Identification of Candidate Agents Active against N. ceranae Infection in Honey Bees: Establishment of a Medium Throughput Screening Assay Based on N. ceranae Infected Cultured Cells

Many flowering plants in both natural ecosytems and agriculture are dependent on insect pollination for fruit set and seed production. Managed honey bees (Apis mellifera) and wild bees are key pollinators providing this indispensable eco- and agrosystem service. Like all other organisms, bees are attacked by numerous pathogens and parasites. Nosema apis is a honey bee pathogenic microsporidium which is widely distributed in honey bee populations without causing much harm. Its congener Nosema ceranae was originally described as pathogen of the Eastern honey bee (Apis cerana) but jumped host from A. cerana to A. mellifera about 20 years ago and spilled over from A. mellifera to Bombus spp. quite recently. N. ceranae is now considered a deadly emerging parasite of both Western honey bees and bumblebees. Hence, novel and sustainable treatment strategies against N. ceranae are urgently needed to protect honey and wild bees. We here present the development of an in vitro medium throughput screening assay for the identification of candidate agents active against N. ceranae infections. This novel assay is based on our recently developed cell culture model for N. ceranae and coupled with an RT-PCR-ELISA protocol for quantification of N. ceranae in infected cells. The assay has been adapted to the 96-well microplate format to allow automated analysis. Several substances with known (fumagillin) or presumed (surfactin) or no (paromomycin) activity against N. ceranae were tested as well as substances for which no data concerning N. ceranae inhibition existed. While fumagillin and two nitroimidazoles (metronidazole, tinidazole) totally inhibited N. ceranae proliferation, all other test substances were inactive. In summary, the assay proved suitable for substance screening and demonstrated the activity of two synthetic antibiotics against N. ceranae.

Introduction colonies was effective [37,54]. It was even speculated that fumagillin may contribute to increased prevalence and pathogenicity of N. ceranae instead of being a curative measure against N. ceranae [53]. This situation makes it even more pressing to find new agents active against N. ceranae.
The quest for new drugs against honey bee pathogens has so far been performed via animal experiments using larvae, adult bees or entire colonies [55][56][57][58]. Such experiments are time consuming, difficult to standardize, and only feasible during the bee season when appropriate bee material is available. In addition, although using invertebrates in experiments does not pose any ethical problem, these experiments still are animal experiments and it is desirable to replace them. Cell culture based assays are a solution to these problems. They can be standardized by using a cell line. They are independent from the bee season and rather fast allowing testing thousands of substances per year. Promising candidate substances will still have to be tested in vivo, however, the number of bee experiments can be reduced to the absolute necessary minimum. Therefore, such an assay could provide a major progress in honey bee drug discovery and testing.
Here we present our results on establishing such an in vitro assay for medium throughput screening of substances with putative activity against N. ceranae. The assay is based on our recently developed cell culture model for Nosema spp., the lepidopteran cell line IPL-LD 65Y. This cell line originating from Lymantria dispar has been shown to be susceptible to N. ceranae infection and to support the entire life cycle of N. ceranae [20]. We combined the cell culture assay with quantitative detection of N. ceranae in infected cells via an RT-PCR-ELISA (reverse transcriptase-polymerase chain reaction-enzyme linked immunosorbent assay) protocol and adopted the assay for the 96-well microplate format to enable automated analysis. We first verified our test system using fumagillin as positive control, paromomycin as negative control, and clioquinol as a highly cytotoxic agent, to cover the possible range of effects on both cells and microsporidia. We then tested several commercially available agents for cytotoxicity and inhibition of N. ceranae intracellular development and demonstrate the efficacy of the synthetic antibiotics metronidazole and tinidazole against N. ceranae. cells was also tested. Different working solutions of each substance were prepared to ensure that always only 1 μl had to be added to the cells in order to achieve the desired final concentrations.
For in vitro-cell viability tests a colorimetric assay (MTT test) was used. Cells were harvested at exponential growth phase, transferred to a 50 ml Falcon tube (VWR International, Darmstadt, Germany), and cell concentration was determined using a hemocytometer (Neubauer-Improved, VWR, Darmstadt, Germany). Cells were pelleted by centrifugation at 210 g for 10 min, the medium was aspirated, and fresh TC-100 cell culture medium, supplemented with 11% FCS, 250 μg/ml penicillin/streptomycin and 250 μl antibiotic/antimycotic-solution (Sigma Aldrich, Taufkirchen, Germany), was added to achieve a final cell density of 2.5E+05 cells/ml. 100 μl (2.5E+04 cells) were transferred into each well of a 96-well microplate. Subsequently, 1 μl of the appropriate working solution of each substance was added to the cells and cells were incubated for 72 hour at 27°C.
After 72 h, the microtiter plates were centrifuged at 210 g for 10 min to pellet the cells before the medium was carefully aspirated. Subsequently, 100 μl of 3-(4,5-dimethylthiazol-2-yl)-2,5diphenyltetrazolium bromide (MTT, Carl Roth, Karlsruhe, Germany) at a concentration of 0.5 mg/ml TC-100 cell culture medium (supplemented with 11% FCS, 250 μg/ml penicillin/ streptomycin and 250 μl antibiotic/antimycotic-solution) were added and cells were incubated for 3 h at 27°C. Viable cells convert the water soluble MTT into the water insoluble formazan. For formazan quantification, the cells in the microtiter plate were pelleted by centrifugation at 210g for 10 min. Subsequently, the cells were washed with 100 μl 1 x phosphate buffered saline (PBS) to remove residual medium. Cells were lysed and formazan was solubilized by adding 100 μl lysis-buffer (99.4% DMSO/0.6% acetic acid/10% SDS). Proportion of dead cells was analyzed at 595 nm excitation wavelength using an ELISA-reader (Synergy HT, BioTek, Bad Friedrichshall, Germany).

Infection of IPL-LD 65Y cells and application of test substances
For isolation of N. ceranae spores, infected honey bees were identified via qualitative microscopic analysis and molecular species differentiation was performed essentially as already described [42,59]. Dissection of infected honey bees and isolation of N. ceranae spores from midguts for infection of cell cultures were performed as recently described [20]. Approximately 1E+08 freshly isolated N. ceranae spores were transferred into 1.5 ml reaction tubes (Eppendorf, Hamburg, Germany), dried in a vacuum concentrator (Eppendorf) for 30 min at 30°C, and immediately used for infection [20].
For infection, IPL-LD 65Y cells were harvested at exponential growth phase and pelleted via centrifugation at 210 g for 5 min. The cell pellet was washed twice with 1 ml of freshly prepared 0.1 M sucrose in 1xPBS and resuspended in sucrose buffer at a concentration of 2.5E+07 cells/ml. Immediately prior to infection, germination of the dried spores was initialized by resuspending the spores with 400 μl of 0.1 M sucrose in 1xPBS [42,60]. Germinating spores (1E+08) were resuspended in 100 μl cell suspension (2.5E+06 cells, resulting in a multiplicity of infection (MOI) of 40) and the cell-spore suspension was incubated for 5 min at room temperature. Infected cells were resuspended in 9.5 ml TC-100 cell culture medium supplemented with 11% FCS, 250 μg/ml penicillin/streptomycin, and 250 μl antibiotic/antimycotic-solution (Sigma Aldrich). Finally, 100 μl of the cell suspension (2.5E+04 infected cells) were carefully transferred into each well of a 96well microplate (VWR).
For substance evaluation, 1 μl of each substance was added to the infected cells to achieve the desired final concentrations. Cells were incubated for 72 h at 27°C and, subsequently, infection status was determined independently both via PCR-ELISA and microscopic analysis. Each substance was tested in three independent replicates for both approaches.

Molecular analysis of the infection status of IPL-LD 65Y cells via RT-PCR and RT-PCR-ELISA
For molecular analysis of the N. ceranae infection status of IPL-LD 65Y cells, the infected cells were transferred into a 1.5 ml PCR-clean reaction tube (Eppendorf) and stored at -70°C for further analysis. For RNA extraction, cell suspensions were resuspended in 300 μl RLT buffer (QIAGEN, Hilden, Germany) containing 3 μl β-mercaptoethanol (Roth). The cell homogenate was transferred to a QIAshredder tube (QIAGEN) and centrifuged at 16,100g for 4 min. The supernatant was gently mixed with 1 volume of 70% ethanol, transferred to an RNeasy spin column (QIAGEN), and centrifuged for 1 min at 8000g. The column was washed with 600 μl of RW1-buffer and again centrifuged for 30 sec at 8000g. To remove residual DNA, 80 μl of RNase-free DNase (QIAGEN) was added directly to the column, followed by incubation for 2 h at room temperature. The column was washed with 600 μl RW1-buffer and subsequent RNA extraction was performed essentially according to the manufacturer's protocol.
One-step RT-PCR reactions were performed according to standard protocols (One-step-RT-PCR kit, QIAGEN) using a reaction volume of 25 μl, 20 ng RNA per reaction, and the following temperature scheme: 30 min at 50°C, 15 min at 95°C followed by 35 cycles with 1 min at 94°C, 1 min at 58°C (optimal for all primer sets), 1 min at 72°C, each, including a final elongation step for 10 min at 72°C. PCR reaction mixes were stored at 4°C until further analysis. Reaction products obtained with unlabeled primers were analyzed via agarose gel electrophoresis, stained with ethidium bromide, and visualized under UV-light. Reaction products obtained using biotin-and digoxigenin-labeled ptp2-primers were quantified via PCR-ELISA.
For quantification of N. ceranae in infected cells, a recently published protocol for the quantification of Leishmania parasites in host tissues [61] was modified. Streptavidin-coated 96well microplates (VWR) were washed four times with ELISA-washing buffer (0.5% Tween 20 in 1xPBS). 1 μl of the RT-PCR reaction mix obtained by using ptp2 primers labeled with biotin and digoxigenin (see above) was resuspended with 99 μl of 2% FCS in 1xPBS and transferred into the wells. After incubation for one hour, the wells were washed six times with ELISA-washing buffer and incubated for 45 min with 100 μl anti-digoxigenin-peroxidase fabfragments (Roche Diagnostics, Mannheim, Germany, 0.15 U/ml) in 2% FCS/PBS-buffer. The wells were washed eight times with ELISA-washing buffer and dried at room temperature. 100 μl of the peroxidase substrate 2,2'-Azino-di-[3-ethylbenzthiazoline sulfonate (6)] diammonium salt (ABTS; Roche Diagnostics) at a concentration of 0.3 mg/ml ABTS in 100 mM citric acid (Carl Roth, pH 4.35) with 0.1% H 2 O 2 (Carl Roth) were added to each well and cells were incubated in the dark to develop color reaction. The color reaction was stopped after 5 min by adding 50 μl of 500 mM oxalic acid (Carl Roth). Arbitrary units for gene expression of the N. ceranae polar tube protein were obtained by reading the absorbance values of the plates at 405 nm excitation wavelength with 620 nm reference wavelength using an ELISA-reader. Arbitrary units of gene expression were compared and statistically analyzed by ANOVA with Kruskal-Wallis nonparametric multiple comparison test with α<0.05 using the GraphPad Prism 6.01 software package. Statistical differences were classified: not significantly different: n.s., p ! 0.05; significantly different: Ã, 0.05 > p ! 0.01; ÃÃ, 0.01 > p ! 0.001.

Microscopic analysis of the infection status of IPL-LD 65Y cells
For microscopic analysis of the N. ceranae infection status of IPL-LD 65Y cells, the infected cells were centrifuged onto glass slides (VWR) with a cell spin (Tharmac, Waldsolms, Germany) at 60 g for 5 min. The cells were stained with Giemsa and microscopically analyzed essentially as already described [20].

Results and Discussion
Development of an in vitro screening assay for candidate agents against N. ceranae The availability of a cell culture model for Nosema spp. infections [20] opened the possibility of developing a cell culture-based screening assay for substances active against Nosema spp. or specifically against N. ceranae. For such an assay, quantification of N. ceranae infection is a prerequisite. To this end we adapted a recently published PCR-ELISA protocol, which had originally been developed for detection and quantification of Leishmania parasites in host tissues [61], to N. ceranae. The described PCR-ELISA relies on the use of one biotin-and one digoxigenin-labeled primer per PCR-product and on the detection of the PCR products by sandwich ELISA. The PCR products are bound to streptavidin coated microplates by virtue of their incorporated biotin and they are visualized by anti-digoxigenin antibodies binding to the incorporated digoxigenin. Primer molecules that are not part of the newly generated PCR products will not produce any background because they are not detected by this ELISA. A prerequisite for establishing an RT-PCR-ELISA for the detection and quantification of N. ceranae was the selection of appropriate primers detecting genes expressed during infection. Previous studies had shown that 72 hours after infection merogony is nearly completed and sporogony, i.e. the development of new spores marking the final phase of the vegetative cycle, is initiated [20]. Therefore, we first analyzed the expression pattern of several N. ceranae genes during infection of IPL-LD 65Y cells via RT-PCR over 72 hours (Fig. 1). Primers (Table 1) were designed on the basis of the published N. ceranae genome sequence data [62].   Table 1. shown to be ineffective against microsporidia because these parasites lack an essential binding site for this drug [63] although one historical study suggested a possible effect of paromomycin on N. apis infection in honey bees [64]. Fumagillin has been demonstrated to be active against N. apis and has been widely used to control nosemosis in bee colonies [48,[50][51][52]. Clioquinol is a chelating agent for copper, zink, and iron with cytotoxic activity [65,66] and, therefore, was used as control to demonstrate the robustness of the assay even for cases when test substances applied to the infected cells turn out to be cytotoxic.
First we tested the effect of the test substances on the viability of IPL-LD 65Y cells via MTT tests. DMSO (0.99%) serving as solvent for all substances had no cytotoxic effect on the cells ( Table 2). Fumagillin was cytotoxic when used at concentrations of 0.1 mg/ml and 0.05 mg/ml but showed no cytotoxicity at 0.01 mg/ml and 0.005 mg/ml (Table 2). Therefore, we used 0.01 mg/ml fumagillin in our in vitro-infection assay. Paromomycin had no cytotoxic activity on IPL-LD 65Y cells in the concentrations tested while the cytotoxicity of clioquinol for IPL-LD 65Y cells was evident at both 1 mg/ml and 0.1 mg/ml ( Table 2).
Analysis of ptp2 mRNA expression at 72 hours post-infection using the RT-PCR-ELISA protocol demonstrated expression of ptp2 in infected, untreated cells while no ptp2 expression (0 ± 0.1 arbitrary units) was detectable in mock infected control cells ( Fig. 2A). This difference was highly significant (p-value < 0.0001). The slight decrease in ptp2-expression in infected, DMSO-treated cells was not significant compared to infected, non-treated cells (p-value > 0.9999). Infected cells treated with fumagillin showed only residual ptp2 expression not significantly different (p-value = 0.9999) from mock infected control cells ( Fig. 2A) consistent with the approved efficacy of fumagillin in the treatment of N. ceranae infections [48]. Treatment of infected cells with paromomycin resulted in ptp2 mRNA steady state levels which did not differ significantly (p-value = 0.9999) from both infected, untreated cells and infected, DMSO-treated cells ( Fig. 2A) confirming that paromomycin indeed has no activity against N. ceranae. In infected cells treated with clioquinol, no ptp2 expression was detectable ( Fig. 2A) as a result of the cytotoxic activity of the substance which killed nearly 90% of the cells (Table 2). Microscopic analysis of Giemsa-stained cells for presence or absence of intracellular stages of N. ceranae visually confirmed the results obtained via the RT-PCR-ELISA protocol (Fig. 2B). Mock infected control cells as well as N. ceranae-infected, fumagillin treated cells did not contain any vegetative or spore stages of N. ceranae. In contrast, N. ceranae infected cells as well as infected cells treated with paromomycin were filled with blue-stained spore stages of N. ceranae as expected at 72 hours post infection [20]. In the few cells still viable after clioquinol In Vitro-Screening Assay for Agents Active against N. ceranae treatment no vegetative stages of N. ceranae were detectable consistent with the lack of detectable ptp2 mRNA expression.
These results indicated that N. ceranae infected IPL-LD 65Y cell cultures combined with an RT-PCR-ELISA protocol for the quantitative detection of ptp2 expression constitutes a well suited assay for the identification of novel drugs with in vitro activity against N. ceranae. The 96-well-format of the RT-PCR-ELISA will allow medium throughput screening of substances thus providing a state-of-the-art tool for anti-nosemosis drug discovery.
Identification of substances active against N. ceranae through the novel in vitro screening assay We next used our newly established screening assay and tested several antimicrobial agents for their activity against N. ceranae in infected cell cultures aiming at identifying novel agents with anti-nosemosis activity comparable to that of fumagillin. The substances to be tested were selected based on their known inhibitory effect on other microsporidia, fungi or intracellular parasites. We tested the synthetic antibiotics albendazole, ornidazole, tinidazole, and metronidazole because both in vitro and in vivo studies have demonstrated that benzimidazoles and nitroimidazoles and its derivatives have activity against microsporidia in vertebrates as well as in invertebrates [67][68][69][70][71]. Furthermore, we tested quinine, which was found to be effective against Encephalitozoon species in grasshoppers [72], and surfactin because it was shown to influence N. ceranae development when fed to infected honey bees [55]. For all agents, the effect on cell viability (MTT test) and the activity against N. ceranae infection (cell culture based assay) were evaluated. DMSO was used as solvent for all agents and, hence, infected cells treated with DMSO served as control and reference for all tested substances (Fig. 3A). DMSO in the tested final concentration of 0.99% was neither cytotoxic (Table 2) nor active against N. ceranae ( Figs. 2A and 3B) and thus did not influence the analysis of the test substances.
Albendazole was applied to uninfected cells with final concentrations of 0.1 mg/ml and 1.0 mg/ml. At the higher concentration, MTT tests revealed 28% dead cells indicating some cytotoxic activity at this concentration. No cytotoxicity was observed when 0.1 mg/ml albendazole were applied (Table 2). Despite a ten-fold difference in albendazole concentration, the ptp2 mRNA steady state levels were not significantly different (p-value > 0.9999) between the two treatments. When compared to infected, DMSO-treated cells, the reduction in ptp2 gene expression was not significant for both concentrations (p-value = 0.4818 for 0.1 mg/ml and pvalue > 0.9999 for 1 mg/ml) suggesting a still ongoing infection (Fig. 3A). Microscopic analysis of Giemsa-stained cells confirmed this result and revealed numerous infected cells filled with blue stained spore stages (data not shown). Therefore, it was not possible to cure infected cells with non-cytotoxic concentrations of albendazole ( Fig. 2A-B).
This lack of in vitro activity of albendazole against N. ceranae is in contrast to its activity against N. bombycis both in vitro and in vivo. In Spodoptera frugiperda cells experimentally infected with N. bombycis, a final concentration of as little as 0.0053 mg/ml albendazole already caused 57.7% reduction in the proportion of infected cells (40.7% reduced to 26.9%; [69]). In vivo assays performed with 2.5 mg/ml, 5.0 mg/ml, and 10 mg/ml albendazole fed to N. bombycis infected silkworm larvae (Bombyx mori) revealed 86%, 96%, and 100% disease reduction, respectively [70]. Hence, N. bombycis in cell culture is highly sensitive towards albendazole. However, 470-to 2000-fold higher concentrations were needed for effective disease reduction in vivo.
We next tested the nitroimidazoles ornidazole, tinidazole, and metronidazole which had also shown in vivo activity against N. bombycis in silkworm larvae and resulted in 80% to 90% disease reduction when orally applied at 2.5 mg/ml to 10 mg/ml, respectively [70]. In our in vitro cell culture assay, ornidazole was highly toxic for IPL-LD 65Y cells (89% dead cells) when applied at 10 mg/ml (Table 2). When used at 1 mg/ml, a non-cytotoxic concentration ( Table 2), no significant reduction in ptp2 mRNA steady state levels compared to infected cells (p-value > 0.9999; Fig. 3A) was observed. Accordingly, no obvious effect against N. ceranae was observed in Giemsa-stained cells (data not shown). Therefore, ornidazole like albendazole has no significant in vitro activity against N. ceranae.
Tinidazole and metronidazole were weakly cytotoxic at 2 mg/ml with 19% and 7% dead IPL-LD 65Y cells, respectively. All cells survived when 0.2 mg/ml were applied ( Table 2). Obviously, cytotoxicity of metronidazole is highly dependent on the cells used because cell viability of uninfected vertebrate cells (Madin-Darby Canine Kidney epithelial (MDCK) cells) exposed to 0.050 mg/ml metronidazole was already reduced to 52% [67]. Steady state levels of ptp2 mRNA were not significantly reduced in infected IPL-LD 65Y cells by 0.2 mg/ml tinidazole (p-value > 0.9999) or metronidazole (p-value > 0.9999) compared to DMSO-treated infected cells (Fig. 3A). However, when tinidazole and metronidazole were used at concentrations of 2 mg/ml, ptp2 mRNA expression was no longer detected (Fig. 3A). Steady state levels of ptp2 mRNA were significantly different between DMSO-treated, infected cells and metronidazoleor tinidazole-treated, infected cells (p-value = 0.007 or p-value = 0.0377, respectively) (Fig. 3A). In addition, there was no statistically significant difference between mock infected cells and metronidazole-or tinidazole-treated, infected cells (p-value > 0.9999 or p-value > 0.999, respectively) (data not shown). In accordance with these molecular data (Fig. 3A), no single infected cell could be observed by microscopic analysis of Giemsa-stained cells (Fig. 3B). These results indicated that at a concentration of 2 mg/ml both tinidazole and metronidazole can completely inhibit N. ceranae infection of IPL-LD 65Y cells and, therefore, were as effective as fumagillin in vitro. However, both substances will hardly have a future as anti-nosemosis drugs in honey bees. Nitroimidazoles are antibiotics and the use of antibiotics in honey bee colonies and in particular the use of nitroimidazoles in food animals and animal food is prohibited in many countries [73] including the European Union (Commission Regulation (EU) No 37/ 2010 of 22 December 2009).
Next, we evaluated the in vitro activity of quinine against N. ceranae. Tested at 2 mg/ml, quinine was highly cytotoxic for IPL-LD 65Y cells (88% dead cells, Table 2). Reducing the final quinine concentration to 0.2 mg/ml resulted in a markedly reduced cytotoxicity with only 14% dead cells (Table 2). However, there was no significant difference in ptp2 mRNA steady state levels between quinine-treated and DMSO-treated infected cells (p-value > 0.9999) indicating that quinine had no effect on N. ceranae infection in cell culture (Fig. 3A). This was not entirely in contrast to results obtained with of Encephalitozoon sp. in infected grasshoppers. While intra-hemocelic injection of quinine resulted in spore counts of about 0.11 x 10 6 spores/mg tissue/grasshoppers which were significantly different from control (0.4 x 10 6 /mg tissue/grasshopper; [72]), oral treatment with about 1.68 mg quinine/day had resulted in a non-significant 53% reduction in spores [74]. Hence, quinine seemed not to be generally active against Encephalitozoon sp. in grashoppers but only under certain conditions whereupon intra-hemocelic injection is rather artificial and the more workable oral application did not result in a significant disease reduction.
Another promising anti-nosemosis agent was the bacterial peptide secondary metabolite surfactin because of its reported effect on N. ceranae development in infected bees [55]. In Vitro-Screening Assay for Agents Active against N. ceranae However, surfactin is able to lyse cells and, hence, a drawback of surfactin is its non-specific cytotoxicity which has already been demonstrated in several vertebrate cell lines [75]. Concentrations of less than 0.025 mg/ml had no effect on vertebrate cell viability but when exposed to concentrations greater than 0.073 mg/ml no cells survived. Testing the cytotoxicity of surfactin for IPL-LD 65Y cells revealed that applying 0.2 mg/ml surfactin already resulted in 90% dead cells and even with 0.02 mg/ml still 35% of the cells died (Table 2). These results suggested that invertebrate cells are as sensitive or even more sensitive towards surfactin than vertebrate cells. When the lower surfactin concentration of 0.02 mg/ml was tested for its activity against N. ceranae in infected cells, we observed no significant reduction in ptp2 mRNA steady state levels in comparison to DMSO-treated infected cells (p-value > 0.9999) (Fig. 3A). This was substantiated by microscopic analysis of Giemsa-stained cells which revealed numerous infected cells filled with blue stained spore stages despite surfactin treatment (Fig. 3B).
These results were contradictory to the published effect of two different surfactin preparations (S1 and S2) produced by different strains of Bacillus subtilis on N. ceranae infection in honey bees [55]. It was reported that 5 mg/ml surfactin supplied ad libitum before and after N. ceranae infection resulted in an increased proportion of bees exhibiting low (0-1E+06) spore intensity and a decreased proportion of bees with high (>1E+07) spore intensity when compared to untreated bees. However, the effect was significant only for the S2-surfactin preparation. In addition, this S2-surfactin preparation also had a significant effect on spore viability because incubation of spores for 40 hours with 10 mg/ml S2-surfactin resulted in an increased proportion of bees exhibiting medium (1E+06-1E+07) intensity of spores in the midgut when compared to untreated bees (23.2% to 58.8%) while at the same time the percentage of bees with high (>1E+07) spore intensity decreased from 63.2% to 23.5%. However, the proportion of bees with low (0-1E+06) spore intensity did not change upon treatment. These results are remarkable in that they suggest that there is no general effect of surfactin on N. ceranae infections in bees but the effect is dependent on the surfactin preparation used. This might be due to the fact that the cyclic lipopeptide surfactin produced by B. subtilis has been shown to be a mixture of isoforms as a result of amino acid substitutions in the peptide ring and variations in the chain length and branching of its hydroxy fatty acid component [76,77]. These isoforms are also reported to have slightly different properties [77][78][79][80] which might explain the differences in efficacy between different surfactin preparations observed in infected bees [55]. Hence, it is conceivable that the commercially obtained surfactin used in our cell culture assay is a less active isoform like the S1-surfactin preparation used by Porrini and co-workers [55]. A more thorough analysis of surfactin activity against N. ceranae including the identification of the structural characteristics correlating with anti-nosemosis activity and the assessment of cytotoxicity of orally administered surfactin in honey bees is necessary to finally decide on the applicability of surfactin as anti-nosemosis drug in honey bees.

Conclusion
We here present a screening assay for the identification of substances exhibiting activity against N. ceranae, which is based on N. ceranae infected cultured cells and coupled with an RT-PCR-ELISA protocol for determining N. ceranae infection intensity. This novel screening assay can also be performed with N. apis infected IPL-LD 65Y-cell cultures [20] and, hence, provides a so far unavailable screening platform to test many substances at a time for their potential antinosemosis activity. Laborious and time-restricted animal experiments and in vivo-tests will be reduced to a minimum because only promising candidates will be tested in bees and in the field. With the help of this cell culture based medium throughput screening system, the analysis or development of new anti-nosemosis substances will be accelerated and simplified in the future.