The present study was focused on identifying functional SNPs located within AD-linked loci [
8] that are likely to modulate miRNA binding. To predict the putative effects of AD-associated SNPs, we used the following criteria to select miRNA target site identification algorithms: (1) availability as a stand-alone program, (2) ability to query the latest version of the miRBase database, and (3) ability to handle an independent set of 3′-UTR sequences. The “gold standard” TargetScan, the less stringent miRANDA, and the machine learning algorithm TargetSpy met these criteria [
21,
26,
27]. We could also have used algorithms (such as miRTarBase and miRecords) that take account of the miRNAs’ reported functionality by consulting collections of miRNA target data from both low- and high-throughput experiments [
39,
40]. Although this approach would have increased the likelihood of finding legitimate sites, it would also have ruled out associations that have yet to be discovered. We therefore decided not to include these algorithms in our analysis.
We next defined a stringent approach for identifying miRNAs and PolymiRTSs with TargetScan, miRANDA, and TargetSpy, based on the alignment scores for miRNAs and their targets. In line with the thresholds used in other studies, we filtered our results with a TargetScan “context +” score of −0.318 [
40‐
42]. In contrast, the miRANDA alignment scores used in the present study were far stricter than those reported in the literature [
25,
27,
43]. We confirmed that three of the eight selected miRNA sites had a functional impact in two unrelated cell lines. Although we are aware that our approach may have excluded potentially interesting miRNAs and PolymiRTSs, we chose to focus on genomic sites and miRNAs that had functional importance in both cell lines [
44]: the miRNAs miR-4504, miR-3945, and miR-1185-3p and the corresponding sites in the 3′-UTRs of the
FERMT2,
MS4A2, and
NUP160 genes. Our data showed that rs2847655 did not affect the decrease in luciferase activity observed when
MS4A2 3′-UTR reporter constructs were coexpressed with miR-3945 (relative to coexpression with SCR control miRNAs). Furthermore, the AD-associated rs7043400-T allele was associated with lower
FERMT2 3′-UTR luciferase reporter levels in HeLa cells in response to miR-4504 (compared with SCR control miRNAs). In HEK293 cells, the rs7043400-T allele was similarly associated with miR-4504-mediated repression, even though the rs7043400-G allele was also found to be regulated by miR-4504. Although the results in the two cell lines differed slightly, the presence of the rs7043400-T allele was always associated with decreased expression of the
FERMT2 3′-UTR-dependent luciferase activity following miR-4504 overexpression. Since the rs7143400-T allele is associated with an increase in AD risk (OR 1.09, 95 % CI 1.04–1.15), our data suggest that low
FERMT2 expression might contribute to the development of AD. This hypothesis is in line with the results of recent screening experiments in
Drosophila, where the authors identified the FERMT2 orthologs Fit1 and Fit2 as regulators of Tau toxicity; low expression of Fit1 or Fit2 exacerbated Tau toxicity in the
Drosophila eye, whereas elevated expression resulted in the opposite phenotype [
45]. Following an extensive analysis of the literature, we noted that many miRNAs predicted to target the FERMT2 mRNA are reportedly downregulated in AD (Table
2). Although the spatiotemporal expression pattern of FERMT2 in the brain and its regulating miRNAs have not yet been determined, this observation supports a hypothesis whereby dysregulation of miRNA expression and/or binding (due to polymorphisms such as rs7043400) may favor
FERMT2 underexpression and thus Tau pathology.
Table 2
Alzheimer’s disease-associated deregulation of microRNAs targeting FERMT2 and NUP160
FERMT2
| hsa-miR-29b-3p | Downregulated | Cogswell et al. [ 32], Hebert et al. [ 33], Nunez-Iglesias et al. [ 48], Geekiyanage et al. [ 49], Hebert et al. [ 35], Kiko et al. [ 50], Leidinger et al. [ 37], Tan et al. [ 51], Villa et al. [ 52], Denk et al. [ 34] |
hsa-miR-107 | Downregulated | Hebert et al. [ 33], Wang et al. [ 53], Leidinger et al. [ 37], Muller et al. [ 54] |
hsa-miR-15a-5p | Downregulated | Cogswell et al. [ 32], Hebert et al. [ 33], Nunez-Iglesias et al. [ 46], Leidinger et al. [ 37], Denk et al. [ 34] |
hsa-miR-144-5p | Downregulated | Leidinger et al. [ 37], Denk et al. [ 34] |
hsa-miR-103a-3p | Downregulated | Cogswell et al. [ 32], Hebert et al. [ 33], Hebert et al. [ 35], Leidinger et al. [ 37], Denk et al. [ 34] |
hsa-miR-582-3p | Downregulated | |
hsa-miR-498 | Not Altered | |
hsa-miR-29a-5p | Not Altered | Hebert et al. [ 35], Denk et al. [ 34] |
hsa-miR-222-3p | Not Altered | Hebert et al. [ 33], Lau et al. [ 38], Denk et al. [ 34] |
hsa-miR-424-5p | Upregulated | Cogswell et al. [ 32], Hebert et al. [ 33], Lau et al. [ 38], Denk et al. [ 34] |
hsa-miR-3163 | Upregulated | |
NUP160
| hsa-miR-1185-1-3p | Downregulated | |
hsa-miR-126-5p | Downregulated | Cogswell et al. [ 32], Hebert et al. [ 35], Leidinger et al. [ 37], Denk et al. [ 34] |
hsa-miR-133b | Downregulated | Cogswell et al. [ 32], Hebert et al. [ 33], Denk et al. [ 34] |
hsa-miR-323b-3p | Upregulated | |
Last, we identified rs9909-C. This allele (1) is known to reduce AD risk (OR 0.93, 95 % CI 0.90–0.96) and (2) affected miR-1185-3p downregulation of the
NUP160 3′-UTR luciferase construct. Our data indicate that increased NUP160 levels might be protective against the development of AD. The
NUP160 gene is located within the
CELF1 locus, and NUP160 is part of a protein family involved in nuclear transport. Interestingly, alterations in nuclear transport have been described as a possible mechanism in the pathogenesis of neurodegenerative diseases [
46,
47]. Through extensive analysis of the literature, we noted that many miRNAs targeting the NUP160 mRNA (including miR-1185-3p) are reportedly downregulated in AD (Table
2) [
35,
38], which might reflect an attempt of neurons to restore normal nuclear transport. Accordingly, it will be important to determine whether the
NUP160 gene accounts for the GWAS signal observed in the
CELF1 locus.