Construction of Opa-Positive and Opa-Negative Strains of Neisseria meningitidis to Evaluate a Novel Meningococcal Vaccine

Neisseria meningitidis is a major global pathogen causing invasive disease with a mortality of 5–10%. Most disease in developed countries is caused by serogroup B infection, against which there is no universal vaccine. Opacity-associated adhesin (Opa) proteins are major meningococcal outer membrane proteins, which have shown recent promise as a potential novel vaccine. Immunisation of mice with different Opa variants elicited high levels of meningococcal-specific bactericidal antibodies, demonstrating proof in principle for this approach. Opa proteins are critical in meningococcal pathogenesis, mediating bacterial adherence to host cells, and modulating human cellular immunity via interactions with T cells and neutrophils, although there are conflicting data regarding their effects on CD4+ T cells. We constructed Opa-positive and Opa-negative meningococcal strains to allow further evaluation of Opa as a vaccine component. All four opa genes from N. meningitidis strain H44/76 were sequentially disrupted to construct all possible combinations of N. meningitidis strains deficient in one, two, three, or all four opa genes. The transformations demonstrated that homologous recombination of exogenous DNA into the meningococcal chromosome can occur with as little as 80 bp, and that minor sequence differences are permissible. Anti-Opa bactericidal antibody responses following immunisation of mice with recombinant Opa were specific to the Opa variant used in immunisation. No immunomodulatory effects were observed when Opa was contained within meningococcal outer membrane vesicles (OMVs), compared to Opa-negative OMVs. These observations support the incorporation of Opa in meningococcal vaccines.


Introduction
Neisseria meningitidis causes up to 500,000 cases of meningitis and septicaemia worldwide annually, with a mortality rate of approximately 10% [1]. Most cases of disease are caused by 5 of the 13 meningococcal serogroups: A, B, C, Y and W135. Proteinpolysaccharide conjugate vaccines are now available for all of these serogroups except serogroup B, since epitopes of this polysaccharide capsule are cross-reactive with the human neural cell adhesion molecule [2], and it is therefore not immunogenic in humans [3]. Serogroup B organisms are currently the major cause of disease in most temperate countries [4,5,6,7,8]. A number of vaccines based on different combinations of subcapsular antigens are currently in development for the prevention of serogroup B disease [9], including different outer membrane vesicle (OMV) vaccines using genetically modified meningococci [9,10,11,12].
The Opacity-associated (Opa) adhesin proteins are some of the major proteins found in the outer membrane of N. meningitidis and N. gonorrhoeae, which can express up to four or eleven different Opa variants respectively, encoded at loci dispersed throughout the genome [13,14,15]. There are at least 338 allelic variants (www. neisseria.org, accessed 17th April 2012) due to sequence variability in three of the four surface-exposed loops [16,17]. Opa proteins play a critical role in meningococcal pathogenesis by mediating bacterial adherence to the nasopharynx and modulating human cellular immunity via interactions with T cells and neutrophils [18,19]. Opa proteins are a potential novel vaccine candidate for the prevention of meningococcal disease caused by all serogroups, and have the potential to protect against the hyperinvasive isolates which are responsible for most of the disease burden [20,21]. Bactericidal antibodies are currently accepted as the correlate of protection against meningococcal disease, and anti-Opa bactericidal antibodies have been demonstrated in patients following infection with N. meningitidis and in recipients of serogroup B OMV vaccines [22,23,24,25,26,27,28,29]. Immunisation of mice with recombinant Opa proteins or Opa-containing liposomes has also elicited the production of high levels of bactericidal antibodies [21,30]. One obstacle to human trials of an Opa vaccine is an observation that these proteins might inhibit CD4 + T cell proliferation under certain conditions in vitro, although there are conflicting published data about the validity of this observation [31,32,33,34,35]. The potential expression of multiple Opa proteins from any given isolate has made it difficult to study the immunoprotective or immunomodulatory effects of specific Opa proteins expressed on the bacterial surface.
In order to individually assess different Opa variants as potential vaccine candidates, we constructed Opa-positive and Opanegative meningococcal strains, from the same parent strain, expressing different Opa proteins. A library of strains was derived from N. meningitidis strain H44/76 in which one, two, three or four opa genes had been disrupted, for further evaluation of Opa proteins as a potential meningococcal vaccine candidate. These strains were utilised to examine the specificity of the anti-Opa response following immunisation of mice with recombinant Opa protein and Opa-positive or Opa-negative OMVs.

Construction of Dopa plasmids
Locus-specific Dopa plasmids were designed to facilitate sequential, targeted disruption of the four opa genes (opaA, opaB, opaD and opaJ) of N. meningitidis strain H44/76. These Dopa plasmids each contained a disrupted opa gene flanked by upstream and downstream sequences specific for the relevant opa locus, with or without an antibiotic resistance cassette (for selection following transformation) (figure 1). However, some of the cloning steps were unsuccessful; it was not possible to construct locus-specific DopaB plasmids, and insertion of an antibiotic resistance cassette was only possible for the DopaJ plasmid. An alternative strategy was devised based on the finding that the four opa genes of strain H44/76 possess 96% sequence identity for the 253 bp at the 59 end and 93% for the 228 bp at the 39 end, with 99% similarity between opaA and opaJ, and 97% between opaB and opaD within these regions. Generic Dopa plasmids were therefore constructed, without the flanking locus-specific regions, to enable non-specific disruption of opa genes (figure 1). Locus-specific plasmids all included the suffix -nmb. PCR primers are listed in table 1.

DopaJ plasmids
The 59 and 39 ends of opaJ were amplified separately along with adjacent sections of the neighbouring genes nmb0925 and pip, introducing SalI and BamHI or SacII and BamHI restriction sites at the ends of the amplicons. These were cloned separately into pCR2.1-TOPO before being excised with the relevant restriction enzymes and ligated together in the plasmid vector pBluescript II (SK-) to construct pBS-DopaJ-nmb. An erythromycin resistance cassette (EryR) containing ermC was amplified from the plasmid pER2 [36], using primers ery-bamf and ery-bamr to introduce BamHI sites flanking ermC. The resulting amplicon was cloned into pCR2.1-TOPO, excised with BamHI and inserted into the BamHI site of pBS-DopaJ-nmb, resulting in pBS-DopaJ::ery-nmb. DopaJ::-ery was amplified from this plasmid using primers OpaFSalI and OpaRSacII to exclude adjacent locus-specific genes. This amplicon contained DopaJ::ery and the immediately adjacent opa homologous regions, including a downstream DNA uptake  (table 1) along with the adjacent locus-specific genes, depicted as nmb01 (blue) and nmb03 (green). Novel SalI and BamHI or BamHI and SacII restriction sites were introduced at the ends of the amplicons. Each PCR product was then cloned separately into pCR2.1-TOPO (not shown). The two ends of each opa were excised from the pCR2.1 plasmids and cloned sequentially into pBluescript, resulting in locus-specific plasmids pBS-DopaA-nmb, pBS-DopaD-nmb and pBS-DopaJ-nmb. These plasmids therefore contained a modified opa gene, Dopa, which contained a 185 bp deletion. (ii) A 1,192 bp BamHI fragment carrying ermC was cloned into pBS-DopaJ-nmb to produce pBS-DopaJ::ery-nmb. (iii) DopaJ::ery was amplified from pBS-DopaJ::ery-nmb using primers OpaFSalI and OpaRSacII (table 1), excluding opaJ locus-specific regions. The resulting amplicon was cloned into pCR2.1-TOPO to generate the generic plasmid pCR-Dopa::ery. (iv) DopaA and DopaD were amplified from their respective pBS-Dopa-nmb plasmids using primers OpaFSalI and OpaRSacII. Each modified opa was cloned into pCR2.1-TOPO, resulting in the generic plasmids pCR-DopaA and pCR-DopaD. (v) DopaA and DopaD were excised from the pCR2.1 plasmids by double digestion with SalI and SacII and cloned into pBluescript which had been similarly prepared, to create pBS-Dopa plasmids. (vi) A 1,194 bp SgrAI fragment carrying ermC was cloned into pCR-DopaD to produce pBS-DopaD::ery. (vii) A kanamycin (KanR) or tetracycline (TetR) resistance cassette was introduced into pBS-DopaD, or KanR was introduced into pBS-DopaA, resulting in generic Dopa plasmids containing selectable markers. doi:10.1371/journal.pone.0051045.g001 sequence (DUS), and novel SalI and SacII sites at either end. This was cloned into pCR2.1-TOPO to construct pCR-opaJ::ery.
DopaD plasmids pBS-opaD-nmb was constructed in a similar fashion to pBS-opaJ-nmb. DopaD was amplified from this plasmid using OpaFSalI and OpaRSacII and cloned into pCR2.1-TOPO, resulting in pCR-DopaD. DopaD was excised using SalI and SacII and ligated into pBluescript, which had been similarly prepared, to create pBS-DopaD. EryR was inserted into pCR-DopaD by amplifying ermC from pER2 using primers ery-sgrf and ery-sgrr, to introduce SgrAI sites flanking ermC, before cloning it into the SgrAI site within opaD, to obtain pCR-DopaD::ery. A kanamycin resistance cassette (KanR) was excised from pUC4-kan (GenBank accession number X06404) using BamHI. A tetracycline resistance cassette (TetR) was constructed by excising a HindIII-SalI fragment from pHVT1 [37], which was cloned into a plic2A vector and amplified by PCR using primers NmnDUS5 and NmnDUS3 (table 1). This reduced the size of TetR by several hundred base pairs, and introduced novel PstI sites and neisserial DNA uptake sequences. This amplicon was cloned into a TA vector, excised with PstI and inserted into pUC4-kan which had been similarly prepared, thus replacing KanR with TetR to produce pUC4NmDUS. pBS-DopaD::kan and pBS-DopaD::tet were constructed by insertion of either KanR or TetR, respectively, into the BamHI site of pBS-DopaD.
DopaA plasmids pBS-DopaA-nmb was constructed as described for pBS-DopaJnmb. pBS-DopaA and pBS-DopaA::kan were derived in a similar fashion to the equivalent DopaD plasmids.

Construction of opa-deficient meningococci
N. meningitidis strain H44/76 was sequentially transformed with different Dopa plasmids to create a library of 15 new opa-deficient strains, each theoretically able to express different combinations of the four Opa proteins ( figure 2 and figure 3). Plasmids containing different antibiotic resistance cassettes were used to enable selection after each disruption. Plasmids containing DopaA or DopaJ were able to target both of these genes and those containing DopaD disrupted opaB and opaD. To disrupt all four opa genes with only three selectable markers, one of the antibiotic resistance

1776II
Restriction sites within primer sequences are underlined. nmb nomenclature as defined in the published sequence of N. meningitidis strain MC58 (GenBank accession number AE002098), which was used to design primers in this study since the sequence of H44/76 had not been published. *nmb0443 is adjacent to opaA but no unique primer site could be identified within nmb0443 so nmb0444 was used; { nmb1635 is adjacent to opaB but the putative coding sequence is only 222 bp so nmb1634 was used; { nmb1466 is adjacent to opaD but primers within ppx have been published following successful use; cassettes had to be removed to be re-introduced at a different opa locus. Strain M011 was obtained by transformation of M009 using pBS-DopaJ-nmb to replace DopaJ::kan with DopaJ. A single transformed clone was obtained after screening 10,000 colonies following transformation. This strain was then used to produce two additional strains with KanR introduced within opaA or opaD and in which opaJ remained disrupted. KanR only conferred kanamycin resistance in N. meningitidis if inserted into opa in the same orientation as the opa gene.
pBS-DopaA::kan resulted in integration of DopaA::kan into the opaA locus on 70/73 (96%) occasions, and into opaJ following other transformations. Similarly, pCR-DopaJ::ery only targeted opaJ and opaA, disrupting opaJ 53/68 times (78%). A total of 81 clones were identified where generic plasmids based on opaD had undergone homologous recombination. This occurred at the opaD locus on 67 occasions (83%), at the opaB locus 10 times (12%), and at the opaA locus in four clones (5%). Four of the events where integration occurred into the opaD locus were single crossover events, with both KanR and TetR being present at the same locus. Although this has been described [38,39], it is not thought to be a common occurrence in N. meningitidis. All other successful transformations were confirmed by PCR to be double crossover events. Transformation rates varied between 10 25 and 10 27 , consistent with previous studies [39,40].

DNA sequence analysis of Dopa genes
Nucleotide sequence analysis was performed on multiple loci where recombination occurred between two different opa genes, one located on the plasmid and one on the chromosome. Crossover events were observed throughout the homologous flanking regions, including within the first 81 bp of homology, close to the site of the primer used to amplify the opa genes, and also within 80 bp of the antibiotic resistance cassette (figure S1). Data from opa pentanucleotide coding repeat (CR) sequences, contained within the 59 end of the Opa open reading frame, revealed that the length of the CR tracts had altered in some opa genes during construction of the strain library (table S1). Based on the length of the CR tract, the mature Opa protein should only be expressed on the bacterial surface if the number of CR sequences is a multiple of 3. Therefore, the wild type strain used in this study is not expected to express any Opa proteins, whereas M001 should express both OpaA and OpaD. Only OpaD is predicted to be expressed by M002, M005 and M007. None of the other strains are expected to express any Opa proteins, based on the nucleotide sequence, although they would have the capacity for differential expression of Opa by phase variation (PV) at opa loci which had not been disrupted.
Characterisation of Opa-positive and Opa-negative meningococci H44/76, M001, M002 and M014 were characterised further and used to provide additional information on the immunogenicity of Opa in the mouse model. Immunodot-blotting of ethanol-fixed bacteria using mAbs 15-1-P5.5 and MN20E12.70 confirmed the expected Opa phenotype based on DNA sequencing data, with expression of OpaD for M001 and M002 and OpaA from M001 only (figure S2). SDS-PAGE and immunoblotting of OMVs produced from these strains confirmed Opa expression consistent with the results of DNA sequence analysis and immunodotblotting ( figure S2). Protein profiles of other major surfaceexpressed proteins were similar between strains, and based on immunoblotting with specific monoclonal antibodies there was no effect on the expression of PorA, PorB, RmpM or factor H binding Bactericidal antibody responses following immunisation of mice with recombinant Opa proteins and Opa-positive and Opa-negative OMVs The immunogenicity of recombinant forms of OpaA and OpaD from H44/76 were tested using the selected Opa-positive and Opa-negative strains as target strains in the serum bactericidal antibody (SBA) assay. The recombinant proteins resulted in titres of 1:256 when the target strain expressed the same Opa variant used for immunisation (table 2). SBA titres of ,1:4 were observed when the target strain in the SBA assay did not express the Opa variant used for immunisation. Immunisation of mice with any of the four OMVs elicited a bactericidal response against all four target strains, demonstrating little difference between the SBA titres (table 2), with a titre of 1:2048 for most sera against the majority of target strains.

Discussion
This is the first description of the construction of a library of opadeficient meningococci from a single parent strain, enabling further evaluation of Opa proteins as a potential novel meningococcal vaccine, including investigation of their immunomodulatory effects. This study has also demonstrated significant, specific bactericidal anti-Opa responses following murine immunisation, supporting the pursuit of Opa as a potential vaccine candidate. These strains will be a valuable tool in the assessment of the role of Opa proteins in the pathogenesis of neisserial infection. They will also allow investigation of different Opa variants from the same parent strain to identify any similarities and differences in their functions, and investigate their underlying molecular basis.
Sera from mice immunised with OpaA or OpaD from strain H44/76 were only bactericidal when the same Opa was expressed on the surface of the target strain in the SBA assay. Although previous studies have demonstrated that Opa proteins elicit bactericidal antibodies in mice, the Opa-positive and Opanegative strains constructed in this study enabled an accurate assessment of specificity of the anti-Opa SBA response for the first time. Opa cross-reactive antibodies have previously been found following immunisation of mice, and additional investigations utilising a larger panel of Opa variants is required to fully explore the occurrence of any cross-reactivity. This would ideally include study of both mouse and human sera in those who have either received Opa-containing vaccines (including OMVs) or following invasive disease or meningococcal carriage. SBA titres following immunisation with recombinant proteins were between 4-and 16fold lower than following OMV immunisations, which may be expected given that OMVs contain a large number of antigens, including the immunodominant PorA [24,41,42]. This is also the likely explanation for the similarity of responses following immunisation with Opa-positive or Opa-negative OMVs, with the inclusion of Opa not conferring either advantage or disadvantage with respect to SBA response. However, given the lack of bactericidal activity demonstrated against strains not expressing the same Opa, it is likely that all of the SBA activity observed after protein immunisation was Opa-mediated, providing further support for pursuing Opa as a vaccine candidate. Importantly, the expression of Opa within the OMVs did not lead to any inhibition of the SBA response as has been found in some in vitro studies [34,35], although these effects may not necessarily be apparent in mice due to differences between the human and mouse CEACAM repertoire.
Inclusion of Opa proteins in future vaccines will require good quality epidemiological data to identify common Opa variants in circulation for inclusion in the vaccine. This would ideally include additional information on cross-reactivity of the bactericidal response following immunisation and Opa PV, which would influence the potential effectiveness of any Opa-containing vaccine. The lack of Opa expression in some strains constructed in this study which retained wild-type opa genes highlights the challenge that PV poses in developing Opa as a vaccine candidate. This would appear to suggest that four Opa proteins are required in a vaccine to ensure coverage against a single strain, in order to circumvent the problem of PV. However, bactericidal epitopes for anti-Opa antibodies have been described in two of the four surface-exposed loops containing hypervariable (HV) regions [17], and there is limited diversity of these HV regions among opa genes from hyperinvasive meningococcal isolates [43]. This lends credence to the possibility of providing broad coverage with a small number of Opa variants, despite the occurrence of PV. The importance of Opa to adhesion of meningococci to epithelial cells in the nasopharynx as the precursor to invasive disease mean it is unlikely that no Opa proteins would be expressed within a population of colonising organisms and provides the additional potential that an Opa vaccine might have activity against colonisation through induction of anti-adhesion antibody. No effect of anti-Opa antibodies (from post-mouse immunisation serum or using anti-Opa mAbs) on adhesion was demonstrated in this study using encapsulated bacteria (data not shown) [44,45], and although Opa-mediated adhesion of apparently fully encapsulated meningococci to epithelial cells has been demonstrated [46,47,48], such interactions occur most effectively with acapsulate organisms [49,50], which may represent a significant proportion of meningococci in the nasopharynx [51,52].
DNA sequence analysis following transformation of meningococci revealed that homologous recombination of exogenous DNA can occur with just 80 bp of homology in the flanking region. N. meningitidis is a naturally transformable bacterial species, having the ability to take up extracellular DNA efficiently, and then incorporate it into the chromosome by homologous recombination [53]. This mechanism can be manipulated in vitro to modify expression of specific proteins, and has been utilised recently in the development of several meningococcal vaccines to delete, overexpress or modify surface components and therefore increase the immunogenicity of OMVs derived from these strains [9,10,11,12]. However, the minimum requirement for sequence homology has not been clearly defined, although most studies suggest 500 bp-1,000 bp is required either side of the selectable marker to enable efficient recombination [39,54,55]. Successful transformation has been achieved with 280-290 bp of homology [38,56] and studies of N. gonorrhoeae have reported unpublished observations that recombination can occur with as little as 100 bp of homology, and that minor modifications in the flanking sequence are permissible [57,58]. This study provides clear evidence that only 80 bp is sufficient to enable homologous recombination to occur. Even with this relatively small amount of flanking sequence, minor differences were tolerated during the recombination process, although it is much more likely to occur if there is 100% sequence identity, as demonstrated by the higher rates of recombination at the same opa locus as that carried on the plasmid.
There are currently a limited number of selectable markers available for use in Neisseria, and this study highlights some important factors to optimise performance of these markers and cloning methods. We found that for KanR, the orientation must be the same as the target gene to obtain the resistance phenotype in N. meningitidis [59], although either orientation will confer kanamycin resistance in E. coli. For EryR or TetR, insertion can be in either orientation, which makes these markers easier to use for genetic manipulation of meningococci. Additional unpublished observations from our laboratory have shown that the concentration of kanamycin required in selective media following meningococcal transformation is strain dependent. The data also suggest that removal of a selectable marker is a feasible strategy for the manipulation of meningococcal DNA, when efficiency of transformation and recombination is high. This increases the number of genes which can be altered in a single isolate, but a nonantibiotic marker which is easily selectable when it is inserted or removed would greatly facilitate construction of mutant strains, either for the development of new vaccines or investigation of the biology of the organism. This would also ideally involve production of strains which are able to constitutively express different phase variable proteins. Construction of a strain which constitutively expresses Opa has been described for an opa gene of N. gonorrhoeae [60], but this has a number of inherent difficulties, including the presence of a repeat sequence of variable length within the open reading frame, a previous finding that high-level expression of Opa within E. coli was toxic to the host cells [60] and the requirement for high fidelity at each stage of DNA manipulation.
Our library of opa-deficient strains will allow further assessment of Opa as a potential vaccine candidate against meningococcal disease caused by all serogroups. These mutants have a potential use to further investigate the immunomodulatory effects of Opa, in order to identify the most immunogenic variants for use in a future meningococcal vaccine. In addition it will be possible to further investigate the role of Opa proteins during adhesion and invasion, including differences between Opa variants and the effects of expression of different combinations of Opa proteins on the bacterial surface. We have demonstrated that transformation of N. meningitidis is possible with very short regions of homology between plasmid and chromosomal DNA. Such shorter fragments of DNA should be easier to manipulate during cloning, so this finding will ease the construction of genetically modified meningococci in the attempt to develop a broadly protective meningococcal vaccine. These mutants demonstrate there is a significant and specific anti-Opa bactericidal antibody response following murine immunisation with recombinant protein, and no inhibition of the response when Opa is contained within OMVs supporting further investigation of Opa as a component of future meningococcal vaccines.

Ethics Statement
Animal studies were conducted according to the UK Home Office regulations and were approved by the local ethics committee at the National Institute for Biological Standards and Control (Home Office Project Licence Number 80/2157). Samples were obtained following terminal general anaesthesia and all efforts were made to minimise suffering.

Transformation of Escherichia coli and Neisseria meningitidis
Chemically competent E. coli DH5a were prepared using calcium chloride [61,62] and transformation of E. coli for the propagation of plasmids was performed using standard methods [62]. Transformed cells were plated onto selective LB agar and incubated at 37uC for 16-18 hours, or 36-40 hours when selecting for erythromycin resistance. N. meningitidis was transformed using the spot transformation technique [57]. Briefly, 10 ml (approximately 10 8 colony forming units [cfu]) of bacterial suspension from overnight growth was incubated with approximately 1 mg of linearised or supercoiled plasmid DNA or chromosomal DNA and plated over a 1-2 cm diameter region on BHI agar. Reactions were incubated at 37uC, 5% CO 2 for 4-8 hours before bacteria were plated onto selective BHI agar and incubated for a further 16-18 hours (36-40 hours for the introduction of erythromycin resistance). When transforming N. meningitidis to remove an antibiotic resistance cassette, transformation reactions were plated onto non-selective BHI agar to achieve 500-1,000 cfu per plate before incubation. After 16-18 hours, a nitrocellulose membrane (pore size 0.45 mm) was used to transfer colonies to selective BHI agar plates. After 16-18 hours paired plates were visually screened to identify colonies which were present on non-selective media only. Colonies were re-grown on both selective and non-selective media to confirm loss of antibiotic resistance.

Construction of meningococcal outer membrane vesicles
Bacteria were grown on BHI agar at 37uC, 5% CO 2 for 16-18 hours. A few individual colonies were picked and resuspended in 200-1000 ml of PBS, before plating 25-50 ml of this suspension onto BHI agar. OMVs were produced from this suspension as previously described [63] with the exception that thiomersal was excluded from all buffers.

Recombinant DNA methods
Plasmid DNA was isolated from E. coli using the QIAprep Miniprep kit (Qiagen, Crawley, UK) before being screened by polymerase chain reaction (PCR) using the universal primers M13 forward  and M13 reverse  and restriction enzyme digestion. Chromosomal DNA from N. meningitidis was isolated using the QIAamp DNA Mini kit (Qiagen). Following transformation, the identity of each new meningococcal transformant was confirmed by growth on selective media and analysis by PCR (table 3). PCR was carried out using primers at a concentration of 0.2 mM per reaction (Sigma-Aldrich). Each reaction contained 1.25 units Taq DNA Polymerase (Qiagen) and 200 mM each of dATP, dCTP, dGTP and dTTP in a total volume of 50 ml. PCR conditions were: initial denaturation at 95uC for 5 minutes; 30 cycles of denaturation at 95uC for 30 seconds, annealing at 55uC for 30 seconds, extension at 68uC for 3 minutes; final elongation at 68uC for 3-10 minutes. TA cloning was performed using the vector pCR2.1-TOPO (Invitrogen, Paisley, UK) according to the manufacturer's instructions. Ligation reactions were performed using standard methods [62].

DNA sequencing
Dye-terminator DNA sequencing was used following amplification of opa genes by PCR. Amplicons were purified by precipitation with polyethylene glycol 8000 (20% w/v)/NaCl (2.5 M), followed by centrifugation at 2,7506 g for 1 hour at 4uC and washing with ethanol (70%). Sequencing reactions contained 1.3 pmol oligonucelotide primer (Sigma-Aldrich) and 0.5 ml BigDye Terminator Ready Reaction Mix (Applied Biosystems, Paisley, UK) in a total of 10 ml. The following conditions were used: 30 cycles of 95uC for 10 seconds, 50uC for 6 seconds and 60uC for 2 minutes. Sequencing primers were O85 (59-GGCA-TAATCTGCCGCTATCC-39) [69] and OpaFwdII (59-TA-TATTGTGTTGAAACATCG-39), which was based on O3510 [20]. Extension products were purified by precipitation with ethanol (100%)/sodium acetate (115 mM, pH 4.6) and centrifugation as above. Labelled extension products were separated by capillary electrophoresis on a 37306l DNA Analyzer (Applied Biosystems) at the Department of Zoology Sequencing Facility, University of Oxford. Sequence trace chromatograms were assembled into consensus sequences using the Staden sequence analysis package [70], using Gap4 shotgun assembly. Sequences were aligned using Clustal X (version 2.0) [71] and visualised with GeneDoc (version 2.7.000) [72]. No new sequence data were generated in this study.

Preparation of recombinant Opa proteins
Refolded recombinant Opa proteins were produced in E. coli inclusion bodies from the pET22b(+)/opa plasmids as previously described [21,73] with an additional purification step. Ion exchange chromatography was performed following heparin affinity chromatography using a Resource S column (GE Healthcare, Buckinghamshire, UK) equilibrated in sodium acetate (50 mM, pH 4.6) containing LDAO (0.1%). Opa protein was then eluted by application of a gradient from 0 to 1 M NaCl. Fractions containing Opa proteins were pooled, dialysed into PBS containing LDAO (0.1%) and stored at 220uC.

Immunisation of mice
Groups of ten 6-7 week old female BALB/c mice (Charles River, Margate, UK) were immunised subcutaneously with 5 mg of recombinant Opa protein or OMVs (2.5 mg total protein) on days 0, 21 and 35. The oil-in-water emulsion Sigma Adjuvant System (Sigma-Aldrich), in a total volume of 2.4 ml of antigen solution per vial, was reconstituted with the antigen on the day of immunisation. Blood was collected by cardiac puncture on day 42 and serum separated by centrifugation at 16,1006 g for 10 minutes. Following each transformation, three PCR reactions were carried out at each opa locus to confirm recombination. The antibiotic resistance cassette was amplified using internal primers to confirm it had inserted into the genome. The antibiotic resistance cassette was amplified with one internal primer and one primer within a locusspecific gene adjacent to opa to confirm insertion had occurred within the target opa gene. Finally, the entire opa locus was amplified using primers in adjacent genes to confirm a double crossover event had occurred. *wt = wild-type. Serum bactericidal assay Bactericidal activity in mouse serum pooled within each immunisation group was quantified by SBA based on a published method [74]. Briefly, pooled murine sera was heated at 56uC for 30 minutes to deactivate endogenous complement and then diluted to give a range from 1:4 to 1:1024. Diluted sera was incubated with mid-log phase meningococci (125 cfu) and baby rabbit complement (lot number 11330, PelFreez Biologicals, Rogers, AR) at a final concentration of 25% (v/v). Reactions were incubated for 60 minutes at 37uC in a humidified 5% CO 2 atmosphere. A sample of each reaction was spread by tilting onto BHI agar plates, which were incubated overnight. The bactericidal antibody titre was reported as the reciprocal of the highest serum dilution at which 50% bacterial survival was observed. Each serum sample was analysed in duplicate against each target strain. When SBA titres were too high to be determined, additional assays were performed with dilution of sera to 1:16,384.