The European honey bee Apis mellifera is the most important pollinator present in the agricultural (Alberoni et al., 2016) environmental and economic sectors (Klein et al., 2017). Over the last few decades, major mortalities in honey bee colonies have been reported globally (Fairbrother et al., 2014; Vanegas, 2017). With the intensive exploitation of honey bees for the pollination of agricultural crops, bee colonies mortality is more accentuated in some area as Europe and US than in others part of the world. Although the stock of honey colonies is increasing, it fails to meet our actual needs (Aizen and Harder, 2009; Smith et al., 2014). Numerous studies agree that synergetic interactions between multiple abiotics (Doublet et al., 2014; Alburaki et al., 2015; Kakumanu et al., 2016; Poquet et al., 2016; Li et al., 2017; López et al., 2017) and biotics stressors (Goulson et al., 2015; Dussaubat et al., 2016) are involved as major causes of bee colonies decline (see also reviews by Fairbrother et al., 2014; Alberoni et al., 2016; Sánchez-Bayo et al., 2016; Klein et al., 2017). Synergies between many of these factors have a strong negative impact on immune defense (Nazzi et al., 2012; Di Prisco et al., 2013), metabolism (Koch and Schmid-Hempel, 2011; Dussaubat et al., 2013; Bordier et al., 2017), and on bees cognitive mechanisms (Klein et al., 2017).

Now, honey bee gastrointestinal microorganisms (i.e., gut microbiota) are known to play a critical role in regulating specific functions associated with metabolism and immune response (Evans and Lopez, 2004; Alberoni et al., 2016; Hyrsl et al., 2017). Microbial symbionts participate in various processes such as digestion of food (Koch and Schmid-Hempel, 2011), detoxification of harmful molecules, supply of essential nutrients, participation in the host's defense system (Lemaitre and Hoffmann, 2007; Hooper et al., 2012), and protection against pathogens and parasites (Evans and Lopez, 2004; Flint et al., 2012; Engel and Moran, 2013; Alberoni et al., 2016; Chaplinska et al., 2016). Although our understanding of mechanisms underlying interactions between the intestinal microbiota and parasites is still partial, it is widely acknowledged that the composition of the intestinal microbiota is a key factor in controlling infection dynamics (Berrilli et al., 2012). To this respect, members of the gut microbiota, including parasites, can be either beneficial, neutral or harmful to their host (Engel and Moran, 2013). Furthermore, according to the host physiological state, some strains can shift from opportunistic to infectious when the host is stressed (Boutin et al., 2013), and potentially favor parasite infections. Nosema ceranae is an obligate intracellular microsporidian parasite, which is the causative agent of one of the most prevalent honey bee disease, called nosemosis (Higes et al., 2008). Honey bee's infection occurs by proliferation of N. ceranae spores in the midgut after the incorporation of infected food (Ptaszynska et al., 2014) and spore accumulation is often associated with mortalities (Higes et al., 2007, 2008, 2009; Martín-Hernández et al., 2007).

Nosemosis disease is involved in physiological and behavioral perturbations in the honey bee colony (Leoncini et al., 2004; Goblirsch et al., 2013) like the disruption of the pheromone production, the ethyl oleate (EO). Ethyl oleate (EO) is involved in the behavioral maturation, in the transition from hive work to foraging activities (Leoncini et al., 2004; Dussaubat et al., 2010, 2013) and suppression of the cellular immune response (Antúnez et al., 2009). Nosemosis is also correlated with a modification of the natural feeding behavior of honey bees, making them less cooperative to participate in trophallaxis process (Naug and Gibbs, 2009). Nosema ceranae infection mainly impacts colony robustness (Cox-Foster et al., 2007; Antúnez et al., 2009): infected honey bees are more susceptible to other pathogens (Mayack and Naug, 2009).

Furthermore, spore accumulation is suspected to cause gut microbiota dysbiosis and to impair the defense system of the host (Alberoni et al., 2016). Dussaubat et al. (2012) showed that N. ceranae triggers inhibition of gene expression involved in the regeneration of the gut epithelial tissue (e.g., Wnt signaling pathway), which in turn could affect the ecological niche of beneficial host microbes. Gut epithelial tissue damages are symptoms of intestinal diseases, including diarrhea. Overall, intracellular intestinal protozoan parasites impact can be extremely damaging, especially in individuals who have weakened immune system (Vitetta et al., 2016). Given that modern agroenvironments exert a strong negative impact on immune defense, (Alaux et al., 2010; Di Prisco et al., 2013), it is not surprising that nosemosis prevalence is rapidly increasing (Higes et al., 2008).

Current treatment against nosemosis homologated in many countries (for example USA and Canada) is the antibiotics Fumagillin-B® and Fumidil-B® (Van Den Heever et al., 2015). Unfortunately, Nosema spp. strains are showing various levels of antibiotic-resistance (Huang et al., 2013) and several studies have measured reduced Fumagillin-B® efficacy on both N. apis and N. ceranae infections (Pajuelo et al., 2008; Williams et al., 2011). For this reason, Fumagillin-B® was recently banned in the European Union (Gisder and Genersch, 2015) like in France in 2002 (Fernandez and Coineau, 2007). Furthermore, antibiotics showed to disrupt host microbiota equilibrium by killing endogenous bacteria (Aguilera et al., 2013; Li et al., 2017). Moreover, antibiotics were observed to increase the susceptibility to Nosema infection, thus inducing a negative impact on the honey bee lifespan (Li et al., 2017).

Therefore, there is a need to develop effective and sustainable alternative strategies to maintain or restore intestinal flora homeostasis. Alternatively, beneficial microorganisms (i.e., probiotics) were successfully used as therapeutic tools in animal production (Hamdi et al., 2011; Alberoni et al., 2016). Oral administration of probiotics has showed a protective effect on the intestinal microbiota of the host (Gismondo et al., 1999). In murine models, both bacterial and yeast probiotic strains have also proven their effectiveness against intracellular parasites and showed positive improvement of the health of host (Alak et al., 1997, 1999; Benyacoub et al., 2005; Humen et al., 2005; Shukla et al., 2008; Dinleyici et al., 2011). Concerning honey bees, endogenous gut bacteria belonging to Lactobacilliaceae, Bifidobacteriaceae, and Acetobacteraceae families have been found to play an antagonistic role against N. ceranae, either in caged bee trial or in field conditions, by reducing the spore load (Sabaté et al., 2012; Audisio et al., 2015; Baffoni et al., 2016; Corby-Harris et al., 2016).

Our research goal focused in comparing the probiotic potential of two endogenous strains [Parasaccharibacter apium (PC1 sp.), Bacillus sp. (PC2 sp.)] and two commercial probiotics strains (Bactocell® and Levucell SB®, Lallemand Inc.) as an alternative treatment against N. ceranae infections in honey bees. These four probiotic candidates were targeted either for their proven or putative probiotic properties. The endogenous strains [P. apium (PC1 sp.) and Bacillus sp. (PC2 sp.)] are respectively closely and distantly related to strains that have been successfully used to inhibit honeybee disease (Sabaté et al., 2009; Yoshiyama and Kimura, 2009; Corby-Harris et al., 2016) and to enhance colony performance (Sabaté et al., 2012). Parasaccharibacter apium (PC1 sp.) belongs to the Acetobacteraceae family, which is particularly abundant in honey crop, hypopharyngeal glands, royal jelly, and larval gut through nurse worker bees feeding behavior (Corby-Harris et al., 2016). Overall, commensal Acetobacteraceae affect tissue development in insects such as in Anopheles mosquitoes (Chouaia et al., 2012; Mitraka et al., 2013) and control maturation of gut immunity, such as in Drosophila melanogaster (Ryu et al., 2008). Then, Bacillus sp. (PC2 sp.) belongs to the phylum Firmicutes and more specifically to the Bacillus genus. Bacillus species are dominant in the honeybee stomach (Wang et al., 2015). This genus can survive at high temperatures and pH and is usually harmless, with the exception of B. anthracis and B. cereus (Sabaté et al., 2012). Administration of Bacillus strains helped the development of bee colonies by enhancing the brood and also honey yield (Sabaté et al., 2012). Concerning the commercial strains, Bactocell® (Pediococcus acidilactici), is a Lactobacillaceae that is used for a wide range of host species: pigs, chickens, shrimps, and salmonids (EFSA, 2012). Bactocell® was demonstrated to improve the laying performance of hens (Denev et al., 2013), the survival rate, and the growth of the shrimp L. Stylirostris (Castex, 2009). Lactobacillacea strains are used as probiotics formulations in animals (Nikoskelainen et al., 2003; Bovera et al., 2012; Piccolo et al., 2015; Billiet et al., 2017) and in honey bees (Evans and Lopez, 2004). Interestingly, a decrease of Lactobacillaceae sp. was correlated with unhealthy honey bees (Cox-Foster et al., 2007; Olofsson and Vásquez, 2008), thus highlighting the important role of this bacterial family in honey bee health. Levucell®SB (Saccharomyces cerevisiae boulardii) is used for monogastrics (swine and chicken) to improve growth performance, gut histology (Le Bon et al., 2010), decrease carriage of foodborne pathogens (Mountzouris et al., 2015) and improve immunity response (Collier et al., 2011). Moreover, other studies showed positive impact of commercial probiotic on bee host (Kazimierczak-Baryczko and Szymas, 2006; Ptaszynska et al., 2016). Consequently, we decided to test the impact on honey bee health of two commercial strains, Bactocell® and Levucell®SB that showed beneficial effects on other animals.

The main objective of the present study was to compare the survival of young caged bees fed with two endogenous honey bee gut bacteria (PC2 and PC1) and two commercial probiotics (Bactocell®, Levucell SB®, two trademarks of Lallemand Inc.), experimentally infected by N. ceranae. This work is one of the very first studies comparing commercial probiotics, which usually have a broad host range, with honey bee endogenous strains, to assess whether beneficial members of the commensal gut microbial community perform better both in prophylactic and curative conditions. The motivation of this study is that probiotic curating strategies based on broad range commercial probiotic strains such as acid bacteria commercial strains such as L. rhamnosus and Lactobacillus sp. were not able to prevent nosemosis, and disrupted the honeybee immune system (Andrearczyk et al., 2014; Ptaszynska et al., 2016). Moreover, Andrearczyk et al. (2014) observed an increase of Nosema spp. infections in young honeybees fed with commercial Lactobacillus sp. and Saccharomyces cerevisiae probiotic strains. Finally, some cases of commercial probiotic induced microbiota-host perturbations were documented (Doron and Snydman, 2015; Durchschein et al., 2016). Taken together, this suggests that some broad range probiotics are not necessary suitable for the honey bees nutrition (Andrearczyk et al., 2014).

The present study shows that both endogenous bacterial strains were at least as much efficient and safe as broad spectra probiotics in significantly increasing survival in the context of experimental infection with N. ceranae.

The results clearly showed that both prophylactic and curative administration of all the four probiotics candidates, either with broad host range and endogenous, significantly increased honey bees survival rates from 20 to 40% when compared to the non-infected control group. For a prophylactic administration, highest survival rates were obtained for (in decreasing order): PC1, Bactocell®, Levucell SB®, and PC2. For the curative administration, highest survival rates were obtained for (in decreasing order): Levucell SB®, PC1, PC2 and Bactocell®. However, there was no significant difference in terms of survival improvement between commercial and endogenous probiotic candidates.

Regarding the patterns of survival rates, a drastic decrease was observed at day 14 in both control and Nosema control. From day 14 to the end of the trial (day 27), the Nosema control group survival was always lower than the control without Nosema, thus confirming that Nosema infection exerted a negative effect on survival. Furthermore, this drastic decrease did not occurred in the groups fed with probiotics, either with or without Nosema infection. This result strongly suggests that probiotics were very efficient to improve bee survival both in curative and prophylactic mode of action. Regarding the cause itself of this drastic survival decrease at day 14 in both control and Nosema control, we observed a similar pattern of sudden survival decrease in a preliminary trial (data not shown), although to a lower extent. We would hypothesize that maintaining honey-bees in cages fed without protein intake (i.e., pure 1:1 sugar syrup) triggered microbiota dysbiosis and in turn gut inflammation. In the probiotic fed groups, it is very likely that both anti-microbial effects and positive immune response stimulations, which were demonstrated for both Bactocell^®; and Levucell SB®, in alternative host organisms (Qamar et al., 2001; Standen et al., 2013), mitigated gut dysbiosis, thus increasing survival rate. Then, from day 15 onwards, we observed significant differences regarding survival differential between curative groups (i.e., Nosema + probiotic vs. Nosema control) and prophylactic groups (i.e., probiotic vs. control; Table 7). Basically, survival differential between curative and prophylactic groups varied from one experimental group to another, and from one sampling time to another. This result suggest that under curative use, at least three out of the four tested probiotic strains had a non-additive effect on survival relative to prophylactic use. Therefore, although it may suggest these three probiotic strains exert a specific effect on Nosema activity, further experiments are needed to univocally assess whether there is a direct functional interaction with the parasite.

Finally, despite of the significant increase of survival rate in curative groups, no significant reduction of Nosema spores was observed in any of the four probiotic groups. At first sight, this result may be surprising as significant decreases of N. ceranae spore loads were recorded in previous studies: Baffoni et al. (2016) combined several strains of microorganisms isolated from the honeybee gut and found that both Bifidobacteriaceae and Lactobacilliaceae strains were correlated with a significant decrease of N. ceranae spore loads. Maggi et al. (2013) also reported a 52% reduction of N. ceranae spores per bee following application of the bacteria L. Johnsonii metabolites in caged bees. Finally, Corby-Harris et al. (2016) documented a significant reduction of Nosema spores following the administration of the endogenous strain P. apium. However, two key parameters from these previous studies are different from those of the present work, and could explain, at least partly, contrasted results in terms of spore reduction. First, relatively low titers (ranging from 1.10³ to 1.10⁴ spores per microliter of sugar syrup) of N. ceranae spores were administrated to caged honey bees to induce nosemosis (Maggi et al., 2013; Baffoni et al., 2016; Corby-Harris et al., 2016). Here, administrated N. ceranae spore titer was 100 to 1,000 times higher (1.10⁶ spores per microliter of sugar syrup). Second, probiotic titers administrated in the present study were 100 to 1,000 times lower to previous studies (Baffoni et al., 2016; Corby-Harris et al., 2016). Interestingly, Corby-Harris et al. (2016) administrated a closely related strain of the endogenous symbiont P. apium at 2.10⁶ CFU per mL, which is two thousand times higher than what was administrated in the present work. Despite of this, N. ceranae titers remained high enough, that the statistically significant reduction of spores recorded in those previous studies was not biologically relevant (Alberoni et al., 2016) to clearly reflect an antagonistic effect on nosemosis, as reduction of spore load did not exceeded 40%, when compared to infected control group. Indeed, workers from infected colonies exhibited mean nosema counts ranging from 0.04 to 1.87.10⁶ spores per bee (Traver and Fell, 2011). In the caged bee trial of Corby-Harris et al. (2016) however, the probiotic fed group housed a mean of 951.10⁶ spores per bee. Also, no survival improvement was recorded in all experimental groups (Corby-Harris et al., 2016). This is not surprising when considering our results as significant differences in survival were observed after 14 days, whereas the duration of the challenge with Nosema spores in Corby-Harris et al. (2016) was 10 days. A peak of bee mortality at day 14 post infection was previously observed (Dussaubat et al., 2012) thus corroborating our results regarding N. ceranae infection dynamics.

More importantly, the probiotic feed additive administration protocol used by Corby-Harris et al. (2016) was quite different from the present study. Corby-Harris et al. (2016) administrated P. apium probiotic strain (2.10⁶ CFUs) during 7 weeks into the pollen patty on colony overwintering, and thus, prior to the cage bee trial. In our experiment, probiotics strains (1.10³ CFUs per mL) were administrated via sugar syrup during the cage trial. Then, in the caged bee trial, young honeybees had ad libitum access to unsterilized pollen patty. Such protocol could have further improved overall survival, by providing bees with both proteins and other bacterial strains. Furthermore, Corby-Harris et al. (2016) used young spring bees from the overwintering hives for their cages experience, whereas we used winter bees. Taken together, the higher load of Nosema spores combined with lower concentration of probiotic candidates and the lack of protein supply in the present work, likely explain the absence of spore load decrease. However, all the four probiotic candidates significantly improved bee survival by 20–40%. Although it is yet not possible to state whether probiotic effects could directly or indirectly interfere with Nosema functional activity, it is reasonable to hypothesize that probiotic effect allows honey bee to better tolerate high loads of Nosema spores. Functional interaction of N. ceranae with honey bee was essentially investigated by focusing on enzyme regulation involved in detoxification process, immune response, and the metabolism activity. At the gene-expression level, honey bee midgut responded to N. ceranae infection by increasing oxidative stress pathway with the concurrent synthesis of antioxidant, defense, and protective molecules (Dussaubat et al., 2012; Di Pasquale et al., 2013). At the proteomic level, differential abundance of proteins associated to metabolism, response to oxidative stress, and apoptosis were measured between sensitive and resistant honey bees (Kurze et al., 2016). Other studies focused on other genetic mechanisms such as microRNA, which were suspected to regulate host honey bee metabolism related genes in response to N. ceranae infection (Huang et al., 2015, 2016). Overall, honey bee health is a complex system resulting from a combination of functional interacting networks between the bee host, the symbiotic gut microorganisms (Anderson et al., 2011) and the pathogens (Schwarz et al., 2015). To this respect, symbiotic gut microorganisms play a central role in modulating energy reserves and immune system development, which are key actors for host defenses against parasites (Habtewold et al., 2017; Knutie et al., 2017). Immune resistance mechanisms can either prevent parasites and pathogens from invading or eliminate them after invasion. Contrastingly to resistance mechanisms, immune tolerance mechanisms primarily involve both a lowered responsiveness to a given immune stimulus, and the initiation of tissue repair processes to protect the host from damage caused by the pathogen (Habtewold et al., 2017). Honey bee tolerance regarding high N. ceranae spores was documented in Huang et al. (2012), where a lower mortality rates were correlated to an up-regulated immune response in tolerant honey bee colonies day 6 post infection, when compared to control groups. Therefore, a balance between immune resistance and tolerance mechanisms ensures protection from parasites and/or pathogens and, at the same time, limits the tissue inflammation that may arise as a result of excessive immune effector production. As it is now suspected that host microbiota is fully involved in both resistance and tolerance mechanisms (Vitetta et al., 2016), it is worth reviewing known prophylactic and curative effects of both commercial probiotic strains tested in the present study. Commercial strains are administrated to a wide panel of host organisms as food additives to increase zootechnical performances by improving both metabolism and immune response (Denev et al., 2013; Schoster et al., 2013; Hou et al., 2015). To this respect, Bactocell® (P. acidilactici) was demonstrated to exert a protective action by stimulating the intestinal mucosal cells activity inducing a positive response of the immune system of the tilapia (Oreochromis niloticus; Standen et al., 2013). Overall, administration of lactic acid bacteria (LAB) acts as an inhibitor of bacterial infection propagation by activating the host immune system (Servin, 2004). Interestingly, it was observed in mammals that administration of LAB has the potential to alter gut microbial diversity, affect intestinal barrier function, and modulate innate immune response (e.g., via NF-κB pathway regulation) by increasing mucosal barrier function, thus leading to a decrease in translocation of bacteria and a subsequent decrease in the ability of pathogenic bacteria to attach to the gut mucosa (Reiff and Kelly, 2010). As insects and mammals share the same NF-κB mediated induction of innate immunity (Hoffmann et al., 1999), one of the putative mechanism underlying survival improvement would be that Bactocell® (P. acidilactici) may exert an effect on the honey bee immune system itself rather than directly interfere with Nosema activity. Then, other mechanism favoring honey bee survival may be involved. Lactic Acid bacteria (LAB) produce active antimicrobial molecules such as bacteriocins, antibiotics, hydrogen peroxide, to name a few. Those excreted molecules can control growth and/or survival of surrounding microorganisms, including viruses, bacteria, yeast, fungi, and protozoan parasites (Cleusix et al., 2007). Also, by lowering the local intestinal pH with lactic acid, LAB are able to inhibit growth of acid-sensitive organisms (Wohlgemuth et al., 2010). Saccharomyces cerevisiae boulardii strains, to which belong Levucell SB®, are nonpathogenic yeast which have proven experimental effectiveness in preventing and curating diseases with a predominant inflammatory component, thus indicating this probiotic might interfere with inflammatory induced cellular signaling pathways (Palma et al., 2015). Saccharomyces cerevisiae boulardii strains have a broad anti-pathogenic action, due to its outer surface molecular profile, which attaches to pathogenic bacteria such as Escherichia coli and Salmonella. Furthermore, S. cerevisiae boulardii is able to effectively compete with and displace parasitic yeast strains such as Candida (Tomičić et al., 2016). S. cerevisiae boulardii also produces anti-fungal substances such as capric, caprylic, and caproic acids, which were proven to inhibit pathogenic yeast growth in its close environment. Finally, S. cerevisiae boulardii promotes the secretion of immune molecules (e.g., IgA), which decrease the action of pro-inflammatory cytokines pathways generated in response to pathogenic bacteria, toxins and antigens (Qamar et al., 2001). Although little is known about the efficacy of S. cerevisiae boulardii against protozoan parasites, three studies documented a significant effect of this probiotic against amoebiasis (Dinleyici et al., 2009), giardiasis (Besirbellioglu et al., 2006), and infection with Blastocystis hominis (Dinleyici et al., 2011). Finally, Lee et al. (2007) documented the combination of P. acidilactici and S. cerevisiae boulardii on coccidiose. However, the molecular mechanism remains to be characterized.

Overall, experimental studies on both bacterial and yeast probiotic action against parasites involving the secretion of an active principle that can inhibit parasite development was only documented for Cryptosporidium, Giardia, and Eimeria, even though the molecular nature of the active component is still to identify (Travers et al., 2011).

This study confirms the potential of endogenous strains isolated from the honey bee as efficient probiotic candidates to improve honeybee survival. Then, we can conclude that the mortality of honey bees was not directly related to N. ceranae spore loads. Furthermore, as the experiment duration was longer enough to cover up to two parasite life cycles (6 days), it strongly suggests that all of the four probiotic strains did not exert direct antagonism effect on the parasite development. Our results, combined with the previous documented probiotic effects of strains related to our four candidates, suggest that mechanism of action is more related to tolerance, rather than resistance against the parasite, potentially by improving the immune system and tissue repair processes to protect the host from damage caused by the parasite. Also, our results, combined with the other experimental infection trials with N. ceranae suggest that both Nosema inoculated titers and probiotic strains concentrations have to be taken into account to properly interpret results, as different mechanisms of tolerance or resistance are potentially involved. Finally, these encouraging preliminary results for the four probiotic candidates were obtained in vivo with minimal nutritional inputs (e.g., no proteins); therefore in situ validation is necessary to fully assess the potential of these probiotic candidates as sustainable preventive and curative tools against nosemosis.

Extended studies are definitely needed to understand the mechanisms underlying the functional interaction between the honey bee, its microbiota, the probiotic strain and the parasite. The metatranscriptomic approach, which consists in our case to measure the gene expression level of all organisms inhabiting the honey bee gut microbiome, including host epithelial cells, is straightforward to better understand the molecular mechanisms underlying the improvement of honey bee survival.

PG and ND wrote the grant, conceived the experiments, supervized all the experiments, and wrote the manuscript; SE performed data analyses, bacterial cultures, endogenous strains molecular identification, and wrote the manuscript; AR performed the caged bee trials and wrote the manuscript; AL performed Nosema spore counts and statistical analyses; BC supervized statistical analyses; SB performed bioinformatic analyses to identify endogenous probiotic strains and designed primers for PCR detection; VD and MC provided their expertise in probiotic development along the project and co-write the manuscript; P-LM was involved in every steps of molecular biology protocols (bee dissection, DNA extraction, PCR, sequencing) and probiotic cultures.