In the first mapping stage of CGS, no regions with significantly stronger labeled SS14 DNA signals were observed, indicating no increase in gene copy number in the SS14 genome. Regions giving significantly weaker SS14 signals indicated 1731 candidate regions of variation encompassing 1 or more overlapping oligonucleotide targets. The sequencing data identified 213 SNPs in the SS14 genome. An additional 17 questionable SNPs were suggested in repeated sequences of the genome but did not score well in a SNP uniqueness algorithm 4, and thus could represent false positives due to cross hybridization with the other repeats. DDT sequencing of 12 such regions revealed 5 real SNPs, 6 false positives, and one position with 2 alleles within the SS14 population (intrastrain heterogeneity). Therefore these questionable SNPs were not included in the final sequence, unless they were verified by DDT sequencing (data not shown). An additional 62 positions out of the 213 SNPs identified by CGS were DDT sequenced. 60 SNPs were confirmed (Tables 1, 2, 3 last column) and 2 false positives were found.

1674 out of 1731 candidate regions were identified as SNPs in the second sequencing stage but there were 57 regions encompassing 124 oligonucleotide targets where sequence changes could not be determined. These represented possible hypervariable regions with multiple differences from Nichols in the sequencing 29 mers. DNA regions comprising these sites were grouped into 38 larger regions (29–1507 bp), amplified by PCR and DDT sequenced. In 21 of the 38 cases, mostly closely spaced SNPs and/or short insertions or deletions (indels ranging from 1 to 7 nts) were found while no changes were seen in 17 cases (Table 1, column 7), which is in agreement with data obtained by others 12. DDT sequencing of hypervariable regions suggested by the first phase of CGS identified nucleotide changes in these regions (Table 1, column 7) and also in the vicinity of these regions, where results of the first CGS phase suggested no changes (Table 1, column 8), indicating the need for extension of DDT sequenced regions of at least 100 bp in both directions. Additional short indels were discovered during DDT sequencing of regions identified by WGF (Table 3). Altogether, 2 false positive SNPs (data not shown), 19 false negative SNPs and an additional 16 indels (Tables 1, 3) were found in these DDT sequenced regions (42,344 bp). The overall confirmation of data suggests that repeated regions of the genome are limitations for SNP discovery and almost half of possible hypervariable regions are false positive results.

The accuracy of CGS was determined by comparison to the results of DDT sequencing for 27 regions encompassing 27,141 bp (Table 2). Selection of these regions was focused on possible variable regions in SS14/Nichols hybridization experiments (D. Šmajs, G. M. Weinstock, unpublished results) using a microarray of TPA coding sequences 18 and thus was not completely random. These regions included 5 SNPs and no false positive or false negative SNPs/indels were found. These results indicate an error frequency comparable to or lower than that of high quality finished DDT sequence.

To test the reproducibility of the method, the genome of TPA SS14 was sequenced twice with the CGS approach, using 2 independent DNA isolations from two subsequent inoculations of rabbit testes (i.e. 4950 and 4951, respectively). When most of the variable genomic regions were excluded from the analysis (CGS cannot identify closely spaced SNPs and/or short indels), CGS discovered 198 SNPs in each DNA preparation. The experiments agreed at 192 SNPs (97%), and 12 SNPs were predicted by only one CGS experiment. Out of these 12 SNPs, 7 were found to be real, as shown by DDT sequencing (data not shown), three loci showed intrastrain heterogeneity in one of the two SS14 DNA isolations, with one allele identical to the Nichols genome sequence and a second allele identical to the base change found by CGS. Two SNPs were predicted falsely, and in both cases the false SNP was located next to a real SNP. The reproducibility of the CGS method is thus likely to be limited by the presence of SNP clusters and influenced by genetically different subpopulations in the test sample.

To verify the complete sequence of SS14 strain, to screen for possible discrepancies in cross-reacting repeat regions (tpr genes) and insertions of unique sequences, WGF was performed. This physical mapping approach showed the order of the ORFs along the chromosome is identical to Nichols genome and 4 large indel regions were identified. A 64 bp deletion upstream of TP0136 was found by both CGS and WGF methods. Three additional indels were found only by WGF, two insertions (between genes TP0126–TP0127 and within overlapping genes TP0433–TP0434) and one deletion (in TP0470) (Table 3). A deletion in TP0470 and an insertion in TP0433–TP0434 comprised tandem repeats of 24 and 60 bp, respectively. Similar analysis of the Nichols strain revealed length differences in genes TP0433–TP0434 compared to the published sequence [GenBank:AE000520] as described previously 19. Moreover, intrastrain heterogeneity in the Nichols strain was observed in regions comprising TP0126–TP0127 and upstream of TP0136 with one allele identical to the published sequence. In the Nichols BAC library 20, similar intrastrain heterogeneity was found in the vicinity of gene TP0126. This region comprises a 1255 bp insertion between genes TP0126 and TP0127 in SS14 strain. A similar region was previously described in another syphilitic strain (Chicago, [GenBank:AY587909]) and was found to contain a sequence similar to tprK and is believed to be recipient site of the tprK conversion 21. Altogether, three large indels were not detected by CGS. We suggest probable reasons for this fact are (1) the length of the repeats is similar to/longer than oligonucleotides used on the array and (2) sequence changes were found in Nichols DNA when compared to published complete sequence used for mapping array design [GenBank:AE000520].

When results of CGS, WGF and DDT sequencing were combined, 327 SNPs (224 transitions and 103 transversions), 14 deletions and 18 insertions were identified (Fig. 1). Sequence changes of variable regions V1–V7 of TP0897, tprK, were not included, because sequences of these regions were found to differ greatly in both length and sequence within the SS14 population, in agreement with investigations published previously 222324. Obtained data have been used to compile the sequence of the SS14 genome [GenBank:CP000805]. The GenBank entry contains Ns in the positions of variable regions V1–V7 of tprK gene. All discovered sequence changes are listed in Table S1 (See Additional file 1: Supplemental data).

Interstrain sequence heterogeneity discovered between strains Nichols and SS14 included silent mutations, amino acid alterations/indels, gene fusions, and truncations and elongations of open reading frames due to indels. Among the SNPs found by CGS was an adenine to guanine transition in both copies of 23S rDNA in SS14 strain. This sequence change was previously described in association with the SS14 erythromycin resistance 25. Many discovered indels did not disrupt the open reading frames and represented variable number of nucleotides in homopolymeric tracts (e.g. in TP0012, TP0127), variable number of short motif repeats of 3 and 6 nucleotides (e.g. in TP0136, TP0304, TP0544, TP0668, TP0865), and variable number of longer motif repeats of 60 and 24 nts (TP0433–TP0434, TP0470).

Frameshift mutations and other changes affecting protein length are presented in Table 4. Besides 11 hypothetical proteins (including two possible surface proteins – Tp75 and p83/100), FlaB1 and Tex protein were affected. Sequence changes in four cases led to fusion of ORFs (TP0006 and TP0007 – elongation of Tp75 protein; hypothetical proteins TP0433 and TP0434, TP0597 and TP0598; conserved hypothetical proteins TP0468 and TP0469). Three of these genes (TP0006, TP0470, TP0486) were predicted to code for possible surface protein virulence factors 26. Moreover, antigen p83/100, hypothetical gene TP0127, conserved hypothetical gene TP0470 and the fused proteins TP0433–TP0434 and TP0468–TP0469 were described to be antigenic in rabbits 27. Two of the frameshift changes were confirmed to be present in the Nichols strain genomic DNA (TP0486 and TP0598).

SNPs in SS14 were found to be non-uniformly distributed with the number of SNPs per ORF varying from 0 to 49. Hypervariable regions are listed in Table 5 and include ORFs encoding 3 hypothetical proteins, Tpr proteins (TprC, TprL) and outer membrane protein TP0326. TP0326 was predicted to be a virulence factor 26 and was experimentally verified to be an antigen 27. It is of interest that the most variable region of the genome represents TP0136 (and sequence upstream of this gene) which encodes a protein that is antigenic in both rabbit and human infections 2728 and was found to serve as fibronectin and laminin binding protein 29.

The distribution of SNPs in coding and non-coding sequences of SS14 was not significantly different. ORFs represent 92.9% of total genomic sequence; 94.8% of all SNPs were in coding sequences corresponding to 310 SNPs in genes (212 transitions and 98 transversions) and 17 SNPs (5.2%) in intergenic regions (12 transitions, 5 transversions). The frequency of SNPs was different among putative protein classes (Table 6). The highest frequency of SNPs was in hypothetical genes, lowest in housekeeping genes. In addition, housekeeping genes had the lowest number of SNPs altering amino acid sequences indicating conservation of these gene products.

Because DDT sequencing of some PCR products did not result in an unambiguous sequence, WGS-DDT sequencing of small insert libraries was performed. Analysis of libraries and PCR products revealed multiple (intrastrain) sequence variants in TP0117 (tprC), TP0402 (coding for flagellum-specific ATP synthase), TP0620 (tprI), TP0621 (tprJ), TP0971 (pathogen-specific membrane antigen) and TP1029 (hypothetical protein) and in the intergenic region between tprI and tprJ. Consensus sequences were mostly identical to the Nichols published sequence, but some positions had minor alternative sequences or vice versa. Altogether, intrastrain genetic heterogeneity comprised polymorphisms in 43 nucleotide positions and one polymorphism in a homopolymeric stretch (Table 7).

TPA strains Nichols and SS14 were maintained by rabbit inoculation and purified by Hypaque gradient centrifugation as described previously 40. Chromosomal DNA was prepared as described previously 3.

100 ng of treponemal genomic DNA was amplified to approximately 100 μg using the REPLI-g kit (Qiagen, Valencia, CA). For each array hybridization, 5 μg of amplified genomic DNA was digested with 0.005 U DNase I in 1× One-Phor-All Buffer (Amersham Pharmacia Biotech, Piscataway, NJ) for 5 min at 37°C, followed by inactivation at 95°C for 15 min. To label the digested DNA fragments, 4 μl 5× Terminal Transferase Buffer (Promega, Madison, WI), 1 nmol Biotin-N6-ddATP, and 25 U Terminal Transferase were added directly to the inactivated digestion mix and incubated at 37°C for 90 min, followed by inactivation at 95°C for 15 min.

Mutation mapping microarrays were designed to map mutations by selecting a 29 mer oligonucleotide every 7 bases for both strands of the complete TPA Nichols genome sequence 3, [GenBank:AE000520]. All 325,138 oligonucleotides were synthesized in parallel as described by others 4142.

Arrays were hybridized to digested, labeled genomic DNA of Nichols and SS14 strains separately and processed as described in 4 with an additional step after second wash in stringent buffer consisting of staining with a solution containing Cy3-Streptavidin conjugate (Amersham Pharmacia Biotech) for 10 min, and washing again with non-stringent wash buffer. The Cy3 signal was amplified by secondary labeling of the DNA with biotinylated goat anti-streptavidin (Vector Laboratories, Burlingame, CA). The secondary antibody was washed off with non-stringent wash buffer, and arrays were re-stained with the Cy3-Streptavidin solution.

Finally, the stain solution was removed, and arrays were washed in non-stringent wash buffer followed by two 30 sec washes in 0.5 × SSC and a 15 sec wash in 70% EtOH. Arrays were spun dry in a custom centrifuge and stored until scanning.

Microarray scanning, data analysis and sequencing microarray design and procedure were described previously 4. The second array designed to sequence SS14 strain contained 392,000 oligonucleotides, with 8 oligos per base position (4 for each strand) and 48,600 bases were sequenced in total. Because mutations are sequenced in step two, inclusion of false positives from the mapping arrays does not affect the final data set.

After the second sequencing stage of the array analysis, some regions (Table 1) of the SS14 genome showed clear differences but SNPs could not be clearly identified. These regions were sequenced by DDT sequencing. Coordinates of these regions were extended with at least 150 bp in both directions and amplified with Taq-polymerase using oligonucleotide primers designed with Primer3 software 43. The resulting PCR products were purified using QIAquick PCR purification Kit (Qiagen) and DDT sequenced using the original amplification primers and internal primers where applicable. Due to sequence similarity between tpr (Treponema pallidum repeat) genes, 3 of the heterologous regions (comprising genes TP0620–TP0621, TP0896–TP0898, TP1029–TP1030) were XL PCR amplified using primers annealing to unique regions in the vicinity of the desired sequence. XL PCR products were purified and mechanically sheared to fragments 500 – 1000 bp in length. These fragments were cloned into the pUC18 vector resulting in small insert libraries and recombinant plasmids isolated from at least 48 colonies were DDT sequenced to multiple coverage using pUC18 primers. All sequence reads were analyzed using Lasergene software (DNASTAR, Inc., Madison, WI).

Whole genome fingerprinting was performed as described previously 44. The chromosomal DNA was amplified in 102 Treponema pallidum interval (TPI) regions with median length of 12,204.5 bp (ranging from 1,778 to 24,758 bp) using the GeneAmp^® XL PCR Kit (Applied Biosystems, Foster City, CA). The primer pairs for these amplifications are listed in Table S2 (See additional file 1: Supplemental data). Each PCR product was digested with BamH I, EcoR I and Hind III (New England Biolabs, Ipswich, MA) or their combinations. To asses the possible presence of indels in restriction fragments ≥ 4 kb, additional digestions using Acc I, Cla I, EcoR V, Kpn I, Mlu I, Nco I, Nhe I, Rsr II, Sac I, Spe I, Xba I or Xho I were performed as needed. The resulting fingerprints of TPA Nichols and SS14 strains were compared.

The complete sequence of TPA SS14 strain was deposited in the GenBank under the accession number CP000805.