Virulence of Melissococcus plutonius and secondary invaders associated with European foulbrood disease of the honey bee

Abstract European foulbrood is a globally distributed brood disease affecting honey bees. It may lead to lethal infections of larvae and, in severe cases, even to colony collapse. Lately, a profound genetic and phenotypic diversity was documented for the causative agent Melissococcus plutonius. However, experimental work on the impact of diverse M. plutonius strains on hosts with different genetic background is completely lacking and the role of secondary invaders is poorly understood. Here, we address these issues and elucidate the impact and interaction of both host and pathogen on one another. Moreover, we try to unravel the role of secondary bacterial invasions in foulbrood‐diseased larvae. We employed in vitro infections with honey bee larvae from queens with different genetic background and three different M. plutonius strains. Larvae infection experiments showed host‐dependent survival dynamics although M. plutonius strain 49.3 consistently had the highest virulence. This pattern was also reflected in significantly reduced weights of 49.3 strain‐infected larvae compared to the other treatments. No difference was found in groups additionally inoculated with a secondary invader (Enterococcus faecalis or Paenibacillus alvei) neither in terms of larval survival nor weight. These results suggest that host background contributes markedly to the course of the disease but virulence is mainly dependent on pathogen genotype. Secondary invaders following a M. plutonius infection do not increase disease lethality and therefore may just be a colonization of weakened and immunodeficient, or dead larvae.

While early studies assumed low genetic diversity in M. plutonius, Allen and Ball (1993) (Budge et al., 2014;Haynes, Helgason, Young, Thwaites, & Budge, 2013;Takamatsu et al., 2014). Sequence types are determined by means of multilocus sequence typing (MLST) and can be detected on both very narrow local scales and widespread across countries (Budge et al., 2014;Haynes et al., 2013;Takamatsu et al., 2014). The different clonal complexes divide M. plutonius strains into two subtypes with CC3 and CC13 (including the type strain ST1) containing the typical strains, and CC12 containing the atypical strains (Budge et al., 2014;Haynes et al., 2013;Takamatsu et al., 2014). Typical and atypical (M. plutonius-like) strains can both be isolated from diseased larvae with clinical signs of EFB. However, they differ in their cultural and biochemical characteristics, and their ability to cause disease after laboratory in vitro cultivation (Arai et al., 2012).
Even though all described M. plutonius strains (belonging to different clonal complexes) were isolated from naturally diseased honey bee larvae, they seem to differ in virulence to their host.
From field pathology data, without experimental testing, CC3 and CC12 are more virulent than CC13, (Budge et al., 2014). Under improved inoculation conditions (potassium-rich medium/diet) and in vitro larvae infection assays, Nakamura and coworkers proved that CC13 is the least virulent (possibly even avirulent) clonal complex followed by CC3, with CC12 being extremely virulent (Nakamura et al., 2016). Strains of CC13 may lose their infectivity/virulence due to in vitro subcultivation which is a well-known phenomenon for decades (Bailey, 1956;Bailey & Locher, 1968;Nakamura et al., 2016).
Though this loss might be rescued following several in vivo passages in honey bee worker larvae (Bailey, 1963).
Here, we investigate the virulence of several M. plutonius isolates (belonging to CC3 and CC13 complex) on different host genetic backgrounds using standard in vitro larvae rearing. Furthermore, we tested the most virulent M. plutonius strain, in combination with two secondary invaders (P. alvei and E. faecalis), for their putative additive pathogenic effects.

| Ethics statement
Endangered or protected species were not used in this study.
Experiments and observations conform with the laws of Germany in relation to animal protection. No specific ethics certification is required for this research.

| Larvae grafting and in vitro rearing
Honey bee larvae (Apis mellifera) were grafted, reared in vitro and received worker larval diet as described in Crailsheim et al. (2013).
To test for variance in M. plutonius CC-type virulence and colony effects, hatched larvae were grafted from two different colonies (Aqueen of Czech-German origin, B-queen of French-German origin), following queen caging for 24 hr, and assigned randomly to one of the four following treatment groups per colony. Treatment group 1: Worker bee larvae were grafted on a 5 µl day 1 diet (Crailsheim et al., 2013) and subsequently fed with 5 μl of a diet-inoculum mix (19:1) containing one of the respective bacterial strain cultures or sterile medium (controls). The absorbance of the bacterial solutions was adjusted to 0.3 (OD 600 nm ) beforehand. From day 2 postgrafting till day 6 postgrafting, bee larvae received a standard diet following Crailsheim et al. (2013).

| Larval survival, weight, and relative M. plutonius infection intensity
Larval survival was monitored on a daily basis until pupation of all larvae of the noninfected control groups. Pupation means that each larva had a clearly recognizable head-thorax-abdomen structure and six legs. Lastly, every individual that did not fully pupate the same day as the control bees was counted as dead. Larvae of all groups that died on day 1 postgrafting were excluded from the survival analysis due to potential grafting errors (Crailsheim et al., 2013). Larval weight was measured on day seven for all larvae of all groups, before placing them into new plates (covered with cellulose) for pupation (Crailsheim et al., 2013).
Larval infection intensity was measured from 8 to 12 white larvae per treatment group (minimum 4 larvae per group and for at least 2 grafting events), randomly selected on day 10 ± 1 postgrafting. White larvae which did not defecate were selected to get a consistent picture of M. plutonius replication by excluding extremes (e.g., brown, gray, and black rotten larvae). DNA isolation was performed following standard phenol-chloroform-isoamyl alcohol (25:24:1) protocol. Initially larvae were homogenized in 400 μl extraction buffer pH 8.0 (100 mmol/L Tris/HCl, 100 mmol/L NaCl, 10 mmol/L EDTA, 1 mmol/L SDS) with 1 μl Tween20, and incubated for 5 min on ice. At the final step of the extraction procedure the dried DNA pellet was resuspended in 20-30 μl elution buffer (10 mmol/L Tris pH 7.4, 1 mmol/L EDTA pH 8.0). DNA quality and quantity were determined on a NanoDrop 1000 (Thermo Fisher Scientific). Samples with more than 1 μg/μl were diluted 1:50 and samples below 1 μg/ μl 1:20 with DEPC water prior to M. plutonius quantification (qPCR) using the CFX Connect Real-Time PCR Detection System (Bio-Rad), SensiMix-SYBR No-ROX Kit (Bioline), and M. plutonius-specific primers (EFB-primer; Budge et al., 2010). Relative M. plutonius bacterial load per larvae (Forsgren et al., 2013) was estimated by normalizing bacterial load to the amount of honey bee mitochondrial DNA (COI-primer) as recommend by Behrens, Forsgren, Fries, and Moritz (2010). The same qPCR protocol was used for each 10 μl reaction (including 1 μl diluted DNA) and one of the two primer pairs (forward and reverse, each with 0.3 μmol/L): initial denaturation at 95°C for 10 min, 40 cycles of 95°C for 15 s, 60°C for 30 s and 72°C for 30 s.
The melting curve of the amplicons was measured from 50°C to 98°C, every 5 s with 1°C increment. Each sample was analyzed in duplicates and repeated if between replicate difference (delta C q ) was higher than 0.5. PCR efficiency for each primer pair was estimated using DNA dilutions: PCR eff (EFB-primer) = 1.93, PCR eff (COIprimer) = 1.89. Finally, data were collected for the two colonies, four treatment groups and for each combination of larvae from two to three grafting events. Details on exact sample sizes are summarized in Table S1.

| Statistics
All statistical analyses were performed using R (ver. 3.3.2). Larval survival was evaluated by means of Kaplan-Meier survival analysis with subsequent log-rank tests, for colony and treatment effects.
Data for larval weight and relative M. plutonius infection intensity were tested for deviation from a normal distribution by Kolmogorov-Smirnov tests. As a normal distribution could not be confirmed for both, nonparametric tests were used to test for significant differences between colonies (Mann-Whitney U test) and treatment groups (Kruskal-Wallis ANOVA). P-values were adjusted for multiple testing using Bonferroni correction.

| Larvae grafting and in vitro rearing
In a second set up we tested the impact of secondary invaders on larval survival and weight. Grafting, rearing, and infection of larvae were performed as described above using larvae from a strong colony (colony C-queen of German origin) of the university stock.
Experiments were run in four replicates. The first replicate con-

| Bacteria cultivation and infection
M. plutonius strain 49.3 was cultivated and applied as described for the first experiment. On day 3 postinfection with M. plutonius strain 49.3, the larvae received 30 μl of standard larval diet (Crailsheim et al., 2013) mixed with vegetative cells of a secondary invader (E. faecalis or P. alvei, approx. 6 × 10 4 CFUs per larvae for each bacteria species) or sterile medium (control-just M. plutonius treatment). Enterococcus faecalis (LMG 7937) and Paenibacillus alvei (LMG 13253) were provided by BCCM/LMG Bacteria Collection (Ghent University, Ghent, Belgium) and grown in specific medium as described elsewhere (Erler et al., 2014). The absorbance of both bacterial cultures was adjusted to 0.3 (OD 600 nm ) and bacteria were fed to larvae as described above (approx. 6 × 10 4 CFUs per larvae).

| Larval survival and weight
Larval survival was determined as described above but the weight was measured on day eight to account for the growth of the secondary invaders.

| Statistics
Survival analysis was performed with a nested Cox regression mixed-effects model (coxme package in R) fitted by maximum likelihood with treatment as fixed effect and replicate as random effect (treatment groups nested in replicates) to account for the variance in the replicates, as replicates 1 and 3 differed significantly from 2 and 4 but not from each other respectively.
Larval weights were analyzed as described previously.

| RE SULTS
Before examining variable virulence of several natural M. plutonius isolates, we examined the in vitro virulence of a commercially available M. plutonius strain (LMG 20360, ST1, CC13, the only one commercially available) in larvae from colonies of different origin (colony A and B). However, neither a significant impact on larval weight on day 7 (MWU test: W = 2447.5, p = .14, Table S3) nor on larval mortality (Kaplan-Meier survival analysis with log-rank test, p = .32, Figure S1) was observed in M. plutonius exposed larvae, but a significant weight difference between larvae of different origin (MWU test: W = 2428.5, p < .0001, Table S3) with larvae of colony B being lighter compared to colony A. For colony A, the difference in mortality was highly significant between strains 119 and 49.3 (log-rank test, p < .001) but not for colony B (log-rank test, p = .017, Bonferroni adjusted significance level: p < .016) (Figure 1). This borderline nonsignificant difference for colony B, following p-value adjustment, indicates potential F I G U R E 1 Larval survival over 12 days (starting from day 1-grafting) for control groups and larvae infected with M. plutonius (strain 49.3, 119 and 4-127) from two unrelated colonies (a) colony A, (b) colony B. Different superscript letters in the legend for treatment groups show significant differences in larval survival, following Kaplan-Meier survival analyses with Bonferroni adjusted posthoc log-rank tests (p < .016) variable virulence for strain 119 depending on host genetic background and/or general health conditions of the tested colonies.

| Impact of co-infection
Although, all infection treatment groups (irrespective of a secondary bacterium combination or solely M. plutonius) had significantly TA B L E 1 Larval weight (mean ± SD) on day 7 (day 6 postinfection) for control groups (uninfected) Table 2) with approx. 15% mean weight reduction in infected individuals (Table 2). Consistently, no weight differences were detected between infected groups (posthoc multiple comparisons, Bonferroni adjusted: M-ME, M-MP, MP-ME p > .05, Table 2).
The CFUs per larva used for infection varied between replicates, especially in E. faecalis treatments (Table S2), although it did not correlate with larval mortality or weight, neither for M. plutonius infected larvae nor larvae treated with an additional infection.

| D ISCUSS I ON
In this study, we observed variation in virulence for four different M.
plutonius strains and a divergent response of hosts depending on genetic background. Here, the determinative host trait for the resulting variation of pathogen impact seems to be the developmental differences (specifically body size) of the honey bee larvae. Host body size and developmental speed may be crucial factors in host-pathogen interactions where the pathogen may be obliged to multiply as fast as possible to overcome clearance by the host (Cable, Enquist, & Moses, 2007). This might particularly be critical for a pathogen infecting a host via the oral-fecal route. In the case of the honey bee larvae, clearance refers to defecation before pupation. The observed weight differences of infected larvae of different origin are negatively correlated with mortality (Pearson correlation: n = 8, r = −.97, p < .0001). At larval stage, the mortality of the lighter phenotype (colony B) was lower compared to the heavier one (colony A) (Figure 1). That lower weight appears to be advantageous, is presumably due to reduced multiplication rates of the pathogen.
However, in the pupal stage, the apparently slower growth rate of larvae from colony B phenotype led to a reduced nutrient supply during metamorphosis and ultimately resulted in a higher mortality of pupae. Malnutrition caused by M. plutonius infections, leading to reduced larval and pupal weight, was already discussed several times (Bailey, 1959(Bailey, , 1960Nakamura et al., 2016). However, at this point no inferences could be made on the cause(s) leading to malnutrition, either due to host-pathogen competition for nutrients or due to less nutrients feed to the larvae in the colony or both.
Mortality of the different honey bee strains used for larval in- picture. In the current study, we exclusively tested typical strains.
From previous studies, we know that both M. plutonius strain types can be detected by duplex PCR and cultivation in the same larvae samples  and several strains (sequence types) might participate in single EFB outbreaks of the same apiary, which, at least in Japan, might be common .
The results of Giersch et al. (2010) (Dickel, Münch, Amdam, Mappes, & Freitak, 2018). This may confirm that we did not find any additive mortality by coinfecting larvae, as previously mentioned by Bailey (1963 For artificial infections, the quality of the larval food (batch of royal jelly) seems to be of high importance for infection success (Giersch et al., 2010). M. plutonius cell numbers can be significantly reduced by a highly antimicrobial royal jelly water extract, based on major royal jelly protein 1, fatty acids and other substances (Vezeteu, Bobiş, Moritz, & Buttstedt, 2017; and references therein).
We did not examine the resistance of M. plutonius strains to larval food used in this study. However, a recent study showed that clonal complex affiliation and number of days precultured are major factors explaining resistance for in vitro growth in medium containing royal jelly or 10-hydroxy-2-decenoic-acid (10-HDA) (Takamatsu, Osawa, Nakamura, Yoshiyama, & Okura, 2017).
Another point of M. plutonius pathogenicity, not analyzed in the current study, is the delay of symptoms under non-natural conditions. Artificially infected colonies show a disease symptom delay of up to 12 days under natural conditions (Bailey, 1957b). Laboratory artificial infections of bee larvae also lead to a delay in defecation and subsequent, if any, pupation (McKee et al., 2004). The same study described that some larvae even survived the infection and reached pupation. This might be the result of M. plutonius variability of multiplication speed and virulence or host susceptibility.

| CON CLUS ION
The environment of the colony has to be mentioned as an important factor which profoundly affects the course of the disease. However, the details are nearly unknown (Bailey & Locher, 1968;White, 1920).
Former studies on M. plutonius epidemiology showed that visible EFB-symptoms can disappear spontaneously from infected colonies after an epidemic peak (about June in the northern hemisphere) which appears to be unrelated to changes in the susceptibility of the larvae as they still can be infected in vitro (Bailey, 1959(Bailey, , 1960Burnside, 1938;White, 1920).
To get a deeper understanding of the pathogenicity of M. plutonius and the nature of EFB it is imperative to study the natural epidemiology of the disease in combination with in vitro and in vivo assays investigating the growth and virulence of the diverse M. plutonius types (typical vs. atypical), clonal complexes and genotypes (sequence types). From the current and all previous studies, the whole infection process seems to be a complex interaction of the honey bees' and M. plutonius genotype, in combination with several secondary invaders.

ACK N OWLED G M ENTS
Financial support was granted by the German Research

CO N FLI C T O F I NTE R E S T
The authors declared no potential conflict of interest.