• Open Access

Genome sequences of the honey bee pathogens Paenibacillus larvae and Ascosphaera apis


  • Re-use of this article is permitted in accordance with the Creative Commons Deed, Attribution 2·5, which does not permit commercial exploitation.

George M. Weinstock. Human Genome Sequencing Center, Baylor College of Medicine, One Baylor Plaza, Alkek N1619, Houston, TX 77025 USA. Tel.: +1 713 798 4357; fax: +1 713 798 4374; e-mail: gwstock@bcm.tmc.edu


Genome sequences offer a broad view of host–pathogen interactions at the systems biology level. With the completion of the sequence of the honey bee, interest in the relevant pathogens is heightened. Here we report the genome sequences of two of the major pathogens of honey bees, the bacterium Paenibacillus larvae (causative agent for American foulbrood disease) and the fungus Ascosphaera apis. (causative agent for chalkbrood disease). Ongoing efforts to characterize the genomes of these species can be used to understand and mitigate the effects of two important pathogens, and will provide a contrast with pathogenic, benign and freeliving relatives.


The completion of the genome sequence of the honey bee Apis mellifera (Honey Bee Genome Sequencing Consortium, 2006) represents a significant contribution toward understanding the social processes in this organism. A full understanding of the genetic implications and mechanisms of sociality requires a complementary description and understanding of the environment in which the organism exists. Infectious diseases are a component of such an environment that must be critically controlled in the dense population structures of a social organism. This is no exception with the honey bee and important pathogens have been described that can be devastating to a colony. Two of the most important are the bacterium Paenibacillus larvae and the fungus Ascosphaera apis. To appreciate the infectious processes of these microbes, and thus the resistance mechanisms developed by the honey bee (Evans et al., 2006), requires a complete elucidation of virulence factors and accessory elements needed for infection. To this end, we report draft genome sequences for these two pathogens as a prelude to more extensive genome annotation and functional analyses.

The spore-forming bacterium Paenibacillus larvae (Genersch et al., 2006) is the agent behind American foulbrood (AFB), a widespread larval pathogen of the honey bee, Apis mellifera. Young larvae (from the first and second instars) are highly susceptible to this disease, with the ingestion of as few as 10 infectious spores from virulent strains being sufficient to cause mortality (Brodsgaard et al., 1998). Infected bees generally die from the disease late in larval development, forming a dry ‘scale’ with approximately 2 × 109 bacterial spores/bee (Shimanuki & Knox, 1997). Despite the observed virulence of P. larvae toward bees, many strains of this bacterium are relatively benign. These strains, once classified as separate subspecies (Heyndrickx et al., 1996), do not reflect distinct genetic lineages and there are great incentives from the standpoint of both diagnostic tests and research questions to identify those traits directly linked to virulence. A congener of P. larvae, P. alvei, is also found in bees but is considered to be a relatively benign species. The genus Paenibacillus as a whole is comprised mainly of soil-dwelling opportunistic bacteria. Among the Bacillales, Paenibacillus appears to be most closely related to the genus Brevibacillus. Bees mitigate this disease both through hygienic behaviour by adult workers and through larval resistance traits.

Ascosphaera apis (Qin et al., 1993; Anderson et al., 1998) is the fungal cause of the honey bee larval disease chalkbrood (Williams, 2000; Hornitzky, 2001). Spores of this fungus germinate within the digestive tract of bees, then begin fungal filamentous (mycelial) growth during the last instar of larval development. Dead larval and pupal bees appear chalky due to growth throughout the bee of mycelia. These chalky ‘mummies’ are highly infectious, and spores of this fungus often reinfect colonies via stored food supplies or direct transport to younger larvae by adult bees working within the nest. Adult bees reduce the effects of this fungus on the colony by frequently identifying and removing diseased individuals. The disease is associated with high brood density (productivity) and cooler outside temperatures.


Draft sequences of P. larvae and A. apis were produced by standard procedures and have been deposited at DDBJ/EMBL/GenBank under the project accession AARF00000000 (P. larvae) and AARE00000000 (A. apis). The versions described in this paper are the first versions, AARF01000000 and AARE01000000, respectively. Whole genome shotgun libraries were constructed in plasmid vectors and paired-end sequences were generated by the Sanger dideoxy-terminator technique and assembled using the Atlas and Phrap assembly programs as described (McLeod et al., 2004). Paenibacillus larvae posed a number of challenges, such as obtaining a pure strain and sequencing difficulties, as manifested by a relatively low rate of successful sequence reads (Table 1). Moreover, after assembling sequences containing 43 Mb of high quality bases, expected to represent about 10× coverage of the genome, many regions were only covered at 6× or lower coverage (as judged by the number of overlapping reads), suggesting selection against cloning these sequences in Escherichia coli. In any case, this assembly produced contigs that spanned over 4 Mb, the expected size for this genome, with an N50 length of 11 kb, which is sufficient contiguity to allow prediction of genes and operons. These contigs were linked into scaffolds by read-pairs that doubled this N50 length. Thus the genome is 4.0 Mb and is likely nearly completely represented in the assembly.

Table 1. Paenibacillus larvae genome assembly v2005–07–06
 Attempted reads79 826
 Passed reads54 073 (68%)
 Average trimmed read length701
 Passed Q20 bases43.0 Mb (10.8× for 4.0 Mb genome)
 Read overlap peak9–12 (coverage 5–6×)
Assembly contigs
 Number of contigs646
 Total contig length4 016 553 bp
 Number of contigs = 2 kb471
 Total length of contigs = 2 kb3 758 779 bp
 Average contig length6217 bp
 N50 contig size11 019 bp
 Largest contig size58 621 bp
 Total assembled reads47 144
 Total assembled Q20 bases33.0 Mb (8.3× for 4.0 Mb genome)
 Number of scaffolds = 1 kb349
 Number of singleton scaffolds197
 Total size of singleton scaffolds982 435 bp
 Number of multiple-contig scaffolds152
 Number of contigs in multiple-contig scaffolds436
 Total size of multiple-contig scaffolds3 033 342 bp
 Total size of scaffold (including = 1 kb singletons)4 015 777 bp
 Average scaffold size11 506 bp
 N50 scaffold size23 797 bp
 Biggest scaffold137 418 bp

This draft sequence was used for gene predictions by two programs, GeneMark (Besemer & Borodovsky, 2005) and GLIMMER2 (Delcher et al., 1999). With GeneMark the computation was performed with four different Bacillus genomes as training sets (B. anthracis, producing 4917 predicted genes; B. cereus, 4915; B. halodurans, 5006; B. subtilis, 5036) yielding about 5000 predicted genes for each. When GLIMMER2 was used (with self-training) 8503 genes were predicted, but almost all of the excess were very small open reading frames, less than 400 bp. This could be the result of frameshift errors in the draft sequence that may have less effect on GeneMark. The predicted genes were compared to sequences in GenBank with Blast and the resulting automated annotation is available at http://www.hgsc.bcm.tmc.edu/projects/microbial.

The draft sequence of A. apis was produced in much the same way (Table 2). In this case a higher rate of successful reads was obtained but there also appeared to be some selection against sequences. A total of over 156 Mb of high quality bases was produced and assembled into contigs that totaled 21.6 Mb in length. This represents 7× coverage but again, a lower value (4×) was obtained by read overlap analysis, indicating possible selection against reads in E. coli. The N50 length for these contigs was low, 3 kb, which can result from repeated sequences and/or low coverage. However these short contigs were joined into very long scaffolds with an N50 of 44 kb. Thus while the contigs are small, the scaffolds are suitable for gene predictions (in progress).

Table 2. Ascosphaera apis genome assembly v2005–07–06
 Attempted reads256 859
 Passed reads204 025 (79.4%)
 Average trimmed read length675 bp
 Total passed Q20 bases156.2 Mb (7.3× for 21.5 Mb genome)
 Read overlap peak7 (coverage 4×)
Assembly contigs
 Number of contigs8092
 Total contig length21 566 106 bp
 Number of contigs = 2 kb4619
 Total = 2 kb contig length16 148 594 bp
 Average contig length2665 bp
 N50 size2967 bp
 Largest contig size16 610 bp
 Total assembled reads168 066
 Total assembled Q20 bases113.4 Mb (5.3× for 21.5 Mb genome)
 Average base coverage from ace7.0×
 Number of scaffolds = 1 kb1627
 Number of singleton scaffolds880
 Total size of singleton scaffolds1 548 023 bp
 Number of multiple-contig scaffolds747
 Number of contigs in multiple-contig scaffolds7172
 Total size of multiple-contig scaffolds19 729 929 bp
 Total size of scaffold (including = 1 kb singletons)21 277 952 bp
 Average scaffold size13 078 bp
 N50 scaffold size44 063 bp
 Biggest scaffold195 321 bp


The genomic view of host–pathogen interactions moves these molecular studies from piecemeal strategies to the systems biology level. With the completion of the sequence of the honey bee, interest in these pathogens is intensified. Knowledge of the genetic makeup of two of the major pathogens of honey bees now allows a full program to be developed.

Analysis of these two sequences will follow familiar pathways. Annotation will include community involvement, using gene lists produced by automated procedures. For P. larvae, prokaryotic gene prediction programs have produced such gene lists already. For A. apis producing comprehensive gene lists is more of a challenge. Because of the diversity of fungal coding sequences, available genomes used to train prediction programs may not be representative. However, more fungal genomes are being sequenced, for example over 25 fungal genomes as part of the Fungal Genome Initiative (http://www.broad.mit.edu/annotation/fgi/). Thus there is likely to be more effective gene prediction options available in the future.

This program will focus on more fully annotating these two pathogen genomes, contrasting their genomic makeup with that of related microbes, and generating functional data related to their survival and pathogenicity in bees. Given the long-term relationship between these pathogens and honey bees, extensive data on pathology, and an opportunity to look at disease transmission and efficacy in a social setting, this system offers excellent opportunities for studying host–pathogen relationships. Moreover, the honey bee has a simpler immune defence system than other insects (Evans et al., 2006; Honey Bee Genome Sequencing Consortium, 2006), and thus offers an opportunity for analysis of a less complex host–pathogen interaction.

Experimental procedures

Source of DNA

The sequenced strain of P. larvae was isolated from scales collected from a single severely diseased colony from Berkeley, CA, USA (isolate BRL-230010). A suspension containing 1 × 108 was heat shocked then used to inoculate replicate plates containing Brain-Heart Infusion agar (Difco, Franklin Lakes, NJ), supplemented with 0.1 mg/ml thiamine hydrochloride. Single colonies from this and two serial replatings were then grown at quantity on the same medium for genomic DNA.

DNA was isolated from scraped cells using an overnight incubation in a final concentration of 100 µg/ml proteinase K, followed by NaCl/CTAB precipitation and phenol/chloroform washes. DNA was precipitated using isopropanol, pulled from solution with a glass rod, and washed gently with EtOH before drying and suspension at approximately 100 ng/µl concentration. Extractions were also carried in situ in agarose gels using pronase. Slices from gel-extracted DNA were equilibrated in TBE buffer for PFGE analyses of genome size (below), while the remaining DNA was eluted by gel digestion (Gelase, Epicentre Biotechnologies, Madison, WI).

For A. apis, chalkbrood mummies were collected from a naturally infected honey bee colony at Weslaco, Texas. A single black mummy was pulverized in 1 ml 0.001% Trition X-100, and 50 µl of this suspension were plated on a YGPS plate (Bailey, 1981) containing ampicillin (100 µg/ml) and streptomycin (12 µg/ml). The plate was incubated for 9 days at 35 °C under 6% CO2, during which time fruiting bodies formed. Two millilitres of Triton X-100 were added to the plate, and the plate was rocked at 25 r.p.m. for 10 min at room temperature. One millilitre of the liquid was then mixed with 0.1 ml of 0.5 mm glass beads, and shaken at 300 r.p.m. for 1.5 h. Dilutions of this spore suspension were plated on YGPS ampicillin streptomycin plates in order to obtain single spore isolates of both mating types. After 3 days incubation, fungal colonies were picked and replated individually on fresh plates. Purified fungal colonies were tested to determine the mating type (Christensen & Gilliam, 1983).

One purified strain of each mating type was used for isolation of DNA for sequencing. One hundred-millilitre batches of liquid YGPS ampicillin streptomycin media were inoculated from purified mycelia growing on plates. Cultures were incubated for 6–9 days at 35 °C with shaking at 200 r.p.m. under normal atmosphere. Mycelial masses were removed from the cultures, and after squeezing out most of the liquid, were separated into aliquots, weighed, and stored at −80 °C until further processing. DNA was isolated from the mycelia by the method of Borges et al. (http://www.fgsc.net/fgn37/borges.html). The two A. apis isolates used in the genome sequencing project, designated 0.5–1A and A10, have been deposited in the USDA-ARSEF collection (Murray et al., 2005), with accession numbers 7405 and 7406, respectively.

Genome size estimation

Ascosphaera apis genome size was estimated using the flow cytometry method (Bennett et al., 2003) and staining nuclei with propidium iodide. Briefly, spores and hyphae of A. apis were collected from plates in microcentrifuge tubes in 0.5 ml 0.2% Triton-X in phosphate-buffered saline (PBS). Nuclei were released by crushing spores and hyphae with a micropestle, stained with PI for at least 20 min at 4 °C. The suspension was filtered through a nylon membrane with a pore size of 15 µm to remove debris. Propidium iodide stained nuclei were analysed using a flow cytometer.


Kevin Hackett (USDA-ARS) guided both sequencing efforts. Spencer Johnston (Texas A & M University) provided the A. apis genome size estimation and S.T. Rogers aided in the isolation and extraction of P. larvae. Cloning, sequencing, and assembly were carried out by members of the BCM-HGSC sequencing team. USDA-ARS provided funding.