zirconia beads (BioSpec Products Inc.). After adding 2 mL of 2-mercaptoethanol and 2 mL of RNase A cocktail (Invitrogen Corp.), samples were bead-beated for 2 Â 2 min. Samples were digested overnight at 50°C and then mixed with 750 mL of phenol-chloroform-isoamyl alcohol (25:24:1, pH 8.0). Samples were inverted five times and centrifuged for 15 min at 4°C and 14,000 rpm. The aqueous layer was transferred to a new vial, and DNA was precipitated at 220°C for 30 min with 700 mL of isopropanol and 70 mL of 3 M NaOAc (pH 5.4). Precipitated samples were centrifuged for 30 min at 4°C and 14,000 rpm, and the supernatant was removed. DNA pellets were washed with 1 mL of cold 75% ethanol and centrifuged for additional 3 min at 4°C. After removing the ethanol wash, and the DNA pellets were dried at room temperature for ;30 min and then resuspended in 50 mL of water. Final DNA samples were stored at 220°C. qPCR analysis. DNA samples from the first trial (48 in total) were 100-fold diluted to be used as the templates for quantitative PCR (qPCR) analysis. For these measures, 10-mL triplicate reaction mixtures were carried out on 96-well plates on an Eppendorf Mastercycler ep realplex instrument using 5 mL of iTaq Universal SYBR green Supermix (Bio-Rad Inc.), 0.05 mL of 100 mM forward and reverse primers (27F, 59-AGAGTTTGATCCTGGCTCAG-39; 355R, 59-CTGCTGCCTCCCGTAGGAGT-39), 3.9 mL of H 2 O, and 1.0 mL of diluted DNA. The cycling conditions consisted of an initial step of 95°C for 3 min, followed by 5 cycles of a three-step PCR (95°C for 5 s, 65 to 60°C for 15 s with decrease of 1°C per cycle, and 68°C for 20 s), then 35 cycles of a second three-step PCR (95°C for 5 s, 60°C for 15 s, and 68°C for 20 s). Total bacterial 16S rRNA gene copies were estimated by standard curves from amplification of the cloned target sequence in a pGEM-T vector (Promega), as described elsewhere (83).
16S rRNA amplicon sequencing. DNA samples from the first trial (48 in total) were used to investigate bacterial community profiling based on sequencing of the V4 region of the 16S rRNA gene, as amplified by PCR primers Hyb515F (59-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGGTGYCAGCMGCCGCGGTA-39) and Hyb806R (59-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGGACTACHVGGGTWTCTAAT-39). Amplification, library preparation, and sequencing (Illumina MiSeq 2 Â 250) were performed by the Genomic Sequencing and Analysis Facility at the University of Texas at Austin.
Bioinformatic analyses. Illumina sequence reads were processed in QIIME 2 version 2019.10 (84). Primer and adapter sequences were removed using the cutadapt plugin (85). Forward and reverse reads were joined and filtered to remove low-quality reads. Joined reads were truncated to a length of 230 bp and denoised, and chimeric reads were removed using the Deblur plugin (86). Taxonomy was assigned to amplicon sequence variants (ASVs) using the SILVA database in the feature-classifier plugin (87). When necessary, BLASTn searches were conducted against the NCBI database (November 2022). Reads with ,0.1% abundance were removed using the feature-table plugin, as were unassigned, mitochondrial, and chloroplast reads using the taxa filter-table plugin. A table of ASVs was generated to investigate changes in microbial abundance and composition between control and treatments (see the supplemental material).
Statistical analyses. Comparisons of survival rates between control and treatment groups during caffeine exposure and Serratia challenge were performed using mixed-effects Cox proportional hazards models, which were fitted using the function coxme in the R package coxme (88). For the first trial, treatment was considered a fixed effect, and bees within cup cages were considered random effects. The following formula was used: survival (start, stop, death) ; treatment 1 [1j(cage/bee)]. For the second and third trials, which were replicates and analyzed together, treatment was considered a fixed effect and bees within cup cages within colonies as random effects. The following formula was used: survival (start, stop, death) ; treatment 1 [1j(colony/cage/bee)]. Survival curves were estimated and plotted using the Kaplan-Meier method and the functions survfit and ggsurvplot implemented in the R package survminer (89). Statistical analyses were performed for the fitted models using the function ANOVA in the R package car (90). If significant, multiple pairwise comparisons were performed using the R package emmeans (91), which uses the Tukey method for P value adjustment.
Microbial diversity analysis was performed to investigate the effect of caffeine exposure on the microbiota of sampled honey bees. Nonmetric multidimensional scaling based on Bray-Curtis dissimilarity was plotted using the R package phyloseq (92). Statistical tests were performed using the Adonis method for permutational multivariate analysis of variance in the R package vegan (93).
Comparisons of changes in the abundance of gut bacteria between control and treatment groups were performed using the nonparametric Kruskal-Wallis test followed by Dunn's multiple-comparison test, if significant, in R version 3.5.2 (94).
Data availability. Sequencing data are available at NCBI BioProject PRJNA905699. Other data are included in the published article and its supplemental material file.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, XLSX file, 0.2 MB.

ACKNOWLEDGMENTS
Thanks to Eli Powell and Tyler De Jong for technical assistance and Kim Hammond for administrative support.
We declare we have no competing interests.