Abstract
APOE is the largest genetic risk factor for late-onset Alzheimer disease (AD) with E4 conferring an increased risk for AD compared to E3. The ApoE protein can impact diverse pathways in the brain including neuroinflammation but the precise impact of ApoE isoforms on inflammation remains unknown. As microglia are a primary source of neuroinflammation, this study determined whether ApoE isoforms have an impact on microglial morphology and activation using immunohistochemistry and digital analyses. Analysis of ionized calcium-binding adaptor molecule 1 (Iba1) immunoreactivity indicated greater microglial activation in both the hippocampus and superior and middle temporal gyrus (SMTG) in dementia participants versus non-demented controls. Further, only an increase in activation was seen in E3-Dementia participants in the entire SMTG, whereas in the grey matter of the SMTG, only a diagnosis of dementia impacted activation. Specific microglial morphologies showed a reduction in ramified microglia in the dementia group. For rod microglia, a reduction was seen in E4-Control patients in the hippocampus whereas in the SMTG an increase was seen in E4-Dementia patients. These findings suggest an association between ApoE isoforms and microglial morphologies and highlight the importance of considering ApoE isoforms in studies of AD pathology.
INTRODUCTION
Microglia are the primary immune cells of the brain whose principal functions include maintaining homeostasis and removing cellular debris and abnormal proteins (1). In Alzheimer disease (AD), microglia surround β-amyloid (Aβ) plaques and produce enzymes known to degrade the Aβ peptide (2–6). The ability of microglia to remove Aβ and the known genetic impact of microglial-expressed genes in AD, emphasize their importance (7, 8). It has been increasingly understood that microglia in disease are likely both detrimental and beneficial by reducing pathology in the brain while also potentially having a negative impact on neurons and other cell types (9–11). For example, activated microglia can exacerbate tau pathology and increase the levels of inflammatory factors, such as proinflammatory cytokines. These proinflammatory factors can, in turn, injure neurons, recruit more microglia to the site, and activate other innate immune cascades including the complement system (12–15).
Rio-Hortega described microglia in 1919 and he has since been dubbed the “Father of Microglia” (16). He determined the distribution of microglia throughout the brain and began ascribing morphological phenotypes to microglia in disease. Ramified microglia are often described as homeostatic or resting microglia due to their predicted function of constantly probing the environment to ensure an optimal microenvironment. Ramified microglia have long thin processes with small cell bodies (16–18). Hypertrophic microglia are often described as reactivated microglia (or “activated” in older literature) as they have transitioned in response to a stimulus. These hypertrophic microglia have shortened enlarged processes and enlarged cell bodies (19). Additional microglial phenotypes include ameboid microglia, which, as the name implies, are essentially a cell body with no processes. These can also be described as the extreme end of the activated microglial spectrum and are often found in regions of extensive tissue damage (17). Finally, rod-shaped microglia are a microglial phenotype that describes polarized microglia with a cell body in the center and long processes coming from 2 sides. Rio-Hortega’s original 1919 manuscript focused on defining these rod microglia. These microglia are thought to be another type of reactivated microglia that are hypothesized to support neurons (20–22).
In studying microglial morphologies, cell-surface markers can be assessed to further our understanding of the function of the microglia in the brain. A marker widely used to label all microglia in the brain is ionized calcium-binding adaptor molecule 1 (Iba1). This is a microglia/macrophage-specific marker on the cell membrane of myeloid cells necessary for microglial structure and function (23). Antibodies to Iba1 label peripheral and infiltrating macrophages; it is a good marker for detecting all microglial morphologies as well as macrophages. To further understand levels of microglia only, anti-purinergic receptor 12 (P2RY12) can be used to identify levels of homeostatic/resting microglia in the brain. P2RY12 is exclusively expressed on microglia in the central nervous system and is required to enable microglia to maintain homeostasis. Surface expression of P2RY12 is decreased upon microglial activation; thus, reduced expression can indicate increased activation of microglia (24).
Research from our group and others has examined the role of the largest genetic risk factor for late-onset AD, apolipoprotein E ε4 (E4), in the context of inflammation. While studies have shown that microglia with E4 have an inherent pro-inflammatory phenotype that is exacerbated in disease, other work has demonstrated significant impairments in inflammatory cascades (25–27). Several studies investigating microglial phenotypes in targeted replacement E4 mouse models have found a reduction in the activation of microglia in traumatic brain injury (TBI) models (27, 28). In a previous study, we also found E4-associated impairments in microglial activation pathways using human autopsy tissue (29).
This study investigated the role ApoE isoforms play on microglial morphology, and, therefore, morphological activation state, in human autopsy tissue. We used 2 independent autopsy cohorts: the first cohort from the Allen Brain Institute (AI-ACT Series) comprised 64 cases with 25 (39.1%) of the cases carrying at least one copy of E4; the second population from the University of Kentucky Alzheimer’s Disease Research Center (UK Series) contained 78 cases with 15 (19.2%) of the cases carrying at least one copy of E4. We studied microglial morphology using Iba1. In the UK Series, we also labeled adjacent tissue sections for P2RY12 (homeostatic microglia). With Iba1, we found an increase in microglial activation in dementia patients in both populations, this aligns with the current literature. When looking at the homeostatic microglial marker, P2RY12, a reduction in expression could only be seen in E3-Dementia whereas both E4-Control and Dementia had elevated levels like E3-Control. When diving into microglial morphologies, an increase in rod microglia was seen in E4-Dementia patients in the UK Series, whereas in the AI-ACT Series, E4-Control patients had a reduction in rod microglia suggesting a regional and ApoE isoform-driven component of rod microglia.
MATERIALS AND METHODS
Studies conformed to the institutional review board protocols regarding the use of human subjects in research.
UK Series: University of Kentucky Alzheimer’s Disease Research Center participant selection
Patient samples were obtained through the University of Kentucky Alzheimer’s Disease Research Center (UK-ADRC) at the Sanders-Brown Center on Aging. 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 (30, 31). Cases were selected for the presence of AD pathology while minimizing additional pathologies. The cases with AD pathology for this study were defined as Braak NFT stage V-VI, Thal Aβ phase 4 or higher, and consensus clinical-pathological diagnosis of AD. Exclusion criteria for the AD cases included substantial cerebrovascular pathology (large and small infarcts, severe arteriolosclerosis) and detectable pathologies for other neurodegenerative diseases in the areas observed with pathological assessments (e.g. TDP-43, α-synuclein). The control cases were matched for age, sex, and postmortem interval and were defined as a Braak NFT stage I-II with normal cognition before death. Dementia diagnosis was based on the clinical dementia rating (CDR), not on neuropathological AD. All E3 cases from the UK Series are E3/E3 and will be referred to as E3-Dementia or E3-Control depending on their diagnosis. All E4 cases will have a minimum of one copy of E4 and will be referred to as E4-Dementia or E4-Control depending on diagnosis. Cases from this series are described in Table 1.
Case series characterization
| Characteristics | UK Series | AI-ACT Series | ||||
|---|---|---|---|---|---|---|
| Number of Cases | 64 | 78 | ||||
| Age | ||||||
| 50–59 | 1 (1.6) | 0 (0.0) | ||||
| 60–69 | 11 (17.2) | 0 (0.0) | ||||
| 70–79 | 13 (20.3) | 8 (10.3) | ||||
| 80–89 | 28 (43.8) | 34 (43.6) | ||||
| 90–99 | 11 (17.2) | 31 (39.7) | ||||
| 100+ | 0 (0.0) | 5 (6.4) | ||||
| Sex | ||||||
| Male | 26 (40.6) | 47 (60.3) | ||||
| Female | 38 (59.4) | 31 (39.7) | ||||
| ApoE status | ||||||
| ApoE4 | 25 (39.1) | 15 (19.2) | ||||
| No ApoE4 | 39 (60.9) | 63 (80.8) | ||||
| Clinical dementia status | ||||||
| All cases | Non-E4 | E4 | All cases | Non-E4 | E4 | |
| Dementia | 28 (43.8) | 16 (41.0) | 12 (48.0) | 42 (53.8) | 31 (49.2) | 11 (73.3) |
| No dementia | 36 (56.2) | 23 (59.0) | 13 (52.0) | 36 (46.2) | 32 (41.0) | 4 (26.7) |
| Braak NFT status | ||||||
| All cases | Non-E4 | E4 | All cases | Non-E4 | E4 | |
| 0/I/II | 32 (50.0) | 22 (56.4) | 10 (40.0) | 21 (26.9) | 20 (31.7) | 1 (6.7) |
| III/IV | 5 (7.8) | 2 (5.1) | 3 (12.0) | 34 (43.6) | 26 (41.3) | 8 (53.3) |
| V/VI | 27 (42.2) | 15 (38.5) | 12 (48.0) | 23 (29.5) | 17 (27.0) | 6 (40.0) |
| CERAD rating | ||||||
| All cases | Non-E4 | E4 | All cases | Non-E4 | E4 | |
| None | 27 (42.2) | 20 (51.3) | 7 (28.0) | 18 (23.1) | 14 (22.2) | 4 (26.7) |
| Sparse | 4 (6.3) | 4 (10.3) | 0 (0.0) | 19 (24.4) | 18 (28.6) | 1 (6.7) |
| Moderate | 11 (17.2) | 3 (7.7) | 8 (32.0) | 21 (26.9) | 18 (28.6) | 3 (20.0) |
| Frequent | 22 (34.4) | 12 (30.8) | 10 (40.0) | 20 (25.6) | 13 (20.6) | 7 (46.6) |
| Characteristics | UK Series | AI-ACT Series | ||||
|---|---|---|---|---|---|---|
| Number of Cases | 64 | 78 | ||||
| Age | ||||||
| 50–59 | 1 (1.6) | 0 (0.0) | ||||
| 60–69 | 11 (17.2) | 0 (0.0) | ||||
| 70–79 | 13 (20.3) | 8 (10.3) | ||||
| 80–89 | 28 (43.8) | 34 (43.6) | ||||
| 90–99 | 11 (17.2) | 31 (39.7) | ||||
| 100+ | 0 (0.0) | 5 (6.4) | ||||
| Sex | ||||||
| Male | 26 (40.6) | 47 (60.3) | ||||
| Female | 38 (59.4) | 31 (39.7) | ||||
| ApoE status | ||||||
| ApoE4 | 25 (39.1) | 15 (19.2) | ||||
| No ApoE4 | 39 (60.9) | 63 (80.8) | ||||
| Clinical dementia status | ||||||
| All cases | Non-E4 | E4 | All cases | Non-E4 | E4 | |
| Dementia | 28 (43.8) | 16 (41.0) | 12 (48.0) | 42 (53.8) | 31 (49.2) | 11 (73.3) |
| No dementia | 36 (56.2) | 23 (59.0) | 13 (52.0) | 36 (46.2) | 32 (41.0) | 4 (26.7) |
| Braak NFT status | ||||||
| All cases | Non-E4 | E4 | All cases | Non-E4 | E4 | |
| 0/I/II | 32 (50.0) | 22 (56.4) | 10 (40.0) | 21 (26.9) | 20 (31.7) | 1 (6.7) |
| III/IV | 5 (7.8) | 2 (5.1) | 3 (12.0) | 34 (43.6) | 26 (41.3) | 8 (53.3) |
| V/VI | 27 (42.2) | 15 (38.5) | 12 (48.0) | 23 (29.5) | 17 (27.0) | 6 (40.0) |
| CERAD rating | ||||||
| All cases | Non-E4 | E4 | All cases | Non-E4 | E4 | |
| None | 27 (42.2) | 20 (51.3) | 7 (28.0) | 18 (23.1) | 14 (22.2) | 4 (26.7) |
| Sparse | 4 (6.3) | 4 (10.3) | 0 (0.0) | 19 (24.4) | 18 (28.6) | 1 (6.7) |
| Moderate | 11 (17.2) | 3 (7.7) | 8 (32.0) | 21 (26.9) | 18 (28.6) | 3 (20.0) |
| Frequent | 22 (34.4) | 12 (30.8) | 10 (40.0) | 20 (25.6) | 13 (20.6) | 7 (46.6) |
A total number of cases and percent of total cases from each series are shown in the table. UK Series tissue was collected from the University of Kentucky Alzheimer’s Disease Research Center (ADRC) brain bank. AI-ACT Series patient breakdown was collected from a publicly available source (http://aging.brain-map.org/donors/summary). Braak neurofibrillary table (NFT) and Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) information was collected from both brain banks.
Case series characterization
| Characteristics | UK Series | AI-ACT Series | ||||
|---|---|---|---|---|---|---|
| Number of Cases | 64 | 78 | ||||
| Age | ||||||
| 50–59 | 1 (1.6) | 0 (0.0) | ||||
| 60–69 | 11 (17.2) | 0 (0.0) | ||||
| 70–79 | 13 (20.3) | 8 (10.3) | ||||
| 80–89 | 28 (43.8) | 34 (43.6) | ||||
| 90–99 | 11 (17.2) | 31 (39.7) | ||||
| 100+ | 0 (0.0) | 5 (6.4) | ||||
| Sex | ||||||
| Male | 26 (40.6) | 47 (60.3) | ||||
| Female | 38 (59.4) | 31 (39.7) | ||||
| ApoE status | ||||||
| ApoE4 | 25 (39.1) | 15 (19.2) | ||||
| No ApoE4 | 39 (60.9) | 63 (80.8) | ||||
| Clinical dementia status | ||||||
| All cases | Non-E4 | E4 | All cases | Non-E4 | E4 | |
| Dementia | 28 (43.8) | 16 (41.0) | 12 (48.0) | 42 (53.8) | 31 (49.2) | 11 (73.3) |
| No dementia | 36 (56.2) | 23 (59.0) | 13 (52.0) | 36 (46.2) | 32 (41.0) | 4 (26.7) |
| Braak NFT status | ||||||
| All cases | Non-E4 | E4 | All cases | Non-E4 | E4 | |
| 0/I/II | 32 (50.0) | 22 (56.4) | 10 (40.0) | 21 (26.9) | 20 (31.7) | 1 (6.7) |
| III/IV | 5 (7.8) | 2 (5.1) | 3 (12.0) | 34 (43.6) | 26 (41.3) | 8 (53.3) |
| V/VI | 27 (42.2) | 15 (38.5) | 12 (48.0) | 23 (29.5) | 17 (27.0) | 6 (40.0) |
| CERAD rating | ||||||
| All cases | Non-E4 | E4 | All cases | Non-E4 | E4 | |
| None | 27 (42.2) | 20 (51.3) | 7 (28.0) | 18 (23.1) | 14 (22.2) | 4 (26.7) |
| Sparse | 4 (6.3) | 4 (10.3) | 0 (0.0) | 19 (24.4) | 18 (28.6) | 1 (6.7) |
| Moderate | 11 (17.2) | 3 (7.7) | 8 (32.0) | 21 (26.9) | 18 (28.6) | 3 (20.0) |
| Frequent | 22 (34.4) | 12 (30.8) | 10 (40.0) | 20 (25.6) | 13 (20.6) | 7 (46.6) |
| Characteristics | UK Series | AI-ACT Series | ||||
|---|---|---|---|---|---|---|
| Number of Cases | 64 | 78 | ||||
| Age | ||||||
| 50–59 | 1 (1.6) | 0 (0.0) | ||||
| 60–69 | 11 (17.2) | 0 (0.0) | ||||
| 70–79 | 13 (20.3) | 8 (10.3) | ||||
| 80–89 | 28 (43.8) | 34 (43.6) | ||||
| 90–99 | 11 (17.2) | 31 (39.7) | ||||
| 100+ | 0 (0.0) | 5 (6.4) | ||||
| Sex | ||||||
| Male | 26 (40.6) | 47 (60.3) | ||||
| Female | 38 (59.4) | 31 (39.7) | ||||
| ApoE status | ||||||
| ApoE4 | 25 (39.1) | 15 (19.2) | ||||
| No ApoE4 | 39 (60.9) | 63 (80.8) | ||||
| Clinical dementia status | ||||||
| All cases | Non-E4 | E4 | All cases | Non-E4 | E4 | |
| Dementia | 28 (43.8) | 16 (41.0) | 12 (48.0) | 42 (53.8) | 31 (49.2) | 11 (73.3) |
| No dementia | 36 (56.2) | 23 (59.0) | 13 (52.0) | 36 (46.2) | 32 (41.0) | 4 (26.7) |
| Braak NFT status | ||||||
| All cases | Non-E4 | E4 | All cases | Non-E4 | E4 | |
| 0/I/II | 32 (50.0) | 22 (56.4) | 10 (40.0) | 21 (26.9) | 20 (31.7) | 1 (6.7) |
| III/IV | 5 (7.8) | 2 (5.1) | 3 (12.0) | 34 (43.6) | 26 (41.3) | 8 (53.3) |
| V/VI | 27 (42.2) | 15 (38.5) | 12 (48.0) | 23 (29.5) | 17 (27.0) | 6 (40.0) |
| CERAD rating | ||||||
| All cases | Non-E4 | E4 | All cases | Non-E4 | E4 | |
| None | 27 (42.2) | 20 (51.3) | 7 (28.0) | 18 (23.1) | 14 (22.2) | 4 (26.7) |
| Sparse | 4 (6.3) | 4 (10.3) | 0 (0.0) | 19 (24.4) | 18 (28.6) | 1 (6.7) |
| Moderate | 11 (17.2) | 3 (7.7) | 8 (32.0) | 21 (26.9) | 18 (28.6) | 3 (20.0) |
| Frequent | 22 (34.4) | 12 (30.8) | 10 (40.0) | 20 (25.6) | 13 (20.6) | 7 (46.6) |
A total number of cases and percent of total cases from each series are shown in the table. UK Series tissue was collected from the University of Kentucky Alzheimer’s Disease Research Center (ADRC) brain bank. AI-ACT Series patient breakdown was collected from a publicly available source (http://aging.brain-map.org/donors/summary). Braak neurofibrillary table (NFT) and Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) information was collected from both brain banks.
AI-ACT Series: aging, dementia, and TBI study of the Allen Institute for Brain Science participant and tissue selection
Samples were obtained through the Allen Institute (AI) website: (2016 Allen Institute for Brain Science. Aging, Dementia, and TBI. Available from: http://aging.brain-map.org/donors/summary). Of the 107 cases with publicly available data, 86 had a hippocampus stained for Iba1. Seven cases were excluded due to no reported APOE status; one was excluded due to a detectable hemorrhage observed on the provided hematoxylin and eosin stain; no other cases had overt vascular changes on the neuropathology slides. Dementia diagnosis was based on the CDR, not on neuropathological AD. All E3 cases from the AI-ACT study have a minimum of one copy of E3 and will be referred to as E3-Dementia or E3-Control depending on diagnosis. All E4 cases for AI-ACT had a minimum of one copy of E4 and will be referred to as E4-Dementia or E4-Control depending on diagnosis. Due to the availability of this cohort, we were unable to see what type of dementia diagnosis the patients were given. Cases from this series are described in Table 1. High-resolution images of the hippocampus stained for Iba1 were downloaded and used for analyses.
UK Series: tissue processing and immunohistochemistry staining
Human formalin-fixed paraffin-embedded superior and middle temporal gyrus (SMTG) (Brodmann Areas 21/22) tissue blocks were obtained from the Sanders-Brown Center on Aging brain bank and 8-µm-thick sections were cut. Deparaffinization and 10 mM sodium citrate antigen retrieval (microwave 1.5 minutes at power 10, cool for 10 minutes) were performed on all samples before immunohistochemistry. After permeabilization and blocking, samples were incubated overnight using either anti-Iba1 (rabbit polyclonal, 1:1000, ref: 019-19741, FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) or anti-P2RY12 (rabbit polyclonal, 1:1000, Sigma-Aldrich, St. Louis, MO). A biotinylated secondary antibody was used (secondary for Iba1 [anti-rabbit, 1:3000] and P2RY12 [anti-rabbit, 1:200]), followed by an avidin-biotin complex incubation (Vector Laboratories, Burlingame, CA, ref: PK6100). Iba1 was visualized using diaminobenzidine (DAB) with nickel enhancement (Vector DAB Substrate Kit with Nickel, ref: SK4100); P2RY12 was visualized with only DAB (Vector DAB eqVSubstrate Kit, ref: SK4103). Stained sections were dehydrated, cleared in xylene, and coverslipped in dibutylphthalate polystyrene xylene (DPX) mountant (Electron Microscopy Sciences, Hatfield, PA). All slides were scanned using Nikon BioPipeline Slide Scanner (Nikon Instruments Inc., Melville, NY) and analyzed using the Halo Image Analysis System (Indica Labs, Albuquerque, NM) as described below.
UK Series and AI-ACT Series: Iba1 activation and morphology
Microglial activation
Using Iba1 tissue (both UK Series and AI-ACT Series), microglial activation was determined using Indica Labs’ Microglial Activation Analysis Module within the Halo Image Analysis System. Thresholding for both series was done using a blinded analyst who created the thresholding settings using 3 slides to set the parameters for microglial activation versus inactivation. For the UK Series, annotations were made in the SMTG to contain the entire image and grey matter only. For the AI-ACT Series, annotations were made to outline the hippocampus. Thresholding parameters for both series are found in Table 2. The size of the tissue varied between sections and all data generated was normalized to tissue size. The same parameters were set for all samples for a given cohort to minimize batch effect for analysis.
Microglial activation criteria
| UK Series | AI-ACT Series | |
|---|---|---|
| Minimum microglia cell body diameter | 2.75 | 4 |
| Maximum microglia process radius | 27 | 20 |
| Microglia maximum fragmentation length | 5 | 2 |
| Activation process thickness | 1.875 | 3.5 |
| UK Series | AI-ACT Series | |
|---|---|---|
| Minimum microglia cell body diameter | 2.75 | 4 |
| Maximum microglia process radius | 27 | 20 |
| Microglia maximum fragmentation length | 5 | 2 |
| Activation process thickness | 1.875 | 3.5 |
A breakdown of the parameters used on the Halo Microglial Activation Module to determine microglial activation from both the UK Series and AI-ACT Series.
Microglial activation criteria
| UK Series | AI-ACT Series | |
|---|---|---|
| Minimum microglia cell body diameter | 2.75 | 4 |
| Maximum microglia process radius | 27 | 20 |
| Microglia maximum fragmentation length | 5 | 2 |
| Activation process thickness | 1.875 | 3.5 |
| UK Series | AI-ACT Series | |
|---|---|---|
| Minimum microglia cell body diameter | 2.75 | 4 |
| Maximum microglia process radius | 27 | 20 |
| Microglia maximum fragmentation length | 5 | 2 |
| Activation process thickness | 1.875 | 3.5 |
A breakdown of the parameters used on the Halo Microglial Activation Module to determine microglial activation from both the UK Series and AI-ACT Series.
Microglial morphology
Using the same tissue analyzed for microglial activation, microglia morphologies were counted throughout a select region. To determine microglia morphologies, previously published criteria for the morphologies were used as a guide (32). Within each randomly generated box, types of microglia were counted by a blinded viewer (CMK) using the following classifications: ramified, hypertrophic, rod, and ameboid. Each type was counted, and average counts were generated. Grey matter was annotated from the UK Series and the region was used to generate random partitioned areas to count microglia types. Ten boxes sized 1000 µm2 were generated. For the AI-ACT Series, within the annotated hippocampus, the CA1 region was identified and annotated and boxes of 1000 µm2 were generated to allow for a standardized area of the hippocampus.
UK Series: P2RY12 analysis
P2RY12 immunohistochemistry was performed on the UK Series tissue and analyzed using the Indica Labs Object Colocalization module in the Halo Image Analysis System. This module was used to delineate between positive stain and background for a better signal to noise ratio and eliminate background signal. Annotations were made for the entire image and the grey matter. Total Object Counts per area (µm2) were collected and analyzed.
Statistical analysis
JMP Pro Software Version 13.0 was used for statistical analysis. Significance was determined where the p-value was <0.05. One-way ANOVA was used to detect genotype and disease state differences. Availability of confounding variables differed between populations and due to this, CDR or final clinical diagnosis was used to separate groups either control or dementia. For ApoE isoform grouping, differing information was provided by each cohort. In the UK Series, all E3 individuals were homozygotes for E3/E3. E4 individuals had at least one copy of E4. In the AI-ACT Series, the information provided was whether a copy of E4 was present. Therefore, the E4 group had at least one copy of E4, whereas we are unable to determine if the E3 group is homozygous for E3 or has a copy of E2 present. PRISM 8 was used to generate all graphs.
To determine significance between CDR score without looking at ApoE isoform, a Student T-test was used to determine significance between the groups. When ApoE isoforms were considered, the groups were determined independently. Treating E3-Control, E3-Dementia, E4-Control, and E4-Dementia as separate groups, a Tukey’s HSD test was run to determine significance between the groups. These are displayed on the point charts.
RESULTS
Microglial activation is associated with a clinical dementia diagnosis
For both the UK and AI-ACT Series, the level of microglial activation was determined using Iba1 immunohistochemistry and our activation module described in the methods and detailed in Table 2. Clinical dementia diagnosis was used to classify patients either as control or dementia. Participants were separated combining all control and dementia patients as well as separating ApoE isoforms within their groups (E3-Control, E3-Dementia, E4-Control, E4-Dementia).
The AI-ACT Series provided Iba1 immunohistochemical images of the hippocampus for all participants. The Iba1 labeling is seen in Figure 1A, B, with red indicating activated and green indicating inactivated microglia as delineated by the software. In the AI-ACT Series, an increase in activated microglia was seen between control and dementia participants (Fig. 1C). When ApoE isoforms were considered, an increase in activated microglia was only seen in the E4-Dementia group with a modest increase in E3-Dementia (Fig. 1D–F).
Microglial activation in AI-ACT Series. (A) Representative image of the hippocampus from the publicly available images from the AI-ACT Series. Scale bar: 1 mm. (B) Depiction of microglia in hippocampus. Green depicts inactivated microglia and red depicts activated microglia. Scale bar: 100 µm. (C) Activated microglia count for overall control and dementia participants. A significant difference was found using a 2 sample T-test demonstrates a significant difference between control and dementia (p = 0.0254). (D) Total microglial counts, both activated and inactivated, as counted by Halo Microglial Activation Module normalized to the size of tissue (mm2). Statistical analysis was run for multiple comparisons, using Tukey-Kramer HSD, no significance was found between individual groups. (E) Number of Activated Microglia as analyzed by Halo Microglial Activation Module normalized to the size of tissue (mm2). Statistical analysis was run for multiple comparisons. Using Tukey-Kramer HSD, significance was found between E3 Control and E4-Dementia with p = 0.0307. (F) Percentage of microglia detected that were activated in the tissue. Using Tukey-Kramer HSD, no significance was found. *p < 0.05.
The UK Series provided Iba1 immunohistochemical staining of microglia in the SMTG of all participants. Figure 2A, B show the delineation of activated microglia (red) and inactivated microglia (green) measured by the software. Within the UK Series, we analyzed both the entire SMTG and the grey matter of the SMTG. In both cases, we found an increase in activated microglia in the dementia group compared to the control (Fig. 2C, D). When we separated control and dementia groups further by ApoE isoforms, in the entire SMTG, an increase in activated microglia was seen with E3-Dementia, whereas only a modest increase was observed in the E4-Dementia (Fig. 2E, F). When we isolated the grey matter, an increase in total activated microglia was only seen with E3-Dementia (Fig. 2G). When we calculated the percent of microglia that were activated in the grey matter, both E3-Dementia and E4-Dementia were increased compared to their respective controls (Fig. 2H).
Microglial activation in UK Series. (A) Representative image of the SMTG from the UK Series. Scale bar: 500 µm. (B) Depiction of microglia in SMTG. Green depicts inactivated microglia and red depicts activated microglia. Scale bar: 100 µm. (C) Number of Activated Microglia as analyzed by Halo Microglial Activation Module normalized to the size of tissue (mm2) in the gray and white matter in the SMTG region. A significant difference was found using a 2 sample T-test demonstrates a significant difference between control and dementia (p = 0.0007). (D) Number of Activated Microglia as analyzed by Halo Microglial Activation Module normalized to the size of tissue (mm2) in the gray matter of the SMTG. A significant difference was found using a 2 sample T-test demonstrates a significant difference between control and dementia (p = 0.0035). (E) Number of Activated Microglia as analyzed by Halo Microglial Activation Module normalized to the size of tissue (mm2) in the entire SMTG region. (F) Percentage of microglia detected that were activated in the entire SMTG. Using Tukey-Kramer HSD, significance was found between E3-Dementia and E4-Control with p = 0.0194. (G) Number of Activated Microglia as analyzed by Halo Microglial Activation Module normalized to the size of tissue (mm2) in the grey matter of the SMTG. Using Tukey-Kramer HSD, significance was found between E3-Dementia and E4-Control with p = 0.0163 and between E3-Control and E3-Dementia with p = 0.0115. (H) Percentage of microglia detected that were activated in the grey matter of the SMTG. Using Tukey-Kramer HSD, significance was found between E3-Dementia and E4-Control with p = <0.0001, between E3-Control and E3-Dementia with p = 0.0079, and E4-Dementia and E4-Control with p = 0.0190. *p < 0.05, **p < 0.01, ***p < 0.001.
In addition to Iba1 staining, the UK Series was used to investigate homeostatic microglia using P2RY12 in the SMTG. Figure 3A shows representative images of the homeostatic microglia from each of the groups. Sample variability accounts for changes in levels of staining. Figure 3B shows object count data from the grey matter of each group. There was a reduction in homeostatic microglia in E3-Dementia compared to E3-Controls with no significant reductions in E4.
Homeostatic microglia in the UK Series. (A) Representative images of the SMTG from the UK Series. Scale bar: 200 µm. (B) Total area covered by P2RY12 in the grey matter of the SMTG normalized to the size of the tissue (µm2). Using Tukey-Kramer HSD, significance was found between E3-Dementia and E3-Control with p = 0.0402. *p < 0.05.
Microglial morphologies and clinical cognitive status
In the AI-ACT Series, the CA1 of the hippocampus was used to count distinct microglial morphologies using criteria seen in Figure 4A and described previously (32). Within the CA1, randomly generated boxes of equal size were created by the image analysis software. Inside each box, 4 microglial morphologies were counted; ramified (Fig. 4B), hypertrophic (Fig. 4C), rod (Fig. 4D), and ameboid (not pictured). Statistical analyses were first performed comparing the clinical cognitive status of control or dementia without considering the ApoE isoform. The percent of ramified microglia in the hippocampus was reduced in the dementia group compared to controls (Fig. 4E). The percent of hypertrophic microglia and rod microglia in the hippocampus were not changed with dementia (Fig. 4F–G).
Figure 4. Microglial morphologies in AI-ACT Series. (A) Representative drawings of the microglial morphologies adapted from (32) with permission from the publisher. (B–D) Representative images of the type of microglia for each row. Ramified (B), hypertrophic (C), and rod (D) from the AI-ACT Series in the CA1 region of the hippocampus. (E–G) The percent of all microglia counted in the CA1 for a given type of microglia combining all samples. A significant difference was found in percent ramified microglia using a 2 sample T-test demonstrates a significant difference between control and dementia (p = 0.0319). (H–J) Percent of all microglia counted in the CA1 for a given type of microglia segregating out ApoE isoform. Ramified Microglia: Using Tukey-Kramer HSD, significance was found between E4-Dementia and E4-Control with p = 0.0164 and between E4-Control and E3-Dementia with p = 0.0334. Rod Microglia: Using Tukey-Kramer HSD, significance was found between E3-Control and E4-Control with p = 0.0418 Statistical comparisons: Control and Dementia and ApoE isoforms. *p < 0.05. Scale bars: 100 µm.
In the UK Series, the grey matter of the SMTG was used to count the microglial morphologies using the same criteria as the AI-ACT Series. Depiction of the microglia morphologies can be seen in Figure 5A–C. Within the grey matter of the SMTG, a decrease in ramified microglia was seen in the dementia participants compared to control (Fig. 5D). Similar to the AI-ACT Series, no significant change was seen in hypertrophic microglia between groups (Fig. 5E). Interestingly, and in contrast to the hippocampus of the AI-ACT Series, a significant increase of rod microglia was seen in the grey matter of the SMTG of the UK Series dementia participants (Fig. 5I).
Microglial morphologies in UK Series. (A–C) Representative images of the type of microglia for each row: Ramified (A), hypertrophic (B), and rod (C) from the UK Series in the grey matter of the SMTG. (D–F) The percent of all microglia counted in the SMTG for a given type of microglia combining all samples. A significant difference was found using a 2 sample T-test demonstrates a significant difference between control and dementia with percent ramified microglia (p = 0.0299) and percent rod microglia (p = 0.0323). (G–I) Percent of all microglia counted in the SMTG for a given type of microglia segregating out ApoE isoform. Rod Microglia: Using Tukey-Kramer HSD, significance was found between E4-Dementia and E3-Control with p = 0.0058 and between E4-Dementia and E3-Dementia with p = 0.0194. *p < 0.05, **p < 0.01. Scale bars: 100 µm.
Rod-shaped microglia are impacted by the presence of E4 in the SMTG
To understand if the ApoE isoform impacted microglia morphologies in either the hippocampus or SMTG, the following groups were compared: E3-Control, E3-Dementia, E4-Control, E4-Dementia. In the hippocampus of the AI-ACT Series, a reduction in ramified microglia was seen in both E4-Dementia and E3-Dementia compared to their respective control groups (Fig. 4H). This finding, though less striking, was also seen in the grey matter of the SMTG of the UK Series, with a reduction in ramified microglia in E4-Dementia and a modest reduction in E3-Dementia compared to their respective control groups (Fig. 5G). In both the AI-ACT Series and UK Series, no change was seen in hypertrophic microglia when considering ApoE isoform (Figs. 4I and 5H).
As previously mentioned, rod microglia showed an increase in the dementia group in the grey matter of the SMTG of the UK Series (Fig. 5F), while there was no change in the hippocampus of the AI-ACT Series (Fig. 4G). In the hippocampus of the AI-ACT Series, we observed a striking decrease in the levels of rod microglia in the E4-Control participants compared to the other groups (Fig. 4J). In contrast, in the grey matter of the SMTG in the UK Series, we noted a significant increase in the level of rod microglia in E4-Dementia participants compared to the other groups (Fig. 5I). The reduction in rod microglia in the E4-Control group in the AI-ACT Series can be seen more in Figure 6A, while the increase in rod microglia in the E4-Dementia group of the UK Series can be seen more clearly in Figure 6B.
Figure 6. Microglial morphologies breakdown in AI-ACT Series and UK Series. Visual breakdown of the percentage breakdown for all 4 microglial morphologies counted in the CA1 of the AI-ACT Series (A) and gray matter of the SMTG in the UK Series (B).
DISCUSSION
E4 has been implicated in AD through direct impacts on AD neuropathology such as an increase in Aβ plaques and tau tangles (33). This conclusion has been drawn through countless neuropathological studies on human tissue (34–36). However, along with AD pathology, there is also a microglial response to the pathology that might lead to exacerbation or amelioration of plaques and tangles (4, 5, 14, 37). This inflammatory response can be impacted by ApoE isoforms. Our previous work on ApoE isoforms suggested an impaired inflammatory response to AD pathology in E4 autopsy brain tissue using RNA analysis (29). To build on those findings, we wanted to explore the impact of ApoE isoforms on both microglial morphology and activation using histological measures. Few studies have examined microglial activation in human tissue based on ApoE isoform (38, 39).
Previous work using autopsy tissue in AD showed an increase in activated microglia in dementia patients compared to controls (19, 40). Using the Halo Microglial Activation Module, a specific image-analysis algorithm, we were able to confirm this finding in both the AI-ACT Series (Fig. 1C) and UK Series (Fig. 2C, D). Our results validated the microglial activation parameters of our algorithm while also confirming the UK and AI-ACT populations follow previous literature. In the hippocampi of the AI-ACT Series, an increase in microglial activation was seen in the E4-Dementia group, which had the highest level of microglial activation, while only a modest increase in activated microglia could be seen in E3-Dementia (Fig. 1D–F). This finding supports prior studies that reported an increase in MHC-II expression in E4-Dementia brain tissue (38). In the SMTG of dementia patients, an increase in activated microglia was observed in both the entire SMTG and, specifically, the grey matter (Fig. 3G). In the grey matter of the SMTG, we found that the percent of all microglia activated was greater with dementia, but not when considering the APOE genotype (Fig. 3H). Additionally, confirming findings from our previous study, a reduction in homeostatic microglia (measured by P2RY12) was seen in E3-Dementia compared to other groups. Overall, these results characterize the microglial activation in both the AI-ACT Series and UK Series.
Because microglial activation is only one measurement and is based on the cell body and process length, we performed a more careful assessment of microglial morphologies to gain a deeper understanding of microglia in the AD brain. Using Bachstetter et al as guidance (32), ramified, hypertrophic, rod, and ameboid microglia were counted based on morphology. While direct comparisons to what each microglia are doing cannot be made due to the morphological nature of the findings, we can hypothesize about what the morphology means in the brain. To our knowledge, this is the first study classifying and stratifying microglial morphologies using ApoE isoform and clinical dementia.
Ramified microglia are hypothesized to be “resting” microglia responsible for surveilling and probing the microenvironment; as such, an increase in ramified microglia would be expected in cognitively normal control patients over demented patients and this is what was found in both the AI-ACT and UK studies (Figs. 4A–C and 5A–C). Hypertrophic microglia are hypothesized to be increased in a disease like AD, however, in neither of our populations did we observe any change among our groups (Figs. 4D–F and 5D–F). While no change could be seen in hypertrophic, this could be due to differences in staining between slides causing difficulties in the ability to distinguish ramified from hypertrophic. While the ramified microglia are elevated in control groups, the inverse could be seen with dementia patients having less ramified microglia and in turn, having more hypertrophic microglia.
Aside from the hypertrophic and ramified microglia, rod microglia are an additional, distinct microglial morphology. Rod microglia have been described in autopsy studies of AD and TBI (21, 41), but they remain understudied. It has been hypothesized that these rod microglia support neuronal health since studies have identified rod microglia parallel to neurons using co-staining techniques (20). Studies have found an increase in rod microglia after TBI, highlighting the potential implications of rod microglia in neuronal support (22, 42). In our current study, we found that there is no change in amounts of rod microglia with E3 in either the hippocampus of the AI-ACT Series or in the SMTG of the UK Series (Figs. 4G–I and 5G–I). However, with E4, there is both a regional and disease-associated change in rod microglia level. In the hippocampus of the AI-ACT Series, there is a reduction in rod microglia with E4-Control compared to any other group (Fig. 4I). In the SMTG of the UK Series, there is a significant increase in the levels of rod microglia in the E4-Dementia group, potentially suggesting an increase in neuronal damage (Fig. 5I). We hypothesize this could be due to the known increase in neurodegeneration in E4 which suggests the microglia are responsive to this damage (43).
Both the UK longitudinal cohort and the ACT longitudinal cohort recruit cognitively normal individuals and follow them as they age. As a result, these 2 cohorts at baseline have a much higher prevalence of mixed dementia including AD, TDP-43, Lewy body, and vascular contributions to cognitive impairment and dementia (30, 44). In the UK Series, limiting the samples with mixed dementia was done, however, and it is still likely that additional pathologies were in the brains of the dementia patient samples. In the AI-ACT Series, because we did not have the full neuropathological workup, we could not know for certain what pathologies were in the brains based on the publicly available data. We hypothesize that the reason we found statistical significance concerning the clinical syndrome of dementia is that inflammation in response to underlying pathologies like plaques, tangles, TDP-43 aggregates, or cerebrovascular disease, likely culminates in an impact on neuronal function and resulting in clinical dementia. Even when simply considering AD, studies have shown that an increase in Aβ begins at least a decade before any cognitive symptoms occur (45). This is then followed by tau pathology, which is most closely associated with cognitive impairment. It has been suggested that this is also around the time that neuroinflammation increases (14, 43, 46). Our findings here show that cognitive measures are associated with microglial activation and that inflammation may be a valid target to ameliorate clinical symptoms of dementia.
Limitations of the study
Studies with autopsy tissue provide cross-sectional insights with several potential confounders. Although they enable the evaluation of human-specific biological phenomena, they do not allow us to study the dynamic processes that are at play in the aged brain in normal or disease conditions. Understanding what “state” microglia are preferentially labeled with PET ligands like translocator protein (TSPO) might allow for longitudinal tracking of microglial changes in human disease (47). Further, using digital pathology and the HALO Microglial module are both sensitive to local staining intensity and batch effects which are a limitation of these types of studies. Additionally, the HALO module has a miss rate for counting and measuring microglia and could lead to changes in numbers of microglia.
Additional considerations are necessary due to the specific cohorts selected herein. In both cohorts, there were a limited number of ethnoracial diverse subjects. Further, because the AI-ACT Series was accessed via a publicly available platform, this tissue was handled in a different facility, and we did not have control over the immunohistochemical procedures. Additionally, ApoE data collected in the AI-ACT Series only provided information as to whether one copy of E4 was present or not. Because of this, we are not able to see whether homozygosity plays a role in the level of neuroinflammation or rod microglia. Additionally, regarding the ApoE isoforms, there is a low number of E4 Controls (n = 4) potentially impacting our findings. Also, the AI-ACT population had a high proportion of self-reported TBI, which adds a confounding variable. For microglial counts, the variability of staining and image quality varied between populations and could play a role in the overall counts and activation levels. Finally, the populations use different brain regions making comparison difficult between groups. To avoid these being a confound, we have not made statistical comparisons between the AI-ACT and UK Series.
Conclusions and future directions
It has long been understood that E4 increases an individual’s risk for developing AD by promoting AD pathology. To our knowledge, this is the first study to characterize microglia morphologies based on ApoE isoform from both the UK Series and the AI-ACT Series. We have shown that individuals with dementia have an increase in activated microglia regardless of ApoE isoform. We also showed that E4 has a regional impact on rod microglia, with the hippocampus of E4-Control patients having a reduced number of rod microglia, and the SMTG of E4-Dementia patients having elevated numbers of rod microglia. Future studies will explore additional microglial markers such as anti-CD68 to investigate the type of microglial activation using a more specified phenotype marker. Additionally, a deeper understanding of the connection between E4 and rod microglia using a microglial and neuronal co-stain will be performed. Overall, this study highlights the importance of considering ApoE isoforms in autopsy studies and highlights the potential need for ApoE genotype to be considered when inflammation is targeted therapeutically in dementia.
ACKNOWLEDGMENTS
We are sincerely grateful for the research volunteers and clinical colleagues at the University of Kentucky Alzheimer’s Disease Research Center. Images in Figure 4A are adapted from Figure 10 of (32), Bachstetter et al, Acta Neuropathol Commun 2015; 3:32 which was originally published by Springer Nature Group.
FUNDING
Research reported in this article was funded by grants P30-AG028383 (UK-ADRC), 1RF1AG057754-01 (DMW), and 1F31AG069372-01 (CMK) from the National Institute on Aging.
CONFLICT OF INTEREST
The authors have no duality or conflicts of interest to declare.
DATA AVAILABILITY
Raw data can be obtained by contacting the corresponding author.
