Genetic and Biochemical Diversity of Paenibacillus larvae Isolated from Tunisian Infected Honey Bee Broods

Paenibacillus larvae is the causative agent of American foulbrood (AFB), a virulent disease of honeybee (Apis mellifera) larvae. In Tunisia, AFB has been detected in many beekeeping areas, where it causes important economic losses, but nothing is known about the diversity of the causing agent. Seventy-five isolates of P. larvae, identified by biochemical tests and 16S rRNA gene sequencing, were obtained from fifteen contaminated broods showing typical AFB symptoms, collected in different locations in the northern part of the country. Using BOX-PCR, a distinct profile of P. larvae with respect to related Paenibacillus species was detected which may be useful for its identification. Some P. larvae-specific bands represented novel potential molecular markers for the species. BOX-PCR fingerprints indicated a relatively high intraspecific diversity among the isolates not described previously with several molecular polymorphisms identifying six genotypes on polyacrylamide gel. Polymorphisms were also detected in several biochemical characters (indol production, nitrate reduction, and methyl red and oxidase tests). Contrary to the relatively high intraspecies molecular and phenotypic diversity, the in vivo virulence of three selected P. larvae genotypes did not differ significantly, suggesting that the genotypic/phenotypic differences are neutral or related to ecological aspects other than virulence.


Introduction
American foulbrood (AFB), a severe and highly contagious disease affecting the larval and pupal stages of honeybee (Apis mellifera), is caused by the bacterium Paenibacillus larvae [1,2]. AFB is one of the few diseases capable of killing the honeybee colony [3]. Prevention and control of AFB are very difficult because the pathogen produces spores that are resistant to heat and chemical agents and can remain viable for more than 35 years [4,5]. AFB is causing considerable economic loss to beekeepers all over the world [6][7][8][9], and it is classified on list B of the World Organization for Animal Health [10]. In many countries, an eradication strategy exists with isolation and destruction of infected colonies and burning of contaminated equipments [10,11].
Different genotypes of P. larvae have been identified in different regions. By using BOX-PCR three genotypes (A, B, and C) have been identified within a worldwide isolate collection [12]. In Germany, four different genotypes of P. larvae, named AB, Ab, ab and B, have been described by combining BOX A1R and MBO REP1 primers [13,14]. Using the same combination (BOX A1R and MBO REP 1 primers), Loncaric et al. [15] described five different genotypes (ab, aB, Ab, AB, and b). After the reclassification of P. larvae was proposed, as one species without subspecies separation, the use of other techniques as ERIC-PCR for subtyping P. larvae and four different genotypes (ERIC I-IV) were identified [1]. The genotypes ERIC I and II correspond to the former Paenibacillus larvae subsp. larvae and ERIC III and IV to the former Paenibacillus larvae subsp. pulvifaciens. AFB is readily disseminated by honeybees robbing honey from neighboring hives and the larval feeding of sporescontaminated pollen and honey [16] or the reuse of contaminated beekeeping equipments [11]. A role in the spread of P. larvae has been also attributed to Varroa destructor [17] and the hive beetle Aethina tumida [18]. The spores ingested by the newly hatched larvae germinate in the midgut lumen. The vegetative forms of P. larvae penetrate the gut epithelium and spread into the larval tissues [19,20].
Although some studies have investigated the pathogenicity of P. larvae and the virulence factors involved in the infection, the picture of the P. larvae virulence mechanisms is not yet complete. Dancer and Chantawannakul [21] associated the pathogenicity of P. larvae with the secretion of metalloproteases. Antúnez et al. [22] reported the production by P. larvae of an enolase that could have a role in the virulence of the pathogen. Recently, P. larvae virulence has been associated with an S layer protein [23] whose presence determined the difference in the virulence between ERIC I and ERIC II genotypes [24] with the former showing a weaker virulence due to the absence of the specific S-layer [23]. This study evidenced the importance of P. larvae genetic diversity in relation to virulence and highlighted the need for assessing the intraspecies diversity in areas of intensive apiculture.
AFB disease has been reported in Arab countries including North Africa [25] and in Tunisia; it has been detected in many beekeeping areas, where it causes important economic losses. Even though it has been shown that the economic value of pollination in North Africa is among the highest of the African continent [26], very limited knowledge is available on AFB and the genetic diversity of P. larvae.
The aim of the present work was to characterize a collection of P. larvae isolated from Tunisian diseased brood and to study the genetic and biochemical diversity related to these isolates.

Materials and Methods
2.1. P. larvae Isolation. Seventy-five isolates of P. larvae were obtained between 2003 and 2005 from diseased honeybee larvae originating from 15 different hives in the northern part of Tunisia ( Figure 1). The isolates were obtained on Columbia blood agar containing 5% horse blood for 48 h at 37 ∘ C. This step was preceded by a heat treatment at 80 ∘ C for 10 min to eliminate the quick growing bacteria that may outcompete P. larvae on the plates. Nine reference strains of seven Paenibacillus species phylogenetically related to P. larvae were obtained from the Bacillus Genetic Stock Center (BGSC), USA: Paenibacillus alvei 33A3 and 33A4, Paenibacillus polymyxa ATCC842T, Paenibacillus popilliae 2525 and B2519, Paenibacillus vorticalis 30A1, Paenibacillus thiaminolyticus NRRLB-4156T, Paenibacillus dendritiformis T168, and Paenibacillus macerans BKM B-51. All these reference strains were routinely cultivated on nutrient broth and agar at 30 ∘ C for 24 h.

Phenotypic and Biochemical Characterization.
Cell and colony morphologies of all isolates were described, and their biochemical profile was determined according to Gordon et al. [27] with the following tests: catalase test, nitrate reduction, gelatin, starch, and casein hydrolysis, tyrosine and urea degradation, acid from glucose, oxidase test, VP test, production of dihydroxyacetone, and indol and citrate test. The growth was tested at different temperatures (4 ∘ C, 30 ∘ C, 37 ∘ C, and 50 ∘ C) and in media containing 2% and 5% of NaCl. All phenotypic tests were made in triplicate and repeated when inconsistent results were observed. Positive and negative results were coded as 1 and 0, respectively, and cluster analysis was carried out by the unweighted pair group method with arithmetic averages (UPGMA) using the Jaccard coefficient [28].
PCR amplification of the 16S rRNA gene and the BOX gene was performed using the universal primers, S-D-Bact-0008-a-S-20/S-D-Bact-1495-a-A-20 and BOX A1R, respectively [29]. PCRs were performed in a final volume of 25 L containing 0.5 M of each oligonucleotide primer for the 16S rRNA PCR and 1 M for the BOX PCR primer, 200 M dNTPs, 2.5 mM MgCl 2 , and 1U of DNA Taq polymerase. PCR was performed for 35 cycles of 45 s at 94 ∘ C, 45 s at 55 ∘ C/42 ∘ C, respectively, for 16S rRNA PCR and BOX-PCR and 60 s at 72 ∘ C. BOX-PCR products were separated in standard 1.5% agarose gel and in 6% polyacrylamide gel, visualized under UV light, and photographed with a gel doc digital image capture system (Bio-Rad).
Numerical analysis of BOX patterns was performed using the MVSP 3.1 software [32]. Bands from all the gels were manually detected using as markers the 100 bp (Fermentas) or the 50 bp ladders (Promega), allowing the identification of the different BOX genotypes.

16S rRNA Gene Sequencing and Phylogenetic Analysis.
The 16S rRNA gene sequencing was performed at the Primm Biotech (Milano, Italy). Partial 16S rRNA gene sequences (E. coli coordinates nt 52 to 787) of the isolates were compared with 16S rRNA gene sequences available by the BLAST search [33], in the National Centre for Biotechnology Information (NCBI) database (http://www.ncbi.nlm.nih.gov/). Multiple sequence alignments were performed using ClustalW version 1.8 [34]. The method of Jukes and Cantor [35] was used to calculate evolutionary distances. Phylogenetic tree was constructed by the neighbor-joining method [36], and the reliability of the tree topology was evaluated by bootstrap analysis of 500 resampled data sets using MEGA 4.1 software [37,38].
16S rRNA gene sequences of thirteen  [39], with few modifications. The sporulated cultures were centrifuged at 3000 rcf for 15 min, and the spores were washed twice with sterile distilled water. The number of spores in the final suspensions was determined by plate count after 80 ∘ C heat treatment. The spore solutions were further diluted in larval diet to give final concentrations of approximately 5 × 10 3 CFU mL −1 and 10 5 CFU mL −1 .
Honey bee larvae of <24 h (based on body size) were collected from a healthy beehive and reared in U-shaped 96-well plates according to the method of Peng et al. [40]. The grafted larvae were fed with an artificial liquid diet containing 50% of royal jelly, 50% of an aqueous solution of yeast extract, and 12% each of D-glucose and D-fructose, both filtered at 0.2 m [41]. The diet was provided to the larvae with micropipette once a day for six days. For experimental infection and before grafting, each well of the plate was filled with 20 L of artificial liquid diet supplemented with a final P. larvae spore concentrations of 5 × 10 3 CFU mL −1 or 10 5 CFU mL −1 for the exposed groups and without P. larvae spores for the control group. The larvae were exposed to P. larvae spores for 24 h after grafting. Forty-eight larvae per group were used in this exposure test, and the experiments were performed three times. After grafting, the plates containing young larvae were incubated at 35 ∘ C in presence of a saturated solution of K 2 SO 4 to keep the humidity at 96% [41]. During each of the eight days of rearing, the larvae were examined for their vitality, and dead and symptomatic individuals were noted both for the larvae exposed to P. larvae and the nontreated ones.
2.6. Statistical Analysis. Statistics were calculated by using Microsoft Excel software [42]. Mean and standard deviations were determined for three independent experiments, and results were presented as mean ± SD. The Student's t-test was used to test for statistical significance of the difference between the mortality of the three groups of infected larvae with three different strains of P. larvae. A value of less than 0.05 was considered statistically significant.

Biochemical, Physiological, and Morphological Characters.
P. larvae colonies were small (3 mm in diameter), regular, buttery, and greyish. Cells were examined, and all isolates were Gram-positive rods with a width of about 1 m and a length of 3-5 m. Bacteria appeared as single cells or pairs and sometimes as short chains.
All isolates were catalase negative, grew at 30 and 37 ∘ C and in 2% NaCl media, but not in nutrient broth, at 4 ∘ C, at 50 ∘ C, and 5% NaCl. Citrate was not utilized. Isolates were positive for degradation of casein and gelatin and for acid production from glucose and starch. Tyrosine was not degraded. Most of strains reduced nitrate to nitrite. Variable results were obtained for oxidase and methyl red tests, and the strains did not form dihydroxyacetone and indol and were negative for the Voges-Proskauer test (Table 1).

Numerical
Analysis. The dendrogram of the biochemical results of the isolates and the reference strains discriminated

BOX-PCR
Analysis of P. larvae Isolates. BOX-PCR distinguished three genotypes out of 75 P. larvae isolates named A, B, and C (Figure 4(a)). P. larvae isolates presented a specific banding pattern clearly different from the other Paenibacillus species. The presence or absence of bands around 300 and 350 bp distinguished the three genotypes. Genotype A showed six bands of approximate sizes: 280, 300, 350, 650, 700, and 800 bp. Genotype B was characterized by the absence of the 350 bp band, and the genotype C showed only four bands of 280 bp, 650 bp, 700 bp, and 800 bp.
Eleven polymorphic bands in the 200-1000 bp range were detected within the BOX-PCR profiles separated by 6% polyacrylamide gel electrophoresis (Figure 4(b)), some of which could not be seen on agarose gel (Figure 4(a)). Six BOX-PCR genotypes (G1 to G6) were distinguished for the 75 isolates (Table 2). Genotypes G2 and G4 represented the most frequent in the collection, including 50% and 20% of the strains, respectively, while the remaining 30% of the strains were distributed among the other four genotypes (G1, G3, G5, and G6).   tree was selected for testing its virulence against honeybee larvae ( Figure 5). All the three isolates, BMG 93, BMG 184, and BMG 259, determined high mortality rates at 5 × 10 3 CFU mL −1 (50.3 ± 2.05%, 47.33 ± 3.5%, and 49 ± 2.6% mortalities, resp.) and at 10 5 CFU mL −1 (79 ± 3.8%, 73 ± 1%, and 75 ± 1.5 mortalities, resp.). No significant differences were observed between the three isolates based on the t-test (for the treatment with 5 × 10 3 CFU mL −1 , BMG 93 versus BMG 184, = 0.29; BMG 93 versus BMG 259, = 0.56; BMG 184 versus BMG 259, = 0.54; for the treatment with  Figure 5: Larval mortality rate after exposure to the pathogen P. larvae. Graphical representation of cumulative mortality percentage of larvae (± SD), fed with artificial diet supplemented with P. larvae spores at 5 × 10 3 CFU mL −1 or 10 5 CFU mL −1 , during 8 days. In Yaxis the mortality percentage of larvae reported, and in X axis the different P. larvae strains used of larval infection tested at the two spore concentrations are reported.

Discussion
In the dendrogram resuming the P. larvae isolates relationships according to the biochemical features (Figure 2), five branches corresponding to five biochemical phenotypes (P1 to P5) could be distinguished. This clustering was based on the detected polymorphism in several biochemical properties (nitrate reduction, oxidase production, and indol and methyl red tests). The isolates in branch 3, representing 88% of the isolates in the collection, presented typical characteristics of P. larvae [27] being Gram, casein, and gelatin positive, catalase, oxidase and starch negative, capable of using citrate, reducing nitrates to nitrites, and acidifying the medium from glucose without gas and H 2 S production and incapable of growing in media containing 5% NaCl or in nutrient broth. The other branches (12% of the collection) presented variability in four tests: nitrates reduction, methyl red test, and oxidase and indol production. The isolates in branch 1 (BMG 191, BMG 232, BMG 250, and BMG 257) were oxidase positive while isolate BMG 198 in branch 2 was double positive for oxidase and methyl red. The positive response of P. larvae to methyl red and oxidase was not described previously. Isolates BMG 192 and BMG 194 in branch 4 were able to produce indol, and isolates BMG 184 and BMG 189 in branch 5 contained isolates unable to reduce nitrates. These results obtained with isolates retrieved from a relatively small area of northern Tunisia show that P. larvae is not a monoclonal species like several other pathogens supporting previous observations [4,14] of a certain phenotypic variability highlighted in the former subspecies P. larvae subsp. larvae and P. larvae subsp. pulvifaciens.
16S rRNA gene sequencing confirmed the assignment of all the strains to P. larvae but highlighted certain sequence variability among the isolates confirming the lack of a strict clonality in the species according to the biochemical study. However, it was not possible to identify a clear correspondence in the isolate grouping between the phenotypes and the 16S rRNA gene sequence variability.
A relative intraspecific diversity within the 75 Tunisian isolates was further confirmed by BOX-PCR typing which allowed the distinction of P. larvae from the related Paenibacillus species. In addition, BOX profiles showed polymorphic bands specific for P. larvae that could be useful for its identification as in the case of other pathogenic bacilli like B. anthracis [29]. Using BOX-PCR, an unexpected genetic variability was revealed for isolates derived from a relatively small region northern Tunisia. Alippi and Aguilar [12], by typing by BOX-PCR a collection of 100 P. larvae originating from a geographic area much larger than northern Tunisia, detected only three genotypes. BOX-PCR combined to REP-PCR revealed four genotypes within a collection of 105 strains of P. larvae isolated from Germany [13,14]. Similarly, within a collection of 214 P. larvae isolates from Austria only five genotypes were identified by PCR typing using BOX A1R and MBO REP1 primers [15]. The results obtained with the present Tunisian isolate collection suggest that the genetic and phenotypic variability of P. larvae can be larger than that previously estimated.
However, despite the combination of the three approaches, the biochemical, phylogenetic, and molecular typing methods highlighted a relatively high intraspecific diversity of the Tunisian P. larvae collection; a clear correlation and grouping of the isolates according to the three methods was not evidenced. This may indicate slightly distinct evolutionary pathways within the species that apparently remain neutral and not yet clearly evident in distinct coherent phenotypes.
The attempt to search for a possible effect of the different observed phenotypes/genotypes on the level of virulence supports the considerations that the observed differences have no apparent effects on the pathogenicity against the honeybee larvae, at least in the conditions adopted in the study to test the virulence. Our results showed that the three tested isolates of P. larvae, BMG 93, BMG 184, and BMG 259 representing three 16S rRNA gene phylotypes and two BOX-PCR genotypes, presented the same virulence level against honeybee larvae. Such lack of correlation could be due to the procedure adopted, and we cannot exclude that, for instance, the low number of isolates tested in the virulence assays or the limited period (8 days) for observing the mortality may have prevented the observation of virulence differences among the different Tunisian genotypes of P. larvae. Also, we do not know the ERIC type of our isolates since all the PCR attempts to get clear fingerprints with the Tunisian isolates failed. For instance, we could be in presence of a collection of isolates representing a single ERIC type and hence a single virulence type [24]. Similarly we cannot exclude that in the beehive the virulence behavior of the Tunisian isolates may vary [43].

Conclusion
By keeping in mind all the above considerations related to the limitations of the adopted experimental conditions, the present data indicate a relatively high biological variability of P. larvae in northern Tunisia and suggest that the variable phenotypic and genotypic traits observed in the isolate collection apparently have a neutral effect in relation to virulence or affect other ecological aspects of P. larvae nondetectable with the experimental approaches used here.