Inflammatory Pathways Are Impaired in Alzheimer Disease and Differentially Associated With Apolipoprotein E Status

Abstract

Alzheimer disease (AD) is a neurodegenerative disease characterized by a cognitive decline leading to dementia. The most impactful genetic risk factor is apolipoprotein E (APOE). APOE-ε4 significantly increases AD risk, APOE-ε3 is the most common gene variant, and APOE-ε2 protects against AD. However, the underlying mechanisms of APOE-ε4 on AD risk remains unclear, with APOE-ε4 impacting many pathways. We investigated how the APOE isoforms associated with the neuroinflammatory state of the brain with and without AD pathology. Frozen brain tissue from the superior and middle temporal gyrus was analyzed from APOE-ε3/3 (n = 9) or APOE-ε4/4 (n = 10) participants with AD pathology and APOE-ε3/3 (n = 9) participants without AD pathology. We determined transcript levels of 757 inflammatory related genes using the NanoString Human Neuroinflammation Panel. We found significant pathways impaired in APOE-ε4/4-AD individuals compared to APOE-ε3/3-AD. Of interest, expression of genes related to microglial activation (SALL1), motility (FSCN1), epigenetics (DNMT1), and others showed altered expression. Additionally, we performed immunohistochemistry of P2RY12 to confirm reduced microglial activation. Our results suggest APOE-ε3 responds to AD pathology while potentially having a harmful long-term inflammatory response, while APOE-ε4 shows a weakened response to pathology. Overall, APOE isoforms appear to modulate the brain immune response to AD-type pathology.

Alzheimer disease, Apolipoprotein E, Microglia, NanoString, Neuroinflammation, Neuropathology

INTRODUCTION

Alzheimer disease (AD) is the fifth leading cause of death for those older than 65. Currently, AD affects over 6 million Americans and that number is expected to more than double by 2050 if no treatment to delay or prevent onset is found (1, 2). Hallmark pathologies that distinguish AD from other dementia-inducing diseases include Aβ plaques and neurofibrillary tangles (3–5). These lesions are both typically present in brain before cognitive tests can register impairment (6, 7). Of the 2 pathologies, cognitive deficits are most closely associated with tangle pathology (7–10). While the majority of AD cases are considered sporadic, there are genetic risk factors that can increase an individual’s risk of developing AD (11).

The AD risk gene that confers the largest known public health impact is an isoform of apolipoprotein E (APOE). APOE has 3 common isoforms: APOE-ε2, APOE-ε3, and APOE-ε4. The APOE-ε4 allele confers an increased risk of AD and has an allele frequency of 14% in the general population but is present in between 40% and 60% of AD cases (12–15). The primary role of APOE in the central nervous system is to transport cholesterol and lipids to provide neurons and other cells with the energy supply needed to function properly (16–19). In the presence of APOE-ε4, there is a decrease in this energy supply and overall decrease in support APOE can give to the neurons (20, 21). Outside of the primary role of APOE in the brain, APOE-ε4 has many secondary affects that are thought to impact AD pathology and progression. Clearance of intracellular debris and toxic proteins through autophagy has also been shown to be impacted by APOE-ε4 (22, 23). With AD pathology, APOE-ε4 has been shown to increase the amount of Aβ plaques (24–29), possibly due to decreased Aβ clearance from the brain, as well as increase tau tangles (30). APOE genotypes have been shown to impact brain development patterns as early as infancy (31) and APOE-ε4 has been shown to have differential impacts on regional pathology in AD (32, 33).

Inflammation has been associated with AD pathology for decades, and research suggests that this inflammation could be a double-edged sword in AD, exacerbating or ameliorating pathology and disease based on a specific type of response (34, 35). Studies suggest that microglia become activated in AD, releasing cytotoxic factors which may facilitate phagocytosis of Aβ, removing the inflammatory stimuli, eventually leading to induction of anti-inflammatory factors (36, 37). However, this cascade has also been shown to cause detrimental long-term consequences if the stimulus is not removed. Additionally, in AD, studies have suggested that alterations in inflammatory status vary over the disease course and the presence of multiple inflammatory profiles simultaneously (38, 39). In a histological study, a reduced level of IBA-1 has been seen suggesting that with AD pathology, the microglia are unable to provide neuronal support leading to neurotoxic environments (40). In another histological study, researchers found a higher level of microglia in AD brains, specifically in female patients suggesting sex differences in inflammation (41). Furthermore, with a growing understanding of the complexity of inflammation, studies have suggested that inflammation could exacerbate Aβ production, Aβ deposition, and tau propagation (42, 43). It is clear that, functionally, inflammation in AD is multifaceted and can have both beneficial and detrimental consequences depending on when and where the inflammation occurs.

The role of APOE in AD-related inflammation has been examined in vitro and in vivo. Studies using mouse models have suggested that APOE-ε4 induces a proinflammatory state in microglia in the presence of AD pathology and lipopolysaccharides (44–48). Further, a study in human autopsy tissue found that individuals with APOE-ε4 have a pro-inflammatory state of inflammation in their brains regardless of the level of pathology (49, 50). In a human study investigating levels of microglia using the frontal and temporal cortex, researchers found an increased level of microglia in APOE-ε4 brain suggesting APOE plays a role with microglia in the presence of AD pathology (51). Interestingly, recent studies have found impairments in other inflammatory cascades. For instance, reduction in microglial activation upon repeated traumatic brain injury in a humanized APOE-ε4 mouse model compared to APOE-ε3 suggesting an impaired inflammatory response (52). Another study found APOE-ε4 directly impairs autophagy which has been implicated in inflammation (22). With the complexity of the role of APOE-ε4, the present study aims to explore these findings further and expand upon these previous findings.

In the current study, we investigated neuroinflammatory processes, and other pathways related to neuroinflammation, using human brain tissue from the superior and middle temporal gyrus (SMTG) from APOE-ε3 or APOE-ε4 participants with and without AD pathology to characterize the inflammatory processes occurring. Our findings suggest that with APOE-ε3, there is a change in inflammatory profile based on the level of AD pathology in the brain. However, with AD pathology in APOE-e4 patients, there is an inflammatory profile similar to control brains suggesting an impaired inflammatory response to pathology.

MATERIALS AND METHODS

Sample Collection

SMTG samples were obtained through the University of Kentucky Alzheimer’s Disease Research Center (UK-ADRC) at the Sanders-Brown Center on Aging (SBCoA). The UK-ADRC recruitment details have been previously described as well as the pathological assessments that were performed on the tissue at the University of Kentucky (53–55). Sample breakdown can be seen in Table 1. Frozen samples of the SMTG were obtained and are briefly summarized in Figure 1A. A piece of SMTG was pulverized and, therefore, gray and white matter were not separated out for this study. All samples were homozygous for their APOE isoform; therefore, they will be described as follows for this manuscript: APOE-ε3/3 with AD pathology: APOE-ε3-AD; APOE-ε4/4 with AD pathology: APOE-ε4-AD, APOE-ε3/3 control: APOE-ε3-Control.

FIGURE 1.

Participants breakdown/study design. (A) Basic depiction of participant samples selected with ApoE status, Braak stage, postmortem interval (PMI), and gender differences. (B) Graphical depiction of the study design highlighting the region of interest, superior and middle temporal gyrus (SMTG), and important components of NanoString Analysis. Figure generated with BioRender.com.

TABLE 1.

AD Subject Characteristics

			Alzheimer Disease
	Combined	APOE-ε3 Control	APOE-ε3-AD	APOE-ε4-AD	p Value
Number of participants	28	9	9	10
Age at death, years, mean (SD)	82.1 (5.4)	85.2 (5.4)	80.8 (5.1)	80.5 (4.8)	0.583
Male, N (%)	12 (42.8)	2 (22.0)	4 (44.4)	6 (60.0)	0.381
PMI, mean (SD)	3.2 (1.6)	3.2 (2.6)	3.1 (1.0)	3.2 (1.0)	0.395
Cancer, N (%)	13 (46)	6 (66)	4 (50.0)	3 (42.9)	0.240
Average RIN	8.7	8.92	8.51	8.76
Atherosclerosis, N (%)					0.011
0	2 (7.2)	1 (11.0)	0 (0.0)	1 (10.0)
1	6 (21.4)	3 (33.3)	0 (0.0)	3 (30.0)
2	5 (17.9)	3 (33.3)	0 (0.0)	2 (20.0)
3	11 (39.3)	2 (22.2)	7 (77.8)	2 (20.0)
4	4 (14.3)	0 (0.0)	2 (22.2)	2 (20.0)
Arteriosclerosis severity, N (%)					0.017
None	4 (14.2)	4 (44.0)	0 (0.0)	0 (0.0)
Mild	20 (71.4)	5 (56.0)	6 (66.7)	9 (90.0)
Moderate	3 (10.7)	0 (0.0)	3 (33.3)	0 (0.0)
Unknown	1 (3.6)	0 (0.0)	0 (0.0)	1 (10.0)
Total infarcts, N (%)					0.098
0	17 (60.7)	5 (55.6)	4 (44.4)	8 (80.0)
1	6 (21.4)	3 (33.3)	3 (33.3)	0 (0.0)
2	3 (10.7)	1 (11.1)	2 (22.2)	0 (0.0)
>2	2 (7.2)	0 (0.0)	0 (0.0)	2 (20.0)
Neuritic plaques, N (%)					<0.001
None	6 (21.4)	6 (66.7)	0 (0.0)	0 (0.0)
Sparse	3 (10.6)	3 (33.3)	0 (0.0)	0 (0.0)
Moderate	3 (12.5)	0 (0.0)	1 (11.1)	2 (20.0)
Severe	16 (66.7)	0 (0.0)	8 (88.9)	8 (80.0)
Braak stage, N (%)					<0.001
I	6 (21.4)	6 (66.7)	0 (0.0)	0 (0.0)
II	3 (12.5)	3 (33.3)	0 (0.0)	0 (0.0)
V	3 (12.5)	0 (0.0)	0 (0.0)	3 (30.0)
VI	16 (66.7)	0 (0.0)	9 (100.0)	7 (70.0)

Detailed breakdown of AD subject characteristics. Combined shows both AD and control subjects. One-way ANOVA was used for all variables that were reported using means and SDs, Kruskal-Wallis tests were used for atherosclerosis, total infarcts, and Braak stage, and chi-square tests were used on the remaining values. Bolded values indicate those that reached statistical significance.

RNA Isolation

Frozen SMTG was pulverized using a mortar and pestle on dry ice with liquid nitrogen. Brain powder was stored at −80°C. RNA was extracted from approximately 40 mg of frozen pulverized tissue using the E.Z.N.A total RNA kit II (Omega-Bio-Tek, Norcross, GA) according to manufacturer’s instructions. RNA was assessed with Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA) at the University of Kentucky Genomics Core Laboratory. RIN values are shown in Table 1; samples with RIN values lower than 7 were removed from analysis. The Genomics Core ran 20 ng of RNA per sample on the NanoString Technologies nCounter Human Neuroinflammatory Panel (Nanostring Technologies, Seattle, WA). These genes were preselected from NanoString to allow for a relatively unbiased panel. All cases were assigned study identifiers allowing for blinded data analysis.

Statistical Analysis—NanoString Data

NanoString data analysis was performed using 2 software packages, nCounter and the NanoStringDiff package in R (56). Using the raw NanoString data, NanoStringDiff analyses were used to model the unnormalized NanoString reads using a negative binomial generalized linear model which incorporates normalization into its parameter estimation (57). Since many genes were tested simultaneously, a false discovery rate (FDR) of 10% was used to determine statistical significance. To allow for this study to mimic the general population, outliers were not removed from the data analysis. Figure 1B is a graphical representation of these methods. Additionally, a principal component analysis (PCA) was run to identify variation between groups. This was performed by standardizing across genes using a z score on the normalized nCounter data (Fig. 2A).

FIGURE 2.

AD impact on neuroinflammation. (A) Principal component analysis was performed using normalized NanoString counts for each participant that were then z scored (n = 28). Principal component 2 has 2 directions that are distinguishing the clusters: the positive direction represents microglial function, and the negative direction represents neurodegeneration. (B) Normalized NanoString nSolver counts for 4 of the top significant genes of interest. Significance markers: * p = 0.05–0.01, **p = 0.01–0.001, ***p = 0.001–0.0001. (C) Volcano plot showing all 757 genes with colored dots representing statistically significant genes if their p value was less than 0.05 and their log2 (fold change) was either less than −1.5 or greater than 1.5. (D) Of the 210 significant genes (p < 0.05), number of genes up- or downregulated categorized based on NanoString annotations.

Quantitative Real-Time RT-PCR and Statistical Analysis

A subset of genes on the NanoString panel was additionally analyzed using qPCR to verify trends observed (Table 2). RNA previously used for NanoString Analysis was quantified using the Biospec-Nano spectrophotometer (Shimadzu, Kyoto, Japan). About 100 ng of RNA was reverse transcribed to cDNA using the cDNA High-Capacity kit (Life Technologies, Carlsbad, CA) according to manufacturer’s instructions. Real-time PCR was performed using a 96-well plate. In each well, 1 µL of the appropriate gene probe was added to 10 µL of the Fast-TaqMan reagent along with 0.5 µL of cDNA and 6.5 µL of RNase-free water. TaqMan Gene Expression probes used are provided in Table 2 (Life Technologies). Real-time PCR was performed using the ViiA 7 Real-Time PCR system (Applied Biosystems, Grand Island, NY). All genes were normalized to 18S rRNA and the fold change was determined using the −ΔΔCt method (58). Statistical analysis was done using a Student t-test. Outliers were removed using the Grubbs test for a single outlier (59). Results from qPCR can be seen in Supplementary Data Figures S1 and S2.

TABLE 2.

Genes Used for Real-Time PCR

Gene	PMID	Probe ID
APOE	NP_000032.1	Hs00171168_m1
ARC	NP_056008.1	Hs01045540_g1
C6	NP_001108603.2	Hs01110040_m1
CLSTN1	NP_001009566.1	Hs00208929_m1
CX3CL1	NP_002987.1	Hs00171086_m1
FSCN1	NP_003079.1	Hs00602051_mH
GRN	NP_002078.1	Hs00173570_m1
IL-1b	NP_000567.1	Hs01555410_m1
IL1RN	NP_000568.1	Hs00893626_m1
ITGAM	NP_001139280.1	Hs00167304_m1
KDM4A	NP_055478.2	Hs04285637_m1
LINGO1	NP_116197.4	Hs01081148_m1
LTBR	NP_002333.1	Hs01101194_m1
MCM2	NP_004517.2	Hs01091564_m1
NRGN	NP_001119653.1	Hs00183469_m1
P2RY12	NP_073625.1	Hs1881698_s1
PIK3R1	NP_852664.1	Hs00933163_m1
PINK1	NP_115785.1	Hs00260868_m1
PSEN2	NP_000438.2	Hs01577197_m1
SALL1	NP_001121364.1	Hs04285637_m1
STX18	NP_058626.1	Hs01099207_m1
TGFb1	NP_000651.3	Hs00998133_m1
TMEM88B	NP_001140157.1	Hs04188764_m1
TNFa	NP_000585.2	Hs00174128_m1
TREM2	NP_061838.1	Hs00219132_m1

String Pathway Analysis

All pathway analysis was performed through string-db.org (60). In brief, the top 60 significant genes names were input into the multiple proteins search option. The pathway image displayed was used to visualize clusters of genes related to specific functions. The number of genes was capped at 60 to allow for visualization.

Immunohistochemistry

Human SMTG brain sections were stained immunohistochemically for P2RY12 (Purinergic Receptor P2Y12). Formalin-fixed paraffin-embedded tissue was cut into 8 µm-thick sections. Immunohistochemistry (IHC) staining was performed using anti-P2RY12 (rabbit polyclonal, 1:1000 IHC, Lot No. D118480) (Sigma-Aldrich, St. Louis, MO). A biotinylated goat antirabbit IgG (1:200; Lot No. ZG0818) secondary antibody (Vector Laboratories, Burlington, CA) was used with avidin-biotin substrate (ABC solution, Vector Laboratories). Color development was preformed using DAB (DAKO). Stained sections were dehydrated, and cover slipped in DPX (Electron Microscopy Sciences, Hatfield, PA). All slides were scanned using Zeiss Axio Scan Slide Scanner (Carl Zeiss Microscopy, White Plains, NY) and analyzed using the Halo image analysis system (Indica Labs, Albuquerque, NM).

RESULTS

AD Participant Characterization

All participants were selected from the UK-ADRC at the Sanders-Brown Center on Aging (SBCoA). The AD cohort for this study was defined as Braak stage V-VI, Thal Aβ phase 4 or higher, and consensus diagnosis of AD in APOE-ε3 and APOE-ε4 individuals. In the AD cases selected, there was minimal cerebrovascular pathology, and no detectable pathologies for other neurodegenerative diseases in the areas observed with pathological assessments (egTDP-43, α-synuclein proteinopathies). Sample size of APOE-ε4 participants for inclusion in the study was limited by there being few cases that lacked cerebrovascular comorbidities. Matching age, sex, and postmortem interval across groups also limited our APOE-ε3 AD participant size. The non-AD participant cohort was defined as a Braak stage I–II with a consensus clinical diagnosis of normal before death. More information about the cohort is presented in Table 1 and Figure 1A. Due to the prevalence of AD pathology in APOE-ε4 carriers, there was only one APOE-ε4 case that met our non-AD control criteria in the entire UK-ADRC autopsy cohort. For this reason, our control group is only APOE-ε3 participants. All samples were homozygous for their APOE isoform; therefore, they will be described as follows for this study: APOE-ε3/3 with AD pathology: APOE-ε3-AD; APOE-ε4/4 with AD pathology: APOE-ε4-AD, APOE-ε3/3 without AD pathology: APOE-ε3-Control.

Neuroinflammatory Profiles Separate Groups Based on Disease State and APOE Isoform

To characterize the differences between the disease state and APOE isoform, verification that all groups were different was essential. A PCA was run to visualize the variation between the groups. The normalized NanoString counts were used from the SMTG. Figure 2A shows the analysis of the disease states and APOE isoforms. The second dimension of the PCA (PCA2) shows variation between the groups and allows for separation between both APOE isoforms and disease states. Separation in PCA2 is driven by genes involved in microglial function, with higher expression resulting in increased PCA2 levels, as seen in APOE-ε4-AD patients. Additionally, decreased levels of PCA2 were driven by increased expression of genes associated with multiple aspects of neurodegeneration, as seen in APOE-ε3-AD patients.

APOE Isoforms Alter the Neuroinflammatory Profile With or Without AD Pathology

A large number of genes showed significant differential expression between APOE-ε3-Control and APOE-ε3-AD cases. We found 210 genes (p < 0.05) (27% of total genes analyzed) were significantly different at an alpha level of 0.05. After controlling for multiple comparisons, we found 167 genes (FDR q value <0.1) (22% of total genes analyzed) showed significant differential expression. Overall, the data show a significant difference due to AD pathology with APOE-ε3. Figure 2B shows 4 genes that are associated with pathways connected with AD in APOE-ε3/3 carriers that have also been shown to be implicated in AD disease progression. The volcano plot in Figure 2C shows gene changes between APOE-ε3-AD relative to APOE-ε3-Control and Figure 2D highlights the gene categories altered. Overall, this shows significant cellular responses to AD pathology in the APOE-ε3 genotype. We further investigated if these changes were also seen comparing APOE-ε4-AD to APOE-ε3-Control. Of the 757 genes assessed, 101 genes (13.3% of total genes analyzed) were significant with a p value <0.05, and 5 remained significant after correction for multiple comparisons (FDR q value <0.1). The pathways and specific genes altered between the groups show changes in microglial function and immune response. These findings are striking, showing the significant similarities between APOE-ε4-AD with APOE-ε3-Control. However, these changes are not related to the typical proinflammatory phenotype associated with APOE-ε4 (Supplementary DataFig. S2).

APOE Isoforms Influence the Inflammatory Profile in the Presence of AD Pathology

To see if the same pathways affected in AD are impacted by APOE isoforms, APOE-ε3-AD was compared with APOE-ε4-AD. A large number of genes were differentially expressed between APOE isoforms: 94 total genes (12.4% of total genes analyzed) were significantly different (p < 0.05) with 10 genes (1.3% of total genes analyzed) significant after correction for multiple comparisons (FDR q value <0.1). The majority of the differentially expressed genes are involved with microglial function, epigenetic regulation, and oligodendrocyte function (Fig. 3A–D). These data show alterations between the 2 APOE genotypes with AD; however, these changes parallel those seen between APOE-ε3-AD and APOE-ε3-Control. Of the 94 genes that were differentially expressed between the APOE isoforms, 33% of these genes overlap with the genes altered between APOE-ε3-AD and APOE-ε3-Control.

FIGURE 3.

ApoE isoform impact on neuroinflammation with AD. (A) Normalized NanoString nSolver counts for 4 of the top significant genes of interest. Significance markers: *p = 0.05–0.01, **p = 0.01–0.001, ***p = 0.001–0.0001 (all *** have passed FDR). (B) Volcano plot showing all 757 genes with colored dots representing statistically significant genes if their p value was less than 0.05 and their log2 (fold change) was either less than −1.5 or greater than 1.5. (C) Of the significant genes (p < 0.05), number of genes up- or downregulated categorized based on NanoString annotations. (D) string-db.org pathway analysis of the top 60 significant genes.

Association Between APOE Isoforms and Microglial Morphology

To confirm the changes seen in the transcriptional analysis, a limited histological analysis was performed on a select group of samples used in RNA analysis (4 samples/group) using the SMTG. Due to the small sample size for histology, overall trends in the expression level were examined. Tissue was stained with P2RY12, a homeostatic microglial marker, to determine the activation levels of microglia in the SMTG region. The data show a trend with APOE−ε4-AD and APOE- ε3-Controls having higher levels of P2RY12 and APOE- ε3-AD having lower expression levels of P2RY12 (Fig. 4), suggesting activation of microglia in APOE- ε3-AD patients but not in APOE−ε4-AD patients.

FIGURE 4.

Association between APOE isoforms and microglial morphology. (A) Immunohistochemistry staining of P2RY12 was scanned on the Axio Slide scanner and representative images from the regions were selected from Halo. Top row is 4× with a scale bar of 200 mM and bottom row is 20× with a scale bar of 100 mM. (B) Halo analysis was used to determine the quantification of microglia stained with P2RY12 using object analysis. PRISM 8 was used to generate graphical representation of data measuring object count per mm².

DISCUSSION

APOE-ε4 is the most impactful known sporadic AD risk factor that has been implicated in many biological pathways including inflammation (61); however, the impact of APOE genotype on inflammatory profiles in human AD brain has not been fully explored. In the current study, we identified significant differences in several key inflammatory pathways when comparing APOE-ε3 and APOE-ε4 participants with and without AD pathology. The inflammatory pathways in APOE-ε4-AD participants appeared to be impaired and generally resembled a control brain with no AD pathology. This is in contrast to APOE-ε3-AD, which showed significant changes in inflammatory pathways compared to APOE-ε3-Control, suggesting that APOE-ε3 brains have an inflammatory response to AD pathology. We also investigated genes that were differentially expressed between APOE-ε3-AD and APOE-ε4-AD and found genes altered in key pathways such as microglial function, epigenetic regulation, neuronal support, cellular motility, oligodendrocyte function, and many other pathways. While this is not a complete list of pathways that our study investigated, these are key pathways that have been supported by animal and cellular research.

Since we are studying the impact of APOE isoforms, it was important to determine the expression levels of APOE. We found a significant increase in APOE-ε3 mRNA expression levels with AD and, while not statistically significant, a similar trend in APOE-ε4 compared to control participants. APOE gene expression was similar in APOE-ε3-AD and APOE-ε4-AD brains (Fig. 2B) (62). This indicates there is an increase in mRNA levels of APOE in the presence of AD pathology regardless of APOE isoform. While studies have shown a decrease in APOE-ε4 protein compared to APOE-ε3 in AD, we have only quantified transcript levels and have not confirmed with protein levels in this study (45, 63, 64).

APOE- ε4 microglia are known to produce a proinflammatory phenotype upon activation and in AD and the present study was able to confirm the previous literature. However, outside of this response, there are additional inflammatory cascades necessary for phagocytosis and an overall response to AD pathology. Because the gene expression panel we used investigated inflammation, we identified significantly altered genes related specifically to microglia and their role in inflammation. Of interest, spalt like transcription factor 1 (SALL1) is a transcriptional regulator of microglial function and is important in maintaining microglia in a homeostatic and probing state (65). Upon activation, SALL1 expression is downregulated to allow for phagocytosis, proliferation (65), activation (66), and in the presence of neurodegeneration, conversion from resting microglia to a disease-associated microglia (DAM) (67) or the neurodegenerative microglia (MGnD) (68) phenotype. Studies suggest that the DAM/MGnD phenotype is beneficial in the early and middle stages of disease to phagocytize and cluster microglia around plaques, yet this can lead to an increase in neurotoxicity in late stages if they remain in this activation state (69). In the present study, in the presence of AD pathology with APOE-ε3 genotype, SALL1 was downregulated, which would facilitate the switch to an activation state, thus permitting microglia to respond to AD pathology. However, in the comparison between APOE-ε4-AD and APOE-ε3-Control participants, the same change was not seen and the expression of SALL1 between the 2 groups was not significantly changed. This suggests that APOE-ε4 microglia may be unable to switch from a homeostatic state to an activated state necessary to clear the AD pathology, perhaps explaining the increase in pathology seen in APOE-ε4-AD participants (Fig. 3A).

The suggestion that APOE- ε4 microglia cannot respond to AD pathology can further be seen in our immunohistochemical staining for P2RY12. P2RY12 has been classified as a homeostatic microglial marker that becomes downregulated upon microglial activation (68). In the presence of AD pathology, it is predicted that P2RY12 expression levels would decline to allow for a response to the accumulating pathology (70). This was seen in our APOE-ε3-AD samples while our APOE-ε3-Control samples still had an increased expression, as expected with lack of pathology. However, APOE-ε4-AD samples had similar levels to that of APOE-ε3-Control, which suggests that APOE-ε4 microglia cannot respond to AD pathology (Fig. 4). Further studies on SALL1 and P2RY12 are required to determine how APOE-ε4 impacts the DAM/MGnD pathways and overall microglial activation states. Additionally, understanding when the DAM/MGnD pathway and activation is beneficial versus neurotoxic will be necessary in future studies targeting inflammation.

Studies have consistently shown the impact AD pathology has on neuronal connectivity, especially through reduced long-term potentiation (LTP), a major molecular mechanism in memory formation. Our data show that the activity-regulated cytoskeleton-associated protein (ARC) which is an intermediate early gene specifically involved with LTP, is reduced in the APOE-ε3-AD participants relative to APOE-ε3-Control and APOE-ε4-AD tissue (Fig. 3A). ARC impacts recycling of glutamatergic receptors necessary for making synaptic connections and long-term memories and upon overactivation of ARC, can increase Aβ production (71). While previous labs have shown that there is an increase in ARC expression in AD, our data suggest a decrease in the APOE-ε3-AD brain. This suggests there is a potential negative role of AD pathology in the context of LTP with ARC. A caveat for this finding is that depending on the agonal state prior to autopsy, there could be a confounding element impacting expression levels of ARC.

Another transcript that was significantly altered in our study is fascin actin-bundling protein 1 (FSCN1), which is involved in white matter and general motility. FSCN1 is present on all cell types in the brain and is responsible for cytoskeleton remodeling which can allow for an increase in motility. Studies have implicated FSCN1 in white matter dysfunction and in AD as a whole due to the impact on motility. One study specifically highlighted that FSCN1 was downregulated in the white matter of AD participants (72). Our findings show a significant decrease in FSCN1 expression in the presence of AD pathology with APOE-ε3 compared to APOE-ε4-AD (Fig. 3A). FSCN1 is important in motility, and some studies have surprisingly showed an increase in motility of microglia in APOE-ε4 (73), which aligns with the results we have seen where FSCN1 is unchanged in APOE-ε4-AD.

Epigenetics has been an area of research in many diseases, especially cancer research, for many years and is a growing area of study in the neurodegenerative field (74). DNA methyltransferase 1 (DNMT1) is a gene responsible for maintaining methylation patterns after DNA replication. Methylation patterns are necessary to maintain proper transcription. Our results show that DNMT1 expression was decreased in APOE-ε3-AD participants compared to APOE-ε3-Control (Fig. 2B). With the decrease seen in the AD brain, this could lead to altered gene regulation. Studies have shown that in the presence of AD pathology, methylation patterns have been altered suggesting an important role of epigenetics in AD. Another gene of interest from the present study is helicase, lymphoid specific (HELLS), which is responsible for normal development and survival in addition to DNA methylation (75, 76). The present study shows an elevated level of expression in APOE-ε3-AD participants compared to APOE-ε4-AD participants and APOE-ε3-Control (Fig. 2B). While the results of DNMT1 and HELLS are opposite, it suggests a potential implication of DNA methylation alterations between APOE isoforms and AD status. Further investigations into the epigenetic link in AD, especially in the context of DNMT1 and HELLS, are needed to identify a possible connection.

Due to the nature of the study, there are limitations that need to be considered. Biochemical studies of material obtained at autopsy provide inherently cross-sectional insights. As mentioned previously, the human brain tissue samples evaluated comprised a mixture of gray and white matter in the SMTG region, which could skew some results because different samples will have differing proportions of gray or white matter. Additionally, due to the small sample sizes, covariates in our participant selection could not be parsed out. And finally, because we included APOE-ε4/4 individuals, there were relatively many included cases with very severe AD pathology.

It has long been known that APOE-ε4 affects many aspects of AD as well as brain health as a whole. In this study, we have identified areas in which APOE isoforms influence inflammatory pathways in AD. Using both RNA and histology data, our results suggest that APOE-ε4 patients have a reduced inflammatory response to AD pathology. It has previously been found that APOE-ε4 individuals respond differently to anti-Aβ immunotherapies, with increased rates of vasogenic edema, microhemorrhage, and other adverse events (77–79), possibly due to this altered inflammatory response. By understanding the specific differences between APOE-ε3 and APOE-ε4 individuals, especially in terms of inflammatory responses to AD pathology, we can begin to determine insights relevant to future therapies. In addition, these data show that when stratifying and grouping participants, APOE status must be considered carefully as they clearly respond to AD pathology differently.

Research reported in this article was funded by grants P30-AG028383 (UK-ADRC), 1RF1AG057754-01 (DMW), 1T32AG057461-01 (CMK), and 1F31AG069372-01 (CMK) from the National Institute on Aging. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

The authors declare that they have no competing interests.

Supplementary Data can be found at academic.oup.com/jnen.

ACKNOWLEDGMENTS

We are sincerely grateful for the research volunteers and clinical colleagues at the University of Kentucky Alzheimer’s Disease Research Center.