Genome-Wide DNA Microarray Analysis of Francisella tularensis Strains Demonstrates Extensive Genetic Conservation within the Species but Identifies Regions That Are Unique to the Highly Virulent F. tularensis subsp. tularensis

Bacterial strains and genomic DNA.The F. tularensis strains were obtained from the Francisella Strain Collection (FSC) of more than 300 strains at the Swedish Defense Research Agency. All strains have been characterized by means of specific agglutination, by PCR specific for a gene encoding an F. tularensis-specific 17-kDa lipoprotein (40), and by biochemical characterization. Twenty-seven F. tularensis strains, representing all four recognized subspecies, were selected for this study to ensure maximum genetic diversity (Table 1). The F. tularensis strains were grown for 2 days on modified Thayer-Martin agar plates at 37°C in 5% CO₂, harvested by scraping, and suspended in saline at a concentration of 10⁹ CFU/ml. The cells were lysed with guanidine isothiocyanate, and the DNA was captured on silica particles as described previously (2, 17). Avirulent strain Yersinia pestis KIM5 (obtained from R. R. Brubaker, Michigan State University, East Lansing) was included as a control strain. Heat-killed whole-cell bacterial lysates were used as templates to confirm the microarray results by PCR.

Genomic F. tularensis microarray construction.A total of 1,832 inserts cloned into pUC18 from a shotgun DNA library of F. tularensis Schu S4 (34) were selected by using the following criteria: good-quality readings, minimal overlap with other clones, and maximal coverage of the genome. The mean ± standard deviation insert sizes were 1,405 ± 205 bp. The DNA probes used for microarray construction were obtained by amplifying the inserts by PCR with vector-specific primers 5′-TGT AAA ACG ACG GCC AGT-3′ and 5′-ATG TTG TGT GGA ATT GTG-3′, and the sizes were visually verified on agarose gels. The PCR products were purified by using 96-well MultiScreen-PCR filter plates (Millipore), suspended in 3× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate), transferred to 384-well plates, and stored at −20°C. Two replicates of each PCR amplicon were printed onto glass slides (25 by 75 mm; CMT-GAPS Amino Silane-coated slides; Corning) with an SDDC-2 robot (Engineering Services, Inc., Toronto, Canada) equipped with Stealth micropins (ArrayIt; TeleChem). The microarrays, which featured eight grids of 480 probes (20 by 24) with a spot-to-spot distance of 160 μm and a total size of 1.4 cm², were postprocessed with UV cross-linking, blocked by a previously described method (http://cmgm.stanford.edu/pbrown/ ), and stored at room temperature with protection from light.

Microarray hybridization.Three micrograms of bacterial genomic DNA was digested with NlaIII and labeled with Cy-3-dCTP (Amersham Pharmacia Biotech) by using Klenow DNA polymerase and random octamer primers. The labeled DNA was precipitated with isopropanol, dissolved in 20 μl of DIG Easy Hyb hybridization solution (Roche) supplemented with 2 μl of herring sperm DNA (10 mg/ml), and then hybridized immediately. Microarray slides were prehybridized beneath glass coverslips with 20 μl of DIG Easy Hyb solution (which lowers the hybridization temperature as though it contained 50% formamide) and herring sperm DNA for 1 h at 37°C under humid conditions. Hybridization with labeled DNA was done overnight at 37°C. The slides were washed three times in prewarmed 0.1× SSC-0.1% sodium dodecyl sulfate for 10 min at 50°C with occasional agitation, washed in 0.1× SSC, and dried by centrifugation at 46 × g for 5 min. The slides were shielded to avoid exposure to excessive light during washing and subsequent handling.

Scanning and data analysis.Fluorescent scanning was performed with a ScanArray 4000 instrument (GSI Lumonics, Billerica, Mass.) at a resolution of 5 μm and a 16-bit image depth. Laser strength and photomultiplier gain were adjusted in order to give saturated signals for <1% of the probes in the hybridizations. Probe signal intensities were quantified with ImaGene (version 3.04 and 4.0; BioDiscovery Inc., Marina del Rey, Calif.). Anomalous spots were identified by visual inspection and were excluded from analysis. To reduce the impact of inconsistent printing, only the higher-intensity value for each duplicate probe was used for further analysis. The data were globally normalized by dividing the intensity signal by the mean signal intensity for each hybridization experiment. The probe noise for 18 of the probe signal variables that had close to equal hybridization signal levels with reference strain Schu S4 was examined with F. tularensis subsp. tularensis strains (excluding ATCC 6223). Probes hybridizing with weak signal intensities for reference strain Schu S4 were found to have low signal-to-noise ratios and were excluded from further analysis, resulting in a data matrix consisting of 1,635 signal intensity values for each hybridization experiment. Differentially hybridizing probes were identified by using a criterion that required <30% of the signal intensity compared with that for reference strain Schu S4. A binary data matrix was derived for analysis of strain relatedness. Relationships between strains were determined by using hierarchical cluster analysis with complete linkage and the uncentered correlation metric (CLUSTER) (9). The cladogram was visualized by conversion of the output file into parenthetical notation (Newick Standard) and by use of TreeView software (31).

Analysis of RDs and development of a subspecies-specific PCR.The genetic differences between F. tularensis subsp. tularensis strains and non-Japanese F. tularensis subsp. holarctica strains detected by microarray analysis were designated genomic regions of difference (RDs). Each RD was verified by PCR analysis across each region in all strains. PCR amplification was performed with the following flanking primer pairs: 5′-TTT ATA TAG GTA AAT GTT TTA CCT GTA CCA-3′ and 5′-GCC GAG TTT GAT GCT GAA AA-3′ (RD1); 5′-CAT GTC TAT GTT ATT TAT CAC TCA TGA TTT-3′ and 5′-GCT TTA GCA TTC ATT ATC ATG AGA AGT AT-3′ (RD2); 5′-TCA ACT ACA GCA GAG TAT AAG CAA-3′ and 5′-ACA TTT TAA GAG CTC CTT GGT AGA T-3′ (RD3); 5′-TTT TAT AAA TCA AGA ATC AAT ATG CCT AA-3′ and 5′-TGA TGA AGG TCA TTG TCT TAG AGA T-3′ (RD4); 5′-TCA GTT ATC ATT TCA CTA AGT AAA GCA TT-3′ and 5′-GTA GAG TTT AAG GGC GAT GAG TT-3′ (RD5); 5′-TGA TTG TAC GCT TTG TTC TAT GAA TA-3′ and 5′-TTT GTC AAA GCA GGA ATC TTA-3′ (RD6); and 5′-CGA TAT GTT TTG CTA TAG ATA ACC TTG AT-3′ and 5′-CAC TTA TCG TAG CCG CTA AAC AT-3′ (RD7). The PCRs were performed in volumes of 25 μl with an initial denaturation at 94°C for 5 min, followed by 30 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 60 s. The PCR products were sized on agarose gels and visualized by ethidium bromide staining. The amplified DNA fragments from F. tularensis subsp. tularensis strain Schu S4 and the F. tularensis subsp. holarctica live vaccine strain (LVS) were purified with Micro-Spin S400HR columns (Amersham Pharmacia Biotech, Uppsala, Sweden), and both strands were sequenced by using the same primers as those used in the initial PCR amplifications and the Big-Dye terminator cycle-sequencing ready reaction kit on an ABI 377XL DNA sequencer (PE Applied Biosystems). An ∼11.5-kb region, RD8, could not be amplified from any strain of F. tularensis subsp. holarctica; hence, this region was not mapped in detail.

In addition, the highly variable RD1 was analyzed by sequencing the genomic DNA from strains representing each of the four F. tularensis subspecies. The large PCR product corresponding to RD1 of strain FSC040 was sequenced by primer walking.

The open reading frames identified in the sequences of all RDs were compared with the sequences in the GenBank database by using the BLASTX program with standard parameters and low-complexity filtering.

Nucleotide sequence accession numbers.The RD1 sequence data have been submitted to GenBank and assigned accession numbers AF469614 (Schu S4, FSC237), AF469615 (ATCC 6223, FSC230), AF469616 (FSC147), AF469617 (LVS FSC155), AF469618 (FSC075), and AF469619 (FSC040).

F. tularensis whole-genome microarray.The F. tularensis microarray was constructed by using 1,832 clones from a shotgun library of highly virulent strain Schu S4. DNA from the homologous organism hybridized with 98% of the probes. These genome sequence data are available online at http://artedi.ebc.uu.se/Projects/Francisella/ . Since the genome size is not yet precisely determined, the genome coverage was estimated by comparing the size of the present gene assembly with that estimated by separation of the F. tularensis genome by restriction enzyme cleavage and pulsed-field gel electrophoresis (unpublished data). On this basis, it is estimated that the probes of the DNA microarray represent more than 95% of the F. tularensis genome. As a control, 16 probes containing various mouse DNA sequences were included in the microarray. In all experiments, they did not hybridize to F. tularensis DNA. Chromosomal DNA from Y. pestis hybridized with 10 of 1,832 probes.

Genomic differences within the species F. tularensis.Chromosomal DNA from 27 F. tularensis strains (Table 1) showed differential hybridization to certain probes on the microarray. There were fewer differences between strains belonging to the same subspecies than between strains belonging to different subspecies (Fig. 1). A total of 0.1 to 3.7% differentially hybridizing probes (DHPs) were detected for a given strain (Table 1). A hierarchical cluster analysis was performed with the binary-coded signal-intensity microarray data, and a cladogram illustrating the similarity between strains is shown in Fig. 1. Generally, strains from the same subspecies formed clusters. Moreover, the cluster analysis indicated that (i) strains belonging to F. tularensis subsp. mediasiatica show close genetic similarity to strains of F. tularensis subsp. tularensis; (ii) the type strain of F. tularensis, ATCC 6223, is distinct from other strains belonging to F. tularensis subsp. tularensis; (iii) the Japanese strains cluster separately from the European and American F. tularensis subsp. holarctica strains; and (iv) the single representative of F. tularensis subsp. novicida showed a unique localization. Overall, only a few probes discriminated the strains. All DHPs present in LVS FSC155 were present in other European or North American strains of F. tularensis subsp. holarctica. The two most divergent strains were FSC040 (F. tularensis subsp. novicida) and ATCC 6223 (F. tularensis subsp. tularensis, avirulent). The strains showed 3.7 and 2.6% DHPs, respectively, many of which were unique for each strain (Table 1 and Fig. 1). The Japanese strains lacked a number of DHPs present among other strains of F. tularensis subsp. holarctica, which explains their location on the cluster plot (Fig. 1).

• Open in new tab
• Download powerpoint

FIG. 1.

Hierarchical cluster analysis of F. tularensis DNA microarray hybridizations. Strain identity is indicated by FSC number (Table 1) and subspecies.

F. tularensis RDs.Many of the DHPs clustered in contiguous genomic regions and were designated F. tularensis RDs. It was apparent that the presence or absence of F. tularensis RDs reflected the different subspecies of the strains. We focused further studies on eight such regions of 0.6 to 11.5 kb that differentiated strains of the highly virulent subspecies F. tularensis subsp. tularensis (type A) from strains of the moderately virulent subspecies F. tularensis subsp. holarctica (type B, non-Japanese) (Table 2). PCR and sequence analysis of these RDs revealed that in each case a 9- to 38-bp direct repeat motif was present in the corresponding Schu S4 sequence (Table 2 and Fig. 2). The only exception was RD8, in which the presence of such motifs could not be verified. Analysis of F. tularensis RD1 to RD8 identified a total of 21 open reading frames in the sequences absent from F. tularensis subsp. holarctica strains, with several open reading frames sharing high degrees of similarity to known bacterial genes (Table 2).

• Open in new tab
• Download powerpoint

FIG. 2.

Schematic depiction of the variable chromosomal region, RD1, identified among the different F. tularensis subspecies with insertions and deletions. Two direct repeats were found to define the boundaries of each deleted genomic segment. Within RD1, two complete open reading frames, designated A and B, were present. Both open reading frames A and B were found in strains Schu S4, ATCC 6223 (F. tularensis subsp. tularensis), and FSC147 (F. tularensis subsp. mediasiatica), although in strain FSC147 open reading frame A was truncated by 10 amino acids from the C terminus and open reading frame B was truncated by 12 amino acids from the N terminus. Only open reading frame B was found to be present in the LVS (F. tularensis subsp. holarctica), and only open reading frame A was found to be present in strain FSC075 (F. tularensis subsp. holarctica, Japan). Four open reading frames, designated C, D, E, and F, were found in strain FSC040 (F. tularensis subsp. novicida). Searches of the sequences in the GenBank database with the BLAST program could not identify any known gene function of the open reading frames of RD1.

F. tularensis RD1 was analyzed in detail, as it appeared to be highly variable between strains from all four subspecies. Sequence analysis of RD1 showed the following characteristics compared to the sequence of strain Schu S4: strain FSC147 (F. tularensis subsp. mediasiatica) showed a 68-bp deletion beginning after position 262; the LVS, strain FSC155 (F. tularensis subsp. holarctica, Europe), exhibited a 598-bp deletion beginning after position 418; and strain FSC075 (F. tularensis subsp. holarctica, Japan) had a 389-bp deletion beginning after position 1037. Strain FSC040 (F. tularensis subsp. novicida) showed a more complex and longer RD1 sequence (Fig. 2 and 3).

• Open in new tab
• Download powerpoint

FIG. 3.

PCR with primers flanking RD1 resulted in amplicons of different sizes for the different subspecies of F. tularensis. Lanes 1 to 3, strains ATCC 6223, Schu S4, and FSC199 (F. tularensis subsp. tularensis), respectively; lanes 4 to 6, strains FSC147, FSC148, and FSC149 (F. tularensis subsp. mediasiatica), respectively; lanes 7 to 9, strains FSC035, FSC155, and FSC200 (F. tularensis subsp. holarctica, non-Japanese), respectively; lanes 9 to 12, strains FSC021, FSC022, and FSC075 (F. tularensis subsp. holarctica, from Japan), respectively; lane 13, strain FSC040 (F. tularensis subsp. novicida). Molecular sizes (in kilobases) are indicated on the left.

By using the primer pair flanking RD1 for PCR amplification, the four subspecies and, in addition, Japanese strains could be discriminated by their distinct amplicon sizes (Fig. 3). DNA was amplified from all strains tested, and each subspecies showed a unique amplicon size. Moreover, also strains from Japan showed one amplicon of the same size distinct from those of all other strains. In total, 13 strains of F. tularensis subsp. tularensis, 22 strains of F. tularensis subsp. holarctica (non-Japanese origin), 3 strains of F. tularensis subsp. mediasiatica, 1 strain of F. tularensis subsp. novicida, and 4 strains of Japanese origin were tested (data not shown). All 27 strains listed in Table 1 were included in the analysis. The specificity of the PCR assay for F. tularensis was verified by analysis of nine other bacterial pathogens: F. philomiragia, Staphylococcus aureus, Salmonella enterica serovar Typhimurium, Yersinia enterocolitica, Bacillus subtilis, Escherichia coli, Coxiella burnetii, Legionella pneumophila, and Listeria monocytogenes. No visible PCR amplicons were obtained (data not shown).