Neuroinflammation and innate immunity have recently emerged as important contributors to AD pathology. GWAS studies pinpointed the association of immunity-related gene SNPs, including, rs1354106T>G in CD33 rs1846190G>A in HLA-DRB1, with AD. However, these studies were limited in the applicability to non-European populations. Our study reports a significant association of rs1354106T>G with AD in a Middle Eastern population, the Lebanese population, for the first time. This further confirms association results and improves the equity of the previously generated genetic information. On the other hand, the importance of our findings lies in providing further genetic support for the role of immunity-related genes and SNPs in AD. Our study establishes the protective role of rs1354106T>G SNP, in CD33, against AD, previously reported in Sherva et al., 2014 [1] and highlights a potential protective effect of rs1846190G>A in HLA-DRB1 against AD. These protective variants could enhance AD risk assessment in asymptomatic individuals and offer potential drug targets.

Alzheimer’s disease (AD) is the most common neurodegenerative disorder, leading to memory loss and multiple cognitive impairments, and is the fourth leading cause of death worldwide among the elderly population [2]. There are two main forms of AD: familial and sporadic [3]. Familial AD typically presents as autosomal dominant and early onset (EOAD), in individuals under 65 years of age, accounting for 1–5% of all cases. EOAD has been linked to mutations in three genes, the presenilin 1 gene (PSEN1), which is identified in up to 70% of cases with familial AD cases; the presenilin 2 gene (PSEN2) and the Amyloid precursor protein gene (APP) [4]. Sporadic AD, or late-onset AD (LOAD) occurs in individuals older than 65 years, with age being the primary risk factor [5]. LOAD is a complex disorder with several identified risk factors including female sex, traumatic brain injury, depression, environmental pollution, physical inactivity, social isolation, low academic level, and metabolic syndrome [6]. Genetic susceptibility also plays a significant role, particularly the ε4 allele of apolipoprotein E (APOE) [7]. The heritability of LOAD is estimated to be between 60–80% [8]. AD is associated with the presence of β-amyloid (Aβ)-containing extracellular plaques and tau-containing intracellular neurofibrillary tangles in the brains of patients [9]. However, the utility of Aβ as a biomarker for AD has faced challenges, with its detection in about 30% of cognitively normal elderly individuals and with the absence of significant clinical improvements after removing Aβ from the brains of AD patients [10–12]. Neuroinflammation, triggered by pathological damage in the central and peripheral nervous system, is recognized as a significant contributor to AD pathogenesis [13]. This leads to the release of proinflammatory cytokines, chemokines, complement cytokines, and small molecule messengers like prostaglandins, nitric oxide (NO), and reactive oxygen species (ROS) [14]. In addition, persistently activated microglia produce high levels of proinflammatory cytokines and chemokines, leading to neuronal dysfunction [15]. Furthermore, microglia are implicated in synaptic loss, tau phosphorylation, and cognitive decline [16]. Genome-wide association studies (GWAS) indicate that a large percentage of AD risk genes are associated with innate immunity and inflammation, highlighting the critical role the immune system plays in AD pathology [17–19].

The cluster of differentiation 33 gene, CD33, on chromosome 19p13.3, is one of the top-ranked AD risk genes identified by genome-wide association studies (GWAS) and has been replicated in numerous genetic analyses [20, 21]. CD33 belongs to the sialic acid-binding immunoglobulin (Ig)-like family and is a myeloid cell receptor, exclusively expressed by myeloid cells and microglia. It has several functions in cell adhesion, anti-inflammatory signaling, and endocytosis [22]. Clinical and biochemical evidence implicates CD33 in Aβ-associated pathology by affecting microglia-mediated Aβ clearance [23–25].

CD33 has been implicated in modulating AD susceptibility and the pathology of late-onset Alzheimer’s Disease (LOAD) [25–27]. Higher CD33 expression in the parietal lobe is shown to be associated with more advanced cognitive decline or disease status [24]. Other studies show that reduced expression of CD33 allows more efficient phagocytic clearance of pathogenic Aβ by microglia and thus protects against AD [25].

HLA, located within the major histocompatibility complex (MHC) on chromosome 6p21, consists of several highly polymorphic and tightly linked genes [28]. Numerous association studies have confirmed significant associations between certain HLA gene variants within MHC class I and II regions and AD [29]. The upregulation of HLA class II antigens is widely accepted as a definitive marker of activated microglia, which are implicated in the formation of lesions characteristic of AD [30].

The mechanism by which HLA may contribute to Alzheimer’s disease (AD) involves the recognition and processing of pathological protein deposits, such as Aβ peptides, by microglia. Once engulfed by microglia, these proteins are broken down and presented to T lymphocytes in conjunction with specific HLA class I or II molecules. This process triggers B lymphocytes to produce antibodies against Aβ peptides, while activated T lymphocytes target cells producing excessive Aβ for elimination [31]. While this immune cascade is a natural defense mechanism against harmful protein accumulation, excessive reactions may lead to detrimental effects [32, 33]. Consequently, an immune response’s severity, scope, and duration can vary depending on the expression of HLA molecules. Individuals carrying certain pathogenic HLA alleles are at a higher risk of developing specific immune-mediated diseases compared to those lacking these alleles [34].

A large GWAS study, including 1,126,563 individuals 90,338 (46,613 proxy) cases and 1,036,225 (318,246 proxy) controls, identified 38 AD risk loci including CD33 and HLA-DRB1 with SNP variants (RS1354106T>G) and (RS1846190G>A) consecutively [20]. In this report, we aimed to investigate the correlation between these SNPs and AD in a sample of 644 Lebanese individuals, including 127 AD patients and 250 controls.

The characteristics of all study participants are described in Table 1. The average age is 61, with 37.4% being females. Of 612 participants, 28.1% had normal weight, 32.4% were overweight, and 242 (39.5%) were obese. Education levels varied also as 25.9% had no formal education, 59.0% attended some school, 3.3% completed high school, and 12.0% attended university. Additionally, 38.1% of the participants were smokers. Blood pressure and lipid measurements were also recorded.

The characteristics of AD patients and controls are described in Table 2. The mean age of AD patients (80.99 ± 7.94) was significantly greater than the mean age of controls (70.06 ± 8.82) (p < 0.001). Moreover, there were significant differences between AD subjects and controls in terms of marital status, number of smokers.

The SNP allele frequencies detected in our study showed minimal variation from the allele frequencies in the Middle Eastern populations (GnomAD) (Table 3). The minor allele frequencies ranged from 0.23 to 0.49, suggesting that these alleles were relatively common in the studied population. The observed genotype frequencies of rs1846190G>A and rs1354106T>G did not show significant deviations from the Hardy-Weinberg equilibrium (HWE). AG and AA carriers of the rs1846190G>A SNP had a decreased risk of AD (OR = 0.042, p = 0.026 and OR = 0.052, p = 0.031 respectively), indicating a much lower likelihood of developing Alzheimer’s disease. Likewise, the rs1354106GT genotype had a lower risk (OR = 0.173, p = 0.005) compared to the TT genotype, indicating a significantly lower risk of Alzheimer’s disease in the studied population.

Assessment of the association between the six SNPs and the likelihood of developing AD, while adjusting for age, gender, BMI, educational status and smoking showed a significant association with AD for rs1846190G>A (AG; OR = 0.042, P = 0.026 and AA; OR = 0.052, P = 0.031) in HLA-DRB1 and rs1354106T>G (GT; OR = 0.173, P = 0.005) in CD33 (Table 4). When applying Bonferroni correction, only rs1354106T>G in CD33 remained significant thus showing a robust association with AD.

In our study, among the six SNPs analyzed, only rs1846190G>A, a regulatory region variant in HLA-DRB1, and rs1354106T>G, an intron variant in CD33, showed a significant association with AD in the Lebanese population. Following Bonferroni correction, only rs1354106T>G in CD33 remained significant, which highlights the potential importance of this gene in the pathogenesis of AD.

SNPs have the potential to alter CD33’s expression level, structure, and function, altering how microglia clear amyloid β [25, 40, 41]. Two previously reported SNPs in CD33, rs3865444 and rs12459419, have shown a protective effect against AD [42]. The protective allele of the rs3865444, located in the promotor region, plays a role in the reduction of both CD33 expression and insoluble Aβ42 levels in AD brain, especially in the microglial cells [25]. Similarly, rs12459419, located in exon 2, and in linkage disequilibrium with rs3865444, exhibits a protective effect by enhancing exon skipping and promoting the production of a short isoform of CD33, known as human CD33m [43]. Recent studies using cell and animal models have highlighted the functional significance of human CD33m, as a gain-of-function variant that enhances Aβ1–42 phagocytosis in microglia [41].

Conversely, a recent computational analysis investigating the 3D structures of CD33 with rs2455069 A>G SNP suggests a potential increase in the risk of Alzheimer’s disease. The study proposes that over time, the CD33-R69G variant, which binds to sialic acid, could boost CD33’s ability to inhibit the breakdown of amyloid plaques [44].

Our study further explored the association of rs1354106 T>G with AD, revealing a protective effect in Lebanese patients (GT; O. R = 0.173 CI = 0.058–0.586, P = 0.005). This finding notably aligns with the findings from a previous study which utilized a Bayesian longitudinal low-rank regression (L2R2) model to explore the impact of single nucleotide polymorphisms (SNPs). Their results revealed that rs1354106 was associated with a reduced rate of decline in the AD assessment scale cognitive score [1]. Moreover, in the same study, the effect of this SNP on the longitudinal trajectories of the hippocampi was investigated. Results revealed that the minor allele significantly slowed hippocampal atrophy compared to the major allele. This suggests a potential protective effect associated with the minor allele of rs1354106 in patients with Alzheimer’s disease and mild cognitive impairment [45]. This is validated by our findings, which indicated a protective role of the rs1354106 T>G in Lebanese AD patients (GT; O. R = 0.173 CI = 0.051–0.586, P = 0.005).

The association between HLA gene variants and Alzheimer’s disease (AD) risk has been extensively explored across diverse populations. Our study on the Lebanese population, first revealed a protective effect of rs1846190G>A, of HLA-DRB1 but the association did not stand after Bonferroni correction. HLA-DRB1 13:02 protects against age-related neural network deterioration and mitigates the deleterious effects of apoE4 on neural network functioning [46]. Furthermore, a recent study, conducted on the Japanese population, identified a significant association between the HLA-DRB109:01-DQB1*03:03 haplotype and LOAD risk in APOE ε4–negative individuals [47]. Moreover, studies have emphasized the protective function of HLA-DRB1*04 against AD, as its presence is correlated with lower CSF tau levels and fewer neurofibrillary tangles in AD subjects [48]. Conversely, HLA-DRB1*03 was identified as a risk factor for late-onset AD (LOAD) in the German population [31]. Additionally, the SNP rs9271192 in HLA-DRB5–DRB1 region has been found to influence AD risk through large meta-analyses of genome-wide association studies (GWAS) in Caucasian populations [48]. These findings have been replicated successfully in two large-scale studies conducted on the Chinese population [49, 50].

A recent study examined global cortical amyloid PET burden, incorporating the 38 gene variants, from the GWAS study, using PRSice-2, to assess overall phenotypic variance in two cohorts [20]. The analysis revealed a strong association between AD risk variants (such as APOE, PICALM, CR1, and CLU) and amyloid PET levels in both cohorts. Importantly, neither CD33 rs1354106T>A nor HLA-DRB1 rs1846190G>A demonstrated an association with amyloid PET levels in this study [51]. This underscores the alignment of our findings with existing evidence concerning the protective effect of both variants against Alzheimer’s disease risk.

In conclusion, understanding protective variants could refine AD risk assessment in asymptomatic individuals, aiding AD prevention. Furthermore, identifying genetic variants that confer protection via a loss-of-function or gain-of-function offers potential drug targets. Most drug candidates never reach the clinic, but those with the same mechanism as protective variants have a higher success rate. Our current study has provided convincing statistical support for an association between CD33 polymorphisms and LOAD. Specifically, the carriage of GT alleles rs1354106 T>G in CD33 is linked to a protective effect against LOAD in the Lebanese Population. The main limitation of this study is the sample size used, probably affecting the statistical significance of rs1846190 SNP and HLA-DRB1 association with AD after Bonferroni correction. Further investigations involving larger sample sizes and diverse ethnic groups are needed to validate the role of rs1354106 and examine the potential role of rs1846190 in LOAD.