2.1 Samples

The samples originated from three geographically separate outbreaks of AFB in Sweden during the summer of 2018. Bee inspectors collected approximately 300 adult honey bees per colony from apiaries in the outbreak areas. The samples were sent to the Swedish National Reference Laboratory for Bee Health and analyzed for P. larvae using standard cultivation methods according to Swedish legislation (Forsgren & Laugen, 2014; Nordström & Fries, 1995) (Table 1). Five bees from each 41 samples collected were pooled and used for analyses. The samples were classified according to four categories: honey bee colonies with disease (AFB1, N = 3 [one from each separate disease outbreak]), disease-free colonies from the same apiaries as AFB1 (AFB2, N = 9 [3 from each separate disease outbreak]), disease-free apiaries belonging to beekeeping operations with AFB (AFB3, N = 16 [4–6 per outbreak]), and disease-free beekeeping operations in the AFB-outbreak areas (AFB0, N = 13 [3–5 per outbreak]) (Figure 1, Table 1). Sample information is summarized in Table 1. All honey bee samples were collected during the same period (early May until early June 2018).

2.2 DNA extraction from bee samples

DNA was extracted in duplicates as described by Engel et al. (2013), with minor modifications, using the DNeasy Blood and Tissue-kit (Qiagen, Hilden, Germany). The bees were washed in a 1% aqueous solution of Klorin Original (Colgate-Palmolive AB, Stockholm, Sweden) for 2–3 min before being rinsed three times in sterile water. Heads and thoraxes were removed and abdomens were placed in a 2 mL tube (one tube per pooled sample, each pooled sample consisting of five bees) with 5–6 glass beads and 500 μL RLT-buffer (Qiagen) with beta-mercaptoethanol (10 μL/mL). The samples were homogenized using Precellys Evolution (Bertin Instruments, Montigny-le-Bretonneux, France) with a pre-set protocol (2 × 5400 rpm for 90 s, 30 s break between runs). The homogenization program was repeated twice and the samples kept on ice between runs. The samples were centrifuged for 2 min at 5000 rpm and 200 μL of the resulting supernatant was mixed with 200 μL 99% ethanol. Two hundred μL of the mixture was used for DNA recovery following the protocol “Purification of Total DNA from Animal Tissues” in the DNeasy Blood and Tissue-kit (Qiagen, starting at step 4 in DNeasy Blood & Tissue Handbook 07/2006). The DNA was eluted in 60 μL AE-buffer and the quality and yield of the DNA was estimated using Nanodrop. All DNA samples were diluted to 20 ng/μL and stored at −20°C until analyzed.

2.3 DNA extraction from lactic acid bacteria

Ten isolates of hbs-LAB species (Lactobacillus apis, Apilactobacillus kunkeei, Lactobacillus kullabergensis, Lactobacillus kimbladii, Lactobacillus helsingborgensis, Lactobacillus melliventris, Apilactobacillus apinorum, Bombilactobacillus mellis, Bifidobacteria asteroides, and Bifidobacteria coryneforme) (Lamei et al., 2020) were cultured in Man, Rogosa and Sharpe broth (Oxoid, Basingstoke, UK) supplemented with 2% fructose (Merck, Darmstadt, Germany) and 0.1% l-cysteine (Sigma-Aldrich, St. Louis, Missouri, USA) (sMRS-broth) as described by Lamei et al. (2017). Approximately, 10⁹ bacteria of each isolate were harvested and used for DNA extraction. The bacteria were lysed in lysis buffer (20 mM Tris, 2 mM EDTA, 1.2% Triton x-100) on a shaker at 37°C for 18 h. DNA was purified from the lysate using a QIAcube extraction robot (Qiagen) and the Qiagen DNA extraction protocol “Purification of bacterial or yeast DNA with enzymatic lysis V1,” eluting in 60 μL AE-buffer (Qiagen). The quality and concentration of the DNA was estimated using Nanodrop.

2.4 Assay design

Primers for five Lactobacilli, two Apilactobacilli, one Bombilactobacilli, and two Bifidobacterium were designed to fit sequences available through the National Center for Biotechnology Information (NCBI) (Table 2). The primers were designed using the Primer-BLAST function on NCBI and CLC Main Workbench 8.1 (Qiagen Bioinformatics). The primers were evaluated using qPCR and DNA from reference isolates using SsoFast EvaGreen Supermix (Bio-Rad) according to the manufacturer's recommendation with an annealing temperature of 60°C followed by a melt curve analysis (hot start 2 min 98°C; 35 cycles 5 s 98°C, 10 s 60°C; melting curve 65–95°C 0.5°C/5 s). To verify the specificity of the primers, the assays were tested with conventional PCR using HotStarTaq Plus DNA Polymerase (Qiagen) according to manufacturer's recommendations. The PCR products were visualized by gel electrophoresis and Sanger-sequenced (Macrogen Europe, Amsterdam, the Netherlands) to confirm their identity.

2.5 qPCR analyses

The qPCR analyses were performed using a Biomark HD system (Fluidigm) following the manufacturer's protocol for gene expression analysis and as described by D'Alvise et al. (2019). To evaluate the quantitative performance of the assays, the assays were tested on FlexSix GE integrated fluidic circuits (IFC) or on 96 × 96 IFCs for gene expression (Fluidigm, San Francisco, CA, USA) using 10-dilution series of extracted hbs-LAB DNA as template. Eight to twelve replicates of the assays were matched with a six- to eight-step dilution series of each hbs-LAB with known DNA concentration to determine detection limit, linear dynamic detection range, variation at detection limits, and PCR efficiency (all above 95%). All assays had a PCR efficiency above 95% and were able to detect as few as 40–850 bacteria per reaction. The final qPCRs were performed on 96 × 96 IFCs for gene expression (Fluidigm) using the manufacturer's standard qPCR protocol for fast PCR followed by a melt curve analysis (thermal mix: 40 min 70°C, 30 s 60°C; hot start: 60 s 95°C; 30 cycles 5 s 96°C, 20 s 60°C; melting curve: 60–95°C 1°C/s). Each duplicate of the extracted DNA was analyzed in triplicates and repeated twice. The qPCR data was checked using the automated quality control in the software Fluidigm Real-Time PCR analysis and further checked and revised manually. Only Cq-values derived from reactions that showed a specific melting temperature peak of the product were used in the analysis.

2.6 Data analysis and statistics

Logarithmic linear regression was used for the conversion of Cq-values to number of bacteria per reaction and converted to bacteria per honey bee by back-calculation of dilutions. Means of the bacterial concentrations per honey bee was determined in duplicate, and was logarithmically (log₁₀) transformed for use in the statistical analysis. The diversity of hbs-LAB species was estimated for each individual sample using the calculated mean bacterial numbers using Shannon's diversity index.

The statistical software SPSS statistics (IBM SPSS Statistics for Windows, Version 26.0) was used for data analysis and significance testing between groups. Normality and homogeneity of variances of the dataset was checked using Shapiro–Wilk's test for normality and Levene's test for homogeneity before the groups' mean values were compared. Since Shannon's diversity index data met the requirements for normality and homogeneity, an ANOVA was used to compare the means of three groups or more (AFB status and outbreak region) and independent t-tests for comparing two groups (visual AFB symptoms in the beekeeping operation, in the apiary, or in the colony, as well as P. larvae presence or absence).

The abundance of the respective hbs-LAB was not normally distributed between individuals. A Kruskal–Wallis with Dunn's post hoc test was therefore used to compare the means of three groups or more (AFB status and outbreak region) and Mann–Whitney U test for comparison between two groups (visual AFB symptoms in beekeeping operation, apiary and colony, and P. larvae presence or absence).

The program R-4.2.2 (Innocent and Trusting; R Core Team, 2022) was used for determining community composition differences between AFB groups. A Bray-Curtis matrix was created on normalized species counts using the vegan package.

3.1 Distribution of honey bee-specific lactic acid bacteria by outbreak

One honey bee-specific lactic acid bacterial species (hbs-LAB), A. apinorum, was not detected in any of the samples and was therefore excluded from the statistical analyses (Table 3, Appendix 1). All other hbs-LAB (L. apis, A. kunkeei, L. kullabergensis, L. kimbladii, L. helsingborgensis, L. melliventris, B. mellis, B. asteroides, and B. coryneforme) were found in all three outbreaks and in all four AFB groups (Table 3, Appendices 1 and 2). Four were detected in all samples: L. apis, L. kullabergensis, L. melliventris, and B. mellis, (Table 3, Appendices 1 and 2). Four additional bacterial species were found in a majority of the samples: L. helsingborgensis, L. kimbladii, B. asteroides, and B. coryneforme (97.6%, 85.4%, 95.1%, and 68.3% respectively; Table 3, Appendix 1). There were some differences in distribution between outbreaks; L. helsingborgensis and B. asteroides were found in all samples (AFB0–AFB3) from outbreaks 1 and 2, but for outbreak 3 one AFB0 sample was negative for L. helsingborgensis, and one AFB 1 sample and one AFB 2 sample were negative for B. asteroids (Table 3, Appendix 1). Conversely, L. kimbladii was found in all samples from outbreak 3 but was not present in one sample from outbreak 1 and 5 samples from outbreak 2. B. coryneforme was found in all but one sample from outbreak 1 (93.3%) compared to in eight (53.3%) and six (54.5%) from outbreaks 2 and 3, respectively (Table 3, Appendix 1). A. kunkeei was detected in 12 (29.3%) of the overall samples and only in one sample from outbreak 3 (40% of outbreak 1 samples and 33.3% of outbreak 2 samples) (Table 3, Appendix 1).

3.2 Abundance of hbs-LABs

The abundance of hbs-LABs was compared between different AFB status groups as well as different outbreaks. While no significant differences were found between the different AFB status groups (Figure 2a), a Kruskal-Wallis comparison of samples from the three outbreaks revealed differences in four bacterial species: L. apis (p = .019), L. kullabergensis (p = .030), B. mellis (p = .008), and B. asteroides (p < .001). The post hoc analysis revealed that samples from outbreak 1 had higher abundance of B. mellis and B. asteroides than samples from outbreak 2 (p = .015 and .049 respectively) (Figure 2b) and a higher abundance of L. kullabergensis, B. mellis, and B. asteroides (p = .013, p = .004 and p < .001, respectively) than samples from outbreak 3 (Figure 2b). The abundance of L. kullabergensis was significantly higher in honey bees from outbreak 2 than in 3 (p = .030, respectively) (Figure 2b). Samples from outbreak 3 had a higher abundance of L. apis than both outbreak 1 and 2 (p = .032 and p = .007, respectively) (Figure 2b). Additionally, a significantly higher abundance of B. asteroides was found in samples from beekeeping operations with colonies without visible AFB symptoms (n = 13, p = .041, data not shown).

Overall, the community composition of the different bacterial communities was different between AFB groups. The group with visible AFB symptoms (AFB1) was moderately similar to the other three groups (Bray–Curtis dissimilarity index 0.52–0.70, Table 4), while the other three groups (AFB0, AFB2, and AFB3) were not very similar to each other (Bray Curtis dissimilarity index 0.1–0.27, Table 4).

3.3 Diversity of hbs-LAB

A comparison of the diversity between samples from P. larvae culture-positive (n = 20, Tables 1 and 5) and culture-negative (n = 21, Tables 1 and 5) colonies revealed that the mean Shannon's diversity index for the hbs-LAB was higher in culture-positive samples (p = .040, Figure 3c). However, no difference in diversity, as measured by Shannon's diversity index, was identified between sample groups when the three outbreaks and the different AFB status were compared (Figure 3a,b). There was almost no difference in community composition between the samples that were P. larvae positive compared to those that were not.