The diversity and dynamics of Neisseria meningitidis populations generate a requirement for high resolution, comprehensive, and portable typing schemes for meningococcal disease surveillance. Molecular approaches, specifically DNA amplification and sequencing, are the methods of choice for various reasons, including: their generic nature and portability, comprehensive coverage, and ready implementation to culture negative clinical specimens. The following target genes are recommended: (1) the variable regions of the antigen-encoding genes porA and fetA and, if additional resolution is required, the porB gene for rapid investigation of disease outbreaks and investigating the distribution of antigenic variants; (2) the seven multilocus sequence typing loci–these data are essential for the most effective national, and international management of meningococcal disease, as well as being invaluable in studies of meningococcal population biology and evolution. These targets have been employed extensively in reference laboratories throughout the world and validated protocols have been published. It is further recommended that a modified nomenclature be adopted of the form: serogroup: PorA type: FetA type: sequence type (clonal complex), thus: B: P1.19,15: F5-1: ST-33 (cc32).
The precise characterization of Neisseria meningitidis samples from cases of invasive disease, both microbiological isolates and clinical specimens, is essential for informed public health responses and the management and control of meningococcal disease ( Stuart et al. , 1997 ; Caugant, 1998 ). The first typing methods to gain wide acceptance were immunological approaches ( Frasch et al. , 1985 ) that identified variants in the meningococcal capsule (serogroup; Vedros, 1987 ), outer-membrane proteins (OMPs) (serotype and serosubtype) ( Frasch & Chapman, 1972 ) and lipooligosaccharide (immunotype; Tsai et al. , 1987 ). The availability of monoclonal antibody panels enhanced the utility of these methods ( Poolman & Abdillahi, 1988 ; Nato et al. , 1991 ; Scholten et al. , 1994 ), which proved valuable in a variety of applications.
The immunological typing schemes exhibit a number of limitations however ( Maiden & Feavers, 1994 ), including: (1) incomplete coverage and difficulties in production and provision of reagents, resulting in an increasing number of isolates not typeable by these means ( Feavers et al. , 1996 ; Sacchi et al. , 1998 ); (2) reliance on expression of the typing target ( van der Ende, 1995 ); (3) inconsistent correspondence with the genetic relatedness of isolates, at least partially due to horizontal genetic exchange ( Feavers et al. , 1992a ; Maiden, 1993 ), combined with the high levels of positive selection experienced by meningococcal surface components ( Feavers et al. , 1992a ; Urwin et al. , 2002 ). Finally, it has proved difficult to apply immunological methods to nonculture diagnosis and typing ( Eldridge et al. , 1978 ), which is increasingly important given early treatment with antibiotics, resulting in culture-negative specimens ( Kaczmarski, 1997 ).
The application of multilocus enzyme electrophoresis (MLEE; Selander et al. , 1986 ) to meningococci ( Caugant et al. , 1987a ) was invaluable in improving our understanding of meningococcal disease, especially in mapping the global spread of particular meningococcal genotypes ( Caugant et al. , 1987b ; Olyhoek et al. , 1987 ; Achtman, 1990 ; Wang et al. , 1992 ) and investigating the underlying biology of meningococcal disease ( Achtman et al. , 1997 ; Caugant, 1998 ). It played an important role in the design of tailor-made vaccines targeted against serogroup B meningococci ( Bjune et al. , 1991 ). MLEE did not, however, gain wide acceptance as a routine typing technique due its inherent technical complexity, which discouraged many reference laboratories, and – perhaps more importantly – difficulties in comparing results among, or even within, laboratories ( Maiden et al. , 1998 ).
The advent of widely available molecular techniques ( Maiden & Frosch, 2001 ) has resolved these problems by enabling the comprehensive, rapid, inexpensive ( Diggle & Clarke, 2002 ) and reproducible characterization of essentially any genetic locus and hence the encoded protein. Furthermore, PCR-based techniques are especially suited to nonculture diagnosis and molecular techniques are generic and do not require specific training or reagent banks for their implementation. The ease with which molecular techniques can be developed and applied has the potential disadvantage, however, that different laboratories may introduce alternative and incompatible schemes, leading to a fragmentation rather than a unification of typing ( Achtman, 1996 ).
Here we propose a series of recommendations for the molecular characterization of meningococci by nucleotide sequence-based approaches, covering both target choice and nomenclature, that is comprehensive and expandable but also compatible with earlier MLEE and serological typing schemes.
The Neisseria meningitidis polysaccharide capsule, which defines the meningococcal serogroup, is an important virulence determinant, preventing opsonophagocytosis ( Unkmeir et al. , 2002 ). Only encapsulated meningococci regularly cause invasive disease and, of the 13 immunologically distinct capsules described in the literature ( Vedros, 1987 ), only five (corresponding to serogroups A, B, C, Y and W-135) are responsible for the great majority of cases ( Peltola et al. , 1992 ). The capsule is also an important vaccine component, with polysaccharide vaccines available against serogroups A, C, Y and W-135, but not serogroup B ( Jodar et al. , 2002 ). Determining the serogroup of an invasive meningococcus is therefore the first priority of the local or reference laboratory, as the result is likely to have important implications for outbreak management ( Stuart et al. , 1997 ).
While serogroup determination is easily achieved by conventional immunological typing where an isolate is available ( Vedros, 1987 ) and is likely to remain a mainstay of routine characterization, molecular typing has advantages in terms of speed and for nonculture diagnosis ( Guiver & Borrow, 2001 ). We, however, leave a detailed discussion of this to the accompanying paper by Taha & Fox (2006) .
Recommendation 1: antigen encoding genes
It is recommended that: (1) the variable region of the antigen-encoding genes porA and fetA are targeted for the rapid investigation of disease outbreaks and for investigating the distribution of antigenic variants, and (2) in cases where additional resolution is required, porB may also be examined.
Although the classification of meningococci into serogroups provides essential epidemiological information, it is not sufficient to define outbreaks, as most disease in many countries is caused by a limited number of serogroups, and often one or two are responsible for the majority of cases ( Caugant, 1998 ). To provide further epidemiological discrimination, the characterization of ‘subcapsular’ antigens, protein and lipooligosaccharide components of the outer membrane, was introduced ( Frasch et al. , 1985 ). The subcapsular antigens are also potential vaccine components employed to overcome difficulties in the development of vaccines based on the serogroup B capsules ( Jodar et al. , 2002 ) and a characterization of the distribution of variants among epidemiological samples is therefore valuable in vaccine design ( Feavers et al. , 1996 ; Urwin et al. , 2004 ).
Immunological schemes defining serotype, serosubtype and immunotype have already been described ( Frasch et al. , 1985 ). The serotype and subtype schemes both target OMPs, and the PorB and PorA proteins respectively, while the immunotype scheme targets lipooligosaccharide variants. As the variation in lipooligosaccharide expression is controlled by multiple synthetic genes ( Jennings et al. , 2001 ), it is not easily transferred to a molecular method for routine typing. As a consequence, we make no recommendations here, as molecular typing is likely to be too complex for routine applications and no such methods have been published or implemented as routine procedures. The protein antigens PorA and PorB, being encoded by single loci, are amenable to analysis by molecular methods, as is a further OMP, the iron regulated FetA protein ( Thompson et al. , 2003 ), which is recommended here as a novel additional typing target.
The meningococcal serosubtyping antigen, the class 1 OMP or PorA ( Hitchcock, 1989 ), is a porin that is a major constituent of the outer membrane of most meningococcal isolates. Nucleotide sequence analysis of porA genes of serological reference strains of the meningococcus ( Maiden et al. , 1991 ) demonstrated that the antigenic diversity recognized by the serosubtyping system resided in regions of variable peptide sequence of PorA. These variable regions (VRs) corresponded to putative membrane exposed loops in the proposed PorA protein structure ( McGuinness et al. , 1990 , 1993 ; Maiden et al. , 1991 ; van der Ley, 1991 ). There are two major (VR1 and VR2) and one minor, or semi-variable (VR3 or sVR), variable regions of this protein; this observation explains why meningococci can exhibit multiple serotypes. PorA is expressed by most meningococcal isolates and is a leading candidate vaccine component, although its use in this role is complicated by the large number of PorA variants found within meningococcal populations ( Maiden et al. , 1991 ; Feavers et al. , 1992a , 1996 ; Suker et al. , 1994 , 1996 ). Nevertheless, it has been an important component of vaccines that target particular epidemic strains ( Bjune et al. , 1991 ; Sierra et al. , 1991 ; Oster et al. , 2005 ) and is included in a number of other vaccine formulations ( Rouppe van der Voort, 2000 ).
The extensive variations in PorA has created problems for the serosubtyping scheme in two respects: (i) the production and provision of reagents covering all possible variants, and (ii) the inconsistent reaction of immunological reagents with the variants ( Suker et al. , 1996 ). This has been exacerbated by the fact that PorA variation occurs among and within groups of related VR peptide sequence ‘families’. For example, more than 135 unique peptide sequences have been reported for VR1, grouped into 10 families, while for VR2, the more diverse variable region, over 375 unique peptide sequences divided into 18 families have been reported; many of these are not recognized by existing monoclonal antibody panels and indeed the provision of a comprehensive panel would be impractical. Determination of the nucleotide sequence of those parts of the porA gene that encode the variable regions (VR1 and VR2) resolves these issues, enabling the peptide sequence of all variants to be deduced ( Suker et al. , 1996 ; Russell et al. , 2004 ). The employment of the deduced peptide sequences for typing retains a link with the serosubtyping scheme and provides information relevant to vaccine design whilst retaining high levels of discrimination.
A single primer pair can be employed to amplify all the VRs in one gene, but because of the distance between them (∼500 bp from the start of VR1 to the end of VR2) the two regions may need to be sequenced separately depending on the sequencing technology being used. The VR peptide sequences are designated in accordance with a standard nomenclature ( Russell et al. , 2004 ) that retains features of the serosubtyping system whilst providing flexibility for the accommodation of new variants which are continually being described. Each unique sequence is assigned to a VR family, indicated with a number, with the particular variant identified by a further number, separated from the family number by a hyphen. Where appropriate, the families are named following the previous serosubtyping nomenclature. The subvariant numbers are arbitrary and assigned in order of description. Whether a variant is located in VR1 or VR2 is indicated by the order of the designation following the prefix P1 (for porin 1, a convention also maintained from the serosubtyping scheme), thus P1.VR1, VR2 e.g., P1.5-2,10-1. The nomenclature is described and maintained in an online database hosted at http://neisseria.org/nm/typing/ ( Russell et al. , 2004 ) containing details of all known variants and providing search tools incorporating the blast algorithm ( Altschul et al. , 1997 ), allowing sequences for both variable regions to be queried simultaneously. Details of representative isolates with matching PorA variants are returned by the web database, where available, and the PubMLST isolate database is simultaneously searched so that as much publicly available information concerning matching isolates is retrieved.
It has been proposed to incorporate the sVR into this scheme ( Clarke et al. , 2003 ). However, as this region is shorter and less diverse and immunogenic than VR1 and VR2, incorporating these data adds little additional information. Therefore, although sVR or VR3 sequencing has and can be employed ( Molling et al. , 2000 , 2001 ; Clarke et al. , 2003 ), it is not commonly performed for routine surveillance and is not recommended here as it adds relatively little additional information for the additional expense and effort involved.
FetA (formerly designated FrpB or sometimes as 70 kDa iron regulated protein) is an iron-regulated meningococcal OMP that has homology to the Escherichia coli receptor for the ferric enterobactin siderophore ( Ala'Aldeen et al. , 1990 ; van der Ley, 1996 ; Carson et al. , 1999 ). This protein was not employed as a target for routine serological typing, mainly because although it is a major component of the outer membrane in vivo , and few isolates lack the gene that encodes it, it is only expressed under conditions of iron limitation and so that while it will be expressed in the human host, it is only expressed in vitro if conditions of iron limitation are imposed. Since its importance was recognized, however, it has been of interest as a vaccine candidate following the observations that: (1) FetA antibodies are present in sera from convalescent patients of meningococcal disease ( Black et al. , 1986 ); (2) antibodies against FetA have been observed in vaccines ( Wedege et al. , 1998 ); and (3) anti-FetA monoclonal antibodies exhibit bactericidal activity specific to the strain to which they were raised ( Pettersson et al. , 1990 ).
Nucleotide sequencing of the fetA gene from representative meningococcal isolates and the deduction of the peptide sequences enabled a topology model of FetA to be proposed, predicting 13 surface-exposed loops that are potentially exposed to the immune system ( Pettersson et al. , 1995 ; Thompson, 2001 ). One of these (loop VII) is variable among isolates, corresponding to a single variable region (FetA VR), somewhat similar to VR1 and VR2 of PorA. This VR includes epitopes for many of the monoclonal antibodies raised against the protein ( van der Ley, 1996 ). A systematic nucleotide sequence-based survey of the diversity of the whole gene showed that the FetA VR peptides could be categorized into five (now expanded to six) families of related peptide sequences in a similar way to the PorA VR families ( Thompson et al. , 2003 ).
The sequence diversity and length of the FetA VR make it suitable for use as a molecular marker. It is especially attractive given the similarities with the VRs of the PorA protein and the fact that the porA and fetA genes are distant from each other on the meningococcal chromosome ( Parkhill et al. , 2000 ). As with PorA VRs, a database is hosted at http://neisseria.org/nm/typing/ containing sequences of all known FetA variable region peptides, and can be searched using the blast algorithm. Currently, 186 FetA VR peptides have been identified. Individual variable region sequences are named with an ‘F’ followed by their family and variant numbers (designated in the order of discovery), e.g., F5-1. The database contains an isolate table so that representative isolates for the FetA variants can be found and it is also linked into the PubMLST isolate database to allow wider searching.
The meningococcal serotyping antigen is the PorB OMP ( Hitchcock, 1989 ; Feavers et al. , 1992b ), which like the PorA protein is a porin and present in most isolates; indeed these two proteins are related ( Derrick et al. , 1999 ). Unlike the porA and fetA genes, the porB gene is constitutively expressed. PorB proteins can be divided into two classes, class 2 and class 3 meningococcal OMPs ( Frasch et al. , 1986 ), a distinction originally based on their size as measured by SDS-PAGE, but these classes, which have also been named PorB2 and PorB3 are also homology groups on the basis of nucleotide and amino acid sequences ( Derrick et al. , 1999 ). As with serosubtyping based on PorA, a range of serological reagents have been developed that detect meningococcal subtypes ( Frasch et al. , 1985 ; Abdillahi & Poolman, 1988 ; Abdillahi et al. , 1988 ).
In contrast to PorA, the sequence variations of the smaller PorB proteins that lead to immunological variation is dispersed among six of the predicted surface-exposed loops of the predicted OMP structure and, although some of the serotype epitopes can be equated to particular peptide sequences ( Urwin, 1998 ; Urwin et al. , 1998a , b ), many cannot ( Zapata et al. , 1992 ). Consequently, the serotyping scheme cannot be translated to a peptide sequence-based nomenclature in the same way as the serosubtyping system. For this reason it is recommended that porB sequencing be used as a supplement to PorA and FetA VR when necessary. This is relatively time consuming for the information gained, as the complete porB gene needs to be sequenced and assembled, requiring at least four sequencing reactions in order to obtain reliable data from both strands.
An online database hosted at http://neisseria.org/nm/typing/ contains all known alleles for the porB gene as well as the known peptides described for the six variable loops ( Urwin et al. , 2004 ). At the time of writing, there were nearly 100 porB2 and 175 porB3 alleles. blast searching with a nucleotide or peptide sequence can query against all the loop peptides and whole-gene nucleotide alleles simultaneously. For typing purposes, the identification of the loop sequences provides more useful information than the allele number and is more analogous to the use of monoclonal antibodies that recognize particular epitopes on the loops, although, as stated, multiple loops may be involved in forming each epitope.
Recommendation 2: multilocus sequence typing
It is recommended that seven-locus MLST be employed to determine the sequence type and clonal complex of invasive and carried meningococci, at least for a representative selection of each country or for situations with enhanced regional surveillance. This information is essential for the national and international management of meningococcal disease, as well as being invaluable in studies of meningococcal population biology and evolution.
Multilocus sequence typing (MLST) was proposed as a portable replacement of MLEE for the identification of bacterial genotypes ( Urwin & Maiden, 2003 ), and was first developed and implemented for Neisseria meningitidis ( Maiden et al. , 1998 ; Jolley, 2001 ). As for MLEE, MLST indexes the variation at housekeeping genes that are under stabilizing selection, but identifies this variation by nucleotide sequence determination. MLST is consequently of higher resolution than MLEE and only seven alleles are required to characterize meningococci and identify hyperinvasive lineages, compared to the 15–20 loci employed in MLEE studies ( Urwin & Maiden, 2003 ). MLST has been employed to resolve meningococcal epidemiology at a variety of levels and is also suitable for investigating the population biology and evolution of the organism.
To ensure an accurate identification of MLST alleles, it is essential that sequencing be performed on both strands of the DNA. Having two strands enables any discrepancies to be highlighted automatically by the assembly tools. Using strands in opposite directions ensures that any artefacts that may be introduced by the combination of a particular sequence and the sequencing chemistry or technology used can be readily resolved.
It is also recommended that a nested sequencing strategy is employed, where separate primers are used for amplification and sequencing. Using two sets of primers means that lower stringency conditions can be employed for the initial amplification, since only specific products will be sequenced. This increases the sensitivity of the reaction, enhancing amplification from clinical or low copy number samples ( Kriz et al. , 2002 ; Diggle et al. , 2003 ; Birtles et al. , 2005 ). Primers recommended for amplification and sequencing are listed on the Neisseria MLST website ( http://pubmlst.org/neisseria/ ).
Recommendation 3: nomenclature
It is recommended that a modified nomenclature be adopted of the form: serogroup: PorA type (formerly assessed by serological typing as serosubtype): FetA type: sequence type (clonal complex), thus: B: P1.19,15: F5-1: ST-33 (cc32).
A standard scheme for the designation of serological variants of meningococci was proposed as long ago as 1985 ( Frasch et al. , 1985 ), this took the form of serogroup: serotype: subtype: immunotype, e.g., B:15:P1.7,16: L3,7,9. The group (capsule) and serotype (PorB) had no prefixes and the subtype (PorA) and immunotype (lipooligosaccharide) variants were indicated with the prefixed P1. and L respectively. Standard names for meningococcal genes have also been proposed ( Hitchcock, 1989 ). Various nomenclatures were adopted to identify hyperinvasive lineages from MLEE data, with different schemes for the mainly serogroup B and C meningococci isolated in Europe and the Americas, using the terms clonal complex or cluster (e.g., cluster A4, ET-37 complex) and the serogroup A meningococci isolated mainly in Africa and Asia (subgroups I through VIII) ( Olyhoek et al. , 1987 ; Wang et al. , 1992 ). To date no formal attempt has been made to unify these two separate schemes into a single nomenclature ( Caugant et al. , 1987a ). Both MLEE and immunological typing have been superseded by the advent of sequence-based typing for antigens and MLST for the genetic characterization of meningococci ( Maiden et al. , 1998 ).
Nomenclature schemes need to be modified in response to changes in understanding and technology, but it is desirable to maintain as much consistency as possible between older and newer schemes. The largest changes proposed – in the transition from an essentially phenotypic to mainly genotypic based scheme – are the inclusion of the MLST designation in the standard description of an isolate and the replacement of the serotype character with the FetA VR. It is proposed that we maintain the same general approach so that the new nomenclature is redolent of the earlier version and can at least to an extent, be compared with it, but to increase the quantity and quality of the data conveyed. The proposed scheme includes ST, clonal complex where known, the serogroup, PorA type and FetA type. Collectively these data provide a high discrimination for the purposes of strain typing and epidemiology but are also informative as to the fundamental biology of the meningococcus and also valuable in studies aimed at developing novel vaccines.
A major disadvantage with serological schemes was the large and increasing proportion of isolates described as not groupable (NG), not typeable (NT) or not subtypeable (NST). The move to genetic techniques means that meningococci will only be not typeable if the gene is not present. Where this has been determined it is suggested that the Greek letter Δ be inserted in the relevant part of the scheme, demonstrating that that particular character is absent. If a character cannot be determined for other reasons, for example if the ST has not yet been assigned to a clonal complex, it is recommended that a hyphen (-) be inserted, for example for an isolate that has an ST not yet assigned to a clonal complex and in which the porA gene is deleted, the designation would be: B: P1.Δ,Δ:F1-8: ST-16 (-). A final recommendation is that the letters ND (for ‘not determined’) are used where a character is unknown because the typing has not been performed.
During the last 15 years there has been a change in emphasis in meningococcal typing, with phenotypic and serological approaches being increasingly replaced by molecular techniques aimed at determining genotypes ( Maiden & Frosch, 2001 ). As molecular technology has become less expensive and widely implemented this has resulted in high resolution typing becoming more standardized, reproducible, and available. The portability of both techniques and information has been greatly enhanced by and the availability of curated databases via the Internet that describe nomenclature schemes and isolate collections in detail ( http://neisseria.org/nm/typing/ ). This has required some changes in targets and nomenclature, but efforts have been made to ensure the nomenclature schemes are as compatible as possible with previous schemes, an example being the naming of the PorA VR families, which although currently defined on the basis of peptide sequence retain the serosubtype name, where this existed ( Russell et al. , 2004 ).
Perhaps the most radical proposal made here is the removal of the serotype character from the scheme, at least for most routine typing, as the serotype is difficult to translate into a rapid molecular method and the effort of sequencing the entire porB gene is not warranted for routine typing given the information that it provides. The serotype character, especially the 2a and 2b serotypes, has been very useful in the past in identifying meningococci responsible for particular epidemics. This is because a number of the peptide sequences of the putative loops of the PorB proteins, especially the larger PorB2 (class 2 OMP) variants, are conserved within the clonal complex. For example, the majority of ST-11 complex meningococci possess identical PorB2 variable loops that correspond to serotype 2a ( Urwin et al. , 2004 ), whereas the ST-8 complex is characterized by serotype 2b. These serotype reagents were therefore useful largely because they identified particular genotypes, a characteristic that can now be more reliably identified by MLST. The utility of the 2a and 2b serotypes also depends on the particular meningococci causing the disease in a given locale at a particular time. The 2a and 2b serotypes are also not absolute, as ST-11 or ST-8 complex meningococci may not possess them and these serotypes are also found on other unrelated meningococci.
One of the perceived difficulties in the adoption of genotypic schemes for antigen genes is that this information does not indicate the expression of a particular character; this is thought to be especially relevant in the case of vaccine antigens. The same can also be said, however, for serological methods as it is difficult to determine whether the expression of antigen genes has been lost during the subculture of meningococci in the laboratory. In practice this concern is not too serious as the expression can be reversibly changed for the capsule, porA , and fetA genes and the existence of the gene in a given meningococcus usually indicates the potential to express that gene. Isolates, especially those from cases of invasive disease, rarely lack the genes proposed here and mostly express the antigens. Even in those cases where expression is permanently down-regulated, the meningococcus is likely to have a recent ancestor that expressed those genes. Finally, hyperinvasive meningococci deleted for these major antigen genes rarely spread in the population.
In conclusion, molecular typing methods have matured into precise, reproducible, and portable means of characterizing meningococci from both microbiological cultures and clinical specimens. The data generated enable epidemiological analysis at a variety of levels (local, national and global) and academic and applied research into the biology and control of this globally important pathogen.
M.C.J.M. is a Wellcome Trust Senior Research Fellow in Basic Biomedical Sciences. This work was funded by the European Union as part of the EU-MenNet project (QLK2-CT-2001-01436) and the Wellcome Trust.