Article,
Distribution and genetic variability of three vaccine components in a panel of strains representative of the diversity of serogroup B meningococcus
Vaccine, 27 (21): 2794--2803 (May 2009)
DOI: 10.1016/j.vaccine.2009.02.098
Abstract
aNovartis V&D, via Fiorentina 1, 53100 Siena, Italy bÖrebro University Hospital, SE-701 85 Örebro, Sweden With the aim of studying the molecular diversity of the antigens of a new recombinant vaccine against meningococcus serogroup B, the three genes coding for the main vaccine components GNA (Genome-derived Neisseria Antigen) 1870 (fHbp, factor H Binding Protein), GNA1994 (NadA, Neisseria adhesin A) and GNA2132 were sequenced in a panel of 85 strains collected worldwide and selected as representative of the serogroup B meningococcal diversity. No correlations were found between vaccine antigen variability and serogroup, geographic area and year of isolation. Although a relevant clustering was found with MLST clonal complexes, each showing an almost specific antigen variant repertoire, the prediction of the antigen assortment was not possible on the basis of MLST alone. Therefore, classification of meningococcus on the basis of MLST only is not sufficient to predict vaccine antigens diversity. Sequencing each gene in the different strains will be important to evaluate antigen conservation and assortment and to allow a future prediction of potential vaccine coverage. Keywords: Meningococcus serogroup B; Gene sequencing; Antigen variability; Antigen diversity In the past, most pathogenic strains were classified by Multi Locus Enzyme Electrophoresis (MLEE) into several hypervirulent lineages: Electrophoretic Types ET-37, ET-5, cluster A4, lineage 3, subgroups I, III, and IV-1 2. Multi Locus Sequence Typing (MLST), an approach based on the comparison of the nucleotide sequences of seven housekeeping genes 3 and 4, is now the gold standard for meningococcal typing. This method classifies the above lineages into complexes ST-11, ST-32, ST-8, ST-41/44, ST-1, ST-5, ST-4, respectively 5. Effective protein–polysaccharide conjugate vaccines are available against serogroups A, C, W-135 and Y. However, no capsular vaccine exists against serogroup B meningococcus, whose polysaccharide is an α (2–8) linked polysialic acid, chemically identical to the polysialic acid found in human glycoproteins. As a consequence, the MenB (meningococcus B) polysaccharide is poorly immunogenic since recognized as self 6. In the absence of a polysaccharide-based vaccine, many attempts have been followed to develop protein-based vaccines. The most successful vaccines developed so far are composed of Outer Membrane Vesicles (OMV) obtained by detergent extraction from whole bacteria. These vaccines are highly effective but they have been proven to be useful against clonal outbreaks only. As an example, an OMV vaccine has been successfully used to control a MenB epidemic in New Zealand 7 and 8. However, Porin antigen A (PorA) is the main immunogen 9 of the OMV-based vaccines and they are effective only against strains containing the same PorA as the OMV used for immunization. An analysis of the strain diversity in the USA showed that a OMV-based vaccine for this country was not feasible, because at least 20 OMV, with different PorA, had to be pooled in order to cover 80\% of the strains 10. Recently, the 5 Component Vaccine against Meningococcus B (5CVMB), a new vaccine based on novel antigens identified by reverse vaccinology, was described 11. The vaccine is composed of five proteins, GNA1870 (or fHbp), 1994 (or NadA), 2132, 1030 and 2091, discovered by mining the genome of a serogroup B isolate. When used to immunize mice with Freund adjuvant, it evoked a bactericidal response against 97\% of the strains included in a panel of 85 representative of meningococcal diversity. When the vaccine was used to immunize mice with less potent adjuvants that are suitable for human use, such as alum or MF59, it raised protection against 78\% and 94\% of the strains, respectively. FHbp, NadA and GNA2132 were the antigens that induced bactericidal activity against most of the meningococcal strains tested. FHbp is a surface-exposed lipoprotein which induces high levels of bactericidal antibodies which are also protective in vivo, following passive immunization 12. Three genetic, and antigenically poorly cross-reactive, variants (1–3) of this protein have been described. The 5CVMB vaccine contains variant 1 antigen, which is the most abundant in MenB strains. This protein was named fH-binding protein (fHbp), because it binds human factor H, a key regulator of the alternative complement pathway 13. The adhesin NadA 14 belongs to the Oligomeric Coiled-coil Adhesin (OCA) class of trimeric proteins 15. NadA induces a strong bactericidal response, and is protective by passive immunization in the infant rat model. The nadA gene is present in approximately 50\% of pathogenic MenB strains 14. More specifically, nadA is present in almost all strains belonging to the hypervirulent clonal complexes ST-32, ST-11 and ST-8, whereas it is always absent in the ST-41/44 clonal complex, in Neisseria gonorrhoeae and in the commensal species Neisseria lactamica and Neisseria cinerea. Three well-conserved forms (NadA-1, NadA-2 and NadA-3) have been described. Antibodies against each NadA form are equally bactericidal against other forms of the protein. Moreover, the nadA gene segregates differently in strains isolated from healthy individuals and from patients. A fourth form of the protein, NadA-4, was described to be associated only to carrier strains 16. GNA2132 is a lipoprotein that induces bactericidal antibodies in mice, and is protective by passive immunization in the infant rat model. Protection in the infant rat was observed also in the absence of complement-mediated bactericidal activity 17. Despite the sequence diversity observed, the induced bactericidal antibodies are cross-protective against most strains tested. In this work we studied the gene distribution of fHbp, nadA and gna2132 in a previously described panel of pathogenic strains, selected as representative of the serogroup B meningococcal diversity in terms of genetic features, geographic area and year of isolation 11. We have been looking for correlations between the diversity of the vaccine components and standard typing methods. In particular, we analysed the antigen variant distribution in the meningococcal clonal complexes, in order to evaluate the possibility of a prediction of the vaccine component assortment, based on MLST. In order to represent the worldwide epidemiologic situation, invasive isolates were collected in collaboration with the Centers for Disease Control and Prevention (CDC) in Atlanta USA, and the main meningococcal national surveillance centers in Europe (i.e. HPA in Manchester, UK; NIPH in Oslo, Norway; Institut Pasteur in Paris, France; Istituto Superiore di Sanità in Rome, Italy; Instituto de Salud Carlos III in Madrid, Spain; Institut für Hygiene und Mikrobiologie in Würzburg, Germany; Reference Laboratory for Bacterial Meningitis in Amsterdam, the Netherlands; National Public Health Institute in Helsinki, Finland; National Reference Laboratory for Pathogenic Neisseria in Orebro, Sweden). Strains were also collected from other regions of the world (Canada, New Zealand, Australia, and South America). Africa and Asia, where mostly serogroup A strains are endemic, were the less represented countries. All strains were serologically classified on the basis of serogroup and the PorB/PorA sero-subserotypes. Strains were genetically typed by MLST according to procedures available at: http://pubmlst.org/neisseria/ 5. 85 pathogenic strains were selected as representative and used for following studies. These strains originated from 16 different countries, USA and Europe being the most represented geographic areas, and were isolated starting from the 1960s. 20 strains were isolated after 2000. The panel also included eight serogroup C-associated strains and one strain for each serogroup A, W-135 and Y. Genomic DNA was prepared by culturing bacteria overnight at 37 °C in atmosphere humidified with 5\% CO2 in GC (Gonococcus) agar (Difco) supplemented with Kellog's solution (0.22 M d-glucose, 0.03 M l-glutamine, 0.001 M ferric nitrate and 0.02 M cocarboxylase) from Sigma. Each strain was grown until stationary phase in 5 ml of GC medium. 1.5 ml of the culture was centrifuged at 16,060 × g for 15 min, and chromosomal DNA was prepared using the NucleoSpin Tissue Kit (Macherey-Nagel) according to the manufacturer's instructions. DNA concentration was calculated by optical density determination at 260 nm. About 10 ng of chromosomal DNA was used as template for the amplification of all genes. The amplification enzymes used were High Fidelity Taq DNA Polymerase (Invitrogen), in the case of fHbp and gna2132, and AccuPrime Taq DNA Polymerase System (Invitrogen) for nadA. All genes were amplified using primers external to the coding region. Primer sequences are reported in Table 1. fHbp gene was amplified using primers A1 and B2. PCR conditions were: 30 cycles of denaturation at 94 °C for 30 s, annealing at 58 °C for 30 s, elongation at 68 °C for 1 min. Sequences were performed using forward primers A1, 22, and reverse primers B2, 32. nadA PCR primers were A and B. PCR conditions were: 94 °C for 30 s, 56 °C for 30 s, 68 °C for 1 min, 30 cycles. Forward sequencing primers were A, 1, 2, 3, reverse primers were B, 4, 5, 6. gna2132 gene was amplified using primers 1 and 6. PCR conditions were: 94 °C for 30 s, 55 °C for 30 s, 68 °C for 1 min, 30 cycles. Forward sequencing primers were 1, 22, 3, 4, 5, reverse primers were 6, 7, 82, 92. PCR products were purified with QIAquik PCR purification kit (QIAGEN) and sequenced using the ABI 377 automatic sequencer (Applied Biosystems). Sequences were assembled and analysed using the Vector NTI Suite 9 and the GCG Wisconsin package. Nucleotide sequences were aligned using Clustalw v1.83. The number of mutation sites and their statistics were evaluated with DNASP v4.10 18 and MEGA v3.1 19. Mutation rates were evaluated measuring the nucleotide diversity per site (π) and the Watterson estimator (theta) 20. The minimum number of recombination events 21 (R) per locus and the corresponding recombination rate between adjacent sites (r = R/L, where L is the average nucleotide distance between the most distant sites under recombination) were measured with DNASP. Recombination rate per site has also been simulated using the programs CONVERT and INTERVAL from the LDHAT package 22 and 23. Tajima D statistical test 24 was applied to evaluate the presence of evolutionary forces deviating each antigen from neutrality. An average value for D was estimated as well as using a sliding window of 100 bp for each locus. Non-synonymous/synonymous mutation rate ratios (dN/dS) were measured in average, and in a 100 bp sliding window for each locus, using DNASP. dN/dS = 1 is the ratio indicating neutral selection, dN/dS \textless 1 means purifying selection, whereas dN/dS \textgreater 1 indicates a gene diversification under positive selection. Phylogenetic trees of the 7 MLST concatenated genes and of each antigen gene were generated using MEGA. In detail, Maximum Parsimony (MP) 20 phylogeny were reconstructed using the Close-Neighbour-Interchange (CNI) algorithm (search level parameter = 3), which is a branch swapping method that begins with a given initial tree. The initial tree was produced by adding a randomly selected sequence to the growing tree on a randomly selected branch (the number of repetitions were 100). MP reconstructions were compared with Unweighted Pair Group Method with Arithmetic Mean (UPGMA) reconstructions, which is a distance-based method. Sequence bootstrap method with 500 repetitions was applied in order to test the branch stability and tree topology structure. From multiple nucleotide alignments of the antigen genes and the phylogenetic reconstructions the variants were grouped into main variants, each corresponding to major clades. Multiple r × c contingency tables of main variants (r) versus clonal complexes, serogroups, country of origin and year of isolation (c) were analysed using R v2.4.0 and the vcd package v1.0–3 for the analysis of categorical data. Pearson chi-squared p-values and Cramer's V coefficients were reported to test the null hypothesis of independence and to measure the strength of the observed correlations 25. The same approach has been applied to test the association between the variants of each antigen and other tracers like serogroup, geographical provenience and year of isolation. Nucleotide sequences of antigen genes were classified on the basis of the main clades, and by assignment of a progressive numerical identifier to each allelic variant. A Multi Locus Sequence Typing like procedure was followed. Each allelic variant was indicated in the form gene-allele (i.e. fHbp-1 indicates allelic variant 1 of the gene fHbp). The haplotypic diversity (H) of each gene, and the standardized index of association (ISA) of the observed allelic combinations were measured using LIAN v3.5 26. This was done in order to quantify the level of allelic diversity and the presence of linkage disequilibrium, respectively. In detail, the null hypothesis of linkage equilibrium was tested with LIAN by a Monte Carlo simulation (104 iterations) comparing a free recombining sample against the observed allelic combinations. Similar to gene classification, each protein main variant was classified on the basis of the main clades. Each sub-variant is named with a first number indicating the main variant and a progressive number after the dot (i.e. fHbp-1.2 indicates variant 1, sub-variant 2, of fHbp). The level of linkage between the protein variants of each antigen and clonal complexes was evaluated with Fisher's exact statistics on 2 × 2 contingency tables. The study was performed starting from a worldwide collection of disease-related strains set up in collaboration with several meningococcal surveillance laboratories (in USA and Europe). A sub-collection of 85 strains was selected on the basis of, first, genetic diversity on the basis of MLST. Almost all clonal complexes mainly described to be associated with disease were included in the panel. Each clonal complex was represented by a number of strains proportional to the worldwide importance and diversity of the complex. All strain belonging to the same clonal complex were isolated in different countries, and/or at different times. Fifty-six of 85 (66\%) strains belonged to the four hypervirulent clusters ST-8, ST-11, ST-32 and ST-41/44, which are responsible for epidemic meningococcal disease worldwide. Among ST-11 strains, eight belonged to the ET-15 clone, that was first detected in Canada and described to be more virulent than other members of the same clonal complex as associated with severe clinical infections and high mortality rates in several countries 27. Twenty-nine of 85 (34\%) strains represented the recently emerged ST-213, ST-269 clonal complexes and the rather widespread complexes ST-60, ST-18, ST-23, ST-35, ST-162, ST-37, ST-103, ST-167, ST-231, ST-1 and ST-5. Also some STs not yet assigned to any clonal complex were included. When feasible, the most represented PorA/B serosubtypes and serotypes were included. In general, all strain of the panel represented different meningococcal emergences, be those epidemics, endemic situations and local outbreaks. The overall composition of the strain panel is represented in Table 2. Full-length nucleotide sequence of the fHbp and gna2132 genes was determined in all 85 strains. nadA nucleotide sequence was determined for the 39 strains harbouring the gene. The three genes were found to have an overall nucleotide conservation of 86, 92 and 95\%, respectively (nucleotide diversity value π = 0.1404 ± 0.0076; π = 0.0824 ± 0.0047; and π = 0.0499 ± 0.0027) (Table 3). There were 34 alleles at the fHbp locus, 13 at the nadA locus and 31 at the gna2132 locus (Table 3). The reduced number of alleles for all antigens suggests a non-randomly structured population and the presence of linkage disequilibrium (Fig. 1). The structured state of the population was confirmed by the allelic diversity analysis H, which was 0.947, 0.895, 0.768 for fHbp, nadA and gna2132, respectively (Table 3). The diversity value for the 7 MLST gene fragments was similar (H = 0.859 ± 0.012). A further confirmation of the population structure was provided by the standardized index of association (ISA), which is expected to be 0 in a situation of free recombination. This was found to be significantly different from 0, being 0.210 for the three genes, 0.535 (p-value \textless 10−4) for the 7 MLST genes and 0.395 (p-value \textless 10−4) for the combination of the MLST and antigen genes. The analysis of the recombination rate showed that recombination parameter estimates were R = 2.9 (L = 825 bp), 0.001 (L = 1154 bp) and 15.7 (L = 1390 bp) for fHbp, nadA and gna2132, respectively. The recombination rate per adjacent nucleotide (r), which is computed by dividing R by the average nucleotide length of the loci subjected to recombination, indicates the probability of changing a nucleotide by recombination and was found to be r = 0.0035, \textless10−5 and 0.0113, for fHbp, nadA and gna2132, respectively. These minimal recombination rates were also compared with the values computed by Monte Carlo simulations, performed by the program LDHAT, which were rs = 0.0308 (c.i. 0.016–0.047) for fHbp, 0.0159 (c.i. 0.014–0.022) for nadA and 0.028 (confidence interval 0.021–0.037) for gna2132, respectively. Overall, the observed and simulated minimal recombination rates were relatively low if compared to the mutation rate (π). Non-synonymous versus synonymous polymorphisms average ratio (dN/dS) (Table 3) were 0.51 ± 0.07 for fHbp, 0.88 ± 0.14 for nadA and 0.57 ± 0.08 for gna2132. nadA is close to the value of 1, which indicates neutral selection. The other two genes appear to be under purifying selective pressure, as they have an average number of synonymous nucleotide polymorphisms higher than the number of non-synonymous nucleotide mutations (dN/dS \textless 1). This indicates sequence conservation, and suggests that mutations at important functional sites are not favoured by natural selection. However, as evidenced for example on the variable regions of the PorA/B molecules 28, 29, 30 and 31, selection could be particularly strong only on limited portions of the antigens favouring non-synonymous mutations (dN/dS \textgreater 1) and these molecular evidences should be further investigated. Dendrograms were built using both MLST genes (Fig. 2A) and each of the three vaccine genes (Fig. 2B–E). As shown in Fig. 2A, the strains that belong to the same clonal complex usually clustered together. ST-11, ST-41/44, ST-8 and ST-32 clonal complexes represent the major clusters of the tree. The phylogenetic tree obtained using fHbp identifies the three main variants previously described, named 1, 2 and 3 (Fig. 2B and C). nadA clusters in four main variants, three of which previously described as associated to invasive strains, 1, 2, and 3 and a new variant, named variant 5 (Fig. 2D). This variant, present in two ST-213 strains, has not been identified previously. Variant 4, described to be carrier-associated, was not found in this panel of strains, consistent with the absence of carrier strains in the panel used. The dendrogram of gna2132 (Fig. 2E) shows clustering but not obvious subdivision into a small number of main variants. Each of the main branches of the dendrograms, which are 3 for fHbp, 4 for nadA and 14 for gna2132, were assigned to a group, which was tested for association with clonal complexes, serogroups, country of origin and year of isolation. Both the association probability test (Pearson p-value) and Cramer's V index, which measure the strength of correlations, indicate an absence of correlation between antigen main variants and serogroups, country of origin and year of isolation. In other words, there was not predominance of a given main variant between different serogroups and countries during the study period (since 1960s). Conversely, the values indicate a strong association between antigen alleles and clonal complexes (Table 3). In agreement with the phylogenetic trees in Fig. 2B–E, fHbp, NadA and GNA2132 were classified into three, five and fourteen main variant families, respectively. In the case of nadA, the gene missing was included as variant 0. Each variant was then subdivided into peptidic sub-variants on the basis of the amino acid substitutions and named with progressive numbers. fHbp variant 1 includes sub-variants 1.1–1.15, fHbp variant 2, sub-variants 2.1–2.12 and fHbp variant 3, sub-variants 3.1–3.4. NadA is represented by 5 variants (0, corresponding to the absence of gene, and 1, 2, 3, 5) and includes 12 sub-variants. GNA2132 contains 14 variants and 29 sub-variants (Table 4). The comparison of variants and sub-variants of fHbp, which, unlike variants of proteins NadA and GNA2132, were described to be poorly cross-protective 12, is shown in Fig. 3. The amino acid sequence of MC58 is considered as the reference sequence (top line) for variant 1 and named 1.1 in Table 4. The residues shown in the sequence alignment in Fig. 3 are only those which change in any variant or sub-variant analysed. As referred in the previous paragraph, the presence of an association between protein variants and clonal complexes implies that variants are not randomly distributed, but tend to associate with the clonal complexes. As shown in Table 4, in most cases, each of the sequence variants is associated with only one clonal complex. A quite low number of sub-variants is shared by more than one clonal complex. For instance, in the case of protein fHbp sub-variant 1.1 was present in all ST-32 complex strains only, whereas sub-variants 2.1 and 2.8 were mainly associated with ST-8 complex strains. Sub-variants 1.4, 1.14, 2.4, 2.5, 2.9 and 3.1 were associated with ST-41/44 complex. Sub-variants 1.2, 1.3, 1.6, 1.9, 1.10, 1.11, 2.3, 2.7, 2.12, 3.2 and 3.4 were associated with ST-11 complex. Some sub-variants were associated with strains belonging to clonal complexes classified in Table 4 as “other” and including ST-213, ST-269, ST-60, etc. A similar non-random association was found in the case of NadA. Here ST-32 was still the most homogeneous complex harbouring variant 1.1 only (with the only exception of one strain that did not have the gene). ST-8 complex carried sub-variants 2.2, 3.1 and 3.2 while complex ST-11 was still quite heterogeneous, being associated with seven different sub-variants. The absence of the gene (variant 0) was totally associated with complex ST-41/44, as already described 14. Similar findings were also seen for GNA2132. In this case, ST-32 complex was associated with 3 sequence sub-variants, ST-8 complex with 2, ST-11 complex and ST-41/44 complex with 5, while other strains were associated with other sub-variants. However, as shown in Table 4, the correlation between protein variants and clonal complexes was not always complete, and each clonal complex could share more than one combination of the three antigen variants. In the present study, the diversity of the loci corresponding to the three main vaccine components of 5CVMB was determined in a panel of 85 meningococcal strains. The diversification of these proteins was compared between strains belonging to different clonal complexes and strains belonging to the same clonal complex. The observed number of alleles for each of the three antigens was significantly lower than the number of strains. This indicates that not all of the alleles predicted to be generated by random association were actually selected and found in nature, and that there is a non-randomly structured population. A phenomenon of linkage disequilibrium has driven the spread of allelic variants and this is in agreement with what already described for other meningococcal antigens, like PorA, PorB, FetA 32, Opa 33 and 34. The allele variants of the three genes generated a number of combinations comparable to the one of the MLST genes of the same strains. In spite of the different features of the vaccine antigens which are known to be surface-exposed and the MLST genes which encode for housekeeping proteins, the results suggest a similar population structure. The analysis of the phylogenetic trees built for each of the three genes (Fig. 2B–E) usually showed a complexity that was lower than that of the MLST tree. Furthermore, there was a good clustering between antigen variants and clonal complexes, although the correlation may not be complete. From the point of view of the vaccine components, ST-32 is the most homogenous clonal complex, showing almost complete clustering in all dendrograms. The clustering is also present in the case of ST-41/44 strains although more than one variant is harboured by this clonal complex. ST-11 is the clonal complex showing the most variability. It is interesting to note that the diversification described so far for other proteins in this same clonal complex is quite different. For instance, the ST-32 complex shows a more heterogeneous repertoire as for PorA/PorB/FetA, whereas ST-11 complex normally shows a limited PorA/PorB repertoire. The highly significant correlation between clonal complexes and antigen variants, regardless of geography and year of isolation, suggest that the vaccine component repertoire is highly structured within each hyperinvasive clonal complex, and that this is true overtime and independent by the country of isolation. Moreover, the absence of non-random associations with geography and year of isolation confirms that the strain panel can be considered as representative of meningococcal diversity. Although a substantial clustering between each protein variant and clonal complexes has been found, the combination of antigen variants in each clonal complex is not always predictable. For instance, while strains belonging to the ST-32 clonal complex have one variant only of fHbp, strains belonging to the ST-11 complex have 11 forms of the same antigen. In the case of GNA2132, ST-32 complex strains have three variants, whereas ST-11 has five variants. Even if the present work focuses mainly on the molecular aspects of the 5CVMB vaccine components without considering functional data, it becomes clear that the presence of such a wide repertoire of sub-variant combinations, also within the same clonal complex, could be responsible for different bactericidal titers. As reported in a previous work with regards to all 85 strains of the same panel 11, the protection induced in mice by 5CVMB vaccine formulated with aluminium was complete (100\%) against strains belonging to ST-32 and ST-8 clonal complexes, almost complete (95\%) against the complex ST-41/44 and substantial against strains of the ST-11 complex (65\%). These different percentages could be explained partially by the presence of different combinations of the sub-variants of the three antigens in strains belonging to the same clonal complex. In fact, ST-32 and ST-8 complexes, which were associated with one or few variants of each vaccine component, were those against which the protection was complete. On the contrary, ST-11, showing the most variable repertoire of antigen sub-variants, was the complex with the lowest percentage of strains killed by bactericidal assay. In order to predict the potential coverage of the 5CVMB vaccine in a given strain panel, it is necessary to know the repertoire of the three vaccine components, as different antigen combinations could induce different bactericidal titers. The antigenic repertoire is not predictable on the basis of MLST alone. Therefore additional effort in sequencing will be required to define the antigen variability and assortment in each strain. Such molecular studies will allow the assessment of the epitopes targeted by protective antibodies, whose variability makes strains diversely susceptible to the immune response elicited by the vaccine 35. Also the analysis of the antigen expression level, combined with functional immunological assays, like the Serum Bactericidal Activity test, will be important in order to evaluate the universality of the vaccine. It is clear that the selection of representative strain panels is of primary importance. Additional panels will be selected in order to study vaccine coverage and to evaluate the vaccine component variability over time. To this aim, active and global surveillance of disease and accurate strain characterisation are needed. We analysed the gene and amino acid sequence variability of the main components of a new meningococcal vaccine, i.e. proteins fHbp, NadA and GNA2132. Several approaches and algorithms were used for the interpretation of the diversity and variability. The most important results and considerations arising from this analysis were: This work made use of the Multi Locus Sequence Typing website (http://pubmlst.org/neisseria/) developed by Jolley et al. 36 and hosted at the University of Oxford. We are grateful to Martin C. J. Maiden and Kate L. Seib for critically reading the manuscript. We are also grateful to Leonard W. Mayer, Ray Borrow, Dominique A. Caugant, Muhamed-Kheir Taha, Paola Mastrantonio, Julio A. Vazquez, Matthias Frosch and Arie van der Ende for providing the majority of strains. We also thank Catherine Mallia for editing the manuscript.
