Introduction

Alzheimer’s disease (AD), the leading cause of dementia in the elderly, is growing in its public health implications. AD is divided into early-onset AD and late-onset AD (LOAD), and both have a genetic component [1]. LOAD is typically defined as onset after 65 years of age, which is widely considered as the proportion of disease vulnerability attribute to heritable genetic factors. Moreover, these susceptibility genetic factors may contribute as much as 80 % to the risk of LOAD [2]. However, to date, apolipoprotein E (APOE) gene is the only established susceptibility gene for LOAD, and variations at the APOE loci account for only 50 % or less of LOAD risk [3, 4], suggesting more additional risk loci remains to be identified.

AD and Parkinson’s disease (PD) are the two most common age-related neurodegenerative disorders. AD is characterized by progressive impairment in memory, judgment, orientation to physical surroundings, and language, and the pathologic features of AD are neuronal loss and the deposition of extracellular plaques containing amyloid β (Aβ) and neurofibrillary tangles (NFT) containing tau [1]. PD manifests as a movement disorder and a distinct form of cognitive impairment with typical pathologic characteristics of the deposition of α-synuclein in multiple brain regions [5]. Although AD and PD are clinically distinct entities, the two diseases appear in a close association in pathological evidence. Aβ, one important hallmark of AD pathology, has been reported present in patients with PD [6–9]. Similarly, PD pathologies have also been reported in some AD cases, for example, pathological evidence indicated Lewy body deposition also exited in AD cases [9, 10]. In addition, AD and PD may also share an overlapping genetic background, such as APOE and MAPT gene which have been reported linked to the presence of Aβ aggregates [11] and increased tau protein expression [12], and they have been widely accepted as risk genes in AD according to the Alzgene database (http://www.alzgene.org/). Noticeable, genetic variants at the two genes have also been verified to increase PD risk [13, 14]. Taken as a whole, these studies indicated a possible pathological overlap and a potential hereditary association between AD and PD. Hence, it would be meaningful to replicate PD-related genetic variants in AD population to discover more candidate loci which may alter AD risk.

Recently, Nalls and his colleagues performed a large-scale meta-analysis of genome-wide association study (GWAS) with a common set of 7,893,274 variants across 13,708 PD cases and 95,282 controls to identify PD risk loci [15]. Among these genetic variants, nine SNPs (rs11724635 of BST1, rs12637471 of MCCC1, rs15553999 of TMEM229, rs17649553 of MAPT, rs34311866 of TMEM175-GAK-DGKQ, rs356182 of SNCA, rs6430538 of ACMSD-TMEM163, rs76904798 of LRRK2, and rs823118 of RAB7L1-NUCKS1) were included in our study, they achieved a widely accepted genome-wide P value threshold of 5 × 10−8 under quality control, and they were never explored in AD genetic studies. Based on the above evidence, the current study was conducted to explore whether these newly identified GWAS-linked PD genetic variants also influenced on LOAD susceptibility.

Methods

Subjects

A total of 2350 samples comprising with 992 sporadic LOAD patients (age at onset, ≥65 years) and 1358 gender- and age-matched control subjects were enrolled. All included LOAD patients and control subjects were unrelated northern Han Chinese residents originally from Shandong province. The LOAD patients were recruited from the Department of Neurology at Qingdao Municipal Hospital and several other hospitals in Shandong province, and all patients were subjected to neuropsychological examination and brain structural neuroimaging. A consensus clinical diagnosis of probable AD must fulfill the criteria of the National Institute of Neurological and Communicative Disorders and Stroke/Alzheimer’s Disease and Related Disorders Association [16]. And the LOAD patients were defined as sporadic when no member of their first-degree relatives had dementia in their family history. The information (age at onset and family history) of enrolled LOAD patients were obtained from themselves or their caregivers. We included age- and gender-matched healthy control subjects from the Health Examination Center of each collaborating hospital, and these control subjects were confirmed healthy and neurologically normal according to medical history, general examination, laboratory examination, and Mini Mental State Examination (MMSE) (MMSE > 28). The characteristics of the study groups were summarized in Table 1. We obtained informed consents from each subject or a guardian, and this study protocol was approved by the institutional ethics committees of Qingdao Municipal Hospital. The study group described in this study has not been reported before.

Table 1 Characteristics of the study groups
Full size table

SNP Selection and Genotyping

A total of nine SNPs from a GWAS which reported PD-related genetic variants-rs11724635 (BST1), rs12637471 (MCCC1), rs1555399 (TMEM229B), rs17649553 (MAPT), rs34311866 (TMEM175-GAK-DGKQ), rs356182 (SNCA), rs6430538 (ACMSD-TMEM163), rs76904798 (LRRK2), and rs823118 (RAB7L1-NUCKS1) were selected [15]. Based on the widely accepted genome-wide P value threshold of 5 × 10−8, these nine SNPs were proved to link to PD risk in discovery phase (13,728 cases and 95,282 controls), replication phase (5353 cases and 5551 controls) and joint phase (19,081 cases and 100,833 controls) under quality control. Moreover, these SNPs have never been explored in AD population. Two SNPs within the APOE gene (rs429358 and rs7412) were also analyzed in this study.

The genotyping methods have been described in our previous literatures [17–21]. Briefly, we isolated genomic DNA from peripheral blood leukocytes with the Wizard Genomic DNA Purification Kit (Promega). Genotyping of rs34311866 (TMEM175-GAK-DGKQ), rs356182 (SNCA), rs6430538 (ACMSD-TMEM163), rs76904798 (LRRK2), rs823118 (RAB7L1-NUCKS1), rs429358, and rs7412 (APOE) were performed by the polymerase chain reaction–ligase detection reaction (PCR–LDR) using the improved multiplex ligase detection reaction (iMLDR) method with technical support from the Shanghai Genesky Biotechnology Company, and the details of the primer sequences used for PCR were shown in Supplementary Table 1. In addition, SNP genotyping of rs11724635 (BST1), rs12637471 (MCCC1), rs1555399 (TMEM229B), and rs17649553 (MAPT) was conducted with a custom-by-design 2- × 48-Plex SNPscan™ kit (Genesky Biotechnologies Inc., Shanghai, China) [22]. This kit was made according to a patented SNP genotyping technology by Genesky Biotechnologies Inc., which was based on double ligation and multiplex fluorescent PCR.

Statistical Analysis

We excluded deviations using Hardy–Weinberg equilibrium (HWE) in control subjects by HWE version 1.20 (Columbia University, New York, NY). Univariate analyses of allele and genotype distributions between LOAD patients and control subjects were examined with the χ 2 test or Fisher’s exact test. Then, multivariate analyses of the association between SNPs and AD risk were evaluated with the odds ratio (OR) with 95 % confidence intervals (CIs) by logistic regression, adjusting for age of onset (age at examination for control subjects), gender, and APOE ε4 status (presence or absence of an ε4 allele). Moreover, the SNPs with minor allele homozygote counts of 14 or more were calculated in three kinds of logistic regression models (dominant, additive, and recessive). When SNPs with minor allele homozygote counts are less than 14 and the number of minor allele homozygotes and heterozygotes are more than 14, only the dominant genetic model could be examined. The interaction of SNPs and APOE status was also evaluated by logistic regression. The STPLAN 4.3 software was performed to estimate the statistical power. All statistical analyses were conducted with SPSS 17.0 for Windows. The statistical results were considered significant when the Bonferroni-corrected P value was lower than 0.05 (based on the number of SNPs and logistic regression models analyzed).

Results

Characteristics of the Study Groups

A total of 2350 northern Han Chinese subjects comprising 992 LOAD patients and 1358 controls subjects were included in our study (Table 1). No statistically significant differences found in age (age at onset for LOAD patients and age at examination for control subjects) (P = 0.189) or gender (P = 0.067). The percentage of APOE ε4 carriers was statistically significant between LOAD patients and healthy controls (P < 0.001). In addition, MMSE scores were significantly lower in LOAD patients than healthy controls (P < 0.001).

Univariate Analysis

From the available data in HapMap database, genotype and allele frequencies of included SNPs (except for rs15553999, rs34311866, and rs76904798 which have no information about CHB genotype data) in our control group were consistent with CHB genotype data (Supplementary Table 2). And our included SNPs all passed a 95 % call rate threshold, and the details were seen in Supplementary Table 2. The information of included SNPs (chromosome, position, minor allele, alternate allele, HWE value, and minor allele frequency) was summarized in Table 2. The distributions of the included nine SNPs were in HWE for control subjects (P > 0.05). As shown in Table 2, most gene locations of these SNPs (rs11724635, rs12637471, rs17649553, rs356182, rs6430538, rs76904798, and rs823118) are intron variant; rs1555399 is in 5′-flanking in TMEM229B gene; rs17649553 is in intron1 in MAPT gene and rs34311866 is in nonsynon_11 in TMEM175-GAK-DGKQ. These included SNPs are not expression quantitative trait loci (eQTLs) according to the eQTL database (http://www.hsph.harvard.edu/liming-liang/software/eqtl/). These SNPs are in linkage disequilibrium with other potentially functional variants according to the results of 1000 Genomes database (http://browser.1000genomes.org/index.html), and they may influence on AD risk with these functional variants.

Table 2 Characteristics of included SNPs
Full size table

In total sample, although significant differences were confirmed in allele or genotype of rs12637471, rs356182, rs6430538, and rs76904798 frequencies between LOAD patients and control subjects (Table 3). However, only the distribution of rs356182 allele remains significant after Bonferroni correction (OR = 0.748; 95 % CI 0.661–0.845; Bonferroni corrected P = 0.000), suggesting the A allele of rs356182 is the protective allele.

Table 3 Distribution of included SNPs in LOAD patients and control subjects
Full size table

Multivariate Analysis

We revaluated the association of the included SNPs with LOAD risk under three genetic models using logistic regression adjusting for age (age at onset in LOAD patients and age at examination in control subjects), gender, and APOE ε4 status to rule out confounding factors in our initial analyses. As Table 4 shows, the minor alleles at rs12637471 and rs6430538 were discovered to alter the risk of developing LOAD under a dominant model (adjusted OR = 1.242, 95 % CI 1.049–1.469, P value 0.012; adjusted OR = 0.289, 95 % CI 0.116–0.721, P value: 0.008); however, these results failed to remain significant after Bonferroni correction. Interesting, the distribution of genotype in rs76904798 significantly reduced LOAD risk in comparison with its wild-type homozygotes under a dominant mode adjusting for age, gender, and APOE ε4 status, and the results were still significant even after Bonferroni adjustment (adjusted OR = 0.616, 95 % CI 0.446–0.849, Bonferroni-corrected P value 0.027).

Table 4 Single-nucleotide polymorphism association with Alzheimer’s disease according to different genetic models of inheritance
Full size table

Interactions with APOE Genetics

Afterwards, we stratified the subjects according to APOE ε4 status to explore whether APOE ε4 affected the connection between the included SNPs and LOAD risk via using logistic regression adjusting for age and gender (Table 5). After stratifying the subjects according to APOE ε4 status, a significant association was only observed between the T allele at rs76904798 and LOAD risk under a dominant model in non-APOE ε4 carriers (adjusted OR = 0.396, 95 % CI 0.258–0.607, Bonferroni-corrected P value 0.000). In addition, as shown in Table 5, rs76904798 within LRRK2 gene showed a significant APOE interaction in the logistic regression under a dominant model indicating a synergistic effect of LRRK2 and APOE on AD risk.

Table 5 Interaction analysis between the SNPs and APOE ε4 carrier status
Full size table

Discussion

GWAS is a powerful tool for genetic association studies, and independent replication remains the first step in distinguishing true positive from false-positive genetic association findings [23]. Furthermore, since there is a possible pathology overlap and a potential genetic association between AD and PD, it is important to replicate GWAS-reported PD risk loci in AD population to discover new AD-related genetic variants. In this study, we performed a comprehensive analysis of the association between LOAD risk and nine SNPs which were reported to have close connections with PD risk in a recent GWAS. Our results showed a successful replication of the association of rs76904798 of LRRK2 with LOAD risk in this independent case–control study. This is the first study to indentify rs76904798 of LRRK2, the PD-related locus, also significantly associated with AD susceptibility in a large northern Han Chinese population.

AD and PD are two most common neurodegenerative diseases with a large population of sufferers. Aggregates of insoluble Aβ and NFT consisting of hyperphosphorylated tau protein are classic pathological features of AD [2, 24]. Pathologically, PD is characterized with degeneration of dopamine neurons in the substantia nigra pars compacta and the presence of proteinaceous inclusions immunoreactive for α-synuclein in surviving neurons [25, 26]. Although AD and PD have distinct mechanisms of etiology, different brain regions, and distinct clinical features, they have much overlap in the development of neurodegeneration and share a great similar pathway to induce the occurrence of disease as shown in Fig. 1, including intracellular mechanisms, local tissue environment, systemic environment, and aging [27]. In addition, except for a possible pathological overlap between AD and PD, a growing evidence indicated a potential genetic association also exists in AD and PD. Summarizing and analyzing the numerous pathways which are implicated in PD and AD, many genes have been confirmed to function in both Mendelian and sporadic forms of AD or PD, such as PD-related genes (MAPT, SNCA, GBA, LRRK2, PM20D1, GAK, MCCC1, STK39, BST1, GPNMB) [26, 28–35], top 10 genes from the AlzGene database (APOE, BIN1, CLU, ABCA7, CR1, PICALM, MS4A6A, CD33, MS4A4E, CD2AP) [36] and a small number of other genes (APP, PSEN1, PSEN2, DJ1, HIP1R, PARK2, SYT11, UCHL) [37]. Allowing for the overlap of pathology and genetic background between AD and PD, it may be an appropriate method to disclose AD-related genetic factors from the GWAS which explored PD risk loci. In this study, nine SNPs (rs11724635 of BST1, rs12637471 of MCCC1, rs1555399 of TMEM229B, rs17649553 of MAPT, rs34311866 of TMEM175-GAK-DGKQ, rs356182 of SNCA, rs6430538 of ACMSD-TMEM163, rs76904798 of LRRK2, and rs823118 of RAB7L1-NUCKS1) in a recent GWAS which showed close connections with PD risk were enrolled to identify their relationship with AD susceptibility. Finally, our study, for the first time, successfully replicated the association between rs76904798 of LRRK2 and AD susceptibility in a northern Han Chinese people and indicated a genetic overlap between AD and PD susceptibility at rs76904798 of LRRK2.

Fig. 1
figure 1

The potential schematic profile for Alzheimer’s disease and Parkinson’s disease. Alzheimer’s disease (AD) and Parkinson’s disease (PD) share a similar pathway to induce the occurrence of disease, including intracellular mechanisms, local tissue environment, systemic environment, and aging

Full size image

Mutations in LRRK2 have been accepted as the most frequent cause of sporadic cases of PD [26, 38–40]. To date, many studies have paid attentions to its role in AD pathology, and LRRK2 is considered involved in AD-related pathology, such as tau, inflammatory response, oxidative stress, mitochondrial dysfunction, synaptic dysfunction, and autophagy–lysosomal system [26]. Now, LRRK2 is widely accepted as a potential candidate locus influencing AD risk in recent studies [41–44]. And our results revealed variants at rs76904798 of LRRK2 would significantly decrease AD risk in comparison with its wild-type homozygotes after analysis in logistic regression adjusting for age, gender, APOE ε4 status, and the strong association may result from its potential influence on AD pathology. Furthermore, screening these genetic variants with positive results maybe a promising method to detect and diagnose AD in a northern Han Chinese people, and more experiments in vivo and in vitro are needed to verify these hypotheses. Our results showed rs76904798 of LRRK2 is a genetic overlap between AD and PD susceptibility, which may come from the influence of LRRK2 on AD and PD shared mechanism, such as mitochondrial function, oxidative damage, ubiquitin, neurotransmission, and inflammation system (Fig. 1) [26]. In the subgroups stratified by the presence or absence of the APOE ε4 allele, our results showed rs76904798 of LRRK2 alter LOAD susceptibility in APOE ε4-negative subjects. In addition, the P value for APOE interaction showed that the interaction of each SNP and APOE genotypes was observed between rs76904798 of LRRK2 and APOE. These results indicated a synergistic effect of LRRK2 and APOE on AD risk. Our study of failing to detect an interaction of APOE and other SNPs mainly attributes the age structure of the current study because the effect of APOE on AD risk is much stronger in young patient populations [45]. However, previous studies about the LRRK2 gene in AD have not provided a consistent conclusion. It was negatively associated with AD in Taiwan [46] and southwestern China population [47]; while it was positively associated with AD in a Singaporean population [44] and in our study. This may be caused by the different population structures in these studies, and more studies are needed to be performed to confirm this relation.

Except for the successful replicated SNP, other SNPs were inconsistent with the results of previous mentioned GWAS. Moreover, the reasons of failing to disclose perfect replications may be attributed to several factors. First, as shown in Table 3, many SNPs had a power less than 0.8. This may partly explained by our negative findings. Second, because of gene–gene or gene–environment interactions, the magnitude of the influence of included genetic variants may differ between populations, and our study was performed in the northern Han cohorts who may contain a different allelic heterogeneity compared to white samples. Third, variants at these SNPs did not alter the expression of related protein to impact in LOAD-related pathology so as to alter AD susceptibility. Last, the differences of endogenous genetic background of AD and PD may be the basic and the most important factors of these inconsistent results.

In summary, our study repeated the reported SNPs which were considered to affect PD risk, and we first demonstrated that rs76904798 of LRRK2 significantly influenced LOAD risk. However, the other genetic variants might not play major roles in the genetic predisposition to LOAD in a northern Han Chinese people. However, these genetic variants have been reported to significantly link to AD; for example, genetic variations at SNCA gene have been proved to play a key role in PD pathology [48–50]. A recent study also indicated a close connection between SNCA polymorphism and NFT (tau) pathologies which are critical factors in AD pathology [51]. Although our study did not discover significant results, these conclusions need more functional genetic analyses and independent replications across different ethnicities and districts to elucidate the epidemiologic relevance and the potential biochemical mechanisms.