Tetracycline Exposure Alters Key Gut Microbiota in Africanized Honey Bees (Apis mellifera scutellata x spp.)

Honey bees play a critical role in ecosystem health, biodiversity maintenance, and crop yield. Antimicrobials, such as tetracyclines, are used widely in agriculture, medicine, and in bee keeping, and bees can be directly or indirectly exposed to tetracycline residues in the environment. In European honey bees, tetracycline exposure has been linked with shifts in the gut microbiota that negatively impact bee health. However, the effects of antimicrobials on Africanized honey bee gut microbiota have not been examined. The aim of this study was to investigate the effects of tetracycline exposure on the gut microbial community of Africanized honey bees (Apis mellifera scutellata x spp.), which are important pollinators in South, Central, and North America. Bees (n = 1,000) were collected from hives in Areia-PB, Northeastern Brazil, placed into plastic chambers and kept under controlled temperature and humidity conditions. The control group (CON) was fed daily with syrup (10 g) consisting of a 1:1 solution of demerara sugar and water, plus a solid protein diet (10 g) composed of 60% soy extract and 40% sugar syrup. The tetracycline group (TET) was fed identically but with the addition of tetracycline hydrochloride (450 μg/g) to the sugar syrup. Bees were sampled from each group before (day 0), and after tetracycline exposure (days 3, 6, and 9). Abdominal contents dissected out of each bee underwent DNA extraction and 16S rRNA sequencing (V3-V4) on an Illumina MiSeq. Sequences were filtered and processed through QIIME2 and DADA2. Microbial community composition and diversity and differentially abundant taxa were evaluated by treatment and time. Bee gut microbial composition (Jaccard) and diversity (Shannon) differed significantly and increasingly over time and between CON and TET groups. Tetracycline exposure was associated with decreased relative abundances of Bombella and Fructobacillus, along with decreases in key core microbiota such as Snodgrassella, Gilliamella, Rhizobiaceae, and Apibacter. These microbes are critical for nutrient metabolism and pathogen defense, and it is possible that decreased abundances of these microbes could negatively affect bee health. Considering the global ecological and economic importance of honey bees as pollinators, it is critical to understand the effects of agrochemicals including antimicrobials on honey bees.


39
Bees play a critical role as pollinators in ecosystems across the globe, contributing to the 40 maintenance of biodiversity on Earth (Kevan and Viana, 2003;Michener, 2007). In addition to this 41 important ecological function, bees are also essential as pollinators in agriculture systems (Gisder and 42 Genersch, 2017; Hung et al., 2018). Honey bees (Apis spp.), specifically, are the top crop pollinators and 43 directly enhance crop yields (Gisder and Genersch, 2017). The Africanized honey bee (Apis mellifera 44 scutellata x spp.), a crossbreed between European honey bees (Apis mellifera sspp.) and African honey 45 bees (Apis mellifera scutellata), emerged in the late 1950's in Brazil (Winston, 1992). African honey bees 46 adapted and spread widely across the Americans because of their reproductive traits and superior ability 47 to colonize tropical ecosystems compared with European bees. Some of the traits include improved 48 thermoregulation capacity, greater resistance to diseases, increased egg-laying rates, more frequent queen 49 replacement, and shorter developmental time (Guzmán-Novoa et al., 2011). 50 In spite of their great economic and biological importance, bee populations across the planet have 51 been under increasing threat due to human population expansion, habitat destruction, and the use of 52 agrochemicals including pesticides and antimicrobials. The use of such compounds has been associated 53 with an increased occurrence of Colony Collapse Disorder (CCD), a phenomenon characterized by the 54 disappearance of worker bees and compromise of the honey bee colony (Caires et  class among the 116 countries that provided data. Moreover, tetracyclines represented approximately 35% 61 of the antimicrobial use in these countries, including use for growth promotion in feed animals, which is 62 an ongoing practice in many countries. Recently, tetracyclines were also highlighted as an option for the 63 treatment and prophylaxis of COVID-19, and tetracycline use has increased significantly in some 64 hospitals during the pandemic ( (Hopkins, 1979). In these cases, oxytetracycline is sprayed over orchards or vineyards, and 75 oxytetracycline concentrations on plant tissues can range from 100 to 4,166 ppm (Chanvatik et al., 2019). 76 Bees can be indirectly exposed to antimicrobials while foraging in these agricultural or urban 77 environments that contain tetracycline residues (Lau and Nieh, 2016). Bees can also be directly exposed 78 to tetracyclines in the course of treatment for European and American foulbrood, bacterial diseases that 79 cause severe losses in hives and honey production (Doughty et al., 2004;Martel et al., 2006). To treat 80 foulbrood, oxytetracycline is applied directly onto the hives at doses ranging from 500 (Dinkov et al.,81 2005) to 5900 ppm (Kochansky, 2000). Antimicrobials can disturb gut microbial communities and affect 82 their overall structure and function (Blaser, 2014 Considering the widespread prevalence of tetracycline in the environment due to its use in 91 agriculture, medicine, and in relation hive health, and evidence of gut microbiome disturbances in 92 European honey bees due to antimicrobial exposure, the aim of this study was to investigate the effects of 93 tetracycline on the gut microbiota of Africanized honey bees (Apis mellifera scutellata x spp.) in tropical 94 conditions. 95

96
Experimental design and sampling  On Day 0 (D0), approximately one thousand bees were collected from five outdoor hives at LABE, 102 placed into ten plastic chambers that were kept in an incubator at 32C and 66% relative humidity (TE-103 371, Tecnal, Piracicaba, Brazil) (Figure 1). The plastic chambers measured 176.71 cm 2 and were covered 104 with a nylon screen. The bees were divided into two groups: The control group (CON), was fed daily with 105 10g of syrup consisting of a 1:1 solution of demerara sugar and water. Sterile cotton balls were soaked 106 into the syrup and then placed into the bee chambers daily. Bees were also fed a solid protein diet (10g) 107 composed of 60% soy extract and 40% demerara syrup solution. The tetracycline group (TET) was fed 108 identically except that syrup contained 450 µg/g (equivalent to 450 ppm) tetracycline hydrochloride 109 (Tetramed, Medquímica, Brazil). This dose reflects what honey bees may be exposed to within some 110 agricultural environments and is also within the range of hive dosing for the treatment of foulbrood 111 (Raymann et al., 2017). 112 A 9 cm 2 -piece of brood comb was placed in each chamber. Five replicates of 20 bees each were 113 collected from each group at each sampling point including: day 0 (D0, pre-treatment) and days three 114 (D3), six (D6), and nine (D9) (Figure 2). Bees were placed in sterile tubes containing 70% alcohol, 115 transported to the lab and stored at -20C until extraction. All procedures performed were approved by the 116 Biodiversity Authorization and Information System -SISBIO (Protocol #: 71750-1, approved on 117 09/19/2019). 118 DNA extraction, library preparation, and sequencing 119 Prior to extraction, bees were placed on sterile filter paper for 10 minutes for defrosting and alcohol 120 evaporation. Bee intestines were dissected by using a sterile pair of scissors to make a cross-sectional cut 121 across the last segment of the bee abdomen. With sterile tweezers, abdominal content was collected out of 122 the abdomen and transferred into microtubes. Abdominal contents from 20 bees were pooled into a single 123 tube for DNA extraction, which was performed using a commercial kit (PowerSoil DNA Isolation kit, 124 Qiagen, Germany) following the manufacturer's protocol. After extraction, DNA was electrophorized in 125 agarose gel for quality analysis and quantified using a microvolume spectrophotometer (Colibri LB 915, 126 Titertek-Berthold, Germany). DNA concentrations were quantified by fluorometry (Qubit 2.0, Life 127 Invitrogen, USA) before further processing steps. 128 The V3-V4 region of the microbial 16S rRNA gene was amplified by PCR using 2.5 μL template 129 DNA (5 ng/μL), 5 μL forward primer, 5 μL reverse primer, and 12 μL 2X KAPA HiFi HotStart 130 ReadyMix (

Sequence processing and statistical analyses 146
The raw demultiplexed paired-end sequences were processed using QIIME 2-2020.2 (Bolyen et al., classifier with a 99% sequence similarity threshold for V3-V4 reference sequences (SILVA-132-99-nb-152 classifier.qza) and the "qiime feature-classifier classify-sklearn". Negative control samples were 153 examined for potential contaminant taxa. No taxa overlapped between negative control and true samples. 154 Microbial diversity was quantified using Pielou's (evenness) and Shannon (richness and abundance) 155 diversity indices. ANOVAs were used to compare diversity between groups in R 4. employed as recommended (Anderson, 2001) to test for differences in microbial composition between 161 experimental groups (Pre-treatment vs. CON vs. TET) and over time (Day 0pre-treatment, and Days 3, 162 6, 9). 163 Differentially abundant taxa between groups were identified using an analysis of composition of 164 microbiomes (ANCOM) (Mandal et al., 2015). We also performed a core microbiota analysis in QIIME2, 165 to identify taxa present in 95% of the samples. The relative abundances of core microbes were then 166 compared by treatment and time using two-way ANOVAa after testing for normality using a Shapiro-167 Wilk test. A P-value < 0.05 was used in the statistical tests for significance. 168

169
16S rRNA sequencing reads 170 We obtained a total of 3,575,254 raw reads across all samples, ranging from 10, 268 Fig. 3, Fig. S1). Notably, the interaction of treatment and 179 time was also significant (Adonis: R2= 0.035, p-value= 0.024), and the effect of treatment increased over 180 time (Fig. 3, Fig. S1). Microbial diversity also differed significantly by time but not by treatment (Two-181 way ANOVA: Shannon Index treatment p = 0.295, time p = 0.042; Fig. 4a). No pairwise comparisons 182 were significant; although, microbial diversity differences on Day 9 (p = 0.081) were greater than at 183 previous timepoints, with the tetracycline group having lower diversity than the control group. This 184 suggests that diversity was being impacted gradually after continual exposure to tetracycline. Microbial 185 community evenness (Pielou's Index) did not differ significantly by treatment or time (Two-way 186 ANOVA: Pielou's Index treatment p = 0.457, time p = 0.061; Fig. 4b) 187 Core microbiota and differentially abundant taxa 188 A core microbiota analyses identified eight genera that were present in 95% of the samples across 189 all treatments and times including: Lactobacillus, a taxon from the class Gammaproteobacteria, 190 Bifidobacterium, Snodgrassella, Gilliamella, a taxon from the family Rhizobiaceae, Apibacter, and 191 Commensalibacter (Fig. 5). These taxa accounted for 22% of all genera in the dataset. We then used a 192 two-way ANOVA to compare relative abundances of these taxa by treatment and time.  Fig. 5a,b). Bifidobacterium was also increased in the tetracycline group (p = 0.029); although, 197 abundances did not change over time (Fig. 5c) Fig. 5d,e,f). Apibacter was also significantly decreased in the tetracycline group; 201 although only at the early time points (treatment p < 0.004; time p = 0.834; interaction p = 0.004; Fig.  202 5g). There were no significant differences in the relative abundances of Commensalibacter between 203 groups or over time (p > 0.05; Fig. 5h). 204 An ANCOM identified five differentially abundant taxa by treatment at the L6 genera level including 205 Bombella and Fructobacillus, a taxon in the family Enterobacteriaceae, Idiomarina, and taxon in the class 206 Gammaproteobacteria (Fig. 6, 5b) tetracycline group at all time points (Fig. 6a,b,c). Idiomarina differed significantly by treatment 211 (Idiomarina p = 0.0002) and by time (Idiomarina p < 0.0001;), and there was a significant interaction 212 between treatment and time (Idiomarina p = 0.005;) as both Idiomarina and the Gammaproteobacteria 213 taxa increased over time particularly in the tetracycline group (Fig. 6d, 5b). We also performed an 214 ANCOM analysis at the L7 (roughly species) and amplicon sequencing variant levels and produced 215 similar results in terms of differentially abundant microbes: Fructobacillus (W=30) and Bombella 216 (W=30) species were decreased in the tetracycline group, while 2 Gammaproteobacteria ASVs (W=194, 217 W=179) increased over time in the tetracycline group. 218

219
Our results demonstrated that tetracycline exposure was associated with alterations in Africanized 220 honey bee gut microbial composition but not diversity over time. We further identified shifts in core and 221 non-core microbiota by treatment and time. These tetracycline-linked gut microbial changes suggest 222 negative implications for honey bee nutrient metabolism and pathogen resistance. 223 Core microbial taxa and tetracycline treatment taxon of the Gammaproteobacteria classin bees exposed to tetracycline has also been observed in 230 previous studies on bees exposed to chemical compounds or in compromised hives. For instance, 231 increased relative abundances of Lactobacillus were reported in bees exposed to the herbicide glyphosate, Taken together, these results suggest that Lactobacillus, Bifidobacterium, and Gammaproteobacteria may 237 be positively associated with exposure to agrochemicals. Notably, our results differ from a study on 238 European honey bees that reported decreases in several Lactobacillius ASVs following exposure to 239 oxytetracycline (Daisley et al., 2020).  Pectin-rich pollen is large part of the honey bee diet, but bees do not produce pectinases and must rely on 253 gut microbes like Gilliamella for pectin metabolism. Like Snodgrassella and Gilliamella, Apibacter also 254 colonizes the gut wall (Kwong et al., 2018), and some strains of Apibacter encode a type VI secretion 255 system (T6SS) (Kwong et al., 2018), which promotes colonization resistance through the delivery of toxic 256 antibacterial proteins into neighboring cells (Steele et al., 2017). Decreased abundances of Snodgrassella, 257 Gilliamella, and Apibacter could impact nutrient metabolism and pathogen defense in Africanized honey 258 bees. 259 Differing sensitivity to tetracycline could explain the taxonomic shifts we observed with 260 tetracycline exposure. Gram positive and gram negative bee gut bacteria reportedly have different 261 sensitivities to host-produced antimicrobial peptides including apidaecin and hymenoptaecin. In a 262 previous study by Kwong et al. (2017), gram-positive species (Lactobacillus Firm-5, Bifidobacterium sp.) 263 were highly resistant to apidaecin and hymenoptaecin, while gram-negative species, particularly 264 Snodgrassella alvi, were more sensitive to hymenoptaecin. It is possible that gram positive bacteria, such 265 as Lactobacillus and Bifidobacterium, which dominate the hindgut, are less sensitive to tetracycline, 266 while gram negative bacteriasuch as Snodgrassella, Gilliamella, Apibacter, and Rhizobiaceae, which 267 are more common in the ileumare more sensitive to tetracycline and therefore decreased in abundance 268 following tetracycline exposure while Lactobacillus and Bifidobacterium increased (Powell et al., 2014;269 Kwong and Moran, 2016; Kešnerová et al., 2020). Commensalibacter was the only core microbe that did 270 not vary in relative abundance after tetracycline exposure; however, these bacteria do vary by season and 271 age in honey bees (Ellegaard and Engel, 2019). In sum, alterations in the core microbiota following 272 tetracycline exposure, and particularly decreased abundances of Snodgrassella, Gilliamella, 273 Rhizobiaceae, and Apibacter, suggest a reduced capacity for pathogen defense and nutrient metabolism 274 which could potentially increase the susceptibility of Africanized honey bees to parasites or infections. 275

Differentially abundant microbes by treatment 276
Among the five differentially abundant taxa identified between treatment groups, three 277 (Bombella, Fructobacillus, an Enterobacteriaceae taxon) were decreased in abundance in bees exposed to 278 tetracycline, while two were increased (Idiomarina, and a Gammaproteobacteria taxon, which was also 279 identified as a core bacteria). inoculation of core microbes into larvae and developing worker bees (Rokop et al., 2015). As such, 287 decreased abundances of Bombella and Fructobacillus due to tetracycline exposure could negatively 288 affect Africanized honey bee larval development. 289 To our knowledge, this is the first study characterizing the gut microbiota of Africanized honey 290 bees in relation to tetracycline exposure. However, this study had several limitations. While the microbial 291 shifts we observed suggest negative implications for bee health, we do not have associated 292 immunological, behavioral, fitness, or production data to explicitly support these implications. Secondly, 293 in this study, we selected a tetracycline concentration consistent with that reported in some agricultural or 294 hive applications. However, quantifying the concentration of tetracycline to which bees are actually 295 exposed under natural conditions is challenging and likely highly variable. Third, we observed a shift in 296 the gut microbiota between pre-treatment and both the CON and TET groups, suggesting either an age or 297 "incubator effect" due to an altered diet and environment; although, we attempted to replicate natural 298 temperature and humidity conditions as closely as possible within the incubator. Despite this, there were 299 still clear differences between the CON and TET groups over time. Finally, the function of some of the 300 differentially abundant microbes we identified, such as Apibacter, have yet to be elucidated. As such, 301 deeper sequencing and associated studies with metabolomics or transcriptomics are necessary to clarify 302 the role of these microbes in the bee gut. 303 304

305
Tetracycline exposure altered gut microbial composition in Africanized honey bees (Apis mellifera 306 scutellata x spp), and was specifically associated with decreased abundances of Bombella, Fructobacillus,307 Snodgrassella, Gilliamella, Rhizobiaceae, and Apibacter. These microbes play a key role in nutrient 308 metabolism and pathogen defense, and reduced abundances of these microbes could potentially have 309 negative impacts on bee health. Considering the global ecological and economic importance of honey 310 bees as pollinators, it is critical to understand the effects of antimicrobials widely used across agriculture, 311 medicine, and in bee keeping, on honey bee health, as bees can be directly or indirectly exposed to these 312 drugs in many of their foraging environments. Future studies assessing bee fitness, behavior, immune 313 response, and disease susceptibility in relation to agrochemical exposure will further elucidate the impacts 314 of these gut microbial changes. Understanding how chemicals, like antimicrobials, affect bees is essential 315 to guide agricultural practices that effectively support ecosystem health. 316

320
The authors declare that the research was conducted in the absence of any commercial or financial 321 relationships that could be construed as a potential conflict of interest. 322  Pollination by honey bees is critical for ecosystem health and crop production around the globe. 350

Author Contributions
However, bee populations have been threatened by anthropogenic activities worldwide. Antimicrobial 351 drugs, such as tetracycline, are widely used in agriculture and medicine, and bees can be directly or 352 indirect exposed to residues of these drugs in the environment. Previous studies have shown that 353 antimicrobial exposure can cause gut microbial disturbances and affect the health of European honey 354 bees. This study reports the effects of tetracycline exposure on the gut microbiota of Africanized honey 355 bees, which are important pollinators across South, Central, and North America. 356 357 Figures   Figure 1. a) Approximately 100 bees were housed in each plastic chamber along with a piece of brood comb. Sterile cotton soaked in water or sugar syrup and a solid protein diet were also included in each chamber, and chambers were covered with Nylon screen. B) All chambers were placed in an incubator that was maintained at 32°C ± 1.45 temp and 66% ± 5.34 relative humidity for the duration of the experiment. Figure 2. Experimental design. Five replicates of 20 bees were collected from the control (CON) and tetracycline (TET) groups at each time point including Day 0 (D0 -Pre-treatment), and Days 3 (D3), 6 (D6), and 9 (D9).

Figure 3.
Bee gut microbial composition (Jaccard) based on treatment (Pre-treatment, Control, Tetracycline) and time (Day 0pre-treatment, Days 3, 6, 9). Microbial composition was significantly altered by treatment (PERMANOVA: p = 0.001) and time (PERMANOVA: p = 0.001; also Supp. Fig. 1)   Figure 4. Microbial diversity and evenness by treatment and time. Box plot shows outliers, first and third quartiles (lower and upper edges), and highest, lowest and median values (horizontal black dash) for Control (Con) and Tetracycline (Tet) groups. a) There were significant differences in diversity (Shannon Index) by time (p = 0.042) but not treatment (p = 0.295); although, no pairwise comparisons were significant. b) There were no significant differences in evenness (Pielou's Index) by time (p = 0.061) or by treatment (p = 0.457).

Figure 5.
Relative abundances of core microbiota (genera) that were present in 95% of all samples: A) Lactobacillus, B) Gammaproteobacteria, C) Bifidobacterium, D) Snodgrassella, E) Gilliamella, F) one taxon from the family Rhizobiaceae, G) Apibacter and H) Commensalibacter. Box plot shows outliers, first and third quartiles (lower and upper edges), and highest, lowest and median values (horizontal black dash) for Control (Con) and Tetracycline (Tet) groups. (ANOVA: *p < 0.05; ** p < 0.01 and *** p < 0.001) Figure 6. Relative abundances of differentially abundant genera (ANCOM) by treatment and by time. A) Bombella, B) Fructobacillus, C) a taxon in the family Enterobacteriaceae and D) Idiomarina. A Gammaproteobacteria taxa was also identified as a core microbe and a differentially abundant microbe between Con and Tet groups and is shown in Fig. 5b. Box plot shows outliers, first and third quartiles (lower and upper edges), and highest, lowest and median values (horizontal black dash) for Control (Con) and Tetracycline (Tet) groups. (ANOVA: * p < 0.05; ** p < 0.01 and *** significant at p < 0.001)