Gonorrhoea, caused by the bacterium Neisseria gonorrhoeae, remains one of the most common sexually transmitted diseases contributing a substantial burden of morbidity, mortality and infertility worldwide. The disease is treatable and curable, but no vaccine is available. Consequently the control of this important disease depends on the identification and treatment of infected individuals and their contacts in transmission networks. High-resolution and reproducible typing methods for clinical isolates of the gonococcus are therefore central to the control of gonococcal infection. Knowledge of the gonococcal strains circulating both locally and globally, and of temporal changes in the prevalence of these strains, would identify transmission patterns and may assist in prevention and control of this disease.

Many typing schemes have been developed for N. gonorrhoeae but no single typing scheme has been generally adopted, and the lack of a single, generally accepted typing method has impeded the sharing of epidemiological data between laboratories. Auxotyping and serotyping are often applied to gonococci and these techniques are frequently combined, but they do not always provide sufficient resolution to distinguish between epidemiologically related and unrelated isolates [1].

Molecular based typing schemes [2–6] provide better discrimination among isolates. One method, multilocus enzyme electrophoresis (MLEE), which indexes variation in housekeeping genes, has been utilized to characterize gonococci, and has shown that strains with an AHU- (arginine, hypoxanthine and uracil requiring) auxotype are uniform, despite frequent recombination among gonococci [7]. AHU- isolates have been linked to disseminated gonococcal infection (DGI) [8], which is related to the penicillin sensitive phenotype usually found among these isolates [9].

Methods that use nucleotide sequencing, however, [10–13], are more portable, have greater definition, and make data storage in globally accessible databases via the internet easier. One method, based on the nucleotide sequence fragments from two gonococcal antigen genes under diversifying immune selection: por and tbpB (N. gonorrhoeae multi antigen sequence typing, NG-MAST) [14, 15], provides a high level of discrimination. However the NG-MAST database only includes genotypes, consisting of two number allelic profiles and the nucleotide sequences, with no isolate data available.

One established method for the characterization of bacteria is multilocus sequence typing (MLST), a development of MLEE, and a highly discriminatory system for indexing the relatedness among isolates based on genetic variation present in genes under stabilising selection for conservation of metabolic function [16]. It is employed for the characterisation of many bacterial species, including the closely related pathogen Neisseria meningitidis and the commensal Neisseria lactamica [16–21].

An intriguing feature of gonococcal biology is the very close relationship of this bacterium to N. meningitidis and N. lactamica, which also have an obligate association with humans, but inhabit the mucosal surface of the nasopharynx rather than the urogenital tract. Application of the same MLST scheme to N. gonorrhoeae, is therefore advantageous as it can be used to analyse genetic relationships among gonococcal isolates, as well as among the neisseriae [22]. Another advantage of MLST is its ability to discriminate among species, facilitating species identification and the detection of mixed bacterial cultures. This paper describes a N. gonorrhoeae typing scheme that exploits the existing globally accessible Neisseria MLST database [23, 24], which provides publicly available isolate information as well as nucleotide sequence data.

It has been suggested that MLST of the pathogen N. gonorrhoeae would not provide sufficient discrimination between strains [27], due to the uniformity of its housekeeping genes [28]. The present study has shown that N. gonorrhoeae can be typed effectively using the same MLST scheme employed to characterize N. meningitidis [16] and N. lactamica [20], with a genotypic diversity comparable to that found among meningococcal isolates [25]. Despite high levels of horizontal genetic exchange among gonococci [2], MLST is robust because it is based on data from seven genetic loci distributed around the chromosome and indexes variation that is subject to stabilising selection. It appears to provide a level of discrimination comparable to the NG-MAST typing scheme [15], although this has not been formally validated as different datasets have been used. MLST, however, has the advantage that isolate information is available alongside genotypic data in an established, publicly accessible database [23, 24]. Unlike schemes that rely on antigen gene variation [5, 15, 29], which is subject to diversifying immune selection, MLST data can also be used to examine the evolutionary relationships among strains.

A total of 149 gonococcal isolates were typed by MLST in the present study. While only 27 polymorphisms were detected among the seven loci, a total of 66 unique allele combinations, or STs, were recorded. The low level of nucleotide diversity among gonococci inevitably results in a tighter clustering of these isolates in phylogenetic trees compared to meningococcal and N. lactamica isolates when concatenated sequences are analysed. However, the use of allelic profiles demonstrates a comparable level of discriminatory power to MLST of N. meningitidis and N. lactamica. The gonococcal STs were well differentiated with some showing temporal and geographic persistence. For instance, isolates with ST-1596 and ST-1583 were first isolated in Liverpool during the 1980s and have since been isolated in 2000, suggesting they may have a fitness advantage that has enabled them to persist in the population for over a decade. A total of 31 STs were represented by more than one isolate, with one group, ST-1579 represented by 10 isolates from three different countries, suggesting that isolates were distributed widely and not structured geographically. No temporal structuring was evident either, as the isolates collected in Liverpool between 1981–1989 were not significantly different from those collected in the same location between 2000 and 2001 (p > 0.05). While there was no evidence of geographical or temporal structuring in the gonococcal populations, UK isolates predominated in this study and some were from undefined locations, which may have influenced the outcome of the analysis. Structuring may be evident if more geographically and temporally diverse isolate collections were examined by MLST.

MLST provides a useful tool to study both the local and global distribution of isolates such as those with the AHU- phenotype, making it possible to track particular variants and examine transmission patterns. Of the 13 AHU- gonococci in the present study, 12 had identical genotypes with one differing by a single synonymous mutation at one locus. This illustrates the close relationship of this group and the ability of MLST to differentiate isolates with this auxotype. Further validation of the method would be required before MLST was used to resolve questions related to an outbreak situation and it may be necessary to complement it with antigen gene sequencing, as used in meningococcal epidemiology [30].

As the gonococcal MLST scheme uses nucleotide sequence data from exactly the same gene fragments as the meningococcal scheme, it can be used to compare MLST data from different Neisseria species, allowing phylogenetic and population genetic inferences to be made. The gonococcal MLST data were compared to data from studies of N. meningitidis and N. lactamica isolate collections previously published by the authors [16, 20, 25]. The use of these data, as opposed to the entire Neisseria MLST database was preferred as they had been extensively characterized and their provenance could be confirmed.

Like the N. lactamica and N. meningitidis alleles, the d _N /d _S ratios of the gonococcal alleles suggest that these loci evolve slowly and are not affected by diversifying selection, making them suitable for analysing evolutionary relationships among these species. However, this ratio could be affected by the small number of polymorphisms present within the collection. The MLST data for the three individual species were examined using: (1) the allelic profiles, (2) the individual alleles at each locus, and (3) the concatenated sequences for each allelic profile. When the MLST profiles were compared, the STs were unique to each species. The allele sequences were also species specific and no alleles were common among the neisseriae when 149 gonococcal, 324 meningococcal, and 103 N. lactamica isolates were examined. When additional N. lactamica isolates from the carriage study [20] and from German and Czech collections (unpublished data, not shown) were included in the analysis, only two alleles were common to more than one species. These alleles, at the pgm locus, were present among both N. gonorrhoeae and N. lactamica isolates. Alleles at this locus are among the least variable in gonococci. Thus, it seems more probable that these common alleles are a result of shared ancestry rather than interspecies recombination.

One of the advantages of a common MLST scheme for the neisseriae is that it can be used to distinguish between the Neisseria species and to identify unknown or misclassified isolates. Both neighbour-joining and split decomposition methods, using the concatenated MLST data, clustered the isolates into three distinct groups. The clustering of the STs into groups suggests that minimal recombination occurs among the housekeeping genes of these three Neisseria. This is confirmed by the F _ST analysis, which suggests low levels of recombination among the species, the high number of fixed differences, the low number of shared polymorphisms, and the lack of alleles shared among species. Interestingly, the split decomposition analysis clustered the gonococcal sequences within the diversity of the meningococcus, reflecting the close ancestral relationship between these bacteria [31].

Although genetic recombination has been reported among N. gonorrhoeae, N. lactamica and N. meningitidis [32, 33], the physical and temporal separation of these species within the human host is likely to contribute to a low frequency of interspecies recombination. N. gonorrhoeae, which colonises the urogenital tract, is rarely found in children as it is sexually transmitted and is only occasionally found in adult throats; N. meningitidis is carried in the throats of approximately 10% of the adult population [34] but is rarely carried by young children and found infrequently in the urogenital tract; N. lactamica is carried by only about 2% of adults [35] but is highly prevalent in young children with carriage rates of around 40 % [36, 37]. While this limits opportunity for interspecies recombination, it does not affect intraspecies recombination, which may occur frequently creating an increasing number of STs from the available pool of alleles for each species, as has been observed in meningococci [25].

The results of the present study are inconsistent with a previous report that the Neisseria housekeeping alleles used in MLST were widely distributed among the neisseriae due to frequent interspecies recombination [38]. This was an in silico study that compared sets of 500 meningococcal STs downloaded from the Neisseria MLST database with all STs assigned to other named Neisseria species in the same database. The data that were analysed were not verified experimentally and access to the original samples was not requested. The present study did not include any of the apparently hybrid STs as they did not form part of the coherent populations analysed. We investigated these apparent hybrid STs present in the database for which samples were available. In all cases these were STs generated from historical freeze-dried cultures from which it was impossible to grow live organisms. Further analysis of the DNA samples suggested these were from mixed cultures and that the hybrid STs were a consequence of differential amplification of some loci. In conclusion, the present study finds little experimental support for extensive interspecies recombination among housekeeping genes in the Neisseria.

The lack of congruence among all of the phylogenetic trees may be a consequence of either shared ancestry or infrequent genetic exchange among the species. The relatively short lengths of the individual sequences used would also reduce any phylogenetic signal and therefore concatenated sequences were used to improve resolution. Only the tree for the alleles at the gdh locus produced a tree congruent with that obtained from the concatenated nucleotide sequences. The gdh locus may have evolved more rapidly than the other loci as these species diverged away from the ancestral population, creating gdh alleles that appear highly distinct for each species as shown in the gdh gene tree.

The species specificity of the gdh alleles and the congruence of the gdh gene tree with that produced from the concatenated sequences suggest that analysing sequences at this locus alone may be useful in differentiating among these three species and might help identify misclassified isolates. Occasionally Neisseria are misidentified [39, 40], therefore a typing tool that can be exploited to differentiate species using either MLST profiles, allele sequences at particular loci or the concatenated gene sequences, could prove extremely helpful alongside traditional microbiological methods. This is especially important if commensal species are misidentified as N. gonorrhoeae, which could lead to serious social, legal and medical consequences [41].

Although a number of other commensal neisseriae have been typed (unpublished results from the MLST database [23, 24]), these were not included in this study as too few isolates of these species have been typed for robust, meaningful analyses. MLST of representative collections of these other commensals, in particular Neisseria polysaccharea and Neisseria cinerea, which are closely related to the pathogenic Neisseria [42] would be advantageous, as knowledge of the genotypes of these species could be applied to species definitions and could facilitate identification of misclassified isolates.

This analysis has shown that MLST can be used effectively to characterise N. gonorrhoeae collections, obtained both locally and globally, and has demonstrated a level of discrimination that appears comparable to that determined for the meningococcus using MLST [16, 25] and the gonococcus using the NG-MAST scheme [15]. As an identical scheme has been used to characterize both N. meningitidis and N. lactamica, these data can be exploited to help define the three species, using either STs, individual alleles, in particular those at the gdh locus, or by concatenating the MLST data.