|Home | About | Journals | Submit | Contact Us | Français|
► Hypothesis tested: common sequence variants in complement are associated with aHUS. ► SNPs in complement genes were genotyped in two independent aHUS cohorts. ► Only SNPs within CFH, CD46 and the CFHRs showed a replicable association with aHUS. ► Complement regulators variability has the main role in determining the disease.
It is well established that common genetic variants in CFH, CD46 and the CFHRs are additional risk factors for the development of aHUS. To examine the hypothesis that common variants in other complement genes have a similar effect we genotyped 501 SNPs in 47 complement genes in 94 aHUS patients from Newcastle, 126 aHUS patients from Paris, 374 UK controls and 165 French controls. We replicated the associations in CFH, CD46 and the CFHRs but found no association with any other complement gene. The strongest associations replicated in both cohorts were found for four SNPs within CD46 (p-value < 10−3) and five SNPs within CFH (p-value <5 × 10−3). Significant replicable associations with single SNPs in CFHR2, CFHR4 and an intergenic SNP (CR1–CD46) were also found. Analysis of the Paris cohort showed that the association with CD46 SNPs was only present in those patients with complement mutations. Haplotype analysis showed at-risk and protective haplotypes in both CD46 and CFH. The CD46 haplotype was only disease-associated in those patients with mutations.
Atypical hemolytic uremic syndrome (aHUS) is a disease characterised by excessive complement activation in the microvasculature (Noris and Remuzzi, 2009). Inherited and acquired abnormalities affecting components of the alternative complement pathway are found in ~70% of patients (Noris et al., 2010). These include mutations in the genes encoding both complement regulators (factor H (Warwicker et al., 1998; Caprioli et al., 2001; Perez-Caballero et al., 2001; Richards et al., 2001; Sanchez-Corral et al., 2002; Venables et al., 2006), factor I (Fremeaux-Bacchi et al., 2004; Kavanagh et al., 2005, 2008; Caprioli et al., 2006; Nilsson et al., 2010), membrane cofactor protein (Noris et al., 2003; Richards et al., 2003) and thrombomodulin (Delvaeye et al., 2009)) and activators (factor B (Goicoechea de Jorge et al., 2007) and C3 (Fremeaux-Bacchi et al., 2008)); and autoantibodies against factor H (Moore et al., 2010). The penetrance of aHUS in the familial form of the disease is ~60% (Caprioli et al., 2006). This is because multiple hits are necessary for the disease to manifest including a trigger, mutations (rare genetic variant) and at-risk haplotypes (common genetic variant) in complement genes. To date common genetic variants in CFH, CD46 and the CFHRs (which all lie with the regulators of complement activation – RCA gene cluster at 1q32) have been reported to be risk factors for the development of aHUS (Caprioli et al., 2003; Esparza-Gordillo et al., 2005; Fremeaux-Bacchi et al., 2005; Pickering et al., 2007; Abarrategui-Garrido et al., 2009). In this study we have examined the hypothesis that common genetic variants in other complement genes apart from these have a similar effect. To do this we have genotyped tagged single nucleotide polymorphisms (SNPs) in 47 complement genes in two cohorts of aHUS patients. We have not found a significant association in any other complement gene apart from CFH, CD46 and the CFHRs.
Atypical HUS was diagnosed clinically within two unrelated cohorts. The study was approved by the Northern and Yorkshire Multi-Centre Research Ethics Committee and informed consent obtained. The first cohort of patients comprised 94 aHUS cases from Newcastle upon Tyne, United Kingdom while the second included 126 aHUS patients from Paris, France. Clinical and demographic details of the subjects were recorded within two different databases. Mutation screening was performed on patients in local clinical laboratories.
We genotyped 374 DNA samples from healthy individuals within the Wellcome Trust Case Control Consortium (Burton et al., 2007; WTCCC, 2007) as a control population for the Newcastle cohort. As a control for the French aHUS cohort DNA samples from 165 healthy controls from France were used.
We selected 501 SNPs within 47 complement genes (Supplementary Table 1). For tagging purposes, we clustered together genes with inter-gene distances less than 200 kb and used a hybrid tagging-plus-putative-function selection strategy similar in design to Cavalleri et al. (2007). This strategy allowed us to maximise the efficiency of SNP coverage. First, using TAMAL v2 program (Hemminger et al., 2006) SNPs with supposed functional effect and with a minor allele frequency (MAF) greater than 5% in the HapMap Phase 2 European-ancestry sample were chosen. SNPs were also selected if they met at least one of the following criteria: (1) non-synonymous coding SNP; (2) SNP within predicted promoter region; (3) SNP within evolutionarily conserved region; (4) SNP within predicted transcription factor binding site; (5) SNP within conserved miRNA target; (6) SNP within a splice region. In addition, 106 putative functional SNPs were added following a literature search.
All selected SNPs were tested for linkage disequilibrium using the Tagger program (de Bakker et al., 2005) and further SNPs were chosen to ensure an r2 greater than 0.9. Genotyping was performed by Sequenom using the iPlex platform, which utilises a multiplex PCR system followed by a single base pair extension step. The primer extension products are then analysed by MALDI-TOF mass spectrometry.
All statistical analyses were done with Plink Software version 1.05 (http://pngu.mgh.harvard.edu/~purcell/plink/) and R statistical Software version 2.11 (http://www.r-project.org/). Linkage disequilibrium patterns between the SNPs were analysed using the Software HAPLOVIEW (Barrett et al., 2005). Genotyped SNPs and DNA samples were subjected to quality control (QC) procedures.
Sample duplication was assessed by estimation of pair wise identity by descent in PLINK (Pi-hat statistic). If a duplicate pair was detected (Pi-hat bigger than 0.9) both samples were excluded. Samples with a genotyping call rate below 95% were excluded as were SNPs with a missing genotype rate more than 5% and SNPs with an exact-test p-value for departure from Hardy–Weinberg equilibrium in controls with a p-value < 10−3. SNPs with a minor allele frequency less than 0.02 in our data set were also excluded.
All SNPs and samples that passed the quality control steps were assessed for association analysis. Statistical analysis on the Newcastle and Paris datasets was performed separately, and then data combined for a stratified analysis. A further statistical analysis was performed only in the Paris cohort for cases with and without complement gene mutations. The Cochran–Armitage trend test was used to establish genetic association in each cohort. Stratified analysis was calculated using the Cochran–Mantel–Haenszel (CMH) test. A Breslow–Day test was carried out to assess homogeneity of the common odds ratio across the two cohorts generated by the CMH test. Centre of origin (Newcastle/Paris) was added as a covariate in the stratified analyses. Haplotype-based association testing was performed using the --hap-logistic option in Plink. This option allows the estimation of an Odds Ratio for each haplotype relative to all other haplotypes combined.
We undertook a candidate genes analysis of 501 SNPs in 47 genes of the complement cascade. We genotyped 220 aHUS patients and 539 controls. After stringent quality control filtering, 444 SNPs within 419 samples (82 cases, 337 controls) in the Newcastle dataset and 461 SNPs within 251 (111 cases, 140 controls) samples in the Paris dataset were analysed. The association between each SNP and aHUS risk was tested using the Cochran–Armitage trend test and possible population stratification was corrected using genomic control (Devlin et al., 2001). The genomic inflation factor for the trend test was calculated for both cohorts separately (λGC = 1.28 for Newcastle cohort; λGC = 1.24 for Paris cohort).
Single marker association analysis in the Newcastle and Paris cohorts was performed on the two datasets separately. When common SNPs typed in both cohorts were compared only SNPs located in or close to CD46, CFH and the CFHRs showed a replicable association with aHUS.
CD46 showed a stronger association with aHUS than CFH. Four out of eight SNPs genotyped (rs2796275, rs2796278, rs10449303 and rs7144) displayed evidence of association with a p-value of < 10−3 in both cohorts.
We genotyped twenty SNPs within CFH, and five (rs1329423, rs12405238, rs3753396, rs424535, and rs1065489) showed an allelic association with aHUS with a p-value < 5 × 10−3 for the Newcastle cohort and a p-value <5 × 10−5 for the Paris Cohort.
In addition we found a common association for a marker in CFHR2 (rs9427934, p-value = 4.24 × 10−3 in Newcastle and p-value = 5.54 × 10−5 in Paris cohorts) and another in CFHR4 (rs3795341, p-value = 2.13 × 10−3 in Newcastle and p-value = 1.52 × 10−5 in Paris cohorts).
The most significant association, however was found for a SNP mapping to an intergenic position on chromosome 1 between CR1 and CD46 (rs2761434, p-value of 3.05 × 10−5 for the Newcastle cohort and a p-value of 8.53 × 10−6 for the Paris cohort). A Manhattan plot for the association study confirmed multiple associations at chromosome 1 in both cohorts (Fig. 1). The complete analysis of the Newcastle and Paris cohorts is provided in Supplementary Tables 2 and 3 respectively.
A quantile–quantile plot of adjusted observed p-values for association between cases and controls showed a remarkable deviation from the null distribution within both cohorts (Fig. 2), which could be ascribed to the strong association observed within SNPs in strong linkage disequilibrium.
In order to further validate these associations a stratified analysis was carried out. Cases and controls from the two cohorts were combined together and stratified according to the geographic origin. Results are shown in Table 1.
To investigate the interaction between SNPs and non-synonymous mutations in complement genes, we considered patients with and without identified mutations separately. For this analysis we focused on the Paris cohort because all the individuals within this cohort had been screened for mutations in CFH, CFI, CD46, CFB and C3. Table 2 summarises the mutations detected.
After application of filtering criteria (as described in Section 2.3), we analysed 467 SNPs in 75 cases with mutations and 470 SNPs within 36 cases without mutations. Both sets were compared with 140 controls from France.
Significant differences between the two subsets were observed. For SNPs within CD46 the strength of the association with aHUS increased when the patient cohort with mutations was analysed separately. In contrast no significant association between aHUS and CD46 SNPs was found in cases without a complement gene mutation (Table 3). An association was found for SNPs within CFH, CFHR2 and CFHR4 in both subsets.
In the Newcastle cohort (data not shown) the prevalence of mutations in aHUS patients was lower (Paris 66%, Newcastle 33%).
Consistent with the Paris cohort for all of the SNPs in CD46, which were associated with aHUS the p-values were lower (10–100 fold) in the group with mutations, suggesting that the CD46 SNPs have a stronger effect in the presence of a known mutation. This is consistent with the data from the Paris cohort.
No association between aHUS and a SNP in any other complement gene was replicated in both populations. In the Newcastle cohort there were two SNPs both within the CD11b gene, which were associated with aHUS (rs9937837 and rs7499077; p-value < 10−3). In the Paris cohort there were additional SNPs both in and outside the RCA cluster that associated with aHUS but were not replicated in the Newcastle dataset. CFHR4 contained 4 further SNPs associated with aHUS (p-value < 5 × 10−3) while CFH showed a further 7. Associated SNPs were also found within CR2 (rs12032512; p-value = 1.74 × 10−3) and C8G (rs2071006; p-value < 10−2).
A list of 10 SNPs with the lowest p-values within each cohort is shown in Table 4.
Linkage disequilibrium (LD) analyses were performed with the two control groups both separately (data not shown) and combined. As the results of both analyses were similar, only the results from combined analysis are discussed.
Linkage disequilibrium between the 21 SNPs within CFH and 10 SNPs within CD46 gene was estimated. The LD plot for CD46 (Fig. 3) showed high correlation (r2 > 0.6) in a region that spans approximately 12 kb on chromosome 1. This region contains three SNPs (rs2796278, rs10449303 and rs7144) strongly associated with aHUS. If D′ as a linkage disequilibrium coefficient is used the LD plot shows one block spanning over 50 kb on chromosome 1 (from rs2761434 to rs7144).
The LD plot for CFH (Fig. 4), showed a region of 19 kb with high correlation (r2 > 0.6) containing three SNPs (rs7524776, rs6680396 and rs800292). No other region displayed a correlation coefficient > 0.6 between three or more SNPs. A reason for the low correlation coefficients may be the initial choice of the SNPs to be genotyped. If the D′ coefficient is used a block spanning over 63 kb can be defined (from rs1329423 to rs1065489). This block, as shown in Fig. 4, contains the 5 SNPs associated with aHUS.
Haplotype analysis of the SNPs in CD46 and CFH associated with aHUS was performed using the --hap-logistic option in Plink for the two case control cohorts. Five possible haplotypes for CD46 were observed in both cohorts. The odds ratio for developing aHUS was calculated for each haplotype. A significant difference between cases and controls were observed for two haplotypes in both cohorts. The haplotype CD46ACAGC (Table 5) is strongly associated with the risk of the disease (odds ratio 2.4 and 2.6 in Newcastle and Paris cohorts respectively; p-value < 10−6) while the haplotype CD46GTCAT has a protective effect (odds ratio 0.5 in both cohorts; p-value < 5 × 10−4) (Table 5). As would be predicted from the single SNP analysis the presence of the ‘at risk’ haplotype was only important in the presence of a complement gene mutation (Table 6).
A similar trend was found in cases and controls of both cohorts for the CFH haplotype. We identified one haplotype block CFHGTGAT strongly associated with the risk of aHUS (odds ratio 2.2 and 2.9 in Newcastle and Paris cohorts respectively; p-value < 10−4) and one haplotype CFHAGATG with a protective effect (odds ratio 0.5 and 0.4 in Newcastle and Paris cohorts respectively; p-value < 10−3) (Table 5).
In this study we have examined the hypothesis that common genetic variants in complement genes apart from CFH, CD46 and the CFHRs are associated with aHUS. In two independent cohorts of aHUS patients we have genotyped 501 SNPs in 47 complement genes. We were unable to find a replicable association in the two cohorts in any other complement gene apart from CFH, CD46 and the CFHRs. In particular there was no replicable association with CFI, C3 and CFB despite these being genes, like CFH and CD46, where aHUS associated mutations have been detected. In both cohorts we have confirmed that CFH and CD46 alleles are significantly more frequent in aHUS patients. Caprioli et al. (2003) first reported this for CFH in 2003 showing that three CFH SNPs −331C>T (rs3753394), c.2016A>G Gln672Gln (rs3753396) and c.2808G>T Glu936Asp (rs1065489) were associated with aHUS. In 2005 we confirmed these findings in two independent cohorts of patients and also examined the allele frequency of six CD46 SNPs (Fremeaux-Bacchi et al., 2005). One of these (CD46 c.4070T>C, rs7144) was associated with aHUS in both cohorts. Subsequently Esparza-Gordillo et al. (2005) defined a haplotype in CD46 (CD46GGAAC) that was associated with an increased risk of aHUS. CD46GGAAC is defined by the following SNPs −652A>G (rs2796267), −366A>G (rs2796268), IVS9 −78G>A (rs1962149), IVS12 +638G>A (rs859705) and c.4070T>C (rs7144) where the at-risk alleles are in bold. Pickering et al. (2007) also subsequently defined a haplotype in CFH (CFH-H3) associated with an increased risk of aHUS. This was subsequently updated by Rodriguez de Cordoba and de Jorge (2008). CFH-H3 is defined by the following SNPs −331C>T (rs3753394), c.184G>A Val62Ile (rs800292), c.1204T>C p.Tyr402His (rs1061170), c.2016A>G p.Gln672Gln (rs3753396), IVS15 −543G>A intron 15 (rs1410996) and c.2808G>T p.Glu936Asp (rs1065489) where the at-risk alleles are in bold. This includes the three SNPs originally reported by Caprioli et al. (2003). In the same study Pickering et al. (2007) also defined a further CFH haplotype (CFH-H2) that protects against the development of aHUS. CFH-H2 is defined by the following SNPs −331C>T (rs3753394), c.184G>A Val62Ile (rs800292), c.1204T>C p.Tyr402His (rs1061170), c.2016A>G p.Gln672Gln (rs3753396), IVS15 −543G>A intron 15 (rs1410996) and c.2808G>T p.Glu936Asp (rs1065489) where the protective alleles are in bold.
In this study we have confirmed the presence of an at-risk CD46 haplotype but in addition have identified a protective haplotype. It is probable that the at-risk CD46 haplotype that we have defined in this study is the same as the previously described CD46GGAAC. However, only one of the SNPs (CD46 c.4070T>C; rs7144) that define CD46GGAAC also defines our at-risk haplotype. While our at-risk haplotype is, like CD46GGAAC, defined by the C allele of rs7144 there are two other rare haplotypes that are also defined by the C allele. Both CD46 haplotypes occur in a region in strong linkage disequilibrium spanning over 59 kb of the CD46 gene (Supplementary Fig. 1).
Like Pickering et al. (2007) we have identified both an at-risk and a protective CFH haplotype. Two of the CFH SNPs (rs3753396 and rs1065489) that define our at-risk and protective CFH haplotypes were also used by Pickering et al. to define CFH-H3 and CFH-H2. The alleles for both rs3753396 and rs1065489 that define the at-risk and protective CFH haplotypes are the same in our study and that of Pickering et al. Thus, it is again probable that these are the same haplotypes. In addition to the SNPs in CD46 and CFH there were SNPs within CFHR2 and CFHR4 (rs9427934 and rs3795341 respectively) that demonstrated a replicable association with aHUS. This would suggest that other genes in proximity to CD46 and CFH within the regulators of complement activation (RCA) cluster of genes at 1q32, particularly the CFHRs, are susceptibility factors for aHUS. In this study we did not examine the allele frequency of the novel CFHR1 polymorphism described by Abarrategui-Garrido et al. (2009) where the at-risk CFHR1*B has greater sequence similarity to factor H and may thus compete with it. Neither have we examined the frequency of the well described CFHR3/1 deletion that is associated, in homozygosity, with factor H autoantibodies (Moore et al., 2010). In contrast there were no replicable associations between SNPs outside the RCA gene cluster with aHUS. It is particularly interesting that we found no replicable association between aHUS and variants in CFI, CFB and C3. CFB and C3 are genes in which aHUS mutations have been found but they are also all genes, which encode complement activators. We speculate that it is naturally occurring variability in complement regulators rather than activators that has the predominant effect in determining manifestation of aHUS.
When data from the two populations are combined, statistically significant associations between other SNPs and aHUS can be identified. Whether these are real associations require further replication studies with greater power.
When a subgroup analysis in patients with or without complement gene mutations was performed, SNPs within CD46 and CFH behaved differently. The significance of SNPs within CD46 increased in the group of patients with a gene mutation but significance was lost in patients without mutations. This suggests that the “at risk” SNPs (and haplotype blocks) within CD46 are most important when they co-exist with a complement gene mutation (Esparza-Gordillo et al., 2005). In contrast SNPs within CFH were statistically associated with aHUS in both populations. This is at variance with our previous study (Fremeaux-Bacchi et al., 2005) which showed that the association between variations in CFH and CD46 and aHUS was present irrespective of the presence of a mutation but is consistent with the findings of Esparza-Gordillo et al. (2005) who showed that the association with CD46 was due almost exclusively to the aHUS patients with mutations in CFH, CD46 and CFI.
In this study we have, therefore, confirmed and emphasised the pivotal importance of common genetic variants in CFH and CD46 in predisposing to aHUS. That common genetic variants in complement genes can collectively also functionally determine a disease such as aHUS has recently been established. Heurich et al. (2011) have shown that non-synonymous variants encoding isoforms in C3 (p.Arg102Gly), factor B (Arg32Gln) and factor H (Val62Ile) can at a protein level combine functionally to influence susceptibility to complement driven diseases such as aHUS. This functional combination they suggest determines an individual's “complotype”. While we have not in this study established that common genetic variants in CFB, CFI and C3 on their own are associated with aHUS we cannot exclude the possibility of an additive effect of such variants. That the functional activity of the multiple activators and regulators of the complement pathway can combine to determine susceptibility to diseases such as aHUS is a concept that is worthy of further study.
This study was supported by grant provided by the Medical Research Council (G0701325). The funding source had no involvement in study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the paper for publication.
L.E. performed the research, collected data, performed statistical analysis and co-wrote the paper. T.H.J.G. designed the research, interpreted data and reviewed the manuscript. L.S. performed research and collected data. M.E.W. was involved in the development of the complement SNP panel and contributed analytical tools. S.H. Sacks was involved in the development of the complement SNP panel and contributed analytical tools. H.J.C. analysed and interpreted data and reviewed the manuscript. V.F.-B. interpreted data, provided vital reagents and reviewed the manuscript. N.S.S. designed the research, interpreted data and co-wrote the manuscript.
The authors declare no competing interests.
We acknowledge use of DNA from The UK Blood Services collection of Common Controls (UKBS collection), funded by the Wellcome Trust grant 076113/C/04/Z, by the Juvenile Diabetes Research Foundation grant WT061858, and by the National Institute of Health Research of England. The collection was established as part of the Wellcome Trust Case-Control Consortium.