- Top of page
- Materials and methods
Genes involved in cellular mechanisms to repair oxidative damage are strong candidates as etiologic factors for Alzheimer’s disease (AD). One important enzyme involved in this mechanism is superoxide dismutase 2 (SOD2). The gene for this enzyme lies within a single haplotype block at 6q25.3, a region showing evidence for linkage to AD in a genome scan. We genotyped four single nucleotide polymorphisms (SNPs) in SOD2 in families of the National Institute of Mental Health-AD Genetics Initiative (ADGI): rs2758346 in the 5′ untranslated region (UTR), rs4880 in exon 2, rs2855116 in intron 3 and rs5746136 in the 3′UTR. Under a dominant model, family-based association tests showed significant evidence for association of AD with the first three loci in a candidate gene set of families with individuals having age of onset of at least 50 years and two affected and one unaffected sibling, and in a late-onset subset of families (families with all affected individuals having age of onset of at least 65 years) from the full ADGI sample. The alleles transmitted more frequently to cases than expected under the null hypothesis were T, C, G, and G. Global tests of the transmission of haplotypes indicate that the first two loci have the most consistent association with risk of AD. Because of the high linkage disequilibrium in this small (14 kb) gene, and the presence of 100 SNPs in this gene, 26 of which may have functional significance, additional genotyping and sequencing are needed to identify the functionally relevant SNP. We discuss the importance of our findings and the relevance of SOD2 to AD risk.
There is a significant body of evidence indicating that oxidative damage is involved in the aging process, and in particular neurodegenerative diseases such as Alzheimer’s disease (AD) (Beal et al. 1997). The brain, with its high oxygen consumption, high content of easily oxidized lipids, and shortage of strong antioxidant defenses, is particularly susceptible to oxidative injury and stress (Smith et al. 1996, 1998). All levels of cell metabolism in the AD brain appear to be more affected by free-radical damage than in the normal aging brain (see Christen 2000; Markesbery 1999; Miranda et al. 2000; Varadarajan et al. 2000, for review). There is increased oxidative damage to nucleic acids and mitochondrial dysfunction in AD brains (Coskun et al. 2004; Davies 1995; Gabbita et al. 1998; Mecocci et al. 1994; Nunomura et al. 1999; Szczesny et al. 2003). Oxidation of proteins by free radicals can result in cross-linking, aggregation and fibril insolubility, like that seen in the aggregated form of beta amyloid (Aβ) (Dyrks et al. 1992), and the hyperphosphorylation of tau in neurofibrillary tangles (NFTs) (Smith 1997).
Oxygen-utilizing cells have evolved complex enzymatic and nonenzymatic defenses such as superoxide dismutases (SODs) and peroxiredoxins to offset oxidative challenges in different subcellular compartments (Fridovich 1978; Krapfenbauer et al. 2003). Superoxide dismutases are the key enzymes involved in the detoxification of the superoxide radical (Fridovich 1995). Levels of antioxidant enzymes, like SODs, increase with age in cortical tissues of Tg2576 mice, an AD transgenic mouse model (Apelt et al. 2004). One SOD, SOD2 or Mn SOD, is located in the mitochondria and is the only isoform that is induced and regulated by reactive oxygen species (Pinteaux et al. 1996; Warner et al. 1996; Weisiger & Fridovich 1973). The Sod2 null mice develop a severe neurological phenotype and neuronal cell death as a result of mitochondrial oxidative stress (Hinerfeld et al. 2004; Lynn et al. 2005), while reduction of Sod2 in transgenic mice carrying amyloid precursor protein (APP) mutations triggers exacerbation of neuronal and vascular AD pathology (Esposito et al. 2006). The SOD2 is expressed in most human tissues (Wan et al. 1994), including regions of the brain that are affected in AD (Zhang et al. 1994).
The SOD2 gene is located at 6q22 in a candidate region of interest we identified in a recent genome scan of the National Institute of Mental Health (NIMH)-AD Genetics Initiative (NIMH-ADGI) set of families (Blacker et al. 2003) and has been identified in other genomic scans of AD (Ashley-Koch et al. 2005; Kehoe et al. 1999; Lee et al. 2006; Pericak-Vance et al. 2000). We now present results using family-based association testing to determine whether four nucleotide changes located in known and possible functional areas of SOD2 are associated with increased risk of AD in the NIMH-ADGI families.
- Top of page
- Materials and methods
The minor allele frequencies of each of the four SNPs in SOD2 in the CGS (203 families) are shown in Table 1 (the allele frequencies in the LOAD subsample were not appreciably different from the frequencies of the CGS). These values are calculated from all the typed individuals in the respective sample using the AFREQ function available in fbat. Single-locus association results for both the CGS and the LOAD subset assuming a general genetic model are shown in Table 2. These results indicate that the alleles resulting in heterozygotes at each locus are consistently, although not always significantly, transmitted more frequently to cases than expected under the null hypothesis of no association between the locus and occurrence of AD (CGS: P = 0.010, 0.037, 0.064 and 0.343, respectively; LOAD: P = 0.007, 0.018, 0.044 and 0.104, respectively). At all but the 3′UTR locus, the major allele homozygous genotype (genotype C/C, T/T, and T/T respectively) is transmitted less frequently to cases than expected under the null hypothesis (CGS: P = 0.009, 0.026 and 0.029, respectively; LOAD: P = 0.026, 0.025 and 0.031, respectively). The transmission disequilibria of the minor allele homozygous genotypes at these three loci (genotypes T/T, C/C and G/G, respectively) are not significant, nor are any of the results at the fourth locus with this general genotypic model. Based on these results, all haplotype analyses were done assuming a dominant genetic model.
Table 1. Genotyped loci in the CGS
|Locus||Name||Location||Major allele||Minor allele||Minor allele frequency|
|2||rs4880||Exon 2 (nt1183)||T||C||0.493|
|3||rs2855116||Intron 3 (nt5777)||T||G||0.464|
Table 2. Family-based, single-locus association test results with general genotype model
|Locus||Genotype||Full CGS sample||LOAD subsample|
The results of single-locus, family-based tests of association, assuming a dominant genetic model, are shown in Table 3. It is worth noting that restricting the analysis to the LOAD group resulted in similar results, except now the associations for all four SNPs are significant, even though the sample sizes are smaller. When the analyses are repeated for the APOE-adjusted traits, we noted little qualitative difference in the results (data not shown), so only the results based on the affection status are reported.
Table 3. : Family-based, single-locus association test results
|Locus||Allele*||CGS sample||LOAD subsample|
|Families†||P value||Families†||P value|
When these tests were performed in the NH4 subgroup, there were no qualitative differences in the results. None of the association results were significant in the HM4 subgroup, possibly because of the loss of power in this much smaller group, but the direction of the association was consistent with the above results. Several ad hoc adjustments using APOE genotype were attempted, and none resulted in any qualitative difference in results.
Haplotype frequencies, estimated within the fbat program, were very similar between the LOAD subset and those of the CGS (see Table 4). The haplotype and single-locus allele frequencies were used to form an estimate of the measure of linkage disequilibrium (LD) between pairs of loci. Specifically, in the CGS, the value of D′ between loci 1 and 2 is 0.980 and r2= 0.960; between loci 2 and 3, D′= 0.996 and r2= 0.897; and between loci 3 and 4, D′= 0.972 and r2= 0.372. Within the LOAD subset, between loci 1 and 2, D′= 0.984 and r2= 0.960; between loci 2 and 3, D′= 1.0 and r2= 0.908; and between loci 3 and 4, D′= 0.972 and r2= 0.355. We, thus, have very strong LD among the first three loci, and slightly less LD with the fourth locus. Data available from the HapMap project (HapMapConsortium 2003) indicate that this entire gene lies within a single haplotype block.
Table 4. Family-based haplotype association test results
|Loci||Haplotype||CGS sample||LOAD subsample|
Results of association tests between haplotypes among the four loci and development of AD are summarized in Table 4. Because of the high LD in this region, we cannot view these tests as independent. For this reason, we do not feel that a conservative multiple test correction, such as the Bonferroni correction, would be appropriate. The global tests of haplotypes reported in Table 4 allow us to partially offset the multiple testing problem. Although every possible combination of loci has a haplotype that is preferentially transmitted to cases under a dominant model, the global tests indicate that the most promising results include only the first three loci.
When the data were analyzed with the gist program, we found that only genotypes at the fourth locus explained a significant proportion (P = 0.004) of the linkage signal at 6q. Genotypes at the first two loci explained a marginal proportion of the linkage signal (P = 0.064, and 0.065, respectively). The full four-locus haplotype (TCGG vs. all others) also explained only a marginal proportion of the linkage signal (P = 0.058).
- Top of page
- Materials and methods
Oxidative damage is believed to be an important factor in the pathogenesis of AD, and SOD2 plays a key role in repairing oxidative damage. All four SNPs in SOD2 examined are located near known or potentially functional regions of the gene and thus could possibly contribute to the associated risk. For example, locus 1 (rs2758346) is located in the SOD2 promoter, 65 bp upstream of a nuclear factor κB (NF-κB) transcription factor binding site (TFBS) (Xu et al. 2002). The TFBS for other transcription factors are located within 1 kb of this SNP (Wan et al. 1994; Xu et al. 2002) Locus 2 is the well-characterized C T (Ala16Val) SNP, rs4880, in the mitochondrial targeting sequence (MTS) of SOD2. The C allele, which retains the alpha helical structure of the protein for normal activity of the enzyme, is over-represented in the patients. The presence of the Ala variant in the MTS of SOD2 for normal insertion and activity in the mitochondria, in addition to other genetic changes, could result in increased activity of SOD2 and induction of oxidative stress (Shimoda-Matsubayashi et al. 1997). The third SNP, rs2855116, is located in intron 3 and is ∼6.5 kb downstream from an additional regulatory region located in intron 2 of SOD2, the tumor necrosis factor responsive element (TNFRE), that is a potent transcription enhancer with binding sites for several transcription factors, including tumor necrosis factor, NF-κB (Guo et al. 2003; Jones et al. 1997; Mao et al. 2006). Another TFBS that plays a role in glial cells, PAX 6, is located within 100 bp of rs2855116 (Heins et al. 2002; Nacher et al. 2005), and nearby intronic sequences are specific to Homo sapiens only (Karolchik et al. 2003; Kent et al. 2002). The last SNP, rs5746136, is located ∼65 bp downstream of the poly A site. It is also <1 kb upstream from SP1 and NF-κB transcription element sequences(Wan et al. 1994).
The LD information in this region indicates that the tests of association among the first three SNPs are almost redundant because of the high values of both D′ and r2. The global tests of association performed to control for multiple testing of haplotypes support this observation. The results from gist indicate that one or more SNPs within our four SNP haplotype may account for a part of the evidence for linkage seen in this sample.
The haplotype apparently associated with AD in this sample could be in LD with other genetic variations that affect function of this gene. According to the dbSNP database build 125 (http://www.ncbi.nlm.nih.gov/projects/SNP/), there are over 100 SNPs in the gene, including the 3′ and 5′ UTRs, which may be functional, including eight coding SNPs, six of which are nonsynonymous, eight additional SNPs located within the 5′ regulatory region or the TNFRE and over a dozen additional SNPs in enhancer splicing elements (ESEs) or within 100 bp of intron/exon junctions.
The significant amount of high LD across this relatively small gene (∼14 kb) presents a challenge to determining a specific causative polymorphism. Any one of four SNPs we genotyped, or an SNP that is in LD, which is within or close to a functional region of the gene (e.g. exon, TFBS, ESEs, exon/intron junctions), could enhance or reduce expression of SOD2 and/or the activity of the gene product. We are currently performing association mapping using a dense SNP strategy, excluding redundant SNPs. In addition, we are sequencing AD cases from our most informative families segregating the high-risk haplotypes in the hopes of determining the functional causative mutation (Fearnhead et al. 2004; Pritchard & Cox 2002).