Targeting Neuroplasticity, Cardiovascular, and Cognitive-Associated Genomic Variants in Familial Alzheimer’s Disease

The identification of novel genetic variants contributing to the widespread in the age of onset (AOO) of Alzheimer’s disease (AD) could aid in the prognosis and/or development of new therapeutic strategies focused on early interventions. We recruited 78 individuals with AD from the Paisa genetic isolate in Antioquia, Colombia. These individuals belong to the world largest multigenerational and extended pedigree segregating AD as a consequence of a dominant fully penetrant mutation in the PSEN1 gene and exhibit an AOO ranging from the early 1930s to the late 1970s. To shed light on the genetic underpinning that could explain the large spread of the age of onset (AOO) of AD, 64 single nucleotide polymorphisms (SNP) associated with neuroanatomical, cardiovascular, and cognitive measures in AD were genotyped. Standard quality control and filtering procedures were applied, and single- and multi-locus linear mixed-effects models were used to identify AOO-associated SNPs. A full two-locus interaction model was fitted to define how identified SNPs interact to modulate AOO. We identified two key epistatic interactions between the APOE*E2 allele and SNPs ASTN2-rs7852878 and SNTG1-rs16914781 that delay AOO by up to ~ 8 years (95% CI 3.2–12.7, P = 1.83 × 10−3) and ~ 7.6 years (95% CI 3.3–11.8, P = 8.69 × 10−4), respectively, and validated our previous finding indicating that APOE*E2 delays AOO of AD in PSEN1 E280 mutation carriers. This new evidence involving APOE*E2 as an AOO delayer could be used for developing precision medicine approaches and predictive genomics models to potentially determine AOO in individuals genetically predisposed to AD. Electronic supplementary material The online version of this article (10.1007/s12035-018-1298-z) contains supplementary material, which is available to authorized users.


Introduction
The prevalence of Alzheimer's disease (AD) continues growing at an alarming pace. In 2006, the number of patients with AD was reported to be over 26.6 million worldwide, and it could rise by approximately fourfold to over 106.2 million by 2050 [1]. This neurodegenerative condition is incurable and constitutes a massive burden for patients, their families, and the public health system.
Genetic isolates have shown to be a powerful tool for the genetic mapping of inherited diseases [2]. For more than three decades, we have been studying the world's largest known pedigree segregating AD in which the E280A (p.Glu280Ala) mutation in the Presenilin-1 (PSEN1) gene causes early-onset AD [3,4]. This pedigree is genetically homogeneous, exhibits a high degree of endogamy, and originated centuries ago as a consequence of a founder effect during the colonizing of Colombia by Spaniards [2][3][4][5][6]. To date, more than 5000 individuals descend from the original founder, 1784 have been enrolled in a comprehensive ongoing clinical monitoring study, and 1181 individuals have been genotyped (459 carry the PSEN1 E280A mutation) [3]. Although the median Alzheimer's disease age of onset (ADAOO) in this mentioned pedigree is~49 years [3], it varies from the early 30s to the late 70s in some individuals [3,[7][8][9][10]. It is hypothesized that this substantial variation in the ADAOO is the result of interactions between PSEN1 and other key genes to modify ADAOO, and that this modification results in some members of this pedigree developing signs and symptoms of AD at an earlier or later age than other members (that is, these gene interactions with PSEN1 either accelerate or decelerate ADAOO).
Since cognitive function and decline are highly polygenic traits where a large number of genetic factors of small effect are involved, it is difficult to find associations between these factors and clinical outcomes assessing cognition or cognitive decline [13,14]. One of the standard methods to overcome this issue is to increase the sample size and subsequently increase the power to detect small effect sizes. Another possible approach is to perform targeted analysis by employing specific genetic markers that could be relevant to AD.
In the present study, we screened 78 individuals from the above-described pedigree and genotyped 65 single nucleotide polymorphisms (SNPs) previously reported to be associated with dementia and cognition. These SNPs showed association with neuroanatomical differences in brain areas that play essential roles in cognition such as the hippocampus, or that were related with hypertension because common genetic links appear to occur between AD and cardiovascular disease (Supplementary Table 1). We successfully replicated the association between the APOE*E2 allele and ADAOO, found two novel variants that also delay the age of onset of this pathological condition, and identified epistatic interactions between the APOE*E2 allele and variants within the Astrotactin 2 (ASTN2) and Syntrophin, Gamma 1 (SNTG1) genes that dramatically delay the ADAOO in PSEN1 E280A mutation carriers.

Subjects
Seventy-eight individuals with AD (47 [60%] women, 31 men [40%]) carrying the PSEN1 E280A mutation from the Metropolitan Area of Medellin in Antioquia, Colombia, were included in this study. Genetic studies have shown that this community has not been subject to microdifferentiation [2,5]. Clinical, neurological, and neuropsychological assessments at the Group of Neurosciences AD Clinic used a Spanish version of The Consortium to Establish a Registry for Alzheimer's Disease (CERAD) evaluation battery [15] adapted for the cultural and linguistic characteristics specific to this population [3,[16][17][18]. Mild cognitive impairment (MCI) and AD affection status were defined based on Petersen's and DSM-IV criteria, respectively [19,20]. The Ethics Committee of the University of Antioquia approved this study (Protocol 1115-408-20543). Informed consent was obtained from all participants.

DNA Extraction and SNP Genotyping
Genomic DNA was extracted from peripheral blood, and whole-genome amplified, fragmented, hybridized, fluorescently tagged, and scanned using the Infinium assay [21].
Sixty-four SNPs were selected based on previous associations with dementia, cognition, neuroanatomical differences, and blood pressure (Supplementary Table 1), and further selected in our sample. Genomic DNA was normalized to a concentration of~50 ng/μl, and 2.5 μL of genomic DNA was mixed with 2.5 μ L TaqMan OpenArray Master Mix. The resulting samples were dispensed using the OpenArray® AccuFill™ System onto OpenArray plates with each plate containing 48 samples and 65 SNP assays per sample. The QuantStudio™ 12K Flex instrument (Applied Biosystems, Carlsbad, CA, USA) was used to perform the real-time PCR reactions on the loaded OpenArray plates. The fluorescence emission results were read using the OpenArray® SNP Genotyping Analysis software v1 (Applied Biosystems), and the genotyping analysis performed using TaqMan® Genotyper v1.3 with the auto call feature and the default settings.

Genetic Association Analysis
Genotypes for the selected SNPs were processed, subject to quality control and association analysis performed using Golden Helix® SNP Variation Suite (SVS) 8.3.2 (Golden Helix, Inc. Bozeman, MT, USA). Quality control exclusion criteria included (i) deviations from Hardy-Weinberg equilibrium with P < 0.05/m (where m is the number of markers included for analysis), (ii) a minimum genotype call rate of 90%, (iii) the presence of one or more than two alleles, and (iv) a minor allele frequency (MAF) < 1% to exclude rare variants [22]. Genotype and allelic frequencies were estimated by maximum likelihood, and the identity by descent (IBD) matrix between all pairs of individuals was used for quality control.
Single-and multi-locus additive, dominant and recessive linear mixed-effect models (LMEMs) with up to 10 steps in the backward/forward optimization algorithm [23][24][25] were used to study the association between ADAOO and the aforementioned SNPs. The advantage of these models is the inclusion of both fixed (sex and years of education) and random effects, the latter to account for potential inbreeding (which, in our case, was estimated using the IBD matrix described above). A single-locus LMEM assumes that all loci have a small effect on the trait, while a multi-locus LMEM assumes that several loci have a large effect on the trait [25]. The optimal model was selected using a comprehensive exploration of multiple criteria (see [8][9][10] for more information). After the estimation procedure completed, the P values associated with the LMEM coefficientsβ 1 ;β 2 ; …;β m were extracted and corrected for multiple testing using the false discovery rate (FDR) [26] and a method based on extreme-values theory [27].

Effect of SNP × SNP Interactions on ADAOO
We evaluated potential SNP × SNP interactions between markers modifying ADAOO in carriers of the E280A mutation using a modified version of the full two-locus epistatic model [28][29][30]. Conceptually, the analysis of SNP × SNP interactions intends to determine whether the joint effect of two SNPs on the ADAOO is greater than that of either marker alone. For each pair of markers found to modify ADAOO in our patients, the ADAOO was compared at each genotype combination after correcting for potential confounding variables. Since the maximum number of genotype combinations is nine, it is likely that the sample size at each of these combinations is small. To overcome this, a nonparametric bootstrap [31,32] procedure with B = 10,000 replicates was implemented to derive permutation-based P values for these comparisons.

ADAOO-Associated SNPs
A dominant multi-locus LMEM with three steps in the forward/backward selection algorithm [25] was selected based on the mPPA and pseudo-heritability criteria. This oligogenic model includes variants rs7412 (APOE, P = 1.94 × 10 −4 , P FDR = 9.34 × 10 −3 , Table 1), rs7852878 (ASTN2, P = 1.94 × 10 −4 , P FDR = 9.34 × 10 −3 , Table 1), and rs16914781 (SNTG1, P = 1.94 × 10 −4 , P FDR = 9.34 × 10 −3 , Table 1), which explains 43% of the ADAOO variance. The proportion of the ADAOO variance explained by each marker is~24,~13, and~8% for rs7412, rs7852872, and rs16914781, respectively. No gender-or education-specific effect of these SNPs was found (Table 1). Because all estimated β coefficients from this model are positive (Table 1), these alleles delay the ADAOO in our sample of PSEN1 E280A mutation carriers. In particular, individuals with the C/T genotype in APOE-rs7412 (that is, the APOE*E2 allele) have an To identify these groups, a mixture of two Gaussian distributions was fitted as implemented in the mixtools [33] package for R [34]; the number of hidden groups was determined based on the log-likelihood criterion (the lowest the better). The blue vertical line is at~48 years, which corresponds to the average ADAOO in our sample. Box and violin plots for the ADAOO by b gender, c early onset, and d education group.
Only differences in the average ADAOO were found by AD status. e ADAOO as a function of the years of education. AD Alzheimer's disease, ADAOO Alzheimer's disease age of onset ADAOO~8 years later than that of individuals with the C/C genotype (β ¼ 8:21;ŜEβ = 1.5; Table 1a and Fig. 2a). Likewise, PSEN1 E280A mutation carriers with C/G or G/G in ASTN2-rs7852878 have an ADAOO~3.7 years later compared to that of C/C individuals (β ¼ 3:68;ŜEβ = 0.88; Table 1 and Fig. 2a). In addition, members of this pedigree with the G/G genotype in SNTG1-rs16914781 have a3 .3 years delay in the ADAOO compared to those with A/A or A/G (β ¼ 3:27;ŜEβ = 0.872; Table 1 and Fig. 2a).

Discussion
In this study, we targeted neuroanatomical, cardiovascular, and cognitive-associated markers in familial AD from the Paisa community, a genetic isolate from Antioquia, Colombia. Even though several GWAS studies have provided a potential list of a handful of putative candidate genes for sporadic AD (i.e., an age of onset > 65 years), most of those genes failed in their replication. It is well known that heterogeneity of genetic and environmental background could largely account for this apparent discrepancy. Thus, to increase power in our analyses, our approach was aimed at performing a targeted analysis in a multigenerational family from a local community that is exposed to a quite homogenous environment. More specifically, we employed 65 genetic markers related to Alzheimer's disease in a large family from the local Paisa community that originated from a common ancestor from Northern Spain during the 1500s. In this community, Alzheimer's disease is quite common as a result of the high frequency of the autosomal dominant and fully penetrant PSEN1 E280A allele. Our main goal was to shed light on the genetic underpinning that could explain the large spread of the age of onset of AD that ranks from the early 1930s to late 1970s. This cohort was also subjected to two earlier preliminary studies in which smaller sample sizes were employed, and different outcomes were observed [35,36]. Since the time those studies were performed, more E280A carriers have been identified. Hence, here we expanded the sample to detect new genes that could explain the widespread of the ADAOO observed in E280A carriers. Our present data show that the presence of the APOE*E2 allele confers protection by delaying the ADAOO by~8.2 years (95% CI 5.2-11.2, P = 4.21 × 10 −5 ; Fig. 2a), which confirms our most recent reported finding in a sample of 71 PSEN1 E280A mutation carriers displaying an extreme ADAOO [8]. Basically, by increasing the sample size to 78 patients carrying the E280A mutation, in the present study, we corroborated the decelerating APOE*E2 effect on ADAOO previously shown in individuals from the Paisa community [8]. Power analyses indicate that, overall, the ADAOO can be safely tested using our current sample size (see Supplementary Material).
Collectively, previous and current work in this genetic isolate suggests that the ADAOO accelerating and decelerating effects conferred by the APOE*E4 and APOE*E2 alleles, respectively, become evident. Therefore, our results provide convincing evidence that not only does the APOE*E2 allele exert a protective role in the onset of AD in sporadic patients [37,38], but also in the PSEN1 E280A familial cases.
The role of beta-amyloid (Aβ) in AD has been openly challenged [39][40][41]. One of the primary reasons is that there is evidence showing that Aβ deposition rises with healthy aging and its increase is not necessarily correlated with the onset of dementia and the progression to AD [41,42]. However, it is noteworthy to remark that patients with familial Alzheimer's disease display fibrillar Aβ pathology several years before symptoms onset [43]. For instance, by employing florbetapir PET analyses, Fleisher et al. showed that individuals from the Antioquia cohort carrying the PSEN1 E280A mutation showed evident accumulation of fibrillar Aβ at a mean age of 28.2 years, which was approximately 16 and 21 years before the expected MCI and dementia onset, respectively [44]. Thus, it appears that fibrillar Aβ pathology could represent an early preclinical stage of AD. Another piece of evidence supporting that Aβ is involved in the pathogenesis of AD is the fact that the three well-known genes that cause a dominant Mendelian form of familial AD (APP, PSEN1, and PSEN2) are involved in the processing of Aβ peptides [45][46][47]. Aβ peptides vary between 37 and 43 amino acids in length depending on the γ-secretase cleavage site. Mounting evidence suggests that the majority of early-onset familial AD mutations in APP, PSEN1, and PSEN2 elevate the Aβ1-42:Aβ1-40 ratio, which favors the aggregation of neurotoxic oligomeric assemblies of Aβ. It is considered that Aβ1-42 is more amyloidogenic than other Aβ peptides, which assemble into soluble Aβ oligomers that are thought to cause synaptic loss and a progressive cognitive decline in AD [48]. Aβ1-42 oligomers can elicit an inflammatory cascade by triggering the activation of microglia [49]. Moreover, Aβ oligomers associate with membrane proteins in synapses [50] and astrocytes [51]. In post-synaptic, neurons increase the Ca 2+ concentration causing inflammation and cell death [50]. Post-mortem studies carried in brain tissue from the E280A kindred suggest that their PSEN1 mutation selectively increases the processing of the amyloidogenic peptide Aβ1-42 [47]. Mounting evidence suggests that there are links between Aβ and tau in the pathogenesis of AD [52][53][54]. Aβ promotes abnormal tau phosphorylation and aggregation into neurofibrillary tangles, which is associated with neuronal toxicity and impaired cognition in AD. For instance, in functional studies employing transgenic animal models and neuronal cell culture, it was found that a 56-kDa amyloid oligomer elicited an influx in intracellular Ca2+ that triggered phosphorylation of tau at a site that promoted its aggregation [55]. This recent finding expands previous evidence supporting a possible link between Aβ and tau in the pathogenesis of AD [52][53][54].
In this context, it can be argued that the APOE*E2 variant might cause a beneficial impact on AD by improving the clearance of central Aβ, and consequently delay the onset of AD [56]. On the other hand, the APOE*E4 variant accelerates the ADAOO since it performs poorly in the clearance of Aβ peptides thereby favoring the formation of aggregates and the occurrence of the disease [57,58].
Marker rs7852878, harbored in ASTN2, was also found to delay ADAOO in individuals with AD carrying the E280A mutation. ASTN2 is an integral membrane protein that participates in glial-guided neuronal migrations and is largely expressed within the hippocampus [59]. Genomic variants in genes engaged in neuronal migration processes have been linked to several neurocognitive and psychiatric disorders. For instance, genes casually linked to schizophrenia such as Disrupted in schizophrenia-1 (DISC1), Reelin, neuroregulin (NRG), and its receptor, ERBB4, control neuronal migration during brain development [60]. Likewise, genes linked to ADHD (LPHN3) [61,62], autism (YWHAZ) [63], and depressive behavior (BDNF) [64] also control neuronal fate within different brain regions. Interestingly, SNPs within ASTN2 have been associated with cognitive decline and reduced hippocampal volume [65,66] and several psychiatric conditions such as schizophrenia [67,68], ADHD [69], and bipolar disorder [68]. More recently, genetic variants within ASTN2 have been associated with ADAOO in late-onset AD [70].
We found that marker rs16914781 within SNTG1 delays ADAOO by~3.2 years in individuals carrying the PSEN1 E280A mutation (Table 1). SNTG1 belongs to the syntrophin family; it is an adapter protein that participates in the subcellular organization of several proteins. It also mediates gammaenolase trafficking to the plasma membrane and is involved in neurotrophic signaling [71]. SNTG1 is expressed exclusively in neurons, including Purkinje cells, hippocampal pyramidal cells, and in multiple cortical regions, where it could be playing important roles in the pathophysiology of AD and other neurodegenerative/neuropsychiatric conditions [59,72]. SNTG1 has been reported as a highly penetrant recessive locus in schizophrenia [72], and as AOO modifier gene in AD [7]. More recently, a circular RNA hotspot involving SNTG1 has recently been identified in multiple system atrophy (MSA) [73], a neurodegenerative disorder causing parkinsonism, cerebellar ataxia, and autonomic, urogenital, and pyramidal dysfunction in various combinations. Previously, a case report displayed an association of MSA and AD [74]. SNTG1 has also been implicated in obstructive sleep apnea [75], a condition that is highly prevalent in patients with Alzheimer's disease [76].
To the best of our knowledge, we are the first to demonstrate a significant association between variants within ASTN2 and SNTG1, and ADAOO in individuals with familial AD caused by a fully penetrant mutation. Our study suggests that the genetic variants described here exert a protective effect by delaying ADAOO up to~3.7 years (Table 1); this value increases to~11 years when the APOE*E2 allele is present (Fig. 2a). Future studies need to be performed to address the underlying action mechanism describing the interaction between ASTN2 and PSEN1, and between STNG1 and PSEN1.