GAB2 gene does not modify the risk of Alzheimer's disease in Spanish APOE 4 carriers

Abstract

Objectives

The genetic basis of Alzheimer's disease (AD) is being analyzed in multiple whole genome association studies (WGAS). The GAB2 gene has been proposed as a modifying factor of APOE ε4 allele in a recent case-control WGAS conducted in the US. Given the potential application of these novel results in AD diagnostics, we decided to make an independent replication to examine the GAB2 gene effect in our series.

Design

We are conducting a multicenter population-based study of AD in Spain.

Participants

We analyzed a total of 1116 Spanish individuals. Specifically, 521 AD patients, 475 controls from the general population and 120 neurologically-normal elderly controls (NNE controls).

Methods

We have genotyped GAB2 (rs2373115 G/T) and APOE rs429358 (SNP112)/rs7412 (SNP158) polymorphisms using real time-PCR technologies.

Results

As previously reported in Spain, APOE ε4 allele was strongly associated with AD in our series (OR=2.88 [95% C.I. 2.16–3.84], p=7.38E-11). Moreover, a large effect for ε4/ ε4 genotype was also observed (OR=14.45 [95% C.I., 3.34–125.2], p=1.8E-6). No difference between the general population and the NNE controls series were observed for APOE genotypes (P>0.61). Next, we explored GAB2 rs2373115 SNP single-locus association using different genetic models and comparing AD versus controls or NNE controls. No evidence of association with AD was observed for this GAB2 marker (p>0.17). To evaluate GAB2-APOE gene-gene interactions, we stratified our series according to APOE genotype and case-control status, in accordance with the original studies. Again, no evidence of genetic association with AD was observed in any strata of GAB2-APOE loci pair (p>0.34).

Conclusion

GAB2 rs2373115 marker does not modify the risk of Alzheimer's disease in Spanish APOE ε4 carriers.

Introduction

Aetiologically, complex diseases are caused by an admixture of genetic and non-genetic (mainly exposures) factors working together to provoke the phenotype. These multiple factors interact disrupting the physiology and/or anatomy of target tissues and causing the clinical phenotype as a consequence of the process. The number of factors involved in each pathology and the nature of the interaction among them remain largely unknown for all complex diseases. Genetic loci involved in common diseases are generally considered unnecessary and insufficient protective or predisposing factors that influence the apparition of the phenotype, and are necessarily modelled by environmental exposures (gene-environmental interactions) or other genetic elements (epistasis or gene-gene interactions).

Alzheimer’s disease (AD) is the most prevalent neurodegenerative disorder afflicting elderly populations (1). AD is insidious, usually starting with a slight or mild cognitive decline evolving progressively to Dementia in the course of the disease. It is well established that AD is due to severe synaptic transmission impairment in the brain as a consequence of a progressive neuronal cell loss in the central nervous system (CNS). Key pathological findings in brains of AD patients are intracellular neurofibrillary tangles (NFTs) composed by the microtubule-associated tau protein aggregates and extracellular accumulation of beta-amyloid deposits (2).

APOE 4 allele is a major risk factor involved in Late Onset Alzheimer’s Disease (LOAD) (3). It is well established that individuals carrying one copy of this allele have 2 to 4- fold risk of AD compared with non-carriers. Furthermore, individuals carrying two copies of the 4 allele multiply their AD risk by 15 (4). Although the APOE 4 effect size may vary depending on ethnic background, (5) its association with the disease have been confirmed in many populations and it is recognized as one of the best characterized non-mendelian genetic factors for any complex disease (6).

Beyond APOE involvement in LOAD, multiple genetic markers have been proposed as candidate risk factors for LOAD. However, most of them remain questioned by the scientific community or simply have never been replicated by independent research groups (4). This situation has decreased the credibility of genetic association studies, and the suggestion that APOE is the main genetic risk factor involved in LOAD (3).

However (and by definition) if the APOE locus is firmly associated to the disease, the existence of differences in APOE

4 allele effect size among populations can only be explained by gene-gene or gene-environmental interactions that are modifying or modulating the effect of APOE alleles. The investigation and isolation of APOE interacting factors could provide significant clues in LOAD aetiology and pathogenesis.

Genetic factors modifying APOE have been intensely researched. Early attempts based on candidate gene approaches have been published but to date, most of them have not been validated sufficiently to be considered proven (7, 8, 9, 10, 11, 12). An Interesting exception could be the ESR1-APOE genetic interaction which has been observed by independent research teams (13, 14, 15). In addition, a regulatory pathway involving oestrogen hormone has been demonstrated for APOE gene transcription and CNS APOE protein levels (16, 17). However, further research on this genetic interaction is necessary to delineate definitive conclusions.

Beyond candidate gene approaches, the improvement of DNA chip-based whole genome association (WGAS) technologies could be useful in identifying APOE gene modulating factors. In fact, during the last years an extraordinary effort has been made to isolate novel genes related to AD using high-density whole genome studies (6, 18, 19, 20). Using this strategy, a new locus, named GRB-associated binding protein 2 (GAB2), has been proposed as a modifying factor for Alzheimer’s disease risk in APOE 4 carriers (6). Given the potential application of these novel results in AD diagnostics, we decided to make an independent replication to examine the GAB2 gene effect in our series. In the present work we are reporting the results obtained for GAB2-APOE genetic interaction analysis in Alzheimer’s disease patients and unrelated controls from Spain.

Methods

Subjects

We are conducting a multicenter population-based study of AD in Spain. We are collecting blood samples as well as clinical and epidemiological data from each of the patients using a protocol developed by the clinical researchers involved in this study. Written informed consent was obtained from all of the individuals included in the present study. The referral centres’ ethics committees and NeoCodex have approved this research protocol, which it was in compliance with national legislation and the Code of Ethical Principles for Medical Research Involving Human Subjects of the World Medical Association.

We have collected 1116 individuals for this study. Ethnically, these individuals are white Mediterranean with registered Spanish ancestors (two generations). Cases and controls were recruited by Fundació ACE, Institut Català de Neurociències Aplicades de Barcelona (Catalonia, Spain), Unidad de Memoria, Hospital Universitario La Paz-Cantoblanco and Hospital Clínico San Carlos. (Madrid, Spain) and Unidad de Demencias, Hospital Virgen de la Arrixaca and Fundación Alzheimur (Murcia, Spain). The study comprised 475 healthy unrelated controls obtained from the general population and 521 unrelated late onset sporadic AD cases (LOAD). All AD patients are diagnosed in accordance with National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA) criteria for possible and probable AD (21). The Spanish control population has been described previously (22). We also studied 120 neurologically-normal elderly controls. All of them were screened for the absence of cognitive impairment by a structured interview including neurological mental status examination, category fluency test, and Folstein Mini-Mental Status Examination. Mean age at recruitment (standard deviation) was 79 (8) years in cases, 52 (8) years in general population controls, and 72 (9) years in elderly controls. The mean age at AD diagnosis was 77 years (8). The respective proportions of female sex in these groups were 68%, 46%, and 54%.

DNA extraction

We obtained 5 ml of peripheral blood from all individuals to isolate germline DNA from leukocytes. DNA extraction was performed in a MagNa Pure LC Instrument (Roche Diagnostics, Germany), using MagNa Pure LC DNA Isolation kit (Roche Diagnostics, Germany) in accordance with the manufacturer’s instructions. To perform polymerase chain reactions (PCRs), we prepared aliquots of DNA at a concentration of 5 ng/μl. The rest of the stock was cryopreserved at –20°C.

APOE genotyping

Commercial kits for APOE rs429358 (SNP112) and rs7412 (SNP158) genotyping are available from Roche Diagnostics (Germany). Specifically, APOE alleles were amplified using LightCycler ApoE Mutation Detection Kit (Roche diagnostics, Germany) and detected using real-time PCR technology (Lighcycler® 480 System, Roche Diagnostics, Germany) following manufacturer’s instructions. To check the quality of the results, different compound heterozygotes for APOE SNPs were verified in an independent research laboratory.

GAB2 genotyping

GAB2 (rs2373115 G/T) genotypes were obtained using Real-time PCR. Primers and probes employed for this genotyping protocol are summarized in supplementary table 1. The technique was performed in the LightCycler ® 480 System (Roche Diagnostics). Briefly, a final volume of 20 μl using 10 ng of genomic DNA, 0.1 μM of forward amplification primer, 0.5 μM of reverse amplification primer, 0.2 μM each detection probe, and 4μl of LC480 Genotyping Master (5X, Roche Diagnostics, Germany). We used an initial denaturation step of 95 C for 7 min, followed by 45 cycles of 95° C for 30 sec, 54° C for 30 sec, and 72° C for 30 sec. Detection probes for rs2373115 polymorphism were designed using the LightCycler Probe Design Software 2.0. Melting curve analyses: after an initial denaturation at 95°C for 2 minutes at a ramp rate of 4.4°C/second, the sample were incubated at 67°C for 30 sec and 56°C for 30 sec at a ramp rate of 2.2°C/sec, and then the temperature was dropped to 40°C at a ramp rate of 1°C/second and finally led to 80°C with one acquisition per °C. Supplementary figures and further information about this technique are available on request. In order to confirm genotypes selected PCR amplicons were bi-directionally sequenced using standard capillary electrophoresis techniques.

Supplementary Table 1.

Primers and probes employed for Real-time detection of GAB2 rs2373115 marker

PCR Primers	Forward	5’CTCATTCAATGTCATTCACAA 3’
	Reverse	5’GGGCAGTTATTATGTAGGCA 3’
FRET Probes	Sensor	5’Cy5-TATAGTCCGGCACTCTCTT3-Phosphate 3’
	Anchor	5’GCTTGTAGACTTATGCGGACATGG-Fluorescein 3’

Statistical Analysis

To explore genotype distribution and test for deviation of Hardy-Weinberg Equilibrium, we employed the on-line resource at the Institute for Human Genetics, Munich, Germany (http://ihg.gsf.de). Age and sex adjusted effect sizes for single markers, genotype patterns or marker combinations were calculated by unconditional logistic regression analysis using SPSS 13.0. software.

We performed multilocus analyses to detect and model the interaction between selected loci and AD using WGA raw genotypic data freely available from TGEN (6) and those genotypes obtained during this work. We employed different approaches to perform such multilocus analyses. To select and rank interactions, we have employed fast epistasis algorithm as implemented in the PLINK software (23). Furthermore, for selected interactions we calculated Synergy index, AP, RERI, strata effect size, and case-only parameter estimates using Alambique software developed by us (unpublished data). The algorithms included in this software have been previously described (24, 25). Power calculations were performed using PS Power and Sample Size Calculations software (http://medipe.psu.ac.th/episoft/pssamplesize/). In the worst scenario (I.e.: using only NNE controls APOE carriers as controls and 231 AD cases carrying APOE 4), we have 79.1% power (Genetic Pattern exposure=0.68, alpha=0.05 and OR=2.88, for details see table 3).

Table 3.

GAB2 (rs2373115 G/T) genotypes in Spanish series

GAB2 Genotypes	LOAD cases	Controls	NNE controls
GG	352 (67.6%)	325 (68.3%)	85 (70.2%)
GT	157 (30.1%)	139 (29.1%)	31 (25.6%)
TT	12 (2.3%)	12 (2.5%)	5 (4.13%)
Totals	521	476	121

NNE controls: Neurologically-normal elderly controls

Results

We genotyped APOE SNPs 112 and 158 in 1116 individuals. Using these data we re-constructed ϵ2, ϵ3 and ϵ4 APOE alleles and genotypes in our series (Supplementary table 2). As previously reported in Spain, (26) the APOE 4 allele is strongly associated with AD in the Spanish population (OR=2.88 [95% C.I. 2.16- 3.84], p=7.38E-11). Moreover, a huge effect for the ϵ4/ ϵ4 genotype is also observed (OR= 14.45 [95% C.I., 3.34-125.2], p=1.8E-6). No differences between general population and NNE controls series were observed for APOE genotypes (P=0.61). This APOE association persists after age and sex adjustments using regression based analyses (data not shown). Given these results, we concluded that APOE genotype is clearly associated to Alzheimer’s disease in our series. Furthermore, the results underscore the suitability of our population to explore APOE-related or modifying factors.

Supplementary Table 2.

APOE genotypes in Spanish series

APOE Genotypes	LOAD cases	Controls	NNE controls
Genotype ϵ2 / ϵ2	0	4 (0.8%)	0
Genotype ϵ2 / ϵ3	24 (4.6%)	48 (10.1%)	12 (10.0%)
Genotype ϵ2 / ϵ4	9 (1.7%)	11 (2.3%)	0
Genotype ϵ3 / ϵ3	266 (51.0%)	320 (67.4%)	83 (69.2%)
Genotype ϵ3 / ϵ4	192 (36.8%)	90 (18.9%)	25 (20.8%)
Genotype ϵ4 / ϵ4	30(5.7%)	2 (0.4%)	0
Totals	521	475	120

NNE controls: Neurologically-normal elderly controls

To explore GAB2 locus involvement in AD, we genotyped the rs2373115 SNP. This marker was associated to AD in a neuropathological series from the U.S. and Netherlands (6). However, that observation was only consistently replicated in APOE ϵ4 carriers in two neuropathological or clinical AD case-control studies. We observed an identical minor allele frequency (MAF=0.17) for rs2373115 Τ allele in genotyped AD, control and NNE controls series (supplementary table 3). Separate and joint Hardy-Weinberg equilibrium (HWE) analyses in our population indicated no deviations for rs2373115 marker (P>0.24, Pearson’s goodness-of-fit chi-square, calculated for 2232 alleles). Next, we explored GAB2 rs2373115 SNP single-locus association to AD using different genetic models, and compared AD versus controls or NNE controls. No evidence of association with AD was observed for this marker (p>0.17). We concluded that the rs2373115 marker is not associated with AD in our series.

To study GAB2-APOE gene-gene interactions, we stratified our series according to APOE genotype grouping AD cases and controls in accordance with original investigations previously published (6). Again, no evidence of genetic association with AD was observed in any strata of GAB2-APOE loci pair (Table 1). In order to explore alternative interactive models, we calculated synergy and interaction indices for this marker pair using several statistical algorithms and compared our results with those obtained by Reiman et al. (Table 2). No evidence of genetic interaction was observed in our series. We concluded that GAB2 rs2373115 marker does not modify Alzheimer’s risk in Spanish APOE ϵ4 carriers (Tables 1, 2).

Table 1.

GAB2 LOAD Odds ratios in Spanish APOE ϵ4 carriers and non-carriers

APOE 4 group	rs2373115 genotype	Controls (n)	Cases (n)	% LOAD (OR, CI))*
APOE ϵ4 -	GG	253	201	44.3% (0.94, CI 0.67-1.33)
	GT/TT	119	89
APOE ϵ4+	GG	71	151	68% (1.18, 0.69-2.00)
	GT/TT	32	80
All Samples	GG	325	352	52% (1.03, 0.78-1.36)
	GT/TT	151	169

*All rs2373115 ORs are calculated using GG versus GT/TT (Groups are split according to Reiman et al. 2007)

Table 2.

Interaction analysis of GAB2-APOE gene pair using different statistical models

	TGEN NPD	TGEN NPR	TGEN CR	NEO GP	NEO SC
Risk estimates
None*	1	1	1	1	1
GAB2 only	1.02	0.66	0.91	1.06	0.90
APOE only	4.23	1.74	2.04	3.34	2.70
GAB2&APOE	11.39	4.74	5.75	2.84	2.86
Interaction indexes
S index	3.19	9.42	5.02	0.77	1.17
	(1.56-6.51)	(0.16-552.12)	(0.76-33.43)	(0.37-1.59)	(0.27-4.99)
AP	0.63	0.71	0.66	-0.2	0.09
	(0.41-0.84)	(0.39-1.02)	(0.34-0.98)	(-0.8-0.4)	(-0.72-0.91)
RERI	7.13	3.34	3.8	-0.56	0.27
	(1.89-12.37)	(0.04-6.64)	(0.23-7.38)	(-2.21-1.1)	(-2.14-2.68)
p interaction	0.01	0.01	0.03	0.47	0.74
case-only	0.742	0.825	0.060	0.340	0.340
control-only	0.002	0.003	0.201	0.872	0.462

TGEN NPD: Neuropathological Discovery series (Reiman et al., 2007; http://www.tgen.org/neurogenomics/data). TGEN NPR: Neuropathological replication series (Reiman et al., 2007; http://www.tgen.org/neurogenomics/data). TGEN CR: Clinical replication series (Reiman et al., 2007; http://www.tgen.org/neurogenomics/data). NEO GP: LOAD cases versus general population (present study). NEO SC: LOAD cases versus Neurogically-normal Elderly Controls (present study). * none: patients without APO ϵ4 and GAB2 Τ allele (reference group, OR=1). GAB2 only: patients with GAB2 Τ allele without APOE ϵ4 (OR, compared to none*). APOE only: patients carrying APOE ϵ4 without GAB2 Τ allele (OR, compared to none*). GAB2&APOE: patients carrying APOE ϵ4 plus GAB2 Τ allele (OR, compared to none*). Synergy index (S index), Attributable Proportion due to interaction (AP) and Relative Excess Risk due to interaction (RERI) according to Hosmer and Lemeshow (1992). P interaction (p value obtained using binary logistic regression based analysis). P value for Case and control only according to Yang et al. (1999).

Discussion

Definitive proofs of genetic involvement in AD have been obtained during the last decades. Specifically, in early-onset and familial Alzheimer’s disease (EOFAD) several mendelian loci have been identified by diverse genetic analysis techniques. Different germline mutations have been observed in presenilins (PSEN1 and PSEN2) or APP. These findings demonstrate the importance of genetics in AD aetiology.

However, the vast majority of AD cases appear in elderly individuals and are sporadic (27). This data implies that a clear Mendelian segregation pattern for the phenotype never appears in LOAD. LOAD is postulated to be a pure complex phenotype. This hypothetical complex model implies that genetic and non-genetic risk factors have to work together causing the disease. The isolation of these factors and the delineation of the relationships among them is the true challenge of the postgenomic era and will revolutionize AD diagnostics and therapeutics in a near future.

Apolipoprotein E plays a central role in LOAD susceptibility as well as in other cardiovascular related traits, specially in those related to plasma lipoprotein metabolism (28). Association studies have firmly demonstrate its relation to LOAD aetiology in Caucasian, Japanese or African-American populations among others. However, APOE ϵ4 associated risk is not uniform among series and its association with LOAD is not completely universal (5). These observations have not obtained a fully satisfactory explanation. It has been argued that APOE gene could contain thrifty alleles selected during deprivation periods (mainly glaciations). In this way, the exposition of APOE alleles (the ancestral allele for APOE gene is ϵ4) to contemporary environmental conditions would provoke predisposition to AD and other conditions affecting elderly populations (5).

Although this explanation might be satisfactory, the differences between APOE ϵ4 effect size among populations could also be related to a different distribution of epistatic or modifying alleles in specific ethnic groups that might modify or even mask the APOE ϵ4 allele effect. This phenomenon might also explain the lack of universality of APOE ϵ4/LOAD association.

The GAB2 gene was the first APOE modifier locus isolated using a whole genome association strategy. To check these interesting results we have replicated the experiment in our series of controls and cases for AD. Unfortunately, we failed to replicate GAB2 locus findings. Because we had employed unselected general population controls in our first replication that might be contaminated by hidden AD cases, we decided to re-analyse the results using a series of age-matched neurologically normal individuals. Again, a lack of replication compared to original findings was observed. It is important to mention that this reanalysis is strictly concordant with previous comparison having 79.1% power to detect described effect. Supporting our observations, observed allele frequencies and genotype distribution of GAB2 rs2373115 in the Spanish population are in agreement with an independent replica conducted in French and UK series (27). In this new work, appeared during the redaction process of this manuscript, a lack of replication for GAB2 gene results have also been obtained. Therefore our results agree with those presented by Chapuis et al. (2008) indicating a non significant effect of GAB2 alleles in LOAD. Furthermore, a complete absence of statistical interaction (epistasis) with the APOE ϵ4 allele was also observed in this independent study.

The observed lack of replication of GAB2 gene findings can be explained in different ways. On one hand, the existence (by random chance or by hidden stratification) of a false positive or negative result in the American or European series, respectively, cannot be ruled out. On the other hand, an overestimation of GAB2 locus effect in AD in the original investigation could be the reason for our lack of power to detect that hypothetical very small effect. In this way, whenever negative associations are reported, the power of the study is usually questioned. However, we must state that according to the patient/control ratio, sample size and the allele prevalence for each marker, we should be able to detect positive OR ≥2 with a power of 80% in most of studied scenarios (table 4). Therefore, the effect of these GAB2 and APOE digenic patterns on AD risk, if present in our population, would be very low compared to original findings (OR=2.88 for APOE ϵ4 carriers).

Table 4.

Power calculations for AD vs controls or NNE controls

Power for stratified cases (based only in APOE ϵ4 carriers). Alpha 0.05, GAB2 GG genotype frequency 0.68.
Effect size (OR)
	2.25	2.5	2.88*	3	3.5
AD vs controls	83	90	95.9	96	99
AD vs NNE controls	60	69	79.1	81	88
Power for GG GAB2/APOE ϵ4+ stratum (based on the entire series information). Alpha 0.05, Pattern frequency 0.11.
Effect size (OR)
	2.5	5	6.95*	7.5	9
AD vs controls	99.9	100	100	100	100
AD vs NNE controls	90	100	100	100	100

*Observed effect size in original investigations (Reiman et al. 2007)

Another possibility could be the existence of differences in linkage disequilibrium between GAB2 markers between Spanish and North America/Netherlands populations (although this point is extremely unlikely taking into account LD patterns observed in Spanish population haplotypes for GAB2 locus, for details see supplementary figure 1). Finally, differences in the genetic background (i.e APOE genetic modifiers) between populations could also explain observed discordances.

Supplementary Figure 1.

Linkage Disequilibrium Structure for the region involving GAB2 gene in Spanish Population

However, a detailed analysis of the original series genotypes revealed that the observed GAB2-APOE ϵ4 interaction is only attributable to a clear distortion in the digenic genotype distribution within controls instead of in LOAD cases in the Dutch/American series (Table 2). Moreover, this distortion is not homogeneous between the neuropathological discovery and the neuropathological and clinical replication series studied (Table 2). Given the observed big effect and the over representation of APOE ϵ4 allele in LOAD, it is markedly implausible that a true interacting gene for APOE does not distort its distribution in LOAD series. In fact, one would expect an important swept of “protective” genotypes in Alzheimer’s disease series and vice versa for markers potentiating APOE ϵ4 effect.

Finally the original study included neuropathologically confirmed cases and controls as opposed to clinically confirmed in our study. Therefore our study might exhibit a greater degree of misclassification and this could potentially dilute the effects of AD risk factors. However, it is reassuring that our study successfully replicated the magnitude of effect reported for APOE, suggesting that this misclassification if present had little effect in our study.

Our result does not support the involvement of the GAB2 gene in AD. However, we are convinced that the existence of large whole genome association datasets publicly available for further re-analysis and independent replication will accelerate the discovery and validation of novel genes involved in Alzheimer’s disease. Moreover, the systematic search and discovery of new potential interactions among genetic markers will be the best way to gain insight into the aetiology of complex diseases.

Acknowledgements: neoCodex SL has been partially supported by The Ministerio de Educacion y Ciencia. Spain, The Corporacion Tecnológica de Andalucía (CTA). The Agencia IDEA, Consejería de Innovación (Junta de Andalucía) and The Consejería de Salud. Comunidad Autónoma de la Región de Murcia (CARM) and Alzheimur Foundation RRL, MES, MG, JG, AGP salaries has been partially supported by the Programa Torres Quevedo. Ministerio de Educación y Ciencia. Spain

Financial disclosure: None of the authors had any financial interest or support for this paper.