Customize your JAMA Network experience by selecting one or more topics from the list below.
Cellini E, Tedde A, Bagnoli S, et al. Implication of Sex and SORL1 Variants in Italian Patients With Alzheimer Disease. Arch Neurol. 2009;66(10):1260–1266. doi:10.1001/archneurol.2009.101
To investigate the association of genetic variants in sortilin-related receptor (SORL1), which has been proposed as an important genetic contributor to late-onset Alzheimer disease (LOAD).
We analyzed 13 SORL1 single-nucleotide polymorphisms (SNPs) and the relative haplotypes in a case-control association study.
The sample included 708 Italian subjects: 251 unrelated, sporadic patients with LOAD, 99 sporadic patients with early-onset Alzheimer disease (AD), and 358 healthy controls.
Main Outcome Measures
We analyzed the 13 SNPs in the SORL1 gene that had been studied in previous reports using case-control methods and included sex, apolipoprotein E (APOE) genotype, and age at AD onset as covariates.
The SNPs 4 (rs661057), 7 (rs12364988), and 10 (rs641120) were significantly associated with LOAD compared with controls. We found an association between these 3 variants and sex, suggesting that SORL1 may possibly affect LOAD through a female-specific mechanism. Of interest, the association of these SNPs with LOAD was confined to APOE ε4 noncarriers. Several haplotypic associations at the 5′ end of SORL1 were found, including the previously associated CGC haplotype at SNPs 8 through 10.
Our results confirm the association of SORL1 with AD and show a possible effect of female sex, suggesting that this gene may be a promising susceptibility factor for LOAD. Further studies to detect pathogenic variants and further elucidate the effect of SORL1 on the development of AD are necessary.
Multiple susceptibility genes have been implicated as possible risk factors for late-onset Alzheimer disease (LOAD), but only the apolipoprotein E (APOE; NCBI Entrez Gene NM_000041.2) ε4 allele is recognized worldwide as a major contributor to the disease. Recently, sortilin-related receptor (SORL1; NCBI Entrez Gene NM_003105.4) has been proposed as another important genetic candidate for LOAD.1SORL1 is involved in intracellular trafficking and recycling the amyloid β (Aβ) peptide into the endocytic pathway.1 Patients with Alzheimer disease (AD) have lower SORL1 expression levels, whereas SORL1 knockout animal models present increased brain Aβ levels, suggesting that SORL1 has an inhibiting effect on the Aβ protein precursor amyloidogenic pathways.2,3
Several different association studies4 (Table 1) have linked genetic variants within the SORL1 gene (11q23–q24) with LOAD. Two clusters of different single-nucleotide polymorphisms (SNPs) and haplotypes located at the 3′ and 5′ ends of the SORL1 gene were associated with risk of developing familial and sporadic forms of AD in samples of different ethnic origins, suggesting a marked allelic heterogeneity at this locus. In particular, 8 studies1,5-11 reported positive allelic and haplotypic associations between AD and SORL1 variants. In addition, one study12 found that SORL1 variants contributed to the subclinical phenotypes (ie, preclinical cognitive dysfunction), whereas Kölsch et al13 suggested that SORL1 SNPs may relate to altered cerebrospinal fluid Aβ levels in patients with AD.
Two recent studies14,15 did not confirm this finding, although the data from Shibata et al14 have been recently reanalyzed17 and 2 SNPs were significantly associated with the Japanese AD population. A large genome-wide association study16 failed to detect an association between SORL1 and AD; however, it was likely confounded by allelic/nonallelic heterogeneity owing to multiple ethnic origins of the data set. Intriguingly, this study suggested that another gene that is functionally and structurally related to SORL1 (SORCS1) may be associated with AD. A systematic meta-analysis18 of family-based data (http://www.alzforum.org) failed to confirm the association of the SORL1 gene with AD risk, whereas a recent study19 reporting an association of some genetic variants with cerebrovascular and neurodegenerative changes related to AD supports the idea that multiple areas in SORL1 are of functional importance.
To date, the exact identity of the pathogenic variants of SORL1 remains to be identified. To further investigate the proposed influence of the SORL1 gene on AD, we analyzed the distribution of 13 different SNPs and their correlating haplotypes, which have already been reported as being involved in LOAD, in a case-control study of an Italian population.
The Italian sample included 708 subjects: 251 unrelated, sporadic patients with LOAD (157 women and 94 men; mean (SD) age at onset, 71.8 [5.2] years); 99 sporadic patients with early-onset AD (mean age at onset, 55.5 [4.9] years); and 358 healthy controls (248 women and 110 men; mean age, 83.4 [17.9] years). The patients were consecutively enrolled from the outpatient clinics connected with the Department of Neurological and Psychiatric Sciences, University of Florence. All cases were clinically evaluated according to published guidelines, and each AD diagnosis fulfilled Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) criteria.20,21 None of the patients with AD had a documented family history of dementia. The ethics committee approved the study protocol, and informed, written consent was obtained from each patient or, when appropriate, from a caregiver. Healthy participants were recruited from the same region and underwent a rigorous diagnostic evaluation to exclude any possible neurological disorders.
Genomic DNA was extracted from blood samples, and genotyping of the SORL1 SNPs was carried out by polymerase chain reaction and high-resolution melting analysis22 using the real-time cycler (Rotor-Gene 6000; Corbett Life Science, Mortlake, Australia).
APOE was genotyped with polymerase chain reaction and restriction fragment length polymorphism. For each SNP, the χ2 test was used to evaluate the risk across the genotype categories of each polymorphism and SORL1 haplotype (Epi Info software, version 3.3.2; http://www.cdc.gov/EpiInfo/).
Statistical analysis was performed using SPSS statistical software, version 15 (SPSS Inc, Chicago, Illinois). The effects of APOE ε4 allele status and SORL1 genotype alone and in combination were assessed using a logistic regression analysis adjusted for age and sex. Age at onset between the genotype groups was compared using analysis of variance. P < .05 was considered statistically significant.
Haplotype frequencies were calculated using PowerMarker software, version 3.25 (http://statgen.ncsu.edu/powermarker/).23 The P values for the overall significance of association for all haplotypes were obtained using a permutation-based algorithm implemented with the software CLUMP, thereby negating the need for Bonferroni correction.24 Power analysis was performed using COMPARE2 software (http://www.brixtonhealth.com/pepi4windows.html).
We genotyped a total of 13 SNPs (4-10, 17, 19, 22-25) in the SORL1 gene, numbered according to the previous publication by Rogaeva et al.1 According to previous reports and to the HapMap project, the analyzed SNPs are adequately representative of the block structure of the 2 clusters (5′ and 3′) in the SORL1 gene that have been associated with AD.
Genotype and allele frequencies for the SORL1 polymorphism are shown in Table 2. In both the AD and control populations, none of the SNPs deviated from the Hardy-Weinberg equilibrium.
The individual SNP analysis revealed that SORL1 SNP 4 (rs661057) was significantly associated with LOAD compared with controls for both the T allele (68.5% vs 59.6%; P = .002; odds ratio [OR], 1.47 [95% confidence interval (CI), 1.15-1.89]) and T/T genotype (47.0% vs 34.3%; P = .002; 1.7 [1.20-2.39]).
Our data indicate a sex difference for SNP 4: women with LOAD predominantly account for the SNP 4 association for both genotype (T/T genotype, P <.001; OR, 2.08 [95% CI, 1.35-3.19]) and allele (T allele, P <.001; 1.83 [1.33-2.53]), whereas in men these associations did not reach statistical significance (Table 3). Of interest, the LOAD association for this SNP was confined to APOE ε4 noncarriers for both the T/T genotype (P = .002; OR, 1.86 [95% CI, 1.22-2.85]) and T allele (P = .002; 1.61 [1.18-2.21]) (Table 4). Among APOE ε4 carriers, the differences were not statistically significant. In addition, no effect of SNP 4 on age at onset was detected in patients with LOAD.
Among APOE ε4 noncarriers, women had a significantly higher frequency of the T/T genotype (P < .001; OR, 2.41 [95% CI, 1.38-4.19]) and T allele (P < .001; 2.0 [1.31-3.06]), which confirms a sex effect for this SNP (Table 5).
We also found a marginally significant association with LOAD for SORL1 SNP 7 (rs12364988) with the G allele (59.3% vs 53.4%; P = .04; OR, 1.27 [95% CI, 1.00-1.61]) and G/G genotype (38.2% vs 27.9%; P = .007; 1.60 [1.12-1.29]) and for SORL1 SNP 10 (rs641120) with the C allele (63.3% vs 57.5%; P = .04; 1.28 [1.00-1.62]) and C/C genotype (41.4% vs 32.9%; P = .03; 1.44 [1.02-2.04]) (Table 2). Among women, this association with LOAD was limited for both the SNP 7 G/G genotype (P = .02; OR, 1.64 [95% CI, 1.06-2.54]) and G allele (P = .01; 1.46 [1.08-1.98]) and for both the SNP 10 (rs641120) C/C genotype (P = .04; 1.54 [1.00-2.37]) and C allele (P = .02; 1.42 [1.05-1.93]) (Table 3). For SNP 7, when we stratified for the APOE genotype, the association remained significant among APOE ε4 noncarriers for both the G/G genotype (P = .02; OR, 1.66 [95% CI, 1.07-2.57]) and G allele (P = .03; 1.38 [1.02-1.86]), whereas for SNP 10, the association for APOE ε4 noncarriers was significant only for the C allele (P = .02; 1.42 [1.05-1.93]) (Table 4). A sex effect (Table 5) among APOE ε4 noncarriers was found for SNP 7; women showed a significantly higher frequency of the G/G genotype (P = .04; OR, 1.72 [95% CI, 0.99-3.01]) and G allele (P = .008; 1.66 [1.12-2.47]), whereas, for SNP 10, women had only a significantly higher frequency of the C allele (P = .03; 1.52 [1.02-2.28]).
Among men, no differences were found in the overall distribution of SNPs 4, 7, and 10 for patients with LOAD vs controls. In addition, differences in allele and genotype frequencies between patients and controls did not reach statistical significance for the other SNPs in the SORL1 gene (Table 2).
Haplotype analysis, performed using a 3-marker contiguous sliding window, demonstrated several haplotypic associations clustered at the 5′ end of SORL1 (Table 6). Different at-risk haplotypes covering SNPs 4 through 9 were found, whereas the CAA haplotype (SNPs 4-6) and the ACG haplotype (SNPs 7-9) had a protective effect. The CGC haplotype at SNPs 8, 9, and 10 was significantly associated with LOAD (P = .03; OR, 1.29), as previously reported. Intriguingly, we were not able to replicate, in our case-control sample, the reported association of haplotypes at the 3′ region of SORL1, reflecting the lack of association of the single SNP analysis in this region.
Based on previous genetic and functional evidence suggesting an important biological role for the SORL1 gene in AD pathogenesis, we analyzed the possible association of some genetic variants in an Italian case-control sample.
Our findings of an association with 3 variants (SNPs 4, 7, and 10) within the 5′ region confirm the results of prior studies supporting a positive contribution to AD by this region. In addition, we found an association of both the TT genotype (P = .002; OR, 1.7) and T allele (P = .002; OR, 1.47) at SNP 4 (rs661057) with the likelihood of developing AD. This SNP has previously been detected with a similar OR and P value in combined Mayo Clinic and white case-control data sets from the first report on SORL1,1 and SNP 10 was associated with AD in earlier reports.1,9
Although the association of SNP 4 with LOAD is stronger, the association of SNPs 7 and 10 cannot survive a multiple-testing correction for the number of analyzed markers. We should comment that this correction could produce inappropriate results6; we attempted to replicate previous investigations, and the associated polymorphisms were in the same region and in a high rate of linkage disequilibrium with SNPs involved in previous studies.4 Consequently, they could not be considered independent tests.
A similar haplotypic association subsists in the 5′ region. No other marker or haplotype reached significance in our sample, including the SNPs in the 3′ cluster, which are reported to influence AD in white populations in many,1,5-9 but not all, of the published studies, suggesting allelic heterogeneity in this SORL1 region.
Moreover, we had a power range of 98.3% to 99.5% to detect an OR of 2.0 for the examined SNPs (power range, 69%-80% for a 1.5-fold increased risk), assuming an α level of .05. Thus, this study should be sufficiently powerful to verify other positive associations, although we cannot exclude that the other analyzed SNPs (including the 3′ region) could confer only a very low effect on risk, or small effects at the observed minor allele frequencies, that was undetectable within this sample. Comparing controls with a group of 99 patients with early-onset AD, allele and genotype distributions for all of the genetic variants did not differ (data not shown), confirming that the SORL1 pathogenic effect affects only LOAD.
Although most studies1,5-13,17 reported a genetic contribution of SORL1 to the disease, not all studies have found a univocal association with AD. To date, no common coding polymorphisms in SORL1 exons have been correlated with functional effects; therefore, we could hypothesize that SNP 4, localized in the noncoding region of this gene, might have a regulatory role. It is also possible that this polymorphism may be in linkage disequilibrium with a near SNP that is contributing to the risk of AD.
In addition, our data confirm the association between the APOE ε4 allele and LOAD (P < .001; data not shown). We performed stratification of SORL1 SNPs by APOE genotype and observed an increased risk of AD in APOE ε4 noncarriers, suggesting that the risks associated with SORL1 and AD were independent of APOE ε4 status.
Among APOE ε4 carriers, SNP 4 was previously weakly associated with AD risk15; SNP 8 was more prevalent among APOE ε4 noncarriers,17 but no other interactions were reported between SORL1 variants and APOE. The association of SORL1 and AD independent of APOE ε4 allele status remains controversial and requires further investigation.
The pathogenic mechanism of SORL1 that leads to AD could be independent of APOE; therefore, in patients not carrying the APOE ε4 allele as a genetic risk factor, SORL1 variants could be important determinants for disease susceptibility.
In addition, we found an association between SORL1 variants and sex; our data suggest that SORL1 may affect LOAD through a female-specific mechanism. The overall risk for AD seems to be higher among women than men.25 However, it is not clear whether this is because of the longer life expectancy among women at ages when AD is common or other confounding factors, such as education, occupation, and some lifestyle variables.26
Nevertheless, the increased risk conferred by SORL1 variants among women, suggested by our data, requires further confirmation because it may simply reflect a smaller sample size for men. The possible effect of these SNPs in relation to sex and the APOE-stratified analysis should be verified in other samples, including a larger sample of men.
Our results replicate in an Italian population the findings that SORL1 genetic variants may modify the LOAD risk, although with a marked allelic heterogeneity. However, taken together, the results of genetic and expression studies of SORL1 show that the association with AD should not be considered spurious; SORL1 could be one of the most attractive candidate genes for AD pathogenesis, but its effect might not be as strong as initially observed.
Further efforts should be carried out to detect the real pathogenic variants associated with AD and to determine whether they contribute to the AD risk in a sex-influenced way. Identification of the functional causative variants influencing SORL1 expression and/or levels may help clarify their role in the pathogenesis of LOAD and may point to possible future therapeutic strategies.
Correspondence: Elena Cellini, PhD, Department of Neurological and Psychiatric Sciences, University of Florence, Viale Pieraccini 6, 50139 Florence, Italy (firstname.lastname@example.org).
Accepted for Publication: February 15, 2009.
Author Contributions: Drs Cellini and Nacmias had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Cellini, Sorbi, and Nacmias. Acquisition of data: Cellini, Bagnoli, Pradella, and Piacentini. Analysis and interpretation of data: Cellini and Tedde. Drafting of the manuscript: Tedde, Bagnoli, Pradella, Piacentini, and Sorbi. Critical revision of the manuscript for important intellectual content: Cellini and Nacmias. Statistical analysis: Tedde and Bagnoli. Obtained funding: Sorbi and Nacmias. Administrative, technical, and material support: Bagnoli and Pradella. Study supervision: Cellini and Piacentini.
Financial Disclosure: None reported.
Funding/Support: This study was supported by grant 2007HJCCSF_003 from the Italian Ministry of Instruction, University and Research; grant 734 from Regione Toscana; and grant 1070IT/cv2007.0548 from the San Paolo Company.
Create a personal account or sign in to: