Abstract
A number of Alzheimer’s disease (AD) susceptibility loci are expressed abundantly in microglia. We examined associations between AD risk variants in genes that are highly expressed in microglia and neuropathological outcomes, including cerebral amyloid angiopathy (CAA) and microglial activation, in 93 AD patients. We observed significant associations of CAA pathology with APOE ɛ4 and PTK2B rs28834970. Nominally significant associations with measures of microglial activation in white matter were observed for variants in PTK2B, PICALM, and CR1. Our findings suggest that several AD risk variants may also function as disease modifiers through amyloid-β metabolism and white matter microglial activity.
INTRODUCTION
Alzheimer’s disease (AD) is the most common cause of dementia in the elderly [1]. AD pathology is characterized by extracellular plaques composed primarily of amyloid-β protein (Aβ) and intracellular neurofibrillary tangles (NFTs) composed primarily of hyperphosphorylated tau [2]. Most AD patients also have some degree of amyloid deposition in walls of cerebral blood vessels, known as cerebral amyloid angiopathy (CAA) [3]. Additionally, increases in microglia and morphologic changes suggestive of activation are associated with neurodegeneration and usually are increased in and around amyloid plaques [4].
The strongest genetic risk factor for late onset AD (LOAD) is the APOE ɛ4 allele [5]. Genome-wide association studies (GWAS) have identified a number of other LOAD susceptibility loci, including ABCA7, CD33, CR1, PICALM, and others [6–12]. These findings implicate molecular pathways in AD pathogenesis, such as Aβ metabolism, endocytosis, synaptic function, lipid transport and metabolism, and immune response [13]. Importantly, GWAS and systems-biology approaches have identified microglial immune response as an important factor in determining risk of AD; indeed, many AD risk genes identified by GWAS are expressed abundantly in microglia [14].
In addition to altering susceptibility to AD, several previous studies have demonstrated that some AD risk variants also modify neuropathological features of AD; however, these studies have primarily focused on associations of AD susceptibility loci with amyloid plaques and NFTs, and have not evaluated microglia [9, 16]. Furthermore, these investigations have predominantly included heterogeneous series of subjects ranging from cognitively normal to AD, thereby not addressing the question of whether AD genetic risk variants may also act as disease-modifiers in AD. Given the significant microglial contribution to AD pathology, in the current study we focused on AD risk variants in genes that are highly expressed in microglia and aimed to investigate whether variation in microglial AD risk genes is associated with microglial activity or other neuropathological features of AD (CAA pathology, tau pathology, or Aβ pathology) in AD cases.
MATERIALS AND METHODS
Case materials
A total of 93 autopsy-confirmed cases of AD from the Mayo Clinic Jacksonville brain bank for neurodegenerative disorders were included. This brain bank operates under procedures approved by the Mayo Clinic Institutional Review board, and all autopsies were obtained after informed consent of the next-of-kin or one with legal authority to grant permission for autopsy. Cases were included if they had been genotyped for the AD risk variants from a previous GWAS [17] and had available brain tissue for immunohistochemical analyses. In most cases, the left hemibrain had been fixed in 10% formalin and the right hemibrain was frozen at –80°. Formalin-fixed tissue was sampled with standardized dissection methods and embedded in paraffin blocks. All AD cases were unrelated non-Hispanic Caucasians. Median age at death was 75 years (range: 60–80 years) and 41 cases (44%) were male.
Neuropathologic assessment and outcome measures
Gross and macroscopic neuropathologic assessment was performed by a single neuropathologist (DWD) using systematic and standardized sampling and tissue processing. For microscopic analyses in this study, formalin fixed, paraffin-embedded tissue samples from the middle frontal gyrus were cut at 5-μm thickness and mounted on glass slides. Thioflavin-S fluorescent microscopy was used to count senile plaques (10×objective) and NFT densities (40x objective), and also to assess the severity of CAA. The density and distribution of NFT was assessed to determine Braak NFT stage [18]; Thal amyloid phase was also assigned based upon distribution of senile plaques [19]. The severity of CAA was scored on a 5-point scale (0–4) [20], where a higher score indicates a greater severity of CAA pathology (Table 1, Fig. 1). These outcome measures are summarized for the 93 AD patients in Table 1. Given the nature of the study on only AD cases, there was lack of variation in Braak stage and Thal phase; therefore, these two outcomes were not assessed for association with genetic variants.
Neuropathologic characteristics of AD patients
NFT, neurofibrillary tangle; CAA, cerebral amyloid angiopathy. The sample median (minimum, maximum) is given for continuous variables.
Immunohistochemistry for microglial markers and thioflavin S fluorescent microscopy are illustrated. For microglial phenotyping, anti-IBA1, anti-CD68 and anti-HLA-DR were used, for amyloid plaque, anti-pan-Aβ, for tau pathology, anti-phosphorylated tau were used. Thioflavin-S fluorescent microscopy was used to assess the severity of CAA.
Immunohistochemistry was performed on 5-μm thick sections after deparaffinization in three 5-min washes in xylene, rehydrated with three 2-min washes of ethanol (100, 100, 95%), and thoroughly washed in distilled water. Sections were immunostained using anti-pan-Aβ (Ab 33.1.1; human Ab 1–16 specific [21]) and anti-phosphorylated tau (CP-13, a gift from Dr. Peter Davies, Feinstein Institute for Medical research) for Alzheimer-type pathology. Sections were also immunostained for microglial markers: anti- IBA-1 (a pan-microglial marker, WAKO), CD68 (a marker for activated microglia/macrophages, DAKO) and LN3 (HLA-DR antibody for activated microglia [22], eBioscience) (Fig. 1). All immunohistochemistry was performed in batches on a DAKO AutostainerPlus (DAKO, Carpinteria, CA, USA) with the DAKO EnVisionTM + system-HRP (diaminobenzidine as chromogen). Normal goat serum (1:20 in TBST; Sigma, St Louis, MO, USA) used to block nonspecific antibody binding.
Image analysis was used to measure burden of tau, Aβ, and the various microglial markers. Digital microscopy methods have been previously described [23]. Briefly, immunostained sections were scanned on an Aperio ScanScope XT slide scanner (Aperio Technologies, Vista, CA, USA) producing high resolution digital images. The brain region included for image analysis was middle frontal gyrus. Several regions of interests were outlined on each slide. Regions of interest included the entire thickness of the cortical gray matter, the superficial white matter (in the region of arcuate fibers beneath the cortical ribbon) and deeper cerebral white matter (in the white matter deeper than the sulcal depths). Digital image analysis was performed using Aperio ImageScope software. A color deconvolution algorithm was used to count the number of pixels that were strongly immunostained by the DAB chromogen. The output variable was percentage of strong positive staining (in the region of interest). Summaries of percentage of strong positive staining for IBA1, CD68, and LN3 in the cortex, superficial white matter, and deep white matter for the 93 AD patients are listed in Table 1.
Genetic analysis
All genotyping was performed using Custom TaqMan SNP genotyping assays in an ABI PRISM 7900HT sequence detection system with 384-well Block Module from Applied Biosystems, CA, USA. The genotype data was analyzed using SDS software version 2.2.3. We selected a total of 10 AD risk variants for inclusion in this study (Supplementary Table 1). First, 9 AD risk variants in 7 different genes (CR1, BIN1, PTK2B, MS46A6A, PICALM, ABCA7, and CD33) that are highly expressed in microglia that have been identified in previous AD-GWAS investigations were included. For clarity, these genes are all highly expressed in microglia, and some (BIN1, PTK2B, ABCA7, and PICALM) are also expressed in other cell types such as neuron and astrocyte. Second, we also included APOE (which is mainly expressed in astrocyte, but also microglia) ɛ4 as a positive control given its strong association with AD and AD-neuropathology measures [24]. All genotype call rates were >95%, and there was no evidence of a departure from Hardy-Weinberg equilibrium (all p > 0.01). Genotype counts and frequencies are displayed in Supplementary Table 2.
Statistical analysis
Associations of genetic variants with NFT counts were examined using negative binomial regression models; multiplicative effects and 95% confidence intervals (CIs) were estimated and are interpreted as the multiplicative effect on the mean NFT count. Associations of variants with continuous variables (tau burden, amyloid burden, IBA1, CD68, and LN3) were evaluated with linear regression models; regression coefficients and 95% CIs were estimated and are interpreted as the additive increase on the mean value of the given neuropathologic measure. Associations between genetic variants and CAA score were examined using proportional odds logistic regression models; odds ratios (ORs) and 95% CIs were estimated and are interpreted as the multiplicative increase in the odds of a higher CAA score. All regression models were adjusted for age at death and sex.
Genetic variants were examined under an additive model (i.e., effect of each additional minor allele) unless the number of rare homozygotes was <5, in which case a dominant model (i.e., presence versus absence of the minor allele) was utilized. Several neuropathologic outcome measures were examined on the logarithm scale owing to skewed distributions. We applied a Bonferroni correction for multiple testing separately for each neuropathologic outcome measure, after which p < 0.005 was considered as statistically significant. Statistical analyses were performed using R Statistical Software (version 3.2.3).
RESULTS
Associations between each AD genetic risk variant and neuropathological outcome measures are displayed in Table 2 (significant or nominally significant [p < 0.05]) associations) and Supplementary Table 3a-d (all associations). There were two associations that survived correction for multiple testing, and both of these associations involved CAA score. Specifically, the APOE ɛ4 allele was associated with a higher CAA score (OR = 2.32, p = 0.004), while the minor allele of PTK2B rs28834970 was associated with a lower CAA score (OR = 0.43, p = 0.004).
Summary of all statistically significant or nominally significant associations between Alzheimer’s disease genetic risk variants and neuropathological outcomes
MA, minor allele; MAF, minor allele frequency; CI, confidence interval. For associations with CAA score, odds ratios, 95% CIs, and p-values result from proportional odds logistic regression models that were adjusted for age at death and sex. Odds ratios are interpreted as the multiplicative increase in the odds of a higher CAA score corresponding to each additional minor allele. For associations with deep white matter IBA1, superficial white matter LN3, and deep white matter LN3, regression coefficients, 95% CIs, and p-values result from linear regression models that were adjusted for age at death and sex. Regression coefficients are interpreted as the additive increase in the mean outcome measure (on the natural logarithm scale for superficial white matter LN3 and deep white matter LN3) corresponding to each additional minor allele. p-values <0.005 were considered as statistically significant after applying a Bonferroni correction to account for the 10 variants that were examined for association with each outcome measure.
Additionally, although not significant after multiple testing correction, nominally significant associations were identified between PICALM rs3581179 and increased superficial white matter LN3 (p = 0.016) and deep white matter LN3 (p = 0.017), between PTK2B rs28834970 and increased deep white matter IBA1 (p = 0.040), and between CR1 rs3818361 and deep white matter LN3 (p = 0.035) (Table 2). No other nominally significant associations were observed (Supplementary Table 3a-d).
DISCUSSION
In the current study, we evaluated associations between AD GWAS risk variants in genes that are highly expressed in microglia and neuropathological outcomes in pathologically confirmed AD patients. We identified associations with severity of CAA pathology for APOE ɛ4 and an inverse association with PTK2B rs28834970 that withstood correction for multiple testing. In addition, nominally significant associations of microglial markers were observed with PICALM, PTK2B, and CR1 variants. Though validation of our findings in larger series of AD patients will be important, these results indicate that several AD risk variants may act as AD disease modifiers through Aβ metabolism and microglial activity, especially in white matter.
The strongest associations were observed with severity of CAA. Specifically, the APOE ɛ4 allele was associated with a higher CAA score and therefore more severe CAA pathology, whereas the minor allele of PTK2B rs28834970 was associated with a lower CAA score and therefore less severe CAA pathology. The association between APOE ɛ4 and more severe CAA has been demonstrated previously [25, 26], and is consistent with the well-known detrimental effect of ɛ4 on AD-related neuropathology. It is thought that APOE plays a role in clearance of Aβ from brain parenchyma through vessels and contributes to CAA formation [24, 28]. Of note, APOE ɛ4 was not significantly associated with NFT count, tau burden, or amyloid burden; however, this likely reflects the lack of heterogeneity in these measures in this relatively homogenous AD autopsy series. Interestingly, the association between PTK2B rs28834970 and less severe CAA pathology is discordant with what has been previously observed regarding this variant and susceptibility to AD, where the minor allele confers an increased AD risk [8]. Therefore, the association between this allele and a lower severity of CAA pathology is somewhat counterintuitive and should be studied further.
Though only nominally significant and therefore requiring validation, the associations of PTK2B, CR1, and PICALM variants with microglial density deserve brief mention. As PTK2B, CR1, and PICALM are highly expressed in microglia [14], our results suggest that variants in these genes may potentially have an effect on microglial activity related to white matter damage/repair in AD pathology. Cerebral white matter is affected by wallerian degeneration in regions with significant cortical neuronal loss, but it is also vulnerable to hypoxic-ischemic injury, both of which are associated with microglial activation [29]. Vascular cognitive impairment, which is increasingly recognized as a concomitant pathologic process in AD [30, 31], may account for the observed association of increased burden in white matter microglia with AD genetic risk variants. Further studies are needed to explore the mechanisms of these associations.
Acknowledging the relatively small sample size of our study, our results confirm an association with severity of CAA pathology for APOE ɛ4 and also indicate that PTK2B variation may also modify severity of CAA pathology in AD. Additionally, our findings suggest that PTK2B, CR1, and PICALM variants may be associated with microglial activation, particularly in cerebral white matter. These findings, though exploratory in nature, provide hints at possible pathomechanisms that implicate interaction between microglial activity, Aβ metabolism, white matter pathology and vascular abnormalities in AD.
Footnotes
ACKNOWLEDGMENTS
We thank the patients and their families who donated brains to help further our knowledge of neurodegeneration. We are grateful to Linda Rousseau and Virginia Phillips for histological support; and Monica Castanedes-Casey (Mayo Clinic) for immunohistochemistry support. Support by the Robert E. Jacoby Professorship of Alzheimer’s Research and Mayo Clinic Alzheimer Disease Research Center (P50 AG16574).
