Although the presence of tandem repeats in the M. genitalium genome has been noted in previous studies 131516, a comprehensive list of tandem repeats with accurate information on the location and repeat unit sequence is not available. We performed a computerized search of the M. genitalium G37 genome 17 and identified 18 loci containing tandem repeat units ranging from 1 to 5 bases in length (Table 1). None of these loci had repeat unit lengths longer than 5 bases (therefore they are all STRs). The repeat copy number for the 18 STR loci varied from 4 to 26. The majority of these STRs (15/18) consisted of a trinucleotide repeat unit. For the remaining three loci repeat units consisted of a mononucleotide, tetranucleotide or pentanucleotide in one each. Nine STRs were located in coding regions while all others fell in non-coding regions. Six of the non-coding region STRs were located in MgPa-related repetitive elements (MgPars) 1718, including MgPar 1, MgPar 2, MgPar 8 and MgPar 9 19. A trinucleotide AGT repeat motif was present in 6 loci, with one in the coding region of MG307, two in the coding region of the adhesin MgPa operon (MG191 or mgpB and MG192 or mgpC genes), and three in MgPar regions (including MgPar 2, MgPar 8 and MgPar 9). While the AGT repeat unit has been previously annotated as TAG repeat by other investigators based on analysis of the G37 genome sequence only 151619, our analyses of multiple M. genitalium strains in the present study and in an earlier study 20 have demonstrated that the correct annotation is AGT. The repeat unit TCT identified in MgPars 1, 8 and 9 in this study has been annotated differently in previous reports 1516, either as CTT or AAG (complementary to CTT). We have found evidence that the TCT repeat region in these three MgPars can recombine with the MG192 gene in which they encode a stretch of serines 20. Thus we believe that the correct repeat unit sequence should be TCT. The STR sequence in MG309 was shown to be a mixture of AGT and AAT trinucleotides in our previous study 13 rather than a hexanucleotide repeat (TATTAC in antisense direction) as previously reported 16. All STR regions except for those in MG192 and MgPars were present in a single copy in the genome.

Previous studies have found that the STRs in MG191, MG192 and MgPars undergo extensive intrastrain variation due to DNA recombination and/or other mechanisms [2021, Ma et al., unpublished data], indicating that these STRs are unsuitable for strain typing. Because our earlier study showed that the STR in the MG309 PLP gene was highly variable among clinical strains 13, we targeted the other three PLP loci (MG185, MG307 and MG338) for further investigation. The remaining 6 STR loci were not evaluated due to limited volume of the DNA extract available from the specimens. Thirty-one unrelated New Orleans specimens were studied initially which revealed significant variation in MG307 and MG338 but none in MG185. Therefore the latter locus was excluded from further study.

The variability of MG307, MG309, and MG338 STRs was determined in a total of 44 unrelated New Orleans patients and 30 unrelated Scandinavian patients (Figure 1). Among the 74 patient samples studied, the number of repeats varied from 4 to 10 for the MG307-STRs, 7 to 17 for the MG309-STRs, and 3 to 17 for the MG338-STRs. Twenty-eight (37.8%) patients contained a single allele at all three STR loci. Sequence analysis revealed a mixture of two or more STR alleles in the three PLP loci as follows: MG307 – 30%, MG309 – 23%, and MG338 – 32%. A total of 46 (62%) of the specimens had mixed STR sequences in at least one of the STR-containing loci. In all mixed sequences, the number of repeats varied by only one or two repeat units. The prevalence of mixed sequences in both MG307-STRs and MG338-STRs significantly increased as the repeat copy number became larger (Spearman's correlation coefficient = 0.87 to 0.95, P < 0.0001) while this correlation was not significant for the MG309-STRs (Spearman's correlation coefficient = 0.26, P > 0.45. See additional file 1). The STR numbers in specimens with mixtures of sequences were determined readily by direct sequencing of the PCR products in the forward and reverse directions, in which individual sequences were read by direct observation of the sequence chromatograms (Figures 2A and 2B).

To verify the presence of mixed sequences as determined by direct sequencing, we performed subcloning of PCR products from 13 selected specimens containing mixed sequences, followed by sequencing of 4–19 individual plasmid clones per specimen. The presence of a mixture of 2 or more sequences as determined by direct sequencing was verified in all 13 specimens (See additional file 2). Plasmid cloning revealed the presence of a third and sometimes a fourth sequence in addition to those identified by direct sequencing in a few cases. In one sample that showed no evidence of a mixed STR population by direct sequencing, the 12 plasmid clones that were sequenced also showed no evidence of sequence mixtures (specimen no. 123.1; see additional file 2). In addition, for 20 selected specimens showing mixed sequences we repeated PCR using DNA polymerases with proof-reading activity and obtained sequences by direct sequencing that were the same as those provided by Taq DNA polymerase. This provided evidence that mixed PLP STR sequences cannot be attributed to PCR artefacts.

Because MG309-STRs contain two different repeat units, AGT and AAT, we examined the distribution patterns of these units among different specimens. Among 57 unrelated patient specimens (including 33 New Orleans patients and 24 Scandinavian patients), which showed no evidence of sequence mixtures, the number of MG309 repeats varied from 7 to 17. Two to four different AGT/AAT distribution patterns were observed in specimens having the same number of STRs with the exception of the five specimens with 9 repeats, which were all identical. When both the repeat number and distribution pattern variations were taken into account, there were 29 unique MG309 sequence types among these 57 patients. Figure 3A shows an example of the variation in the distribution of the AGT and AAT repeat units among patient specimens containing 13 and 14 repeats. The distribution patterns of these repeat units were stable in sequential specimens from the same patients (Figure 3A) and in plasmid clones that had an identical number of repeats and were obtained from the same patients (Figure 3B).

Overall rRNA-SNPs types 2 and 3 are most common (52.7% and 40.5%, respectively). Types 4 (5.4%) and 5 (1.4%) are relatively uncommon. As in our previous study, Type 1 is not present. This type is uniformly found in the currently available ATCC strains but apparently nowhere else. Stratifying the data geographically revealed a significant difference between rRNA-SNP types in New Orleans and Scandinavia (Figure 4A). In New Orleans Type 2 is most common (65.9%) while in Scandinavia 3 is the predominant type (66.7%) (P < 0.0004 by Fisher's exact test). Thus far types 4 and 5 are seen only in New Orleans. Of interest is that none of the other typing loci studied here differ in prevalence between the two locations.

To increase efficiency of rRNA-SNP locus typing we developed a simple and rapid method based on single-strand conformational polymorphism (SSCP) analysis to detect the rRNA-SNPs. As shown in Figure 4B, Types 2, 3 and 4 exhibited different SSCP patterns. Type 2 showed three bands which were well separated. In contrast, types 3 and 4 showed two bands with different mobility. Subsequently, we applied the SSCP assay blindly to all additional specimens used in this study. The typing results using the SSCP assay were completely consistent with the sequencing data for these three genotypes.

This typing method is based on sequence analysis of the polymorphic sites in an approximately 280-bp fragment of the MG191 conserved AB region 10. Previous study of 267 specimens by this method has identified 56 unique genotypes, designated type 1 to 56 10. In this study, we observed 19 and 15 different sequence types in 44 unrelated New Orleans patients and 30 unrelated Scandinavian patients, respectively (Figure 1). When the results for both patient populations were combined, there were a total of 27 unique sequence types. Twenty-three of these were identical to those reported previously by Jensen et al. 10. Among 11 previously unstudied specimens from New Orleans patient four new types (types 57 to 60) were identified, which have been deposited into GenBank with accession numbers EU131379 to EU131382.

Two sequential specimens were obtained from each of 10 New Orleans patients at an interval of 10 to 36 days (Table 2). None of the 10 patients showed a change in MG309-STRs, rRNA-SNPs or MG191-SNPs. Two patients had minor changes in the STR number at MG307 and/or MG338. These changes involved either contraction or expansion of only one repeat unit.

The discriminatory index (DI) for each marker was calculated using the 28 specimens having a single allele at all five loci studied (See additional file 3). Of the five markers evaluated, the MG191-SNPs and the MG309-STRs loci had the highest DIs (0.9392 and 0.9153, respectively), followed by the MG338-STRs and MG307-STRs (0.8730 and 0.7381, respectively). The rRNA-SNPs had the lowest DI (0.5820). When the variation in both the repeat number and the distribution of the AGT and AAT repeat units at the MG309 locus was considered, the DI increased to 0.9471. We analyzed the DI for various combinations of two of the 5 markers and found that the combination of MG309-STRs and MG191-SNPs gave the highest (0.9894). However, some specimens containing identical MG309-STR and MG191-SNP genotypes (including three sets of two specimens and one set of three specimens), still could be distinguished by MG307-STRs, MG338-STRs or rRNA-SNPs (Figure 1).

To study the sexual transmission of M. genitalium infection, we analyzed the genotype profiles at the five variable loci described above in specimens obtained from 31 concurrently infected couples (18 couples from Scandinavia and 13 from New Orleans) in comparison with the 74 unrelated patients. Figure 1 shows the genotype profile at 5 loci for 74 unrelated patients. For each patient, the genotypes at all 5 loci were used to define the multi-locus genotype profile. Patients were classified as having matching profiles if they had identical alleles at all loci or if they had identical alleles at rRNA-SNP and MG191-SNP loci and shared at least one common allele at each of the three variable STR loci when mixed STR alleles were present. Among these 74 patients, only four sets of two patients had matching profiles and all the remaining patients had mismatching profiles. Among the 31 couples studied (Figure 5), matching genotype profiles were identified within 27 couples. Among the four couples without complete matches, one (no. 10) had a mismatch of one repeat unit at a single PLP locus (MG338) while three (nos. 12, 97 and 115) had mismatches at two or three loci. For all couples showing matching genotypes at MG309 STRs, the distribution patterns of the AGT/AAT units were also identical.

In an attempt to increase the repertoire of previously established STR markers 13, we screened the complete genome of the M. genitalium type strain G37 17 for tandem repeats using the Tandem Repeats Finder program 33. The program generates an output file giving the repeat location, the repeat unit length, the copy number, and the nucleotide composition. We used the selection criteria recommended by the program as follows: repeat unit length, 1–100 bp; minimal copy number 11 for single base repeats, 5 for 2-bp repeats, 4 for 3- to 9-bp repeats, 3 for 10- to 100-bp repeats; percent matches, 95–100.

Two sets of M. genitalium positive specimens were used in this study (Figure 1 and Figure 5). The first set was obtained between 1997 and 2002 from patients in Scandinavia, including 18 concurrently infected couples (one female had two male partners) and 30 unrelated patients (including the female from each couple), who had been described previously in a MG191 sequence-based typing study 10. The second set was obtained between 2002 and 2005 from patients in New Orleans, Louisiana, USA, including 13 concurrently infected couples and 44 unrelated patients (including the female from each couple). Thirty-one of the 44 unrelated patients were described along with their rRNA-SNPs and MG309 STRs genotypes in our earlier study 13. For the New Orleans patients, specimens were first voided urine from men and cervical swabs from women. The average time between recovery of specimens from the 31 couples was 7.5 days (ranging from 0 to 90 days), and 74.2% of the specimens were recovered within a week of each other. Genomic DNA was extracted directly from clinical specimens by using the Chelex 100 Resin (Bio-Rad, Hercules, CA, USA) or the High Pure PCR Template Preparation Kit (Roche Applied Science, Indianapolis, IN) as described elsewhere 1231. This study was approved by the relevant ethical committees.

PCR amplification and sequence analysis of the MG309-STRs in 31 of the 44 unrelated patients from New Orleans was reported in an earlier study 13. In the present study we used the same primers and conditions described previously to amplify the MG309-STRs in the new specimens included in this study. To amplify the STRs in MG185, MG307 and MG338, primers were designed based on the sequences flanking the STRs of these three loci by using the Primer3 software 34 (Table 3). To obtain PCR products in sufficient quality and quantity for sequencing, we carried out a primary PCR for all samples and a nested PCR for those samples for which the initial PCR products were not visible or very faint on agarose gels. The primers used to amplify the MG185 STR region were 185F1 and 185R1 for the primary PCR and 185F1 and 185R2 for the nested PCR. The primers used to amplify the MG307 STR region were 307F6 and 307R3 for the primary PCR and 307F5 and 307R4 for the nested PCR. The primers used to amplify the MG338 STR region were 338F1 and 338R5 for the primary PCR and 338F4 and 338R4 for the nested PCR. Amplification was conducted on a PTC-200 programmable thermal controller (MJ Research, Inc., Watertown, MA, USA). PCR conditions were same as previously described for amplifying the MG309-STRs 13, except that the annealing time was reduced to 1 min for all reactions.

Each PCR run included a sample containing DNase-free water as a negative template control. The risk of sample cross-contamination was minimized by following standard PCR procedures. All PCR products were purified by use of the DNA Clean and Concentrator-5 Kit (Zymo Research, Orange, CA, USA) and sequenced directly. In direct sequencing of samples containing two or more trinucleotide repeat populations, the minority populations are three bases out of alignment with the majority population, which can be directly observed on the sequence chromatograms (Figure 2). The sequences flanking the repeat region were determined accurately by sequencing from two directions. To further confirm the presence of heterogeneous repeat populations, some PCR products were sequenced after subcloning by use of the pPCR-Script Amp Cloning Kit (Stratagene, La Jolla, CA, USA). Recombinant plasmids were transformed into SURE competent cells (Stratagene), which permit cloning of unstable DNA. For specimens containing mixed STR alleles, the proportion of individual alleles were arbitrarily estimated by the height of the sequence chromatogram peaks and/or by the cloning results if available.

We previously identified 6 M. genitalium genotypes based on sequence analysis of two fragments of the rRNA operon, including a fragment of approximately 800 bp covering parts of the 16S and 23S rRNA genes and the internal transcribed spacer between them (16S-ITS1-23S) and a fragment of approximately 450 bp covering parts of the 23S and 5S rRNA genes and the ITS between them (23S-ITS2-5S) 13. Types 1–5 were identified by four SNPs located within the 16S-ITS1-23S fragment. Type 6 was differentiated from Type 2 by a single SNP at position 124 in the ITS2 region and was present in only 2/31 patients. We amplified and sequenced the 23S-ITS2-5S region in 10 Scandinavian patients with the type 2 genotype and none of them had the SNP characteristic of Type 6. Therefore, given the low discriminatory value of Type 6, in this study we sequenced only the 16S-ITS1-23S fragment in order to identify Types 1–5. Details of the PCR reaction and sequencing methods have been published 13.

Because we found that rRNA-SNPs typing may be useful for the study of the geographic distribution of M. genitalium, we developed a SSCP method to permit rapid detection of the 3 most common types. The SSCP method was developed using the specimens from the 31 unrelated New Orleans patient isolates for which the 16S-ITS1-23S sequences had been determined previously as noted above 13. SSCP conditions were optimized using the GeneGel SSCP Starter Kit according to the instructions of the manufacturer (Amersham Pharmacia Biotech, San Francisco, CA, USA). We performed nested PCR for all specimens, using primers Mg16-1240F and Mg23-2033R for the first round of amplification and primers Mg16-1381F 13 and ITS1-107R for the second round (Table 3). Four-microliter aliquots of the nested-PCR products (246 bp) were mixed with 4 μl of loading buffer containing 90% formamide, 0.025% (wt/vol) bromophenol blue, 0.025% (wt/vol) xylene cyanol, and 3% (vol/vol) glycerol. This mixture was heated at 95°C for 5 min and then chilled in an ice water bath. Five-microliter mixture was loaded on a precast GeneGel SSCP gel (Amersham Pharmacia Biotech). Electrophoresis was performed using the temperature-controlled Gene-Phor Electrophoresis System and SSCP running buffer C, pH 8.3 (Amersham Pharmacia Biotech). The optimal conditions included a running temperature of 12°C, and a running voltage at 80 V for 20 min followed by 510 V for 140 min. The gels were stained by using the PlusOne DNA Silver Staining Kit (Amersham Pharmacia Biotech).

PCR amplification and sequencing of the MG191 conserved region AB (approximately 280 bp) were performed as described previously 10.

Sequence analysis was done using the Chromas 1.45 (Technelysium Pty Ltd, Tewantin, Australia) and CLC Combined Workbench 1.0.2 (CLC bio, Aarhus, Denmark). Discriminatory index was calculated according to the method described previously 35. To evaluate the relatedness of M. genitalium strains, clustering analysis was performed by use of the BioNumerics 5.0 software (Applied Maths, Sint-Martens-Latem, Belgium), in which heterozygous genotypes and the frequency of each genotype are taken into account. Dendrograms were constructed by the unweighted pair-group method using arithmetic averages (UPGMA) method. All statistic tests were carried out using SAS 9.1 (SAS Institute Inc., Cary, NC). The distribution of M. genitalium genotypes between New Orleans and Scandinavia was analyzed by Fisher's exact test. The correlation between the repeat copy number and prevalence of mixed alleles of PLP STRs was assessed by Spearman's correlation test.

Representative nucleotide sequences obtained in this study have been deposited in the GenBank under accession numbers DQ470670 and DQ470674 for MG307-STRs, DQ470672 and DQ470673 for MG338-STRs, and DQ470671 and DQ470675 for MG185 sequences, and EU131379 to EU131382 for the sequence types 57 to 60 of the MG191 conserved AB region.