Introduction

Alzheimer’s disease (AD) is the most common neurodegenerative disease and represents a serious public health problem worldwide (Cacabelos et al. 2005; Defina et al. 2013). AD is categorized into two major groups, early-onset (EOAD) and late-onset Alzheimer’s disease (LOAD), differing in terms of their symptomatic, biological, genetic, and neurophysiological characteristics (Biagioni and Galvin 2011). EOAD corresponds to <1–6 % of all cases of Alzheimer disease (Avramopoulos 2009). This condition is inherited in an autosomal dominant pattern and caused by highly penetrant mutations in one of the three known genes (PSEN1, PSEN2, and APP). Affected people with this type of disease have the disease onset prior to age 65 (Goate et al. 1991; Levy-Lahad and Bird 1996; Sherrington et al. 1996).

In contrast to EOAD, sporadic or late-onset Alzheimer’s disease has heterogeneous etiology and affects individuals older than 65 years with modest or no familial clustering (Bird 2008; Cacabelos et al. 2005; Zetzsche et al. 2010).

Although LOAD patients exhibit no clear pattern of inheritance, twin and family-based studies have suggested a significant genetic component in the etiology of LOAD (Gatz et al. 2005, 2010). Results of scientific research have implicated complex interactions between different genetic variants, age, gender, and other environmental factors that modulate LOAD risk (Goldman et al. 2011; Reitz et al. 2011).

According to these studies, the estimated heritability in LOAD ranges from 25 to 75 % (Wilson et al. 2011). In addition to the ε4 allele of the apolipoprotein E gene (ApoE), which is the only genetic factor that unambiguously confers increased risk of LOAD, recent GWASs (genome-wide association studies) have implicated loci of other genes with more modest effects on this disease (Table 1).

Table 1 Investigated SNPs, previously explored in different genome-wide association studies, case–control studies, and family-based studies

Some of these loci are mapped to genes with established roles in Aβ metabolism [APOE, apolipoprotein J (clusterin), phosphatidylinositol-binding clathrin assembly protein], neuroinflammation (complement receptor 1, tumor necrosis factor-α, T cell receptor, CRR2, CRR5), calcium signaling (calcium homeostasis modulator 1), and hormonal regulation (estrogen receptor 1), and all are recognized as novel putative LOAD risk loci (www.alzgene.org).

APOE exerts its effect on the risk and age of onset distribution in a dose-related manner (Sando et al. 2008). APOE protein plays a critical role in Aβ (amyloid β) homeostasis, promotes the proteolytic clearance of amyloid β, and functions as a chaperone (Liu et al. 2009; Schellenberg and Montine 2012). Clusterin is a multifunctional glycoprotein. Both APOE and clusterin molecules are involved in Aβ metabolism and deposition and are protective agents against Aβ neurotoxicity (Koren et al. 2009). Both PICALM and BIN1 are ubiquitously expressed genes, most abundantly in the brain, and are involved in intracellular trafficking of proteins such as vesicle-associated membrane protein 2 (VAMP2) and neurotransmitters through clathrin-mediated endocytosis (CME) (Ando et al. 2013; Harel et al. 2008; Parikh et al. 2014; Xiao et al. 2012).

Neuroinflammation as a common feature of the brain pathology presents in Alzheimer’s disease is associated with immune mediators’s up-regulation. TNFα, a pro-inflammatory cytokine, a well-known immune mediator with modulating effects on memory and synaptic function, is up-regulated in AD patients and animal models of the disease (McAlpine et al. 2009; Tweedie et al. 2012).

CCR2, CCR5, and their related ligands are involved in the accumulation of microglia at sites affected by neuroinflammation (Galimberti et al. 2004; Navratilova 2006). CCR2 and its ligand CCL2 (MCP-1) is involved in the metabolism of Aβ, and CCR5 receptor has role in the regulation of brain immune responses in AD (Harries et al. 2012). A valine to isoleucine substitution at codon 64 within the first membrane region of CCR2, CCR2V64I, causes a controversial influence on the expression of CCR2 on the cell surface, while CCR5∆32 polymorphism creates a premature stop codon and has a protective effect toward inflammatory diseases such as AD (Sezgin et al. 2011).

Toll-like receptor 2 (TLR2), located in the Alzheimer dementia (AD) linkage region on 4q, is involved in the microglia-mediated inflammatory response (Blacker et al. 2003; Landreth and Reed-Geaghan 2009). TLR2 is a member of pattern recognition receptors in the innate immune system. TLR2 may also be essential for Aβ clearance and in that way provides neuroprotection in AD (Bsibsi et al. 2002; Richard et al. 2008). Calcium homeostasis modulator 1 (CALHM1), mainly expressed in the hippocampus, encodes a multipass transmembrane glycoprotein that controls cytosolic Ca2+ concentrations and Aβ levels (Shoji et al. 2005). A non-synonymous polymorphism, rs2986017 (p.P86L) in the CALHM1 gene, was reported to affect calcium homeostasis and promotes Aβ accumulation via a loss of CALHM1 control on Ca2+ permeability and cytosolic Ca2+ levels and increases the risk of AD (Dreses-Werringloer et al. 2008).

ESR1 is one of the receptors through which estrogen exerts its biological effects and mediates the effects of estrogen on AD (Sundermann et al. 2010). According to some reports, AD is more prevalent in women, and the use of estrogen by women after menopause is associated with a lower risk of AD or delayed disease (Ma et al. 2009). Numerous laboratory studies have demonstrated inhibitory effects of estrogen on Aβ plaque formation and its antioxidant and anti-apoptotic functions (Ba et al. 2004; Chiueh et al. 2003).

We aimed to examine the possible associations of 19 different risk factors for LOAD, identified recently by GWAS, in different genetic models, and their possible interaction with ApoE genotype conditions in LOAD within the Azeri Turkish population of Iran.

Materials and Methods

A total of 160 patients, 66 males and 94 females (age range between 65 and 99 and mean age 76.06 ± 7.75 years), with LOAD were recruited from northwest of Iran. The patients were screened by their physicians in the Neuroscience Research Center between February 2010 and July 2014 according to the DSM-IV diagnostic criteria (American Psychiatric Association 1994; Alberoni et al. 2000) and were referred to our laboratory for genotyping of candidate genes. The control group consisted of 163 ethnically, sex-matched 68 male and 95 female (age range between 65 and 89 and mean age 75.29 ± 6.75 years) participants. They underwent neurological and medical examinations, which showed that they were free of any symptoms suggestive of cognitive decline (Gharesouran et al. 2014). Among different racial or ethnic groups living in Iran, including Persian (51 %), Azeri Turk (24 %), Kurd (7 %), Arab (3 %), and other minorities, we limited our investigation to 15–24 million Azeri Turks in northwestern Iran.

Excluding patients with disease onset age before 65 years and from families with two or more affected people within more than one generation, ensured the sporadic form of the disease (Ray et al. 1998). Written informed consent was obtained from each participant, or their legally authorized representatives, who were accepted previously by the ethics committee of special clinics at the Tabriz University of Medical Sciences. This study was approved by the review board of the Neuroscience Research Center and Immunology Research Center at the Tabriz University of Medical Sciences.

Variants on CR1, PICALM, BIN1, and APOE genes were genotyped by PCR and direct sequencing. Genotypes related to polymorphisms on TNF α, CCR2, CLU, TLR1, and ESRα genes were determined by PCR–RFLP reaction, and the CCR5Δ32 genotype was determined by PCR without RFLP. The purified PCR products of these genes from AD cases and healthy controls were randomly sequenced bidirectionally (Gharesouran et al. 2014).

The Hardy–Weinberg equilibrium (HWE) was estimated using the Chi-square test. Differences in allele and genotype distribution between the LOAD patients and healthy controls were analyzed using the Chi-square and Fisher’s exact tests. We used the Bonferroni method to adjust for multiple statistical tests. A Bonferroni-corrected p value of 0.05 (based on the number of SNPs analyzed and genetic models) was regarded as statistically significant. The odds ratio (OR) was calculated at 95 % confidence interval (CI), whenever possible.

To assess the role of interaction of the APOE ɛ4 allele with these polymorphisms, the stratified analysis was performed regarding the existing of ApoE ɛ4 allele. To adjust case and control for APOE ε4 status, subjects were divided into the APOE ε4-positive and APOE ε4-negative subgroups. In addition, the significance of the SNPs association with LOAD was tested in the dominant, additive, and recessive models for SNPs with minor allele homozygote counts of more than 14.

Results

Allele and genotype distribution of the investigated SNPs is shown in Table 2. The distributions of investigated markers were in HWE for both AD patients and controls (p > 0.05, corrected p = 0.003). Allele distributions in CCR2 (rs1799864), TNF α (rs1800629), PICALM (rs541458), BIN1 (rs744373), APOE, TLR2 (−196 to −174 Del), ESRα (PvuII), and ESRα (XbaI) (8 of the 19 investigated polymorphisms) were significantly different between the LOAD and control groups. However, CALHM1 (rs2986017) showed nominally significant association with an increased risk of LOAD, but missed criteria for significance after Bonferroni correction (p = 0.01, corrected p = 0.19).

Table 2 Genotype frequencies, allele frequencies, and statistical analysis results for 19 investigated polymorphisms of different genes among LOAD patients and control subjects

Significant differences were revealed in genotype distribution of CCR2 (rs1799864), CALHM1 (rs2986017), TNF α (rs1800629), PICALM (rs541458), APOE, CLU (rs11136000), and TLR2 (−196 to −174del) polymorphisms among the LOAD and control groups in this ethnic group (Table 2). CALHM1 (rs2986017), CLU (rs11136000), BIN1 (rs744373), ESRα (PvuII), and ESRα (XbaI) genotype distributions among the LOAD and control groups were only significant before Bonferroni correction (Table 2).

As shown in Table 3, seven markers, CCR2 (rs1799864), CLU (rs11136000), ESRα (PvuII), ESRα (XbaI), TLR2 (−196 to −174 del), TNF α (rs1800629), and APOE had minor allele homozygote counts more than 14, and the significance of their associations with LOAD was tested in a dominant, additive, and recessive model.

Table 3 Re-evaluation of the association of 7 markers with minor allele homozygote counts more than 14, in dominant, recessive, and additive models

Among them, CCR2 (rs1799864), ESRα (PvuII), ESRα (XbaI), TNF α (rs1800629), and APOE were correlated with the LOAD in three different statistical models. However, ESRα (PvuII) in dominant model and ESRα (XbaI) in dominant and recessive models missed their correlations with the LOAD after Bonferroni correction (corrected significant p = 0.002).

TLR2 (−196 to −174 del) was associated with the risk of LOAD only in the additive and dominant models (recessive model p = 0.073), and CLU (rs11136000) showed no association with LOAD in additive and dominant models (p = 0.27 and p = 0.16, respectively). CLU (rs11136000) association with LOAD in recessive model (p = 0.0329) missed after Bonferroni correction (Table 3).

After adjusting for APOE, statistical analysis of the allelic distribution showed an association with PICALM (rs541458), BIN1 (rs744373), CCR2 (rs1799864), TNF α (rs1800629), TLR2 (−196 to −174 del), and ESRα (PvuII) only among subjects without the APOE ε4 allele. The association with CCR5, ESR α (XbaI), and TNF α (rs1800630) was evident only among subjects with the APOE ε4 allele.

Among them PICALM (rs541458), CCR5 (rs333), and ESR α (XbaI) missed their association after Bonferroni correction (corrected significant p = 0.003).

The interactions of other SNPs with the APOE ε4 allele were not statistically significant. The genotype distribution of the investigated SNPs after adjusting for APOE is shown in Table 4.

Table 4 Allelic and genotypic distribution of investigated polymorphisms in cases and controls stratified by APOE ε4 allele status

Discussion

Alzheimer’s disease is clinically characterized by a progressive decline of memory with pathogenic features including amyloid plaques, oxidative stress, dysfunctional calcium homeostasis, hormonal dysregulation, and decline in immune system function (Melesie and Dinsa 2013; Wuwongse et al. 2010). AD, particularly multifactorial type, LOAD, is immensely complex on the molecular level (Bertram and Tanzi 2012).

In this article, we report the results of case–control investigations of potential risk factors for LOAD in northwestern Iran of patients with Azeri Turkish origin (Gharesouran et al. 2013). We also re-examined the SNPs association with LOAD in different genetic models. We then compared allele and genotype distribution of the variants after adjustment for APOE status. Overall in the present study, 19 of the most replicated markers association with LOAD were assessed (Table 2).

The ε4 allele of apolipoprotein E accounts for 20–70 % of the LOAD risk and offers odds ratios (ORs) ranging from 6 to 30 in the APOE ε4/ε4 genotype carriers for disease association (Ertekin-Taner 2007, 2010; Corneveaux et al. 2010). APOE ε4, compared with APOE ε3 and APOE ε2, has less three-dimensional folding stability (Koren et al., 2009; Mahley and Huang 2006; Hatters et al. 2006). APOE ε4 is also not as efficient at repairing neuronal damage, transporting cholesterol, and delivery of cholesterol to neurons as ApoE ε3 (Liu et al. 2009; Kok et al. 2011).

According to our statistical results, APOE ε4 is a LOAD risk factor in the additive model (OR 8.430; 95 % CI 4.784–14.854), dominant model (OR 6.619; 95 % CI 3.617–12.109), and recessive model (OR 17.821; 95 % CI 4.175–76.075). In a previous study in Iran, the difference between individuals with and without a ε4 allele regarding the susceptibility to LOAD was a risk factor of 6.5 (Gozalpour et al. 2010). In the Gyungah et al. study on 7070 cases with AD and 8169 cognitively normal controls (13 cohorts), including white, African-American, and Caribbean Hispanic individuals, a APOE ε4 significant association with AD was revealed (ORs 1.80–9.05) in all groups except the Amish and Arabs (Gyungah et al. 2010). Although the contribution of APOE allele is known to make the strongest association with LOAD susceptibility, much remains to be learned about the contributions of loci with more modest effects identified by genome-wide association studies (GWASs).

Failing in localize additional highly penetrant LOAD susceptibility genes and regarding to pathogenic features of AD, it seems that multiple low penetrate alleles, with modest effects and related to genes with roles in calcium homeostasis, in hormonal regulation, and in the immune system, can be considered to be possible risk factors for LOAD.

CALHM1 plays a role in controlling cytosolic Ca2+ levels and APP processing (Marambaud et al. 2009). The P86L polymorphism minor allele in this gene was reported to be a predisposing factor in LOAD. Unlike the Japanese, the T allele distribution was increased in AD cases as compared to controls with ORs ranging from 1.29 to 1.99 in the combined population from USA, France, UK, and Italy (Dreses-Werringloer et al. 2008; Inoue et al. 2010). Although the minor allele (A) at SNP rs2986017 within CALHM1 showed a nominally significant association with an increased risk of LOAD (nominal p = 0.011; OR 2.232; 95 % CI 1.223–4.073) suggesting that the A allele is the risk allele, this did not remain significant after Bonferroni correction.

Some studies propose a gender- and race-specific pattern for ESR1-mediated hormonal effects on clinical outcomes in LOAD (Sundermann et al. 2010; Xing et al. 2013). The recent reports demonstrate a protective effect of the x and p alleles in addition to the px haplotype of the Xbal and PvuII SNPs against LOAD in the Italian and Caucasian population (Becherini et al. 2000; Corbo et al. 2006; Lambert et al. 2001). In a case–control study involving Japanese individuals, a greater prevalence of the X and P alleles in patients versus controls was revealed. Our results present evidence that indicates the risk of dementia associated with X and P alleles and their related genotypes and haplotypes. The association was also replicated in three different genetic models, although not all of these associations were statistically significant after adjustment for multiple comparisons (Table 3). A significant association was observed between increased risk of LOAD and X allele in participants who carried the APOE ε4 without Bonferroni correction (p = 0.0279), whereas the risk of LOAD regarding PP genotype and P allele was significant for participants who did not carry the APOE ε4 allele (p = 0.002 and p = 0.0005 for allelic and genotypic distribution, respectively) (Table 4).

TNFα, CCR2, TLR2, BIN1, and PICALM contribute to the LOAD pathogenesis in different ways, especially with their role in the immune system. Among different polymorphisms in the promoter of the TNFα gene with effects on the transcription rate and susceptibility to different diseases, rs1800629 associates with an elevated transcriptional activity. Significant elevated serum concentrations of TNFα have been reported in Alzheimer patients compared to controls (Gezen-Ak et al. 2013). In southern China, Spain, and US populations, the effect of this variant was revealed on the age of onset of LOAD (Randall et al. 2009; Alvarez et al. 2002). Similarly, the allele and genotypic distribution between case and control groups provided an evidence of the possible role of rs1800629 in LOAD pathogenesis in the Azeri Turkish population (p < 0.001; OR 0.12.36; 95 % CI 7.473–20.44). Allelic but not genotypic distribution in the APOE ε4 adjusted subgroups revealed a negative interaction between this marker and the ε4 allele of APOE (Table 4).

TLR2 is essential for Aβ clearance, activation of microglia, and induction of phagocytosis. TLR2 deficiency in transgenic AD mice could increase Aβ deposition and accelerate cognitive decline (Bsibsi et al. 2002; Richard et al. 2008). Assessing the involvement of −196 to −174 Del in TLR2 in developing LOAD suggests a significant association between this variant and the risk of LOAD in our investigation as in the Chinese population (Yu et al. 2011).

Considering adjustment for multiple comparisons, this association was confirmed in additive and dominant models. After adjustment for APOE ε4, frequency distribution of −196 to −174 Del genotypes but not alleles differed significantly between patient and control groups among non-APOE ε4 carriers (Tables 3, 4).

CCR2 is believed to mediate blood monocyte extravasation for sites of inflammation and be involved in macrophage recruitment to the injured peripheral system (Vande et al. 2003). In our study population, the occurrence of rs1799864 CCR2-64I allele polymorphism is decreased in LOAD patients (p < 0.001; OR 0.222, 95 % CI 0.142–0.347) and also a low frequency of the genotype 64I/64I in AD patients proved a real protective effect of this polymorphism on AD (2.5 vs 12.8 %) (p < 0.001). Our finding about this polymorphism is in agreement with the results of Galimberti et al. carried out on the Italian population (p = 0.037; OR 0.65; 95 % CI 0.41–1.03) (Galimberti et al. 2004). Our findings showed the association in dominant, recessive, and additive models. After adjustment for APOE, and considering Bonferroni correction, the association limited only in the group without carrying the APOE allele (Table 4).

BIN1 has a key role in regulating endocytosis and endolysosomal trafficking pathways that, through which APP, Aβ, and ApoE are all internalized, suggests a potential association with AD pathology (Tan et al. 2013). Higher levels of BIN1 expression have recently been reported to be associated with later age at onset and shorter disease duration in AD patients (Karch et al. 2012). In Seshadri et al. study, three-stage meta-analysis (8,371 LOAD cases and 26,965 controls) on variant rs744373 offered the strongest association with LOAD after ApoE, CLU, and PICALM (p = 1.6 × 10−11, OR 1.15) (Seshadri et al. 2010). Lambert et al. replicated the association of the BIN1 rs744373 variant with the risk of AD in three European populations (p = 2.9 × 10−7; OR 1.26; 95 % CI 1.15–1.38) (Lambert et al. 2010). In our research, only BIN1 rs744373 marker allelic association was replicated in the investigation of possible associations among three selected SNPs on BIN1 and LOAD (allelic distribution: p < 0.0001, OR 95 % CI 2.847 (1.562–5.187); genotypic distribution p = 0.006). After adjustment for APOE status, both allelic and genotypic distribution of this variant showed association, only among APOE ε4 non-carriers, with LOAD.

Significant associations between LOAD and several SNPs close to PICALM have been demonstrated, including rs12800974, rs17159904, and rs541458 (www.alzgene.org). Harold et al. reported the first significant evidence for these SNPs association (rs541458: p = 8.3 × 10−10; OR 0.86) with LOAD (Harold et al. 2009; Piaceri et al. 2011). In our study, among three variants on PICALM, rs541458 was associated with an increased risk for LOAD occurrence.

After adjustment for APOE status, both the allelic and genotypic association observed for this variant limited only in APOE ε4 non-carriers. This association did not remain significant after Bonferroni correction (p = 0.0265; p = 0.0366 for allelic and genotypic distribution, respectively).

Similar to other multifactorial diseases, the absence of the risk genes decreased the likelihood but does not completely rule out the disease occurrence. However, finding the LOAD associated markers in an individual raises the chances of disease from a priori risk for the general population to higher risk and increases the odds with the age of presentation. In conclusion, we have replicated CCR2 (rs1799864), ESRα (PvuII), ESR1α (XbaI), TNFα (rs1800629) as well as APOE, associations with LOAD susceptibility in Azeri Turk ancestry populations in dominant, recessive, and additive models.

Our results reveal after adjusting for APOE and considering Bonferroni adjustment for multiple testing, the association with TNFα (rs1800630) (regarding allelic distribution) was evident only among subjects with the APOE ε4 allele, whereas TNFα (rs1800629) (allelic distribution), TLR2 (genotypic distribution) and CCR2, ESR1α (PvuII), and BIN1 (rs744373) (both allelic and genotypic distribution) differences between case and control groups were restricted to non-APOE ε4 carriers. It seems that the genetic effect of these variants is relevant in predisposing to LOAD only in the absence of the APOE ε4 allele, while in ε4 carriers the genetic effect is determined by this robust susceptibility factor.