Clinical association studies implicate SORL1
variants as an AD risk factor [12
]. This potentially important finding must be validated by identifying the polymorphisms responsible for altered SORL1
functions. Because the implicated polymorphisms are frequent but do not include nonsynonymous SNPs that could have affected protein functions, and since low SORL1
expression was associated with AD, we suspected that any genetic variants would affect transcription and mRNA processing, including maturation, splicing, and turnover.
We have applied allelic mRNA expression analysis to SORL1
in prefrontal cortex tissues from AD patients and controls. This brain region is typically involved in late stage AD, and therefore relevant for assessing cis
-acting regulatory polymorphisms that may vary from one tissue to another. In a series of previous studies we have demonstrated that allelic mRNA ratios can be measured accurately in human autopsy brain and yield important information on the presence or absence of cis
-acting regulatory polymorphisms [11
]. The results of the present study failed to detect any substantial allelic expression imbalance (less than 20% deviation from unity), except in one subject not previously diagnosed with AD. As we had measured AEI in 32 subjects total (70 chromosomes), the frequency of a possible allele causing this substantial expression difference could be as high as 2–3%, but probably much less in this population. Nevertheless, given the potential importance of SORL1
in the etiology of AD, a larger study is needed to establish allele frequency, underlying mechanism (gain or loss of function) and possible clinical relevance.
The absence of AEI in all other subjects strongly argues against a role of the frequent candidate SNPs in mRNA expression. Were any of these SNPs functionally relevant, we would have expected detectable AEI, resulting from several possible mechanisms, for example gene regulation, mRNA processing, and mRNA turnover. However, we cannot exclude the possibility that these events can occur in other brain regions.
While the present study was relatively small, the group of 26 AD subjects, and 16 with AEI measurements, it was sufficient to address possible functionality of the frequent SORL1
candidate SNPs. For each of the 7 candidate SNP analyzed, there were several subject heterozygous for one or more of them (, Supplemental table 2
), yet no AEI was present. Measuring allelic mRNA expression ratios obviates the influence of trans
-acting factors and mitigates problems with post-mortem degradation, assuming both alleles degrade at similar rates. In several previous studies where we had detected AEI, we were able to identify regulatory polymorphisms consistent with these results [5
]. The results therefore argue strongly against any of the tested SNP to affect transcription, mRNA processing and turnover in the tissues studied.
Measuring total mRNA levels (relative to β-actin), we did not find any association between mRNA levels and genotype, as expected because mRNA levels are subject to multiple trans-acting factors, and post-mortem degradation. The lower SORL1 levels in AD subjects relative to controls were offset by lower β-actin levels that we used for normalization. Possibly, mRNA levels were generally lower in these AD tissues.
The SNPs analyzed in this study have shown haplotypic association with AD in the Northern European data sets [12
]. Four SNPs at the 3′ end of SORL1
showed overlapping haplotypes of CTT at rs1699102, rs3824968, rs2282649 and TTC at rs3824968, rs2282649, rs1010159 associated with Alzheimer’s disease in the Caucasian data set. These haplotypes were also present in our study, indicating that AEI could also not have arisen from a combination of SNPs as suggested by the haplotype associations. Since two clinical studies failed to reveal an association of SORL1
with AD [16
], the question remains open as to whether and to what extent SORL1
variants contribute to disease risk.
Our analysis argues against regulatory mechanisms but leaves open a number of possible molecular genetic mechanisms that still need to be explored. First, we cannot exclude that alternative splicing has been affected by a cis-acting polymorphism, leading to mRNA variants that have equal turnover, thereby failing to result in detectable AEI. However, inspection of EST databases did not reveal the presence of substantial splice variants. Neither of the two exonic SNPs is associated with exonic splicing enhancer sites (using the online RESCUE-ESE tool). Second, any polymorphism, including synonymous and nonsynonymous SNPs, and SNPs in the 3′ and 5′ untranslated regions in mRNA, can affect the rate of translation and thus protein levels, by several mechanisms including altered codon usage or other regulatory effects. For example, the two selected marker SNPs rs3824968 and rs12364988, while being synonymous, are present in the mature mRNA and could alter translation. Measuring SORL1 protein levels in association with genotype will require a larger sample cohort and precisely specified brain regions because protein levels are also subject to trans-acting factors, generating more variability farther downstream of genetic variants in SORL1. Nevertheless, the previous suggestion that SNPs occurring in at least two different clusters of intronic SNPs regulate expression of SORL1 mRNA is not supported by the present study.