Haemolytic-uraemic syndrome (HUS) is characterised by the triad of thrombocytopenia, Coomb’s test negative microangiopathic haemolytic anaemia, and acute renal failure.^1,2 HUS is classified as either (D+) when it is associated with a preceding diarrhoeal illness which in most is cause by infection with Ecoli O157 or less commonly as non-diarrhoeal associated (D−) (also called “atypical”, hence aHUS). aHUS may be sporadic or familial. In both types, mutations have been reported in regulators of the complement pathway including factor H (CFH), membrane cofactor protein (MCP, CD46),^{3,4,5,6,7,8,9,10,11} and the serine protease factor I (IF).¹² Such mutations result in impaired protection of host surfaces against complement activation^13,14 and it is likely that they predispose to rather than directly cause a thrombotic microangiopathy. In this situation endothelial activation secondary to injury is maintained by excessive complement activation.¹⁵ A characteristic feature of both MCP and CFH associated HUS is variable penetrance and inheritance. The penetrance of the disease phenotype in our panels of families is approximately 50%. One possible explanation for this is that functional variants in complement proteins act as modifiers. In support of this is the finding of an association between CFH single nucleotide polymorphisms (SNPs) and aHUS,¹⁶ and a recent study which identified a specific SNP haplotype block spanning MCP which is over-represented in aHUS patients.¹⁷ To extend these observations we have examined CFH and MCP SNPs in two cohorts of aHUS patients.

METHODS

A clinical diagnosis of aHUS was made in all the patients included in this study. Appropriate ethics approval was given for both cohorts to be studied and all subjects gave informed consent. There were 75 patients in the Newcastle cohort and 77 in the Paris cohort, some of whom have been reported previously.^{3,5,6,9,11,12,18–24} Of the 75 patients in the Newcastle cohort there were four families in which more than one family member was included in the cohort. In three families there were two siblings included and in one there were three siblings. Of the 77 patients in the Paris cohort there were again four families in which more than one family member was included in the cohort. In three families there were two siblings included and in one there were three siblings.

In the Newcastle cohort 15 had a CFH mutation, four had an MCP mutation, and three had an IF mutation; in the Paris cohort 21 patients had a CFH mutation, eight had an IF mutation, 10 had an MCP mutation, and one had anti-CFH antibodies. Of the 21 patients with a CFH mutation, eight had ∼50% and three had no detectable antigenic level. Eight patients had a mutation located in the C-terminal region, with normal CFH levels. While the functional significance of each of these mutations has not been examined, their location would suggest that they act in a way similar to those previously studied in detail.^13,14 Of the 18 patients with either an FI or MCP mutation, in ∼60% there was evidence of a decreased antigenic level of the protein.

One hundred locally recruited normal controls were studied for the Newcastle cohort and 84 for the Paris cohort. Both cohorts and their respective control populations were predominantly white (Newcastle cohort 65/75; Paris cohort 74/77).

The SNPs analysed in the Newcastle cohort were CFH (c.−257C→T, promoter; c.2089A→G, synonymous; c.2881G→T, E963D) and MCP (IVS12+638A→G; c.2232T→C, 3′ UTR). DNA was prepared from peripheral blood according to standard procedures. Regions of DNA containing the SNPs of interest were amplified by polymerase chain reaction (PCR). Genotyping was carried out through a primer extension reaction, using the ABI PRISM SNaPshot ddNTP primer extension kit (Applied Biosystems, Courtaboeuf, France) and for subsequent detection we employed an ABI PRISM 3100 genetic analyser (Applied Biosystems). SNaPshot reactions were carried out on pooled PCR products according to the manufacturer’s instructions, using pooled SNaPshot primers at the concentrations described in table 1. Primer sequences and conditions for PCR and SNaPshot analysis are also shown in table 1. In addition, ambiguous SNPs were verified by direct sequencing using DYEnamic ET dye terminator cycle sequencing (Amersham, UK), as described by the manufacturer.

The SNPs analysed in the Paris cohort were CFH (c.−257C→T, promoter; c.2089A→G, synonymous; c.2881G→T, E963D) and MCP (c.−547A→G, promoter; c.−261A→G, promoter; IVS8+23T→G; IVS9-78G→A; c.2232T→C, 3′ UTR). DNA was extracted from whole blood using the proteinase K/phenol method.²⁵ Genomic DNA was amplified using the primer sequences and conditions shown in table 2.¹¹ PCR products were purified using Multiscreen plates according to the manufacturer’s instructions (Millipore, Molsheim, France). Direct DNA sequencing of the purified PCR products was then carried out by the Dye terminator cycle sequencing method (Applied Biosystems, Courtaboeuf, France) using a 96 capillary Sequencer 3700. Sequence analyses were done using Sequencher software (Gene Codes Corporation, Ann Arbor, Michigan, USA).

The NCBI SNP ID numbers, in parentheses, are CFH −257C→T (rs3753394), CFH c.2089A→G (rs3753396), CFH c.2881G→T (rs1065489), MCP-547A→G (rs2796267), MCP c.IVS8+23T→G (rs2724374), MCP c.IVS9-78G→A (rs1962149), MCP c.IVS12+638 (rs859705), and MCP c.2232T→C (rs7144).

The differences in genotype frequency between HUS patients in the two cohorts and the respective control individuals were tested using either χ², or if there were less than five observations in a cell, Fisher’s exact test. As the SNPs are not assorting independently it would not be appropriate to apply a Bonferroni correction equal to the total number of SNPs studied in each cohort. Haplotype analysis (FUGUE) showed that there were three CFH and two MCP haplotypes with a frequency greater than 10% in both cohorts. We therefore applied a Bonferroni correction of (0.05/5) giving a significance value of 0.01.

Comparison of haplotypes was undertaken using FUGUE-CC (Goncalo Abecasis, Center for Statistical Genetics, University of Michigan, USA). This allows significance to be computed by analysis of random permutations of the data.

RESULTS

The genotype and allele frequency for the two cohorts is shown in table 3. In both cohorts there was an association between CFH alleles and HUS. The same CFH SNPs were analysed in both cohorts. There was also an association between MCP alleles and HUS in both cohorts. The results for the one SNP (c.2232T→C) analysed in both cohorts showed a strong association (p<0.001) in the Paris cohort but failed to reach a significance value of <0.01 in the Newcastle cohort (p = 0.012).

The allele frequency in those with and without known mutations in CFH, MCP, and IF is shown in table 4. For CFH in the Newcastle cohort c.−257T c.2089G, c.2881T were associated with atypical HUS in those not known to have a mutation but not in those known to have a mutation. For CFH in the Paris cohort c.−257T c.2089G, c.2881T were associated with atypical HUS only in those known to have a mutation. For MCP the association was present in both those with and those without an identified mutation for all the SNPs in the Paris cohort. In the Newcastle cohort c.IVS12+638A was associated with atypical HUS in those without a mutation.

The haplotypes generated by FUGUE are shown in tables 5–7. For both the Newcastle and the Paris cohort there was a significant difference for both CFH and MCP haplotype frequency in the HUS patients and controls.

The genotype and haplotype results for both cohorts are internally consistent in that an increase in frequency of the rarer allele and haplotype for both MCP and CFH was associated with HUS.

DISCUSSION

In this study we showed a significant difference in two independent cohorts of atypical HUS patients in both allele frequency and haplotypes for two complement regulatory genes, CFH and MCP. Mutations in both these genes have been described in aHUS patients. However, the inheritance and penetrance seen with mutations in both these is variable. For instance in two families with the same MCP mutation (S206P), only homozygotes are affected in one whereas in the other heterozygotes are affected.⁹ Moreover, the series reported to date show that the penetrance of CFH associated HUS is approximately 50%. This suggests that other factors are modifying the inheritance and penetrance. Both CFH and MCP belong to a cluster of genes located at 1q32 which are involved in complement regulation, the so called RCA (regulators of complement activation) cluster. Other members of this group include decay accelerating factor (DAF), complement receptor 1 (CR1), C4 binding protein (C4BP), and five factor H related proteins (FHR1–5). To date mutations have only been found in CFH and MCP but it is possible that genetic variability in these other regulators could be acting as modifiers for the development of HUS.

Caprioli et al have previously reported that the CFH alleles −257T, c. 2089G, and c. 2881T are significantly more common in patients with atypical HUS. The results from both the Newcastle and Paris cohorts support this observation.¹⁶ They found that this was true both for patients with a CFH mutation and for those without. In the Newcastle cohort we found that −257T c.2089G, and c.2881T were more frequent in those without known mutations in CFH, MCP, and IF. In the Paris cohort, the reverse was seen in that −257T, c. 2089G, and c. 2881T were more frequent in those with mutations. Caution must therefore be exercised in interpreting subgroup analysis such as this where numbers may be inadequate. It is not yet known whether c.−257C→T or c.2881G→T are functionally significant. c.2089A→G is a synonymous change. c.−257T is located in a putative NF –κB binding sequence of the CFH promoter²⁶ and it is known that CFH expression is upregulated by interferon γ, providing a possible link. c.2881G→T changes a glutamate to an aspartate in CCP16 of CFH. We have now not only confirmed Caprioli’s observation but also extended it in two independent cohorts to show that genetic variability in MCP is also associated with atypical HUS. This confirms the recent finding by Esparza-Godilla et al that a specific SNP haplotype block which includes MCP was associated with aHUS.¹⁷ In the Paris cohort this strong association was present in both those with and those without known mutations in CFH, MCP, and IF. In contrast, the association was only seen in those with known mutations in the study of Esparza-Godilla. This discrepancy emphasises the need to be cautious in interpreting such data. In all, six MCP SNPs were examined in the two cohorts; of these three are intronic, two are in the promoter, and one is in the 3′ UTR. It is possible that the promoter SNPs or the 3′ UTR SNP are functionally significant. In support of this is the recent study from Esparza-Gordillo which showed that MCP −261G disrupts a potentially functional CBF-1/RBP-Jk binding site. Transient transfection showed that this was associated with a 25% lower transcriptional activity.¹⁷ It is also possible that a combination of factors within the haplotype block results in a functional effect. Alternatively, these markers may simply be surrogates for another untested SNP in the vicinity.

Conclusion

This study emphasises the importance of variability in CFH and MCP as a modifier for the development of atypical HUS. The results suggest that complement regulatory genes in the RCA cluster are acting in a coordinated manner to prevent host cell damage and that perturbations of this network in the face of endothelial injury will lead to a thrombotic microangiopathy.