From a shotgun clone library of strain Agy99, 352 Escherichia coli plasmids (pCDNA2.1, Invitrogen, Basel, Switzerland) were randomly selected. Each plasmid contained an M. ulcerans DNA fragment of ≈2.3–2.7 kb. Given a genome size of 5,806 kb (18), this set of plasmid inserts represents a theoretical genome coverage of ≈10%. Plasmid DNA was prepared by using a Biomek 2000 Workstation (Beckman Coulter, Krefeld, Germany) and dissolved at a concentration of 150 ng/µL in 3× SSC (20× SSC stock solution is 3 M sodium chloride, 0.2 M sodium citrate, pH 7.0). The DNA samples were loaded on a piezo-dispensing head that contained 24 channels and spotted onto glass slides coated with poly-L-lysine (Superfrost Plus, Menzel, Braunschweig, Germany) by using a Topspot spotter (Biofluidix, Freiburg, Germany). Slides were incubated at 4ºC overnight and rehydrated under 50%–60% humidity for 1 h at room temperature. The spots resulting from a volume of ≈1 nL had an average diameter of 270 µm and were 500 µm apart from each other. The microarray layout displayed 2 identical fields—for hybridization with 2 different probes—that consisted of 2 replicates each, both of which contained 32 controls and 352 plasmids.

M. ulcerans clinical isolates used in this study are listed in Figure 1. Bacterial pellets of about 60 mg (wet weight) were heat inactivated for 1 h at 95°C in 500 µL extraction buffer (50 mmol/L Tris-HCl, 25 mmol/L EDTA, 5% monosodium glutamate) and sequentially treated with lysozyme (2 h, 37°C, 17 M lysozyme) and proteinase K (overnight, 45°C, 0.3 M proteinase K in proteinase K buffer: 1 mmol/L Tris-HCl, 5 mmol/L EDTA, 0.05% sodium dodecyl sulfate [SDS], pH 7.8). After digestion the samples were subjected to bead beater treatment (Mikro-Dismembrator, Braun Biotech International, Berlin, Germany) with 300 µL of 0.1-mm zirconia beads (BioSpec Products, Bartlesville, OK, USA) for 7 min at 3,000 rpm. DNA was extracted from the supernatants by phenol-chloroform (Fluka, Buchs, Switzerland) extraction and ethanol precipitation. Seven micrograms of M. ulcerans genomic DNA was digested with 3 U of Sau3A1 (New England Biolabs, Hitchin, UK) for 2 h at 37ºC and biotinylated according to Pollack et al. (19) using a BioPrime kit (Gibco/BRL, Gaithersburg, MD, USA). The biotinylated DNA was purified by using a Microcone YM-30 filter (Amicon/Millipore, Bedford, MA, USA), and its concentration was measured by optical density at 260 nm (GeneQuant spectrophotometer, Cambridge, UK).

Five micrograms of biotinylated DNA was mixed with 30 µg human Cot-1 DNA (Roche Applied Science, Indianapolis, IN, USA) and 100 µg yeast tRNA (Gibco/BRL). The hybridization mix was concentrated with a Speed Vac Concentrator System (Eppendorf, Basel, Switzerland), resolved in 3× SSC, 0.3% SDS, denatured for 3 min at 95ºC, and incubated for 30 min at 37ºC before hybridization. Microarray slides were cleaned with a nitrogen flow, exposed to UV light in a Stratalinker 2400 (Stratagene, La Jolla, CA, USA) at 650× 100 µJ, and heated for 5 min to 95ºC before application of 13 µL of the hybridization mix on each array field. Hybridization occurred for 20 h at 65ºC in a hydration chamber. Hybridized slides were washed once with 2× SSC, 0.03% SDS for 5 min at 65ºC, twice with 1× SSC for 5 min at room temperature, and finally with 0.2× SSC for 5 min at room temperature. The coloration step was performed with 2 mL staining solution containing 50% caseine, 1× maleic acid buffer (Roche Applied Science), and 2 µg Streptavidin Cy3 Fluorolink (Amersham, Piscataway, NJ, USA) for 30 min at room temperature, followed by additional washings for 5 min with 1× TBS (0.15 M sodium chloride, 0.02 M Tris, pH 7.5) as well as 0.1× TBS and drying with a nitrogen flow. DNA of all 30 M. ulcerans strains was processed under identical conditions and hybridized at least twice, which yielded 4 sets of data for each strain. Human Cot-1 DNA and plasmid DNA without insert as well as a hybridization mix without DNA served as negative controls for hybridization. A 500-bp β-lactamase gene fragment and Cy3-labeled random oligonucleotides (Microsynth, Balgach, Switzerland) were used as positive controls and for estimation of the amount of spotted DNA.

Images of the microarrays were acquired by using a laser microarray scanner (GenePix 4100A, Axon Instruments Inc., Foster City, CA, USA) with an excitation wave length of 532 nm, an emission wavelength of 570 nm, and standardized measurement parameters. The resulting image was analyzed by the software GenePix Pro 4.1 (Axon Instruments Inc.), which enabled assignments of mean intensity values used for data interpretation. To select spots to be included in the analysis of genomic diversity of M. ulcerans strains, replicates of 10 hybridizations were performed by using M. ulcerans Agy99 genomic DNA. All spots that showed a signal lower than twice that given by the negative control plasmid without insert were rejected, as were all spots for which coefficient of variation was >30%. Further analysis used 232 spots that had an average signal above the threshold and sufficient signal stability. For each plasmid, we calculated the average signal value, standard deviation, and coefficient of variation and assessed a signal ratio in comparison with the reference strain. Outlier spots with a ratio higher than U2 (U2 = upper quartile + 3× interquartile) were identified through a box-plot analysis.

Microarray data that indicated the presence of a deletion were verified by PCR analysis, which used primer pairs that spanned the insertion sequences of the respective plasmids, the flanking regions, or both. The 5′ and 3′ limits of the confirmed genomic deletions with respect to the genome of strain Agy99 were determined by PCR analysis, which used multiple sets of primers complementary to flanking genomic regions. PCR analyses that bridged the genomic breakpoints were performed by using a long-range PCR polymerase mix (Fermentas, St Leon-Rot, Germany) according to the manufacturer’s description. PCR products were cloned into pGEM-T (Catalys AG, Promega, Wallisellen, Switzerland) and sequenced using an ABI PRISM 310 genetic sequence analyzer (Perkin-Elmer, Waltham, MA, USA).

We constructed a microarray based on a random selection of 232 Escherichia coli plasmids obtained from a shotgun sequence library of the M. ulcerans isolate Agy99 from Ghana. Genomic DNA hybridization signal intensities from 30 M. ulcerans clinical isolates of worldwide distribution (Figure 2) were compared with those obtained with strain Agy99. Box-plot analysis (Figure 3) identified plasmids that yielded outlier signals with respect to strain Agy99. For 19 of 20 plasmids, PCR analysis confirmed an association of the outlier signal with a genomic deletion. Only 1 low hybridization signal represented a false-positive result (p188 from strain 940511, Côte d’Ivoire; Figure 3). The number of confirmed outlier plasmids per isolate ranged from zero for most African isolates to 9 for isolates from Suriname and French Guiana (Figure 1).

Of the 19 plasmid inserts that yielded confirmed outlier signals, 3 (p111, p299, and p341) contained sequences from the virulence plasmid pMUM001 of M. ulcerans. Of the 16 plasmids derived from the M. ulcerans chromosome, some contained fragments that overlapped the same region (Figure 4). Hybridizing regions were almost identical for p60 and p61. Both plasmids yielded outlier values with the isolates from Suriname and French Guiana. A cluster of overlapping inserts was observed for p88, p153, and p360; these produced outlier values for both of the Mexican isolates. The same pattern was seen with p124 and p291, which have inserts that are located in close proximity to each other in the genome (Figure 3). These results from related inserts demonstrated the reproducibility of the differential hybridization analysis. Because the inserts p60–p61, p88–p153–p360, and p124–p291 were part of the same deletion in regions of difference (RDs) 4, 5, and 8, respectively; (Figure 4), altogether 12 chromosomal RDs were identified.

The 5′ and 3′ limits of the genomic deletions with respect to the genome of strain Agy99 were determined by PCR analysis that used multiple sets of primers complementary to plasmid inserts and to flanking genomic regions. The size of the deletions ranged from 1.8 kb to 53.1 kb (Table).

In 3 of the 12 RDs (RD3, 9, and 12), 2 distinct types of overlapping deletions (designated A and B) were observed, leading to a total of 15 large deletions. The overlapping deletions shared neither common 5′ nor 3′ end sequences. The strains from Australia had a 3.5-kb deletion in RD3; strains from Suriname and French Guiana had a slightly larger (3.8-kb) deletion. The isolates from Suriname and French Guiana had a larger (25.4-kb) deletion in RD9 than the isolates from Japan and China (17.7 kb). The largest deletion (53.1 kb) was designated RD12A and was observed in strains from Japan and China. Isolates from Suriname and French Guiana had a significantly smaller deletion in RD12 (35.2 kb). The 19.7-kb deletion 6 was found in isolates from 2 different regions (Mexico and Japan/China, respectively). All other deletions were observed in 2 isolates from the same region (Table).

To assess whether polymorphisms undetected by the microarray analysis would frequently occur in the identified RDs, we performed a detailed PCR analysis in all 30 M. ulcerans strains included in this study for 2 randomly selected RDs (RD5 and 12). We used 4 distinct primer pairs to span the insert sequence plus 5′ and 3′ flanking sequence stretches. For RD12, the PCR analysis confirmed the presence of a deletion in the 4 strains that had outlier signals in the microarray analysis, but no evidence for deletional polymorphism was obtained in the other strains. For RD5, PCR analysis confirmed the presence of a deletion in the 2 Mexican strains that had outlier signals (not shown). In addition, this PCR analysis identified the presence of an insertion in strains from Japan, China, Suriname, and French Guiana. The sequence of this 765-bp DNA insert was identical for all 4 strains. Its G+C content was 64%, and BLAST searches showed 98% identity with a sequence stretch of the M. marinum genome (www.sanger.ac.uk/cgi-bin/blast/submitblast/m_marinum) but no significant homology with sequences in the National Center for Biotechnology Information BLAST databases (www.ncbi.nlm.nih.gov/blast).

Of the 15 identified genome rearrangement events, 1 (deletion 3A observed in 2 Australian isolates) was found to be a deletion, with the genomic sequences flanking the 5′ and 3′ borders of the 3,451-bp deletion being directly joined (Figure 5). Analysis of the other 14 deletions showed that the loss of DNA in a given strain with respect to the genome of Agy99 was associated with the insertion of substituting sequences of varying sizes unrelated to the deleted regions. As an example, the larger (3,784-bp) deletion 3B found in the isolates from Suriname and French Guiana was associated with the insertion of an unrelated DNA fragment, which comprised the 1,368 bp of IS2404 (20) plus an additional DNA stretch of 163 bp (Figure 5). For most of the other deletions, 1 of the 2 highly abundant insertion sequence elements (IS2404 or IS2606) was situated in either the genomic sequences that flanked the deletion or that were in the deleted parts or in the substituting sequence stretches (as for deletion 3B).

The 15 deletions identified contained 52 pseudogenes and 185 predicted protein-coding sequences (CDSs), which represent 5.7% of the annotated 4,143 CDSs in the genome of the M. ulcerans strain Agy99 (18). The number of deleted CDSs and pseudogenes ranged from 2 (RD4) to 50 (RD8 and 12A) and averaged 18.6 per deletion (Table). CDSs were classified into 11 functional categories (17). When compared with the gene composition of the entire Agy99 genome, the following functional categories were overrepresented among the 185 deleted CDSs: insertion sequences, unique hypothetical genes, and predicted proteins involved in detoxification (Figure 6). Also overrepresented was the deletion of the 52 pseudogenes that contain frame shift mutations and premature stop codons or that are disrupted by an insertion sequence. In contrast, genes involved in intermediary metabolism, information pathways, and cell wall/cell processes were underrepresented among the deleted CDSs (Figure 6). Of the 185 deleted functional CDSs, 89 had orthologs with >50% amino acid sequence identity to proteins from the M. tuberculosis H37Rv genome. A tendency for gene categories to cluster within the RDs was found. RD2 comprises 2 PPE genes: RDs 1, 12A, and 12B are predominantly CDSs involved in lipid metabolism, and RDs 9A and 11 include mainly transcriptional regulators. However, overall M. ulcerans lineages from distinct geographic origin (Africa, Australia, Asia, South America, Mexico) did not differ markedly in the categories of deleted genes. RD8 (deleted in the Mexican strains) is particularly interesting because it contains a cluster of proteins of the mammalian cell entry mce3 operon and associated regulators thereof. The transcriptional repressor, Mce3R, is considered to be an essential gene required for growth of M. tuberculosis (21). In addition, RD8 comprises a collection of CDSs of almost every functional category (Appendix Table). The spectrum of RD8-associated CDSs involved in detoxification included the multidrug transport protein mmr, the epoxide hydrolase EphB, the thiol peroxidase Tpx, and the alkyl hydroperoxide reductase C protein AhpC.

Although CDSs involved in intermediary metabolism were underrepresented among the deleted genes, 21 (42%) of deleted CDSs of this category were dehydrogenases (such as acyl-CoA short-chain alcohol, saccharopine, and aldehyde dehydrogenases), which are central enzymes in anaerobic metabolism (22) and important for survival in poorly oxygenated environments such as soil (23). In addition, other genes associated with anaerobic respiration, such as nitroreductases and electron transfer proteins, were found among the deleted CDSs.