SORL1 gene, plasma biomarkers, and the risk of Alzheimer’s disease for the Han Chinese population in Taiwan

The present study confirmed SORL1 as a susceptible gene for LOAD and MCI in the Han Chinese population in Taiwan. The SNP rs1784933 located in the 3? region of the SORL1 genome and the nonsynonymous SNP rs2298813 were most significantly associated with AD and MCI. A lower plasma level of A?42 was found in individuals carrying the minor allele G of rs1784933 in comparison with those without the G allele. A similar trend of reduced plasma levels of A?40 was also observed in the G allele carriers, but this finding was not significant. Neither MMSE scores nor any test of the six cognitive domains differed among SORL1 genotypes.

In the Taiwanese population, SNP rs1784933 (SNP26) is most significantly associated with AD/MCI susceptibility, and its minor allele G exerts a protective effect against disease. Consistent with our findings, in a study of persons of Han Chinese descent in mainland China, researchers found that the G allele of rs1784933, but not the other two tested SORL1 SNPs was related to a reduced risk of AD [16]. Although the associations between SORL1 polymorphisms and AD have been replicated in several studies [14, 15], the regions tagged by most significant SNPs vary across different ethnic groups. For whites, Caribbean Hispanics, and Israeli Arabs, SNPs located in the 5? end of the SORL1 genome (i.e., SNPs 8–10) are most strongly associated with AD [12, 14]. However, SNPs near the 3? region of the SORL1 genome (i.e., SNP19 and SNPs 22–25) are more significantly related to AD in the Chinese, Japanese, and African American populations [13, 33, 34]. The consistent findings between our study and other Asian groups imply a pathogenic role of the 3? region of SORL1 in AD, especially for Asian populations. In addition, the different haplotype frequency and LD conformation between Han Chinese and CEU populations (Fig. 1) further explain why the most significant SNPs vary across populations.

It is worthwhile to note that the nonsynonymous SNP rs2298813 (A528T), causing an amino acid substitution from alanine to threonine at the 528th residue of SORL1 protein, was significantly associated with MCI in our population. A similar but insignificant effect of rs2298813 on LOAD was also observed. Interestingly, rs2298813 was rarely found significant in previous GWAS of LOAD, but this coding variant segregates with disease status in familial AD [30]. The results of an in vitro study suggest that this coding variant has a direct and deleterious impact on AD pathogenesis because HEK293 cells expressing A528T mutant SORL1 could not physiologically interact with APP, which subsequently increased the secretion of A?42, soluble APP?, and APP? [30].

The SORL1 protein regulates APP trafficking and processing, which subsequently influences the formation of A? [9]. Researchers in several studies explored the relationship between SORL1 polymorphisms and CSF levels of A?42 and A?40, but their work led conflicting results [17–19, 35, 36]. Concordant with our findings that subjects carrying the minor allele of rs1784933 have reduced plasma levels of A?42, investigators in several studies found that SNPs located at the 3? region of SORL1 were associated with lower concentrations of A?42 in CSF [17–19]. A trend for reduced CSF concentrations of A?40 was also observed in these studies, but without statistical significance. Because SORL1 regulates the APP processing pathway upstream from the catalyzation of ?- and ?-secretases, insufficient SORL1 activity would not change the ratio of A?42/A?40 concentrations.

To our knowledge, the present study is the first investigation of the influence of SORL1 polymorphisms on plasma concentrations of A?42 and A?40. Although plasma A? concentrations might be confounded by age, disease duration, and other factors [37, 38], they are more easily accessible than CSF A? levels as a surrogate marker of AD pathology. Notably, the association between plasma A? levels and SORL1 rs1784933 derived mainly from patients with AD rather than from patients with MCI (Table 3). There might be two reasons for such a discrepancy. First, only 30–60% of patients with MCI have a neurodegenerative and progressive course, with the remainder having nondegenerative (or reversible) causes [39, 40]. The MCI group consists of heterogeneous entities, including AD and other pathogenesis, which might account for the insignificant correlation between SORL1 polymorphisms and plasma A? concentrations. Second, the smaller sample size of the patients with MCI with available plasma A? levels may have limited our power to detect a significant correlation. We did not measure the plasma concentrations of SORL1 protein, because it is undetectable in the circulation, according to a previous study [41].

The relationship between SORL1 variants and cognitive function has been investigated. Reynolds et al. found that markers at the 5? region of SORL1 tended to be associated with verbal function decline and that SNPs near the 3? end were more related with episodic memory impairment [42]. However, in a large cohort with a sample size up to 9624 participants, researchers did not find any correlation between SORL1 variants and different domains of cognitive function [43]. The present study also does not demonstrate any association between SORL1 SNPs and the six cognitive domains.