Putative determinants of virulence in Melissococcus plutonius, the bacterial agent causing European foulbrood in honey bees

ABSTRACT Melissococcus plutonius is a bacterial pathogen that causes epidemic outbreaks of European foulbrood (EFB) in honey bee populations. The pathogenicity of a bacterium depends on its virulence, and understanding the mechanisms influencing virulence may allow for improved disease control and containment. Using a standardized in vitro assay, we demonstrate that virulence varies greatly among sixteen M. plutonius isolates from five European countries. Additionally, we explore the causes of this variation. In this study, virulence was independent of the multilocus sequence type of the tested pathogen, and was not affected by experimental co-infection with Paenibacillus alvei, a bacterium often associated with EFB outbreaks. Virulence in vitro was correlated with the growth dynamics of M. plutonius isolates in artificial medium, and with the presence of a plasmid carrying a gene coding for the putative toxin melissotoxin A. Our results suggest that some M. plutonius strains showed an increased virulence due to the acquisition of a toxin-carrying mobile genetic element. We discuss whether strains with increased virulence play a role in recent EFB outbreaks.


Introduction
The Western honey bee (Apis mellifera L.) is an insect of major worldwide ecological and economic importance. Honey bees produce honey and wax and pollinate many economically important crops [1]. The recent loss of managed honey bee colonies in several regions of the world threatens the ecological services provided by this pollinator [2][3][4][5]. Consequently, honey bee health has become a major concern not only for scientists, but also for the public and policy-makers [6][7][8]. Current research indicates that pathogens are a major cause of colony losses. Honey bee pathogens include viruses [9], protozoa [10], fungi [11], parasitic mites [12] and bacteria [7,9,13,14]. One of the most detrimental bacterial diseases affecting honey bees is European foulbrood (EFB). EFB is reported worldwide [15] and has emerged as an infectious disease since the mid-eighties in the United Kingdom, since the year 2000 in Switzerland and since 2010 in Norway [13,14,16,17]. High numbers of clinical cases have also been reported from Finland, France, Greece, Holland, Czechia and Italy [18][19][20], making EFB an economically important veterinary disease [18,21].
The pathogenic agent of EFB, Melissococcus plutonius (Lactobacillales, Enterococcaceae) [22], enters the intestinal tract of honey bee larvae through contaminated food provided by adult bees [23]. Once ingested, M. plutonius rapidly multiplies in the mid-gut lumen, possibly depriving the host of nutrients [24]. Diseased larvae typically change from white to a yellowish color, become flaccid and die 4-5 days after infection [23,25]. The massive loss of brood resulting from severe infection weakens the colony and can lead to its collapse [25].
To date, only broad-spectrum antibiotics such as oxytetracycline are available to treat EFB-affected colonies. Due to the risk of antibiotic resistance development [26,27] and an accumulation of residue in honey, the use of antibiotics is not a sustainable method to control EFB and is banned in some countries. In the absence of efficient treatment and given the severity of EFB outbreaks, 79 countries worldwide have classified EFB as a notifiable disease (World Animal Health Information Database, OIE [28]). In 22 of these countries, veterinary authorities destroy symptomatic colonies and monitor neighboring apiaries to avoid the further spread of the pathogen, which is costly and time-consuming. This situation calls for new control strategies, but their development is constrained by the limited knowledge of the pathogenesis of EFB [23].
Virulence is central to pathogenesis. A better knowledge of the mechanisms that determine this trait could contribute to the design of improved control methods. Indeed, virulence factors are promising targets for specific drugs or management measures [29]. Such measures include curing honey bee colonies hosting less virulent pathogens and restricting the use of destructive control methods to treat infections caused by more virulent pathogens [30]. An assessment of the extent and causes of virulence variation in strains of M. plutonius may thus prove useful in controlling EFB more efficient and more sustainable to improve honey bee health.
Early reports [31,32] suggested that M. plutonius isolates were genotypically and phenotypically homogeneous. As a result, variation in virulence was not expected. The discovery of genetic differences between isolates [33][34][35][36][37][38][39][40] has challenged this view. Differences in virulence have indeed been documented within M. plutonius, with strains defined as atypical by Arai et al.,2012 [34], killing a higher proportion of hosts in a shorter time period than typical strains. Differences in virulence also occur between typical strains, depending on experimental conditions [41,42]. However, no virulence factor has yet been clearly identified.
Several factors are likely to influence the virulence of pathogens. One is the impact of secondary agents [43]. In the case of EFB, saprophytic species such as Paenibacillus alvei, Enterococcus faecalis, Brevibacillus laterosporus or Achromobacter eurydice cause secondary bacterial infections that might increase damage to larvae [23,44]. However, their influence in EFB is debated [42,[45][46][47]. Another potential cause of variation in virulence is the growth dynamics of the bacterial pathogen. Bacterial strains that multiply rapidly and reach high densities can cause more damage to the host [48,49]. Other major factors influencing virulence include the production of biologically active compounds, such as adhesins improving attachment to host cells, enzymes degrading host tissues, or toxins disturbing the physiological processes of the host [50][51][52][53][54][55].
Experiments aimed at quantifying the virulence of M. plutonius and identifying factors that influence virulence have been hampered by the complex social environment of the honey bee colony that affects the spread and growth of the bacteria [56,57], and by the legal obligation to destroy colonies showing disease symptoms [14]. These constraints can be overcome by experimentally infecting honey bee larvae reared in vitro. In the past, such assays were considered impossible due to M. plutonius loss of infectivity during bacterial sub-cultivation [46,56,58]. More recently, several studies using the in vitro larval rearing method [59] have circumvented this problem [34,41,42,47,[60][61][62], making standardized quantification of virulence possible without any sanitary risk to the colonies in the field.
In this study, we reared honey bee larvae in vitro and infected them with M. plutonius to investigate the causes and extent of variation in virulence. We screened 17 M. plutonius isolates collected from five European countries for genetic differences and measured their virulence over the entire developmental period of worker brood. We examined whether virulence was associated with M. plutonius multilocus sequence type (MLST), coinfection with P. alvei, differences in growth dynamics in a culture medium and presence of a putative toxin-coding gene [39]. We discuss whether M. plutonius strains with high virulence play a role in recent EFB outbreaks and whether they could be targets for more sustainable control measures of the disease.

Virulence of M. plutonius strains
The M. plutonius isolates collected in various regions of Europe belonged to six multilocus sequence types and two clonal complexes (CC 3 and CC 13, Table 1) of the typical form. After excluding cases with excessive control mortality, we obtained enough replicates to quantify the virulence of 16 out of the 17 isolates tested. The virulence of these isolates in terms of honey bee brood mortality in vitro varied greatly ( Figure 1). Three Swiss isolates (CH 21.1, CH 49.3 and CH 60) were highly virulent, causing over 80% mortality in infected brood up to the imago stage ( Table 1, Table. S2, Figure. S3). These isolates caused significant mortality compared to the controls (Figure 1; pairwise log-rank tests, Bonferroni-Holm corrected, p < 0.001, Table. S1, Figure. S3 15 to 55%, which was significantly higher than that of the controls (Figure 1; pairwise log-rank tests, Bonferroni-Holm corrected, p < 0.001, Table. S1 and S2, Figure.  3) was not significantly higher than that of the non-infected control brood. These isolates were therefore categorized as avirulent ( Figure 1; pairwise log-rank tests, Bonferroni-Holm corrected, n.s., Table. S1 and S2, Figure. Table. S2 and S4; pairwise log-rank tests, Bonferroni-Holm corrected p = 0.004; Mean Henderson-Tilton corrected mortality 23.1%). However, no significant increase in mortality was observed after co-infection with M. plutonius and P. alvei compared to single infections with one bacterial species alone (Figure 2 Tab. S4, pairwise log-rank tests, Bonferroni-Holm corrected p > 0.01).

Growth dynamics and virulence of M. plutonius
The M. plutonius isolates varied in their in vitro growth dynamics ( Figure 3). Some bacterial isolates multiplied rapidly to high cell densities, while others multiplied slowly and reached low final cell densities in liquid basal medium at the end of the observation period.
Other strains showed mixed features, multiplying rapidly to low final cell densities and inversely. Four isolates of low to intermediate virulence (FR 27.1, CH 45.1, NO 764-5B and NO 765-6B) and one avirulent indicates survival of non-infected controls. N = number of larvae tested. Dashed lines represent 95% confidence intervals. Significant differences in the survival of brood due to treatments (uncorrected for control mortality) are indicated by different letters (pairwise log-rank tests, Bonferroni-Holm corrected, p < 0.001).

Figure 2.
Survival of in vitro reared honey bees exposed to M. plutonius isolate CH 90, orange curve) or to P. alvei reference strain DSM29 (purple curve) only, or after co-infection with M. plutonius CH 90 and P. alvei DSM29 (green curve). The black curve indicates the survival of non-infected controls. Dashed lines represent 95% confidence intervals and N designates the number of larvae tested. Significant differences in the survival of brood due to treatments are indicated by different letters (pairwise log-rank tests, Bonferroni-Holm corrected, p < 0.01).  Table. S5). M. plutonius isolates causing over 40% mortality (corrected through a Henderson-Tilton calculation) in the in vitro larval infection assay, also reached high final densities in basal medium with OD 600 values at 0.7 or above, and five out of the seven most virulent isolates reached OD 600 values above 0.8 (Tab. S5).  the generation time of the tested M. plutonius isolates during their exponential growth phase in artificial medium, which ranged from 3 to 6 h ( Table 1). In contrast, final cell density (at 84 h) was positively and significantly correlated with mortality (N = 15, Kendall's τ = 0.641, p < 0.005, Tab. S7).

Discussion
Virulence varies among M. plutonius isolates Mortality of the honey bee brood caused by the M. plutonius isolates tested varied greatly. The variation in virulence among the 16 isolates was independent of their multi-locus sequence type and clonal complex affiliation. Variation in virulence among strains of the so called "atypical" M. plutonius belonging to different clonal complexes has recently been documented [41].
Our results indicate that variation in virulence also  [39]. The post-infection mortality rate is expressed as a Henderson-Tilton corrected percentage. Avirulent isolates did not cause significant mortality compared to non-infected controls (Tab. S1) and Henderson-Tilton corrected mortality rates are inferior to 10%. Low to intermediate virulence degrees correspond to mortality rates in the range of 15-55%. High virulent isolates cause mortality rates above 80% (see Fig. S3). occurs in clonal complexes belonging to the "typical" strains of M. plutonius as observed in [42]. Mortality rates also varied among replicates of in vitro infection tests using the same M. plutonius isolate, but larvae of different colonies. This is in line with the results from two recent studies [41,42] and may be due to variation in host susceptibility. Although all but one isolate (CH MeplS1) tested in this study came from honey bee colonies with explicit symptoms of an acute EFB infection, four isolates (UK 36.1, UK 31.1, CH 82 and IT 1.3) did not cause elevated larval mortality in vitro, and were therefore ranked as avirulent. It is possible that this discrepancy is a methodological artifact due to the loss of the plasmid carrying the putative virulence factor during cultivation on artificial media. Although the tested isolates were subjected to the same laboratory procedures, random loss of the plasmid cannot be excluded. Alternatively, negative effects on the development or physiology of the larvae not detected in our assay could have triggered increased removal by adult hygienic workers, generating the typical spotty brood pattern in EFB-diseased colonies. It is also possible that bacterial isolates less virulent at individual level are highly virulent at colony level, and vice versa, as was hypothesized for Paenibacillus larvae [63]. Larvae infected with low virulence isolates might be removed slower than larvae affected by high virulence isolates. A low removal rate could facilitate the production and spread of the pathogenic bacteria within the colony. It is highly likely that the in vivo virulence of M. plutonius at colony level is modulated by the effect of the social immunity of the colony [64], as proposed for P. larvae [63,65] and be different from virulence quantified in vitro.

No evidence that co-infection with P. alvei increases mortality
Our results indicate that the presence of the secondary agent P. alvei is not required to induce brood mortality. We observed high larval mortality after single infection with some of the M. plutonius isolates, in line with former studies [34,41,42,62]. To our knowledge, this is the first report on single infections with P. alvei in an in vitro infection assay. In this study, single infection with P. alvei strain DSM29 caused significant mortality to the honey bee brood. In contrast, co-infection with M. plutonius did not increase brood mortality beyond the effect of single infections with the tested M. plutonius isolate (CH 90). These results are similar to those of Lewkowski and Erler, 2018 [42], but differ from those of Giersch et al., 2010 [47], probably because of methodological differences (e.g., different infection time, higher concentration of P. alvei used in their study). To better understand the role of secondary agents, other bacteria associated with EFB should be tested singly and in co-infections with a larger set of M. plutonius isolates in vitro [42] as well as in vivo in honey bee colonies.

Relationship between M. plutonius growth dynamics and virulence
The growth dynamics of M. plutonius isolates in the culture medium were highly variable. The bacterial generation time in the artificial medium did not predict mortality in the honey bee brood. The final bacterial density may play a more important role in determining pathogenicity (sensu [66]) than the generation time. All isolates defined as being of intermediate or high virulence reached high final densities in the culture medium, and isolates with lower final densities caused low mortality. This pattern obtained in vitro suggests that strains must reach a damage threshold to induce high mortality. Although the growth dynamics of M. plutonius isolates in culture medium as observed in this study may be different from that in honey bee larvae, the hypothesis that a high bacterial load is a prerequisite for high virulence is supported by the occurrence of a positive correlation between bacterial numbers and virulence in vivo [41,62]. However, some low virulence isolates reached high final densities in the culture medium at 84 h, suggesting that factors other than the number of M. plutonius bacteria are involved in causing high mortality in honey bee brood.

Role of melissotoxin A in virulence
The melissotoxin A gene was restricted to three highly virulent Swiss isolates (CH 21.1, CH 49.3 and CH 60) and the presence of this gene was a significant predictor of honey bee brood mortality (Tab. S5). In our experiment, the bacterial isolate CH MeplS1 was avirulent. This isolate originated from a bacterial culture of the highly virulent isolate CH 49.3 and lost plasmid pMP19 (19.4 kbp; GenBank: JSBA01000009.1) after repeated sub-cultivation [39]. Plasmid pMP19 encodes melissotoxin A (GenBank: KMT29105) and another putative virulence factor, the extracellular matrix-binding protein [39]. While the additional loss of genome-encoded virulence factors in CH Mepl S1 cannot be excluded, the simultaneous loss of pMP19 and virulence suggests a causal link.
A further piece of evidence for the possible implication of the melissotoxin A gene in virulence is that it is expressed by M. plutonius during infection in vivo [39,62]. This gene shares a high sequence similarity with the epsilon toxins of the ETX/MTX2 family (pfam03318 of Clostridium perfringens, 33% amino acid sequence identity [39] and of Bacillus pumilus with 48% sequence identity, 90% coverage and an Expect (E)-value of 7e -73 , NCBI query; https://www. ncbi.nlm.nih.gov/). Epsilon toxins change the cell permeability for ions by forming large membrane pores, causing cell death and edema in animal models [67]. It was recently suggested that a protein of the ETX/MT2 family is an important virulence factor in two subtypes of P. larvae, the causative pathogen of American foulbrood [68]. It is thus plausible that melissotoxin A increases the virulence of M. plutonius by corrupting the cells of the larva's digestive tract.
The melissotoxin A gene is situated on plasmid pMP19 [39]. Plasmids can easily be exchanged between individual bacterial cells within and between species [69]. The plasmid pMP19 found in certain M. plutonius strains, including atypical isolates from Japan [40,62], may originate from other bacterial invaders co-existing with M. plutonius in the intestinal tract of honey bee larvae. The fact that the toxin gene is located on a mobile genetic element can explain the absence of association between brood mortality and the sequence type or clonal complex of the isolates used in this study.
The conventional explanation for the negative effect of M. plutonius is a competition for nutrients in the gut [25]. Although not mutually exclusive, our finding that a toxin-carrying mobile genetic element could confer high pathogenicity to certain M. plutonius isolates suggests a direct detrimental effect on gut cells. This mode of action is in line with the findings of McKee et al., 2004 [46] and with evidence that M. plutonius still caused high larval mortality in in vitro experiments where food was given in excess [34,41,42,46,47,[60][61][62].
Further experiments will be needed to demonstrate the direct role of melissotoxin A in the pathogenicity of M. plutonius. Such experiments include producing genetically modified strains of M. plutonius, for example by removing plasmid pMP19 from highly virulent strains, knocking-out melissotoxin A gene on pMP19 (pMP19Δmelissotoxin A gene), transforming wildtype pMP19 and pMP19Δmelissotoxin A gene into low virulent isolates of M. plutonius, as well as producing the putative toxin in the laboratory. The impact of mutant strains or the synthesized toxin on honey bee larvae could then be directly tested in vitro. Curing pMP19 from Japanese M. plutonius isolates belonging to CC 3 resulted in a loss of virulence [62], supporting our hypothesis for the role of the melissotoxin A gene. However, transformation experiments aimed at introducing the plasmid back in the cured isolates did not result in recuperation of virulence [62]. A reduced growth of these isolates due to the transformation protocol was provided as an explanation [62], in line with our suggestion that rapid multiplication is also required to produce highly virulent phenotypes. The role of the putative virulence gene on pMP19 could not be confirmed in a Japanese isolate of CC 13 [62]. A reduced growth of this isolate could also explain its low virulence despite the gene's presence.
Is the recent emergence of EFB due to highly virulent strains?
The question then arises as to whether the recent emergence of EFB is linked to the occurrence of bacterial strains of particularly high virulence at the colony level. Virulence trade-off models predict that the virulence of a pathogen is inversely correlated to its transmission, because excessive virulence causes early host death, hindering the spread of the pathogen amongst the population [70][71][72][73]. However, human-induced changes in the rate of horizontal transmission may have altered the adaptive compromise between virulence and transmission in M. plutonius [17,74,75]. Increased contact between honey bee colonies due to human management may allow for the spread of highly virulent strains unable to spread to new hosts under more natural settings.
More extensive field screening of the prevalence of different M. plutonius sequence types and of the plasmid carrying the melissotoxin A gene will reveal if highly virulent strains of M. plutonius arose by acquiring this toxin-carrying mobile genetic element. This will also contribute to a better understanding of the origin of recent epidemic outbreaks of EFB. If a link between virulence and disease outbreaks in the field can be established, strain-specific, virulence-based control methods can be promising avenues to better control the spread of EFB [29,30]. The melissotoxin A gene could provide a putative marker for such specific control methods.

Conclusion
Infection with M. plutonius alone caused European foulbrood symptoms in honey bee brood reared in vitro. Bee mortality varied substantially among 16 investigated M. plutonius isolates, independently of their sequence type. Mortality did not increase when larvae were infected with the potential secondary agent P. alvei in combination with a M. plutonius isolate of low to intermediate virulence. Variation in the virulence of M. plutonius isolates was not significantly linked to growth dynamics. A high final density in the culture medium was reached by highly virulent isolates, but was not sufficient to explain the high mortality caused by these M. plutonius isolates. High virulence was associated with the presence of a plasmid carrying the putative toxin melissotoxin A, suggesting that this mobile genetic element is a major factor in the virulence of M. plutonius. These new insights into the pathogenicity of a poorly understood, but important honey bee disease may serve as a basis for the development of more sustainable control methods.

Experimental Procedures
Origin and culture of M. plutonius isolates Diseased larvae were collected from European foulbrood (EFB) outbreaks in several European regions (Table 1) and sent to our laboratory for the purpose of M. plutonius isolation. The only exception was the Norwegian samples, which were processed elsewhere. Melissococcus plutonius was isolated from smears of diseased larvae on basal medium. The medium contained 1% yeast extract, 1% glucose, 1% saccharose, 0.04% L-cysteine and 0.1 M KH 2 PO 4 in distilled water, with a pH adjusted to 6.7 with 5 M KOH. The medium was solidified with 18 g agar/liter and autoclaved at 121°C for 18 min [58,76]. After incubation for four days at 36°C under anaerobic conditions (GENbox anaer, bioMérieux), individual bacterial colonies identified as M. plutonius based on colony morphology were picked from the Petri dishes and inoculated in a liquid basal medium or on plates. The plates were incubated anaerobically for another four days at 36°C. Isolate stock solutions were supplemented with 15% glycerol and stored at −80°C until further use.

PCR identification of M. plutonius and detection of the gene encoding melissotoxin A
Isolates were confirmed as M. plutonius by PCR as described by Govan et al., 1998 [74]. The strains were further identified as either "typical" or "atypical" following the protocol described in Arai et al., 2014 [77]. The multi-locus sequence type and clonal complex of each isolate were assessed as described in Haynes et al., 2013 [36]. We screened for the presence of melissotoxin A gene (GenBank: KMT29105) [39] using specific primers (tox_MEPL_for: 5ʹ-GCTCAAGCAGCAACTTTT ACG-3ʹ and tox_MEPL_rev: 5ʹ-TTCCCCTGGTAT TACTTGTAGATG-3ʹ; fragment size approx. 1.36 kbp) in a conventional PCR reaction using KAPA2 G Fast DNA Polymerase (KAPA Biosystems). DNA was extracted with the NucleoSpin Tissue kit from Macherey-Nagel according to the manufacturer's instructions, the DNA-concentration was measured with NanoDrop® ND-1000 spectrophotometer (NanoDrop, Thermo Fischer Scientific, USA) and each DNA-extract diluted to 5 ng DNA per µl. Each reaction in the PCR consisted of 2 µl template DNAextract, 13 µl KAPA2 G fast ready mix (2x), 1 µl forward primer (10 µM), 1 µl reverse primer (10 µM), and 8 µl water added to a total reaction volume of 25 µl. The PCR started with an initial denaturation step at 95°C for 3 min, followed by 40 cycles of denaturation at 95°C for 15 s, annealing at 58°C for 15 s and DNA extension at 72°C for 20 s, and a final extension step at 72°C for 2 min (TProfessional Basic Thermocycler, Biometra). PCR products were visualized under UV light after staining the 1.5% agarose gel with GelRed™ (Biotium).

Infection assays of honey bee larvae reared in vitro
Honey bee brood originating from healthy, queenright colonies was used in the infection assays. Same aged larvae were obtained by confining queens to empty combs for the purpose of egg-laying, using excluder cages. After 36 hours, the queens were removed from the cages for three days, until first instar larvae hatched. First instar larvae were grafted individually in plastic queen starter cells (Nicoplast™) that had been sterilized in 70% ethanol for 30 min. The plastic cells were placed in the cells of 48-well tissue culture plates, on top of a piece of wet dental roll imbibed with 15.5% glycerol in 0.4% methyl benzethonium chloride to prevent unwanted microorganism growth. The larvae were reared in vitro, according to standard methods [59]. In brief, the culture plates were placed in a hermetic desiccator containing a dish filled with saturated K 2 SO 4 solution, which ensured a relative humidity of 95% needed to prevent dehydration of larvae. The desiccator containing the larvae was placed into an incubator at 34.5°C for the first six days of larval development. At day 7, when the larvae defecate and start to pupate, the plates were moved into a desiccator containing a dish filled with saturated NaCl solution to ensure an optimal relative humidity of 75%. The larvae were fed daily with pre-warmed diet (34.5°C), according to the following feeding program: on the day of grafting, larvae were provided with 10 µl of diet A (1.2 g glucose, 1.2 g fructose, 0.2 g yeast extract and 8.4 g pure water, filter sterilized (0.2 µm) and mixed with 10 g pure royal jelly). On day 3, larvae were fed 20 µl of diet B (1.5 g glucose, 1.5 g fructose, 0.3 g yeast extract and 8 g pure water, filter sterilized (0.2 µm) and mixed with 10 g pure royal jelly). On days 4, 5 and 6, larvae were fed with 30, 40 and 50 µl of diet C, respectively. Diet C consisted of 1.8 g glucose, 1.8 g fructose, 0.4 g yeast extract and 7.45 g pure water, filter sterilized (0.2 µm) and mixed with 10 g pure royal jelly [59,78]. The royal jelly was obtained from healthy colonies and stored at −20°C.
For in vitro infection experiments, M. plutonius isolates were cultivated in a liquid basal medium at 36°C under anaerobic conditions for four days. In order to standardize the number of bacteria administered in the infection assays, we determined the concentration of viable bacterial cells using colony forming units (CFUs) counting. The CFUs from a serial dilution spread on basal medium agar plates were counted after four days of incubation at 36°C under anaerobic conditions [76]. During this time and before administration, bacterial cultures were stored at 4°C.
Paenibacillus alvei type strain DSM29 was obtained from the German Culture Collection (DSMZ, Braunschweig, Germany) and cultivated on caseinpeptone agar plates under aerobic conditions at 30°C for three weeks, until spores formed. Spores were harvested from agar plates by scraping them off the plate and suspending them in saline (0.9% NaCl in distilled water). The spores were washed once with saline, diluted in saline and heat treated for 5 minutes at 90°C . The total P. alvei spore concentration was determined via 10-fold dilutions plated on casein-peptone agar plates and CFU counting [79]. Spore suspensions were stored at 4°C.
Infection with M. plutonius was induced by administering a droplet of 10 µl diet A spiked with 10 7 CFU ml −1 of the respective M. plutonius isolate (1:9 M. plutonius inoculum-diet mix) on day 1, within two hours of grafting. Hence, each larva was fed 10 5 CFU of M. plutonius. Larvae were infected with P. alvei on day 1 by receiving 10 µl of diet A spiked with 10 5 spores of P. alvei per ml of saline (i.e. 10 3 P. alvei spores per larva). Co-infections were performed by feeding 10 µl diet A spiked with a mix of 10 5 P. alvei DSM29 spores ml −1 and 10 7 CFU ml −1 of M. plutonius isolate CH 90. This isolate was selected for co-infection assays for its relatively low virulence, which facilitates the detection of a putative increased mortality due to P. alvei. Non-infected control larvae received 10 µl of diet A mixed with sterile saline. We administered a lower number of P. alvei spores per larva (10 3 ) than in other studies (6 x 10 4 ; [42,47]). This lower dose is closer to the one eliciting American foulbrood symptoms in infections by the congeneric bacteria P. larvae (20 spores/larva; [80]), and comparable to the number of P. alvei spores larvae were exposed to in a recent study (8.3 x 10 3 ; [42]).
Honey bee larvae were subjected to single infections with M. plutonius, single infections with P. alvei and coinfections with M. plutonius and P. alvei. We monitored brood survival and compared brood mortality caused by the isolates, alone or in combination. The status of each larva (dead or alive) was recorded every 24 h by observing it under a microscope. Larvae or pupae without signs of respiration or reaction to mechanical stimulus were recorded as dead and removed from the plates [78]. In contrast to other studies [34,41,42,46,47], we monitored brood survival until completion of development (i.e. until imaginal stage) to take into account the whole brood developmental period to assess its mortality. For each strain or combination of strains, 2-6 replicates were performed, each with 24-84 larvae produced by 2-8 queens (Tab. S2).

Growth dynamics of M. plutonius isolates
For each M. plutonius isolate used in the infections assays (17 isolates), ten replicates (two runs with five replicates each) of 20 ml liquid basal medium were inoculated with a single M. plutonius colony precultured on an agar plate for three days and incubated under anaerobic conditions at 36°C. Bacterial growth was monitored by measuring optical density at 600 nm (OD 600 ) with a spectrophotometer (DR/2000, HACH), every 6 h until 96 h past inoculation [81]. OD 600 was averaged over the ten replicates. Based on this data, we calculated the generation time and compared the growth dynamics of the isolates (see below).

Statistics
Survival differences in honey bee brood after experimental infection with M. plutonius or P. alvei, or with a combination of M. plutonius and P. alvei, were illustrated with Kaplan-Meier survival curves with 95% confidence intervals [82]. Differences in survival between hosts infected by various isolates, and between the latter and uninfected controls were tested using pairwise log-rank tests (Mantel-Haenszel test [83]) and adopting a significance level α of 0.05, corrected by a Bonferroni-Holm procedure for multiple comparisons [84]. We calculated the Henderson-Tilton corrected mortality rate of individuals dead until day 21 of development, when the imago stage is reached, as follows: mortality ratio = 1-((number of live test bees after treatment * number of live control bees after treatment -1 ) * (number of live control bees before treatment * number of live bees before treatment -1 )) [85], and state the resulting mortality rate caused by each tested M. plutonius isolate in Table 1.
We used the R package growthrates [86] to compute the growth constant k for each isolate, by fitting segments of linear models to the log-transformed OD 600 values during the exponential growth phase [81]. The generation time (g) of each isolate was then calculated [87] as g = ln (2)/k. As recommended by Hall et al., 2013 [81], we verified the fit to the exponential phase generated by the algorithm implemented in Petzoldt, 2016 [86]. For isolate CH 54.1, the algorithm did not identify the exponential phase correctly, due to a long and irregular lag phase. We obtained a corrected dataset by excluding the first six points for this isolate well ahead of the exponential phase, which led to correct fitting. The overall correlation between growth rate and mortality, as well as the results of the regression model, were similar between the corrected dataset, the original dataset, and after excluding this isolate from the dataset (Tab. S7). We chose to present conservative results of the regression model, based on the dataset excluding CH 54.1.
Generation time only considers the exponential phase of the bacterial growth cycle. To analyze the growth dynamics over the entire experiment (96 h), we used permutation tests with 100,000 iterations to conduct pairwise comparisons between growth curves [88,89]. These statistical analyses were done in R (R Foundation for Statistical Computing, Vienna, Austria) with the statmod package [90].
A multiple regression analysis [91] was conducted to examine the relationship between the Henderson-Tilton corrected mortality and the potential virulence factors multi-locus sequence type, clonal complex, generation time, final density (density reached at 84 h) and presence of melissotoxin A gene. For this model, we used SPSS Statistics, version 21 (IBM Corp.).