Lactobacillus spp. attenuate antibiotic-induced immune and microbiota dysregulation in honey bees

LX3 enhances larval pathogen eradication by antibiotics

Prophylactic administration of OTC to honey bees is a common practice in beekeeping for the prevention of AFB. To evaluate the efficacy of this long-standing apiculture management strategy, we monitored a 2-week treatment regimen with OTC under natural field conditions in honey bee hives experiencing low-grade chronic infection with P. larvae (Fig. 1a). Using a qPCR-based approach to enumerate pathogen load, P. larvae abundance was found to be significantly lower in honey bee larvae (primary target of AFB) at week 1 and week 2 of OTC treatment (Kruskal–Wallis with Dunn’s multiple comparisons, P = 0.0071 and P = 0.0005, respectively) compared to baseline levels at day 0 (Fig. 1b). In contrast, no observable differences in P. larvae abundance were found in adult honey bees (active vector of AFB) at any time point during this treatment (Kruskal–Wallis with Dunn’s multiple comparisons, P = 0.9999, P = 0.6367, respectively; Fig. 1c).

A robust body of scientific evidence shows that supplementation with probiotic Lactobacillus spp. can augment the effects of certain antibiotics and attenuate antibiotic-induced dysbiosis in humans and other animals^26,27,28,29. Testing this in honey bees, it was found that larval samples from LX3-treated hives exhibited significantly lower levels of P. larvae at week 2 and week 4 compared to both no treatment control (NTC; P < 0.0001 for both) and vehicle-treated (P = 0.0011 and P = 0.0014) hives, respectively (two-way analysis of variance [ANOVA] with Sidak’s multiple comparisons; Fig. 1d). The NTC group also demonstrated a trend towards higher P. larvae loads in larval samples compared to the vehicle group at week 4 (two-way ANOVA with Sidak’s multiple comparisons, P = 0.0568). Similar results were observed in adults with samples from LX3-supplemented hives demonstrating a significantly lower P. larvae load compared to both the NTC group (P = 0.0002 and P = 0.0003) and vehicle-treated group (P < 0.0001 for both) at 2 and 4 weeks, respectively (two-way ANOVA with Sidak’s multiple comparisons; Fig. 1e).

To compare the effects of OTC and LX3 treatments on overall hive health, the coverage of capped brood on hive frames (an established metric for assessing colony strength and reproduction³⁰) was measured weekly during experimentation. No observable differences were found in capped brood counts following 1 week of OTC treatment, whereas a significant reduction was found after 2 weeks (one-way ANOVA with Sidak’s multiple comparisons, P < 0.9999 and P = 0.0041, respectively) compared to pretreatment baseline measurements (Fig. 1f). In contrast, capped brood counts were significantly higher in LX3-treated hives at week 3 and 4 of the supplementation period (two-way ANOVA with Sidak’s multiple comparisons, P = 0.0071 and P = 0.0055, respectively), while no differences were found in NTC and vehicle-treated hives at comparable time points (Fig. 1g).

Antibiotics reduce key immune regulators in the adult gut microbiota

Given the broad-spectrum activity of tetracyclines, we evaluated how OTC exposure might influence the symbiotic bacterial communities associated with honey bees. Total bacterial loads, as determined by qPCR-based molecular quantification, were significantly reduced in adult bees following 1 week of OTC treatment (P < 0.0001) and in larvae (P = 0.0421) after 2 weeks of treatment (Kruskal–Wallis tests with Dunn’s multiple comparisons; Supplementary Fig. 2). The nurse-aged adult bees sampled at the experimental start point (pre-OTC exposure) and on the final day of OTC treatment (post-OTC exposure) were chosen for 16S rRNA gene sequencing-based microbiota analysis due to their close physical proximity with brood, passive carriage of P. larvae, and well-balanced representation of overall hive microbial diversity³¹.

Bar plots shown in Fig. 2a, b visually represent the relative proportion and absolute abundance (adjusted according to qPCR-based quantification of total bacteria) of taxa in samples from pre- and post-OTC exposure, respectively. Adult gut samples collected post-OTC exposure revealed a significant reduction in a single amplicon sequence variant (SV) of Frischella perrera (SV19; Wilcoxon test with Benjamini–Hochberg [BH] multiple comparisons, P = 0.0043) and three unique SVs of Lactobacillus Firm-5 (SV01, SV08, and SV10; Wilcoxon tests with BH multiple comparisons, P = 0.0151, P = 0.0295, and P = 0.04217, respectively) compared to samples collected pre-OTC exposure (absolute effect >0.5; Fig. 2c). SV74 (Lactobacillus Firm-4), SV63 (Bartonella apis), and SV54 (Lactobacillus Firm-4) showed a trend towards decreased relative abundance following OTC exposure (absolute effect <0.5; Fig. 2c).

The compositional 16S rRNA gene sequencing dataset also demonstrated differences in alpha diversity (intraindividual) and beta diversity (interindividual) metrics of the adult gut microbiota following OTC exposure (Fig. 2d, e). Specifically, Shannon’s H index (balanced alpha diversity metric taking species abundance and evenness into account) was determined to be significantly lower following 2 weeks of OTC treatment (Wilcoxon test, P = 0.0273; Fig. 2d). These findings also corresponded with a decrease in microbiota stability as demonstrated by a significant increase in Aitchison distance (i.e. Euclidean distance after center log-ratio transforming SV read counts) between adults within the same treatment group (Wilcoxon test, P < 0.0001; Fig. 2e).

Next, to determine how the standard practice of treating hives with OTC may influence accumulation of antibiotic-resistance genes in the honey bee gut microbiota, we screened seven tetracycline-resistance loci that have been repeatedly detected in honey bee guts⁶. These loci included five tetracycline efflux pump genes (tetB, tetC, tetD, tetH, and tetY) and two ribosomal protection protein-encoding genes (tetM and tetW). The abundance of tetB was significantly higher (two-tailed t test, P = 0.0092) whereas tetY showed a trend towards increased abundance (two-tailed Mann–Whitney test, P = 0.0887) in post-OTC adult gut samples (two-tailed Mann–Whitney tests; Fig. 2f). No observable change in abundance were found for any of the remaining five tetracycline-resistance genes examined in this study (Fig. 2f). A positive association was identified between total Gammaproteobacteria (Pearson correlation, r = 0.888, P = 9.25 × 10⁻¹⁷) and Gilliamella apicola (Pearson correlation, r = 0.719, P = 1.23 × 10⁻⁰⁸), but not Frischella perrera (Pearson correlation, r = −0.228, P = 0.124), and the presence of tetB (Supplementary Fig. 3).

LX3 improves adult microbiota recovery post-OTC exposure

In humans, probiotic therapy helps to encourage healthy remodeling of the microbiota and can improve recovery following antibiotics³². Here, we tested the ability of LX3 supplementation in honey bees to reduce P. larvae levels and restore microbiota homeostasis in adults and larvae following OTC exposure. As expected, principal component analysis showed clear separation between the microbiota composition of adult and larval samples (Fig. 3a). The largest influencers of separation that were positively associated with adult samples included core microbiota members such as G. apicola, Snodgrassella alvi, F. perrara, Commensalibacter, Lactobacillus Firm-4, and Lactobacillus Firm-5. In contrast, larval samples showed a positive association with mostly opportunistic bacteria including Escherichia/Shigella, Staphylococcus, Enterococcus, Pseudomonas, and P. larvae (Fig. 3a).

Using ALDEx2³³ software, Bland–Altmann-like effect plots were generated to investigate the relationship between differences in relative abundance for each SV between treatment groups and the within-group variance of each SV. While no discernible differences were observed within larval samples with an underdeveloped microbiota composition, a tightly clustered group of 13 SVs with low variance were identified in adult samples (Fig. 3b, c). Notably, 12 out of 13 of these SVs trended towards increased abundance in the microbiota of LX3-supplemented adults (Fig. 3b) and consisted of several well-characterized honey bee core microbiota members including five SVs of Lactobacillus Firm-5, two SVs of S. alvi, two SVs of G. apicola, one SV of Lactobacillus Firm-4, one SV of F. perrara, and one SV of Bifidobacterium (see Supplementary Data 1 for full list of SVs). Corroborating these results, LX3 supplementation demonstrated a rescuing effect on microbiota stability as demonstrated by a significant decrease in compositional differences between adult microbiota samples (measured via Aitchison distance; P = 0.0012), whereas no changes were observed in vehicle-treated adults (Kruskal–Wallis test with BH multiple comparisons, P = 0.7570; Fig. 3e).

No differences in overall alpha diversity, as determined by Shannon’s H Index (accounting for species abundance and evenness), were detectable in either vehicle or LX3 treatment groups (Kruskal–Wallis test with BH multiple comparisons, P = 0.2985 and P = 0.2348, respectively; Fig. 3d). However, further investigation demonstrated that species richness alone (determined using the abundance-based coverage estimator [ACE] algorithm in QIIME2) was significantly lower (P = 0.0428) and that compositional dominance (or unevenness, as determined by Strong’s Dw Index) trended towards being higher (P = 0.0535) in LX3-treated adults compared to vehicle-treated adults (Kruskal–Wallis test with BH multiple comparisons; Fig. 3f, g). Since these data did not provide conclusive evidence as to whether core SVs (dominant microbiota members) or noncore SVs (rare species and transient opportunists) were responsible for the observed differences in diversity indices, a nested compositional analysis was performed on total relative abundance for each group. LX3-treated adults demonstrated a significant enrichment in core SVs (effect size = 1.3631) and a reduction in noncore SVs (effect size = −1.3860; Kruskal–Wallis test with BH multiple comparisons, P = 0.0225; Fig. 3h). In contrast, no change was seen in the total relative abundance of core SVs (effect size = 0.1579) or noncore SVs (effect size = −0.1654) in the vehicle-treated controls (Kruskal–Wallis test with BH multiple comparisons, P = 0.2046; Fig. 3i).

LX3 strains are detectable in-hive members post-supplementation

Since 16S rRNA gene sequencing is unable to resolve bacterial taxonomy at the species level, we performed qPCR-based quantification of Lp39, GR-1, and BR-1 using established primer sets²⁵ to confirm that LX3 strains were being effectively dispersed throughout the hive as intended. LX3-supplemented adults and larvae were found to contain significantly higher levels of L. plantarum (multiple t tests, P = 0.0087 and P = 0.0035, respectively) and L. rhamnosus (multiple t tests, P = 0.0002 for both) compared to vehicle-treated control groups (Fig. 4a). In addition, evaluation of bacterial compositions in honey bee larval samples showed that abundance of P. larvae correlated inversely with L. plantarum (r = −0.442, P = 0.006) and L. rhamnosus (r = −0.456, P = 0.006), but not with L. kunkeei (r = 0.060, P = 0.724, Pearson correlations; Fig. 4b).

LX3 upregulates both head and gut immunity in adult bees

Repeated exposure to antibiotics can weaken immune defenses in honey bees—a phenomenon thought to be facilitated through the reduction of bacterial species important to immunoregulation³⁴. Using an established zone-of-inhibition (ZOI) assay as a crude measure of immune function³⁵, we assessed the inhibitory potential of honey bee-derived hemolymph against Arthrobacter globiformis. The antimicrobial capacity of adult hemolymph was found to be reduced by 31.27% (95% CI = 8.19–54.34%, one-way ANOVA with Sidak’s multiple comparisons, P = 0.0150) following 2 weeks of OTC treatment (Fig. 5a). In contrast, the antimicrobial capacity of hemolymph from LX3-supplemented adults was significantly increased by 121.30% (95% CI = 22.34–220.30%, two-way ANOVA with Sidak’s multiple comparisons, P = 0.0443) after 2 weeks in comparison to vehicle supplemented controls (Fig. 5b).

Previous work has shown that ex situ supplementation of LX3 to honey bee larvae can increase innate immune gene expression of several antimicrobial peptides (AMPs) that control susceptibility to P. larvae infection²⁵. Here, we tested how LX3 supplemented directly to the hive impacted adult bee immunity. Expression of nine well-characterized innate immune- and antioxidant-related genes (defensin-1, defensin-2, hymenoptaecin, apismin, apidaecin, VgMC, catalase, and lysozyme-1) were measured in adult head and gut tissue as their respective anatomical sites are known to play a major role in social and individual immunity³⁶. Exploratory analysis showed that basal gene expression levels in pre-supplemented adult bees demonstrated a vast enrichment of defensin-1 (900.0 ± 163.6-fold higher) and apismin (297,292 ± 63,498-fold higher) in adults heads relative to gut expression, as compared to the other AMP genes examined (one-way ANOVA with Sidak’s multiple comparisons, P < 0.0001 for both; Fig. 5c)—suggesting that changes to the expression of these genes in the head, as opposed to the gut, might produce a more potent immune response and protective benefit at the colony level.

During the supplementation period, head expression of defensin-1, apismin, and apidaecin showed a significant increase over time in only the LX3-supplemented group (two-way ANOVAs with Sidak’s multiple comparisons, P = 0.0406, P < 0.0001, and P = 0.0004, respectively) whereas expression of hymenoptaecin increased in both LX3 and vehicle supplemented groups (two-way ANOVAs with Sidak’s multiple comparisons, P = 0.0240 and P = 0.0365, respectively; Supplementary Fig. 4A, C–E). All experimental groups showed an increase in head expression level of catalase (two-way ANOVA with Sidak’s multiple comparisons, P = 0.0050, P < 0.0001, and P < 0.0001, respectively) over time with no changes in expression of defensin-2 or lysozyme-1 (Supplementary Fig. 4B, H, I). Gut expression of defensin-1, hymenoptaecin, apidaecin, and VgMC were exclusively increased by LX3 supplementation (two-way ANOVAs with Sidak’s multiple comparisons, P = 0.0004, P = 0.0201, P < 0.0001, and P < 0.0001, respectively), whereas lysozyme-1 and catalase expression increased in both vehicle (P = 0.0013 and P < 0.0001, respectively) and LX3 (P < 0.0001 for both) treatment groups (two-way ANOVA with Sidak’s multiple comparisons; Supplementary Fig. 4J, M, N, P–R and Supplementary Data 2). Relative expression of immune- and antioxidant genes in head and gut samples are visually summarized by the PCA plots and heatmaps in Fig. 5d–g.

Correlations between host gene expression and bacterial loads

It has been known for decades now that probiotic bacteria can induce an immune response in honey bees³⁷. In addition, recent evidence shows that host bacterial communities are selectively regulated by the innate immune system of honey bees, and that core microbiota members demonstrate a higher level of resistance to host AMPs than opportunistic bacterial pathogens³⁸. Here, we examine the simultaneous relationship between bacterial abundances and immune- and antioxidant-related gene expression by using a dual extraction-based method to derive RNA and DNA from adult heads and guts. Experimental end point measurements at week 4 of the supplemental period demonstrated a negative Pearson correlation between total abundance of P. larvae and expression of apidaecin (r = −0.589, P = 0.004), apismin (r = −0.483, P = 0.023), hymenoptaecin (r = −0.460, P = 0.036), defensin-1 (r = −0.599, P = 0.004), and catalase (r = −0.650, P = 0.001) in head samples, and with apidaecin (r = −0.559, P = 0.007), hymenoptaecin (r = −0.467, P = 0.029), and catalase (r = −0.589, P = 0.004) in gut samples (Fig. 6).

Assessment between the supplemented strains of lactobacilli and immune or antioxidant gene expression showed a positive relationship (Pearson correlations) between L. plantarum abundance and head expression of apidaecin (r = 0.653, P = 0.001), defensin-1 (r = 0.526, P = 0.014), and catalase (r = 0.435, P = 0.043), as well as gut expression of defensin-2 (r = 0.401, P = 0.008) catalase (r = 0.591, P = 0.004), and VgMC (r = 0.468, P = 0.028; Fig. 6). For L. rhamnosus, bacterial abundance was associated with increased head expression of apidaecin (r = 0.454, P = 0.039) and a trend towards increased gut expression of defensin-2 (r = 0.401, P = 0.072). No correlations were observed between gene expression and abundance of L. kunkeei. Core microbiota members demonstrated varying relationships with head and gut expression of immune- and antioxidant-related genes in adult honey bees. Total abundance of Alphaproteobacteria, Betaproteobacteria, Gammaproteobacteria, Bifidobacterium, and F. perrara clustered together (based on Euclidean distance of Pearson correlation coefficients) and were found to be mostly associated with increased AMP gene expression (Fig. 6). Oppositely, total abundance of Bacteroidetes, Firmicutes, and G. apicola clustered together and were mostly associated with an overall decrease in immune- and antioxidant-related gene expression irrespective of body site. A notable exception to these trends was the gut expression of apismin and defensin-1 as well as the head expression of defensin-2 and lysozyme-1, which clustered together (Euclidean distance matrix) based on similarities in gene expression patterns relative to bacterial abundances (Fig. 6). Pearson correlation coefficients (and associated statistics) for all relationships between bacterial abundances and host gene expression are provided in Supplementary Data 3.

Source: Ecology - nature.com