Immune system stimulation by the native gut microbiota of honey bees

Gut microbial communities can greatly affect host health by modulating the host's immune system. For many important insects, however, the relationship between the gut microbiota and immune function remains poorly understood. Here, we test whether the gut microbial symbionts of the honey bee can induce expression of antimicrobial peptides (AMPs), a crucial component of insect innate immunity. We find that bees up-regulate gene expression of the AMPs apidaecin and hymenoptaecin in gut tissue when the microbiota is present. Using targeted proteomics, we detected apidaecin in both the gut lumen and the haemolymph; higher apidaecin concentrations were found in bees harbouring the normal gut microbiota than in bees lacking gut microbiota. In in vitro assays, cultured strains of the microbiota showed variable susceptibility to honey bee AMPs, although many seem to possess elevated resistance compared to Escherichia coli. In some trials, colonization by normal gut symbionts resulted in improved survivorship following injection with E. coli. Our results show that the native, non-pathogenic gut flora induces immune responses in the bee host. Such responses might be a host mechanism to regulate the microbiota, and could potentially benefit host health by priming the immune system against future pathogenic infections.

primer sequences were trimmed from joined reads, which were then quality filtered by applying cutoffs of phred > 29, max N = 0, and length between 230 bp and 270 bp (expected size ~250 bp); OTUs at 97% identity clustering were picked de novo using the default uclust algorithm. OTUs accounting for < 0.5% of reads for each sample were removed, as were OTUs without matches to 16S rDNA sequences or with matches to mitochondria or chloroplast. OTUs were assigned taxonomy by BLASTN searches against the Genbank nr database.
Quantitative PCR qPCR was carried out on an Eppendorf Mastercycler ep realplex (S. alvi treatment experiment) or a Life Technologies ViiA 7 System (gut-fed treatment experiment). Tenmicroliter reactions were set up in 96-well plates using 5 µl of iTaq Universal SYBR Green Supermix (Bio-Rad Inc.), 2 µl H2O, 1 µl each of 5 µM forward and reverse primers, and 1 µl of 1:10 diluted cDNA template. The cycling conditions consisted of 95°C for 30 s; 10 cycles of 95°C for 5 s, 55°C for 15 s (-0.5°C per cycle), and 68°C for 20 s; 35-40 cycles of 95°C for 5 s, 50°C for 15 s, and 68°C for 20 s. A melting curve analysis was also conducted to ensure the PCR products were the correct sizes.

Detection of peptides in gut fluid and hemolymph
Newly eclosed bees were fed 5 µl of a single whole gut crushed in 500 µl 1:1 (w/v) sucrosewater. The guts were from adult hive bees (microbiota treatment) or newly eclosed bees lacking microbiota (control). After 5 days, hemolymph from the thorax and gut fluid from the rectum lumen were withdrawn using micropipettes after puncture by glass needles. Approximately 3 µl of each type of fluid was obtained per bee, and fluid was pooled from 5 -13 bees to obtain sufficient concentrations for analysis (~13-18 µl per sample). Samples were added to 100 µl of 0.1% (v/v) trifluoroacetic acid and boiled at 100°C for 10 min, then centrifuged at 15,000 g for 10 min. The supernatant was transferred to a new tube and boiled at 100°C for another 10 min, then spun at 15,000 g for 10 min. The supernatant was taken and frozen at -80°C until further enrichment of apidaecin by method of Danihlík et al. [3]. For peptide enrichment, Oasis WCX 1cc Vac Cartridges with 30 mg sorbent, 30 µm particle size (Waters Corp.), were conditioned with 1 ml methanol, then equilibrated with 1 ml HPLC-grade water. Samples were added to 900 µl 5% (v/v) formic acid and drawn through the sorbent via vacuum pressure. Cartridges were washed twice with 500 µl of 30 mM NH4HCO3 (pH 9), and then washed twice with 500 µl methanol. Peptides were subsequently eluted with 400 µl 50% (v/v) acetonitrile containing 5% (v/v) formic acid.
Prior to analysis, samples were concentrated by SpeedVac and resuspended in 12 µl 0.1% (v/v) formic acid. Samples were analyzed by reverse phase LC-MS/MS using a Dionex Ultimate 3000 Nano UPLC coupled to an Orbitrap Elite mass spectrometer (Thermo Fisher Scientific Inc.). The LC utilizes an Acclaim PepMap100 C18 Nano-Trap pre-column (75 µm × 2 cm, 3 µm particle size, 100 Å pore size) and an Acclaim PepMap Nano column (75 µm × 15 cm, 3 µm particle size, 100 Å pore size) for separation. Peptides were eluted from the column using a gradient of 0.1% (v/v) formic acid and water and 0.1% (v/v) formic acid and acetonitrile over 60 minutes. The Orbitrap Elite was operated using a targeted MS/MS method with precursor ions of 703.40 m/z, 665.70 m/z, 422.44 m/z, and 1054.59 m/z being detected in the orbitrap at 60,000 resolution. Fragment ions were captured in the orbitrap and also detected at 60,000 resolution. Following acquisition, data were analyzed using Proteome Discoverer 1.4 (Thermo Fisher Scientific Inc.) with the SEQUEST HT search algorithm and target decoy method. Results reported are filtered for high (1% false discovery rate) confidence.
Bacterial challenge of bees E. coli ATCC 25922, a strain commonly used for antimicrobial susceptibility assays, was grown in tryptic soy broth (BD Difco) at 37°C overnight to an OD600 of 0.5. Cells were centrifuged, washed with PBS, and then diluted to set concentrations (10 3 -10 5 cells/µl) in PBS. Five days after inoculation with S. alvi or G. apicola, bees were chilled on ice in preparation for bacterial injection, and then 1 µl of PBS or E. coli was injected into the hemocoel laterally between the second and third tergite with a fine-tipped glass capillary. After infection, bees were returned to incubation at 34°C with sucrose-water and sterile pollen.

Measurement of E. coli survival in bee hemolymph
At 2 hours and 6 hours post-injection, hemolymph was collected from each bee and 0.5 µl aliquots of undiluted and 1:100 diluted hemolymph were plated onto tryptic soy agar. After overnight incubation at 37°C, the number of colony-forming units (CFUs) per agar plate was counted, multiplied by the dilution factor, and adjusted to reflect the number of surviving E. coli cells per µl of hemolymph.

Survival assays
We monitored and recorded the number of deceased bees in each cup cage approximately every 3 hours over 2 days. The initial number of bees in each cup was recorded at the time of bacterial injection. Survival curves were generated in Prism 6 (GraphPad Software Inc.). Multiple comparison tests were performed using the log-rank test.

G. apicola colonization trials
To test the effectiveness of gut colonization with G. apicola, preliminary trials were conducted with two G. apicola strains, wkB1 [4] and wkB7 [5]. Two methods of inoculation were used, hand-feeding and placing the cultured strains into the pollen food supply. Cultivation and feeding methods are as described in materials and methods. Sterile sucrose solution was used for the uninoculated controls. After 5 days, DNA was extracted from a subset of bees (Cohort 1) as previously described [1], and qPCR was performed using Gilliamella-specific primers as described in [6] to quantify colonization.
To determine if transmission of G. apicola directly from gut material improves colonization compared to using in vitro cultures, 5 µl of resuspended crushed guts from Cohort 1 bees were fed to a new set of newly eclosed workers. After 5 days, DNA was extracted from these bees (Cohort 2) and quantified by qPCR as above.

Microbiota composition of gut-fed bees
For the gut-fed qPCR experiment, two cohorts of the "full microbiota" treatment were used, each fed gut material from a single adult hive bee ("Gut A" and "Gut B"). For the uninoculated controls, bees were fed the gut material of a newly eclosed worker to control for the effect of ingesting gut material on gene expression. Microbiome composition was determined for 9 bees from each treatment, taken from the same cup-cages as bees kept for the qPCR experiment (three cup-cages for each of "Gut A", "Gut B", and "Control" groups; 3 bees taken from each cup-cage for DNA extraction/16S rDNA community analysis, 7-9 bees taken from each cup-cage for RNA extraction/qPCR analysis). Microbiome composition was also determined for the initial gut material used for inoculating "Gut A", "Gut B", and "Control" treatment groups.
Gut-fed "full microbiota" workers have microbiotas that largely resembled that of the source material (figure S1), although the relative proportions varied among individuals. Most individuals were successfully colonized with all members of the core microbiota (S. alvi, G. apicola, F. perrara, Bifidobacterium spp., and Lactobacillus Firm-4 and Firm-5). The inoculum for the control group had a low number of reads (290) compared to the "full microbiota" samples (average 23,731), suggesting that very few bacteria were present. The bees inoculated with the control gut material likewise lacked the characteristic bee gut microbial community (figure S1); however, some control individuals became colonized with Lactobacillus Firm-5. Although not all bees in this group may indeed be fully germfree, reduced PCR amplification was observed for many control samples, suggesting a lower absolute bacterial load.

E. coli survival in hemolymph
Two of three trials showed a significant decrease of E. coli in the hemolymph of bees inoculated with S. alvi or G. apicola (figure 4). The third trial showed no difference; this trial was conducted on late-season bees (November 2015), whereas the other trials were on bees collected in the spring (May 2015). There may be seasonal immunological differences responsible for the observed differences between trials [7, 8], although this will need to be further tested.

G. apicola colonization trials
Up to 10 9 Gilliamella 16S rRNA gene copies were detected in bees monoinoculated with G. apicola strains. Colonization by strain wkB7 was more reliable than strain wkB1 (figure S5); however, hand-feeding of wkB7 resulted in lower colonization efficiencies than leaving cultured bacteria in the pollen food supply. We continued using hand-feeding in our survival trials, as it allowed for inoculation of a known quantity of bacteria per individual. Feeding of homogenized guts containing G. apicola resulted in better colonization than using in vitro cultures (figure S5, Cohort 2 treatments), but this method still did not allow for reliable colonization by strain wkB1. Based on these results, we decided to use strain wkB7 rather than wkB1 in our survival and E. coli-challenge experiments.  Figure S1. Gut microbiome composition of bees in the gut-fed qPCR experiment (figure 1). Bacterial taxonomic abundances for each individual bee were averaged to obtain relative proportions. Two cohorts of the "full microbiota" treatment were used, each fed gut material from a single adult hive bee ("Gut A" and "Gut B"). Input, the microbiome composition of the source gut. Output, the averaged microbiome from 9 individuals of treatment bees.