The algorithm for Alzheimer risk assessment based on APOE promoter polymorphisms

We report a 9.1 % incidence of the APOE E4 variant in our sample derived from the Polish population, which is in line with previously published values [14]. Similarly to previous studies, we were able to show that the presence of at least one E4 variant confers almost a five times higher risk of AD development and that this correlation is gender-dependent [2]. For women, the presence of E4 variant conferred almost an eight times higher risk, while for men the association was not statistically significant (data not shown). In contrast to previously published results, no correlation between the presence of the E4 variant and the age of AD onset was observed [15]. However, our study group did not include many individuals diagnosed when they were older than 80 years of age, as late-onset AD was not a priority in our study. This might have resulted in the lack of such a correlation in our data.

Despite the reproducibility of the APOE E4 variant association with AD development, also shown in our study, the pathomechanism of such a correlation remains largely unsolved. Hence, an intense search for additional modifying factors, both intragenic and located at other loci of the genome, has been conducted worldwide. Of these, two APOE promoter SNPs—rs449647 (?491A/T) and rs405509 (?219T/G)—were shown to have predictive value for AD development [1–3, 5, 6, 9, 16–22]. However, not all studies were able to reproduce the protective effect in their populations, for instance in French [10], Irish [23], Finnish [24] or Japanese [25]. Hence, the exact role of these SNPs and their interaction with epsilon variants—in other words, their predictive character being independent or derivative of epsilon status—remains a subject of intensive debate.

In the present study, we were able to show a protective effect of ?219G (rs405509) allele in a sample derived from the Polish population, while the correlation with the ?491T (rs449647) allele did not reach the level of significance. This is interesting in light of an MRI study by Chen et al. [4], who showed an accelerated age-related reduction of thickness in the left parahippocampal gyrus in ?219 TT carriers and suggested it as the neural substrate underlying a faster decline in cognition in individuals with this genotype. Hence, carrying a G allele might offer protection against such a process and act protectively in terms of cognition. The Chen et al. [4] study, as well as a few others (see, e.g., [9]), showed the effect of this polymorphism to be age-related. As most individuals (95 of 110) in our study were mostly older than 60 years of age, this effect might have been more pronounced.

In addition, we observed that both alleles, ?219G and ?491T, were significantly underrepresented in the carriers of the E4 variant. Such characteristics render the variants plausible candidates for independent risk/protective factors for AD development. However, the latter was not confirmed in the multivariable logistic regression analysis.

The inconsistent findings in previous studies regarding the presence of LD between promoter polymorphisms and the APOE epsilon variant coding SNPs were explained by the suggestion that LD may vary significantly depending on ethnic background. Here, we have shown the presence of a significant LD in the coherent group of Polish patients, being a large ethnic group with a low rate of consanguinity. The analysis of data retrieved from the HapMap project showed no haplotype blocks within the APOE gene. However, various differences in LD between individual SNPs have been observed in all ethnic groups. Lescai et al. [13] performed a similar, large-scale haplotype analysis in a group of more than 1000 Italian subjects and showed the presence of individual LDs but not haplotype blocks. In their study of the prognostic role of APOE SNPs on AD risk, the only significant additive effect has been observed for the rs405509 allele T/E4 haplotype present in phase (i.e., ‘in cis’). They did not perform a more comprehensive analysis that would include other in phase SNP loci. In our present study, we were able to show that the presence of the ?491AA/?219TT/E4 genotype appears to be even more informative. It has been shown that lower ApoE levels are associated with a higher risk of developing AD. As ?491A, ?219T and E4 alleles have all been shown to independently decrease APOE gene expression, and hence ApoE levels [5–7], a haplotype containing all three alleles might confer a higher risk for AD because of the synergistic action of all three variants. Promoter SNPs confer a mild but additive effect on APOE expression. We showed that the ‘preventive score’ correlated significantly with protein ApoE levels, while this was not the case when individual SNPs were considered. It is possible that the SNPs need to be assessed jointly for their impact to reach significance, which might be one of the possible explanations for discrepancies in the previous reports. The results of the analysis of the above-tested polymorphisms in the promoter region of the APOE gene, in addition to increasing knowledge of the allelic variants of the gene, can be important for determining the risk of the incidence of dementia.

We believe that one of the most important findings of this study is the identification of the ‘preventive score’. Its importance derives from the fact that it takes into account an additive risk related to the presence of alleles and genotypes, which on their own may have too weak an impact to reach the level of significance. This has potentially practical implications, as it may help improve the informative potential of APOE testing in a clinical setting. The APOE testing, limited to the evaluation of the E4 variant, was not found to be sufficiently informative to be included in the current European Federation of Neurological Societies guidelines for AD diagnosis and management [26]. Here, we show that the ‘preventive score’ has an independent prognostic value, regardless of E4 status. Moreover, it has been shown to be more informative than serum ApoE levels, the ultimate marker of gene expression. The latter, owing to the complexity of protein-protein interactions and the impossibility of their direct measurement in the tissue of choice (i.e., brain), appears to be a less preferable parameter.

One important issue is the relationship between the ‘preventive score’ and epsilon variants, namely whether the ‘preventive score’ is independent of their effect, and especially the effect of the E2 allele. We attempted to tackle this issue by performing additional statistical analyses using the logistic regression model in the subgroups of patients resulting from inclusion/exclusion of the E2 and E4 allele carriers and in the subgroup homozygous for the E3 allele. They have shown that the ‘preventive score’ allows for further delineation of the eventual risk of AD development in the subgroup of E4 carriers and supports the role of the ‘preventive score’ as an additional diagnostic tool. However, these results need to be viewed with caution, as the sizes of the subgroups after the removal of the E2 and E4 alleles carriers were small (for example, only 5 % of patients with AD and 12 % of control subjects were E2 carriers), and a well-discriminating distribution of the epsilon alleles was not possible to obtain; for example, E2 homozygotes were lacking in both cohorts.

We believe that the utility of the ‘preventive score’ is an interesting and potentially useful finding; however, our results need to be replicated to confirm the validity of the score. Hence, future studies by independent research groups with larger numbers of patients are warranted.