The meta-analysis of the five GWASs (Murcia, ADNI, GenADA, NIA, and TGEN) included a total of 3,009 cases and 3,006 controls. A total of 696,707 SNPs were common to all GWAS whereas 1,098,485 SNPs were common to at least four. Figure shows a Manhattan plot with the results of this GWAS meta-analysis. We identified several signals, most of them found in previously reported AD loci (Additional file 2
). The only GWAS-significant result (P
= 4.71 × 10-15) corresponded to rs10402271 in chromosome 19, a marker located 78 kb upstream of the APOE
locus. Other suggestive signals were located in chromosome 2 (rs7561528, located 25 kb downstream of the BIN1
locus), chromosome 22 (rs7561528 and rs13447284), and multiple regions within chromosome 11. In fact, among the top 100 markers, 45 were located on chromosome 11 (Additional file 3
). Chromosome 11 contains several independent suggestive association signals, including the HBG2
(hemoglobin, gamma G) locus (peak association at rs10838245, P
= 1.04E-5), MSE4A
gene family cluster (peak association at rs7626344, P
= 5.48E-6), GAB2
(GRB2-associated binding protein 2; rs450128, P
= 2.79E-6), downstream PICALM
= 1.50E-4), and putative downstream gene BC038205 (rs7935502, P
Manhattan plot of meta-analysis of five GWASs (Murcia, ADNI, GenADA, NIA, and TGEN), including a total of 3,009 cases and 3,006 controls. A total of 696,707 SNPs were common to all GWASs whereas 1,098,485 SNPs were common to at least four studies.
We then conducted an 'in silico
' replication of our results using aggregated data from Harold et al
] (which includes the top 731 signals from their study, many of them also located in chromosome 11) and Hu et al
] (a comprehensive rank of 451,001 SNPs genotyped in their GWAS). Although limited by the number of SNPs available from these studies, the new meta-analysis yielded quite interesting results, with a total of 17 markers above the GWAS significance level (Additional file 4
). Several signals belonged to known AD loci: APOE
with eight SNPs, PICALM
(three SNPS, the most significant being rs536841, P
= 2.96E-9), CLU
= 4.11E-8), and BIN1
= 2.13E-9). Most important, we found four SNPs that belong to a region in chromosome 11q12 not previously reported as GWAS significant for AD. The new peak for AD is located within the MS4A
cluster and the most significant SNP was rs1562990 (OR 0.87; P
Since we have previously published replication studies of APOE
signals in the Spanish population [8
], we decided to replicate only rs1562990 in 2,200 cases and 2,301 controls from the this population. Importantly, the result of this new independent replica was fully consistent, yielding a significant OR of 0.90 (95% confidence interval (CI) 0.83 to 0.98; P
= .01). Detailed results for the original Spanish GWAS dataset, Spanish replica sample, and the combined Spanish dataset are described in Additional file 5
. We fitted a multivariate logistic regression model for the combined Spanish sample in which we adjusted for age, sex and APOE
. The adjusted OR estimate was virtually unchanged (OR 0.87; 95% CI 0.74 to 1.04; P
= 0.12), suggesting that the observed effect is not influenced by age, sex or APOE
in our series.
Finally, combining this new replication in a final meta-analysis together with the five original GWASs and the two 'in silico' replications yields an OR of 0.88 (95% CI 0.85 to 0.91; P = 4.4E-11), which exceeds the accepted threshold for testing multiple comparisons (that is, P < 5E-8). A total of 10,181 cases and 14,341 controls are included in this combined analysis. The magnitude of effect is consistent across studies, with all ten estimates between 0.74 and 0.91 (Figure ).
Figure 2 Meta-analysis and Forest plot of rs1562990, reporting odds ratio (OR) with 95% confidence interval (CI), study-specific weight, sample size and minor allele frequency (MAF) in cases and controls, for each study. The figure shows the remarkable consistency (more ...)
Our results point to the existence of a new AD locus located within the MS4A
cluster at 11q12. Coincidentally, during the drafting of this manuscript two independent articles emerged reaching similar conclusions regarding MS4A
cluster involvement in AD [25
]. Certainly, the SNP markers described in the three studies are different, but they are only 83,871 bp apart. However, our signal is closer to rs4938933 (reported by Naj et al
]), which is only 9 kb centromeric to rs1562990. In any case, peak markers observed in these studies are located in the same haplotypic block and have identical effect size and direction, which strongly suggest that they are tracking the same functional variant.
It is important to mention that sample overlapping exists between these studies. Nonetheless, at least three full datasets contained in our study (comprising 7,809 individuals, 31%) do not overlap with previous published works. Importantly, meta-analysis using only these non-overlapping samples also rendered a significant association with the MS4A region (OR = 0.897; 95% CI 0.838 to 0.961; P = 0.0018). Therefore, our study could be considered an independent replication of the involvement of the MSA4A gene cluster in AD. The concurrence of three independent studies reaching the same conclusion by employing different SNP platforms, imputation methods and datasets underscores the strength and consistency of this new AD locus, at least in European populations. Further studies will be necessary to corroborate its involvement in AD etiology in other ethnic groups.
family includes at least 16 paralogues. Each gene has been probably generated by an ancestral cascade of intrachromosomal duplications during vertebrate evolution. Unfortunately, this gene family is poorly characterized, although a role in immunity has already been shown for several members this cluster, including MS4A1
]. However, the function in humans of many other members remains obscure and a more general involvement of MS4A
family members as ion channel adaptor proteins in non-immune tissues is suspected [28
The rs1562990 marker maps between MS4A4E
members of the cluster. However, we detected a critical linkage disequilibrium haplotype block spanning 163 kb that comprises three members of the family (MS4A2
, and MS4A4
) and the top four meta-GWAS-significant markers (Additional file 4
). With the available data it is difficult to determine the precise location of the functional variant associated with AD, or even which gene could be the best candidate for AD etiology. Furthermore, it may be the case that a functional non-coding variant within the cluster might alter, by cis
-regulation, the function of other members of the cluster simultaneously. Re-sequencing and functional studies of candidate mutations could help resolve this question.
The most centromeric gene within the critical block, MS4A2
, encodes a protein that binds to the Fc region of immunoglobulin epsilon. MS4A2
seems responsible for initiating the allergic response by binding of allergen to receptor-bound IgE, which leads to cell activation and the release of mediators (such as histamine). This signal cascade is responsible for the manifestations of allergy [29
]. Indeed, polymorphisms within the MS4A2
gene have been associated with susceptibility to aspirin-intolerant asthma [30
], and some epidemiological studies suggest a link between asthma and AD [31
]. Consequently, a hypothetical link between MS4A2
and AD would add new evidence in favor of the AD neuroinflammatory hypothesis, suggesting a role for the immune system in the pathogenesis of AD. The other genes within the candidate block are poorly characterized and it is not easy to delineate a plausible hypothesis for them yet.
GWAS data from Spanish patients is available for qualified researchers after institutional review board approval by the Comunidad Autónoma de la Región de Murcia (Spain). Send requests to Dr Carmen Antúnez Almagro mcarmen.antunez/at/carm.es.