Susceptibility to Invasive Meningococcal Disease: Polymorphism of Complement System Genes and Neisseria meningitidis Factor H Binding Protein

Background Neisseria meningitidis can cause severe infection in humans. Polymorphism of Complement Factor H (CFH) is associated with altered risk of invasive meningococcal disease (IMD). We aimed to find whether polymorphism of other complement genes altered risk and whether variation of N. meningitidis factor H binding protein (fHBP) affected the risk association. Methods We undertook a case-control study with 309 European cases and 5,200 1958 Birth Cohort and National Blood Service cohort controls. We used additive model logistic regression, accepting P<0.05 as significant after correction for multiple testing. The effects of fHBP subfamily on the age at infection and severity of disease was tested using the independent samples median test and Student’s T test. The effect of CFH polymorphism on the N. meningitidis fHBP subfamily was investigated by logistic regression and Chi squared test. Results Rs12085435 A in C8B was associated with odds ratio (OR) of IMD (0.35 [95% CI 0.19–0.67]; P = 0.03 after correction). A CFH haplotype tagged by rs3753396 G was associated with IMD (OR 0.56 [95% CI 0.42–0.76], P = 1.6x10−4). There was no bacterial load (CtrA cycle threshold) difference associated with carriage of this haplotype. Host CFH haplotype and meningococcal fHBP subfamily were not associated. Individuals infected with meningococci expressing subfamily A fHBP were younger than those with subfamily B fHBP meningococci (median 1 vs 2 years; P = 0.025). Discussion The protective CFH haplotype alters odds of IMD without affecting bacterial load for affected heterozygotes. CFH haplotype did not affect the likelihood of infecting meningococci having either fHBP subfamily. The association between C8B rs12085435 and IMD requires independent replication. The CFH association is of interest because it is independent of known functional polymorphisms in CFH. As fHBP-containing vaccines are now in use, relationships between CFH polymorphism and vaccine effectiveness and side-effects may become important.


Introduction
The complement system is a fundamental part of the innate immune response. This pathway harms unprotected surfaces by a powerful positive feedback cycle that injures cells by perforating them with circular polymers (the membrane attack complex) and by activating further immune response by releasing opsonins and anaphylatoxins [1][2][3]. Complement activation can cause harm to both unprotected self and foreign cell surfaces [1,2].
Pathogenic bacteria evade the complement system by mimicking or binding to protective host proteins [4]. Human complement factor H (CFH) is the major inhibitory regulator of the complement system. Polymorphism of CFH and the adjacent homologous CFHR1-5 genes is associated with susceptibility to several inflammatory diseases [5][6][7][8][9][10]. A genome-wide association study of susceptibility to invasive meningococcal disease identified a major risk association at the CFH and CFHR3 locus [11]. The report noted that the associated variants are in strong linkage disequilibrium with the minor allele of rs1065489 (D936E) in the CFH gene, but evidence that this is the functional cause is lacking. Unexpectedly, the associated CFH polymorphism as one with no known functional effect and is not one associated with other inflammatory diseases.
Neisseria meningitidis infection causes sepsis and meningitis, with death in approximately 10% of cases [12]. Factor H-binding protein (fHBP) and Neisserial Surface Protein A bind host CFH to protect N. meningitidis [13][14][15]; Neisserial fHBP is critical for meningococcal survival in blood [16]. It binds short consensus repeats 6 and 7 of human CFH, which is a region of CFH that also binds to self-surface membranes [17]. It may cause its severe systemic effects by sequestering host CFH, leaving self surfaces unprotected [17,18]. The common CFH Y402H polymorphism (rs1061170), which is a major risk factor for age-related macular degeneration, is adjacent to the fHBP binding site, but does not affect binding to fHBP [17].
Factor H binding protein has been a recent focus of interest because it is now a component of vaccines against serogroup B N. meningitidis [19,20], one of which is already used in outbreak control [21] and is likely to be added to the UK childhood immunisation schedule [22], which might result in meningococcal disease becoming rare.
Polymorphism of fHBP can be categorized by two different systems of nomenclature. Fletcher et al. use a system of two subfamilies, A and B [19]. Masignani  three variant groups (1, 2 and 3) [20]. Disease severity is associated with polymorphic variation affecting the five segments that make up the modular structure of fHBP [23][24][25]. Previous studies of fHBP sequenced the whole gene from cultured N. meningitidis isolates, and then defined the subfamily, variant group or modular group using only a small number of sequence features. Our study is the first to define fHBP type directly using DNA isolated from patient blood. This may avoid bias due to variation in the success in culturing different strains of N. meningitidis. The aim of this study was to explore the relationship between invasive meningococcal disease and variation of the human complement system. We sought to refine the risk association at the CFH locus and to investigate whether variation of N. meningitidis fHBP affects this association. We explored other variations of the complement system, including terminal pathway genes where deficiency of proteins has been associated with susceptibility to recurrent meningococcal disease [26][27][28][29] and two complement inhibitors to which N. meningitidis binds: CD46, which encodes membrane cofactor protein, a membrane-bound complement inhibitor [30,31] and C4 binding protein, which inhibits the classical pathway [32].

Ethics Statement
The study was approved by the Office for Research Ethics Committees Northern Ireland (part of the UK National Research Ethics Service; study reference: 10/NIR03/24). Following formal proposal to its project advisory group, the Health Protection Agency (now part of Public Health England) provided anonymised residual clinical diagnostic DNA samples from PCRconfirmed cases of invasive meningococcal disease collected in 2009 and 2010. No identifying information was provided. Informed consent was not required for these samples because only anonymised demographic data and residual clinical DNA samples were provided. The Wellcome Trust Case Control Consortium (WTCCC) 1958 Birth Cohort and National Blood Service samples were collected with informed consent and used with permission of the WTCCC.

Study Population
The case population characteristics have been described in detail previously [33]. The cases were 309 European individuals with PCR-confirmed invasive meningococcal disease. We did not have access to details of the clinical features or demographic details other than age at time of illness. The N. meningitidis serogroups were: B, 292; C, 3; W, 4; Y 4. The ages ranged between one month and 73 years, with a median of two years. European ancestry was ascertained by using an ancestry-informative panel of polymorphisms [34] and cluster analysis, as described previously [33].
The control population comprised 5,200 individuals with European ancestry from the United Kingdom 1958 Birth Cohort and National Blood Service (NBS) cohort, for whom microarray genome-wide data were provided by the Wellcome Trust Case-Control Consortium (WTCCC). The WTCCC exclusions were applied, and validated by principal components analysis, as described previously [33]. The case population does not overlap with any group tested in the Davila et al. genome-wide association study. The control group is identical to that used in the genome-wide association study [11]. The median age of NBS participants was 45 years. The 1958 Birth Cohort participants were aged 52 years at the time of genotyping.

Detection of N. meningitidis
Confirmation of invasive meningococcal disease was based on a positive diagnostic Taqman assay for capsular transfer gene (ctrA) at the Health Protection Agency (now Public Health England) Meningococcal Reference Unit using an Applied Biosystems 7700 sequence detection system, as described previously [35]. The cycle number at which each positive sample was detected was reported. Cycle threshold, which is inversely correlated with bacterial load, has been used as an indicator of disease severity in other studies [36,37].

Genotyping of CFH Polymorphisms
Six SNPs in CFH were genotyped by SNaPshot primer extension methodology, involving PCR to amplify sequence fragments containing polymorphisms of interest, ExoSAP-IT to neutralise unincorporated dNTPs, fluorescent primer extension, further shrimp alkaline phosphatase clean-up, and analysis on an ABI 3100 genetic analyser. The SNPs genotyped were rs1061170 (Y402H), rs800292 (I62V), rs6677604 (which is in full linkage disequilibrium with the deletion of CFHR3-CFHR1), rs3753396 (which is in full linkage disequilibrium with rs1065489, D936E), rs419137, and rs2284664. SNPs were chosen on the basis of our established protocols for genotyping [8,38]. Oligonucleotides are shown in S1 Table. Genotype calls were made using GeneMarker v1.5.1. The control dataset had been genotyped on an Illumina 1.2M duo SNP microarray. Four core CFH haplotypes were defined using four SNPs (Table 1) based our previous studies (with the very closely related haplotypes 1 and 2 in our haplotype model [38] replaced by haplotype A in the present report, defined by rs1061170) [8,38].

Sequenom iPLEX for complement gene polymorphisms
Polymorphisms of complement pathway genes were investigated using the Sequenom iPLEX platform. PCR and primer extension oligonucleotides were designed using the My Sequenom Online Tools (Sequenom Inc., San Diego, CA, USA), with a target PCR fragment length of between 80 and 120 bp and target primer extension oligonucleotide of between 15 and 30 bases, according to the manufacturer's protocol. Oligonucleotide sequences are shown in S2 Table. Shrimp alkaline phosphatase was used to neutralise unincorporated dNTPs after the PCR reaction and the iPLEX Gold resin used for final conditioning after primer extension. The reaction product was nanodispensed onto an array chip by technical staff in the Queen's University Belfast Genome Core using the MassARRAY nanodispenser prior to operation of the MassAR-RAY mass spectrometer. Genotyping was then carried out using TYPER software (Sequenom). Each SNP cluster plot was inspected individually for quality of clustering and the mass spectrometry plot examined for potentially erroneous base calls, which were recalled or set to missing if a problem was observed. Data were exported from TYPER and converted to PLINK format using Microsoft Excel 2010.

Genotyping of Neisserial fHBP
A method based on the sequence variation described by Pajon et al. [23] was developed and validated by sequencing fHBP from eight cultured isolates of N. meningitidis. We used a onestep duplex of two PCR reactions in the same experiment. Subfamily A and B fHBP were  distinguished by detection of a ten-base insertion/deletion polymorphism in module A of the gene. One pair of primers, one of which was fluorescently Fam-labelled, was used to make a product of either 157 or 167 bases in length, which was detected on the ABI 3100 Genetic Analyser. Further distinguishing between variant 2 and variant 3 was by the use of three sequencespecific primers, all paired with one common, Fam-labelled primer. The three sequence-specific primers were of different lengths, and bound to the junction between modules C and D, each of which may have one of two variants. According to Pajon et al., three variants are found, and therefore, the sequence-specific primers were designed to bind to these combinations of polymorphic modules C and D. Oligonucleotide sequences are shown in S3 Table. In addition to the subfamily and variant classifications of fHBP, the method can determine the most common modular groups described in Pajon et al. [23]. The rare (<0.5%) groups VII, VIII and IX cannot each be distinguished from closely related more common modular groups ( Table 2).

Statistical Analysis
Assessment for deviation from Hardy Weinberg equilibrium was conducted using the method of Wigginton et al. [39] as implemented in PLINK v1.07 [40,41]. A P value for deviation of <0.05 in cases, controls, or overall was considered to be significant. A minimum genotyping rate per SNP of 90% and minimum genotyping rate per individual of 90% were used. Association with individual SNPs was assessed by univariate logistic regression (additive model) in PLINK. Significance for association was accepted at P < 0.05 after Bonferroni correction for multiple testing in the study of the complement pathway. Correction for multiple testing was not used for analysis of CFH as this was a replication of a previous report.
Individuals missing genotypes for any of the four CFH haplotype-tagging SNPs were excluded from haplotype analysis. Multivariate logistic regression was conducted in SPSS v19 for case-control status using the number of copies of each CFH haplotype as covariates, omitting one haplotype (D) as a reference variable to avoid multicollinearity.
Cycle threshold was compared between CFH haplotype B carriers and non-carriers using Student's T test in R v3.1.1. Age was compared between fHBP Subfamily A and fHBP Subfamily B using the independent samples median test in SPSS.

CFH
The genotyping rate was 99.7%. No markers deviated significantly from Hardy Weinberg equilibrium. Two cases failed genotyping entirely, no other individuals were excluded for incomplete genotyping and no markers were excluded because of low genotyping.

Factor H-Binding Protein Characteristics in the Case Population
The frequencies of the Neisserial fHBP variants are shown (Table 5). These are similar to those reported by Pajon et al. [23].
The median age of cases with subfamily A fHBP N. meningitidis was one year and that of cases with subfamily B fHBP N. meningitidis was two years (independent-samples median test P = 0.025). The mean cycle threshold for detection of ctrA for cases with subfamily A fHBP N. meningitidis was 29.7 and that for subfamily B fHBP N. meningitidis was 29.2.

Susceptibility to N. meningitidis with Subfamily A and Subfamily B fHBP
CFH Haplotype B was associated with a statistically significant protective effect against the rarer subfamily A fHBP-expressing N. meningitidis infection, but was not associated with protection against the more common subfamily B fHBP-expressing N. meningitidis (Table 6). However, there was no significant difference between the distribution of fHBP subfamily N. meningitidis infection in haplotype B heterozygote cases and wild-type homozygote cases (Pearson chi-square P = 0.11; Table 7).

Complement Pathway
Twenty nine individuals were excluded because of a low genotyping rate, and no SNPs were excluded because of low genotyping leaving 281 cases and 5,199 controls. The remaining genotyping rate was 99.8%. Five SNPs were excluded due to significant deviation from Hardy Weinberg equilibrium in controls, all of which were in CD46.
One SNP in each of CD46 (rs2796278), C5 (rs17216529), C8A (rs17300936) and C8B (rs12085435) was associated with the phenotype before correction for multiple testing ( Table 8). The association with rs12085435 in C8B was significant after Bonferroni correction for 21 tests (P = 0.03), and the other SNP associations were non-significant. Both C8 SNPs were independently associated with the disease phenotype in multivariate logistic regression (Table 9).

Discussion
Our study explored the characteristics of the association between meningococcal disease and CFH polymorphism, the relationship between fHBP subfamily and this association and the effects of other polymorphisms of complement system genes on disease risk. Our findings provide further evidence supporting the association between CFH polymorphism and invasive meningococcal disease reported by Davila et al. [11] but as we used the same control group as that study, this is not a full independent replication of the finding. Haplotypic exploration revealed that the association is due to only one haplotype (B), which does not carry any of the known major functional variants associated with AMD. The haplotypes that carry deletion of CFHR3-CFHR1 (haplotype C) and Y402H (haplotype A) were not associated in a multivariate logistic regression. The protective haplotype B was not associated with any difference in the bacterial load (measured by cycle threshold), suggesting that while disease risk is altered by carrying this variant, severity of disease is not. There was no significant difference in age of cases between carriers and non-carriers of haplotype B, suggesting that the protective mechanism does not affect the age of onset. There was no significant difference between the distribution of the fHBP subfamilies in individuals who have a copy of CFH haplotype B and those who do not, which suggests that CFH haplotype variation does not alter the risk of infection with meningococci with fHBP subfamilies A and B differently.
The patients affected by N. meningitidis with fHBP subfamily A were significantly younger than those who had infection with N. meningitidis with fHBP subfamily B.
Our investigation of other complement pathway polymorphisms suggests that that a coding polymorphism in C8B may be associated with susceptibility to invasive meningococcal disease. This is consistent with the observation that terminal complement component deficiencies increase risk of invasive meningococcal disease [42]. This possibility requires replication in an independent cohort.
Our study is unique in integrating human and bacteriological genomic information to assess the effects of variation of genes that produce interacting proteins. The most important limitation of this study is the lack of an independent replication group to confirm the association at C8B. Our study used the same control group as Davila et al. [11], meaning that we have not reported a full independent replication of the association at CFH. We conducted our own laboratory experiments to genotype meningococcal disease cases and compared them to controls that were genotyped using a different method, at a different time. This presents risk of a systematic genotyping error that could result in a false association. Independent replication is therefore vital. The exploration of the effect of CFH haplotype on disease severity (indicated by CtrA cycle threshold) and of fHBP on severity and age, are not affected by these limitations because these are analyses of cases only. The functional basis for the relationship between CFH polymorphism and susceptibility to invasive meningococcal disease is not yet understood: It is likely that it relates to the interaction between CFH or CFH-related proteins and N. meningitidis. As meningococcal disease becomes less common, the focus of research may change to questions of immunity and vaccination: It would be most interesting to understand whether vaccine fHBP interacts differently with different variants in host CFH and CFHR proteins following vaccination in humans, and whether vaccination results in complement activation through transient sequestration of CFH. Costa et al. recently suggested that fHBP with low affinity for CFH should be explored in future development of fHBP-based vaccines to increase immunogenicity and reduce the chance of autoantibody formation to CFH [43]. It is conceivable that carriage of the protective CFH haplotype could influence the effectiveness or side-effects of the vaccine, such as fever, which is common following meningococcal B vaccination [44].
The association between C8B rs12085435 and risk of meningococcal disease in our study is in keeping with the effect of inherited terminal complement deficiencies on susceptibility to meningococcal disease [42]. Independent replication will be key to establishing whether C8B is a second complement gene associated with meningococcal disease risk.
Supporting Information S1

Acknowledgments
We thank Dr Steve Gray for preparation of samples.
This study makes use of data generated by the Wellcome Trust Case Control Consortium. A full list of the investigators who contributed to the generation of the data is available from www.wtccc.org.uk.