New genetic players in late-onset Alzheimer's disease: Findings of genome-wide association studies.

Late-onset Alzheimer's disease (LOAD) or sporadic AD is the most common form of AD. The precise pathogenetic changes that trigger the development of AD remain largely unknown. Large-scale genome-wide association studies (GWASs) have identified single-nucleotide polymorphisms in multiple genes which are associated with AD; most notably, these are ABCA7, bridging integrator 1 (B1N1), triggering receptor expressed on myeloid cells 2 (TREM2), CD33, clusterin (CLU), complement receptor 1 (CRI), ephrin type-A receptor 1 (EPHA1), membrane-spanning 4-domains, subfamily A (MS4A) and phosphatidylinositol binding clathrin assembly protein (PICALM) genes. The proteins coded by the candidate genes participate in a variety of cellular processes such as oxidative balance, protein metabolism, cholesterol metabolism and synaptic function. This review summarizes the major gene loci affecting LOAD identified by large GWASs. Tentative mechanisms have also been elaborated in various studies by which the proteins coded by these genes may exert a role in AD pathogenesis have also been elaborated. The review suggests that these may together affect LOAD pathogenesis in a complementary fashion.

Introduction

The first case of Alzheimer's disease (AD) was described more than 100 years ago, but the precise pathogenetic changes leading to the development of AD are still a matter of considerable controversy. Based on the age of onset and heredity, AD is classified into early-onset AD (EOAD), late-onset AD (LOAD) and familial AD. LOAD or sporadic AD is the most common form of AD, accounting for about 90 per cent of cases and usually occurring after the age of 65 yr1. Neurofibrillary tangles of phosphorylated tau protein and senile plaques composed of amyloid β (Aβ)-protein are the two characteristic pathological hallmarks of AD; however, there exists controversy in how well these correlate with AD phenotype as some AD brains on post-mortem examination reveal minimal plaques and tangles2.

The protein apolipoprotein E (ApoE) is the only well-established genetic risk factor for LOAD. The APOE gene consists of four exons and three introns, with a total of 3597 base pairs, and is mapped to chromosome 19. ApoE is polymorphic with three major isoforms, ApoE2, ApoE3 and ApoE4. High frequency of the APOE4 allele is found in patients with AD than in the general population3. ApoE4 is known to inhibit neurite outgrowth, disrupt neuronal cytoskeleton4, stimulate tau phosphorylation and cause neurodegeneration5. However, neither is the APOE4 variant present in all AD cases nor is it absolutely essential for AD pathogenesis6. Multiple rare mutations in the amyloid precursor protein gene (APP), PSEN1 gene and PSEN2 gene cause early-onset AD7. However, a large case-control study (3940 cases and 13,373 controls) reported that common variants in these genes were not likely to make strong contributions to susceptibility for LOAD8.

Recent efforts have been focussed on conducting genome-wide association studies (GWASs) to identify newer risk genes for LOAD. Multi-stage meta-analytic reports by different groups documented the association of single-nucleotide polymorphisms (SNPs) in 10 genes with AD; these being ABCA7, bridging integrator 1 gene (BIN1), triggering receptor expressed on myeloid cells gene (TREM), CD33, clusterin gene (CLU), complement receptor 1 gene (CR1), ephrin type-A receptor 1 gene (EPHA1), CD2AP, membrane-spanning 4-domains, subfamily A (MS4A) gene cluster and phosphatidylinositol binding clathrin assembly protein gene (PICALM)9101112.

In 2009, Lambert et al13 published an open letter of two-stage GWAS performed on AD subjects and controls. The three-city study identified two new susceptibility loci: CLU and CR1. They also detected evidence for the association of PICALM with AD13. A collaborative consortium from Europe and the USA [European AD Initiative 1 (EADI 1)] also performed a GWAS over 16,000 individuals with AD and controls. They identified two novel loci CLU and PICALM, significantly associated with AD. They also observed one more associated locus BIN114. In 2010, Seshadri et al15 performed a three-stage analysis of GWAS data to identify additional loci associated with LOAD. In their gene discovery phase, they concluded that BIN1 showed association with AD in GWAS. They also confirmed the association of two reported loci; CLU and PICALM with LOAD15. Hollingworth et al10 undertook a combined analysis of four independent genome-wide studies- GERAD1, TGEN1, ADNI and EADI1 - to identify new susceptibility loci of AD. Their data provided significant evidence for the association of ABCA7, MS4A gene cluster with AD at stage one. In stage two, they observed association of more suggestive loci; CD33 and EPHA1 with AD10. To identify newer susceptibility loci for AD, the AD Genetic Consortium (ADGC) group conducted a three-staged association study on AD patients and provided compelling evidence for the association of MS4A4A, EPHA1 and CD33 with AD. They also replicated previous associations of CR1, CLU and PICALM with LOAD11. Advances in sequencing techniques of entire genomes identified rare variants in those patients, in whom linkage analysis cannot be done. TREM2 is one of the variants that increase the risk of AD12.

Fig. 1 gives a schematic representation of the multiple research groups who worked to find new susceptibility genes for LOAD and also the different loci which affect LOAD pathogenesis.

Fig. 1

Schematic representation of multiple organizations who worked to find new genome-wide association study loci and how different loci are connected with each other. The gene loci found as a result of meta-analyses belong to three broad functional categories: immune response, synaptic function and cholesterol metabolism. GWAS, genome wide association studies; GERAD1, genetic and environmental risk for Alzheimer's disease consortium 1; EADI1, European Alzheimer's disease initiative 1; CHARGE, Cohorts for Heart and Aging research in genomic epidemiology; TGRI, Translational Genomics Research Institute; ADGC, Alzheimer's disease genetic consortium; LOAD, late onset Alzheimer's disease.

Alzheimer's disease (AD) pathogenesis as the cumulative effect of multiple genetic risk factors

Large-scale GWASs have identified SNPs in ten genes: ABCA7, BIN1, TREM2, CD33, and PICALM which may participate in the pathogenesis of AD by several functional pathways that are affected9101112. These genes may be categorized on the bases of their involvement in cellular pathways:

Immune response and inflammation: CR1, MS4A family, EPHA1, CD33, TREM91012.

Lipid (cholesterol) metabolism: CLU and ABCA7910.

Endocytosis and synaptic function: PICALM, BIN1, CD2AP and EPHA1910.

It is hypothesized that these gene SNPs identified by GWAS influence their respective interconnected cellular processes to cause AD. The exact pathogenesis of AD is still unclear, and it is possible that not all of the above processes are deranged in each case of LOAD. Either of the three may dominate in or solely contribute to LOAD in individual patients. Further, the exact links between the pathways still need to be worked out. However, the common pathways through which these act are widely believed to be the amyloidogenic pathway and the tau hyper-phosphorylation pathway1.

Fig. 2 represents how the various genetic risk factors may be interconnected and contribute to LOAD risk by ultimately inducing amyloid and hyper-phosphorylated tau protein accumulation.

Fig. 2

Interconnected responsible pathways to cause amyloid and tau accumulation. Gene involved in AD pathogenesis can be broadly grouped into 3 categories; immune response (CR1, MS4A, TREM2, CD33, EPHA1), cholesterol metabolism (APOE, CLU, ABCA7), synaptic function (PICALM, CD2AP, BIN1). The cumulative effect of all these genes is manifested through the final common pathway of amyloid and tau cascade.

Functional significance of new genetic loci associated with LOAD

Genes associated with lipid metabolism: CLU codes for the secretory hetero-dimeric 75-80 kDa CLU also known as apolipoprotein J16. This gene encodes a 2 kb mRNA which translates into a 449 amino acid primary polypeptide chain17. CLU is a highly conserved chaperone protein that is found in the cell cytosol under some stress conditions18. It is expressed in most mammalian tissues19, and has been reported to be involved in neurodegeneration and hypoxic-ischaemic neuronal death20. Elevated level of CLU has been found in post-mortem AD brains and also in the brains of ApoE4 carriers21. CLU is involved in the regulation of Aβ. This has been demonstrated in guinea pig brain perfusion model where apolipoprotein J interacts with the soluble form of Aβ in a specific and reversible manner and forms complexes in the brain, facilitating the transport of soluble Aβ across the blood-brain barrier22. In transgenic mouse model (clu− and clu+), it has been seen that Aβ deposits in clu− mice are significantly reduced as compared to clu+ which indicates that CLU has a role in Aβ fibril formation and neurotoxicity23. Plasma CLU level was reported to be associated with rapid clinical progression in AD, suggesting its possible use as a biomarker of AD24. GWASs found a significant negative association [odds ratio (OR)=0.86] between an SNP within the CLU, rs11136000 and the risk of having AD14. This association was found in both APOE4 carriers and non-carriers15.

ABCA7 is a member of the superfamily of ATP-binding cassette (ABC) transporters, which transport various molecules across extra- and intra-cellular membranes. These transporters are divided into eight distinct subfamilies. ABCA7 is a member of the ABC1 subfamily25. This gene codes for a membrane protein which is expressed in the myelolymphatic tissues, brain and trachea26. Analysis of isolated foetal human brain cells has shown that microglia express the highest level of ABCA7 mRNA27. This gene is also involved in AD pathogenesis28. It regulates the phagocytosis of apoptotic cell debris inside the brain. Protein products of these loci bind with APOA1 and contribute to the apolipoprotein-mediated phospholipid efflux mechanism in cells29.

In stage 1 meta-analysis of GERAD1, TGEN1, ADNI and EADI1, evidence was found for the positive association (OR=1.22) of SNP of ABCA7 (rs3764650) with AD. This has further been proven in stages 2 and 3 meta-analysis10. Another SNP variant of ABCA7, i.e. rs3752246, was found to be associated with AD in stage 2 meta-analysis (OR=1.17). However, association of rs3764650 with ABCA7 expression was not observed30.

Genes associated with inflammatory response: CD33 is located on chromosome 19q13.3 in humans and codes for the 67kDa CD33 protein31. CD33 belongs to the sialic acid-binding immunoglobulin-like lectins (Siglecs) family32. It is expressed in microglial cells in the human brain33. The Siglecs family mediates cell-cell interaction through glycan recognition34. They also play an important role in the regulation of functions of innate and adaptive immune cell systems35. CD33 is expressed by haematopoietic and phagocytic cells and participates in adhesion processes of human primary immune cells36. It appears to inhibit the production of pro-inflammatory cytokines [such as interleukin-1β, tumour necrosis factor alpha (TNF-α)] by monocytes37. Being an inhibitory receptor in immune response, it also regulates cell growth and survival and also induces apoptosis38.

CD33 inhibits Aβ clearance in LOAD39. It has been seen that levels of CD33-positive microglial cells are increased in brains of AD patients, and play a direct role in the progression of AD. The CD33 SNP rs3865444, which confers protection against AD, has been seen to be associated with reductions in both CD33 expression and insoluble Aβ42 levels in AD brain33. Various SNPs of CD33 such as rs3826656 and rs3865444 are found to be associated with AD40.

CR1 found on chromosome 1q32 codes for the complement regulatory protein, CR1 or CD35 which is expressed widely on a number of blood cells41 and can also be found dissolved in the blood plasma42. CR1 induces phagocytosis by forming a complex with C3b/C4b. Extracellular domain of CR1 is composed of long homologous repeats (LHRs). Genetic duplications and deletions result in increased number of LHR regions, which result in the formation of four co-dominant alleles of CR1. Frequencies of the four alleles vary only slightly between populations43. The increased number of LHRs means that the larger alleles have additional C3b/C4b-binding sites44.

The classical complement pathway has been long known to play a protective role in AD by acceleration of clearance of the Aβ plaques. Aβ interacts with C1q of the classical complement pathway45. This results in the activation of the membrane attack complex comprising C3b/C4b, which results in activation of glial cells46. CR1 helps in this process by providing multiple C3b/C4b-binding sites47. Lambert et al13 found an SNP variant of CR1, rs6656401 (OR=1.12) with a strong association with LOAD.

EPHA1 also known as eph is located on chromosome 7q34.1. The protein product belongs to the tyrosine kinase receptor family48 and the ephrin receptor subfamily. The ligand for the EphA receptor is ephrin-A, which is anchored to the cell membrane through a glycosylphosphatidylinositol linkage49. Eph receptors and ephrins are expressed in endothelial and epithelial cells50, and guide the migration of cells during embryonic development and also have a role in cytoskeletal organization of neuronal processes51. They play a role in synaptic development and plasticity52. Additional roles in apoptosis and inflammation exist53. AD patients with an allele of EPHA1 (A allele) having enhanced rate of cerebral metabolism for glucose in the right lateral occipitotemporal gyrus and inferior temporal gyrus may not have hippocampal atrophy54. Combined result of the meta-analysis of the GERAD consortia with the ADGC GWAS shows that the rs11767557 SNP of the EPHA1 gene is negatively associated with AD (OR=0.90)10.

MS4A encodes several proteins including CD20. This gene family is further divided into at least 12 subgroups from MS4A1 to MS4A1255. CD20 expressed by B-lymphocytes56 forms a hetero-tetrameric complex on the cell membrane that regulates Ca2+ influx downstream57. This regulation of calcium signalling may have an important role in neurodegeneration and AD pathogenesis58. Several members of this cluster (such as MS4A1, MS4A2 and MS4A4B) have an important role in immunity59. MS4A4B appears to have a role in Th1 development, CD8+ memory T-cell function and modulation of regulatory T-cell signalling60. MS4A2 mediates interactions with IgE-bound antigens that lead to cellular responses such as the degranulation of mast cells61.

Meta-analysis data of GWAS by ADGC suggested two SNPs of the MS4A gene cluster: rs610932 and rs670139 to be associated with LOAD10. Another independent GWAS study on the Spanish population revealed the association of rs1562990 SNP of MS4A with AD62.

TREM2 codes for a membrane glycoprotein, consisting of an extracellular immunoglobulin-like domain and a cytoplasmic tail that is involved in receptor signalling complex along with the DAP12 and TYRO binding proteins63. This protein functions in the immune response and may be involved in chronic inflammation64. In brain cells, TREM2 is primarily expressed on microglia6566. Microglia stimulate the proliferation of CD4+ T-cells, as well as the secretion of TNF and CCL267. Microglia have phagocytic role on amyloid plaques68. In a study, reduced phagocytic activity was found in microglial cells to phagocytose β amyloid fragment of AD brain in TREM2 knockdown mice in comparison with mice expressing TREM269. A rare missense mutation (rs75932628) in the TREM2 results in an R47H substitution which has been found to confer a significant risk of AD. This may be because of the inability of the brain to clear Aβ toxicity65.

Genes associated with endocytosis: PICALM codes for PICALM which can influence the risk of AD through modulation of APP processing via AP2-dependent clathrin-mediated endocytotic pathways, resulting in changes in Aβ level70. PICALM initiates clathrin polymerization at sites of coated pit formation71. It was seen in cell culture experiments that clathrin-mediated endocytosis (CME) retrieved full length APP from the cell surface, thus promoting the intracellular accumulation of amyloid72. In the endosome, full length APP is cleaved in to Aβ by β-secretase (BACE) and this is released into the brain interstitial fluid. Increased number of endosomes formed by CME drives more APP into the cell73, resulting in an increase of Aβ production74. Synaptic vesicles limit the dispersion of neurotransmitter at the pre-synaptic plasma membrane. It was seen in live cell image of hippocampal neurons that synaptic vesicle containing VAMP2 on surface helped in diffusing neurotransmitters along the axonal membrane75. PICALM may also be involved in directing the trafficking of VAMP276. The SNP of PICALM which has been found to be most significantly protective against LOAD is rs3851179 (OR=0.86)14.

BIN1 codes for Myc box-dependent-interacting protein 1. It is a nucleo-cytoplasmic tumour suppressor adaptor protein77. Isoforms of this protein expressed in the central nervous system are involved in synaptic vesicle endocytosis78. The BIN1 is identified as the most important genetic susceptibility locus in LOAD after APOE79. Higher BIN1 expression has been reported to be linked with later age at onset and shorter disease duration80. Although the mechanisms are still not fully understood, data suggest that BIN1 affects AD risk primarily by modulating tau pathology. BIN1 also affects other cellular functions including endocytosis/trafficking, inflammation, calcium homoeostasis and apoptosis79. Seshadri et al15 combined the data from CHARGE, TGEN, EADI1 and GERAD1 groups and analyzed by a three-stage sequential meta-analysis. They reported the association (OR=1.13) of the BIN1 SNP rs744373 with LOAD15. Another independent study- The Washington Heights-Inwood Columbia Aging Project and the Estudio Familiar de Influencia Genetica de Alzheimer study also showed positive associations of the BIN1 SNP rs7561528 with LOAD in the ε4 carrier state81.

CD2AP codes for CD2-associated protein which is a scaffolding molecule that regulates the actin cytoskeleton82. It plays a role in receptor-mediated endocytosis. CD2AP contributes to APP metabolism and subsequent Aβ generation83. It regulates the encounter of APP and BACE1 in axonal and dendritic endosomes84. GERAD1, EADI1, deCODE and AD-IG GWAS datasets observed independent evidence for the association of CD2AP gene loci with AD (OR=1.11 for rs9349407 SNP)10.

Racial variation of Alzheimer's disease susceptibility genes

Survival after the diagnosis of AD varies amongst different races, ranging from 3 to 9 years. African American and Latino AD patients have better survival than Caucasian patients and genetic background plays an important role in the progression of AD85. Most GWASs and replication studies of AD have been done in populations of European descent, and non-European genetic studies of new AD-susceptibility loci are limited. Studies that evaluated the association of CLU and CR1 with AD in Asian populations are limited86. Many AD-associated SNPs of CLU, PICALM and BIN1 were not necessarily identical in Caribbean Hispanic individuals compared with a European American data set81. Meta-analytic data showed that CLU, PICALM and CR1 were associated with LOAD in Caucasians subjects, but a study found that investigated SNPs of CR1, CLU and PICALM were not associated with AD in a Polish population87. A study found that in the Korean population, the PICALM is the only AD susceptibility loci in addition to APOE88. ADGC assembled multiple data sets for meta-analysis representing African American older subjects. The data showed another SNP (rs115550680) of ABCA7 (OR=1.79) was associated with AD in comparison to European ancestry89.

Potential therapeutic implications of GWAS loci

Novel loci may exert their effects in a number of pathways such as oxidative balance, protein metabolism, cholesterol metabolism and synaptic function90. Genes with moderate to large effects on LOAD risk are valuable targets for therapeutic development. Neuroinflammation is both a cause and a consequence of AD and treatment with anti-inflammatory agents is likely to be successful if initiated before the onset of neurological symptoms91. Similarly, on the lipid metabolism front, the CLU protein may be targeted to reduce AD risk92. Genes associated with endocytosis and synaptic functions are BIN1, PICALM and CD2AP. Modulating these at the gene-expression level using siRNA or antisense techniques is a valid approach.

New developments

While the present review focuses on the most established gene loci involved in AD pathogenesis as suggested by GWAS, several newer loci have made a foray into the AD scene. Under the supervision and support of International Genomics of AD project, two-stage meta-analysis identified 11 loci which are HLA-DRB5-DRB1 gene, SORL1, PTK2B, SLC24A4, ZCWPW1, INPP5D, MEF2C, CELF1, NME8, CASS4, FERMT2 genes, with suggestive evidence of association with AD93. The Table represents newer loci involved in AD pathogenesis as suggested by GWAS with tentative pathogenic mechanisms94.

Table

New susceptibility gene loci of Alzheimer's disease

Conclusion

GWASs have revealed the association of new gene loci with AD. The first few identified SNPs from GWAS suggest the involvement of different associated pathways with pathogenesis of AD although the exact mechanisms remain unknown. Modification and advancing the research in these pathways may lead to therapeutic intervention for AD. Many of these GWAS loci may serve as biomarkers of AD. The search for additional genetic risk factors requires more large-scale meta-analysis of GWAS and enhanced statistical power as well as replicating these findings at the molecular level. Exciting times await us in AD genetic research and newer paradigms might open in the near future.