Alzheimer's disease (AD) is a complex disease that is likely influenced by many genetic and environmental factors. Recent studies using meta-analyses and genome-wide association studies (GWAS) have provided increasing evidence for new genetic risk factors (Beecham and others 2009; Bertram and others 2008; Carrasquillo and others 2009; Coon and others 2007; Feulner and others 2009; Harold and others 2009; Lambert and others 2009; Li and others 2008). Evidence from AlzGene (alzgene.org) meta-analyses provides support for several risk variants with small effect sizes (Bertram and others 2007). Two recent studies investigated 29 such variants from the Alzgene meta-analyses for association in a large family based sample (Schjeide and others 2009) and in samples in which cerebrospinal fluid (CSF) biomarkers have been measured including amyloid-beta (Aβ) levels (Kauwe and others 2009). Among the consistent findings, one SNP in TF, rs1049296 that results in a missense coding polymorphism (P589S), showed significant association in both studies (Kauwe and others 2009; Schjeide and others 2009). Like many other genetic associations, results from various studies with rs1049296 have yielded both positive (Namekata and others 1997; Robson and others 2004; Schjeide and others 2009; Van Landeghem and others 1998; van Rensburg and others 1993; Zambenedetti and others 2003) and negative results (Blazquez and others 2007; Emahazion and others 2001; Hussain and others 2002; Kim and others 2001; Lleo and others 2002; Reiman and others 2007; Rondeau and others 2006). Such inconsistency may indicate that the association is spurious, or that the studies lack statistical power (Bertram and Tanzi 2004). It has also been suggested that lack of replication in genetic association studies is not surprising given the extent of genetic and environmental heterogeneity (Gorroochurn and others 2007) and may even be a “signature of epistasis” (Moore and Williams 2005; Wade 2001). Evidence for epistatic interaction between APOE e4 and genetic variation in BACE has been consistently replicated, though the nature of the interaction has yet to be characterized (Combarros and others 2008). It has also been reported that a synergy between rs1049296 and rs1800562 in the hemochromatosis gene (HFE) has strong association with risk for AD, with individuals that carry the minor allele at both loci having 5-fold greater risk for disease using both Synergy Factor Analysis (SFA) and logistic regression (Robson and others 2004). Both of these variants are amino acid substitutions (rs1049296 is P589S; rs1800562 is C282Y). In this study we attempt to replicate the report of epistasis between rs1049296 and rs1800562 and association with risk for LOAD in a total of 1166 cases and 1404 controls from three European and European American samples.

The case control series for this study were collected at three different sites. Basic sample characteristics for each series are shown in table I. The Washington University (WU) case-control series used in this study was collected through the WU Alzheimer's Disease Research Center (ADRC) patient registry. Cases in this series received a diagnosis of dementia of the Alzheimer's type (DAT), using criteria equivalent to the National Institute of Neurological and Communication Disorders and Stroke-Alzheimer's Disease and Related Disorders Association, modified slightly to include AD as a diagnosis for individuals aged >90 years (Berg and others 1998; McKhann and others 1984). A total of 331 unrelated DAT cases with a minimum age at onset (AAO) of 60 years were recruited for the study. DNA from 385 age- and sex-matched nondemented controls aged >60 years at assessment were obtained through the ADRC.

We also used clinical data and DNA samples from 631 individuals with late-onset AD and 769 control subjects ascertained from both community and hospital settings in the UK collected as part of the Medical Research Council genetic resource for late-onset AD (MRC Sample). A detailed description of the ascertainment and assessment of this sample has been reported elsewhere (Morgan and others 2007).

Data from 199 AD cases and 188 controls from the Alzheimer's Disease Neuroimaging Initiative (ADNI) were used. Data used in the preparation of this article were obtained from the ADNI database on May 15^th, 2008 (www.loni.ucla.edu\ADNI). The Principle Investigator of this initiative is Michael W. Weiner, M.D., VA Medical Center and University of California – San Francisco. ADNI is the result of efforts of many co-investigators from a broad range of academic institutions and private corporations, and subjects have been recruited from over 50 sites across the U.S. and Canada. The initial goal of ADNI was to recruit 800 adults, ages 55 to 90, to participate in the research -- approximately 200 cognitively normal older individuals to be followed for 3 years, 400 people with MCI to be followed for 3 years, and 200 people with early AD to be followed for 2 years.” For up-to-date information see www.adni-info.org. Finally, genotype counts from Robson et al (2004) were used in our meta-analysis (Robson and others 2004).

Rs1049296 and rs1800562 were genotyped using Sequenom genotyping technology. Single SNP allelic associations were evaluated using Fisher's exact test and genotypic associations were evaluated with logistic regression using the additive, dominant and recessive models. Synergy factor analysis (SFA) and adjusted SFA were used to evaluate the size and significance of the effect of interaction between rs1049296 and rs1800562 and risk for AD with minor allele non-carriers as the reference group (Combarros and others 2008; Cortina-Borja and others 2009; Lehmann and others 2001; Robson and others 2004).

Neither rs1049296 nor rs1800562 showed association with risk for AD in single SNP tests using the additive model (table II). Analyses using the recessive and dominant genetic models and models using APOE e4 as a covariate also failed to detect association in the single SNP tests. Allele and genotype frequencies appeared consistent between males and females. SFA in the WU series was significant with a p-value of 0.0032 and a synergy factor of 5.99 (95% Confidence Interval (CI): 1.82-19.69) for bi-carriers using non-carriers as the reference. SFA in the MRC and ADNI samples was not significant (table III) but showed trends in the same direction. A large number of samples from the WU and MRC were recently included in a genome-wide association study. Extensive analyses using Eigenstrat (Price and others 2006) showed no evidence of population stratification between these two samples (Harold and others 2009). Allele and genotype frequencies for each SNP were very similar between the three samples (table II). A combined analysis of our samples shows significant association with SFA unadjusted for covariates and adjusted SFA including site, gender, age and APOE e4 as covariates (table III). SFA unadjusted for covariates using our three samples and data from the initial report (Robson and others 2004), is also significant (p=5.15*10^-3, OR=2.72; table 3).

Our findings in the WU series and the combined sample support the previous observation of synergy between rs1049296 and rs1800562 as risk factors for AD. While there were differences in the level of association in the individuals sample (possibly due to differences in the sample populations and genetic or clinical heterogeneity; table III) the unadjusted SF in our combined sample for individuals that carry at least one minor allele at each locus is 2.71 (CI: 1.46-5.05) and the adjusted SF including age and APOE e4 as covariates was 2.4 (CI: 1.38-3.94). This is lower than the SF of 5.1 from the original report but still indicates a higher level of risk for the bi-carriers of these alleles. Individuals that carry the minor allele at only one of these loci do not show significantly increased risk for AD (table II). In this study bi-carriers make up about 4% of the AD sample. It has been proposed that these individuals may be at higher risk of AD due to increased redox-active iron and oxidative stress (Lehmann and others 2006; Robson and others 2004). Rs1800562 in HFE is known to cause iron-overload and hemochromatosis in individuals homozygous for the allele (OMIM-235200). Wild type HFE binds transferrin receptor 1 (TfR1). HFE with the minor allele “A” of rs1800562 has much lower affinity for TfR1, leaving the receptor free to bind TF with high affinity (Feder and others 1998). This may result in increased uptake of TF bound iron, causing over-absorption of dietary iron and iron deposition in various tissues (Townsend and Drakesmith 2002). Wild type TF is important for iron transport and may be involved in limiting the amyloid aggregation process (Giunta and others 2004). While rs1049296 does not appear to affect the ability of TF to bind iron (Zatta and others 2005), the minor allele “T” shows significant association with increased Aβ42/Aβ40 ratio in the cerebrospinal fluid (Kauwe and others 2009). Our current knowledge of rs1049296 and rs1800562 implicate both effects on Aβ and iron-overload as possible mechanisms for AD risk. In summary our findings provide support for previous reports of synergy between rs1049296 and rs1800562 as risk variants for AD and support for the hypothesis that iron transport and regulation play a role in AD pathology.