In this preliminary study of 50 neuropathologically well-characterized AD cases and 50 unaffected controls, AD was significantly associated with an IREB2-related haplotype with a potentially functionally relevant SNP in the 5′ upstream region (p < 0.001). DNA sequence analysis of all 22 exons of the IREB2 gene (including splice junction sites) as well as 2,241 bp of the 5′ upstream region containing the promoter revealed the presence of 14 polymorphisms. Two of these, a G/A transition at bp –428 in the 5′ upstream region (rs2656070) and a silent C/T transition at bp 2616 in exon 21 of the IREB2 coding region (rs13180), showed statistically significant distributions between AD patients and unaffected controls via chi-square at the p < 0.05 and p < 0.001 levels (p < 0.035 and p < 0.001 via Fisher’s Exact test), respectively.
analyses revealed rs2656070 lies within a probable promoter and disrupts the binding sites of at least two known transcription factors (AP-1 and Sp1). Although c-MYC has been suggested to play a role in the regulation of IREB2
], a clear understanding of IREB2
transcriptional regulation remains elusive. Thus, the finding that rs2656070 is located in a probable promoter region and disrupts the binding sites of these two different transcription factors is intriguing and warrants further investigation. Though silent and likely not functionally relevant despite a potential role as an exonic splice enhancer, rs13180 shares significant levels of linkage disequilibrium with rs2656070 (D′ > 0.999), creating an IREB2
-related haplotype that is significantly associated with AD (p < 0.001). These results suggest that there have been minimal amounts of historic recombination in the IREB2
genomic region and that there is a strong correlation of alleles at these two sites. Interestingly, rs13180 is located at the very end of a haplotype block in Caucasians (http://www.celera.com
). Also interestingly, this haplotype block ends with rs13180 only in Caucasians, whereas Chinese and African populations display an extended block that includes the nearby locus LOC123688. It is also possible that rs13180 is also in LD with an as yet unidentified SNP and it is known that a group of nicotinic acetylcholine receptor subunits known to be associated with AD [10
] are located about 200 kb 3′ upstream of rs13180 [3
Although the finding of the statistically significant polymorphisms with in silico data suggesting a possible functional role for the SNP is intriguing, findings from our study must be considered exploratory, given the small number of subjects and potential for Type 1 errors. Type I errors are essentially “false positives” that occur when multiple tests are performed, whereby statistical associations are seen purely by chance alone. Using a confidence interval of 95% (p < 0.05), the chances of obtaining a statistically significant association by chance alone is 1 in 20 (5%). That chance increases to 7% in our sample due to the smaller number of tests (i.e. 14 SNPs/tests). To guard against Type 1 errors caused by multiple testing, we applied the Bonferroni correction (dividing the confidence interval by the number of tests) to the two SNPs shown to be significant and found that only rs13180 retains statistical significance (p < 0.001).
It should also be noted that the cases and controls show some disparity in mean age. In order to determine whether or not age had any effect on our genotypic and allelic distributions, we reran the logistic regression analysis, a regression-based analysis testing the association of haplotype and trait (quantitative or qualitative) with one regression coefficient per haplotype permitting the inclusion of covariates and moderator variables (e.g. age, gender, etc), adjusting for age and gender and found that the haplotype remains statistically significant (p = 0.043). To further rule out the possibility that our association is due to solely to age, we removed all young individuals (< 65 yrs. old) from the control cohort and re-analyzed the data (n = 21). Using this model and the Fisher’s Exact test, we find that the significance of the two SNPs is maintained, albeit at lower levels (p < 0.033 for rs2656070 and p < 0.034 for rs13180), due to reduced power.
Because the two significant SNPs are interrelated (D′ = 0.99), we ran forward and backward logistic regression analysis with both SNPs as independent variables using the additive model. This analysis revealed that only rs13180 remained significant at the p < 0.001 level (OR 4.4, chi-square = 10.9) and no significance remained for rs2656070 (p = 0.9), which is expected as rs13180 seems to be driving the association. However, given that LD does not always decay in a linear mode, the analysis of promoter SNPs (i.e. rs2656070) could still be promising. If non-Caucasians are excluded, the picture is the same at the p < 0.004 level due to slightly reduced power. Additionally, in order to further exclude the possibility that our distributions were skewed by ethnicity, we eliminated the 10 Africans, 2 Latinos, 1 Asian, and 1 Native American present in the sample and re-analyzed the data. The statistical significance of the distributions of these two SNPs either increased or remained unchanged with the exclusion of samples that could possibly impose any racial bias.
In conclusion, we find by application of multiple statistical models, that the association of the driving SNP (rs13180) as well as the haplotype to the AD phenotype remains statistically significant despite any potential age and/or racial bias. Though extremely encouraging, these findings need to be confirmed in an independent sample and their functional significance clarified in additional validation experiments. In order to assess the frequency of these polymorphisms/haplotype more precisely, we propose to genotype these SNPs in larger samples of cases and controls. Once haplotype frequencies are established, we propose to determine the functional relevance of this haplotype by cloning haplotype-specific promoter fragments into the pGL3-Basic vector and assaying promoter activity via a dual-luciferase system. If the association between this haplotype and AD can be confirmed in a large, ethnically diverse population, it would provide further support for the role of iron in the pathogenesis of AD and provide additional information about the heritable risk factors associated with this catastrophic and increasingly common neurodegenerative disorder.